Patent Application
20150286759
The present invention relates to the technical field of simulating an electric power distribution network. Specifically, the present invention relates to a computer implemented method for hybrid simulation of an electric power distribution network and an associated communication network connected therewith for determination of a time delay between an event occurring in the power distribution network and a desired effect of a performed measure in the power distribution network, the measure having been decided on by means of a decision making algorithm as a reaction on the event.
The development of the European electrical transmission system is currently driven by several fundamental political decisions at national and international levels. In addition to the liberalization of electricity markets, system operators are facing the challenge of increasingly decentralized and volatile energy production, which increases the complexity of energy forecast and load management leading to an operation of the transmission grid close to its limit. The emerging electric mobility and the management of shiftable loads in so-called smart homes also offer new challenges and approaches for both centralized and decentralized management of the power system. New protection and control algorithms (as described by G. Ziegler, Digitaler Distanzschutz: Grundlagen and Anwendung, Wiley-VCH, May 2008) are here necessary to assure system stability and to provide an automatic and highly dynamic real-time monitoring and management. For this purpose, communication networks became an increasingly important part of the system infrastructure, as they are no longer used for static monitoring and control exclusively, but must also fulfill real-time requirements for a dynamic management (as described in C. Wietfeld, H. Georg, S. Groning, C. Lewandowski, C. Mtiller, and J. Schmutzler, “Wireless M2M Communication Networks for Smart Grid Applications,” European Wireless 2011 (EW2011), Vienna, Austria: IEEE, pp. 275-281, April 2011).
Power systems and especially Smart Grids thus increasingly apply Information and Communication Technologies (ICT) for exchanging information over wide area networks. The performance of the ICT has here become crucial for the development of new Wide Area Monitoring, Protection And Control (WAMPAC) (as described in A. Johnson, J. Wen, J. Wang, E. Liu, and Y. Hu, “Integrated System Architecture and Technology Roadmap toward WAMPAC,” Innovative Smart Grid Technologies (ISGT) 2011, IEEE, pp. 1-5, January 2011) systems. In this context, simulations are a common way for evaluation, but power systems and communication networks are usually analyzed with dedicated simulators. A combined simulation for analyzing the dynamics and mutual impacts of both domains is therefore needed.
Large scale integration of renewable energies and cross-border market integration cause an increase of both base level and volatility of power flows in the European electrical transmission system. It therefore becomes increasingly important to ensure reliable and secure system operation close to admissible limits. A promising approach is the use of dynamic wide-area information, most importantly from time-synchronized Phasor Measurement Units (PMUs), for new monitoring, protection and control applications in order to avoid large-scale system collapses and to increase operational efficiency. The development of such WAMPAC applications needs to take into account the close link to ICT infrastructure that provides the data for all monitoring and decision making processes in power system operation. The impact of underlying ICT systems is nevertheless often neglected, or is modeled to be highly simplified.
Simulations are a common way to evaluate new control and protection algorithms, but both power systems and communication networks are usually analyzed using dedicated simulators. A new simulation platform is here needed for evaluating the mutual impacts on both networks. This platform needs to handle different types of simulators, like the discrete event-based simulation commonly used for the simulation of communication networks as well as discrete time-step-based power system simulation.
In the last few years, several approaches have been developed for evaluating the performance impact of communication technologies on monitoring and control of power systems. The different types of simulators being used for power systems (discrete time-steps) and communication networks (discrete event-based) here pose special challenges for engineers. The simulation approaches proposed for solving this problem in the context of smart grids can be categorized as belonging to two fundamental concepts: comprehensive simulation and co-simulation. Both concepts will be detailed below. A generic approach from the field of distributed computing is also explained, which can also be applied for generic co-simulation.
A first concept is used in Comprehensive Simulation Environments. This concept is the creation of a new simulation for modeling both power system and communication network in one development environment. The main challenge therefor is the combination of both models in one environment, which leads to an implementation of power system simulation techniques in a communication network environment or vice versa. An advantage of this approach is that there is no need for time synchronization as both models are acting in the same time domain using the same logical simulation time. This approach has been presented in K. Mets, T. Verschueren, C. Develder, T. Vandoorn, and L. Vandevelde, “Integrated simulation of power and communication networks for smart grid applications”, IEEE 16th International Workshop on Computer Aided Modeling and Design of Communication Links and Networks (CAMAD), pp. 61-65, June 2011, introducing an integrated smart grid simulation framework using OMNeT++ as development platform. The focus is set on simulating the electrical distribution system, which has been modeled in MATLAB® and linked into the OMNeT++ simulation. This can constitute a suitable approach to stationary power system analyses, such as power flow calculations, but is not easily transferable to dynamic (e.g., electro-mechanical) power system simulations, which are of much higher complexity.
A second concept is used in Co-Simulations. Co-simulating both networks in their specialized simulators is the second approach to enable a combined simulation. Using a co-simulation, the main challenge is to connect, handle and synchronize data objects and interactions between both simulators. The implementation of time advance strategies pursued by the discrete event and time based simulators here in particular forms the main challenge which needs to be addressed. The main advantage of this approach is that there is no need for re-implementing given simulation models on both sides.
The Electric Power and Communication Synchronizing Simulator (EPOCHS) (as described in K. Hopkinson, X. Wang, R. Giovanini, J. Thorp, K. Birman, and D. Coury, “EPOCHS: A Platform for Agent-Based Electric Power and Communication Simulation Built from Commercial Off-The-Shelf Components”, IEEE Transactions on Power Systems, Vol. 21, No. 2, pp. 548-558, May 2006) has been the first approach to combine simulators for power system and communication networks in order to provide a simulation environment for analyzing the mutual impact on both networks. The creators here connected the two commercial simulators PSLF and PSCAD/EMTDC to the open source Network Simulator 2 (NS2). The connection between the simulators is realized using a Run Time Infrastructure (RTI) based on the High Level Architecture (IEEE 1516-2000). EPOCHS has been designed to investige the impacts of multi-agent systems in power systems.
An additional approach using co-simulation has been described by J. Bergmann, C. Glomb, J. Götz, J. Heuer, R. Kuntschke, and M. Winter, “Scalability of Smart Grid Protocols: Protocols and Their Simulative Evaluation for Massively Distributed DERs”, First IEEE International Conference on Smart Grid Communications, pp. 131-136, October 2010. The commercial power system simulator PSSTMNetomac was here connected to NS2 via an “external bridge” and the Java Native Interface (JNI). While the first approach of co-simulation focused on creating a comprehensive simulation environment by using a time synchronized middleware for handling the interactions between simulators, both simulators are here merged together to perform a joint simulation.
A concept to realize a co-simulation is given in Distributed Computer Simulation Systems. Distributed Computer Simulation Systems have been developed driven by military research combining various approaches from the field of distributed computing to enable a joint simulation training. These approaches resulted in the IEEE Standard 1278—Distributed Interactive Simulation (DIS) in 1993, the IEEE 1516—High Level Architecture (HLA) in 2000, and its successor IEEE 1516-2010—HLA evolved in 2010.
Comprehensive simulation environments are useful when the models to be imported from the one simulator into the other simulator are simple. For example, load flow calculations may be integrated into a communication network simulator easily, or simple delays in data transmissions may easily be taken into account in power distribution network simulators. However, if more complex models have to be considered, i.e., electro-mechanical models of power generators in power distribution networks, integrating such models is very laborious and no possibility exists to use available special simulators that are validated and comprise extensive model libraries.
The present invention relates to the second type of simulators, so called co-simulators. The previously described co-simulator EPOCHS is used to simulate communication between software agents in the electrical power network. In EPOCHS, there is no consideration of execution components (in IT hardware equipment), no integration of hardware, there are no communication standards available, and there is no comprehensive possibility to validate real time ability of applications. However, these aspects, as well as a scenario generation on the basis of standardized network models, are essential for a simulator when evaluating Smart Grid applications.
An aspect of the present invention is to provide an integrated, flexible and generic examination platform. An additional aspect of the present invention is to provide a co-simulator for hybrid simulation of an electric power distribution network, and an associated communication network connected therewith, that makes it possible to validate execution and communication components as well as execution and communication processes within the networks.
In an embodiment, the present invention provides a computer implemented method for a hybrid simulation of an electric power distribution network and an associated communication network connected therewith to determine a time delay between an event occurring in the electric power distribution network or in the associated communication network and a desired effect of a measure performed in the electric power distribution network. The method includes electromechanically simulating a dynamic behavior of the electric power distribution network in a first simulator by time-discrete numerically calculating a set of algebro-differential equations describing the electric power distribution network as a model so as to obtain a simulated electric power distribution network. The simulated electric power distribution network comprises at least two nodes and a power transmission line arranged therebetween. Each of the at least two nodes are described by at least a part of the set of algebro-differential equations. One of the at least two nodes corresponds to a generator. One of the at least two nodes corresponds to a load. In a second simulator, event-triggered, protocol-based transmissions of messages are simulated within the associated communication network from a sending communication unit over a transmission medium to a receiving target communication unit, wherein, each of the sending communication unit and the receiving target communication unit is associated to one of at least two nodes of the electric power distribution network. A decision-making algorithm is executed on a first processing unit. The first processing unit is associated to a first node of the at least two nodes of the power distribution network. On the first processing unit or on a second processing unit, the measure that was decided on via the decision-making algorithm in reaction to the event is performed. The second processing unit is associated to a second node of the at least two nodes of the electric power distribution network. The first simulator, the second simulator, the first processing unit and the second processing unit are coordinated by a time-synchronous data delivery via a superior central controlling instance. The first simulator, the second simulator, and the superior central controlling instance, and at least one of the first processing unit and the second processing unit, are respectively configured to run separately in its own process. The first simulator, the second simulator, and at least one of the first processing unit and the second processing unit each comprise a local simulation time. The superior central controlling instance comprises a global simulation time. Status values are calculated with the first simulator in discrete time steps describing a status of the electric power distribution network. The status values are made available to the superior central controlling instance and at least to the first processing unit at a first point of time. The decision-making algorithm is triggered in the first processing unit by the event. The decision-making algorithm is executed using current state values. When a processing time is determined, the decision-making algorithm needs to decide a reaction to the event. The global simulation time is set to a second point of time by adding the processing time to the first point of time. The superior central controlling instance is configured to control the first simulator and the second simulator so that a local simulation time of the first simulator and of the second simulator, respectively, do not overrun the second point of time. Depending on the measure, a communication message is generated by the first processing unit to be transmitted to a target processing unit. The communication message comprises information about the measure on which the decision-making algorithm has decided and information on a target node at which the measure is to be performed, wherein, the target node is the first node or the second node. A data size of the communication message is calculated. A time stamp of the second point of time is associated to the communication message, and the communication message is conveyed to the second simulator via the superior central controlling instance. Via the second simulator, a transmission time needed to transmit the communication message from the sending communication unit of the first node to the receiving target communication unit of the target node. The global simulation time is set to a third point of time by adding the transmission time to the second point of time. The superior central controlling instance controls the first simulator so that its local simulation times does not overrun the third point of time. Via the superior central controlling instance, based on the information on the target node and the information on the measure, an execution of the measure at the third point of time at a target processing unit associated with the target node is initiated. The measure is evaluated via the target processing unit by requesting a database to provide an execution time needed to execute the measure in reality, and executing, via the target processing unit, the measure by updating one or more status values at a fourth point of time obtained by adding the duration of the execution time to the third point of time. Via the first simulator, new status values considering current status values that have been updated by the target processing unit as soon as its local simulation time overruns the fourth point of time are calculated.
The present invention is described in greater detail below on the basis of embodiments and of the drawings in which:
For such a hybrid simulation architecture, the following design criteria have been designed:
In an embodiment, the present invention provides a computer implemented method for hybrid simulation of an electric power distribution network 1 and an associated communication network 2, 5a, 5b, 5c connected therewith, as shown in
j the target processing unit 4a, 4b, 4c evaluates the measure by requesting a database to provide the duration TACTION needed to execute the measure in reality, and the target processing unit 4a, 4b, 4c executes the measure by updating one or more status values at a fourth point of time t3t+T=t2 TACTION obtained by adding the duration TACTION to the third point of time, and thereinafter,
k the first simulator calculates new status values considering the current status values that have been updated by the second processing unit 5a, 5b, 5c, as soon as its local simulation time overruns the fourth point of time t3.
The proposed hybrid simulator architecture will establish an environment for investigating various power system management and protection algorithms. It is therefore kept generic and extensible for future developments in different areas of research in the context of smart grids. “Hybrid” in the meaning of the present invention refers to the integration of two main simulators in one common simulator architecture, including processing units for event recognition and for execution of measures as reaction on an event in the power distribution network that is all centrally controlled by means of the controlling instance.
According to the present invention, the term “event” refers to any incident that may occur in the electric power distribution network or in the communication network. It basically refers to an external incident that is not triggered by an algorithm in one of the simulations nor any other pro-configured portion of a scenario to be investigated. For example, an event can be a short current or a breakdown of a transmission line, e.g., due to a tree having fallen on said line. Alternatively, an event can be the action decided on by the decision making algorithm, e.g., switching off a power transmission line because there was an overcurrent on said line for a predetermined time.
The electric power distribution network can comprise two or more nodes, in particular a plurality of network nodes, the nodes being connected with one another via at least one transmission line. Each node can correspond to one of the following, a generator, a load, a busbar, a transformer, a switch or switching station in particular a power switch, a power plant, or a combination of two or more of these units, e.g., a substation comprising at least a busbar, a transformer and a switch. This means that in the first simulator, each physical node is represented by a model, and its dynamic behavior is described by at least one equation that is part of the set of algebro-differential equations. In the first simulator the models are together simulated by a continuous time-discrete calculation of the set of algebro-differential equations, i.e., the model of the generator, the transformer, the busbar, the load, the substation, the switching station, and/or the power plant.
A electromechanical model is used with regard to the node corresponding to a generator. The generator model can comprise a voltage controller, a turbine controller and/or a power system stabilizer. For other nodes, e.g., a simple load, a mere electrical model is sufficient. An electromechanical model can also be used for a transformer. With respect to the power transmission line, said line can physically be an AC or DC power transmission line, in particular for 60 kV, 110 kV, 220 kV, 380 kV, 400 kV, 500 kV, 700 kV, or 1150 kV. The power transmission line is considered as sub-model as well.
The overall model that is simulated in the first simulator thus consists of interconnected simulated sub-models that altogether form the set of algebro-differential equations. Each sub-model has all information about the physical entity, the node, or the transmission line, needed to describe the dynamic behavior of the physical entity. With the help of all sub-models in the overall model in the first simulator, the dynamic behavior of the electrical power distribution network can thus be simulated. In other words, the sub-models of the nodes altogether form the model describing the behavior of the power distribution network. This model is described by means of a set of algebro-differential equations. “Algebro” in this context means that this set of equations comprises algebraic equations and differential equations. Said equations represent dependencies of the state values with each other the current values of which express the current status of the power distribution network. Calculation of the equations is performed on a time-step basis, i.e., the model equations are repeatedly calculated numerically in discrete time steps with the status values thereby being updated with each time step.
The communication network exists physically parallel to the power distribution network as a completely independent working system. However, its structure may depend on the structure of the electric power distribution network. For example, the communication network can be a Wide Area Network (WAN), like the public internet (World Wide Web). The network comprises communication units each of which is associated with one network node. The amount of communication units is thus equal to the amount of nodes of the power distribution network. The communication network can consist of a computer or messaging server with a user interface and additional communication components, e.g., routers, switches and gateways. The communication units are connected with each other by means of a transmission medium. Said transmission medium can be a copper cable or fiber glass cable thus establishing the communication network as a wired landline network. It can alternatively be a radio transmission over the air, thus establishing the communication network as a mobile radio network. In another alternative, the communication network can consist of a combination of these types of transmission media. This is advantageous in order to select that transmission medium which is the fastest, cheapest and/or most reliable for a message transmission. For example, if a short current occurs in the power network, which is an event according to the present invention, a message about this event must be sent to a target node, i.e., to a target communication unit associated with this target node. This message may have high priority as this event needs immediate action. The message needs a fast and reliable transmission medium in contrast to a message that simply contains measurement data. However, transmitting measurement data may need a fast and reliably transmission medium in some cases as well.
As previously described, each node is associated with a communication unit so that it can send and/or receive communication messages. The first communication unit is associated with the first node, the second communication unit is associated with the second node, and the target communication unit is associated with the target node. As the target node is the first or second node, the target communication unit is the first or second communication unit, respectively. The electrical power distribution network and the communication network are consequently somehow connected.
The transmission of the message from one communication unit to the other naturally takes time. This leads to a time delay. Another time delay is generally involved in every protection mechanism and protection algorithm until the event is firstly recognized, secondly analyzed, and thirdly evaluated, the latter comprising the step of deciding which measure is appropriate to react on the event. For example, a short circuit can be detected by a current exceeding a defined limit, an analysis can then be performed to identify the event, the place where the high current occurred and/or which values other status parameters have. The decision making algorithm decides on the measure to be performed. Several measures can be predefined and, thus, are available for the decision making algorithm. Finally, performing the measure at a target node by means of a controlling mechanism and controlling algorithm, i.e., cutting off a power transmission line from the network, needs time as well.
All these steps lead to an accumulated time delay between the event occurring in the power distribution network and a desired effect of the performed measure in the power distribution network. With the present invention, this time delay can be determined for every protection and/or controlling algorithm.
As the given example shows, the transmission of a message over the communication network is event triggered. A message is sent only if and when an event occurs.
A node of the power distribution network may need to be controlled and/or protected. A controlling and/or protecting means belonging to the secondary equipment of a power network node can thus be present at this node, in general comprising hardware (controlling and/or protecting mechanisms) and software (controlling and/or protecting algorithms). Such means can form an integral part of the node (e.g., controller of a generator or motor, circuit breaker in a sub-station), or it can functionally be connected to the node (e.g., over current protection before a load). These means are thus considered as associated to a node in the following.
According to the present invention, at least two nodes are associated with a processing unit. The processing unit is for local data processing at a node's side. It can be connected to one or more sensors delivering measurement values and/or it can be connected to an actor which is controlled by the processing unit. For example, one or each one of the processing units comprises simulated measuring equipment, protective equipment and/or controlling equipment.
A processing unit is associated with and may in reality be physically connected to a communication unit so that a processing unit can send and receive messages to and from other processing units of other nodes of the power distribution network. A power network entity, for example, a phase shifting transformer, thus consists of the electric hardware (the transformer for transforming a voltage), and, a processing unit (unit for measuring and setting phase shift) and a communication unit (receiving instructions for setting the phase shift, sending measurement values).
A protection algorithm, a control algorithm, a measurement data processing algorithm, or a combination of said algorithms in order to protect and/or control a node of the power distribution network can be executed in a processing unit. In an alternative, an algorithm can be executed partly in one processing unit and partly in another processing unit.
An event occurring in the power distribution network can be recognized by a protection algorithm, for example, on the basis of at least one measured status value of the power distribution network. Within the simulation, said parameter measurement corresponds to the surveillance of the calculated status values. Said protection algorithm can comprise the decision making algorithm. The algorithm evaluates the event on the basis of the status values and decides on which measure is at best be performed to counteract to the event.
The measure is an action to be performed in a node. This node can be the same node in which the event occurred, or it can be another node. Said action is performed by the processing unit associated with that node (target node) in which the effect of the action shall result. This corresponds to a controlling of said node so that in this embodiment the algorithm has a protecting as well as a controlling aspect. However, in another embodiment, an algorithm running on the first processing unit can be a mere protection algorithm or a mere control algorithm, and an algorithm running on the second processing unit can also be a mere protection algorithm or a mere control algorithm.
The measure or action can, for example, be a switching of a power switch, in particular, switching the power off. Within the simulation, this is performed by means of the second processing unit changing a status value indicating the status of the switch, i.e., from the value “1” (switch closed, power turned on) to “0” (switch open, power turned off). The second processing unit can thus access and update certain status values of the first simulator, at least those status values that are associated with actors. Another measure to be performed can be data measurement at a specific node.
Coordinating the first and the second simulator and the first and the second processing unit is performed by means of the superior, central controlling instance. The controlling instance provides time-synchronous data delivery to the first and the second simulator and the first and the second processing unit. It enables the combined simulation of power systems and communication networks in an integrated simulation environment.
The controlling instance combines the hybrid simulator architecture along with the communication and synchronization process between sub-simulators of the nodes and simulated processes for controlling and protecting in the processing units. It administers time synchronicity and data synchronization. The controlling instance can be realized by a IEEE standard 1516—High Level Architecture (HLA) (as described in “IEEE Standard for Modeling and Simulation (M&S) High Level Architecture (HLA)—Framework and Rules”, IEEE Std. 1516-2010 (Revision of IEEE Std. 1516-2000), pp. 1-38, 18, 2010), its core element is a Run-Time Infrastructure. The HLA it is the latest development in this area of research and provides a general, object-oriented approach for realizing a synchronized and distributed simulation environment.
A distributed simulation using HLA is based on the following components: a Run Time Infrastructure (RTI) serving as a central administration instance, the simulators (called federates in context of the HLA), the HLA rules defining and describing the interactions of the federates, interface specifications (called ambassadors) to federates and the RTI, a standardized Object Model Template (OMT) and the customized Federation Object Model (FOM) for the overall simulation, also called federation.
The given hybrid simulator architecture has been developed within the German Research Foundation (DFG) research unit FOR1511 “Protection and control systems for reliable and secure operation of electrical transmission systems” at TU Dortmund University. Within this unit, new protection and control algorithms, emerging from interdisciplinary research from electrical engineering, information technology, statistics and computer sciences, will be developed and analyzed with respect to their real-time capability. For evaluation, focus is set on both new algorithms for protection and control as well as detailed representation of ICT infrastructure. A particular challenge constitutes the combined analysis of discrete time based (with regard to the power systems) and event based (with regard to the information and communication technologies) simulations, which is addressed in the present invention.
A core idea of the present invention is to run the first simulator, the second simulator, both the first and second processing unit together, and the controlling instance, separately in its own process, for example, on its own processor, in particular, on a separate computer system. Four such processors, in particular, four computer systems, can thus be used. Alternatively, one computer system comprising a quad core processor can be used. The first and second processing unit run together on the same processor, in particular, on the same computer system. It is, however, also possible that the first and second processing unit run separately, i.e., in separate processes, on separate processors, in particular, in separate computer systems. The processors or computer systems are connected with each other, in particular, via a local area network.
Using different processors, in particular, different computer systems, results in each simulator, at least one processing unit, and the controlling instance, having its own simulation time. Simulation time of the controlling instance is defined as a global simulation time, whereas simulation time of the first and second simulator as well as of the processing units is defined as local simulation time.
This configuration of the simulation environment anticipated, the method according to the present invention firstly calculates in the first simulator in discrete time steps status values describing the current dynamic status of the power distribution network. The status values correspond to measurement values of physical parameters used in the model and/or in the sub-models to mathematically describe the power distribution network. Calculating the status values by solving the set of equations is performed repeatedly in given intervals (whenever the local simulation time of the first simulator is advanced) as long as the controlling instance has not stopped the simulation.
Possible status values representing physical values are, for example:
Possible status values representing set values as input parameters are, for example:
After having been calculated, the status values are made available to the controlling instance and at least certain of said status values are made available at least to the first processing unit, and, for example, to the second processing unit, especially to other processing units as well. In an embodiment, a global database can be used in which the first simulator stores the status values. This database enables other entities, e.g., the controlling instance, the first or second processing unit, to access the current state values in the global database.
In an embodiment, the status values are only locally available in the first simulator and an external server, in particular, an OPC (OLE for Process Control) server, fetches the status values from the first simulator and actively provides or conveys them to one or more of the processing units and/or the controlling instance. In this embodiment, the other entities are informed about updated status values as soon as they have been calculated and/or updated. If a server, in particular an OPC server, is used, the first simulator comprises a corresponding interface, in particular an OPC interface, to enable the OPC server to access the local status values.
The time at which the status values are available, i.e., calculated or updated, is considered as a first point of time t0. Thus, at this first point time, at least the first processing unit has an amount of information. Said amount comprises at least some of the global status values calculated by the first simulator. The amount of information can additionally comprise that information received by way of communication with other entities, i.e., with other processing units. Furthermore, alternatively or in addition, the amount of information can comprise internal data that was produced by internal evaluation, calculation or execution processes. In another aspect, the amount of information can further comprise a history of the global status values, a history of data received by way of communication, and a history of the internal data. The first processing unit executes the decision making algorithm and/or other algorithms, for example, on the basis of this information amount.
In the first processing unit, an amount of information, in particular, the updated status values, are monitored and checked as to whether an event occurs, e.g., an over current or a short current. If this is the case, the decision making algorithm is triggered by the event and is executed using the current state values. A processing time TALGO is determined the algorithm needs to make a decision on how to react to the event. This can be performed by counting the flops the processor needs to completely execute the algorithm, and converting the flops into a time. The processing time is used for the next time advance of the local simulation time in order to apply the delay to the simulation.
The measure on which the algorithm has decided is defined by a specific action performed in a target node. The specific action can be the opening of a circuit breaker, reduction of power delivery on a specific power transmission line, or any other action mentioned above. In particular, the measure not only comprises the action to be performed in the target node, but also the information as to whether a communication with another processing unit has to be initiated. In some cases, the action must be performed in the same node in which the decision was made. The target node can thus be the first node, the second node, or any other node of the power distribution network.
While the algorithm is executed, the first simulator continues to calculate the status values and its local simulation time proceeds. The global simulation time of the controlling instance is set to a second point of time t0=t0+TALGO by adding the processing time TALGO to the first point of time t0.
For the following method steps, two cases must be distinguished from the two simulators point of view. Either the second point of time t1 is far in the future, i.e., the local simulation times are in the past in comparison with to the advanced global simulation time, or the second point of time t1 is in the past of the local simulation time of the first or second simulator. The controlling instance now controls the first and the second simulator in that their local simulation times do not overrun the second point of time t1. This provides that the time causality is not affected.
In an embodiment, referring to the first case, the local simulation times are advanced to the second point of time t1. This is done if there is meanwhile no incident. It must be noted that, in the interim, the system status is evaluated repeatedly by means of the processing units. E.g., measurement values are transmitted, and it may be that a decision is made somewhere.
In an embodiment, referring to the second case, the first and/or second simulator is stopped, i.e., its simulation time is held so as to prevent the respective simulation time from overrunning the second point of time. This is done because the calculations in other simulators are not yet completed.
Holding the local simulation time must be performed right before the second point of time t1. As the second simulator is event discrete, the simulation time must be stopped at the time stamp of the previous event. This could be very early to the second point of time t1. It can, for example, be performed directly before an event in the first simulator is detected, or when the decision making algorithm has been started, or shortly after this start. When the second point of time t1 has then been calculated, and is in the future in comparison with the stopped local simulation time, and calculating the equations of the model of the power distribution network has not yet been completed, the first simulator will be continued until calculation of the equation is completed. After that, the local simulation time of the first and/or second simulator can be advanced to the second point of time t1.
After the second point of time is determined, the first processing unit generates a message comprising information about the measure on which the decision making algorithm has decided, and comprising information about a target node at which the measure is to be performed. Said message is to be sent to a target communication unit associated with the target node. In some cases, the action must be performed in the same node as the event has occurred. There is thus no need to send a message to another node. However, communication can take place within this node as well, namely, an internal communication. In this case, the message is considered as having being internally sent within the first node.
It is now of interest how long the message will take to arrive at the target node. For this, the data size of the message is calculated. As the communication is performed protocol based, the total number of Bits, Bytes, frames or another unit can be used as a data size of the message.
A time stamp is additionally associated with the message; the time stamp corresponds to the second point of time t1. The time stamp is used to evaluate the end-to-end delay. The message is then conveyed to the second simulator by means of the controlling instance.
Having received the message, the second simulator determines the transmission time TK needed to transmit the message from the first communication unit of the first node to a target communication unit of the target node. As described above, the target node can be the first node, the second node, or any other node of the power distribution network. The target communication unit can consequently be the first communication unit, the second communication unit, or any other communication unit of the communication network.
Determining of the transmission time TK is performed on the basis of the model representing the communication network, in particular, the communication units and the transmission medium (optical fiber, metal cable or radio link). The second simulator takes into account data transfer rate, attenuation, data loss, data delay (jitter), capacity saturation and/or bandwidth of the transmission medium.
The transmission time TK can, for example, be considered to comprise three portions. A first time-portion for preparing the message for data transmission in the sending first communication unit, a second time-portion for preparing the message for processing in the receiving second communication unit, and a third time-portion for transmitting the message via the transmission medium.
With respect to the first and second time-portion of the transmission time TK, these portions can be determined by calculating status models of the sending first communication unit and the receiving second communication unit. With respect to the third time-portion of the transmission time TK, this portion can be determined by calculating an analytic and/or stochastic channel model of the transmission medium.
For the first and the second communication unit, and the target communication unit, respectively, a layered model can be used in order to determine the first and second time-portion of the transmission time TK. For example, the so called OSI (Open System Interconnection Reference Model) model can be used. This model can describe a communication unit as comprising up to seven layers lying hierarchically over each other, one layer serving the layer above and being served by the layer below. However, not every layer must be present. In the layered models, each layer is associated with a specific protocol for data transmission so that each layer of a communication unit can only communicate with the layer of the same hierarchical level of the other communication unit. Only neighbored layers can communicate within a communication unit.
Due to the specific protocols of the individual layers, the transmission of the message from one layer of the sending first communication unit to another layer of the receiving target communication unit is performed by protocol-specific encapsulation of the message on the sender side, and decapsulation of the message on the receiver side. Encapsulation in this context means that a protocol header, and optionally a footer, is added to the message. An encapsulation step is in particular performed for every layer of the model representing the first communication unit, and a decapsulation step is performed for every layer of the model representing the target communication unit. Taking this into account, a specific duration TKi, i=1 . . . n, can be assumed for each encapsulation step and for each decapsulation step. Said durations TKi summed up determine the first and second time-portion. Transmission time TK is thus given by TK=TK1+TK1+TK1+ . . . TKn.
The second simulator can, for example, calculate in discrete time steps the duration of each step of encapsulation and decapsulation. This can, for example, be done by using state transition systems as models for the communication units.
Although each transmission time TKx, x=1 . . . n, will be carried out in local simulation time steps (which will be synchronized to the global simulation time of the controlling instance), only TK is relevant for the first simulator.
The global simulation time is then set to a third point of time t2=t1+TK by adding the transmission time TK to the second point of time t1. The controlling instance again controls the first and/or the second simulator in that its local simulation times do not overrun the third point of time t2.
As described with regard to the second point of time t1, the local simulation times of the first and second simulator can be controlled identically in relation to the third point of time. The local simulation times are thus advanced to the third point of time t2 when the third point of time t2 is in the future of the current local simulation time. On the other hand, the first and/or second simulator is stopped, i.e., its simulation time is held so as to prevent the respective simulation time to overrun the third point of time.
This can, for example, be performed directly after the measure has been decided on, or when the decision making algorithm has been completed, when the calculation of the data size of the message is made. When the third point of time t1 has then been calculated and is in the future in comparison with the local stopped simulation time, and calculating the equations of the model of the power distribution network has not yet been completed, the first simulator will be resumed until calculation of the equations is completed. The local simulation time of the first and/or second simulator can thereafter be advanced to the third point of time t2.
The message is then analyzed by the controlling instance with respect to the kind of measure (open a switch, taking measurements, reduce power, . . . ) and with respect to the target node in which the measure has to be made.
The controlling instance, based on the information about the target node and the information about the measure, delivers to and initiates at a target processing unit associated with the target node to execute the measure at the third point of time t2. As the target node is the first node, the second node or any other node of the power distribution network, the target processing unit is the first processing unit, the second processing unit, or the processing unit of another node to which it is associated. Executing the measure at the third point of time t2 means that this execution is performed immediately after the message would in reality have been provided to the target processing unit from the target communication unit. Execution of the measure thus begins when the local simulation time of the target processing unit reached the third point of time t2. The local simulation time is alternatively set to the third point of time t2.
For execution, the target processing unit analyzes the measures to be performed on its side as well with respect to the action to be taken as defined in the measure. The target processing unit in particular evaluates the measure with regard to its execution time TACTION, i.e., the duration needed to completely execute the action in reality, e.g., the change of a transformer tap-position can take several seconds due to supportive mechanical processes. A database containing execution times for each possible measure the decision making algorithm can decide on can, for example, be used. This database can be queried, and in return it provides the duration TACTION. The execution process (like the supportive mechanical processes for changing a transformer tap) of the measure can alternatively be modeled in detail as part of the models in the first simulator.
The target processing unit then executes the measure virtually. This means the following:
In reality, the measure can, for example, be turning a switch, reducing power for a specified amount, initiating a measurement, or changing the setpoint of a controller. Performing said action naturally needs time.
On the other hand, performing the measure in the simulated processing unit merely corresponds to the change of one or more status values. For example, virtual execution of the measure can be done by changing one status value representing a switch that enables or disables power flow into a transmission line. Changing this status value would have the meaning of opening the switch. If the target processing unit comprises simulated protective equipment and/or controlling equipment, e.g., a power flow controller, the model describing this equipment, e.g., this power controller, can be calculated.
In order to consider the real time duration of the measure, updating the one or more status values is performed at a fourth point of time t3=t2+TACTION obtained by adding the execution time TACTION to the third point of time t2.
The first simulator then calculates new status values considering the status values updated by the target processing unit. It might be that the fourth point of time t3 is in between two regular time steps tE
At this point, every essential entity of the co-simulation architecture has been executed at least once, and the simulation results can be analyzed as to whether the tested decision making algorithm being part of a protection and control algorithm works well or not, i.e., works successfully according to a specific criteria.
Data exchange between the entities of the co-simulator, i.e., between the first simulator and the controlling instance, the second simulator and the controlling instance, the processing units and the controlling instance, and the communication units and the controlling instance, can be performed by means of the controlling instance. Data exchange in this context refers to updating and/or providing status values as well as providing the processing time TALGO, transmission time TK, execution time TACTION, the communication message, and the data size of the communication message, to that respective entity which needs it.
In an embodiment, the models for simulating the electrical power distribution network, and its nodes/components, respectively, can, for example, be generated automatically by means of a model generator. The models are in particular generated on the basis of the Common Information Model (CIM) according to the standard IEC 61970 (IEC International Electrotechnical Commission).
The CIM can be used to set up models for logical local entities on substation level or field level of the power distribution network. The nodes as well as the processing units can thus be modeled by means of the CIM. The CIM can also be used to set up models for local entities of the communication network, i.e., for the communication units.
The Common Information Model provides a flexible and extensible model description, said model description can be generated based on the given power system topology description mapping the physical entities (the nodes) of the power distribution network to an IEC 61850 based model.
For this purpose, the concept of a Substation Data Processing Unit is introduced, which represents entities on substation level by providing an IEC 61850 (as described in IEC TC57, IEC 61850: Communication networks and systems in substations, International Electrotechnical Commission Std.) based model for mapping logical functions for local process control, implementing decentralized protection and control algorithms and providing additional components for the power systems (e.g., Phasor Measurement Units (PMUs) (as described in B. Milosevic and M. Begovic, “Voltage-stability protection and control using a wide-area network of phasor measurements”, IEEE Transactions on Power Systems, Vol. 18, No. 1, pp. 121-127, February 2003). Each of the first and the second processing unit can be such a SSDPU. A SSDPU can contain the modeled hardware for the corresponding local process control and/or logical methods for decentralized protection, as well as software, i.e., a protection and/or control algorithm.
The processing units (SSDPUs) can, for example, be mapped to the power system simulator using OLE for Process Control (OPC) (as described in OPC Foundation, The OPC Foundation, April 2012, http://www.opcfoundation.org/). They can furthermore be mapped to the communication network simulator using the controlling instance.
The communication network is modeled so as to analyze traffic caused by the information transmitted by control and protection algorithms executed in the data processing units of the electrical transmission system. The communication network can, for example, be as well modeled by the Common Information Model (CIM) as defined in IEC 61970 (as described in A. McMorran, “An Introduction to IEC 61970-301 & 61968-11: The Common Information Model”, Institute for Energy and Environment, Department of Electronic and Electrical Engineering, University of Strathclyde, Glasgow, UK, 2007; and in ENTSOE-E, “Common Information Model (CIM) Model Exchange Profile”, First Edition, ENTSOE-E, 5 2009) and in IEC 61850 for describing the communication between power system elements, i.e., between the first communication unit of the first node and the target communication unit of the target node.
The interfaces to the central controlling instance as part of the co-simulator can in particular be programmed in JAVA-/C++. The interface to the first simulator can be realized with the help of the OPC interface. These interfaces can furthermore be used, e.g., to connect further sub-simulators to the central controlling instance. A sub-simulator in this context can be an executable model of every device in or connected to the power distribution network that is not already considered in the first simulator. For example, a processing unit may be described by means of a sub-simulator. The controller can further provide a central programming interface to connect external decision making processes to the co-simulator, in particular to the central controlling instance. The controller can also associate the logical objects of the CIM model with data structures based on IEC 61850.
The co-simulator can further comprise a central scenario configurator to describe an overall simulation scenario on the basis of the Common Information Model (IEC 61970). This means that common scenarios can be generated for power transmission network and communication network.
The co-simulator performing the computer implemented method according to the present invention can be initialized as follows:
Initialization begins with partitioning an overall model of a given overall system structure for simulating a desired scenario into individual sub-models for the entities corresponding to the system components, said individual models being distributed executable, i.e., on different processors, in particular computer systems.
Said overall model of the system structure is given by defining how many entities and what kind of entities are present in the power distribution network and in the communication network, i.e., how many nodes, processing units, communication units, power transmission lines, and communication transmission media, and what kind of nodes (generator, active or reactive load, active or reactive power controller, switching center, energy converter, batteries, etc.), processing units (protection algorithm, control algorithm, combined protection and control algorithm), communication units (router, repeater, hub, gateway), power transmission lines (AC or DC, 110 kV, 220 kV, 380 kV, 400 kV, 500 kV, 700 kV, 1150 kV, etc.) and communication transmission media (copper cable, optical cable, GSM, UMTS, LTE).
Even the overall model can be described by the Common Information Model (IEC 61970). The overall model is analyzed and firstly partitioned into sub-models, each sub-model corresponding to an entity of the station level or field level of the power distribution network, i.e., corresponding to specific network node. Afterwards, the generated sub-models of the network nodes of the CIM model, e.g., sub-stations, are mapped to IEC 61850 based data structures. In a next step, these data structures are used to connect the interface of a node with the corresponding simulator.
After that, the generated data structures (local instances of the previously defined sub models) are associated to HLA object instances (synchronized data objects within the HLA federation) of the second simulator by means of an instantiation with the help of the HLA object model (template for describing the HLA object instances) and assignment of the corresponding OPC elements (data elements taken from the first simulator and stored within the OPC server).
At this point of time, the processing units are thus considered as software instances. These processing units are called Substation Data Processing Units (SSDPU) in the following, as they represent the secondary equipment of the nodes, i.e., the substations, of the power distribution network. The software instances, i.e., the processing units, follow the data structures of IEC 61850. The software instances are furthermore connected to the first simulator for the power distribution network, and to the second simulator for the communication network. Both the first and the second simulator are based on the same overall model.
Further software instances, i.e., further processing units, can be generated that can be set in interaction with the other entities. A further processing unit may be that of a power plant. So, if another processing unit is present, there will normally be present another network node as well, because each processing unit is associated with a network node. The interaction between the further processing unit and the other entities within the co-simulator can be realized, for example, by enabling the further processing unit to change status values and/or to send or receive messages. A further processing unit can be programmed with additional functions, for example, with an intelligent external decision making algorithm on the station level or on control center level.
The central controlling instance (the HLA) provides in this context a time synchronous execution of all entities of the co-simulator. This can be provided by using common time advancing strategies, using a common object model and/or defining common interaction.
Finally, in the initialization phase, the central control instance is configured. The control instance coordinates communication between the individual entities of the co-simulator and processes events, messages and measures time synchronously.
The first simulator simulating the electrical power distribution network uses a certain minimal step size. This minimal step size depends on the desired degree of exactness for calculating the set of algebro-differential equations describing the power distribution network model. This degree of exactness can be specified manually in advance, e.g., by specifying the number of decimal places to be used. The minimal step size can furthermore also be roughly specified in advance. For example, lms could be a step size. The first simulator then calculates for every time step the step size around the desired step size to reach the desired degree of exactness. Elaborating the given example, for having an average time step of 1 ms, the simulator will use a time step size of, for example, 0.978 ms, 0.99 ms, 1.037 ms, 1.1 ms, etc. However, the time step size is not necessarily chosen in that it has a mean value of 1 ms. It can fluctuate simply because of the calculating algorithm for calculating the set of algebro-differential system. Thus, during the simulation, step size can change dynamically for every next time step.
In contrast to the first simulator, the minimal step size in the second simulator simulating the communication network can have any value as it is triggered event-based. In the processing instances, the minimal step size can also have any value.
Further features, aspects and technical effects of the method according to the present invention are describes in the following with regard to advantageous embodiments and the figures enclosed.
In the following, the hybrid simulation structure of the co-simulator as shown in
To describe the generic architecture, functionalities can be divided into three major modules, which consist of the simulation core, the networking layer, and the management layer. The individual functionalities of the modules will be described in detail in the upcoming subsections.
The simulation core is realized in the central controlling instance. It provides the master event and time control for synchronizing the sub-simulators along with the generic network description, providing the overall network and scenario topology. For the master event and time synchronization, the HLA is used, providing time management services for enabling the sub-simulators to control their logical time synchronized with the other sub-simulators in the federation (as described in C. D. Carothers, R. M. Fujimoto, R. M. Weatherly, and A. L. Wilson, “Design and implementation of HLA time management in the RTI version F.0”, Proceedings of the 29th Conference on Winter simulation, Washington D.C., IEEE, pp. 373-380, 1997). At this, advancing the simulators logical time and keeping synchronicity within the sub-simulator is provided using a conservative synchronization algorithm. The second functionality provided is the generic network description. As the hybrid simulator architecture is driven by the power system, a topology export of the power system provided as a CIM model will be used to describe the connectivity in the electrical network.
The networking layer provides the connectivity to the subsimulators, taking into account various network protocols to enable a generic interaction with third party software tools and simulators. Considering the HLA based simulation core, proprietary sockets as well as standardized web services connections are available to realize the connectivity to the subsimulators. For communication, an eventing service for other sub-simulators is provided to transmit attributable updates as well as interaction request and responses taking into account the HLA object models and transportation services.
The management layer provides comprehensive functionalities for the combined simulation environment. Different modules are here available to manage and operate the hybrid simulator regardless of the affected sub-simulator. As each module provides a different functionality, they are described separately as follows:
Main Configuration
Scenario Configuration
Section III-A) and includes additional information for the ICT topology converter. Especially the investigated communication technologies are here specified for automatic scenario conversion.
Database
Statistical Analysis
Event Logging
Incident Generation
This section introduces the communication architecture model for the power system based on the topology of the electrical transmission grid. Especially the ENTSOE-E 2009 profile for the CIM Model is here a common description language for the power system topology. As nowadays different approaches for harmonization of CIM and IEC 61850 are discussed (as described in Santodomingo, R., Rodriguez-Mondejar, J. A., and Sanz-Bobi, M. A., “Ontology Matching Approach to the Harmonization of CIM and IEC 61850 Standards,” 1st IEEE International Conference on Smart Grid Communications, IEEE, pp. 55-60, 2010), this architecture will benefit from ongoing research by using ontology matching approaches for converting the CIM based grid description in the IEC 61850 based model description stored in the SSDPU. For modeling the communication network, presently available ICT network architecture have been evaluated and taken into account. Analyzing the power systems ability for monitoring, protection and control systems along with its arising wide area communication traffic, entities can be mapped to three different layers, described as follows and illustrated in
Centralized Monitoring and Control Layer
Wide Area Communication Layer
Local Process Layer
For normal operation, a full fiber coverage is assumed to exist along the transmission lines which can be used exclusively. Communication nodes are modeled at substation level for representation of decentralized data processes, e.g., in distributed protection and control systems as well as in PMU data streaming (cf. Section VI). All units at substation level are connected to an ethernet-based local area network and routed to the wide area network using Ethernet-over-SDH (EoSDH). For representation of centralized data processing (e.g., by Supervisory Control and Data Acquisition (SCADA) systems and Super Phasor Data Concentrators (Super PDCs)), control centers are modeled separately and connected to the wide area network using EoSDH as well. Additionally, to wired access technologies, wireless broadband technologies and cellular networks (e.g., Tetra, WiMAX, etc.), will be taken into account as possible candidates for fall-back solutions both as dedicated and shared infrastructures.
The following section presents the HLA federates used in the combined simulation architecture for realizing the power system driven co-simulation. because the upcoming analysis integrates heterogeneous tools and simulators from different fields of research, integration concepts depending on the available interfaces are necessary. The tools to be integrated are thus introduced as follows:
For representation of the electrical transmission grid, an electro-mechanical power system simulation with the commercial simulator DIgSILENT PowerFactory (as described in DIgSILENT, DIgSILENT PowerFactory, April 2012, http://www.digsilent.de/) is applied as it constitutes the most adequate type of analysis for the dynamic behavior of large-scale power systems. As this simulation is based on numerically solving differential equations in discrete time steps (currently down to 1 ms), the simulation needs to be synchronized within the time domain of the federation. Furthermore, DIgSILENT does not provide an interface to the HLA, thus the available API needs to be modified to connect, synchronize and manage the power system data objects within the HLA. An OPC interface is here used providing access to the simulation data and triggering events in the power system simulation. In general, also an electro-magnetic power system simulation with smaller step sizes and even complexer models can be integrated if electro-magnetic transients are of interest. Electro-magnetic simulation is, however, much more computationally intensive and requires data rarely available for larger systems.
The simulation of communication networks is performed by a discrete event simulation. OPNET is here used as a simulator deploying the HLA interface to connect the network simulator to the federation. A scenario generation based on the given CIM model, translating the CIM model into a communication topology assuming optical fiber along the transmission lines, is additionally used to automate the scenario development. For the communication network, the power system behaves like a traffic generator that injects a request for transmission directly into the application layer. The cross-platform communication between the power system simulator and OPNET is modeled as HLA interactions. HLA federates are here able to subscribe to interaction events; the ICT simulator will thus be notified when the power system simulator initiates a transmission. After the transmission has been successfully delivered to the destination node, the ICT simulator will notify the power system simulator by sending another interaction in turn.
For development of protection and control systems, an in-depth analysis of the interplay of a variety of applications and the interdependences with the ICT infrastructure needs to be carried out. These systems use measurement data and information gained by communication from other network devices for decision making and resulting operations in the power system simulation. Such applications can be situated both at the decentralized (e.g., distributed control systems in substations) and the centralized (e.g., contingency-analysis at control centres) level (as described in U. Hager, S. Lehnhoff, C. Rehtanz, and H. F. Wedde, “Multi-Agent System for Coordinated Control of Facts Devices”, in 2009 15th International Conference on Intelligent System Applications to Power Systems. IEEE, pp. 1-6, 2009; and V. Terzija, G. Valverde, Deyu Cai, P. Regulski, V. Madani, J. Fitch, S. Skok, M. M. Begovic, and A. Phadke, “Wide-Area Monitoring, Protection, and Control of Future Electric Power Networks”, Proc. IEEE, Vol. 99, No. 1, pp. 80-93, 2011). They can exhibit a wide range of characteristics, from simple algorithms to intelligent and autonomous systems, being modeled in various software tools (e.g., MATLAB or JAVA-based agents) (as described in C. Rehtanz, “Autonomous systems and intelligent agents in power system control and operation”, Berlin and New York, Springer, 2003), which can be integrated in the substation model introduced in Section VI.
New statistical approaches for online analysis of system data are additionally under development and will be integrated in the co-simulation. E.g., a dynamic clustering of nodes based on an analysis of voltage angles is planned to be integrated allowing an identification of homogeneous subsystems, which in turn enables an intelligent aggregation of monitoring information. For this and other monitoring applications, it is of interest to connect simulation environments for statistical computing and data analysis to the simulation framework. By using HLA JAVA or C++ interfaces, both GNU R as well as MATLAB can be integrated in the federation, offering direct access to the HLA objects.
This section describes the implementation approach, considering the Substation Data Processing Unit (SSDPU) model for coupling the simulators described in Section V. SSDPUs represent the nodes of the power system at the substation level, which forms the link to the physical power system and provides access to wide area communication networks. An overview of the SSDPU model is given in detail within
For creating the IEC 61850 based model description within the SSDPU, a CIM model is provided and parsed by the topology parser inside the substation controller. The given power system model is here divided into multiple substation instances containing detailed model information for local process control. Power system nodes and branches are additionally analyzed and translated into a corresponding communication network topology for close to real-world scenarios.
For instance generation, each substation instance described in the CIM model is instantiated in a single SSDPU instance using the previously generated submodel. Information for local process control is here extracted and mapped into an IEC 1850 based model. Finally, for realizing the interfaces to each sub-simulator, attributes inside the local substation model are mapped to corresponding OPC items for accessing the power system simulation. On the side of the communication network simulator, objects within the HLA federation are generated and published according to the Federate Object Model.
After the HLA mapping is finalized, the control interfaces are set up to manage the time synchronization on part of the HLA federation and to realize the update functions on part of the OPC side. Depending on the type of the attributes, a continuous update of each time step, a static update on initialization, or an event driven update is realized by the given protection algorithms. The overall hybrid architecture has been validated as a proof of concept via JAVA implementations. An exemplary message flow generated during co-simulation together with its validation is given in
In an exemplary power system, at time-stamp t0, a power generator accelerates and increases its active power feed-in (1). Its state is determined by measurement values, which are submitted to the SSDPU via the OPC interface and updated within the corresponding HLA objects. A local protection algorithm detects this trend due to the excess of a reference value of 350 MW at time-stamp t1 and triggers a network transmission to the control center by interacting with the network simulator. As a result, both interactions within the co-simulation and network packets within the network simulation are increasing (2). At time-stamp t2, the information arrives at the control center. Now a centralized protection algorithm calculates the state of the power system and, in case an intervention is needed, triggers additional operations for the power system. If so, these operations are available at t3, transmitted via the network simulator in turn and available at the substation at time-stamp t4. Operations are here executed by local process control recovering the generator to a stable condition, which is shown in (3) and achieved at t5. Due to its oscillating behavior, the generator leads to additional interactions, which are shown in (4) and processed as before.
In this description, a novel concept for a generic hybrid simulation environment was introduced and discussed, which is able to run a combined simulation of power systems and communication networks for validating real-time capabilities of power system protection and control algorithms. The hybrid simulator architecture describing the modules and functions for realizing a combined simulation was introduced first. The substation data processing unit, a CIM based model instance, mapping power system entities to the communication network, was then introduced. The SSDPU concept enables the development and representation of power system protection and control applications both centralized and decentralized. The modular design enables a variety of third party software tools (e.g., MATLAB, GNU R) to be integrated into the simulation environment for future enhancements. The design furthermore offers the mapping of all physical levels of power system operations to dedicated models. For realizing the communication between sub-simulators, interfaces and the application flow for initialization have been described in the course of translating the topology description. The interface with the power system simulator is based on OPC and connecting the communication network simulator is realized by HLAbased objects and interactions, transmitting IEC 61850 based information from the power system to the communication network simulator. For describing the co-simulation, a generic network description approach has been introduced based on the CIM Model. This description is used to map power system entities to the corresponding communication nodes based on standardized description models taking into account ongoing research developments by applying ontology matching of CIM and IEC 61850. This will enable a communication network simulation close to reality. The implementation and evaluation of the simulation architecture is currently under strong development and will be extended in future works.
In the following, main components and related concepts for integrated simulation of power and ICT systems are presented. The novel hybrid simulation design based on IEEE 1516-2000 (High-Level Architecture), IEC 61850, OLE for Process Control (OPC) and the Common Information Model (CIM, IEC 61968/61970) is then presented. The time components are then discussed that are considered for analysis of real-time performance and how they are accounted for in the proposed simulation environment. Simulation results for a test case are then presented. Finally, a conclusion is drawn and an outlook regarding further work is given.
As power systems are in a transition phase adopting new highly dynamic equipment (e.g., HVDC, FACTS and PMUs) as well as new ICT solutions (e.g., high-performance computing and communication), it becomes increasingly important to address both domains at the same time and to gain insights regarding the real-time capabilities of new approaches.
A brief overview on core components that need to be covered in an in-depth analysis of dynamic power system operation is first given. Secondly, existing simulative solutions already addressing the integrated analysis of power systems and ICT are presented. Among these, it can be distinguished between comprehensive simulation environments and co-simulation.
Power systems monitoring, protection and control systems can be located at three levels (as described in C. Rehtanz, “Autonomous systems and intelligent agents in power system control and operation”, Berlin and New York, Springer, 2003): at the lowest level, the bay level, measurements and actions in the electrical transmission system are carried out. Here, also local Intelligent Electric Devices (IEDs) such as local protection devices are situated. The bay level is linked with the next higher level, the substation level, via Local Area Networks (LAN). On the substation level, data is processed for all connected bays and a communication gateway to a Wide-Area Network (WAN) is available that might consist of several interconnected communication layers, e.g., WANs based on optical fiber or wireless technology.
Via WAN, substations can communicate with other substations and entities, e.g., central control centers that collect system-wide information and apply a variety of software tools for system management. These software tools can feature a variety of centralized applications for power system management (as described in V. Terzija, G. Valverde, Deyu Cai, P. Regulski, V. Madani, J. Fitch, S. Skok, M. M. Begovic, and A. Phadke, “Wide-Area Monitoring, Protection, and Control of Future Electric Power Networks”, Proc. IEEE, Vol. 99, No. 1, pp. 80-93, 2011). Also, a variety of smart grid members located externally could be connected via WAN. This could, e.g., include communication for power plant management, demand side management, or management of electric vehicles. Decentralized protection and control applications can also be applied at the substation level which are solely based on the interaction of substations without a central control entity. Any entity beyond bay and substation level that has access to the smart grid is in the following referred to as belonging to the wide-area level.
For integrated simulation of several simulators, typically one of two concepts is applied: co-simulation or comprehensive simulation. In the following existing approaches of these concepts are presented that address power systems and ICT simulation.
1) Comprehensive Simulation: A comprehensive simulation for the analysis of both domains combines power system and communication network simulation in one environment. The challenge here is to bring together both system models and solving routines which leads either to integrate power systems simulation techniques into a communication network simulator or vice versa. K. Mets, T. Verschueren, C. Develder, T. Vandoorn, and L. Vandevelde, “Integrated simulation of power and communication networks for smart grid applications”, IEEE 16th International Workshop on Computer Aided Modeling and Design of Communication Links and Networks (CAMAD), pp. 61-65, June 2011 describes such an integrated smart grid simulation framework using OMNeT++ as a development platform. In this approach, the electrical distribution network has been modeled in MATLAB and linked into the OMNeT++ simulation.
2) Co-Simulation: The second concept for integrated simulation of both domains is co-simulation. In co-simulation, networks are analyzed by their own dedicated simulators and are brought together by appropriately designed interfaces as well as coordinated simulation management. The first approach in applying a co-simulation environment for power systems and communication networks has been the electric power and communication synchronizing simulator EPOCHS (as described in K. Hopkinson, X. Wang, R. Giovanini, J. Thorp, K. Birman, and D. Coury, “EPOCHS: A Platform for Agent-Based Electric Power and Communication Simulation Built from Commercial Off-The-Shelf Components” IEEE Transactions on Power Systems”, Vol. 21, No. 2, pp. 548-558, May 2006; and K. Hopkinson, K. Birman, R. Giovanini, D. Coury, X. Wang, and J. Thorp, “Epochs: integrated commercial off-the-shelf software for agent-based electric power and communication simulation,” Proceedings of the 2003 International Conference on Machine Learning and Cybernetics (IEEE Cat. No. 03EX693), IEEE, pp. 1158-1166, 2003). The present invention connected the power systems simulators PSLF and PSCAD/EMTDC to the open source Network Simulator 2 (NS2), realizing the interconnection between simulators using a Run-Time Infrastructure (RTI) based on the High-Level Architecture (IEEE 1516-2000). The system is focused on simulating communication among software agents and does not investigate IT execution and standards for communication in power systems. Furthermore, no information regarding genericity and computational performance has to date been published. It is believed that EPOCHS constitutes the most advanced co-simulation approach so far. Further approaches have been presented for connecting the commercial power system simulator PSS™ Netomac to the Network Simulator 2 (NS2) via a JAVA interface (as described in J. Bergmann, C. Glomb, J. Gotz, J. Heuer, R. Kuntschke, and M. Winter, “Scalability of Smart Grid Protocols: Protocols and Their Simulative Evaluation for Massively Distributed DERs”, First IEEE International Conference on Smart Grid Communications, pp. 131-136, October 2010) and for linking the open source simulators OpenDSS and NS2 in order to simulate dispatching messages in NS2 and to generate scripts in the power system simulation. For these approaches, it appears difficult to expand them by other sub-simulators, and they lack the use of a time-synchronizing middle-ware as in EPOCHS for distributed simulations.
In this section, the novel hybrid simulator design and reference to current implementation is presented. After giving an overview on the architecture, the explanation follows the structure of the three physical levels in power transmission systems discussed in subsection II-A.
The High Level Architecture (HLA) (as described in IEEE Standard for Modeling and Simulation (M&S) High Level Architecture (HLA)—Framework and Rules”, IEEE Std. 1516-2010 (Revision of IEEE Std. 1516-2000), pp. 1-38, 18, 2010; “IEEE Standard for Modeling and Simulation (M&S) High Level Architecture (HLA)—Federate Interface Specification”, IEEE Std. 1516.1-2010 (Revision of IEEE Std. 1516.1-2000), pp. 1-378, 18 2010; “IEEE Standard for Modeling and Simulation (M&S) High Level Architecture (HLA)—Object Model Template (OMT) Specification”, IEEE Std. 1516.2-2010 (Revision of IEEE Std. 1516.2-2000), pp. 1-112, 18, 2010) is a generic approach for realizing a synchronized and distributed simulation environment which has been standardized in IEEE 1516 in 2000. A distributed simulation using HLA is based on the following components: a Run Time Infrastructure (RTI) serving as a central administration instance, the HLA rules defining and describing the interaction of sub simulators (Federates), Interface specifications (RTI and Federate Ambassadors) and an Object Model Template (OMT). Although the concept of the HLA has been driven by the former Defense Modeling and Simulation Office (DMSO, nowadays Modeling and Simulation Coordination Office) of the US Department of Defense and has mainly been focused on military simulation, various approaches have come from the field of distributed simulation and are applicable to both communication network and power system simulation.
In our hybrid simulation architecture, time synchronization and object management have been specified for the use of HLA. The communication network simulator and any communicating entity in the power system are integrated as HLA federates. Federates can interact and exchange objects, and they can be run in a distributed environment enabling computational performance. Time synchronized simulation of the event-based communication network and the discrete time steps of the power systems simulation is provided by the HLA.
Representing the physical power system, at the bay level measurements corresponding to primary equipment are captured, local control and protection are in place, and actions (e.g., change of transformer tap settings) can be executed. The underlying power system can be derived from any Common Information Model (CIM), but needs to be extended by certain dynamic parameters and models of local devices such as PMUs, controllers, and protection devices. Besides the network model, the power system simulator receives a scenario from a central configurator (including eed-ins, loads, events such as short-circuits, etc.) in the initialization phase. All processes at the bay level are simulated in discrete time steps in a power system simulator (realized in steps of 10 ms with DIgSILENT PowerFactory). Every 10 ms, simulation data is provided via an OPC interface to an OPC Server (realized with MatrikonOPC Server, as described in S. C. Mtiller, U. Hager, H. Georg, S. Lehnhoff, C. Rehtanz, C. Wietfeld, H. F. Wedde, and T. Zimmermann, “Einbindung von intelligenten entscheidungsverfahren in die dynamische Simulation von elektrischen Energiesystemen,” in Einbindung von intelligenten Entscheidungsverfahren in die dynamische Simulation von elektrischen Energiesystemen, D-A-CH-Konferenz Energieinformatik 2012, Oldenburg, Germany, 2012). Using this industrial standard offers the flexibility to integrate real substation hardware (e.g., PMUs) in the simulation environment.
In the initialization phase, a Substation Data Processing Unit (DPU) for each node in the CIM model is initialized (implementation in JAVA), representing all functions available at substation level. It is also registered as a federate of the HLA. As a basic function, a data set of current information from the equipment in all bays at that node is created. For this, all equipment at the node is identified by the CIM model and for each piece of equipment a data object based on IEC 61850 is created. Ontology matching of CIM and IEC 61850 is subject of current research (as described in Santodomingo, R., Rodriguez-Mondejar, J. A., and Sanz-Bobi, M. A., “Ontology Matching Approach to the Harmonization of CIM and IEC 61850 Standards”, in 1st IEEE International Conference on Smart Grid Communications, IEEE, pp. 55-60, 2010; and R. Santodomingo, J. Rodriguez-Mondejar, M. Sanz-Bobi, S. Rohjans, and M. Uslar, “Towards the automatic alignment of cim and scl ontologies”, 2011 IEEE International Conference on Smart Grid Communications (SmartGridComm), IEEE, pp. 422-427, 2011). It enables the realistic mapping of substation data to communication protocols and further strengthens the ability to integrate real hardware. The Substation DPU can receive data from the OPC Server and can also trigger events in the power system simulation via OPC. Besides the data set, any functions at substation level can be integrated as distinct modules of the DPU. These include decentralized protection and control systems (e.g, S. C. Müiller, U. Hager, C. Rehtanz, and H. F. Wedde, “Application of self-organizing systems in power systems control”, Lecture Notes in Computer Science, D. Hutchison, T. Kanade, J. Kittler, J. M. Kleinberg, F. Mattern, J. C. Mitchell, M. Naor, 0. Nierstrasz, C. Pandu Rangan, B. Steffen, M. Sudan, D. Terzopoulos, D. Tygar, M. Y. Vardi, G. Weikum, O. Dieste, A. Jedlitschka, and N. Juristo, Eds. Berlin and Heidelberg, Springer, Berlin Heidelberg, pp. 320-334, 2012; and U. Hager, S. Lehnhoff, C. Rehtanz, and H. F. Wedde, “Multi-agent system for coordinated control of facts devices”, 2009 15th International Conference on Intelligent System Applications to Power Systems, IEEE, pp. 1-6, 2009) or concentration of PMU data in a Phasor Data Concentrator (PDC). The modules could also exhibit functionalities from additional software tools such as MATLAB or Multi-Agent-Systems. Regarding external communication, the DPUs can send messages to other DPUs, control centers or any other external entities. For sending a message, a HLA interaction containing the recipient and the message content is triggered which is then processed by the communication network simulator.
All wide-area communication is simulated with an event-based communications network simulator (realized with OPNET). The communication network topology can be solely derived from the CIM model (e.g., when assuming fiber optics along all transmission lines) or be extended by specific topology information including fallback technologies, e.g., dedicated wireless communication networks.
As the core component of system-wide control, a JAVA instance for a control center (or any additional identity) is initialized at the wide-area level and registered as a federate of the HLA. As for substation DPUs, this instance features a data model (based on CIM and IEC 61850 in order to easily process data from substations) and any functionality of interested (possibly by secondary tools). This is the typical point of monitoring and decision making by centralized applications for monitoring, protection and control as well as of a PDC collecting system-wide information (SuperPDC). Messages (e.g., including control and protection actions) are sent to substation by triggering a HLA interaction followed by the simulation of its processing in the communication network. Any additional smart grid participants, such as generation units or loads, can also be initialized at this level.
A main purpose of the hybrid simulation is the evaluation of real-time performance of applications in power systems operation. In particular, the time elapsed between the measurement of a critical value in the power system (e.g., overload) and successful reaction by the application (e.g., protection applications disconnecting a line, control applications relieving overload by controlling power flows) is a subject of interest. It is important to clearly define what is considered to be a successful performance. As an example, in the case of a power flow control application, not the time until the first control response could be relevant but the time until the overload is reduced below a critical level. The relevant times that impact this total time to successful reaction are categorized as follows:
These components are taken into account in the hybrid simulation modules as described in the following and as visualized in
At each IT component (e.g., substation hardware, control center IT systems), times for execution of calculations tIT;calc as well as waiting tIT;wait can also be accounted for by queuing. Execution times for algorithms tIT;calc must be estimated or evaluated beforehand as the algorithm in the hybrid simulation runs in a simulation environment and the performance of the real hardware component is decisive. If the algorithm features time-based requirements for its inputs (e.g., use of measurement data being more recent than an event message), the times for waiting tIT;wait for all necessary input data evolve from the interaction of the hybrid simulation modules involved in providing the measurement data. This could also involve the concentration of PMU data by a PDC.
The timely response of primary power system equipment tPE, e.g., time between receiving the signal for changing a tap position of a transformer and having successfully changed the tap mechanically, is accounted for in the equipment modeling in the power system simulator. The length of tPS results from the power system trajectory evolving due to all module interactions in the dynamic power system simulation and depends on what had been defined as a successful reaction of the application (e.g., also accounting for time until steady state is reached).
Having accounted for the aforesaid timely components, a real-time performance analysis close to reality can be carried out with the presented simulation architecture. A central scenario generator for both simulators based on the CIM system model enables the efficient analysis of a set of scenarios. Practical problems that can be analyzed include:
The simulation results gained from a set of scenarios can furthermore be evaluated by real-time calculus methods and conclusions about performance levels certainly achieved by the applications can be drawn. The inclusion of real-time calculus analysis on base of the results gained with the simulation architecture presented here is subject of ongoing research and will be published in the near future.
In the following, simulation results for a generic scenario are shown as a proof of context and to illustrate fundamental functionalities of the simulator observing a simple example. Detailed analyses of smart grid applications in settings close to reality, in particular of integrated WAMPAC systems, will be published in the future.
For the present invention, the New York Test System (IEEE 39-bus 10-machine system) extended by three Power Flow Contollers (PFCs, in this case Phase Shifting Transformers (PSTs)) and one High-Voltage-Direct-Current (HVDC) line was investigated. The system is depicted in
The scenario is set as follows: at t=2 s the load at node 15 is disconnected causing an overload on transmission line TL0414 between nodes 4 and 14. PST3 is controlled by a decentralized control system located at substation 4 using PMU measurements from substation 14 as input for determining the loading of line TL0414. If the loading of the line exceeds 80%, an intervention is created by sending a control message to the PST (PST3) to increase the tap position by one in order to relief the stress on the line. The PST needs a time of tPE=6 sec. to change the tap position. The data processing unit at substation 14 waits tIT; wait=10 sec. before sending another measurement to substation 4. A successful response of the control system is assumed to be reaching steady state load flow below a loading of 80%.
As an underlying communication network topology, the substation was assumed to contain a substation controller node, a local network switch, and a wide area network access router, which are connected by 10 Mbit/s Ethernet. For wide area connectivity, a DS1 link (digital signal with a data rate of 1.55 Mbit/s between substations) was furthermore assumed. The network protocol applied for message exchange is proprietary and transmits UDP (User Datagram Protocol) based messages with a constant size of 240 bytes. Last but not least, as a worst case scenario, the communication network was assumed not to be initialized so that no network routes exists initially. Times due to substation LAN (tCT;LAN) and IT execution tIT;calc for the simple algorithm are assumed to be very small and are neglected for this example.
The results of the proof of concept scenario given by
At the beginning, the scenario is in a steady state and the communication network is not yet initialized. At timestamp t=2 sec., a load at node 15 is disconnected causing an overload on transmission line TL0414 between node 4 and 14. The PMU at substation 14 determines the state of line TL0414 instantly and sends a message to substation 4 when the loading exceeds 80%. As the communication network has not yet been initialized, this control message is dropped, but routes over the wide-area network are established by routing protocols. Due to the dropped message (and also due to the IT logic of only sending measurements every 10 sec.), a delay of t=10 sec. in the response of the control system is caused (indicated by tCT;WAN in
In this description, a new generic hybrid simulation architecture for smart grids which is able to run a time synchronized and distributed co-simulation of power systems as well as communication networks and to account for IT processes and execution was proposed. In contrast to the most advanced existing co-simulation proposal of (as described in K. Hopkinson, X. Wang, R. Giovanini, J. Thorp, K. Birman, and D. Coury, “EPOCHS: A Platform for Agent-Based Electric Power and Communication Simulation Built from Commercial Off-The-Shelf Components”, IEEE Transactions on Power Systems, Vol. 21, No. 2, pp. 548-558, May 2006; and K. Hopkinson, K. Birman, R. Giovanini, D. Coury, X. Wang, and J. Thorp, “Epochs: integrated commercial off-the-shelf software for agent-based electric power and communication simulation”, Proceedings of the 2003 International Conference on Machine Learning and Cybernetics (IEEE Cat. No. 03EX693), IEEE, pp. 1158-1166, 2003), the hybrid simulation presented in this embodiment takes into account all levels of power system operation as well as the IT processes explicitly. This enables detailed analyses of the real-time performance of smart grid applications in joint interaction of components at different system levels and of different applications. E.g., performance by use of different algorithms, different ICT infrastructure, new power system equipment, as well as impact of additional smart grid members (e.g., electric vehicles) can be evaluated. Benefits of the new approach furthermore include its generic design based on CIM and IEC61850 for the configuration of both simulators and the HLA Federates. This facilitates a quick adaping to new system models. Second, the flexibility for integrating a huge variety of modules by use of the HLA enables the deployment of the simulation environment for various smart grid applications modeled by a wide range of software tools. Third, the observance of industrial standards such as OPC and IEC61850 in the simulator development and ontology matching of CIM and IEC 61850 allows for simulations close to reality and facilitates the integration of, e.g., substation hardware. The implementation and evaluation of this simulation architecture is currently under strong development, and advances as well as reports on performance will be given in future work. As the current focus is set on investigating applications of monitoring, protection and control in transmission systems and their respective real-time performance, exemplary simulation results for a simple communication based control system for PSTs are shown as a proof of concept. In the near future, the simulator will be used for detailed analyses of the joint real-time interaction of WAMPAC applications developed in DFG research unit FOR1511. These will include the investigation of different communication infrastructures including fallback solutions as well as the use of the simulation results for real-time calculus methods to determine worst-case performance levels of the applications.
The present invention is not limited to embodiments described herein; reference should be had to the appended claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 12007104.8 | Oct 2012 | EP | regional |
This application is a U.S. National Phase application under 35 U.S.C. §371 of International Application No. PCT/EP2013/002067, filed on Jul. 11, 2013, and which claims benefit to European Patent Application No. 12007104.8, filed on Oct. 12, 2012. The International Application was published in English on Apr. 17, 2014 as WO 2014/056558 A1 under PCT Article 21(2).
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/EP2013/002067 | 7/11/2013 | WO | 00 |
