This application is the US National Stage of International Application No. PCT/EP2010/058878, filed Jun. 23, 2010 and claims the benefit thereof. The International Application claims the benefits of German application No. 10 2009 030 322.7 DE filed Jun. 24, 2009. All of the applications are incorporated by reference herein in their entirety.
The invention relates to a method for controlling a combustion process, in particular in a firing chamber of a fossil-fuel-fired steam generator, wherein spatially resolved measured values are determined in the firing chamber. The invention further relates to a corresponding combustion system.
In the combustion process of a steam generator the fuel is prepared in a first stage (e.g. pulverizing of the coal in the coal pulverizer, preheating of the heating oil or similar) and then supplied in a controlled manner together with the combustion air to the combustion chamber in accordance with the current heat requirement of the installation. In this case the fuel is introduced into the firing chamber at different points of the steam generator at what are termed the burners. The air is also supplied at different points. A supply of air also takes place at all times at the burners themselves. In addition there can be supplies of air at points at which no fuel flows into the firing chamber.
The object is therefore to manage the combustion process in such a way that it executes in the most efficient manner possible with minimum wear and tear and/or with the lowest possible emissions. The typical key influencing parameters for the combustion process of a steam generator are:
These influencing parameters are usually set at the time of commissioning of the steam generator. At this time, depending on boundary operating conditions, various optimization targets are prioritized, such as maximum plant efficiency, minimum emissions (NOx, CO, . . . ), minimum carbon content in the ash (completeness of the combustion). However, constant monitoring and adjustment of the combustion process is necessary due to the variability of the process parameters over time—in particular the fluctuating properties of the fuel (calorific value, air requirements, ignition behavior, etc.). In industrial installations the combustion is therefore monitored by means of measurement instrumentation and the available influencing parameters are modified by means of closed-loop control interventions in accordance with the currently detected combustion situation.
However, the influencing parameters are varied only to a very limited extent during operation of the plant. The reason for this is that due to the high temperatures, as well as the environment that is characterized by high levels of chemical and mechanical attrition, only few measurement results of adequate quality or even none at all are available from the immediate combustion environment. As a consequence only measured data recorded in the flue gas path far away from the combustion can be called upon for regulating the combustion. The process data is therefore available only with a delay and without specific reference to the individual actuating elements for closed-loop control optimizations. Furthermore, owing to the large dimensions of large-scale firing plants the available point measurements are often not representative and fail to reflect a differentiated picture of the real spatial process situation.
Since in many cases no closed-loop control or optimization of the combustion process is possible, the process parameters (e.g. excess air) are set at a sufficient distance from the technical process limits. This causes losses due to operation at a reduced level of process efficiency, higher levels of wear and tear and/or higher emissions.
A possibly present closed-loop control and optimization of the combustion process is performed according to the present prior art using different approaches:
An even more extensive optimization of the combustion is made possible if a spatially resolving measurement system is used with the aid of which measured data from the immediate vicinity of the combustion can be made available.
It is the object of the present invention to disclose an improved method for controlling a combustion process, wherein spatially resolved measured values in the firing chamber are used. A further object is to disclose a corresponding combustion system.
These objects are achieved by means of the features of the independent claims. Advantageous embodiments are set forth in the respective dependent claims.
The essential features of the invention can be summarized as follows:
The invention therefore employs an improved means of acquiring the current status of firing processes through the use of at least one measurement technology with spatially resolving measurement space for the purpose of quantitatively determining the combustion products following the combustion in the interior of the industrial firing plant in order to achieve a more differentiated and faster closed-loop process control. A significant advantage of the invention resides in the fact that the complex measured value distributions of the spatially resolving measurement technology can be processed through the transformation to simple state or controlled variables with the aid of conventional controllers. Furthermore, as a result of the inverse transformation the output signals of the conventional controllers are distributed among the manipulated variables present in accordance with a predefined optimization target. An optimal interaction is therefore achieved between the newly defined closed-loop control concepts and the installed complex measurement technology. In particular, however, a combustion process executing in the most efficient manner possible with minimum wear and tear and/or with the lowest possible emissions is realized by means of the control structures that have been improved in the manner described.
In a first embodiment variant the state variables are determined on the basis of statistical information of the spatially resolved measured values. This has the advantage that in this case the enormous diversity of the information relating to, for example, the existing temperature or concentration distributions can be compressed. Weightings can be introduced and other image processing methods can be applied. A further advantage is that in this way process variables are produced by means of which the combustion process can be described and controlled.
Further embodiment variants relate to the determination of setpoint values. The advantage in the specification of the setpoint values is that an optimization target can be predefined in concrete terms and in a generally intelligible manner. As a result an unambiguous and reproducible description of the desired optimal plant behavior is obtained. The plant operator then has the possibility at any time to redefine the optimal operating point by varying the setpoint values, e.g. to attach a higher weight to minimum emissions at the expense of a somewhat poorer level of efficiency.
In one embodiment variant the distribution of the controller outputs among the actuating elements is optimized with the aid of a neural network. The corrective control interventions can furthermore be finely adjusted with the aid of the neural network. By this means a particularly intelligent and precise closed-loop control is achieved which is robust against variations in external influencing factors, e.g. variable fuel quality.
The invention is explained in more detail below with reference to an exemplary embodiment illustrated in the drawing, in which the
FIGURE shows a schematic diagram intended to illustrate the closed-loop combustion control system according to the invention.
The firing chamber FR of a power plant or another industrial installation in which a combustion process takes place is equipped with a spatially resolving measurement system (designated by MS in the FIGURE). It is possible here to employ any measurement systems with the aid of which measured data from the immediate vicinity of the combustion is made available. Examples of such measurement systems are:
In selecting the measurement technology it is crucial that it is suitable for determining important properties of the combustion with spatial resolution. In this case measurements are carried out for example on a cross-section of the firing chamber close to the combustion process. The determined measured values characterize the combustion on the basis of properties such as e.g. local concentrations (CO, 02, CO2, H20, . . . ) and temperature.
In all cases a multiplicity of the most disparate measured values is obtained as a function of spatial coordinates. Thus, instead of individual measured values, entire measured value distributions similar to a two- or three-dimensional pattern are present at the input of the closed-loop control system according to the invention.
In the course of a variable transformation VT said data, identified by M measured values MW in the FIGURE, is converted in a first step into state variables which can be used for closed-loop control purposes. In the process the spatial information relating to the combustion chamber is mapped onto individual characteristic parameters and accordingly compressed.
In order to derive the different state variables from the spatial measurement information, the following points are typically evaluated:
a) Weighted average values with accentuation or suppression of parts of the space recorded by the measurement technology means,
b) the average value of the measured variable over the space recorded by the measurement technology means,
c) spatial position of the center of mass of the measured values,
d) statistical characteristic parameters for spatial distribution patterns.
An optimization target can be defined as a setpoint value for the state variables which can be used for closed-loop control purposes. In addition said state variables, in conjunction with conventional measurement and process information that is available for process control purposes, characterize the current operating status of the combustion process.
As a result of the variable transformation VT described an arbitrary number of M measured values MW is accordingly converted into an in turn arbitrary number of N controlled variables RG, where M and N represent natural numbers and N is typically less than M. The controlled variables RG are state variables which are subsequently used as actual values for individual controllers.
The N controlled variables are supplied to N controllers R. This is illustrated in the FIGURE with the aid of the closed-loop control module which contains a subtractor and further modules which can be used for closed-loop control purposes such as a PI controller, for example. In this context said module is a conventional closed-loop control module which may possibly already be present in the industrial installation that is to be controlled. It can also be a multivariable closed-loop control module, depending on embodiment variant. The closed-loop control module under consideration here additionally has an input ESW for the setpoint value of the derived state variable. This is either specified manually, is constant or is specified as a function of load and is intended to characterize the desired operating behavior. In addition to the input ERG for the controlled variable RG there also exists a further input EPG for further arbitrary measured process variables PG which are acquired outside of the spatially resolving measurement system. The deviation between the setpoint and actual value is formed inside the controller, the deviation is varied by means of the further measured process variables, e.g. in order to adjust the controller gain as a function of the current load situation, and supplied to the existing controller (a PI controller in this case) which determines the necessary changes to manipulated variables. This signal is present at the output ARA of the controller.
If there are now N controllers present, then at this point there exist N values for the control outputs RA (cf. FIGURE). The aim now is to convert said signals RA of number N referred to as control outputs in an inverse transformation RT in such a way that a specific number of K actuating elements in each case receive the actuating signal which is necessary for achieving the control target. In other words it is now necessary to derive, from the control outputs RA of the N controllers R, control interventions for different actuating elements by means of which the combustion process can be beneficially influenced. In this case a control intervention can be applied to a plurality of actuating elements at different degrees of intensity.
Examples of actuating elements are the openings of dampers arranged in the combustion chamber.
The allocation of N control outputs to K actuating elements takes place in the calculation unit RT (where N, K are each natural numbers). Measured process variables PG that are acquired outside of the spatially resolving measurement system are also taken into account here. It is of particular advantage in the inverse transformation of the controller outputs to the existing manipulated variables that the controller outputs are allocated to the actuating elements in an optimal manner so that e.g. the emission values can be minimized and yet at the same time a highest possible level of efficiency of the installation is reached. This is achieved in the present exemplary embodiment in that the calculation unit RT is also supplied with optimization values OW from the optimizer OPT. The optimizer receives information from different areas.
In addition to measured process variables that are acquired outside of the spatially resolving measurement system, the optimizer can also receive measurement results of the spatially resolving measuring instruments arranged in the combustion chamber. In the course of the variable transformation VT′ a number M′ of the spatially resolved measured values is converted into an arbitrary number N′ of state variables which are supplied to the optimizer OPT. These can be the same measured values as described hereintofore, although alternatively other measured values can also be used. The optimizer OPT can optionally be connected to a neural network NN.
In this case a hybrid closed-loop control structure consisting of conventional closed-loop control modules and neural networks is realized. The neural network is trained with measured process variables and serves as a specific model for predicting the firing behavior. On the basis of the firing response predicted by the neural network an iterative optimization algorithm determines the optimal distribution of the control interventions among the actuating elements as well as correction values for the actuating elements. By this means the process is optimized in accordance with a predefined target function.
The optimization values OW can also be trim factors, for example. The results of the inverse transformation RT are weighted, shifted and adjusted by means of the trim factors taking into account the optimization process in accordance with the desired control target.
Finally, a total manipulated variable calculation GSB for the K actuating elements present takes place on the basis of the output values of the inverse transformation and where applicable taking further account of the result from the optimization process. The different control interventions applied to different actuating elements by different identified setpoint value deviations are superimposed additively on one another to produce an overall control intervention for each actuating element. At the end of the algorithm, K manipulated variable changes ST are forwarded to the individual actuating elements such as dampers or fuel feed devices.
During the entire closed-loop control method, the speed and magnitude of the individual control interventions are adapted to the given technical boundary conditions and limits of the industrial installation. Limits predefined by the process are not exceeded.
Number | Date | Country | Kind |
---|---|---|---|
10 2009 030 322 | Jun 2009 | DE | national |
Filing Document | Filing Date | Country | Kind | 371c Date |
---|---|---|---|---|
PCT/EP2010/058878 | 6/23/2010 | WO | 00 | 2/7/2012 |
Publishing Document | Publishing Date | Country | Kind |
---|---|---|---|
WO2010/149687 | 12/29/2010 | WO | A |
Number | Name | Date | Kind |
---|---|---|---|
5347446 | Iino | Sep 1994 | A |
5408406 | Mathur | Apr 1995 | A |
5493631 | Huang | Feb 1996 | A |
5822740 | Haissig | Oct 1998 | A |
6532454 | Werbos | Mar 2003 | B1 |
6551094 | Fastnacht | Apr 2003 | B2 |
6553924 | Beaumont | Apr 2003 | B2 |
6712012 | van Kessel | Mar 2004 | B1 |
6752093 | Schirmer | Jun 2004 | B2 |
7200520 | Wolfgang | Apr 2007 | B2 |
7637735 | Wintrich | Dec 2009 | B2 |
8219247 | Esmaili | Jul 2012 | B2 |
20070122757 | Widmer | May 2007 | A1 |
20080147241 | Tsangaris | Jun 2008 | A1 |
20110113993 | Esmaili | May 2011 | A1 |
Number | Date | Country |
---|---|---|
4220149 | Dec 1993 | DE |
102006022626 | Nov 2007 | DE |
1091175 | Apr 2001 | EP |
0815397 | Oct 2003 | EP |
1850069 | Aug 2008 | EP |
2313815 | Dec 2007 | RU |
WO 9840673 | Sep 1998 | WO |
WO 0016010 | Mar 2000 | WO |
Number | Date | Country | |
---|---|---|---|
20120125003 A1 | May 2012 | US |