Patent Application
20240418793
This application claims priority of European patent application no. 23180031.9, filed Jun. 19, 2023, the entire content of which is incorporated herein by reference.
The disclosure relates to health status of a pitch energy storage unit of a wind turbine, a method for charging a pitch energy storage unit of a wind turbine, and a pitch control system of a wind turbine.
Wind turbines are widely known from the prior art and are used to convert wind energy into electrical energy. Among others, wind turbines may include an energy storage unit to provide all or some parts of the wind turbine with electrical energy, for instance, in case the wind turbine does not generate sufficient energy and is at least temporarily disconnected from an external energy source, such as a power grid, or during situations like grid fault ride-through (FRT).
In particular, a wind turbine may include a so-called pitch energy storage unit, which stores sufficient electrical energy to control a pitch angle of rotor blades of the wind turbine. For example, in case a previously deactivated wind turbine needs to be activated or started, a pitch drive may change the pitch angle of the rotor blades to start the wind turbine. Other operations in which power is provided from the pitch energy storage unit may include (self) testing functions of the wind turbine.
Operational phases, in which power is provided by a pitch energy storage unit, are infrequent and relatively short compared to normal operational phases of a wind turbine, in which the wind turbine provides electrical power to a power grid or is deactivated. However, reliable provision of power is crucial in such operational phases, as this enables, inter alia, the safe activation and deactivation of the wind turbine in exceptional circumstances, such as power grid failures and/or extreme weather situations like gusts.
Like many other components, the performance of pitch energy storage units deteriorates over time. For example, rechargeable batteries used as energy store of a pitch energy storage unit have a limited lifetime in terms of age, charging cycles, et cetera. However, detecting defective storage devices well in advance is challenging, and often requires additional hardware, computational capabilities, effort and costs for monitoring them. Thus, in practice, the entire pitch energy storage unit or its internal energy storage components are typically replaced at regular service intervals to avoid unintentional outages of and/or damages to the wind turbine due to insufficient energy being stored in the pitch energy storage unit. Replacing a pitch energy storage unit or its internal energy storage components is costly, resource-intensive and may be dangerous. In addition, it typically requires the wind turbine to be deactivated or stopped for service, which therefore, reduces the amount of energy being provided to the power grid.
Accordingly, there is a need to monitor a health or remaining lifetime of a pitch energy storage unit and/or to detect defective storage components included therein at an early stage to avoid unintentional outages of and/or damages to the wind turbine and, at the same time, prevent unnecessary service and/or replacement of a pitch energy storage unit.
According to a first aspect of the present disclosure, a method for estimating a health status of a pitch energy storage unit of a wind turbine is provided. The method includes:
Among others, the inventor has found that a health status of a pitch energy storage unit can be reliably estimated based on the provision of an equivalent circuit model (ECM) and incremental updating of ECM parameters of the pitch energy storage unit. At the same time, the disclosed method is computationally efficient and can be implemented, for example, using an existing controller of the wind turbine, such as a controller of a pitch converter. Thus, health status monitoring and predictive maintenance can be implemented with no or very limited extra effort.
Contrary to known methods, the disclosed method does not require the provision of additional components, such as specialized sensors or measuring devices, the disconnection of the pitch energy storage unit, for example, to determine an open circuit voltage or resistance of the pitch energy storage unit, or to perform a complete discharge and/or charge cycle of an energy storage element contained therein. Instead, the health status of the pitch energy storage unit can be determined in-situ when the turbine is in operation and, preferably, continuously, so as to minimize and schedule servicing as needed, while preventing unintentional outages.
In at least one implementation, the optimization procedure is based on a predetermined number of sample tuples, each sample tuple including a present voltage sample, a corresponding time derivative du/dt of the voltage u, a present current sample im, and a corresponding time derivative dim/dt of the current im. The optimization procedure is performed when a predetermined number of sample tuples has been stored in a buffer storage, so as to ensure sufficient data for a successful optimization.
In at least one implementation, the method further includes determining whether a present sample tuple meets at least one predetermined filter criterion, in particular, if the present sample tuple differs from all sample tuples contained in the buffer storage. If the at least one predetermined filter criterion is met, adding the present sample tuple to the buffer storage. Otherwise, discarding the present sample tuple. By filtering out unsuitable samples, such as duplicate samples, the quality of the optimization can be improved.
In at least one implementation of the method, the optimization procedure is performed multiple times, and the method further includes: selecting starting values for a set of optimization parameters optimized by the optimization procedure; determining at least one quality parameter, such as a Residual Square Sum RSSnew, associated with updated values for the set of optimization parameters obtained after performing the optimization procedure starting from the starting values; and if the at least one quality parameter RSSnew associated with the set of optimization parameters meets at least one quality condition, computing the updated resistance of the equivalent series resistor and the updated at least one model parameter of the at least one electrical energy storage element, and otherwise maintaining a previously computed resistance of the equivalent series resistor and the at least one model parameter of the at least one electrical energy storage element. By using an iterative optimization method minimizing a quality parameter, the computational method can be reduced in each round. In particular, the parameters of the ECM are only updated when an optimization step has actually resulted in a better approximation of the optimization parameters.
In at least one implementation, the optimization procedure is performed using a gradient optimizer, wherein the gradient optimizer performs up to kmax optimization steps, with kmax≥1; and the gradient optimizer minimizes the residual square sum (RSS). The use of a gradient optimizer is computationally efficient and approximates the modelling parameters particularly fast.
In at least one implementation, the step of estimating the health status of the pitch energy storage unit includes comparing the at least one updated parameter of the at least one electrical energy storage element with a nominal parameter value of the at least one electrical energy storage element, in particular, with a predetermined multiple of the nominal parameter value, and/or comparing the updated resistance Ri with a nominal resistance value of the equivalent series resistor, in particular, with a predetermined multiple of the nominal resistance. Note that the term “multiple” may also refer to a fractional multiple in this context, that is, a case where the respective threshold value is smaller than a given nominal value. By observing characteristic parameters, such as an updated capacity and/or series resistance of a pitch energy storage unit and comparing them with reference values, taken, for example from a data sheet, enables a good estimation of its state of health.
In at least one implementation, the pitch energy storage unit includes at least one capacitor, in particular an ultra-capacitor, for pitch energy storage, the at least one energy storage element of the ECM is an equivalent capacitor, and the updated parameter corresponds to an updated capacitance C of the equivalent capacitor. Capacitor-based pitch energy storage units have many advantages, including a higher number of charging cycles compared with rechargeable batteries, and can be modelled based on their capacity and series resistance using the disclosed method.
In at least one implementation, the method further includes, if the health status indicates that the pitch energy storage unit is at or near its begin-of-life, BoL, then charging the capacitor to a first charging voltage sufficient to provide a predetermined amount of electrical power and/or energy. If the health status indicates that the pitch energy storage unit is at or near its end-of-life, EoL, charging the capacitor to a second charging voltage sufficient to provide the predetermined amount of electrical power and/or energy, wherein the second charging voltage exceeds the first charging voltage. Such an adaptive charging model based on a current health status will extend the service life of capacitors.
In at least one implementation, the health status indicates an estimated capacitance C of the capacitor, and a charging voltage for the capacitor is increased from the first charging voltage to the second charging voltage based on the estimated capacitance C, in particular in an anti-proportional ratio. Such an approach ensures optimal charging voltages in all situations.
According to a second aspect of the disclosure, a pitch control system for a wind turbine is disclosed. The wind turbine includes an energy storage unit and a converter circuit, in particular a pitch energy storage unit and a pitch converter. The energy storage unit includes at least one energy storage device and has at least two power terminals for receiving or providing electrical energy. The converter circuit is connected via the at least two power terminals to the energy storage unit to selectively charge or discharge the at least one storage device of the energy storage unit. The pitch control system further includes at least one controller, in particular a controller of the converter circuit, the at least one controller being configured to implement a method according to the first aspect or any of its implementations.
The invention will now be described with reference to the drawings wherein:
During operation, the rotor 108 is set in rotation by an air flow, for example wind. This rotational movement is transmitted to the generator via the rotor shaft, with or without a gearbox. The generator converts the mechanical energy of the rotor 108 into electrical energy.
To control the rotational speed of the rotor, the rotor blades 110 can be adjusted by rotating them along the longitudinal axis. This rotation is performed by a pitch control system, including one or more pitch drives.
Each pitch drive 122 may include an electrical motor (not shown), such as a servo motor, and a transmission system (not shown) for coupling the electrical motor to at least one of the rotor blades 110. The pitch drives 122 are controlled by the pitch controller 124. The pitch controller 124 may be a separate component or may be an internal controller of another component. It usually includes or has access to the measurement circuitry for measuring voltage and current provided by the pitch energy storage unit 128. Moreover, it usually controls a charging or discharging current of the pitch energy storage unit 128. In the depicted embodiment, the pitch controller 124 forms part of the pitch converter 126. However, this is not limited by the present disclosure. Electrical energy required for the operation of the pitch drives 122 and the pitch controller 124 is provided by the pitch converter 126.
In normal operation, the electrical energy required for the pitch control system 120 may by derived from the generator coupled to the rotor 108. In particular, the electrical energy generated by the generator is supplied and stored in the grid, and the pitch control system 120 may draw power from the grid to operate. However, in case the rotor 108 is not moving or the generator does not provide sufficient electrical energy, all or parts of the electrical energy for the operation of the pitch control system 120 may be provided by other power sources, including the pitch energy storage unit 128. This is important, as the pitch control system is also important for braking the rotor 108, for example, in case of strong winds or emergency shut-down of the turbine. In this regard, the pitch converter 126 and pitch energy storage unit 128 effectively provide an uninterruptable power supply (UPS) functionality, to enable pitch control even in the absence of any wind, or during a power grid failure, or failure of other components of the wind turbine 100, or during circumstances like grid fault ride-through (FRT) characterized by low or high grid voltages.
The pitch converter 126 is also responsible for charging one or more energy storage devices 130, such as rechargeable batteries or capacitors, of the pitch energy storage unit 128. In an embodiment of the disclosure, the capacitor may be categorized as an electric double-layer capacitor, ultra-capacitor (UC) or a super-capacitor (SC) based on their storage capacities and other characteristics. For example, UCs or SCs are high-capacity capacitors with high energy density than the conventional electrolytic capacitors and having a tolerance for charge-discharge cycles much larger than rechargeable batteries. However, for the sake of simplicity, the term capacitor has been used throughout the description which may be interchangeably referred to as an ultra-capacitor or a super-capacitor.
Charging is typically performed in case sufficient electrical energy for charging is available, for example, from the generator or an electricity grid, to which the wind turbine 100 is connected during normal operating states. Preferably, the pitch energy storage unit 128 should always store sufficient electrical energy to enable the pitch control system 120 to at least brake or stop the wind turbine 100 to a standstill in the most adverse conditions. Preferably, it should also hold enough energy to restart the wind turbine when normal conditions are restored.
Due to the uncertainty of such emergency operations, the pitch energy storage unit 128 is used infrequently and in a non-predictable manner. However, if it is relied on, it may need to provide a relatively high amount of electrical power and corresponding high electrical current in a relatively short duration of time for pitching, particularly with a high torque. Thus, energy storage devices 130 with a relatively high storage and discharge capacity are used in pitch energy storage unit 128. Moreover, such energy storage devices 130 need to be replaced before their capacity becomes insufficient to provide the required amount of electrical energy, for example, due to thermal or electrical stresses leading to deterioration of health.
Determining the actual storage capacity and/or momentary electrical charge or other characteristics of the one or more energy storage devices 130 encapsulated in the pitch energy storage unit 128 is a difficult task. For example, electrolytes of energy storage devices 130 like batteries may be affected in different ways by environmental stress. Moreover, the information about the internal state of the one or more energy storage devices 130 constituting the pitch energy storage unit 128 available to the pitch controller 124 may be limited or not reliable enough. Thus, in practice, energy storage devices are often charged using a maximum nominal charging voltage, current or time at regular intervals or whenever they have been used, and are replaced at regular intervals, for example, before their shortest possible service life or when they approach end-of-life.
According to the present disclosure, an improved approach to monitoring the health and charging of a pitch energy storage unit is provided.
As can be seen in
To better assess an actual storage capacity, the disclosed system makes use of an equivalent circuit model (ECM) of the energy storage unit 128.
Nonetheless, the ECM 140 shown in
Note that the nodes 142 and 144 shown on the left of the ECM 140 correspond to the terminals 132 and 134 of the energy storage unit 128. The voltage u between them and the current im measured through, for example, the first terminal 132, can be observed from the outside. It is more difficult to obtain real-time values for the model parameters, in particular the equivalent capacitance C, the equivalent series resistance Ri and the equivalent parallel resistance Rp. Similarly, one cannot directly observe the modelled internal voltages and currents, like the voltage u_C across the capacitor 146 as well as the currents i_C and i_p through the capacitor 146 and the parallel resistor 149, respectively. While these parameters cannot be directly observed, they can be derived from the ECM 140 indirectly, as further detailed below.
In the disclosed embodiment, an estimator is employed to determine at least the capacitance C and series resistances Ri based on multiple samples of the voltage and the current measured at the terminal 132 and 134. Optionally, the estimator may also be used to determine the parallel resistance Rp as further detailed below. The estimator uses an optimizer to match the voltage u and current is of the ECM 140 to those measured at the terminals 132 and 134 of the energy storage unit 128 to determine real-time values of the parameters, like capacitance and internal resistance. The model parameters of the ECM 140 are tuned by the optimizer. The difference between the measured and calculated values provides a figure of merit, which serves a measure of the quality of the estimation.
Attention is drawn to the fact that in the targeted application, the current i_p, which represents the leakage current of the capacitor 146 and the current through the balancing network, may be insignificant compared to the total current i. Thus, the proposed method may be simplified and more stable when the value of the parallel resistance Rp is set to a constant, or is omitted completely, and only the model parameters Ri and C are subject to the optimization. However, as a more general case, in the following the optimization of all three model parameters of the ESM 140 of
It is also possible establish a similar ECM and formulate similar differential equation(s) for solving the voltage parameter (instead of the current parameter as indicated in the embodiment described herein above and below in detail). In that case, a measured voltage um and a calculated voltage us to be determined from the ECM are considered.
The current is through the measurement node 142 calculated from the ECM 140 is characterized by the differential equation:
i
s
:=C/(1+Ri/Rp)du/dt−CR;(1+Ri/Rp)dim/dt+1/(Rp+Ri)u (1)
Where C, Ri, Rp, and u have the regular meanings as described above, dx/dt represents rate of change or time derivative of parameter x, which in this case are applied to the measured voltage u and the measured current im. On the other hand, the current measured at the corresponding first terminal 132 measured is denoted as im herein above and below. The voltage u between the measurement node 142 and the reference node 144 corresponds to the voltage measured across the two terminals 132 and 134.
For each optimization interval, the estimator collects a measurement sample including four values, that is, a rate of a voltage change x1, a rate of a current change x2, a terminal voltage x3 and a current im measured across the terminals:
The time derivatives du/dt and dim/dt are calculated from the measured values for u and im, for example by comparing two subsequent measurements of a time-series of measurements performed by the pitch converter 130 or by any other means. The resulting samples, that is, the quadruples (x1, x2, x3, im) are then stored, for example, in a ring buffer of size N.
To simplify the equations, the following optimization parameters are defined:
Each optimization parameter a1, a2 and a3 may be a characteristic or a function of one or more of C, Ri and Rp.
Using relations (2) and (3), equation (1) can be simplified as:
i
s
=a1x1+a2x2+a3×3
Thus, the error of nth sample, defined as the difference between the calculated current is,n and the measured current im,n is
r
n
=i
s,n
−i
m,n=[(a1x1n+a2x2n+a3x3n)−im,n]
The residual square sum (RSS) for all samples is
Or, in an alternative notation, RSS=∥A a−Im∥2, where A is a matrix including a subset of the first three entries (x1, x2, x3) of each of the N samples, a is a vector of the optimization parameters (a1, a2, a3), and Im is a vector of N including the last entry of each of the N samples corresponding to the measured currents im.
The goal for the optimizer is to minimize RSS by tuning the optimization parameters a1, a2, and a3. This results in an approximation of the calculated current is to the measured current i_m. From the optimization parameters a1, a2, and a3, the model parameters C, Ri and Rp of the equivalent circuit 140 can be calculated based on the equations above.
In method step 152, a suitable ECM is provided. For example, the ECM 140 may be provided for estimating a current capacitance or health state of an energy storage unit 128 including a plurality of capacitors as energy storage devices 130. Alternatively, an ECM based on a different type of energy storage device may be used, such as an ECM including a voltage source and a shunt resistor, which represents an energy storage unit 128 including one or more rechargeable batteries.
In method step 154, initial values are provided for the estimator. For example, known reference values, for example, the values defined in the specification of the original equipment manufacturer (OEM), may be provided for a new energy storage unit 128. Alternatively, the initial values may be calculated based on a known physical configuration and the technical specification of used components taken from corresponding data sheets.
In method steps 156 and 158, a sampling component measures the voltage u and the current im across/at the terminals 132 and 134. Steps 156 and 158 may be performed at regular intervals, for example, once per second or minute, or may be triggered by specific events, that is, the activation of a service routine, a detection of a charge or discharge event, or a combination of both.
As detailed below, method steps 156 and 158 need to be performed repeatedly before the optimization is performed for the first time. In at least some embodiments, steps 156 and 158 are performed throughout the operation of the wind turbine 100.
In subsequent method steps 160 and 162, the sampling component calculates the time derivatives du/dt and dim/dt, respectively, with respect to the previously taken sample. Moreover, the sampling component may weigh and/or select samples as further detailed below. For example, only samples that fulfill certain criteria, such as a predefined quality of the measurement, proximity to a relevant operating point, and/or uniqueness, are considered and placed, for example, into a ring buffer.
When new samples are added in further iterations of the method steps 156 to 162, and in case of a limited data structure such as the ring buffer, the oldest sample may be previously removed or overwritten by the new sample.
Once a sufficient number of samples is available, for example when the ring buffer is full, the optimizer can start working as illustrated by method step 164. The optimizer provides as its output updated values of the model parameters, such as the equivalent capacitance C, the equivalent series resistance Ri and, optionally, the equivalent parallel resistance Rp of the ECM 140.
The optimizer may be implemented as a gradient optimizer as further detailed below. This is beneficial to minimize the computational effort. In each time step, the optimizer takes only a limited number of optimization steps. While it will take some time to find a suitable solution, this reduces the computational effort and provides a filtering effect against measurement and numerical noise.
Alternatively, for example, a more powerful controller or a general-purpose processor may be available for monitoring, and/or other optimizers or optimization methods known to the skilled person may be employed. Such optimization methods may include, among others, simulated annealing, particle swarm optimization, or the use of generic algorithms. While the computational effort of such optimization methods is much higher as in the case of a gradient optimizer, they may be beneficial, for example, to avoid local optima. Such optimization methods may also be useful for optimizing parameters of a more complex ECM, such as a potentially higher order model of a battery based pitch storage unit.
In method step 166, a health status for the pitch storage unit 128 is computed and, optionally, output. This is based on a comparison of the model parameters computed in method step 164 with known reference values at the end of the life of a given pitch storage unit 128. Such a health status value may represent a remaining lifetime as a percentage value and may be used to plan servicing or replacement of the pitch storage unit 128. It may also be used for operation of the pitch control system, such as implementing an adaptive charging of the pitch storage unit 128 as further detailed below.
Attention is drawn to the fact that the described method 150 may be performed in a continuous manner throughout the operation of the wind turbine 100. In that case, method steps 154 to 166 may be repeated in a loop-like structure. Note that in this case, the output model parameters computed in method step 164 may be used as initial ECM parameter values for method step 154 in the next iteration.
Attention is drawn to the fact that the physical parameters of the pitch storage unit 128 corresponding to the model parameters of the ECM typically change much slower than a timeframe, in which the optimizer converges. Thus, after the optimizer has reached a minimum in one cycle or iteration of the optimization procedure, it can follow the change of the physical parameters easily.
The estimation cycle starts at step 172. For example, an update of the health status can be triggered at a regular interval or on occurrence of a specific event. If the method is triggered for the first time, preconfigured values for C, Ri, Rp are loaded into the estimator. The preconfigured values may be taken from the component specification as an initial guess. Otherwise, the estimator loads or maintains the values for the model parameters C, Ri, Rp from a previous estimation cycle.
In a step 174, the voltage u and current im at the terminals 132 and 134 of the pitch energy storage unit 128 are measured. Using the ECM 140 definitions and equations from above, the measured voltage can be used directly as third sample value x3=u. Equally, the first and second sample values are determined by computing the voltage and current change x1=du/dt and x2=dim/dt, respectively.
In case the estimation cycle is performed at regular, relative short intervals, for example every second, the time derivatives du/dt and dim/dt can be approximated by simply computing the differences between the presently measured values for ut and im,t and their respective predecessors ut-1 and im,t-1 from the last estimation cycle, that is x1=du/dt=(ut−ut-1)/ΔT and x2=dim/dt=(im,t−im,t-1)/ΔT, where ΔT is the sampling interval between samples taken at time t−1 and t. In case the cycle is performed at irregular and/or longer intervals, multiple measurements should be taken during step 174 and used to determine the present rates of change for the voltage and current. In other embodiments, numerical schemes for equal and unequal time intervals (for example time steps) may be applied to determine rate of change of voltage and current corresponding to different time instants.
In step 176, the newly acquired samples are conditioned. This may include one or more of the following options.
As a first option, depending on the sampling frequency, filtering may be applied to smooth sampling errors.
As a second option, conditioning is implemented, at least in part, by weighting of the individual sample values, that is quadruples (x1, x2, x3, im). For example, as described above, the influence of the equivalent parallel resistance Rp on the optimization result may be typically small. To make it relevant for the optimization, the possible error is increased by a high weighting factor. Also, the influence of samples with a poor measurement quality can be decreased with a low weighting factor. In the described embodiment, samples with large du/dt or dim/dt get a lower weighting factor. This is partly to reduce the influence of noise, partly because the capacity of the capacitors is strongly frequency dependent.
As a third option, sample tuples (x1, x2, x3, im) that are already contained in an identical form or within a predefined tolerance interval of an existing sample in the ring buffer, may be rejected, that is, not entered into the ring buffer a second time.
In step 178, the method checks whether a list or set of samples contains sufficient sample tuples to perform the optimization procedure. In the described embodiment, a fixed number of samples is used by the optimization procedure, which are stored in a ring buffer of a corresponding size. In this case, in step 178 it is verified whether all storage locations of the ring buffer are filled with a valid sample.
If the list of samples is still incomplete, that is if there are still unused spaces in the ring buffer, in step 180, the newly acquired sample is added to the stored set of samples. Thereafter, the present estimation cycle ends immediately, as there is not enough data available for optimization yet.
Otherwise, that is, if a sufficient number of samples has been stored, the method proceeds in step 182 with removing the oldest previously stored sample from the list of samples. The present sample is then stored at the end of the list. In case of a ring-buffer, this may be achieved by simply overwriting a corresponding storage location based on a cyclic indexing scheme.
In step 184, a starting point is chosen for the optimizer. In the specific implementation, to combine computational efficiency as well as a stable output of the optimizer, this step may be performed in one of two possible ways. In particular, at regular intervals, for example every xth cycle starting from and including the first time the optimizer is started, an initial parameter set may be used as starting point for the optimization. As detailed above, the initial parameter set may be derived from knowledge of the actual structure of the pitch energy storage unit 128. In addition to characteristic values contained in or derived from the components data sheet, such values may also be updated in accordance with a predicted aging, for example, of the energy storage devices 130.
Alternatively, for example, at all other cycles, the parameter obtained from last cycle of the estimation method 170 are used as starting point for a run of the optimizer.
In step 186, the actual optimization takes place.
Note that a gradual optimizer is used, which performs optimization using a discrete number of optimization steps, up to a predefined maximal number of steps, that is, kmax.
In mathematical terms, the optimization process can be expressed as follows:
Then, the RSS, RSS(ak), is computer for step k, based on the optimizer “gradient”, dk, as follows:
The optimizer uses a defined step length ρ to perform the next optimizer step with ak+1=ak+ρ dk.
The step length ρ can either be fixed, or optimized by itself using a back tracking approach. In the described implementation, a fixed step length ρ is used.
Note that step 186 may be performed in a repetitive manner, for example forms an inner loop with the outer loop corresponding to repeated execution of the estimation method 170 as shown in
Attention is drawn to the fact, that only a few or even a single optimization step may be performed in each estimation cycle of the disclosed method 170. For example, in the described implementation kmax=1 is used, meaning that only single optimization step is performed during each execution of method step 186. When there is only one step, the optimizer effectively runs over several cycles of the estimation method 170.
In step 188, a quality parameter for the new parameter estimation is determined. In the described embodiment, the RSS of the last optimization step is used as quality parameter, for example a figure of merit, and stored in a separate variable, that is RSS_new=RSS(ak).
In step 190, the quality parameter RSS_new determined in step 188 of the present cycle is compared with a quality parameter RSS_last determined in step 188 of a previous estimation cycle of method 170.
If it is determined that the newly computed set of optimization parameters a1, a2 and a3 from the optimizer is better than the previously used set of optimization parameters, that is, if RSS_new<RSS_last, the model parameters C, Ri and Rp of the ECM 140 are newly computed based on the above equations.
In the described implementation, an additional plausibility or sanity check on the newly calculated values for C, Ri and Rp is performed. In particular, it can be verified if the newly calculated values for C, Ri and Rp fall into respective, for example preconfigured, ranges of reasonable parameter values.
If it is determined in step 190 that the new parameter values are better and plausible, in step 192, the present optimization parameters a, for example, the vector (a1, a2, a3), and model parameters, for example C, Ri and Rp, are overwritten with the last set of corresponding parameters obtained in steps 186 and 190, respectively.
Optionally, at this stage, an appropriate filter may be applied prior to the updating of the stored parameter values, for example to prevent sudden changes in the respective parameter values.
Otherwise, if it is determined that the last set of parameters are not better than the previously stored set of parameters, the newly computed parameters are rejected in step 194.
Either way, the cycle of the estimation method ends in step 196. If necessary, the present state of the estimator and/or optimizer is stored at this stage and used as initial values in the next estimation cycle.
At this stage, the current model parameters for C, Ri and Rp may be output or logged for use by other components, such as the adaptive charging method described later. Alternatively, or in addition, an aggregate health status may be computed based on one or more threshold values. Such an aggregated health status may be used to schedule and implement predictive maintenance. Note that in general, the capacitance of a capacitor will decrease over its lifetime, while its series resistance will increase. Only if the corresponding model parameters output by the estimator lie outside an allowed range for safe operation of the wind turbine 100, maintenance is required.
For example, beginning-of-life (BoL) and end-of-life (EoL) thresholds may be provided based on nominal values provided in the component's data sheet. In particular, a BoL capacitance CBoL may be provided by the manufacturer of a capacitor serving as the energy storage device 130. Then, a corresponding EoL capacitance CEoL may be computed, for example, as 80% of the CBoL value, that is, CEoL=0.8. CBoL. A current health status of the energy storage device 130 may be expressed relative to these values as a percentage value. Note that in practice, a capacitor may initially have a capacitance even higher than a nominal capacitance indicated in its datasheet. In this case, the health status of the capacitor would be more than 100%. Inversely, if it is not possible to replace the capacitor in good time, its actual capacitance may drop to below CEoL, which would result in a negative health status value. Similar definition and observation also apply with regard to one or more internal resistance values, which can be compared with the series resistance Ri or parallel resistance Rp. For example, a series resistance Ri of 200% of the initial series resistance RBoL of a capacitor may mark its end-of-life value REoL, that is, REoL=2·RBoL.
In the described embodiment, the lower of the two percentage values is taken as the total health status. As soon as one of the two values approaches, reaches or even falls below the respective EoL value, predicted maintenance is scheduled, at which the entire pitch energy storage unit 128 or its internal energy storage device 130 is replaced. Note that a sudden drop below or above the respective threshold may also allow to identify defective energy storage devices 130.
To make sure that the estimator is provided regularly with new, relevant samples, a controller executing the method 170, for example, the pitch controller 124, can request a small charge or discharge of the pitch energy storage unit 130. The frequency of these tests may depend on the quality of the last estimation, which is measured by the RSS as detailed above.
In particular, the upper part of
In the present example, the pitch energy storage unit 128 includes one or several capacitors, which are initially completely discharged, for example on their first connection to the pitch converter 126. Accordingly, the voltage initially increased from zero to a predetermined first target voltage. During this first phase 206, the charging current is more or less constant.
In a second phase 208, the charging current drops to zero, for example due to a deliberate interruption of the charging cycle or due to an accidental grid failure. Note that the pitch drive 122 is not activated in this phase. Accordingly, the voltage u may only drop slowly and insignificantly at this phase, for example, to a self-discharging of the capacitors or due to losses caused by a balancing network.
In a third phase 210, the charging process continues with the same, more or less constant charging current, until the pitch energy storage unit 128 has reached a second target voltage, at which the capacitors are essentially fully charged.
Accordingly, the charging current i drops back to and essentially remains at zero in a subsequent fourth phase 212.
Note that at this stage, the state of health of the energy storage unit 128 may not be known yet, as the estimator may not yet have obtained sufficient data for optimizing the ECM 140. Accordingly, the second target voltage may represent a default value for charging the energy storage unit 128.
In a fifth phase 214, the current i becomes negative, indicative of a discharge operation. This may be caused by the operation of the pitch drive 122, while no or only an insufficient amount of electrical energy is available from the generator or a supply grid. It may also be used to improve the results provided by the estimation method 170, for example by deliberately discharging the capacitor in a controlled manner. Accordingly, the voltage u drops during the fifth stage 214.
In a sixth phase 216, the pitch drive 122 is deactivated and/or the controlled discharging stops, and the pitch energy storage unit 128 is charged again. Accordingly, the charging current i becomes positive and the voltage u increases again, this time to a third target voltage.
Note that at this stage, the state of health of the energy storage unit 128 is already known, as the estimator has now obtained sufficient data for optimizing the ECM 140. Accordingly, the third target voltage may represent an optimized target voltage for charging the energy storage unit 128 based on its current state of health.
The middle part of
It can be seen that the estimator initially used a first capacitance estimate, which may be taken directly from a data sheet of the capacitors included in the pitch energy storage unit 128. However, during optimization, this value is increased to a second capacitance value during the second phase 208. In later phases, the capacitance is approximated further, and eventually sinks to a third capacitance value. Similar changes can be observed for the fourth graph 224 and fifth graph 226, which show the gradual approximation of the respective model parameters as the estimator processes further samples.
The lower part of
A manual comparison with respective parameters determined by direct measurements or retrieved over corresponding field busses of the internal components of the pitch energy storage unit 128 showed a good match of the estimated model parameter values, at least from the fourth phase 212 onwards, for example after one complete charging operation of the pitch energy storage unit 128.
As discussed above, knowing the SoH of the pitch energy storage unit 128 enables, among others, to perform system maintenance, including replacement of the pitch energy storage unit 128 or of the energy storage device(s) 130 contained therein, according to their specific requirements, for example, just-in-time. Thus, early replacement of energy storage device(s) that have been operated, for example, in less extreme conditions, have been used less than anticipated or simply have a better quality than indicated in their datasheet, can be retired later in their life. Conversely, storage device(s) that have been operated, for example, in very demanding conditions, have been used more than anticipated or simply have a lower quality than indicated in their datasheet, can be replaced earlier to avoid equipment failures.
Moreover, according to another aspect of the disclosure, a good knowledge of the actual SoH of the pitch energy storage unit 128 or its internal components may also be used to actually prolong their service life by implementing an adaptive charging scheme as further detailed below.
Depending on the type of energy storage device 130 used, adapting a charging procedure and/or parameters, may extend its lifetime. In the following, an adaptive charging method for a capacitor-based pitch energy storage unit 128 is described as an example. However, similar approaches may also be applicable to other types of energy storage device 130, such as rechargeable batteries.
In step 302, the current SoH of a pitch energy storage unit 128 or of the energy storage device(s) 130 contained therein is obtained. Note that the SoH may be obtained by the method described above, or any other suitable means of providing a good estimate of a SoH, including modelling physical properties of its internal components, based on other data, including monitored environment conditions and usage periods.
In the described embodiment, the state of health of a capacitor-based pitch energy storage unit 128 may be represented by the equivalent capacity C and the equivalent series resistance Ri of the equivalent circuit 140 shown in
Accordingly, in step 304, a determination is performed to establish whether the SoH of the pitch energy storage unit 128 meets at least one health criterion or not. Note that the health criterion may take the form of a single numerical threshold, such as an estimated SoC value indicated as a percentage. Alternatively, step 304 may include a number of individual determinations, such as validating a combination of individual parameters, obtained, for example, during estimation.
In the specific example, the equivalent capacity C and the equivalent series resistance Ri taken from the ECM 140 are compared with corresponding parameter values indicative for the equivalent capacity C and the equivalent series resistance Ri at the end of their service life.
In step 306, a first charging method based on a first charging parameter is used to charge the pitch energy storage unit 128 if the pitch energy storage unit 128 meets the at least one health criterion. The at least one health criterion may indicate that the pitch energy storage unit 128 is close to its beginning of life (BoL), or at least closer than to its end of life (EoL).
In the described implementation, a first charging voltage, for example 430V, is used to charge the pitch energy storage unit 128 in a service period towards the BoL.
Note that in case of a capacitor-based pitch energy storage unit 128, the pitch energy storage unit 128 is typically charged with a constant current until a desired target voltage is measured at the two terminals 132 and 134. Thereafter, the desired target voltage is maintained, but the charging voltage drops to essentially zero.
If pitch energy storage unit 128 does not meet at least one of the health criteria, in step 308, a second charging method based on a second charging parameter is used to charge the pitch energy storage unit 128. Failing the at least one health criterion may indicate that the pitch energy storage unit 128 is close to its end of life (EoL), or at least closer than to its beginning of life (BoL).
In the described implementation, when the equivalent capacity C decreases and/or the series resistance Ri rises over a respective threshold value throughout the lifetime of the pitch energy storage unit 128, the charging voltage is increased. For example, at or near the EoL, a second charging voltage, for example 460V, is used.
Note that in the disclosed implementation, the charging voltage is increased linearly from the BoL to the EoL value. Specifically, as the energy stored in a capacitor is proportional to both its voltage and capacity, the charging voltage is increased in an anti-proportional fashion with respect to a decrease in the equivalent capacity C, such that the amount of energy stored in the fully charged pitch energy storage unit 128 remains essentially constant over its lifetime.
Alternatively, only two distinct charging parameter values, for example charging voltages of either 430 or 460 V, may be used in two corresponding parts of the overall life cycle. In general, any number of charging parameter values may be used for a corresponding number of the overall life cycle, for example charging voltages of 430, 440, 450 or 460 V.
Attention is drawn to the fact that, due to the higher capacity near the BoL, the amount of energy stored in the pitch energy storage unit 128 in the first part of its service life is still sufficient for emergency operation at the lower charging voltage. At the same time, the use of the lower charging voltage reduces the stress on the capacitors contained therein, thus extending the service life of the pitch energy storage unit 128.
At least in case of capacitors being used as storage elements, the capacity C and resistance Ri normally depend on the temperature of the pitch energy storage unit 128. Usually, the capacity 312 and resistance 314 of a capacitor are slightly better at higher temperatures, as shown in
Note that the above methods 150 and 170 for estimating the health status of the pitch energy storage unit 128 disclosed above determines a more or less instantaneous value for the equivalent capacity C and the series resistance Ri. Thus, the adaptive charging method 300 will automatically select a lower charging voltage at higher temperatures (due to the higher capacity and lower resistance obtained from the estimator) and thereby at least partly compensate for the faster aging due the higher temperature. Note that at least in case the capacity C and resistance Ri are obtained from the estimator as detailed above, no knowledge of the actual age, charging and operating history, or temperature is needed to select an appropriate charging voltage.
In the case that the pitch energy storage unit 128 includes at least one rechargeable battery cell for pitch energy storage, charging the pitch energy storage unit 128 typically includes charging the rechargeable battery cell using a battery charging circuit or management system, which may be adapted to the type of the battery being used, for example a lithium battery. The battery charging circuit usually switches off a charging voltage, when a target voltage is reached, and may repeatedly perform short loading cycles to maintain a desired charging level as requested by the battery charging circuit or management system. In this case, a charging current, duration or frequency of the battery charging circuit or management system may be updated based on the determined state of health.
It is understood that the foregoing description is that of the preferred embodiments of the invention and that various changes and modifications may be made thereto without departing from the spirit and scope of the invention as defined in the appended claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 23180031.9 | Jun 2023 | EP | regional |
