Method of controlling a wind power plant

Abstract

A method of controlling a wind power plant (12) comprising energy storage means (44, 46), the wind power plant (12) being connected to an electrical grid (19) and comprising one or more wind turbine generators (1) producing electrical power for delivery to the electrical grid (19), the method comprising: processing input data relating to the electrical grid (19) to determine a probabilistic forecast of future states of the electrical grid (19); and controlling charging and discharging of the energy storage means (44, 46) in dependence on said probabilistic forecast.

Description

Technical Field

[0001] The present invention relates to methods and control systems for controlling wind power plants including energy storage devices, and particularly to the control of the charging state of the energy storage devices. Background Technology

[0002] A wind farm comprises a group of wind turbine generators that convert the energy contained in the wind into electricity, which is typically fed into the power grid. Variations in wind conditions at a wind farm site can result in different power outputs for each generator. Furthermore, the time-varying nature of wind causes corresponding random power outputs for each of the generators.

[0003] Meanwhile, a series of events may occur on the power grid that alter the demand for electricity from wind farms. These events include: changes in the power consumption of loads connected to the grid; planned changes to the grid (such as the addition of new loads); real-time electricity price fluctuations; and critical grid events (including failures such as sudden changes in voltage or frequency on the grid).

[0004] The characteristics of the signals transmitted from wind farms to the grid (particularly the frequency and phase angle of these signals) must be consistent with instantaneous grid requirements. Wind farms must also deliver active and reactive power at grid demand levels. During grid events, these requirements may change abruptly, in which case wind farms may have to respond by modifying their output.

[0005] Therefore, the way the output of each wind turbine generator in a wind farm is processed and fed into the grid must be carefully managed. Consequently, wind farms must incorporate means to dynamically control their output in response to changing grid demands, while also compensating for varying wind conditions.

[0006] One such approach is to include one or more energy storage devices within the wind farm. In fact, if the grid needs to control the wind farm as a virtual synchronous machine (VSM) to provide the desired output unaffected by fluctuations or “virtual inertia” in generated power, then including energy storage in some form may be essential. These energy storage devices are typically chargeable and dischargeable on command and are capable of storing large amounts of charge, allowing them to amplify the power supplied by the wind turbine generators over short but sustained periods. Therefore, these energy storage devices differ from the relatively small smoothing capacitors and inductors typically included in the power converters and filters within the wind farm.

[0007] In practice, the corresponding energy storage device can be integrated into each wind turbine generator to provide inertia for the generator independently. For example, the energy storage device can be coupled to the DC link of the wind turbine generator's power converter. In this case, the energy storage device can be charged or discharged to alleviate the demand on the wind turbine generator. For example, in high wind conditions, the energy storage device can be discharged to allow the generator to reduce its power output to implement damping, thereby reducing structural load. However, the ability of the energy storage device to operate in this way is limited by the need to retain the charge that may be needed during critical grid events.

[0008] Alternatively or additionally, energy storage devices can be attached to a point of common coupling (PCC) to provide virtual inertia for the entire power plant, in which case similar operational constraints apply.

[0009] In either case, controlling energy storage units to optimize operation is challenging given the conflicting demands arising in the service. For example, ensuring sufficient capacity to absorb peak demand during power production must be balanced with constraints on reserving enough charge to compensate for transient drops in power production or to handle fault conditions such as low-voltage ride-throughs. Another factor to consider is that when electricity prices are low, it may be desirable to increase the state of charge of energy storage units in order to generate additional revenue through electricity sales when prices rise. Conversely, when electricity prices are high, a lower state of charge may be expected.

[0010] It was in this context that the present invention was designed. Summary of the Invention

[0011] One aspect of the present invention provides a method for controlling a wind power plant. The wind power plant is connected to a power grid and includes an energy storage device and one or more wind turbine generators that produce electricity for transmission to the power grid. The method includes: processing grid data associated with the grid to determine a probabilistic prediction of the future state of the grid; and controlling the charging and discharging of the energy storage device based on the probabilistic prediction.

[0012] Compared to known methods that seek to predict future inputs to wind farms in a discrete or deterministic manner, the method of this invention bases the control of the energy storage device of the wind farm on a probabilistic prediction of the plant's inputs—specifically, the grid state—representing potential future grid events in a probabilistic manner. In other words, the probabilistic prediction of the grid state provides an indication of the relative probabilities of a series of events or state changes occurring within a specific time window. This approach is less susceptible to inaccuracies than discrete predictions, thus enabling enhanced optimization of the operation of the energy storage device (which typically requires finding the device's optimal charging state).

[0013] This method may include identifying a set of chance-constraints related to the operation of a wind power plant, determining corresponding limits for each chance-constraint in the set, and determining a set of device constraints related to the charging and / or discharging of an energy storage device based on each chance-constraint limit and the probability predictions. The charging and discharging of the energy storage device is then controlled to avoid violations of the device constraints, which should be noted that the device constraints are determined based on the probability predictions. Therefore, such a method enables optimized use of the energy storage device while controlling the risk of violations of device and / or wind power plant operational constraints, thereby enhancing control flexibility compared to conventional methods.

[0014] Such a method may also include solving an optimization problem for controlling the charging and discharging of an energy storage device, which includes device constraints, in which case the charging and discharging of the energy storage device is controlled based on the control output of the optimization problem.

[0015] Predictive algorithms can be used to solve optimization problems. For example, the method may include simulating the operation of a wind power plant using a predictive algorithm. In such an embodiment, the predictive algorithm may be based on a chance-constrained model predictive control strategy. In other words, in such an embodiment, chance-constrained model predictive control can be used to optimize the state of charge of an energy storage device.

[0016] Each constraint can limit the proportion of wind farm operating time that can violate the corresponding opportunity constraint.

[0017] Each of the set of opportunity constraints may optionally include requirements to provide any of the following: virtual inertia; grid frequency control; finite rate of rise and fall of power delivered to the grid; power output profile during a defined voltage ride-through event; reduction of power line flicker; reactive power injection; grid oscillation damping; lateral torque damping; drivetrain torque damping; up yaw control; and noise below a threshold level.

[0018] In some embodiments, the method includes determining a charging state setpoint for an energy storage device, and controlling the charging and discharging of the energy storage device based on the charging state setpoint.

[0019] The probability prediction may optionally include any of the following: a cumulative distribution function; and a probability density function.

[0020] This method may include processing data indicating wind conditions to determine probabilistic predictions of those conditions, and controlling the charging and discharging of the energy storage device based on these probabilistic predictions. Therefore, embodiments of the invention allow control of the energy storage device based on the probability of changes in multiple inputs to the wind power plant, i.e., grid conditions and wind conditions. Probabilistic predictions can also be derived for other inputs, including the status of auxiliary systems at the wind power plant (such as antifreeze systems), which can also be taken into account in the control of the energy storage device.

[0021] The charging and discharging of energy storage devices can be controlled based on the probability of violating one or more power grid requirements.

[0022] Grid data may include any of the following: data indicating grid status; rate limits on power delivery to the grid; requests received from the grid; grid design data; historical grid data; data indicating a weak grid obtained from plant-level and / or turbine-level analysis; electricity price data; and user-input forecasts indicating planned changes to the grid.

[0023] The method may include altering the operation of one or more wind turbine generators in a wind power plant based on the probability prediction or each probability prediction, and controlling the charging and discharging of an energy storage device based on the altered operation of the one or each wind turbine generator and the probability prediction or each probability prediction.

[0024] Energy storage devices can be electrically coupled to a common coupling point where the wind farm connects to the power grid. Alternatively, energy storage devices can be integrated into the wind turbine generators of the wind farm. The wind farm may also include multiple energy storage devices, such as those integrated into the wind turbine generator or each wind turbine generator, as well as other energy storage devices coupled to a common coupling point.

[0025] Another aspect of the present invention provides a control system for a wind power plant. The wind power plant is connected to a power grid and includes an energy storage device and one or more wind turbine generators that produce electricity for transmission to the grid. The control system includes: an input configured to receive grid data associated with the grid; a processing module configured to process the grid data to determine a probabilistic prediction of the future state of the grid and generate a control signal based on the probabilistic prediction, the control signal being arranged to control the charging and discharging of the energy storage device; and an output configured to output the control signal.

[0026] The control system can be configured to perform the methods described above.

[0027] The present invention is also extended to wind power plants that include control systems incorporating the above aspects.

[0028] It should be understood that preferred and / or optional features of each aspect of the invention may also be incorporated, individually or in appropriate combinations, into other aspects of the invention. Attached Figure Description

[0029] To enable a more complete understanding, the invention will now be described by way of example only with reference to the following figures, in which:

[0030] Figure 1 This is a schematic diagram of a wind turbine generator applicable to an embodiment of the present invention;

[0031] Figure 2 It includes multiple such as Figure 1 A schematic diagram of a wind power plant showing a wind turbine generator;

[0032] Figure 3 This is a schematic diagram of the architecture of a wind power plant based on a full-size converter, applicable to embodiments of the present invention;

[0033] Figure 4 It is a graph showing the time curves of control inputs and power plant outputs; and

[0034] Figure 5 It shows the control Figure 3 A flowchart of the process of building a wind power plant. Detailed Implementation

[0035] Generally speaking, embodiments of the present invention provide methods and control systems for optimizing the operation of one or more energy storage devices incorporated in a wind power plant by probabilistically modeling the inputs of the wind power plant and determining the optimal strategy for charging and discharging the device or each device based on the corresponding probabilities of various possible variations in the inputs.

[0036] For example, a predictive algorithm can optionally be used to derive statistical characteristics or probabilistic predictions (such as probability density functions) for future wind conditions and / or grid events. The state of charge setpoint of the energy storage device can then be determined based on these probabilistic variables. This makes the setpoint less susceptible to prediction inaccuracies, thus taking into account the stochastic nature of these inputs to the wind farm.

[0037] Specifically, in the embodiments described below, the prediction algorithm forms part of a chance-constrained model predictive control method, in which constraints are set for chance constraints related to the operation of a wind power plant. These constraints are then translated into hard constraints for the operation of one or more energy storage devices of the wind power plant using a probability density function derived for each input. These hard constraints are then fed into the model predictive control algorithm, which solves an optimization problem to find the optimal state of charge and / or charge / discharge strategy for the energy storage device or each energy storage device.

[0038] In this context, the relevant opportunity constraints are related to changes in the power plant's inputs (particularly changes in wind conditions and grid events), which directly or indirectly reflect situations that may require charging or discharging of the energy storage device, and this particular form of model predictive control considers the chances of each of those situations occurring when determining the strategy for charging and discharging the energy storage device.

[0039] The conflicting requirements associated with each opportunity constraint can be weighed against each other in the context of the corresponding probability of each of those opportunity constraints being violated or must be satisfied in operation, thereby enabling the finding of the optimal state of charge for the energy storage device.

[0040] In a simple example, a pair of conflicting constraints could be the requirement to provide virtual inertia to the energy storage device to absorb excess power when wind energy instantaneously peaks, and the separate requirement to release energy from the energy storage device during a low-voltage ride-through event. The probability of each of these scenarios can be determined by probabilistic modeling of future wind conditions and grid events, and the optimized state of charge can be weighted toward the more likely scenarios, thereby minimizing the overall probability of violating either constraint in the long run and ensuring that violations of the chance constraint are kept within specified limits. In practice, many more variables and possible outcomes will need to be considered, leading to the need to solve a complex optimization problem.

[0041] Because electricity prices can fluctuate significantly in a short period of time, there is potential to increase revenue by optimizing the cost of energy storage devices by modifying their charging / discharging strategies in response to anticipated price changes, while simultaneously meeting minimum grid requirements during this period.

[0042] Other examples of specific constraints that can be considered are described in more detail below, but it should be noted at this stage that probabilistic modeling of various possible scenarios can optimize setpoints or other charging / discharging strategies to a much greater extent than could be achieved in prior art methods that only deal with transient conditions. Even methods that attempt to predict future inputs deterministically are susceptible to their predictive inaccuracies, especially for volatile inputs such as wind conditions or grid events. In contrast to these methods, embodiments of the present invention do not base control on predictive outcomes, but rather on the corresponding chances of a series of outcomes.

[0043] Since any form of forecasting involves uncertainty, previous methods have imposed hard limits or margins on the state of charge of energy storage devices, providing a buffer for suboptimal operations. However, imposing such margins can hinder full utilization of the device's capacity. In this context, embodiments of the present invention offer the advantage that the optimized state of charge setpoint of the energy storage device takes into account the uncertainties of the wind power plant's inputs, thus avoiding the need to apply margins to the device's state of charge and thereby improving the utilization of the device's capacity.

[0044] The following is for reference. Figures 1 to 3 This paper outlines specific implementations of this method, demonstrating how the concepts of the invention can be incorporated into existing wind power plant control architectures. It should be understood that this implementation is described as an example only, and embodiments of the invention will find application in all wind power plant architectures.

[0045] Therefore, in order to provide context for the present invention, Figure 1 A single wind turbine generator 1 of various types, which can be controlled according to embodiments of the present invention, is shown. It should be understood that... Figure 1 The wind turbine generator 1 is mentioned here only as an example, and embodiments of the invention can be implemented in many different types of wind turbine systems.

[0046] The wind turbine generator 1 shown is a three-bladed upwind horizontal axis wind turbine (HAWT), which is the most common type of turbine in use.

[0047] The wind turbine generator 1 includes a turbine rotor 2 with three blades 3, which is typically supported at the front of the nacelle 4. It should be noted that while three blades are common, a different number of blades may be used in alternative embodiments. The nacelle 4 is mounted on top of a support tower 5, which is fixed to a base (not shown) embedded in the ground.

[0048] Cabin 4 contains a generator driven by rotor 2 to produce electrical energy. Figure 1(Not shown in the image). Therefore, the wind turbine generator 1 is able to generate electricity from the wind flow that rotates the blades 3 through the swept area of ​​the rotor 2.

[0049] Figure 2 The image shows a wind turbine generator 1 in the context of a wind power plant 6 having multiple individual wind turbine generators 1; specifically, in Figure 2 The wind power plant 6 shown contains three wind turbine generators 1. Each wind turbine generator 1 has an output line 7a connected to a transmission line 7b, which transmits the electricity generated within the wind power plant 6 to the power grid.

[0050] Each wind turbine generator 1 in the wind power plant 6 is connected to a power plant controller (PPC) 8 that controls the operation of the wind power plant 6. In this embodiment, the PPC 8 is responsible for monitoring operating conditions and for providing a reactive power reference to each wind turbine generator 1 based on active power demand. Therefore, the PPC 8 represents part of a control system for controlling the operation of the wind power plant 6.

[0051] To this end, PPC 8 includes an input terminal 9 at which operational data from each wind turbine generator 1 is received. Input terminal 9 also receives data indicating grid status and data indicating wind conditions. PPC 8 also includes a processor 10, which, among other things, uses the data received at input terminal 9 to determine active and reactive power references for the wind turbine generator 1.

[0052] Now for reference Figure 3 An example of a wind power plant 12 to which the method according to an embodiment of the present invention can be applied is shown. Figure 3 The example shown is based on a full-size converter architecture, although as mentioned above, embodiments of the present invention can be used with other types of converters and in general the present invention is applicable to all topologies. Figure 3 The wind power plant 12 shown can be connected with Figure 2 The wind power plant 6 shown is configured in the same way.

[0053] Figure 3 The components of the wind power plant 12 are conventional and familiar to readers in the art, and will therefore be described only in a general manner.

[0054] Figure 3 The wind power plant 12 shown includes a single wind turbine generator 1 (such as...) Figure 1 The wind turbine shown may actually include other wind turbine generators (such as...). Figure 2 (As shown).

[0055] As already noted, the wind turbine generator 1 includes a generator 20 driven by a rotor 2 to produce electricity. The generator 20 includes a central armature 21 driven by the rotor 2 to rotate within a stator 23. The stator 23 includes one or more sets of three-phase windings (not shown), wherein current is induced in response to changes in magnetic flux generated by the rotation of the armature 21 under the control of the turbine controller 27.

[0056] Wind power plant 12 also includes a low-voltage link 14 defined by a bundle of low-voltage lines 16 terminating at a coupling transformer 18, which acts as a terminal for connecting the wind turbine generator 1 to a grid transmission line, which in turn connects to a grid 19. Electricity generated by the wind turbine generator 1 is transmitted to the grid 19 via the coupling transformer 18, and the electricity transmitted to the grid 19... Figure 3 The middle is represented as P PPC .

[0057] The full-scale generator 20 typically produces multiphase power. In this embodiment, the power produced in generator 20 is three-phase AC, but not in a form suitable for transmission to the grid 19, particularly because it is typically not at the correct frequency or phase angle. Therefore, the wind turbine generator 1 includes a power converter 22 and a filter 24 arranged between generator 20 and coupling transformer 18 to process the generator output into a suitable waveform with the same frequency and appropriate phase angle as the grid 19.

[0058] Power converter 22 provides AC-to-AC conversion by feeding current through an AC-DC converter or "machine-side converter" 26, followed by a series DC-AC converter or "line-side converter" 28. Machine-side converter 26 is connected to line-side converter 28 via a conventional DC link 30, which includes: a switched resistor 32 that acts as a load dumper to release excess energy; and a smoothing capacitor 34 that provides smoothing to the DC output.

[0059] Any suitable power converter 22 can be used. In this embodiment, the AC-DC and DC-AC sections of the power converter 22 are defined by corresponding bridges of switching devices (not shown) (e.g., in a conventional two-stage back-to-back converter configuration). Suitable switching devices for this purpose include integrated gate bipolar transistors (IGBTs) or metal-oxide-semiconductor field-effect transistors (MOSFETs). The switching devices are typically operated using pulse-width modulated drive signals.

[0060] The smoothed DC output of the machine-side converter 26 is received as a DC input by the line-side converter 28, which generates a three-phase AC output to be supplied to the coupling transformer 18.

[0061] The AC output of the power converter 22 is carried by three power lines 16 that collectively define a low-voltage link 14, each line 16 carrying a corresponding phase. The low-voltage link 14 includes a filter 24, which in this embodiment includes a corresponding inductor 38 with a corresponding shunt filter capacitor 40 for each of the three power lines 16 to provide low-pass filtering for removing switching harmonics from the AC waveform.

[0062] As described above, the low-voltage link 14 terminates at the coupling transformer 18, which provides the necessary voltage ramp-up. The high-voltage output from the coupling transformer 18 defines the wind turbine generator terminal 42, which serves as the common coupling point of the wind power plant 12.

[0063] As described above, in the full-scale architecture, the line-side converter 28 is configured to provide a certain level of control over the characteristics of the produced AC power, such as increasing relative reactive power according to grid demand. Note that the output amplitude, angle, and frequency are determined by grid requirements, and the voltage is set at a constant level according to the specifications of the low-voltage link 14; in practice, only the AC output current is controlled, and a converter controller 36 is provided for this purpose. The converter controller 36 and the turbine controller 27, in turn, act upon commands received from the PPC 8. In this respect, Figure 3 The dashed lines in the diagram represent the communication lines between PPC 8, turbine controller 27, and converter controller 36.

[0064] The converter controller 36, turbine controller 27, and PPC 8 together form part of an overall control system for controlling the operation of the wind power plant 12.

[0065] As described above, embodiments of the present invention relate to the control of energy storage devices integrated into wind power plants. In this respect, Figure 3 Two such devices are shown: a first energy storage device 44, which is electrically coupled to the DC link 30 of the power converter 22 and operated by the converter controller 36; and a second energy storage device 46, which is electrically coupled to the wind turbine generator terminal 42 on the grid side of the power converter 22 and controlled by the PPC 8.

[0066] As mentioned above, although Figure 3 Only one wind turbine generator 1 is shown in the figure, but in practice, wind power plants typically include a group of such wind turbine generators, and in such an arrangement, each wind turbine generator may include a corresponding energy storage device.

[0067] Having both a first energy storage device 44 and a second energy storage device 46 maximizes the flexibility of the wind power plant 12 in responding to different operating scenarios. However, in practice, it may be sufficient to provide an energy storage device in only one of these locations, either integrated into the wind turbine generator 1 or each wind turbine generator 1, or coupled to a common coupling point (i.e., the wind turbine generator terminal 42 in this example).

[0068] Various other energy storage topologies are also possible. For example, wind turbine generators in a wind farm can be distributed into two or more subgroups, each with a corresponding energy storage device.

[0069] In principle, various energy storage technologies can be used in either the first energy storage device 44 or the second energy storage device 46. In practice, batteries and large capacitors are possible options.

[0070] The first energy storage device 44 has a state of charge (SoC) 44a, which is as follows: Figure 3 The dashed line in the diagram indicates this. Above this, another dashed line represents the upper margin 44b of SoC 44a, which is implemented to ensure that device 44 always has a reserve capacity to absorb power if needed. Accordingly, the first energy storage device 44 also has a lower margin 44c of SoC 44a, which ensures that device 44 always retains some charge (e.g., for grid events). For example, lower margin 44c could represent 5% charge, and upper margin 44b could correspond to 95% charge.

[0071] Similarly, the second energy storage device 46 is in Figure 3 The SoC is shown as a dashed line with an upper limit 46b of the SoC 46a and a lower limit 46c of the SoC 46a.

[0072] As described above, by probabilistically modeling the input uncertainty, embodiments of the present invention advantageously allow the upper and lower limits 44, 46b, 44c, 46c to be set to 100% and 0% (if desired).

[0073] The first energy storage device 44 and the second energy storage device 46 are each operable to selectively charge and discharge under the control of the converter controller 36 and the PPC 8, respectively. In this embodiment, the converter controller 36 controls the first energy storage device 44 based on commands received from the PPC 8. The charging / discharging strategy adopted for each energy storage device 44, 46 is determined by the PPC 8 using chance-constrained model predictive control to optimize the state of charge of each device 44, 46 at all times, which will be explained in more detail below.

[0074] In this embodiment, each of the first and second energy storage devices 44, 46 is a DC device. The first energy storage device 44 is connected to the DC link of the power converter 22, and therefore can be directly connected. However, the second energy storage device 46 must exchange power with the low-voltage link 14 carrying three-phase AC power. Therefore, an AC / DC converter 48 is provided to act as an interface between the low-voltage link 14 and the second energy storage device 46.

[0075] The first energy storage device 44 and the second energy storage device 46 each have a significantly increased storage capacity relative to the smoothing capacitor 34 or the filter capacitor 40 of the power converter 22. Combined with the ability to selectively charge and discharge, this significantly extends and improves the wind power plant 12's ability to match its output to grid requirements. In other words, the first and second energy storage devices 44, 46 enhance the virtual inertia of the wind power plant 12.

[0076] The location of the first energy storage device 44 within the power converter 22 is also ideal for enhancing the operation of the wind turbine generator 1 (e.g., allowing power to be consumed within the wind turbine generator 1 during high-load conditions to reduce structural load). If the first energy storage device 44 maintains sufficient charge, it provides a power source that can be easily accessed to supplement the output of the generator 20 when necessary, particularly where the generator output is intentionally reduced to implement lateral torque damping and / or transmission torque damping under high loads.

[0077] Accordingly, the second energy storage device 46 is exposed to the combined output of all the wind turbine generators 1 of the wind power plant 12, and is therefore ideally positioned to smooth and adjust the output to match grid requirements.

[0078] Therefore, in this embodiment, the first energy storage device 44 is mainly used to assist the operation of the wind turbine generator 1 in which it is located, especially to reduce the structural load generated in the wind turbine generator 1, while the second energy storage device 46 mainly focuses on making the output of the wind power plant 12 consistent with the grid requirements. However, each energy storage device 44, 46 can fulfill either role.

[0079] As can be seen from the above, the first energy storage device 44 draws power from the DC link 30 when in charging mode, and outputs power to the DC link 30 when in discharging mode. Correspondingly, the second energy storage device 46 draws power from the low-voltage link 14 when in charging mode, and distributes power to the low-voltage link 14 when in discharging mode.

[0080] Figure 3 The power input from generator 20 to machine-side converter 26 is denoted as P. refThe power output from the line-side converter 28 is expressed as P. gref A first energy storage device 44 generates P between the machine-side converter 26 and the line-side converter 28. ref and P gref Different potentials. Therefore, the charging / discharging strategy of the first energy storage device 44 can be used with P ref and P gref The target value is expressed as [value].

[0081] Accordingly, charging or discharging the second energy storage device 46 affects P. gref and P PPC The relationship between these references can be used to define the charging / discharging strategy of the second energy storage device 46.

[0082] Operating either of the first and second energy storage devices 44, 46 in their respective charging modes typically requires utilizing the electricity produced by generator 20, thus reducing the power reaching the wind turbine generator terminal 42. In this scenario, P ref More than P gref , and / or P gref More than P PPC .

[0083] However, each of the first and second energy storage devices 44, 46 can also draw power from the grid 19, thereby advantageously enabling the wind farm 12 to act as a load. This can be useful in various situations, including: restoring power when negative electricity prices occur due to overproduction; charging the device to ensure the ability to perform auxiliary functions or compensate for charge depletion due to wind events; and allowing frequency downsampling. In this scenario, P gref It may exceed P ref And P PPC It may exceed P gref .

[0084] Using the physical hardware described above, the control strategies for the energy storage devices 44 and 46 will now be considered in more detail.

[0085] As already noted, embodiments of the invention attempt to optimize the SoC 44a, 46a of each of the energy storage devices 44, 46 by taking into account uncertainties in the inputs of the wind power plant 12. These uncertainties are combined by probabilistically modeling them to, for example, produce probability density functions representing probabilistic predictions of the inputs. Simultaneously, constraints are applied to chance constraints related to the operation of the wind power plant 12. These constraints are used in conjunction with the probability density function of each input to generate hard constraints, or “device constraints,” for the operation of the first and second energy storage devices 44, 46, which are fed into a model of the wind power plant 12 configured according to the principles of model predictive control. In this respect, while model predictive control is known to be used in the context of operating a wind power plant for other purposes, the introduction of chance constraints allows for further optimization, which is particularly useful for the control of energy storage devices.

[0086] The most relevant inputs with significant uncertainty are wind conditions—or power production dependent on wind conditions—and the state of the power grid 19. In this embodiment, the control of the first energy storage device 44 prioritizes the uncertainties associated with wind conditions because the load effects of wind variations are local to the first energy storage device 44. Accordingly, the control of the second energy storage device 46 prioritizes the uncertainties associated with the state of the power grid 19. In other embodiments, the opposite may be true, and typically each device 44, 46 is controlled to at least partially take into account all uncertainties of the available data.

[0087] The main objective of the optimization is to limit the P required by the grid so that wind power plant 12 can meet the grid requirements. ref and P gref Value, maximizing the production of electricity P PPC And to minimize the structural load on the wind turbine generator 1 or each wind turbine generator 1. Other objectives may include reducing the frequency of charge / discharge cycles of each energy storage device 44, 46 to minimize wear and tear, and avoiding penalties for failing to meet grid requirements.

[0088] Meeting these objectives is equivalent to defining a suitable charging / discharging strategy for each of the first and second energy storage devices 44, 46, thus requiring, at the most basic level, the optimal SoC 44a, 46a to be found for each device 44, 46 at any given time.

[0089] In practice, the output of the optimization process can be a control signal used to induce charging or discharging of the first and second energy storage devices 44, 46. This control signal can be expressed as a charging / discharging command and / or P. ref P gref and P PPCThe target value can also be generated. Control signals can also be generated to adjust operational aspects of the wind turbine generator 1 that affect the operation of energy storage devices 44, 46. For example, PPC 8 can generate setpoint commands to de- or over-rate the wind turbine generator 1, thereby influencing power production in a manner complementary to the charging / discharging strategies of the energy storage device or each of the energy storage devices 44, 46.

[0090] As a skilled reader will understand, in a general sense, the standard optimization problem can be expressed as:

[0091] Minimum ⨍(𝑥, 𝜉)

[0092] Limited by g (𝑥, 𝜉) = 0

[0093] h (𝑥, 𝜉) ≥ 0

[0094] Where 𝜉 is the uncertainty vector. Under the chance-constrained method, the inequality constraint is expressed as:

[0095] P(h (𝑥, 𝜉)) ≥ 0) ≥ p

[0096] in p ϵ[0, 1] is a condition that satisfies the constraint h The probability density function is (☥, ☐) ≥ 0. Therefore, in order to minimize ⨍(☥, ☐), the method sets a limit on the extent to which the chance constraint can be violated within a predetermined time period or "window", which in this case requires finding the optimal SoC 44a, 46a for each of the first and second energy storage devices 44, 46.

[0097] The relevant window for each constraint can be different and can be fixed in advance, for example, if the power grid 19 explicitly or implicitly indicates that the window is part of the constraint. Alternatively, the window can be adjusted as part of the control method. For constraints with a finite tail (e.g., constraints with a severity cap), operation can be performed with zero chance of constraint violation, thus the window is effectively infinite. An example of a constraint with a finite tail is power production, which is limited by the electrical characteristics of each wind turbine generator 1. While the power production of wind turbine generator 1 depends on wind conditions and is therefore indeterminate, it cannot exceed the physical limitations of wind turbine generator 1.

[0098] While the following description reflects how chance-constrained model predictive control methods can be used to solve the optimization problem in the context of the aforementioned wind power plant 12, it should be noted that the optimization problem is solved for each energy storage device 44, 46, and can also be solved if only one such device exists in the wind power plant 12. However, in the presence of multiple energy storage devices, they can be considered collectively in the analysis. For example, the problem can be formulated as finding the optimal collective state of charge among all energy storage devices, thus treating the distributed energy storage devices as a unified energy source. However, this will depend to some extent on the location of the energy storage devices 44, 46 within the wind power plant 12. For example, while treating the line-side devices as a unified device may be relatively straightforward, the storage devices integrated into the power converter may need to respond to the individual demands of their respective wind turbine generators.

[0099] In this embodiment, PPC 8 solves an optimization problem for each energy storage device 44, 46 to determine a charging / discharging strategy for each device 44, 46. The charging / discharging strategy generated for the first energy storage device 44 is then transmitted to the converter controller 36 for implementation. The charging / discharging strategy for the second energy storage device 46 is implemented directly by PPC 8. In other embodiments, the converter controller 36 may perform its own optimization to independently determine the charging / discharging strategy for the first energy storage device 44.

[0100] As previously mentioned, PPC 8 can access data indicating wind conditions and data indicating the status of the power grid 19.

[0101] Data indicating wind conditions may include measurements of instantaneous parameters such as wind speed and direction, turbulence intensity, and shear. PPC 8 can also access data that enhances its ability to assess how wind conditions may change within a predefined window. For example, PPC 8 can receive or store site map data, meteorological data, and data from lidar sensors. It should also be noted that current conditions at one location within wind farm 12 may indicate future conditions at another point within wind farm 12, and therefore these should also be taken into account.

[0102] By analyzing such data, PPC 8 can derive probabilistic predictions, such as probability density functions that indicate the relative probability of a series of variations in wind conditions occurring within a defined window. Variations that can be probabilistically characterized can include changes in turbulence intensity and low-frequency variations in mean wind speed (as well as the occurrence of more extreme wind conditions such as gusts), changes in wind direction, and the occurrence of extreme positive or negative shear. For example, the probability density function calculated by PPC 8 can specify the probability of gusts of different speeds, directions, and / or durations within a two-minute window.

[0103] For grid 19, a key uncertainty that needs to be predicted to optimize the operation of the first and second energy storage devices 44, 46 is the occurrence of critical events. Such events can include various faults, such as the occurrence of weak grid scenarios, or sudden changes in grid voltage in terms of its frequency, phase, or amplitude (including low and undervoltage conditions). Other grid events can include: changes in power consumption of loads connected to the grid; planned changes to the grid (such as the addition of new loads); real-time electricity price fluctuations; and ancillary service demands (such as inertial simulation, primary / secondary frequency control, and rapid rise and fall).

[0104] Another type of uncertainty arises in the form of inter-regional / intra-regional oscillations, which are related to the occurrence of resonance—and consequently to the oscillation power and voltage—as a result of the excitation of intrinsic frequencies or eigenmodes (when signals from different wind turbine generators are combined). In this respect, the signals from each generator or transmission line have characteristics reflecting the individual inertia, time constant, delay, and capacitance of the generator and / or transmission line. For example, a pair of independent wind turbine generators within a power plant may produce slightly out-of-phase outputs.

[0105] To assess the probability of such an event occurring within a predefined time window, PPC 8 can process various types of data, including the current state of grid 19 and / or indications of demand received from grid 19 (such as the design of grid 19), which include known vulnerabilities and historical data indicating grid behavior. Furthermore, plant-level and turbine-level techniques for detecting weak grids are known and can be fed into the analysis performed by PPC 8. Forecasts by grid operators based on planned changes to grid 19 can also be taken into account.

[0106] Similar to other inputs to the analysis, the output of this analysis phase is a probability density function, which indicates the likelihood of a power grid event occurring within a predefined window.

[0107] Another form of uncertainty that can be considered is the need to power ancillary services, such as anti-freeze systems. Such systems may consume a significant amount of power internally, and therefore, if not taken into account, could interfere with power output.

[0108] Modeling the aforementioned inputs of the wind power plant 12 probabilistically within the context of optimizing the SoCs 44a and 46a of each energy storage device 44 and 46 results in a set of opportunity constraints that can be customized for each application. Limitations are applied to these opportunity constraints, each limiting the proportion of time during which the corresponding opportunity constraint may be violated in the long run.

[0109] The opportunity constraints are then transformed into hard device constraints for the operation of energy storage devices 44 and 46 using the input probability density function. These device constraints are then fed into opportunity-constrained model predictive control analysis to determine the optimal values ​​of SoC 44a and 46a for each energy storage device 44 and 46.

[0110] The following are some examples of chance constraints that can be used as precursors in chance-constrained model predictive control analysis.

[0111] The first opportunity constraint may be that the first and / or second energy storage devices 44, 46 have the ability to provide virtual inertia, and a typical limitation of this constraint may be that it must be met more than 99% of the time. This requires devices 44, 46 to have SoCs 44a, 46a at levels between the upper and lower limits 44b, 46b, 44c, 46c, thereby enabling devices 44, 46 to charge and / or discharge as needed to provide inertia (e.g., when there are fluctuations in the power production of the wind turbine generator 1 of the wind farm 12).

[0112] A similar opportunity constraint could be the ability to ensure that the first and / or second energy storage units 44, 46 have grid frequency control for more than 99% of the time, especially where the wind farm 12 helps enable the grid 19 to respond quickly to changes in demand and / or production. Therefore, this opportunity constraint responds to grid events where the grid frequency deviates from the nominal level.

[0113] Other opportunity constraints related to grid events include the ability to ensure that the first and / or second energy storage devices 44, 46 have reactive power injection capabilities for more than 95% of the time, and / or the ability to supply both active and reactive power when needed. In this regard, including the first and / or second energy storage devices 44, 46 is expected to be particularly useful in mitigating the latter of these constraints.

[0114] Other opportunity constraints related to grid events include the ability to ensure grid oscillation damping for more than 99% of the time, the ability to maintain adherence to a specified power output curve for more than 99% of the time during low-voltage ride-through / undervoltage ride-through events, or the ability to limit the output P to grid 19 for more than 99% of the time. PPC The specified rate of increase and decrease.

[0115] Some opportunity constraints related to reducing structural loads on the wind turbine generator 1 include ensuring that the first and / or second energy storage devices 44, 46 have the capability to provide lateral torque damping and / or drivetrain torque damping for more than 98% of the time. Another opportunity constraint related to structural loads is ensuring that the energy storage devices 44, 46 have the capability to provide upward yaw control / hydraulic pressure for at least 5 minutes in the event of grid errors at critical wind speeds for more than 90% of the time. These opportunity constraints are primarily related to the first energy storage device 44, which, as mentioned above, can focus on enhancing the operation of the wind turbine generator 1 to reduce structural loads (particularly in response to grid events). However, the second energy storage device 46 can also be used for this purpose and therefore can also be controlled for these opportunity constraints.

[0116] Other relevant opportunity constraints are ensuring that the first and / or second energy storage devices 44, 46 have the ability to keep power variations or “power line flicker” at certain frequencies on the power spectrum below regulatory limits for more than 99% of the time.

[0117] The last example of an opportunity constraint that can be considered is the requirement to limit noise to a specified threshold at least 95% of the time, which typically requires controlling blade pitch and / or rotor speed to limit noise under certain wind conditions or at specific times.

[0118] The examples above illustrate typical opportunity constraints arising in the context of optimizing the SoCs 44a and 46 of energy storage devices 44 and 46 within a wind farm 12, and how these opportunity constraints can be prioritized by adjusting the corresponding risks of failing to meet each of them. Therefore, opportunity constraints that must be met more than 99% of the time (e.g., those related to auxiliary grid recovery) have a higher priority than those that only need to be met 95% of the time (e.g., noise limits). This prioritization can be adjusted for each application, and typically it will reflect the consequences of violating each constraint.

[0119] The risk of failing to meet each opportunity constraint is related to the storage constraint of each energy storage device. To some extent, failing to meet an opportunity constraint usually means that the device's storage limit has been reached. For example, if devices 44 and 46 are fully discharged, the opportunity constraint requiring energy storage devices 44 and 46 to discharge may not be met. Conversely, a fully charged device cannot consume power to meet the opportunity constraint.

[0120] As can be seen from the above, assigning corresponding constraints to each opportunity constraint can be considered as effectively specifying the probability of meeting grid requirements. Therefore, this probability can be explicitly specified in the optimization problem and thus used as a design parameter that can be adjusted for each application based on the requirements imposed by the grid on the wind farm and / or wind turbine generators. This, in turn, provides greater flexibility in how the wind farm 12 is operated, thereby allowing for enhanced optimization in various aspects. In particular, this method enables the maximization of the power output of the wind farm 12 while controlling the degree of violation of constraints related to grid requirements.

[0121] Figure 4 The above-described optimization process is illustrated graphically to show how it improves the performance of wind power plant 12 in practice. Figure 4 Two time-varying graphs are shown: the lower graph 50, which represents the control input of the power plant 12; and the upper graph 52, which represents the power production P of the wind power plant 12. PPC . Figure 4 The horizontal dashed lines in the diagram represent the limits within which power production must be kept to meet grid requirements.

[0122] The upper curve 52 includes a solid line surrounded by a shaded area. The solid line represents the deterministic equivalent output from wind power plant 12, i.e., the output that will be achieved without uncertainty. The shaded area represents the region within which the actual output may deviate due to a 99% probability of uncertainty. In other words, the shaded area represents the potential impact of chance constraints in the 99% of cases. Therefore, in order to meet grid requirements 99% of the time, this shaded area must remain within the upper and lower limits.

[0123] In this example, the shaded area has been calculated by applying Monte Carlo analysis to the results of a series of simulations of the system, in which different uncertainties are randomly implemented.

[0124] Figure 5 This is a flowchart summarizing the process 60 used to control the wind power plant 12, which in this example is executed by PPC8.

[0125] Process 60 begins with PPC 8 processing data associated with one or more inputs of wind power plant 12 in step 62 to probabilistically model the input uncertainties to obtain a probabilistic prediction for each input of power plant 12. For example, the probabilistic prediction may be in the form of a probability density function. Data associated with the inputs of wind power plant 12 may include data indicating the state of the power grid 19 and / or data indicating wind conditions, which is received at input 9 of PPC 8. In one embodiment, only the state of the power grid 19 may be considered, particularly if only the second energy storage device 46 is present in wind power plant 12.

[0126] Process 60 continues, where PPC 8 determines, in step 64, the relevant opportunity constraints governing the operation of wind power plant 12. Opportunity constraints relate to meeting operational requirements, and the limit for each opportunity constraint is the probability of meeting the constraint, typically expressed as a percentage of operating time during which the constraint will not be violated in the long run. For example, the limit for successfully delivering active power to grid 19 when needed could be set to 99% of the time, meaning that in the long run, wind power plant 12 is allowed to fail to deliver the active power required by the grid for at most 1% of the time.

[0127] Each constraint is determined based on operational objectives and grid requirements, and therefore differs for each application. Determining opportunity constraint constraints may require retrieving them from memory, receiving constraints from an external source such as grid 19, or having the user define constraint constraints via the PPC 8 interface.

[0128] PPC 8 then uses the probability density function generated for the input of wind power plant 12 in step 66 to convert the opportunity constraint limits identified in the previous step into hard device constraints for the SoCs 44a and 46a of each energy storage device 44 and 46. For example, these device constraints may include upper and lower thresholds for the SoCs 44a and 46a of each device, as well as the maximum charge or discharge rate.

[0129] The device constraints are configured such that if they are complied with and the energy storage devices 44, 46 operate within them, each opportunity constraint will not be violated beyond its corresponding limit. Therefore, referring again to the example above, satisfying the device constraints of each energy storage device 44, 46 will ensure that, in the long term, the wind power plant 12 delivers active power to the grid 19 when needed at least 99% of the time.

[0130] The next step in process 60 is to determine a charge / discharge strategy for each energy storage device 44, 46 in step 68 based on device constraints derived from the SoC 44a, 46a for each energy storage device 44, 46. For example, the charge / discharge strategy may take the form of a corresponding setpoint for each energy storage device 44, 46. Alternatively, the charge / discharge strategy may include the charge or discharge rate for each device 44, 46.

[0131] The charging / discharging strategy is determined by solving a back-level optimization problem within a defined window using a model predictive control algorithm to find the optimal SoC 44a, 46a for each energy storage device 44, 46.

[0132] The optimization problem takes into account the initial conditions of the system, a set of equality constraints, a set of inequality constraints, and a cost function. Initial conditions include, for example, current wind conditions, grid conditions, and the SoC 44a, 46a of each storage device 44, 46. Equality constraints define a system model representing the system dynamics. Inequality constraints include device constraints derived in step 66 and other operational constraints (including the storage capacity of each storage device 44, 46). The cost function indicates the form of the solution required for the optimization problem and can be configured in various ways. For example, the cost function can be arranged to consider the difference between power production and the power output relative to the input and demand of the wind farm 12, as well as the deviation of the SoC 44a, 46a of each storage device 44, 46 from the desired value. Alternatively, the cost function can be configured to ensure that the wind farm 12 operates in the most cost-effective manner, for example, by maximizing power production while minimizing the structural load on the wind turbine generator 1 and the charging cycles of the energy storage devices 44, 46.

[0133] Solving the optimization problem generates control actions related to the charging / discharging strategies of energy storage devices 44 and 46, which, as described above, can be in the form of corresponding setpoints of the SoCs 44a and 46a of each device 44 and 46.

[0134] Finally, PPC 8 implements the charging / discharging strategy determined in the previous step and controls the charging and discharging of energy storage devices 44 and 46 in step 70 based on the output of the optimization problem (e.g., the setpoint SoC of each energy storage device 44, 46). Therefore, in a general sense, energy storage devices 44, 46 are controlled based on probabilistic predictions of each input (such as grid conditions).

[0135] It should be noted that control of the first energy storage device 44 typically requires the PPC 8 to issue control commands to the converter controller 36. In this embodiment, the second energy storage device 46 is directly controlled by the PPC 8.

[0136] The process 60 is then repeated continuously, in which PPC 8 first updates the probabilistic model of the input uncertainty, and then updates the chance constraints and device constraints accordingly. The charging / discharging strategy (e.g., the corresponding charging setpoint state) of each energy storage device 44, 46 is thus dynamically updated to reflect any changes in the input data, the charging state of the devices 44, 46, and / or the constraints.

[0137] Those skilled in the art will understand that modifications can be made to the specific embodiments described above without departing from the inventive concept defined in the claims.

Claims

1. A method for controlling a wind power plant (12) including energy storage devices (44, 46), the wind power plant (12) being connected to a power grid (19) and including one or more wind turbine generators (1) that produce electricity for transmission to the power grid (19), the method comprising: Processing grid data associated with the grid (19) to determine a probabilistic prediction of the future state of the grid (19), wherein the probabilistic prediction includes any one of the following: a cumulative distribution function and a probability density function; as well as A set of opportunity constraints related to the operation of the wind power plant (12) is determined, wherein each opportunity constraint has a corresponding limitation; Based on the corresponding constraints of each opportunity constraint and the probability prediction, a set of device constraints related to the charging and / or discharging of the energy storage devices (44, 46) is determined; and By solving an optimization problem for controlling the charging and discharging of the energy storage devices (44, 46), the charging and discharging of the energy storage devices (44, 46) are controlled according to the probability prediction to avoid violating the device constraints, wherein the optimization problem includes the device constraints.

2. The method according to claim 1, comprising: The control output based on the optimization problem controls the charging and discharging of the energy storage devices (44, 46).

3. The method according to claim 2, comprising using a prediction algorithm to solve the optimization problem.

4. The method according to claim 3, comprising using the prediction algorithm to simulate the operation of the wind power plant (12).

5. The method according to claim 3 or claim 4, wherein, The prediction algorithm is based on the chance-constrained model to predict control strategies.

6. The method according to any one of claims 1 to 4, wherein, Each restriction limits the proportion of the operating time of the wind power plant (12) that may violate the corresponding opportunity constraint.

7. The method according to any one of claims 1 to 4, wherein, Each of the set of opportunity constraints includes a requirement to provide any of the following: virtual inertia; grid frequency control; finite rate of rise and fall of power delivered to the grid; defined power output curve during voltage ride-through events; reduction of power line flicker; reactive power injection; grid oscillation damping; lateral torque damping; drivetrain torque damping; up yaw control. And noise below the threshold level.

8. The method according to any one of claims 1 to 4, comprising determining a charging state setpoint of the energy storage device (44, 46), and controlling the charging and discharging of the energy storage device (44, 46) according to the charging state setpoint.

9. The method according to any one of claims 1 to 4, comprising processing data indicating wind conditions to determine a probability prediction of wind conditions, and controlling the charging and discharging of the energy storage device (44, 46) based on the probability prediction of wind conditions.

10. The method according to any one of claims 1 to 4, comprising controlling the charging and discharging of the energy storage device (44, 46) according to a predetermined probability of violating one or more power grid requirements.

11. The method according to any one of claims 1 to 4, wherein, The grid data includes any of the following: data indicating the state of the grid (19); rate limits on the increase or decrease of power delivered to the grid (19); requests received from the grid (19); grid design data; historical grid data; data indicating a weak grid obtained from power plant-level and / or turbine-level analysis; electricity price data; and user-input predictions of planned changes to the grid (19).

12. The method according to any one of claims 1 to 4, comprising changing the operation of one or more wind turbine generators (1) of the wind power plant (12) according to the probability prediction or each probability prediction, and controlling the charging and discharging of the energy storage device (44, 46) according to the changed operation of the one wind turbine generator or each wind turbine generator (1).

13. The method according to any one of claims 1 to 4, wherein, The energy storage devices (44, 46) are electrically coupled to the common coupling point (42) connecting the wind power plant (12) and the power grid (19), or integrated into the wind turbine generator (1) of the wind power plant (12).

14. A control system (8, 27, 36) for a wind power plant (12) including energy storage devices (44, 46), the wind power plant (12) being connected to a power grid (19) and including one or more wind turbine generators (1) that generate electricity for transmission to the power grid (19), the control system (8, 27, 36) comprising: The input terminal is configured to receive power grid data related to the power grid (19); The processing module (10) is configured as follows: Processing the power grid data to determine a probabilistic prediction of the future state of the power grid (19), wherein the probabilistic prediction includes any one of the following: a cumulative distribution function and a probability density function; A set of opportunity constraints related to the operation of the wind power plant (12) is determined, wherein each opportunity constraint has a corresponding limitation; Based on the corresponding constraints of each opportunity constraint and the probability prediction, a set of device constraints related to the charging and / or discharging of the energy storage devices (44, 46) is determined; and By solving an optimization problem for controlling the charging and discharging of the energy storage devices (44, 46), control signals are generated based on the probabilistic prediction to control the charging and discharging of the energy storage devices (44, 46) to avoid violating the device constraints, wherein the optimization problem includes the device constraints; and The output terminal is configured to output the control signal.

15. A wind power plant (12) comprising the control system according to claim 14.

Images