Wind power plant distributed control method, device and equipment based on data driving sensitivity and medium

Abstract

The invention discloses a distributed control method, device, equipment and medium for a wind power plant based on data-driven sensitivity, and relates to the field of wind power plant control, the method comprises the following steps: estimating the state quantity of the wind power plant by using a data-driven model, and calculating the gradient of the data-driven model by using a back propagation algorithm, determining a power data driving sensitivity and a voltage data driving sensitivity; based on the estimated wind power plant state quantity, a mathematical analysis method is adopted to calculate mathematical method voltage sensitivity; calculating loss functions in different control modes based on the power data driving sensitivity, the voltage data driving sensitivity and the mathematical method voltage sensitivity; converting a nonlinear constraint condition into a linear constraint condition based on a variable spacing constraint linearization method, solving the loss function, and determining a control reference value of each wind turbine generator; and controlling each wind turbine generator of the wind power plant based on the control reference value of each wind turbine generator. The distributed control of the wind power plant can be realized, and the control effect is improved.

Description

Technical Field

[0001] The present application relates to the field of wind farm control, and in particular to a method, device, equipment and medium for distributed control of a wind farm based on data-driven sensitivity. Background Art

[0002] As a leading renewable energy source, the rapid development of wind power generation matches the complex power dispatching challenges of wind farms. Research on wind farm optimization control has become the key to improving efficiency, reducing costs, and enhancing the stability and reliability of wind power generation systems. With the continuous increase in the number of wind farm units, the control dimension has also increased significantly. The traditional centralized control method can no longer meet the requirements of large-scale wind farm control. Therefore, it is urgent to achieve rapid control of wind farms. Summary of the invention

[0003] The purpose of this application is to provide a method, device, equipment and medium for distributed control of wind farms based on data-driven sensitivity, which can realize distributed control of wind farms and effectively improve the control effect.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a wind farm decentralized control method based on data-driven sensitivity, comprising:

[0006] Based on the operating data of local wind turbines, the data-driven model is used to estimate the state of the wind farm;

[0007] Based on the estimated state quantity of the wind farm, a gradient of the data-driven model is calculated using a back propagation algorithm to determine a power data-driven sensitivity and a voltage data-driven sensitivity;

[0008] Based on the estimated wind farm state quantity, the voltage sensitivity is calculated by mathematical analysis method;

[0009] Calculate the loss function under different control modes based on the power data driving sensitivity, the voltage data driving sensitivity and the mathematical method voltage sensitivity;

[0010] The nonlinear constraints are transformed into linear constraints based on the variable spacing constraint linearization method;

[0011] Solving the loss function under different control modes based on the linear constraint conditions to determine the control reference value of each wind turbine generator set;

[0012] Each wind turbine generator set in the wind farm is controlled based on the control reference value of each wind turbine generator set.

[0013] In a second aspect, the present application provides a wind farm decentralized control device based on data-driven sensitivity, comprising:

[0014] A wind farm state quantity prediction module is used to estimate the wind farm state quantity using a data-driven model based on the operating data of local wind turbines;

[0015] A data-driven sensitivity determination module, used to calculate the gradient of the data-driven model based on the estimated wind farm state quantity using a back propagation algorithm to determine the power data-driven sensitivity and the voltage data-driven sensitivity;

[0016] A mathematical voltage sensitivity calculation module, used for calculating mathematical voltage sensitivity based on an estimated wind farm state quantity;

[0017] A loss function calculation module, used to calculate the loss function under different control modes based on the power data driving sensitivity, the voltage data driving sensitivity and the mathematical method voltage sensitivity;

[0018] A conversion module, used for converting nonlinear constraints into linear constraints based on a variable spacing constraint linearization method;

[0019] A control reference value determination module, used to determine the control reference value of each wind turbine generator set based on the linear constraint condition and the loss function under different control modes;

[0020] The control module is used to control each wind turbine set in the wind farm based on the control reference value of each wind turbine set.

[0021] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned wind farm distributed control method based on data-driven sensitivity.

[0022] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the above-mentioned wind farm decentralized control method based on data-driven sensitivity is implemented.

[0023] According to the specific embodiments provided in this application, this application has the following technical effects:

[0024] The present application provides a method, device, equipment and medium for distributed control of a wind farm based on data-driven sensitivity. The data-driven sensitivity (including power data-driven sensitivity and voltage data-driven sensitivity) can be determined through the operating data of the local wind turbine and the data-driven model. The voltage sensitivity can be combined with the mathematical method to determine the loss function under different control modes. The control reference value of each wind turbine can be determined by solving the loss function to achieve control of each wind turbine. In the above-mentioned distributed control process, the local wind turbine no longer needs to communicate with the central processor or the controller of the adjacent wind turbine, thereby improving the control speed and effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1 A schematic flow chart of a wind farm decentralized control method based on data-driven sensitivity provided in one embodiment of the present application;

[0027] Figure 2 A principle framework diagram of a wind farm decentralized control method based on data-driven sensitivity provided in one embodiment of the present application;

[0028] Figure 3 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0030] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0031] In an exemplary embodiment, Figure 1-Figure 2As shown, a method for decentralized control of a wind farm based on data-driven sensitivity is provided, which is executed by a computer device, specifically, it can be executed by a computer device such as a terminal or a server alone, or it can be executed by a terminal and a server together. In the embodiment of the present application, the method is applied to a server as an example for explanation, and includes the following steps S1 to S7. Among them:

[0032] S1: Based on the operating data of local wind turbines, the data-driven model is used to estimate the state of the wind farm.

[0033] S2: Based on the estimated state quantity of the wind farm, a back propagation algorithm is used to calculate the gradient of the data-driven model, and a power data-driven sensitivity and a voltage data-driven sensitivity are determined.

[0034] S3: Based on the estimated wind farm state quantity, the mathematical analysis method is used to calculate the mathematical voltage sensitivity.

[0035] S4: Calculating loss functions under different control modes based on the power data driving sensitivity, the voltage data driving sensitivity and the mathematical method voltage sensitivity.

[0036] S5: Nonlinear constraints are converted into linear constraints based on the variable spacing constraint linearization method.

[0037] S6: Solving the loss functions under different control modes based on the linear constraints to determine the control reference value of each wind turbine generator set.

[0038] S7: Controlling each wind turbine generator set in the wind farm based on the control reference value of each wind turbine generator set.

[0039] In a specific embodiment, step S1 specifically includes:

[0040] The data-driven model uses a deep learning model to learn the relationship between the operating data of the local wind turbine and the global state of the nodes and wind turbines in the wind farm at the current moment. This embodiment uses the Seg-RNN model as an example to construct a data-driven model.

[0041] In view of the different characteristics of each operating data, a corresponding mapping model will be built for each variable in the data-driven model. The formula of the mapping model is as follows:

[0042] V g =DDM(V i )

[0043] V i =[U i ,θ i ,P i ,Q i ]

[0044]

[0045] Among them, DDM represents data-driven model, V i Represents the historical operation data of the i-th wind turbine (local wind turbine). g V i The data length. i,l ,θ i,l , P i,l , Q i,l V represents the voltage amplitude, phase angle, active power and reactive power of the i-th wind turbine in time step l. g Represents the wind farm state quantity estimated by the data-driven model.

[0046] In a specific embodiment, step S2 specifically includes:

[0047] This application uses the forward propagation, loss calculation, and back propagation steps of the back propagation algorithm to calculate the gradient of the output of the data-driven model with respect to the input.

[0048] Step 1: Connect V i Input to the data driven model to get V g .

[0049] Step 2: Connect V g Defined as the loss function value.

[0050] Step 3: Back propagation to calculate V g P i and Q i The gradient of V g It can be the voltage of each node in the wind farm or the active and reactive power of each wind turbine.

[0051] Define the voltage of each node in the wind farm as P i and Q i The gradient of is the voltage data driving sensitivity, and the active and reactive power of each wind turbine is related to P i and Q i The gradient is the power data driving the sensitivity.

[0052] This embodiment takes the Seg-RNN model as an example to introduce how to calculate the voltage data driving sensitivity and the power data driving sensitivity.

[0053] The encoder part of the Seg-RNN model retains the transmission structure of the traditional RNN, but changes the input of each recurrent unit to sequence data. Taking the Seg-RNN model encoder using the GRU structure as an example, the formula is as follows:

[0054] z t =σ(W zx ·xt +W zh ·h (t-1) +b z ) ↓

[0055] r t =σ(W rx ·x t +W rh ·h (t-1) +b r ) ↓

[0056]

[0057] Where σ represents the sigmoid function. t is the input sequence data of the GRU unit. t and r t Corresponding to the update gate and reset gate that control the information flow. Represents the candidate hidden state, which is mainly responsible for retaining information. t is the hidden state of the unit output. zx , W zh , W rx , W rx , W hh and W hx is the learnable weight of the corresponding variable, b z 、b r and b h is the learnable bias. “·” denotes matrix multiplication. Represents element-wise multiplication.

[0058] When calculating the GRU unit for the kth time, x t By V i The kth segment of sequence information V i k It is obtained through the fully connected layer, and the formula is as follows:

[0059]

[0060] Among them, W x,k and Indicates that V i k Convert to x t The learnable weights and biases of y k Represents the output of the kth data-driven model and represents the estimated value of the state quantities of other nodes. Denotes the hidden state of the kth data-driven model in the decoder. and Indicates that Convert to yk The learnable weights and biases.

[0061] The i-th local controller uses the back propagation algorithm to calculate the data-driven model output y k For the active power P of the i-th wind turbine i and reactive power Q i The gradient of . The calculation formula of data-driven sensitivity is as follows:

[0062]

[0063]

[0064] The resulting gradient and This is data-driven sensitivity. In this embodiment, a total of four data-driven models are constructed to estimate the voltage and phase of each node in the wind farm, and the active power and reactive power of each wind turbine. k (k=1,2,3,4) represents the global voltage, phase, active power and reactive power output by the data-driven model. k When is voltage, and is the voltage data driver sensitivity, y k When is active power and reactive power, and is the power data driver sensitivity.

[0065] In a specific embodiment, step S3 specifically includes:

[0066] Calculate the voltage U of the i-th node by mathematical analysis i and phase θ i The active power P of the first wind turbine l and reactive power Q l The sensitivity of phase to active power and reactive power is a mathematical relationship used to describe how changes in voltage phase angle (θ) affect changes in active power (P) and reactive power (Q). The formula is as follows:

[0067]

[0068] in, express The conjugation of.

[0069] In a specific embodiment, step S4 specifically includes: calculating the loss function under the first control mode based on the voltage data driving sensitivity and the wind farm state space equation; determining the local prediction model according to the power data driving sensitivity and the wind farm state space equation; calculating the loss function under the second control mode based on the local prediction model and the mathematical method voltage sensitivity; calculating the loss function under the third control mode based on the local prediction model and the voltage driving data sensitivity.

[0070] When the wind farm network parameters are inaccurate but need to be accurately calculated, the loss function under the first control mode is adopted; when the wind farm network parameters are accurate but need to be quickly calculated, the loss function under the second control mode is adopted; when the wind farm network parameters are inaccurate but need to be quickly calculated, the loss function under the third control mode is adopted.

[0071] This embodiment takes the control target of power tracking and maintaining voltage stability as an example to illustrate the construction process of the loss function. This method can also be used in other scenarios.

[0072] Taking into account the time delay of the control system, the active power and reactive power regulation of the wind turbine can be expressed as:

[0073]

[0074] Among them, ΔP WT is the active power increment; s is a complex variable; is the time delay coefficient of active power; is the reference value of active power increment; ΔQ WT is the reactive power increment; is the time delay coefficient of reactive power; is the reactive power increment reference value; is the reference value of active power; P WT (t0) is the measured value of active power at time t0; is the reference value of reactive power; Q WT (t0) is the measured value of reactive power at time t0.

[0075] From the above formula, the regulation model of active power and reactive power of wind turbine can be derived as follows:

[0076]

[0077] in, ΔP WT The first derivative of ; = ΔQ WT The first derivative of .

[0078] Then the state space equation of the wind farm is as follows:

[0079]

[0080] y=Cx

[0081]

[0082]

[0083] Among them, x is the state variable; is the first-order derivative of x; A is the system matrix; B is the control matrix; u is the control amount; y is the response result; C is the output matrix or observation matrix, which is 2N in dimension W *2N W The identity matrix of W is the number of wind turbines; and are the time delay coefficient of active power and reactive power of the i-th wind turbine respectively; and represents the active power and reactive power of the i-th wind turbine, i = 1, 2, ..., N W .

[0084] Using power data to drive sensitivity can build a linear relationship between the power of each unit, the formula is as follows:

[0085]

[0086] in, and represents the increment of active power and reactive power of the i-th wind turbine. and It represents the active and reactive data-driven sensitivity of the j-th wind turbine to the active and reactive data of the ith wind turbine (power data-driven sensitivity), which is obtained by the back propagation algorithm.

[0087] The linear relationship between the power of the local wind turbine and other wind turbines can be obtained as follows:

[0088]

[0089] Among them, S PQ is the data driven power sensitivity matrix.

[0090] Combining the power linear relationship with the wind farm state space equation, the local prediction model can be obtained:

[0091]

[0092] Among them, x i is the state variable of the local prediction model of the i-th wind turbine; For x i The first derivative of ; is the system matrix of the local prediction model; is the control matrix of the local prediction model; u i For x i The amount of control; It is the output matrix or observation matrix of the local prediction model, which is a unit matrix with a dimension of 2*2.

[0093] The proportional distribution strategy is used for power tracking. The active power of the i-th wind turbine should be very close to its proportional distribution reference Therefore, the power tracking loss function Obj P It can be described as:

[0094]

[0095] Among them, N p is the total number of prediction steps in MPC; is the tracking power of the wind farm. is the power increment of the i-th wind turbine when the prediction step number is k, is the power reference value of the i-th wind turbine when the prediction step number is k, is the estimated power of the i-th wind turbine.

[0096] Voltage regulation loss function Obj V It can be described as:

[0097]

[0098] Among them, N WF is the number of nodes in the wind farm; and are the active power increment and reactive power increment of the i-th wind turbine respectively. and is the sensitivity of the voltage of the j-th wind turbine to the active and reactive power of the ith wind turbine. and It can be either the voltage sensitivity calculated by mathematical analysis or the voltage data driven sensitivity calculated by a data driven model. j (t0) Estimated voltage value of the j-th wind turbine generator set, It is the voltage reference value of the j-th wind turbine when the prediction step number is k.

[0099] Power tracking loss function Obj P and the voltage regulation loss function Obj V The weighted addition gives the total loss function Obj:

[0100] Obj=W p Obj P +W v Obj V

[0101] Among them, W p , W v are the weights of the power tracking loss function and the voltage regulation loss function, respectively.

[0102] In this embodiment, the loss functions in the three control modes have the same expression form. The difference is:

[0103] First control mode: loss function Obj V In and The sensitivity is driven using voltage data.

[0104] Second control mode: Substitute the local prediction model into the loss function Obj to obtain a loss function that only contains the local control amount:

[0105] Obj local =W p Obj P,loacl +W v Obj V,loacl

[0106]

[0107] in, and Use mathematical method to calculate voltage sensitivity.

[0108] The third control mode: Change the loss function of the second control mode to and The sensitivity is driven using voltage data.

[0109] In a specific embodiment, step S5 specifically includes:

[0110] In the current control method, due to the rated capacity limit of the wind turbine, the active and reactive power inputs are restricted by quadratic inequality constraints. Directly applying this nonlinear inequality constraint to predictive control may result in too long a calculation time, making it difficult to meet the control requirements. Therefore, the present application proposes a variable spacing constraint linearization method to approximate the nonlinear constraint within a specific region of linearization. The default constraint condition is greater than zero. If the corresponding nonlinear function is a concave function, a hyperplane formed by the endpoints of the region is used as an alternative linear function; if it is a convex function, a hyperplane parallel to the hyperplane formed by the endpoints and tangent to the function is used as an alternative linear function. A growth function is used to determine the radius of the region, gradually increasing from small to large. The small radius region effectively retains the feasible area of ​​the nonlinear constraint, while the large radius region reduces the total number of linear constraints, thereby reducing the computational burden of MPC solution.

[0111] The specific operation process of the algorithm is as follows:

[0112] Step 1: Convert all constraints to a form greater than zero and obtain the corresponding nonlinear functions.

[0113] Step 2: Determine the function used to generate the region radius, the number of region divisions, and the inflection point of the function to divide the linearization region.

[0114] Step 3: Perform linear substitution based on the divided regions and the concavity of the function in each region.

[0115] Step 4: Derive the linearized constrained inequalities.

[0116] In this embodiment, the constraint conditions include active power constraint of the wind turbine generator set and reactive power constraint of the wind turbine generator set.

[0117] The active power constraint of the wind turbine is a linear constraint. The active power constraint of the i-th wind turbine is as follows:

[0118]

[0119] in, is the available active power of the i-th wind turbine.

[0120] The reactive power constraints of the first wind turbine are as follows:

[0121]

[0122]

[0123] The upper and lower limits of the reactive power constraints of wind turbines include active power and are nonlinear constraints.

[0124] Using the proposed variable spacing constraint linearization method, it can be transformed into the following linear constraints.

[0125]

[0126] n=1,…,N n ,m=1,…,N m

[0127] Among them, N n and N m represent the number of upper linear constraints and lower linear constraints respectively, and is the proportionality coefficient, and is a constant term.

[0128] By using optimized linear constraints, the calculation time of the control algorithm can be greatly reduced.

[0129] In a specific embodiment, step S6 specifically includes:

[0130] According to the loss function obtained in step S4 and the linear constraints obtained in step S5, the following quadratic programming problem can be obtained.

[0131] The first control mode is:

[0132]

[0133] ikB ln P i (k)+b ln ≤Q i (k)≤K um P i (k)+b um

[0134] The second control mode is:

[0135]

[0136] ikB ln P l (k)+b ln ≤Q l (k)≤K um P l (k)+b um

[0137] The third control mode is:

[0138]

[0139] ikB ln Pl (k)+b ln ≤Q l (k)≤K um P l (k)+b um

[0140] The above quadratic programming problem can be solved using a standard optimization solver. In the first control mode, the reference values ​​of all fans are obtained, and then the reference values ​​of the corresponding fans are taken as the local control quantities. In the second and third control modes, the reference values ​​of the local control quantities are directly obtained.

[0141] The wind farm decentralized control method based on data-driven sensitivity provided in the present application establishes a data-driven model using a deep learning model, and obtains the gradient of the data-driven model using a back-propagation algorithm, that is, the data-driven sensitivity (including power data-driven sensitivity and voltage data-driven sensitivity) of each wind turbine unit or each node state change in the wind farm to the operating data of the local wind turbine unit. Data-driven sensitivity can be substituted into MPC control to optimize existing control, such as power-to-power sensitivity can reduce the number of variables to be solved in the objective function of the control problem, and reduce the dimension of the objective function of the control problem; voltage-to-power sensitivity can replace the sensitivity based on mathematical models in existing control methods, and can increase stability in situations such as the lack of wind farm network parameters.

[0142] With the decentralized control method proposed in this application, there is no need to communicate with the central processor or the controller of the adjacent wind turbine during the control process of the local wind turbine. At the same time, this method can deploy the centralized control method to the local wind turbine, combined with data-driven sensitivity, while greatly reducing the control dimension and achieving an effect close to centralized control.

[0143] Based on the same inventive concept, the embodiment of the present application also provides a device for implementing the above-mentioned wind farm distributed control method based on data-driven sensitivity. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the embodiments of one or more wind farm distributed control devices based on data-driven sensitivity provided below can refer to the limitations of the wind farm distributed control method based on data-driven sensitivity above, and will not be repeated here.

[0144] In an exemplary embodiment, a wind farm decentralized control device based on data-driven sensitivity is provided, comprising:

[0145] The wind farm state quantity prediction module is used to estimate the wind farm state quantity using a data-driven model based on the operating data of local wind turbines.

[0146] The data-driven sensitivity determination module is used to calculate the gradient of the data-driven model based on the estimated wind farm state quantity by using a back propagation algorithm to determine the power data-driven sensitivity and the voltage data-driven sensitivity.

[0147] The mathematical voltage sensitivity calculation module is used to calculate the mathematical voltage sensitivity based on the estimated wind farm state quantity.

[0148] The loss function calculation module is used to calculate the loss functions under different control modes based on the power data driving sensitivity, the voltage data driving sensitivity and the mathematical method voltage sensitivity.

[0149] The conversion module is used to convert nonlinear constraints into linear constraints based on the variable spacing constraint linearization method.

[0150] The control reference value determination module is used to determine the control reference value of each wind turbine generator set based on the linear constraint condition and the loss function under different control modes.

[0151] The control module is used to control each wind turbine set in the wind farm based on the control reference value of each wind turbine set.

[0152] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program. The computer device may be a server or a terminal, and its internal structure diagram may be as shown in FIG. Figure 3 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data to be processed. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a distributed control method for a wind farm based on data-driven sensitivity is implemented.

[0153] Those skilled in the art will understand that Figure 3The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0154] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0155] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0156] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0157] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0158] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0159] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A wind farm decentralized control method based on data-driven sensitivity, characterized in that: include: Based on the operating data of local wind turbines, the data-driven model is used to estimate the state of the wind farm; Based on the estimated state quantity of the wind farm, a gradient of the data-driven model is calculated using a back propagation algorithm to determine a power data-driven sensitivity and a voltage data-driven sensitivity; Based on the estimated wind farm state quantity, the voltage sensitivity is calculated by mathematical analysis method; Calculate the loss function under different control modes based on the power data driving sensitivity, the voltage data driving sensitivity and the mathematical method voltage sensitivity; Convert nonlinear constraints into linear constraints based on variable spacing constraint linearization method; Solving the loss function under different control modes based on the linear constraint conditions to determine the control reference value of each wind turbine generator set; Each wind turbine generator set in the wind farm is controlled based on the control reference value of each wind turbine generator set.

2. The method for decentralized wind farm control based on data-driven sensitivity according to claim 1, characterized in that: The data-driven model is built based on the Seg-RNN model.

3. The method for decentralized wind farm control based on data-driven sensitivity according to claim 1, characterized in that: The loss functions under different control modes are calculated based on the power data driving sensitivity, the voltage data driving sensitivity and the mathematical method voltage sensitivity, specifically including: Calculate the loss function in the first control mode based on the voltage data driving sensitivity and the wind farm state space equation; Determining a local prediction model based on the power data driven sensitivity and the wind farm state space equation; Calculate the loss function in the second control mode based on the local prediction model and the mathematical voltage sensitivity; A loss function in a third control mode is calculated based on the local prediction model and the voltage drive data sensitivity.

4. The method for decentralized wind farm control based on data-driven sensitivity according to claim 3, characterized in that: The expression of the wind farm state space equation is: y=Cx in, is the first-order derivative of the state variable x, A is the system matrix, B is the control matrix, u is the control amount, y is the response result, C is the output matrix or observation matrix, which is a 2N dimension W ×2N W The identity matrix of is the active power increment of the i-th wind turbine, i=1,2,...,N W , N W is the number of wind turbines, is the reactive power increment of the i-th wind turbine, is the reference value of the active power increment of the i-th wind turbine generator set, is the reactive power increment reference value of the i-th wind turbine group, is the time delay coefficient of the active power of the i-th wind turbine; is the time delay coefficient of the reactive power of the i-th wind turbine.

5. The method for decentralized wind farm control based on data-driven sensitivity according to claim 4, characterized in that: The expression of the local prediction model is: in, is the state variable x of the local prediction model of the i-th wind turbine i The first derivative of is the system matrix of the local prediction model, is the control matrix of the local prediction model, u i For x i The amount of control; is the output matrix or observation matrix of the local prediction model, S PQ Drive the sensitivity matrix for the power data.

6. The method for decentralized wind farm control based on data-driven sensitivity according to claim 5, characterized in that: The expressions of loss functions under different control modes are: Obj=W p Obj P +W v Obj V in, and is the active and reactive sensitivity of the voltage of the j-th wind turbine to the i-th wind turbine, W p is the power tracking loss function Obj P The weight, W v is the voltage regulation loss function Obj V The weight, N p is the total number of prediction steps; is the power reference value of the i-th wind turbine when the prediction step number is k, is the estimated power value of the i-th wind turbine at time t0, is the power increment of the i-th wind turbine when the number of prediction steps is k, N WF is the number of nodes in the wind farm, U j (t0) is the estimated voltage value of the j-th wind turbine generator set at time t0, is the voltage reference value of the j-th wind turbine when the prediction step number is k, and are the active power increment and reactive power increment of the i-th wind turbine when the prediction step number is k, respectively.

7. The method for decentralized wind farm control based on data-driven sensitivity according to claim 6, characterized in that: The linear constraints include wind turbine active power constraints and wind turbine reactive power constraints; The active power constraint of the wind turbine is: in, is the available active power of the i-th wind turbine; The reactive power constraint of the wind turbine is: n=1,…,N n ,m=1,…,N m Among them, P i (k) is the active power control quantity of the i-th wind turbine when the prediction step number is k, Q i (k) is the reactive power control value of the i-th wind turbine when the prediction step number is k, and is the proportionality coefficient, and is a constant term, N n and N m are the number of upper and lower linear constraints, respectively.

8. A wind farm distributed control device based on data-driven sensitivity, characterized in that: include: A wind farm state quantity prediction module is used to estimate the wind farm state quantity using a data-driven model based on the operating data of local wind turbines; A data-driven sensitivity determination module, used to calculate the gradient of the data-driven model based on the estimated wind farm state quantity using a back propagation algorithm to determine the power data-driven sensitivity and the voltage data-driven sensitivity; A mathematical voltage sensitivity calculation module, used for calculating mathematical voltage sensitivity based on an estimated wind farm state quantity; A loss function calculation module, used to calculate the loss function under different control modes based on the power data driving sensitivity, the voltage data driving sensitivity and the mathematical method voltage sensitivity; A conversion module, used for converting nonlinear constraints into linear constraints based on a variable spacing constraint linearization method; A control reference value determination module, used to determine the control reference value of each wind turbine generator set based on the linear constraint condition and the loss function under different control modes; The control module is used to control each wind turbine set in the wind farm based on the control reference value of each wind turbine set.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the wind farm decentralized control method based on data-driven sensitivity according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the decentralized control method for a wind farm based on data-driven sensitivity according to any one of claims 1 to 7 is implemented.

Images