A distributed dynamic control method and system for maximum power output of a wind farm

Abstract

The present invention discloses a decentralized dynamic control method and system for maximum power output at a wind farm. The method comprises constructing a wind farm total power maximization optimization problem and solving the global optimal axial induction factor of each wind turbine; constructing a power maximization optimization problem for the downstream wind turbine cluster affected by its wake for each wind turbine, and solving the axial induction factor of the downstream wind turbine cluster; using wind speed, wind direction, and the axial induction factor of the wind turbine as input, and subtracting the maximum total power of the downstream wind turbine cluster from the total power when the axial induction factor of the wind turbine is 0 as output to generate a training data set for the wind turbine; and training a data-driven model based on a deep learning model to estimate the power loss of the downstream wind turbine cluster affected by its wake, thereby constructing a local decentralized optimal control problem and solving the axial induction factor of the wind turbine. The present invention aims to overcome the shortcomings of traditional centralized control strategies in dynamic control of large-scale wind farms, such as slow computational speed, heavy communication burden, and low fault tolerance.

Description

Technical Field

[0001] The present invention belongs to the technical field of wind power generation, and in particular relates to a method and system for distributed dynamic control of maximum power output of a wind farm. Background Art

[0002] In practical operation, wind farms often face the problem of downstream wake effects caused by wind turbines capturing wind energy. This reduces the inflow wind speed to downstream turbines, reducing their effective output power, significantly impacting the overall efficiency and power generation of the wind farm. Traditional wind farm power optimization methods often use centralized control, which requires a centralized wind farm controller. However, its solution speed decreases significantly with the scale of the wind farm and the number of wind turbines, making it difficult to adapt to time-varying wind conditions. Distributed control and data-driven approaches play an important role in significantly improving computational efficiency and reducing communication complexity. However, their performance is highly dependent on reliable communication links. In the event of communication failures between wind turbines, control performance may degrade significantly. In contrast, decentralized control eliminates the need for a central or distributed controller, enabling each wind turbine to operate independently without communication. This not only further improves computational efficiency but also alleviates concerns about communication failures. However, due to the wake effect, there are complex aerodynamic couplings between wind turbines, making it extremely challenging to account for these interactions in a fully decentralized framework. Summary of the Invention

[0003] Technical problem to be solved by the present invention: In response to the above-mentioned problems of the prior art, a method and system for distributed dynamic control of the maximum power output of a wind farm are provided. The present invention aims to make up for the shortcomings of traditional centralized control strategies such as slow calculation speed, heavy communication burden, and low fault tolerance when dynamically controlling large-scale wind farms, thereby improving the calculation speed of optimized control of large-scale wind farms, eliminating complex centralized communications, and improving fault tolerance.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A method for distributed dynamic control of maximum power output of a wind farm comprises the following steps:

[0006] Under the conditions of different inflow wind speeds and wind directions of the wind farm, the maximum total power optimization problem of the wind farm is constructed, and the optimization algorithm is used to solve the global optimal axial induction factor of each wind turbine under different wind conditions;

[0007] For each wind turbine, a power optimization problem for the downstream wind turbine cluster affected by its wake is constructed under different axial induction factors. The axial induction factor of the downstream wind turbine cluster is solved using an optimization algorithm.

[0008] For each wind turbine, the wind speed, wind direction, and the axial induction factor of the wind turbine are used as inputs. The output is the subtraction of the total power of the downstream wind turbine cluster and the total power when the axial induction factor of the wind turbine is at the preset minimum value. This generates a training dataset for the wind turbine.

[0009] For each wind turbine, a data-driven model for estimating power loss of the downstream wind turbine cluster is constructed. Using the training dataset of the wind turbine, the data-driven model for estimating power loss of the downstream wind turbine cluster is trained.

[0010] For each wind turbine, a data-driven model for estimating the power loss of the downstream wind turbine cluster affected by the wind turbine's wake is combined with a mechanical power model and a data-driven model based on the individual wind turbine to construct a local decentralized optimal control problem. The optimal axial induction factor of the wind turbine is then solved for controlling the wind turbine.

[0011] Optionally, constructing a wind farm total power maximization optimization problem under conditions of different inflow wind speeds and wind directions, and using an optimization algorithm to solve the global optimal axial induction factor of each wind turbine under different wind conditions includes:

[0012] S101, calculate the number of wind turbines in the wind farm according to the following formula i Captured mechanical power :

[0013] ,

[0014] in, ρ is the air density, S=πR ² is the swept area of ​​the fan impeller, R is the radius of the fan impeller, a i 、 v i Fan i The axial induction factor and inflow velocity, For fans i The power coefficient, ;

[0015] S102, use the Jensen wake model to calculate the wake wind speed:

[0016] ,

[0017] ,

[0018] in, For fans i The inflow wind speed, is the inflow wind speed at the most upstream of the wind farm, Indicates the fan jThe collection of wind turbines that produce wake effects, For fans i For fans j The tail wind speed, is the wake expansion coefficient of the Jensen wake model, For wind turbines along the wind direction i and j The geographical distance between For fans i The wake induced by the wind turbine j Coverage area on the impeller plane;

[0019] S103, the fan j Inflow wind speed It is written as the function of the inflow wind speed at the most upstream of the wind farm and the axial induction factor of the upstream wind turbine that affects its wake as shown in the following formula: :

[0020] ,

[0021] The fan j The captured mechanical power Written as a function of the axial induction factor shown below :

[0022]

[0023] in, For fans j Axial induction factor;

[0024] The total power of the entire wind farm is the sum of the powers of all wind turbines Written as a function of the upstream inflow velocity and the axial induction factors of all fans :

[0025] ,

[0026] in, represents the set of all fan axial induction factors, n is the number of wind turbines in the wind farm;

[0027] S104, construct the objective function of the wind farm total power maximization optimization problem:

[0028] ,

[0029] Among them, max represents the function The maximum value of

[0030] S105, using a simulated annealing particle swarm optimization algorithm to solve the global optimal axial induction factor of each wind turbine under different wind conditions.

[0031] Optionally, the step of solving the problem using a simulated annealing particle swarm optimization algorithm includes:

[0032] S201, initializing the parameters of the particle swarm and simulated annealing, wherein the particle swarm is composed of the results of solving the problem;

[0033] S202, update the speed and position according to the iterative process of the particle swarm optimization algorithm:

[0034] ,

[0035] in, and Respectively represent k+ Particles at 1 iteration i speed and position, and Respectively represent k The particle i speed and position, w represents the inertia weight, c 1 and c 2 is a non-negative acceleration coefficient, r 1 and r 2 is a random number in the interval (0,1) used to enhance the randomness of the search process. and Respectively k After iterations, particles i The individual optimal position of and the global optimal position of the particle swarm;

[0036] S203, calculating the wind farm power output using the objective function of the problem to obtain the particle fitness of the current iteration;

[0037] S204, update the individual optimal position and value according to the fitness difference and annealing criterion:

[0038]

[0039] in, p k+1 For the system from k The status at the iteration x k Transfer to k Status at +1 iteration x k+1 The acceptance probability, E k+1 and E k The system isk +1 iteration and k The energy level of the iteration, T k Represents the temperature of the simulated annealing mechanism; if the current particle is better than the global optimal solution, the global optimal position and value are updated;

[0040] S205, reduce the temperature according to the annealing formula:

[0041] ,

[0042] in, and Respectively k+ 1 and k The temperature at the iteration, K is the simulated annealing coefficient, and 0< K <1;

[0043] S206 , determining whether the number of iterations reaches a preset threshold; if so, outputting the global optimal axial induction factor of each wind turbine under different wind conditions; otherwise, skipping to step S202 to continue iteration.

[0044] Optionally, for each wind turbine, constructing a power maximization optimization problem for a downstream wind turbine cluster affected by its wake in scenarios where the wind turbine adopts different axial induction factors, and solving the axial induction factor of the downstream wind turbine cluster using an optimization algorithm includes:

[0045] S301: For each fan, the axial induction factor is divided into M subintervals and randomly select a value from each subinterval, thus obtaining M Axial induction factor;

[0046] S302: For each wind turbine, find the downstream wind turbine cluster affected by the wind turbine wake according to the Jensen wake model and set it as a set D , the wind turbine clusters not affected by the typhoon's wake are set as a set N , the wind turbine is controlled by different axial induction factors, belonging to the set N The global optimal axial induction factor of the wind turbine under different wind conditions is constructed to belong to the set D The objective function of the optimization problem of maximizing the total power of the wind turbine cluster is:

[0047] ,

[0048] in, and Fan i The lower and upper limits of the axial induction factor;

[0049] For each wind turbine, a power optimization problem for the downstream wind turbine cluster affected by its wake is constructed under different axial induction factors. The axial induction factor of the downstream wind turbine cluster is solved using an optimization algorithm.

[0050] S303: For each wind turbine, a simulated annealing particle swarm optimization algorithm is used to solve the axial induction factor of the downstream wind turbine cluster.

[0051] Optionally, generating a training data set for the wind turbine includes: searching for M different axial induction factors that belong to the set D The wind turbine with the largest total power in the cluster is the one with the largest total power. The axial induction factor, wind speed and wind direction in the current scenario are taken as input, and the maximum total power and the axial induction factor of the wind turbine are calculated to be the preset minimum value. D The difference in the total power of the wind turbine cluster is taken as the output and combined into the training data set of the wind turbine.

[0052] Optionally, constructing a data-driven model for estimating power loss of a downstream wind turbine cluster for each wind turbine, and using a training data set of the wind turbine to train the data-driven model for estimating power loss of a downstream wind turbine cluster includes:

[0053] S401, construct the neuron shown in the following formula:

[0054] ,

[0055] in, is the input of the neuron, is the output of the neuron, represents the dimensional space, n x and n y They are x in and z Dimensions, is the weight matrix, is the bias, g represents the activation function;

[0056] S402, construct a multi-layer perceptron model based on neurons as shown in the following formula:

[0057] ,

[0058] ,

[0059] ,

[0060] in, Represents the multilayer perceptron model k Tier j The output of a neuron, is the activation function, the activation function Select the hyperbolic tangent function tanh, Represents the multilayer perceptron model k Tier j The weight of a neuron, Represents the multilayer perceptron model k -1 layer output vector, Represents the multilayer perceptron model k Tier j The bias of a neuron, ~ Represents the multilayer perceptron model k Layer 1~ N k The weight of a neuron, N k Represents the multilayer perceptron model k The number of neurons in the layer, ~ Represents the multilayer perceptron model k Layer 1~ N k The output of a neuron;

[0061] S403: For each wind turbine, construct a wind turbine model based on the multi-layer perceptron model. i Data-driven model for estimating power losses in downstream wind turbine clusters:

[0062] ,

[0063] ,

[0064] ,

[0065] in, Represents the model output of the multilayer perceptron model, the model output is the fan i Estimated power loss of downstream wind turbine cluster, x i =[ d i , v i , a i ] T represents the model input, d i 、 v i 、 ai Respectively represent fans i The inflow wind direction, inflow wind speed and axial induction factor, b 1~ b 3 represents the bias of the 1st to 3rd layers, w 1~ w 3 represents the weight of the 1st to 3rd layers, w k Indicates the k The weight matrix of the layer, b k Indicates the k The bias vector of the layer; ~ Represents the multilayer perceptron model k Layer 1~ N k The weight of a neuron, ~ Represents the multilayer perceptron model k Layer 1~ N k The bias of each neuron;

[0066] S404 : For each wind turbine, use the training data set of the wind turbine to train a data-driven model for estimating power loss of a cluster of wind turbines downstream of the wind turbine.

[0067] Optionally, for each wind turbine, combining a data-driven model for estimating power loss of a cluster of downstream wind turbines affected by the wake of the wind turbine, and constructing a local decentralized optimal control problem based on a mechanical power model and a data-driven model of the individual wind turbine, and solving the optimal axial induction factor of the wind turbine for controlling the wind turbine includes:

[0068] S501, construct the fan shown in the following formula i Capturing mechanical power P i Axial induction factor a i The relationship:

[0069] ,

[0070] in, ρ is the air density, S=πR ² is the swept area of ​​the fan impeller, R is the radius of the fan impeller, a i 、 v i Fan i Axial induction factor and inflow velocity;

[0071] S502, the fani Capturing mechanical power P i Axial induction factor a i The relationship between the two is combined with the power loss estimation data-driven model of the downstream wind turbine cluster to construct the objective function of the local decentralized optimal control problem. :

[0072] ,

[0073] in, x i =[ d i , v i , a i ] T represents the model input, d i 、 v i 、 a i Respectively represent fans i The inflow wind direction, inflow wind speed and axial induction factor, b 1~ b 3 represents the bias of the 1st to 3rd layers, w 1~ w 3 represents the weight of the 1st to 3rd layers, For fans i Rated power, For fans i The inflow wind speed is v i And the axial induction factor when working at rated power is the numerical solution of the following formula:

[0074] ;

[0075] S503, Objective Function of Local Decentralized Optimal Control Problem Axial induction factor a i The objective function of the local decentralized optimal control problem is calculated by the single variable piecewise function of The extreme points in each segment, as well as the extreme points at 0, a u The function values ​​at the three endpoints of , 1 / 3 are taken so that the objective function obj The maximum value corresponds to the axial induction factor a i As a fan i The optimal axial induction factor is used to control the fan i .

[0076] In addition, the present invention also provides a wind farm maximum power output distributed dynamic control system, comprising a microprocessor and a memory connected to each other, wherein the microprocessor is programmed or configured to execute the wind farm maximum power output distributed dynamic control method.

[0077] In addition, the present invention also provides a computer-readable storage medium, in which a computer program or instruction is stored. The computer program or instruction is programmed or configured to execute the wind farm maximum power output distributed dynamic control method through a processor.

[0078] In addition, the present invention also provides a computer program product, including a computer program or instructions, wherein the computer program or instructions are programmed or configured to execute the wind farm maximum power output decentralized dynamic control method through a processor.

[0079] Compared with the existing technology, the present invention can mainly achieve the following beneficial effects: The present invention proposes a data-model hybrid driven decentralized dynamic control method for the maximum power output of a wind farm. The method uses a deep neural network to train a power loss estimation data model of the downstream wind turbine cluster for each wind turbine, and converts the centralized optimization problem of the wind farm into a decentralized control problem for each wind turbine. This method can not only improve the calculation speed of the optimization control of large-scale wind farms, eliminate complex centralized communications, and improve fault tolerance, thereby making up for the shortcomings of traditional centralized control strategies in dynamic control of large-scale wind farms, such as slow calculation speed, heavy communication burden, and low fault tolerance, but also provides a new technical path for the maximum power output control of large-scale wind farms. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] Figure 1 Schematic diagram of the basic process of the method of the embodiment of the present invention.

[0081] Figure 2 Flowchart of the simulated annealing particle swarm optimization algorithm in an embodiment of the present invention.

[0082] Figure 3 It is a simulation system structure diagram of a simulation case in an embodiment of the present invention.

[0083] Figure 4 Graph showing the total power output of a wind farm under different control methods in an embodiment of the present invention.

[0084] Figure 5 Graphs of total power output of a wind farm under different control methods when certain wind turbines fail in an embodiment of the present invention. DETAILED DESCRIPTION

[0085] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention will be further described in detail below with reference to the accompanying drawings in the embodiments of the present invention.

[0086] like Figure 1 As shown, the distributed dynamic control method for maximum power output of a wind farm in this embodiment includes the following steps:

[0087] S1, under the conditions of different inflow wind speeds and wind directions of the wind farm, construct the optimization problem of maximizing the total power of the wind farm, and use the optimization algorithm to solve the global optimal axial induction factor of each wind turbine under different wind conditions;

[0088] S2: For each wind turbine, a power optimization problem for the downstream wind turbine cluster affected by its wake is constructed under different axial induction factors, and the axial induction factor of the downstream wind turbine cluster is solved using an optimization algorithm.

[0089] S3: For each wind turbine, using wind speed, wind direction, and the axial induction factor of the wind turbine as input, the maximum total power of the downstream wind turbine cluster is subtracted from the total power when the axial induction factor of the wind turbine is at a preset minimum value (e.g., 0), generating a training dataset for the wind turbine.

[0090] S4, for each wind turbine, building a data-driven model for estimating power loss of the downstream wind turbine cluster, and using the training data set of the wind turbine to train the data-driven model for estimating power loss of the downstream wind turbine cluster;

[0091] S5, for each wind turbine, a local decentralized optimal control problem is constructed by combining the power loss estimation data-driven model of the downstream wind turbine cluster affected by the wind turbine wake, and the mechanical power model and data-driven model based on the individual wind turbine. The optimal axial induction factor of the wind turbine is solved and used to control the wind turbine.

[0092] In step S1 of this embodiment, constructing a wind farm total power maximization optimization problem under conditions of different inflow wind speeds and wind directions, and using an optimization algorithm to solve the global optimal axial induction factor of each wind turbine under different wind conditions includes:

[0093] S101, calculate the number of wind turbines in the wind farm according to the following formula i Captured mechanical power :

[0094] ,

[0095] in, ρ is the air density, S=πR ² is the swept area of ​​the fan impeller, R is the radius of the fan impeller, ai 、 v i Fan i The axial induction factor and inflow velocity, For fans i The power coefficient is a function related to the axial induction factor and can be expressed as: ;

[0096] S102, due to the complex wake effect in the wind farm, when wind turbine j is affected by the wakes of multiple upstream wind turbines at the same time, its inflow wind speed is calculated using the wake superposition model. The Jensen wake model is used to calculate the wake wind speed:

[0097] ,

[0098] ,

[0099] in, For fans i The inflow wind speed, is the inflow wind speed at the most upstream of the wind farm, Indicates the fan j The collection of wind turbines that produce wake effects, For fans i For fans j The tail wind speed, is the wake expansion coefficient of the Jensen wake model, For wind turbines along the wind direction i and j The geographical distance between For fans i The wake induced by the wind turbine j Coverage area on the impeller plane;

[0100] S103, according to the wake model, the wind turbine j Inflow wind speed It is written as the function of the inflow wind speed at the most upstream of the wind farm and the axial induction factor of the upstream wind turbine that affects its wake as shown in the following formula: :

[0101] ,

[0102] The fan j The captured mechanical power Written as a function of the axial induction factor shown below :

[0103]

[0104] in, For fans j Axial induction factor;

[0105] According to the above two equations, the total power of the entire wind farm is the sum of the powers of all wind turbines, which can be written as a function of the upstream inflow wind speed and the axial induction factor of all wind turbines. Therefore, the total power of the entire wind farm is the sum of the powers of all wind turbines. Written as a function of the upstream inflow velocity and the axial induction factors of all fans :

[0106] ,

[0107] in, represents the set of all fan axial induction factors, n is the number of wind turbines in the wind farm;

[0108] S104, construct the objective function of the wind farm total power maximization optimization problem:

[0109] ,

[0110] Among them, max represents the function The maximum value of

[0111] S105, using a simulated annealing particle swarm optimization algorithm to solve the global optimal axial induction factor of each wind turbine under different wind conditions.

[0112] like Figure 2 As shown, the steps of solving the problem using the simulated annealing particle swarm optimization algorithm in this embodiment include:

[0113] S201, initializing the parameters of the particle swarm and simulated annealing, wherein the particle swarm is composed of the results of solving the problem;

[0114] S202, update the speed and position according to the iterative process of the particle swarm optimization algorithm:

[0115] ,

[0116] in, and Respectively represent k+ Particles at 1 iteration i speed and position, and Respectively represent k The particle i speed and position, w represents the inertia weight, c 1 and c 2 is a non-negative acceleration coefficient, r 1 and r2 is a random number in the interval (0,1) used to enhance the randomness of the search process. and Respectively k After iterations, particles i The individual optimal position of and the global optimal position of the particle swarm;

[0117] S203, calculating the wind farm power output using the objective function of the problem to obtain the particle fitness of the current iteration;

[0118] S204, update the individual optimal position and value according to the fitness difference and annealing criterion:

[0119]

[0120] in, p k+1 For the system from k The status at the iteration x k Transfer to k Status at +1 iteration x k+1 The acceptance probability, E k+1 and E k The system is k +1 iteration and k The energy level of the iteration, T k Represents the temperature of the simulated annealing mechanism; if the current particle is better than the global optimal solution, the global optimal position and value are updated;

[0121] S205, reduce the temperature according to the annealing formula:

[0122] ,

[0123] in, and Respectively k+ 1 and k The temperature at the iteration, K is the simulated annealing coefficient, and 0< K <1;

[0124] S206 , determining whether the number of iterations reaches a preset threshold; if so, outputting the global optimal axial induction factor of each wind turbine under different wind conditions; otherwise, skipping to step S202 to continue iteration.

[0125] In step S2 of this embodiment, for each wind turbine, constructing a power maximization optimization problem for a downstream wind turbine cluster affected by its wake under scenarios where the wind turbine adopts different axial induction factors, and solving the axial induction factor of the downstream wind turbine cluster using an optimization algorithm includes:

[0126] S301: For each fan, the axial induction factor is divided into M subintervals and randomly select a value from each subinterval, thus obtaining M Axial induction factor;

[0127] S302: For each wind turbine, find the downstream wind turbine cluster affected by the wind turbine wake according to the Jensen wake model and set it as a set D , the wind turbine clusters not affected by the typhoon's wake are set as a set N , the wind turbine is controlled by different axial induction factors, belonging to the set N The global optimal axial induction factor of the wind turbine under different wind conditions is constructed to belong to the set D The objective function of the optimization problem of maximizing the total power of the wind turbine cluster is:

[0128] ,

[0129] in, and Fan i The lower and upper limits of the axial induction factor;

[0130] For each wind turbine, a power optimization problem for the downstream wind turbine cluster affected by its wake is constructed under different axial induction factors. The axial induction factor of the downstream wind turbine cluster is solved using an optimization algorithm.

[0131] S303: For each wind turbine, a simulated annealing particle swarm optimization algorithm is used to solve the axial induction factor of the downstream wind turbine cluster.

[0132] In step S3 of this embodiment, generating the training data set for the wind turbine includes: searching for the wind turbine M different axial induction factors that belong to the set D The wind turbine with the largest total power in the cluster is the one with the largest total power. The axial induction factor, wind speed and wind direction in the current scenario are taken as input, and the maximum total power and the axial induction factor of the wind turbine are calculated to be the preset minimum value. D The difference in the total power of the wind turbine cluster is taken as the output and combined into the training data set of the wind turbine.

[0133] In step S4 of this embodiment, for each wind turbine, a data-driven model for estimating power loss of a downstream wind turbine cluster is constructed. Using the training data set of the wind turbine, the data-driven model for estimating power loss of a downstream wind turbine cluster of the wind turbine is trained, including:

[0134] S401, construct the neuron shown in the following formula:

[0135] ,

[0136] in, is the input of the neuron, is the output of the neuron, represents the dimensional space, n x and n y They are x in and z Dimensions, is the weight matrix, is the bias, g represents the activation function; based on the need for explicit deep learning models, a multi-layer perceptron deep learning model is used to construct a power loss estimation model for the downstream wind turbine cluster, whose basic building block is the above-mentioned neuron (artificial neuron);

[0137] S402, passed w and b The weighted bias of the neuron can linearly combine the input and activate the function g The linear combination output can be transformed nonlinearly. Based on this idea, a multilayer perceptron model is constructed based on neurons as shown in the following formula:

[0138] ,

[0139] ,

[0140] ,

[0141] in, Represents the multilayer perceptron model k Tier j The output of a neuron, is the activation function, the activation function Select the hyperbolic tangent function tanh, Represents the multilayer perceptron model k Tier j The weight of a neuron, Represents the multilayer perceptron model k -1 layer output vector, Represents the multilayer perceptron model k Tier j The bias of a neuron, ~ Represents the multilayer perceptron model k Layer 1~ N k The weight of a neuron, N k Represents the multilayer perceptron model k The number of neurons in the layer, ~ Represents the multilayer perceptron model k Layer 1~ N k The output of a neuron;

[0142] S403: For each wind turbine, construct a wind turbine model based on the multi-layer perceptron model. i Data-driven model for estimating power losses in downstream wind turbine clusters:

[0143] ,

[0144] ,

[0145] ,

[0146] in, Represents the model output of the multilayer perceptron model, the model output is the fan i Estimated power loss of downstream wind turbine cluster, x i =[ d i , v i , a i ] T represents the model input, d i 、 v i 、 a i Respectively represent fans i The inflow wind direction, inflow wind speed and axial induction factor, b 1~ b 3 represents the bias of the 1st to 3rd layers, w 1~ w 3 represents the weight of the 1st to 3rd layers, w k Indicates the k The weight matrix of the layer, b k Indicates the kThe bias vector of the layer; ~ Represents the multilayer perceptron model k Layer 1~ N k The weight of a neuron, ~ Represents the multilayer perceptron model k Layer 1~ N k The bias of each neuron;

[0147] S404 : For each wind turbine, use the training data set of the wind turbine to train a data-driven model for estimating power loss of a cluster of wind turbines downstream of the wind turbine.

[0148] In step S5 of this embodiment, for each wind turbine, combining a power loss estimation data-driven model for a cluster of downstream wind turbines affected by the wind turbine wake, and constructing a local decentralized optimal control problem based on the mechanical power model and data-driven model of the individual wind turbine, and solving the optimal axial induction factor of the wind turbine for controlling the wind turbine includes:

[0149] S501, construct the fan shown in the following formula i Capturing mechanical power P i Axial induction factor a i The relationship:

[0150] ,

[0151] in, ρ is the air density, S=πR ² is the swept area of ​​the fan impeller, R is the radius of the fan impeller, a i 、 v i Fan i Axial induction factor and inflow velocity;

[0152] S502, the fan i Capturing mechanical power P i Axial induction factor a i The relationship between the two is combined with the power loss estimation data-driven model of the downstream wind turbine cluster to construct the objective function of the local decentralized optimal control problem. :

[0153] ,

[0154] in,x i =[ d i , v i , a i ] T represents the model input, d i 、 v i 、 a i Respectively represent fans i The inflow wind direction, inflow wind speed and axial induction factor, b 1~ b 3 represents the bias of the 1st to 3rd layers, w 1~ w 3 represents the weight of the 1st to 3rd layers, For fans i Rated power, For fans i The inflow wind speed is v i And the axial induction factor when working at rated power is the numerical solution of the following formula:

[0155] ;

[0156] S503, Objective Function of Local Decentralized Optimal Control Problem Axial induction factor a i The univariate piecewise function of the local decentralized optimal control problem is the hyperbolic tangent function (activation function ) and the exponential function are differentiable, so the objective function of the local decentralized optimal control problem is calculated separately The extreme points in each segment, as well as the extreme points at 0, a u The function values ​​at the three endpoints of , 1 / 3 are taken so that the objective function obj The maximum value corresponds to the axial induction factor a i As a fan i The optimal axial induction factor is used to control the fan i .

[0157] In order to verify the effectiveness of the distributed dynamic control method for the maximum power output of a wind farm proposed in this embodiment, Figure 3The wind farm shown in the figure is tested. The test wind farm includes 7 feeders, each of which connects 7 wind turbines arranged in series, and the distance between each wind turbine is 400 meters in both the longitudinal and transverse directions. The electricity generated by the wind turbines is collected and transmitted to the external grid through 33kV / 155kV and 155kV / 380kV substations. Figure 3 From left to right in the figure, it is parallel to the line connecting wind turbine 1 and wind turbine 43, and the positive direction is defined as the counterclockwise direction when looking down at the wind farm.

[0158] Figure 5 A box plot shows the estimation error of the power loss estimation model for the downstream wind turbine cluster of wind turbines 1 through 42, when the wind direction is between 0° and 45°. Since wind turbines 43 through 49 have no downstream wind turbines, the box plots for these seven wind turbines are not shown. It can be seen that the average output error of the power loss estimation model for the downstream wind turbine cluster for each wind turbine is below 50 kW, with the average for most units below 10 kW, demonstrating that the trained model exhibits good accuracy. The error distribution range for the front-row wind turbines is significantly larger than that for the rear-row wind turbines. This is because the downstream wind turbine clusters of the front-row wind turbines have more wind turbines, resulting in deviations in model accuracy. Furthermore, the root mean square error (RMSE) of the power loss estimation model for the downstream wind turbine cluster for each wind turbine, operating under wind direction conditions of 0° to 45°, is shown in Table 1.

[0159] Table 1 Root mean square error of the downstream wind turbine cluster power loss estimation model

[0160]

[0161] Figure 4 The total power output of the wind farm under different control schemes is shown. Table 2 shows the difference between the total power output of the wind farm under normal operating conditions using the method proposed in this embodiment and the different control schemes. The greedy control method refers to the operation of all wind turbines in maximum power point tracking mode, while the centralized method refers to the centralized optimization problem solution based on the SA-PSO algorithm.

[0162] Table 2 The difference between the total power output of the wind farm proposed by the method in this embodiment and different control methods under normal working conditions

[0163]

[0164] Depend on Figure 4As can be seen from Table 2, compared with the conventional greedy control method (i.e., each wind turbine operates in maximum power tracking mode), the method proposed in this embodiment and the centralized method can significantly improve the total power output of the wind farm. In addition, the total power output of the wind farm of the method proposed in this embodiment is higher than that of the centralized method in certain periods. This is because the centralized method may still fall into the local optimum during these periods, while the method proposed in this embodiment happens to find a better solution than the centralized method. The average computation time of the method proposed in this embodiment is only 4.82×10 -4 seconds, while the centralized method takes 7.02 seconds. This proves that the method proposed in this embodiment can achieve similar or even better control performance than the centralized method based on the extremely short computing time.

[0165] Figure 5 The difference between the total wind farm power output of the proposed method and the other two different control methods is shown when wind turbines 1, 16, 34, and 49 fail, respectively. Table 3 shows the difference between the total wind farm power output of the proposed method and the different control methods under fault conditions.

[0166] Table 3 The difference between the total power output of the wind farm proposed by the method in this embodiment and different control methods under fault conditions

[0167]

[0168] from Figure 5 Table 3 shows that under the four fault scenarios, the maximum, average, and minimum differences between the method proposed in this embodiment and the greedy control method remain significant, while the average values ​​compared to the centralized method are all greater than zero, demonstrating the superiority of the method proposed in this embodiment. Furthermore, when fan 1 fails, the average difference between the method proposed in this embodiment and the centralized method is the highest, while the average difference is the lowest when fan 49 fails. Comparison with Table 2 reveals that when fan 1 fails, the average difference between the method proposed in this embodiment and the centralized method is even higher than under normal operating conditions, demonstrating that the method proposed in this embodiment exhibits greater adaptability to front-row fan failures.

[0169] In summary, the decentralized dynamic control method for maximum power output of a wind farm in this embodiment includes constructing a maximum total power optimization problem for the wind farm under conditions of different inflow wind speeds and wind directions in the wind farm, and using an optimization algorithm to solve the global optimal axial induction factor of each wind turbine under different wind conditions; selecting a certain wind turbine, and constructing a maximum power optimization problem for the downstream wind turbine cluster affected by its wake under scenarios where it adopts different axial induction factors, and using an optimization algorithm to solve the axial induction factor of the downstream wind turbine cluster; taking the wind speed, wind direction and the axial induction factor of the wind turbine as input, and subtracting the maximum total power of the downstream wind turbine cluster from the total power when the axial induction factor of the wind turbine is 0 as output, to generate a training data set for the wind turbine; for each wind turbine, training a data-driven model for estimating power loss of the downstream wind turbine cluster affected by its wake based on a deep learning model; for each wind turbine, constructing a local decentralized optimal control problem based on the mechanical power model and data-driven model of the individual wind turbine. The distributed dynamic control method for maximum power output of a wind farm in this embodiment can dynamically increase the overall power generation of the wind farm while taking into account the influence of the wake effect, thereby bringing higher economic benefits to operators. It does not require a central controller, reducing communication complexity. It has extremely fast calculation speed, can ensure real-time performance in actual engineering applications, and is fault-tolerant.

[0170] In addition, this embodiment also provides a wind farm maximum power output distributed dynamic control system, comprising a microprocessor and a memory connected to each other, wherein the microprocessor is programmed or configured to execute the wind farm maximum power output distributed dynamic control method.

[0171] In addition, this embodiment further provides a computer-readable storage medium, in which a computer program or instruction is stored. The computer program or instruction is programmed or configured to execute the wind farm maximum power output distributed dynamic control method through a processor.

[0172] In addition, this embodiment further provides a computer program product, including a computer program or instructions, which are programmed or configured to execute the wind farm maximum power output distributed dynamic control method through a processor.

[0173] Those skilled in the art should understand that the technical solution provided by the present invention may be in the form of a method, a system, or a computer program product. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The present invention is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the functions described in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including the instruction device, which implements the function specified in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0174] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiment. All technical solutions based on the concept of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for distributed dynamic control of maximum power output of a wind farm, characterized in that: The steps include: Under the conditions of different inflow wind speeds and wind directions, the wind farm total power maximization optimization problem is constructed, and the global optimal axial induction factor of each wind turbine under different wind conditions is solved by an optimization algorithm, which is a simulated annealing particle swarm optimization algorithm. For each wind turbine, a power optimization problem for the downstream wind turbine cluster affected by its wake is constructed under different axial induction factors. The axial induction factor of the downstream wind turbine cluster is solved using an optimization algorithm. For each wind turbine, the wind speed, wind direction, and the axial induction factor of the wind turbine are used as inputs. The output is the subtraction of the total power of the downstream wind turbine cluster and the total power when the axial induction factor of the wind turbine is at the preset minimum value. This generates a training dataset for the wind turbine. For each wind turbine, a data-driven model for estimating power loss of the downstream wind turbine cluster is constructed. Using the training dataset of the wind turbine, the data-driven model for estimating power loss of the downstream wind turbine cluster is trained. For each wind turbine, a local decentralized optimal control problem is constructed by combining a data-driven model for estimating the power losses of the downstream wind turbine cluster affected by the turbine's wake, with a mechanical power model and data-driven model for each wind turbine. The optimal axial induction factor of the wind turbine is then solved and used to control the wind turbine. For each wind turbine, the power maximization optimization problem of the downstream wind turbine cluster affected by its wake is constructed under the scenario of using different axial induction factors, and the axial induction factor of the downstream wind turbine cluster is solved by the optimization algorithm, including: S301: For each fan, the axial induction factor is divided into M subintervals and randomly select a value from each subinterval, thus obtaining M Axial induction factor; S302: For each wind turbine, find the downstream wind turbine cluster affected by the wind turbine wake according to the Jensen wake model and set it as a set D , the wind turbine clusters not affected by the typhoon's wake are set as a set N , the wind turbine is controlled by different axial induction factors, belonging to the set N The global optimal axial induction factor of the wind turbine under different wind conditions is constructed to belong to the set D The objective function of the optimization problem of maximizing the total power of the wind turbine cluster is: ， in, and Fan i The lower and upper limits of the axial induction factor; For each wind turbine, a power optimization problem for the downstream wind turbine cluster affected by its wake is constructed under different axial induction factors. The axial induction factor of the downstream wind turbine cluster is solved using an optimization algorithm. S303: For each wind turbine, a simulated annealing particle swarm optimization algorithm is used to solve the axial induction factor of the downstream wind turbine cluster.

2. The method for distributed dynamic control of maximum power output of a wind farm according to claim 1, characterized in that: The method of constructing a maximum total power optimization problem for a wind farm under conditions of different inflow wind speeds and wind directions, and using an optimization algorithm to solve the global optimal axial induction factor of each wind turbine under different wind conditions includes: S101, calculate the number of wind turbines in the wind farm according to the following formula i Captured mechanical power : ， in, ρ is the air density, S=πR ² is the swept area of ​​the fan impeller, R is the radius of the fan impeller, a i 、 v i Fan i The axial induction factor and inflow velocity, For fans i The power coefficient, ; S102, use the Jensen wake model to calculate the wake wind speed: ， ， in, For fans i The inflow wind speed, is the inflow wind speed at the most upstream of the wind farm, Indicates the fan j The collection of wind turbines that produce wake effects, For fans i For fans j The tail wind speed, is the wake expansion coefficient of the Jensen wake model, For wind turbines along the wind direction i and j The geographical distance between For fans i The wake induced by the wind turbine j Coverage area on the impeller plane; S103, the fan j Inflow wind speed It is written as the function of the inflow wind speed at the most upstream of the wind farm and the axial induction factor of the upstream wind turbine that affects its wake as shown in the following formula: : ， The fan j The captured mechanical power Written as a function of the axial induction factor shown below : in, For fans j Axial induction factor; The total power of the entire wind farm is the sum of the powers of all wind turbines Written as a function of the upstream inflow velocity and the axial induction factors of all fans : ， in, represents the set of all fan axial induction factors, n is the number of wind turbines in the wind farm; S104, construct the objective function of the wind farm total power maximization optimization problem: ， Among them, max represents the function The maximum value of S105, using a simulated annealing particle swarm optimization algorithm to solve the global optimal axial induction factor of each wind turbine under different wind conditions.

3. The method for distributed dynamic control of maximum power output of a wind farm according to claim 2, characterized in that: The steps of solving the problem by using the simulated annealing particle swarm optimization algorithm include: S201, initializing the parameters of the particle swarm and simulated annealing, wherein the particle swarm is composed of the results of solving the problem; S202, update the speed and position according to the iterative process of the particle swarm optimization algorithm: ， in, and Respectively represent k+ Particles at 1 iteration i speed and position, and Respectively represent k The particle i speed and position, w represents the inertia weight, c 1 and c 2 is a non-negative acceleration coefficient, r 1 and r 2 is a random number in the interval (0,1) used to enhance the randomness of the search process. and Respectively k After iterations, particles i The individual optimal position of and the global optimal position of the particle swarm; S203, calculating the wind farm power output using the objective function of the problem to obtain the particle fitness of the current iteration; S204, update the individual optimal position and value according to the fitness difference and annealing criterion: in, p k+1 For the system from k The status at the iteration x k Transfer to k Status at +1 iteration x k+1 The acceptance probability, E k+1 and E k The system is k +1 iteration and k The energy level of the iteration, T k Represents the temperature of the simulated annealing mechanism; if the current particle is better than the global optimal solution, the global optimal position and value are updated; S205, reduce the temperature according to the annealing formula: ， in, and Respectively k+ 1 and k The temperature at the iteration, K is the simulated annealing coefficient, and 0< K <1; S206 , determining whether the number of iterations reaches a preset threshold; if so, outputting the global optimal axial induction factor of each wind turbine under different wind conditions; otherwise, skipping to step S202 to continue iteration.

4. The method for distributed dynamic control of maximum power output of a wind farm according to claim 1, characterized in that: The generating of the training data set of the wind turbine includes: searching for the wind turbine M different axial induction factors that belong to the set D The wind turbine with the largest total power in the cluster is the one with the largest total power. The axial induction factor, wind speed and wind direction in the current scenario are taken as input, and the maximum total power and the axial induction factor of the wind turbine are calculated to be the preset minimum value. D The difference in the total power of the wind turbine cluster is taken as the output and combined into the training data set of the wind turbine.

5. The method for distributed dynamic control of maximum power output of a wind farm according to claim 1, characterized in that: The step of constructing a data-driven model for estimating power loss of a downstream wind turbine cluster for each wind turbine and training the data-driven model for estimating power loss of a downstream wind turbine cluster of the wind turbine using a training data set of the wind turbine includes: S401, construct the neuron shown in the following formula: ， in, is the input of the neuron, is the output of the neuron, represents the dimensional space, n x and n y They are x in and z Dimensions, is the weight matrix, is the bias, g represents the activation function; S402, construct a multi-layer perceptron model based on neurons as shown in the following formula: ， ， ， in, Represents the multilayer perceptron model k Tier j The output of a neuron, is the activation function, the activation function Select the hyperbolic tangent function tanh, Represents the multilayer perceptron model k Tier j The weight of a neuron, Represents the multilayer perceptron model k -1 layer output vector, Represents the multilayer perceptron model k Tier j The bias of a neuron, ~ Represents the multilayer perceptron model k Layer 1~ N k The weights of neurons, N k Represents the multilayer perceptron model k The number of neurons in the layer, ~ Represents the multilayer perceptron model k Layer 1~ N k The output of a neuron; S403: For each wind turbine, construct a wind turbine model based on the multi-layer perceptron model. i Data-driven model for estimating power losses in downstream wind turbine clusters: ， ， ， in, Represents the model output of the multilayer perceptron model, the model output is the fan i Estimated power loss of downstream wind turbine cluster, x i =[ d i , v i , a i ] T represents the model input, d i 、 v i 、 a i Respectively represent fans i The inflow wind direction, inflow wind speed and axial induction factor, b 1~ b 3 represents the bias of the 1st to 3rd layers, w 1~ w 3 represents the weight of the 1st to 3rd layers, w k Indicates the k The weight matrix of the layer, b k Indicates the k The bias vector of the layer; ~ Represents the multilayer perceptron model k Layer 1~ N k The weights of neurons, ~ Represents the multilayer perceptron model k Layer 1~ N k The bias of each neuron; S404 : For each wind turbine, use the training data set of the wind turbine to train a data-driven model for estimating power loss of a cluster of wind turbines downstream of the wind turbine.

6. The method for distributed dynamic control of maximum power output of a wind farm according to claim 1, characterized in that: The method of constructing a local decentralized optimal control problem for each wind turbine by combining a data-driven model for estimating power losses of a cluster of downstream wind turbines affected by the wind turbine wake, and a mechanical power model and data-driven model for each wind turbine, and solving the optimal axial induction factor of the wind turbine for controlling the wind turbine includes: S501, construct the fan shown in the following formula i Capturing mechanical power P i Axial induction factor a i The relationship: ， in, ρ is the air density, S=πR ² is the swept area of ​​the fan impeller, R is the radius of the fan impeller, a i 、 v i Fan i Axial induction factor and inflow velocity; S502, the fan i Capturing mechanical power P i Axial induction factor a i The relationship between the two is combined with the power loss estimation data-driven model of the downstream wind turbine cluster to construct the objective function of the local decentralized optimal control problem. : ， in, x i =[ d i , v i , a i ] T represents the model input, d i 、 v i 、 a i Respectively represent fans i The inflow wind direction, inflow wind speed and axial induction factor, b 1~ b 3 represents the bias of the 1st to 3rd layers, w 1~ w 3 represents the weight of the 1st to 3rd layers, For fans i Rated power, For fans i The inflow wind speed is v i And the axial induction factor when working at rated power is the numerical solution of the following formula: ； S503, Objective Function of Local Decentralized Optimal Control Problem Axial induction factor a i The objective function of the local decentralized optimal control problem is calculated by the single variable piecewise function The extreme points in each segment, as well as the extreme points at 0, a u The function values ​​at the three endpoints of , 1 / 3 are taken so that the objective function obj The maximum value corresponds to the axial induction factor a i As a fan i The optimal axial induction factor is used to control the fan i .

7. A wind farm maximum power output distributed dynamic control system, comprising a microprocessor and a memory connected to each other, characterized in that: The microprocessor is programmed or configured to execute the distributed dynamic control method for maximum power output of a wind farm as recited in any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program or instruction stored therein, characterized in that: The computer program or instruction is programmed or configured to execute the method for distributed dynamic control of maximum power output of a wind farm according to any one of claims 1 to 6 through a processor.

9. A computer program product comprising a computer program or instructions, characterized in that The computer program or instruction is programmed or configured to execute the method for distributed dynamic control of maximum power output of a wind farm according to any one of claims 1 to 6 through a processor.

Images