A method and device for constructing a low voltage ride through model of a wind turbine group

Abstract

The application discloses a kind of wind turbine group low voltage ride through model construction method and device, comprising: obtaining the operating variable dataset of wind turbine group;According to the operating variable dataset, the initial wind turbine group low voltage ride through model constructed is trained, to obtain wind turbine group low voltage ride through model.The application does not need to construct wind turbine group equivalent model and other mechanism model through the physical principle of wind turbine group, avoids the state of each wind turbine in wind turbine group is different to need to model each wind turbine respectively, and directly realizes the construction of wind turbine group low voltage ride through model through the operating data of wind turbine group, and then does not need to determine the parameter information of each wind turbine in wind turbine group, so that the application has high accuracy and high reliability.

Description

Technical Field

[0001] This invention relates to the field of wind turbine cluster characteristic analysis technology, and in particular to a method and apparatus for constructing a low-voltage ride-through model for wind turbine clusters. Background Technology

[0002] With the development of technology, new power systems dominated by new energy sources are increasing rapidly. Wind energy is a safe and clean renewable energy source, and the total installed capacity of wind power is increasing year by year. However, as renewable energy sources such as wind power are continuously connected to the grid, the grid architecture is becoming increasingly complex, bringing new technical challenges to the safe and stable operation of the power system.

[0003] Modeling and simulation are fundamental methods for studying the operating characteristics and stability of new power systems, and accurate and reliable simulation relies on precise dynamic component modeling. Traditional dynamic component modeling methods typically construct mechanistic models, such as equivalent models of wind turbine clusters, based on the physical principles of wind turbine clusters. Since the states of each wind turbine in a cluster differ, each turbine needs to be modeled separately to simulate the power system's operating state when the wind turbine cluster is connected to the grid. This is costly, and the parameter information of each wind turbine in the cluster is difficult to determine, resulting in low modeling accuracy and reliability.

[0004] Therefore, there is an urgent need for a method to construct low-voltage ride-through models for wind turbine clusters that can improve the accuracy and reliability of modeling. Summary of the Invention

[0005] This invention provides a method for constructing a low-voltage ride-through model for wind turbine clusters, in order to solve the technical problems of low accuracy and reliability in existing modeling technologies.

[0006] To address the aforementioned technical problems, embodiments of the present invention provide a method for constructing a low-voltage ride-through model for wind turbine clusters, comprising:

[0007] Obtain the dataset of operating variables for the wind turbine cluster;

[0008] Based on the aforementioned dataset of operating variables, the initial low-voltage ride-through model of the wind turbine cluster is trained to obtain the low-voltage ride-through model of the wind turbine cluster.

[0009] Understandably, compared to existing technologies, this invention obtains a low-voltage ride-through model of the wind turbine cluster by acquiring the operational variable dataset of the wind turbine cluster and training the constructed initial low-voltage ride-through model of the wind turbine cluster. This eliminates the need to construct equivalent models or other mechanistic models of the wind turbine cluster based on its physical principles, thus avoiding the need to model each wind turbine separately due to the different states of each turbine in the cluster. Instead, this invention directly constructs the low-voltage ride-through model of the wind turbine cluster using the operational data of the wind turbine cluster, thereby eliminating the need to determine the parameter information of each wind turbine in the cluster. This results in high accuracy and high reliability for this invention.

[0010] As a preferred embodiment, the initial low-voltage ride-through model of the wind turbine cluster is trained based on the operational variable dataset to obtain the low-voltage ride-through model of the wind turbine cluster, specifically as follows:

[0011] The input variables and environmental variables in the running variable dataset are used as the input of the initial wind turbine cluster low voltage ride-through model, and the output variables in the running variable dataset are used as the output of the initial wind turbine cluster low voltage ride-through model, thereby training the initial wind turbine cluster low voltage ride-through model and obtaining the wind turbine cluster low voltage ride-through model.

[0012] The operational variable dataset includes the input variables of the wind turbine cluster, the output variables of the wind turbine cluster, and environmental variables.

[0013] It is understandable that by using the input variables and environmental variables in the running variable dataset as the input to the initial wind turbine cluster low-voltage ride-through model, and using the output variables in the running variable dataset as the output of the initial wind turbine cluster low-voltage ride-through model, it can be ensured that the constructed and trained wind turbine cluster low-voltage ride-through model can represent the relationship between the input variables, environmental variables, and output variables, thereby obtaining an accurate and reliable wind turbine cluster low-voltage ride-through model.

[0014] As a preferred embodiment, training the initial low-voltage ride-through model of the wind turbine cluster to obtain the low-voltage ride-through model of the wind turbine cluster specifically involves:

[0015] Based on the parameter gradient of the initial wind turbine cluster low-voltage ride-through model, the network parameters of the initial wind turbine cluster low-voltage ride-through model are iteratively adjusted until a preset condition is met, and the wind turbine cluster low-voltage ride-through model is output; wherein, the parameter gradient is the gradient of the loss function of the initial wind turbine cluster low-voltage ride-through model with respect to the network parameters of the initial wind turbine cluster low-voltage ride-through model, which is obtained by solving the running variable dataset.

[0016] Understandably, by using the parameter gradient of the initial wind turbine cluster low-voltage ride-through model, the network parameters of the initial wind turbine cluster low-voltage ride-through model are iteratively adjusted, thereby outputting the wind turbine cluster low-voltage ride-through model. This ensures the accuracy and precision of the wind turbine cluster low-voltage ride-through model, making it highly reliable.

[0017] As a preferred embodiment, the step of obtaining the parameter gradient includes:

[0018] The first ordinary differential equation is solved based on the running variable dataset to obtain the hidden state of the initial wind turbine group low voltage ride-through model at each time step; wherein, the first ordinary differential equation is used to characterize the functional relationship between the input variable, the output variable, and the environmental variable and the derivative of the output variable, respectively;

[0019] The hidden states of the initial wind turbine cluster low voltage ride-through model at each time step are input into the loss function of the initial wind turbine cluster low voltage ride-through model to obtain the loss value of the loss function at each time step.

[0020] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine group low voltage ride-through model at each time step, the parameter gradient is obtained and determined.

[0021] Understandably, by running the variable dataset, the functional relationships between the derivatives of the input, output, and environmental variables and the output variable are solved, thereby obtaining the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step. The loss value is then obtained through the loss function, which can accurately determine the parameter gradient and ensure that the constructed wind turbine cluster low-voltage ride-through model has high precision and accuracy.

[0022] As a preferred embodiment, the step of obtaining and determining the parameter gradient based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine group low-voltage ride-through model at each time step specifically involves:

[0023] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the hidden state gradient of the loss function with respect to the initial wind turbine cluster low-voltage ride-through model at each time step is determined, and used as the state gradient at the corresponding time step.

[0024] The parameter gradient is determined based on the state gradient at each time step.

[0025] It is understandable that by using the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the hidden state gradient can be accurately and efficiently determined, and then used as the state gradient at the corresponding time step. Based on the state gradient at each time step, the parameter gradient can be accurately determined, thereby ensuring the high reliability and high accuracy of the wind turbine cluster low-voltage ride-through model.

[0026] As a preferred embodiment, the step of determining the gradient of the loss function with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, specifically involves:

[0027] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the second ordinary differential equation is solved by backpropagation to obtain the gradient of the loss function at each time step with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model.

[0028] The second ordinary differential equation is used to characterize the state gradient and the functional relationship between the hidden state and the derivative of the state gradient.

[0029] It is understandable that by using the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the functional relationship between the state gradient and the derivative of the hidden state and the state gradient can be solved by backpropagation, thereby obtaining the gradient of the loss function at each time step with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model. At the same time, the second ordinary differential equation can make the solution of the gradient of the loss function at each time step with respect to the hidden state more efficient and accurate.

[0030] As a preferred embodiment, the input variable includes the grid connection voltage of the wind turbine group;

[0031] The output variables of the wind turbine group include the grid connection point output current of the wind turbine group;

[0032] The environmental variables include the wind speed of all wind turbines in the wind turbine group.

[0033] Accordingly, the present invention also provides a device for constructing a low-voltage ride-through model for a wind turbine cluster, comprising: an acquisition module and a training module;

[0034] The acquisition module is used to acquire the operating variable dataset of the wind turbine group;

[0035] The training module is used to train the constructed initial wind turbine cluster low-voltage ride-through model based on the running variable dataset, thereby obtaining the wind turbine cluster low-voltage ride-through model.

[0036] As a preferred embodiment, the initial low-voltage ride-through model of the wind turbine cluster is trained based on the operational variable dataset to obtain the low-voltage ride-through model of the wind turbine cluster, specifically as follows:

[0037] The input variables and environmental variables in the running variable dataset are used as the input of the initial wind turbine cluster low voltage ride-through model, and the output variables in the running variable dataset are used as the output of the initial wind turbine cluster low voltage ride-through model, thereby training the initial wind turbine cluster low voltage ride-through model and obtaining the wind turbine cluster low voltage ride-through model.

[0038] The operational variable dataset includes the input variables of the wind turbine cluster, the output variables of the wind turbine cluster, and environmental variables.

[0039] As a preferred embodiment, training the initial low-voltage ride-through model of the wind turbine cluster to obtain the low-voltage ride-through model of the wind turbine cluster specifically involves:

[0040] Based on the parameter gradient of the initial wind turbine cluster low-voltage ride-through model, the network parameters of the initial wind turbine cluster low-voltage ride-through model are iteratively adjusted until a preset condition is met, and the wind turbine cluster low-voltage ride-through model is output; wherein, the parameter gradient is the gradient of the loss function of the initial wind turbine cluster low-voltage ride-through model with respect to the network parameters of the initial wind turbine cluster low-voltage ride-through model, which is obtained by solving the running variable dataset.

[0041] As a preferred embodiment, the step of obtaining the parameter gradient includes:

[0042] The first ordinary differential equation is solved based on the running variable dataset to obtain the hidden state of the initial wind turbine group low voltage ride-through model at each time step; wherein, the first ordinary differential equation is used to characterize the functional relationship between the input variable, the output variable, and the environmental variable and the derivative of the output variable, respectively;

[0043] The hidden states of the initial wind turbine cluster low voltage ride-through model at each time step are input into the loss function of the initial wind turbine cluster low voltage ride-through model to obtain the loss value of the loss function at each time step.

[0044] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine group low voltage ride-through model at each time step, the parameter gradient is obtained and determined.

[0045] As a preferred embodiment, the step of obtaining and determining the parameter gradient based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine group low-voltage ride-through model at each time step specifically involves:

[0046] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the hidden state gradient of the loss function with respect to the initial wind turbine cluster low-voltage ride-through model at each time step is determined, and used as the state gradient at the corresponding time step.

[0047] The parameter gradient is determined based on the state gradient at each time step.

[0048] As a preferred embodiment, the step of determining the gradient of the loss function with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, specifically involves:

[0049] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the second ordinary differential equation is solved by backpropagation to obtain the gradient of the loss function at each time step with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model.

[0050] The second ordinary differential equation is used to characterize the state gradient and the functional relationship between the hidden state and the derivative of the state gradient.

[0051] As a preferred embodiment, the input variable includes the grid connection voltage of the wind turbine group;

[0052] The output variables of the wind turbine group include the grid connection point output current of the wind turbine group;

[0053] The environmental variables include the wind speed of all wind turbines in the wind turbine group.

[0054] Accordingly, the present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the wind turbine cluster low voltage ride-through model construction method as described in any of the preceding claims.

[0055] Accordingly, the present invention also provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program; wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the wind turbine group low voltage ride-through model construction method as described in any of the preceding claims. Attached Figure Description

[0056] Figure 1 This is a flowchart illustrating the method for constructing a low-voltage ride-through model for wind turbine clusters provided by this invention.

[0057] Figure 2This is a schematic diagram of the structure of the initial low-voltage ride-through model of the wind turbine cluster provided by the present invention;

[0058] Figure 3 This is a schematic diagram showing the comparison between the measured curve and the predicted curve of the d-axis component of the grid-connected output current of a wind turbine group under operating condition 1, provided by the present invention.

[0059] Figure 4 This is a schematic diagram showing the comparison between the measured curve and the predicted curve of the output current q-axis component at the grid connection point of a wind turbine group under operating condition 1, provided by the present invention.

[0060] Figure 5 This is a schematic diagram showing the comparison between the measured curve and the predicted curve of the d-axis component of the grid-connected output current of the wind turbine group under the second working condition provided by the present invention.

[0061] Figure 6 This is a schematic diagram showing the comparison between the measured curve and the predicted curve of the output current q-axis component at the grid connection point of the wind turbine group under operating condition 2 provided by the present invention.

[0062] Figure 7 This is a schematic diagram showing the comparison between the measured curve and the predicted curve of the d-axis component of the grid-connected output current of the three wind turbine groups under the three working conditions provided by this invention.

[0063] Figure 8 This is a schematic diagram showing the comparison between the measured curve and the predicted curve of the output current q-axis component at the grid connection point of the three wind turbine groups under the three working conditions provided by this invention.

[0064] Figure 9 This is a schematic diagram showing the comparison between the measured curve and the predicted curve of the output current d-axis component at the grid connection point of the wind turbine group under working condition four provided by the present invention.

[0065] Figure 10 This is a schematic diagram showing the comparison between the measured curve and the predicted curve of the output current q-axis component at the grid connection point of the wind turbine group under the four working conditions provided by this invention.

[0066] Figure 11 This is a schematic diagram of the structure of the wind turbine group low voltage ride-through model construction device provided by the present invention. Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Example 1

[0069] Please refer to Figure 1 The present invention provides a method for constructing a low-voltage ride-through model for a wind turbine cluster, comprising the following steps S101-S102:

[0070] S101: Obtain the dataset of operating variables for the wind turbine cluster.

[0071] The operational variable dataset includes the input variables of the wind turbine cluster, the output variables of the wind turbine cluster, and environmental variables.

[0072] As a preferred embodiment, the input variables include the grid connection voltage of the wind turbine group; the output variables of the wind turbine group include the grid connection point output current of the wind turbine group; and the environmental variables include the wind speed of all wind turbines in the wind turbine group.

[0073] For example, in this embodiment, the method for constructing a low-voltage ride-through model for a wind turbine group according to the present invention is specifically described using a doubly fed wind turbine group as an example.

[0074] It should be noted that the training dataset is used for model training. In this embodiment, the running variable dataset is the training dataset, including parameter data corresponding to input variables, output variables, and environmental variables. The input and output variables can be determined based on the influencing factors of low-voltage ride-through. For example, input variables may include the components of the grid-connected voltage of the wind turbine group along the d-axis and q-axis in the d-q rotating coordinate system, and output variables may include the components of the grid-connected output current of the wind turbine group along the d-axis and q-axis in the d-q rotating coordinate system. Environmental variables are used to characterize the operating environment of each wind turbine in the wind turbine group. For example, environmental variables that have a significant impact on the state of the input and output variables, such as wind speed, can be selected. In this embodiment, environmental variables include the wind speeds of all wind turbines in the wind turbine group, which can be arranged in a preset order to represent a vector. The parameter data corresponding to the input, output, and environmental variables can be the curves of the input, output, and environmental variables over time, respectively.

[0075] The training dataset can be obtained through simulation or by actual measurement of the power system, depending on the specific needs. For example, based on the CloudPSS cloud simulation platform, an equivalent model of a doubly-fed induction generator (DFIG) wind turbine group can be connected to the power grid, and its low-voltage ride-through characteristics can be simulated to obtain batch simulation data of the grid connection point voltage and grid connection point output current of the DFIG wind turbine group under different wind speed distribution scenarios, which can then be used as the training dataset.

[0076] S102: Based on the aforementioned running variable dataset, train the constructed initial wind turbine cluster low-voltage ride-through model to obtain the wind turbine cluster low-voltage ride-through model.

[0077] It should be noted that the constructed initial wind turbine cluster low-voltage ride-through model is a differential neural network model. The specific structure of the differential neural network model can be set according to actual needs. For example, it can include an input layer, an output layer, and at least one hidden layer. As an optional implementation, the differential neural network model can include an input layer 201, a first hidden layer 203, a second hidden layer 205, and an output layer 207 connected in sequence. The input layer 201 and the first hidden layer 203, the first hidden layer 203 and the second hidden layer 205, and the second hidden layer 205 and the output layer 207 are fully connected through a first activation function 202, a second activation function 204, and a third activation function 206, respectively, as shown in the figure. The number of neurons in the input layer 201, the first hidden layer 203, the second hidden layer 205, and the output layer 207 can be set to 64. The specific type of activation function can also be set according to actual needs. For example, each activation function can use an ELU (Exponential Linear Unit) activation function.

[0078] By training the differential neural network model using a training dataset (running variable dataset), a low-voltage ride-through (LVRT) model for a wind turbine cluster can be obtained. This model characterizes the LVRT dynamics of a wind turbine cluster. The inputs to the LVRT model can be input variables and environmental variables, while the output can be an output variable. Furthermore, the LVRT model can be considered a black box, effectively representing the strongly nonlinear switching process between input and output variables. Therefore, during the LVRT simulation, by inputting the grid connection point voltage and wind speed, the output current at the grid connection point can be obtained through the computation of the differential neural network within the LVRT model.

[0079] Traditional dynamic component modeling methods typically construct mechanistic models, such as equivalent models of wind turbine clusters, based on the physical principles of the wind turbine cluster. Since the states of each turbine in a wind turbine cluster differ, each turbine needs to be modeled separately to simulate the power system's operating state when the wind turbine cluster is connected to the grid, which is costly. Furthermore, the parameter information of each turbine in the wind turbine cluster (e.g., physical characteristics, internal structure, operating principles, etc.) is difficult to determine. For example, the physical characteristics of dynamic components in the power system are still under research, and there is insufficient prior knowledge and complete theoretical support to construct a suitable model. Alternatively, due to considerations such as technical confidentiality, power equipment manufacturers will not share information such as the internal structure, operating theory, and simulation models of the equipment with users, making it difficult to effectively construct low-voltage ride-through models for wind turbine clusters.

[0080] This invention provides an embodiment of the invention that obtains a dataset of operating variables for model training. This dataset includes parameter data corresponding to input variables, output variables, and environmental variables. The environmental variables characterize the working environment of each wind turbine in the wind turbine cluster. Based on the training dataset, a differential neural network model is trained to obtain a wind turbine cluster low-voltage ride-through model that characterizes the low-voltage ride-through dynamics of the wind turbine cluster. During the modeling process, only the low-voltage ride-through model needs to be established for the wind turbine cluster, without the need to model each wind turbine separately. Furthermore, the modeling process only requires the training dataset and does not require knowledge of the parameter information of each wind turbine in the wind turbine cluster. This effectively constructs a wind turbine cluster low-voltage ride-through model and greatly reduces the complexity and modeling difficulty of the low-voltage ride-through model.

[0081] Meanwhile, traditional dynamic component modeling methods cannot accurately match measured data with theoretical derivation models, and lack methods to integrate theoretical analysis models and data-driven models.

[0082] In this embodiment of the invention, the low-voltage ride-through model of the wind turbine group obtained by training a differential neural network model (initial low-voltage ride-through model of the wind turbine group) based on the running variable dataset can be regarded as a black box, which can well characterize the strong nonlinear switching process of input and output variables. Thus, in the low-voltage ride-through simulation process of the wind turbine group, the grid connection point voltage and wind speed conditions are input, and the output current of the grid connection point can be obtained through the operation of the differential neural network inside the low-voltage ride-through model of the wind turbine group. This solves the problem that the model constructed by the traditional method cannot match the measured data well, and greatly reduces the complexity and modeling difficulty of the low-voltage ride-through model.

[0083] In a preferred embodiment, the step of training the initial wind turbine cluster low-voltage ride-through model based on the operational variable dataset to obtain the wind turbine cluster low-voltage ride-through model specifically involves:

[0084] The input variables and environmental variables in the running variable dataset are used as the input to the initial wind turbine cluster low-voltage ride-through model, and the output variables in the running variable dataset are used as the output of the initial wind turbine cluster low-voltage ride-through model, thereby training the initial wind turbine cluster low-voltage ride-through model and obtaining the wind turbine cluster low-voltage ride-through model; wherein, the running variable dataset includes the input variables, output variables, and environmental variables of the wind turbine cluster.

[0085] It should be noted that during the training of the initial low-voltage ride-through model of the wind turbine cluster based on the operating variable dataset, the parameter data corresponding to the input variables and the parameter data corresponding to the environmental variables can be used as the input to the initial low-voltage ride-through model (differential neural network model), and the parameter data corresponding to the output variables can be used as the output to train the initial low-voltage ride-through model of the wind turbine cluster, thus obtaining the wind turbine cluster low-voltage ride-through model. The wind turbine cluster low-voltage ride-through model obtained through training can well characterize the strongly nonlinear switching process of the input and output variables under this environmental condition. By inputting the grid connection point output current of the wind turbine cluster low-voltage ride-through model into the power grid, the simulation results of the power system's operating state when the wind turbine cluster is connected to the grid can be obtained, reducing the amount of simulation calculation.

[0086] It is understandable that by using the input variables and environmental variables in the running variable dataset as the input to the initial wind turbine cluster low-voltage ride-through model, and using the output variables in the running variable dataset as the output of the initial wind turbine cluster low-voltage ride-through model, it can be ensured that the constructed and trained wind turbine cluster low-voltage ride-through model can represent the relationship between the input variables, environmental variables, and output variables, thereby obtaining an accurate and reliable wind turbine cluster low-voltage ride-through model.

[0087] As a preferred embodiment, the step of training the initial wind turbine cluster low-voltage ride-through model to obtain the wind turbine cluster low-voltage ride-through model specifically involves:

[0088] Based on the parameter gradient of the initial wind turbine cluster low-voltage ride-through model, the network parameters of the initial wind turbine cluster low-voltage ride-through model are iteratively adjusted until a preset condition is met, and the wind turbine cluster low-voltage ride-through model is output; wherein, the parameter gradient is the gradient of the loss function of the initial wind turbine cluster low-voltage ride-through model with respect to the network parameters of the initial wind turbine cluster low-voltage ride-through model, which is obtained by solving the running variable dataset.

[0089] It should be noted that the parameter gradient of the initial wind turbine cluster low-voltage ride-through model, i.e., the gradient of the loss function of the initial wind turbine cluster low-voltage ride-through model with respect to the network parameters of the initial wind turbine cluster low-voltage ride-through model, can be set according to the specific structure of the differential neural network model. For example, it can include the weights and biases of the neuron connections.

[0090] The specific method for iteratively adjusting the network parameters of the initial low-voltage ride-through model of the wind turbine cluster can be set according to actual needs. For example, in each round of iterative training, the parameter gradient of the initial low-voltage ride-through model of the wind turbine cluster can be calculated, and the network parameters of the initial low-voltage ride-through model of the wind turbine cluster can be adjusted based on the gradient descent method. The next round of iterative training can then be carried out based on the adjusted network parameters until the preset iteration termination condition is met.

[0091] The preset iteration termination condition can be set according to actual needs. For example, it can be a preset number of iterations, such as terminating the iteration after 100 iterations; or it can be a preset value for the parameter gradient, such as terminating the iteration after the parameter gradient reaches 0.

[0092] Understandably, by using the parameter gradient of the initial wind turbine cluster low-voltage ride-through model, the network parameters of the initial wind turbine cluster low-voltage ride-through model are iteratively adjusted, thereby outputting the wind turbine cluster low-voltage ride-through model. This ensures the accuracy and precision of the wind turbine cluster low-voltage ride-through model, making it highly reliable.

[0093] This invention iteratively adjusts the network parameters of the initial low-voltage ride-through model of the wind turbine cluster based on the parameter gradient of the initial low-voltage ride-through model of the wind turbine cluster until a preset termination condition is met, so as to obtain the low-voltage ride-through model of the wind turbine cluster. This can effectively improve the learning effect of the low-voltage ride-through model of the wind turbine cluster, thereby ensuring the modeling accuracy and modeling efficiency of the low-voltage ride-through model of the wind turbine cluster.

[0094] As a preferred embodiment, the step of obtaining the parameter gradient includes:

[0095] The first ordinary differential equation is solved based on the dataset of operating variables to obtain the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step. The first ordinary differential equation characterizes the functional relationships between the derivatives of the input variables, the output variables, and the environmental variables and the output variables. The hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step are input into the loss function of the initial wind turbine cluster low-voltage ride-through model to obtain the loss value of the loss function at each time step. Based on the loss value of the loss function at each time step and the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step, the parameter gradient is obtained and determined.

[0096] It should be noted that the first ordinary differential equation is used to characterize the functional relationship between the input variable, the output variable, and the derivatives of the environmental variable and the output variable. The number of first ordinary differential equations is determined based on the number of output variables. For example, when the output variables include the components of the grid-connected output current of the wind turbine group in the d-q rotating coordinate system along the d-axis and q-axis, it can include two first ordinary differential equations as shown in equations (1) and (2):

[0097]

[0098]

[0099] In the formula, I d and I q V represents the components of the grid-connected output current of the wind turbine group along the d-axis and q-axis in the d-q rotating coordinate system; d and V q These represent the components of the grid connection point voltage of the wind turbine group along the d-axis and q-axis in the d-q rotating coordinate system; v w Wind speed conditions of each wind turbine in the wind turbine group; and I d and I q The derivative at time t; f() and g() are respectively the derivatives of I. d and I q The derivative function.

[0100] The running variable dataset can be input into the initial low-voltage ride-through model of the wind turbine cluster. The first ordinary differential equation can then be solved using a black-box differential equation solver to obtain the hidden states of the initial low-voltage ride-through model at each time step. The hidden states of the initial low-voltage ride-through model represent the predicted output of the model at that specific time step.

[0101] After obtaining the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step, the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step can be further input into the loss function of the initial wind turbine cluster low-voltage ride-through model to obtain the loss value of the loss function at each time step. Taking the hidden states obtained by solving equation (1) as an example, the solution of the loss value is specifically shown in equation (3):

[0102]

[0103] In the formula, L() is the loss function, h(t0) and h(t1) are the hidden states of the initial wind turbine cluster low-voltage ride-through model at time t0 and time t1, respectively, where t0 is the initial time; ODESolve() is the differential equation solver; θ is the network parameter of the initial wind turbine cluster low-voltage ride-through model; f(h(t),t,θ) represents I when the network parameter is θ. d The value of the derivative function at time t.

[0104] The loss value of the loss function at each time step can be obtained through equation (3). Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine group low voltage ride-through model at each time step, the gradient of the loss function with respect to the network parameters of the initial wind turbine group low voltage ride-through model can be determined, i.e., the parameter gradient.

[0105] This invention solves the first ordinary differential equation representing the functional relationship between the input variable, output variable, and the derivatives of the environmental variable and the output variable based on the running variable dataset. This yields the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step. The hidden states at each time step are then input into the loss function of the initial wind turbine cluster low-voltage ride-through model to obtain the loss value of the loss function at each time step. The parameter gradient is determined based on the loss value of the loss function at each time step and the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step. The parameter gradient determination process is simple and efficient, thereby ensuring the training efficiency of the initial wind turbine cluster low-voltage ride-through model.

[0106] Understandably, by running the variable dataset, the functional relationships between the derivatives of the input, output, and environmental variables and the output variable are solved, thereby obtaining the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step. The loss value is then obtained through the loss function, which can accurately determine the parameter gradient and ensure that the constructed initial wind turbine cluster low-voltage ride-through model has high precision and accuracy.

[0107] As a preferred embodiment, the step of obtaining and determining the parameter gradient based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine group low-voltage ride-through model at each time step specifically involves:

[0108] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the hidden state gradient of the loss function with respect to the initial wind turbine cluster low-voltage ride-through model at each time step is determined, and these gradients are used as the state gradients at the corresponding time steps. Based on the state gradients at each time step, the parameter gradients are determined.

[0109] It should be noted that in determining the parameter gradient, the gradient of the loss function with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step is first determined based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step. This gradient is then used as the state gradient at the corresponding time step. For example, the state gradient a(t) at time t can be expressed as:

[0110] After obtaining the state gradient at each time step, the parameter gradient is further determined based on the state gradient at each time step. The parameter gradient is determined based on the state gradient at each time step as shown in equation (4):

[0111]

[0112] In the formula, For the gradient of the parameters; t end The termination time, Indicates t end Gradient of parameters a at time t θ (t end The value of ) is 0; f(h(t),t) represents I d The value of the derivative function at time t.

[0113] This invention determines the gradient of the loss function with respect to the hidden state of the initial wind turbine group low-voltage ride-through model at each time step based on the loss value of the loss function at each time step and the hidden state of the differential neural network model at each time step. The gradient of the loss function with respect to the hidden state of the initial wind turbine group low-voltage ride-through model at each time step is used as the state gradient at the corresponding time step. The parameter gradient is determined based on the state gradient at each time step, which can effectively improve the effectiveness of the parameter gradient determination result, thereby improving the learning effect of the initial wind turbine group low-voltage ride-through model.

[0114] It is understandable that by using the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the hidden state gradient can be accurately and efficiently determined, and then used as the state gradient at the corresponding time step. Based on the state gradient at each time step, the parameter gradient can be accurately determined, thereby ensuring the high reliability and high accuracy of the wind turbine cluster low-voltage ride-through model.

[0115] In a preferred embodiment, the step of determining the gradient of the loss function with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, specifically involves:

[0116] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the second ordinary differential equation is solved by backpropagation to obtain the gradient of the loss function at each time step with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model; wherein, the second ordinary differential equation is used to characterize the functional relationship between the state gradient and the derivative of the hidden state with respect to the state gradient.

[0117] It should be noted that the second ordinary differential equation is used to characterize the functional relationship between the state gradient and the derivative of the hidden state with respect to the state gradient. That is, the second ordinary differential equation is used to characterize the dynamic process of the state gradient changing with time. Taking the state gradient obtained based on equation (1) as an example, the second ordinary differential equation can be shown as equation (5):

[0118]

[0119] In the formula, ε represents the step size, that is, the interval between adjacent time points; Let be the derivative of the state gradient at time t.

[0120] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine low-voltage ride-through model at each time step, the differential equation solver is used to backpropagate and solve equation (5). The gradient of the loss function at each time step with respect to the hidden state of the initial wind turbine low-voltage ride-through model can be obtained, that is, the state gradient at each time step is obtained, as shown in equation (6).

[0121]

[0122] In the formula, a(t) end ) for t end The state gradient at time step 1.

[0123] It is understandable that by using the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the functional relationship between the state gradient and the derivative of the hidden state and the state gradient can be solved by backpropagation, thereby obtaining the gradient of the loss function at each time step with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model. At the same time, the second ordinary differential equation can make the solution of the gradient of the loss function at each time step with respect to the hidden state more efficient and accurate.

[0124] To further verify the effectiveness of the low-voltage ride-through model construction method for wind turbine clusters according to the present invention, the low-voltage ride-through model of wind turbine clusters constructed by the method of the present invention was tested under four typical operating conditions in the test set. The wind turbine cluster in the test set included three wind turbines.

[0125] Operating Condition 1: When the grid voltage is 110kV nominal voltage and the wind turbine group is operating at a symmetrical wind speed of 5.95m / s (all three turbines are operating at the same wind speed), the comparison results of the measured curves and predicted curves of the d-axis and q-axis components of the output current at the grid connection point of the wind turbine group are as follows: Figure 3 and Figure 4 As shown.

[0126] Operating Condition 2: With a grid voltage of 110kV nominal voltage, and three wind turbines in the wind turbine group operating at asymmetrical wind speeds of 6.39m / s, 9.87m / s, and 10.08m / s respectively, the comparison results of the measured and predicted curves of the d-axis and q-axis components of the output current at the grid connection point of the wind turbine group are as follows: Figure 5 and Figure 6 As shown.

[0127] Operating Condition 3: When the grid voltage is 109.16kV and the wind turbine group is operating at a symmetrical wind speed of 8.71m / s (all three turbines are operating at the same wind speed), the comparison results of the measured curves and predicted curves of the d-axis and q-axis components of the output current at the grid connection point of the wind turbine group are as follows: Figure 7 and Figure 8 As shown.

[0128] Operating Condition 4: With a grid voltage of 115.56kV, and three wind turbines in the wind turbine group operating at asymmetrical wind speeds of 6.00m / s, 7.75m / s, and 9.81m / s respectively, the comparison results of the measured and predicted curves of the d-axis and q-axis components of the output current at the grid connection point of the wind turbine group are as follows: Figure 9 and Figure 10 As shown.

[0129] in, Figures 3 to 10 The measured curves in the test set are the calibration values ​​of the d-axis or q-axis components of the output current at the grid connection point. The predicted curves are the prediction results of the d-axis or q-axis components of the output current at the grid connection point after 400 iterations of the low voltage ride-through model of the wind turbine group constructed by the method of this invention.

[0130] Depend on Figures 3 to 10 It can be seen that after 400 iterations of training, the predicted curve of the low-voltage ride-through model of the wind turbine group constructed by the method of this invention basically matches the measured curve. The prediction errors for the four operating conditions are shown in Table 1.

[0131] Table 1

[0132] Test conditions Prediction error Operating Condition 1 0.000265704 Operating Condition 2 0.000185900 Operating Condition 3 0.000405468 Operating Condition 4 0.000229250

[0133] Meanwhile, the low-voltage ride-through model of the wind turbine group constructed by the method of the present invention was tested through the entire test set. The average error and variance obtained by the entire test set when the low-voltage ride-through model of the wind turbine group was iterated for 100, 200, 300 and 400 cycles are shown in Table 2.

[0134] Table 2

[0135] Cycle number average error variance 100 0.000265 0.000162 200 0.000234 0.000211 300 0.000139 0.000229 400 0.000115 0.000218

[0136] Depend on Figures 3 to 10 As shown in Tables 1 and 2, the prediction curves of the low-voltage ride-through model of wind turbine clusters constructed by the method of the present invention have high consistency and small prediction error, which indicates that the low-voltage ride-through model of wind turbine clusters constructed by the method of the present invention has high generalization ability and effectiveness, thus proving the effectiveness and high reliability of the method for constructing the low-voltage ride-through model of wind turbine clusters of the present invention.

[0137] It is understood that the wind turbine cluster low-voltage ride-through model construction method and apparatus provided by the present invention obtains a dataset of operating variables for model training. The dataset of operating variables includes parameter data corresponding to input variables, output variables, and environmental variables. The environmental variables are used to characterize the working environment of each wind turbine in the wind turbine cluster. Based on the dataset of operating variables, an initial wind turbine cluster low-voltage ride-through model is trained, thereby obtaining a wind turbine cluster low-voltage ride-through model that characterizes the low-voltage ride-through dynamics of the wind turbine cluster. In the modeling process, only the low-voltage ride-through model needs to be established for the wind turbine cluster, without the need to model each wind turbine separately. Moreover, only the dataset of operating variables is needed in the modeling process, without the need to know the parameter information of each wind turbine in the wind turbine cluster. This can effectively construct a wind turbine cluster low-voltage ride-through model and greatly reduce the complexity and modeling difficulty of the low-voltage ride-through model.

[0138] Implementing the above embodiments has the following effects:

[0139] Compared to existing technologies, this invention obtains a low-voltage ride-through model of the wind turbine cluster by acquiring the operational variable dataset of the wind turbine cluster and training the constructed initial low-voltage ride-through model of the wind turbine cluster. This eliminates the need to construct equivalent models or other mechanistic models of the wind turbine cluster based on its physical principles, thus avoiding the need to model each wind turbine separately due to the different states of each turbine in the cluster. Instead, this invention directly constructs the low-voltage ride-through model of the wind turbine cluster using the operational data of the wind turbine cluster, thereby eliminating the need to determine the parameter information of each wind turbine in the cluster. This results in high accuracy and high reliability for this invention.

[0140] Example 2

[0141] Please see Figure 11 The present invention provides a wind turbine cluster low voltage ride-through model construction device, comprising: an acquisition module 301 and a training module 302.

[0142] The acquisition module 301 is used to acquire the operating variable dataset of the wind turbine group.

[0143] The training module 302 is used to train the constructed initial wind turbine cluster low-voltage ride-through model based on the running variable dataset, thereby obtaining the wind turbine cluster low-voltage ride-through model.

[0144] In a preferred embodiment, the step of training the initial wind turbine cluster low-voltage ride-through model based on the operational variable dataset to obtain the wind turbine cluster low-voltage ride-through model specifically involves:

[0145] The input variables and environmental variables in the running variable dataset are used as the input to the initial wind turbine cluster low-voltage ride-through model, and the output variables in the running variable dataset are used as the output of the initial wind turbine cluster low-voltage ride-through model, thereby training the initial wind turbine cluster low-voltage ride-through model and obtaining the wind turbine cluster low-voltage ride-through model; wherein, the running variable dataset includes the input variables, output variables, and environmental variables of the wind turbine cluster.

[0146] As a preferred embodiment, the step of training the initial wind turbine cluster low-voltage ride-through model to obtain the wind turbine cluster low-voltage ride-through model specifically involves:

[0147] Based on the parameter gradient of the initial wind turbine cluster low-voltage ride-through model, the network parameters of the initial wind turbine cluster low-voltage ride-through model are iteratively adjusted until a preset condition is met, and the wind turbine cluster low-voltage ride-through model is output; wherein, the parameter gradient is the gradient of the loss function of the initial wind turbine cluster low-voltage ride-through model with respect to the network parameters of the initial wind turbine cluster low-voltage ride-through model, which is obtained by solving the running variable dataset.

[0148] As a preferred embodiment, the step of obtaining the parameter gradient includes:

[0149] The first ordinary differential equation is solved based on the dataset of operating variables to obtain the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step. The first ordinary differential equation characterizes the functional relationships between the derivatives of the input variables, the output variables, and the environmental variables and the output variables. The hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step are input into the loss function of the initial wind turbine cluster low-voltage ride-through model to obtain the loss value of the loss function at each time step. Based on the loss value of the loss function at each time step and the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step, the parameter gradient is obtained and determined.

[0150] As a preferred embodiment, the step of obtaining and determining the parameter gradient based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine group low-voltage ride-through model at each time step specifically involves:

[0151] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the hidden state gradient of the loss function with respect to the initial wind turbine cluster low-voltage ride-through model at each time step is determined, and these gradients are used as the state gradients at the corresponding time steps. Based on the state gradients at each time step, the parameter gradients are determined.

[0152] In a preferred embodiment, the step of determining the gradient of the loss function with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, specifically involves:

[0153] Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the second ordinary differential equation is solved by backpropagation to obtain the gradient of the loss function at each time step with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model; wherein, the second ordinary differential equation is used to characterize the functional relationship between the state gradient and the derivative of the hidden state with respect to the state gradient.

[0154] As a preferred embodiment, the input variables include the grid connection voltage of the wind turbine group; the output variables of the wind turbine group include the grid connection point output current of the wind turbine group; and the environmental variables include the wind speed of all wind turbines in the wind turbine group.

[0155] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0156] Implementing the embodiments of the present invention has the following effects:

[0157] Compared to existing technologies, this invention obtains a low-voltage ride-through model of the wind turbine cluster by acquiring the operational variable dataset of the wind turbine cluster and training the constructed initial low-voltage ride-through model of the wind turbine cluster. This eliminates the need to construct equivalent models or other mechanistic models of the wind turbine cluster based on its physical principles, thus avoiding the need to model each wind turbine separately due to the different states of each turbine in the cluster. Instead, this invention directly constructs the low-voltage ride-through model of the wind turbine cluster using the operational data of the wind turbine cluster, thereby eliminating the need to determine the parameter information of each wind turbine in the cluster. This results in high accuracy and high reliability for this invention.

[0158] Example 3

[0159] Accordingly, the present invention also provides a terminal device, comprising: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the wind turbine group low voltage ride-through model construction method as described in any of the above embodiments.

[0160] The terminal device in this embodiment includes a processor, a memory, and a computer program and computer instructions stored in the memory and executable on the processor. When the processor executes the computer program, it implements the various steps described in Embodiment 1 above, for example... Figure 1 The steps S101 to S102 are shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiment, such as the training module 302.

[0161] For example, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the terminal device. For example, training module 302 is used to train the constructed initial wind turbine cluster low-voltage ride-through model based on the running variable dataset, thereby obtaining the wind turbine cluster low-voltage ride-through model.

[0162] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the schematic diagram is merely an example of a terminal device and does not constitute a limitation on the terminal device. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the terminal device may also include input / output devices, network access devices, buses, etc.

[0163] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0164] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile terminal, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0165] Wherein, if the modules / units integrated in the terminal device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0166] Example 4

[0167] Accordingly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the wind turbine group low-voltage ride-through model construction method as described in any of the above embodiments.

[0168] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for constructing a low-voltage ride-through model for a wind turbine cluster, characterized in that, include: Obtain the operating variable dataset of the wind turbine cluster; wherein, the operating variable dataset includes the input variables, output variables, and environmental variables of the wind turbine cluster; The input variables and environmental variables in the running variable dataset are used as the inputs to the initial wind turbine cluster low-voltage ride-through model, and the output variables in the running variable dataset are used as the outputs of the initial wind turbine cluster low-voltage ride-through model. Based on the parameter gradients of the initial wind turbine cluster low-voltage ride-through model, the network parameters of the initial wind turbine cluster low-voltage ride-through model are iteratively adjusted until the preset conditions are met, and the wind turbine cluster low-voltage ride-through model is output. The parameter gradient is the gradient of the loss function of the initial wind turbine cluster low-voltage ride-through model with respect to the network parameters of the initial wind turbine cluster low-voltage ride-through model, which is obtained by solving the running variable dataset. The step of obtaining the parameter gradient includes: The first ordinary differential equation is solved based on the running variable dataset to obtain the hidden state of the initial wind turbine group low voltage ride-through model at each time step; wherein, the first ordinary differential equation is used to characterize the functional relationship between the input variable, the output variable, and the environmental variable and the derivative of the output variable, respectively; The hidden states of the initial wind turbine cluster low voltage ride-through model at each time step are input into the loss function of the initial wind turbine cluster low voltage ride-through model to obtain the loss value of the loss function at each time step. Based on the loss value of the loss function at each time step and the hidden state of the initial wind turbine cluster low-voltage ride-through model at each time step, the second ordinary differential equation is solved by backpropagation to obtain the gradient of the loss function at each time step with respect to the hidden state of the initial wind turbine cluster low-voltage ride-through model, and these gradients are used as the state gradients at the corresponding time steps. The second ordinary differential equation is used to characterize the state gradient and the functional relationship between the hidden state and the derivative of the state gradient. The parameter gradient is determined based on the state gradient at each time step.

2. The method for constructing a low-voltage ride-through model for a wind turbine cluster as described in claim 1, characterized in that, The input variables include the grid connection voltage of the wind turbine group; The output variables of the wind turbine group include the grid connection point output current of the wind turbine group; The environmental variables include the wind speed of all wind turbines in the wind turbine group.

3. A device for constructing a low-voltage ride-through model of a wind turbine cluster, characterized in that, include: Acquisition module and training module; The acquisition module is used to acquire a dataset of operating variables of the wind turbine cluster; wherein, the dataset of operating variables includes the input variables, output variables, and environmental variables of the wind turbine cluster; The training module is used to take the input variables and environmental variables in the running variable dataset as the input of the initial wind turbine cluster low-voltage ride-through model, and the output variables in the running variable dataset as the output of the initial wind turbine cluster low-voltage ride-through model. Based on the parameter gradient of the initial wind turbine cluster low-voltage ride-through model, the network parameters of the initial wind turbine cluster low-voltage ride-through model are iteratively adjusted until preset conditions are met, and the wind turbine cluster low-voltage ride-through model is output. The parameter gradient is the gradient of the loss function of the initial wind turbine cluster low-voltage ride-through model with respect to the network parameters of the initial wind turbine cluster low-voltage ride-through model, obtained by solving the running variable dataset. The step of obtaining the parameter gradient includes: solving the first ordinary differential equation based on the running variable dataset to obtain the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step. In this process, the first ordinary differential equation characterizes the functional relationship between the input variable, the output variable, and the environmental variable and the derivative of the output variable, respectively. The hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step are input into the loss function of the initial wind turbine cluster low-voltage ride-through model to obtain the loss value of the loss function at each time step. Based on the loss value of the loss function at each time step and the hidden states of the initial wind turbine cluster low-voltage ride-through model at each time step, the second ordinary differential equation is solved by backpropagation to obtain the gradient of the loss function at each time step with respect to the hidden states of the initial wind turbine cluster low-voltage ride-through model, and these gradients are used as the state gradients at the corresponding times. The second ordinary differential equation characterizes the state gradient and the functional relationship between the hidden states and the derivative of the state gradient. The parameter gradient is determined based on the state gradient at each time step.

4. A terminal device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the method for constructing a low-voltage ride-through model of a wind turbine cluster as described in any one of claims 1-2.

5. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program; wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the wind turbine cluster low-voltage ride-through model construction method as described in any one of claims 1-2.

Images