Power prediction method, device and equipment for wind driven generator cluster, storage medium and product

Abstract

The invention provides a power prediction method, device and equipment of a wind driven generator cluster, a storage medium and a product, and particularly relates to the technical field of power prediction of the wind driven generator cluster, and the method comprises the steps: a processor calculates correlation coefficients of each wind driven generator and other wind driven generators, and obtains an average value, determining the maximum calculation result as a target wind driven generator; the wind speeds of the multiple wind driven generators from the (t-x) th moment to the t moment and the power of the target wind driven generator in the same period are obtained; inputting the historical wind speed data into the first neural network model, predicting the wind speed of the plurality of wind driven generators at the (t + 1) th moment, and inputting the historical power data of the target wind driven generator into the second neural network model to obtain the power of the target wind driven generator at the (t + 1) th moment; and finally, calculating the power of the wind driven generator cluster at the (t + 1) th moment by combining the predicted wind speeds of the plurality of wind driven generators and the predicted power of the target wind driven generator. The method reduces the prediction error of the power of the wind driven generator cluster, and improves the prediction precision.

Description

Technical Field

[0001] The present application relates to the technical field of power prediction for a wind turbine cluster, and in particular to a method, apparatus, device, storage medium, and product for power prediction for a wind turbine cluster. Background Art

[0002] Wind power is currently one of the most important renewable energy sources. Its popularity has rapidly increased in recent years thanks to its clean and sustainable nature. As technology continues to advance, wind power is becoming increasingly efficient and cost-effective, making it an increasingly attractive option for meeting energy needs while reducing our carbon footprint.

[0003] With the rapid growth of installed capacity of wind turbines, accurate power prediction of wind turbine clusters can alleviate the pressure of peak load regulation and frequency regulation of power systems, which is conducive to the flexible scheduling of renewable energy grids.

[0004] However, due to the unpredictability and intermittency of wind energy, the wind speed at the location of the wind turbine has uncertainties. These uncertainties lead to very inaccurate prediction of the power of the wind turbine cluster based on the wind speed at the location of the wind turbine. Summary of the Invention

[0005] In view of this, the purpose of this application is to provide a power prediction method, device, equipment, storage medium and product for a wind turbine cluster to solve the above problems and improve the accuracy of power prediction for a wind turbine cluster.

[0006] In a first aspect, an embodiment of the present application provides a method for predicting power of a wind turbine cluster, the method comprising: The processor calculates, based on the correlation coefficients between every two wind turbines, the correlation coefficients between each wind turbine and the remaining wind turbines, and takes an average value thereof, which is the total correlation coefficient of the wind turbine; compares the total correlation coefficients of each wind turbine to obtain a maximum total correlation coefficient; and determines the wind turbine corresponding to the maximum total correlation coefficient as the target wind turbine; The processor obtains the wind speed of the plurality of wind turbines from time tx to time t and the power of the target wind turbine from time tx to time t; The processor inputs the wind speeds of the multiple wind turbines from time tx to time t into the trained first neural network model to obtain the wind speeds of the remaining wind turbines and the target wind turbine at time t+1; The processor inputs the power of the target wind turbine from time tx to time t into the trained second neural network model to obtain the power of the target wind turbine at time t+1; The processor obtains the power of the wind turbine cluster at the t+1th moment according to the wind speeds of the remaining wind turbines and the target wind turbine at the t+1th moment and the power of the target wind turbine at the t+1th moment.

[0007] Optionally, the processor calculates, based on the correlation coefficient between every two wind turbines, the correlation coefficients between each wind turbine and the remaining wind turbines, and takes an average value thereof, which is the total correlation coefficient of the wind turbine, including: The calculation expression of the total correlation coefficient is:

[0008] in, is the correlation coefficient between the i-th wind turbine and the j-th wind turbine, is the total correlation coefficient of the i-th wind turbine, n is the number of wind turbines in the wind turbine cluster, and i is not equal to j.

[0009] Optionally, the step of obtaining, by the processor, the power of the wind turbine cluster at time t+1 based on the wind speeds of the remaining wind turbines and the target wind turbine at time t+1 and the power of the target wind turbine at time t+1 includes: The processor calculates the wind speed coefficients of the locations of the plurality of wind turbines at the time t+1 according to the wind speeds of the remaining wind turbines and the target wind turbine at the time t+1; The processor calculates the power of the wind turbine cluster at time t+1 according to the wind speed coefficients at the locations of the multiple wind turbines at time t+1 and the power of the target wind turbine at time t+1.

[0010] Optionally, the step of calculating, by the processor, the wind speed coefficients at the locations of the plurality of wind turbines at the time t+1 based on the wind speeds of the remaining wind turbines and the target wind turbine at the time t+1 includes: The calculation formula for the wind speed coefficient at the locations of multiple wind turbines at time t+1 is as follows:

[0011] in, is the wind speed coefficient at the location of the i-th wind turbine at time t+1, is the wind speed at the location of the i-th wind turbine at time t+1, is the wind speed of the target wind turbine at the t+1th moment.

[0012] Optionally, the processor calculates the power of the wind turbine cluster at time t+1 based on the wind speed coefficients at the locations of the multiple wind turbines at time t+1 and the power of the target wind turbine at time t+1, including: The calculation formula of the wind turbine cluster power at time t+1 is as follows:

[0013] in, is the power of the wind turbine cluster at time t+1, is the wind speed coefficient at the location of the i-th wind turbine at time t+1, is the power of the target wind turbine at the t+1th moment.

[0014] The power prediction method for a wind turbine cluster provided in this application brings the following beneficial effects: The present application provides a power prediction method for a cluster of wind turbines, in which a processor first performs correlation modeling on the wind speeds of each wind turbine based on a D-Vine Copula model and calculates the required correlation coefficient. Secondly, the Shapley value method is used to distribute and integrate the weights of the correlation coefficients, and the average of the correlation coefficients between each wind turbine and the remaining wind turbines in the cluster is calculated to determine the target wind turbine. The wind speeds of the target wind turbine and the remaining wind turbines from the txth to the tth moment are input into a trained first neural network model to obtain the wind speeds of the target wind turbine and the remaining wind turbines at the t+1th moment. The power of the target wind turbine from the txth to the tth moment is input into a trained second neural network model to obtain the power of the target wind turbine at the t+1th moment. According to the wind speeds at the locations of multiple wind turbines at the t+1th moment, the wind speed of the target wind turbine at the t+1th moment, and the power of the target wind turbine at the t+1th moment, the power of the wind turbine cluster at the t+1th moment is obtained. The method amplifies or reduces the predicted power of the target wind turbine at the t+1 moment according to the magnitude relationship between the predicted wind speeds at the t+1 moment, effectively reducing the difference between the target wind turbine and the remaining wind turbines, reducing the prediction error of the power of the wind turbine cluster, and improving the prediction accuracy.

[0015] In a second aspect, the present application further provides a power prediction device for a wind turbine cluster, the device comprising: a determination module configured to calculate, based on the correlation coefficients between every two wind turbines, the correlation coefficients between each wind turbine and the remaining wind turbines, and to obtain an average value thereof, which is the total correlation coefficient of the wind turbine; compare the total correlation coefficients of each wind turbine to obtain a maximum total correlation coefficient; and determine the wind turbine corresponding to the maximum total correlation coefficient as the target wind turbine; An acquisition module, configured to acquire the wind speeds of multiple wind turbines from time tx to time t and the power of a target wind turbine from time tx to time t; A data processing module is used to input the wind speeds of the remaining wind turbines and the target wind turbine from time tx to time t into a trained first neural network model to obtain the wind speeds of the remaining wind turbines and the target wind turbine at time t+1; input the power of the target wind turbine from time tx to time t into a trained second neural network model to obtain the power of the target wind turbine at time t+1; and obtain the power of the wind turbine cluster at time t+1 based on the wind speeds at the locations of the multiple wind turbines at time t+1, the wind speed of the target wind turbine at time t+1, and the power of the target wind turbine at time t+1.

[0016] The power prediction device for a wind turbine cluster provided in the embodiment of the present application has the same technical features as the power prediction method for a wind turbine cluster provided in the above embodiment, and therefore can also solve the same technical problems and achieve the same technical effects.

[0017] In a third aspect, the present application provides a computing device, including a memory and a processor; One or more computer programs are stored in the memory, and the one or more computer programs include instructions; when the instructions are executed by the processor, the computing device executes the method as described in any one of the first aspects.

[0018] In a fourth aspect, the present application provides a computer-readable storage medium for storing a computer program for executing the method as described in any one of the first aspects.

[0019] In a fifth aspect, the present application provides a computer program product, which includes one or more computer instructions. When the computer instructions are executed by a computer, the computer executes the method as described in any one of the first aspects.

[0020] Other features and advantages of the present application will be described in the following description, and in part will become apparent from the description, or understood by practicing the present application. The purpose and other advantages of the present application are realized and obtained by the structures particularly pointed out in the description and drawings.

[0021] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the specific implementation methods of the present application or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the specific implementation methods or the description of the prior art. Obviously, the drawings described below are some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0023] Figure 1 A schematic diagram of a power prediction process for a wind turbine cluster provided in an embodiment of the present application; Figure 2 A schematic diagram of a power prediction value of a target wind turbine provided in an embodiment of the present application; Figure 3 A schematic diagram of a wind speed prediction value of a target wind turbine provided in an embodiment of the present application; Figure 4 A schematic diagram of wind turbine cluster power prediction without using wind speed coefficients provided in an embodiment of the present application; Figure 5 A schematic diagram of wind turbine cluster power prediction using wind speed coefficients provided in an embodiment of the present application; Figure 6 A schematic diagram of the structure of a power prediction device for a wind turbine cluster provided in an embodiment of the present application; Figure 7 A schematic diagram of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0025] To facilitate understanding of this embodiment, the embodiments of this application are described in detail below.

[0026] The present application provides a method for predicting the power of a wind turbine cluster. Figure 1 As shown, Figure 1 A schematic diagram of a power prediction process for a wind turbine cluster provided in an embodiment of the present application. The method includes the following steps: S1011 , the processor determines a correlation coefficient between every two wind turbines.

[0027] Specifically, wind power generation is the process of converting wind energy into mechanical energy and then into electrical energy. Wind energy is affected by wind speed, which in turn affects the wind turbine's ability to convert wind energy into electrical energy and electrical energy into mechanical energy. The formula for calculating the power of a wind turbine is as follows:

[0028] in, is the power of the wind turbine, is the wind speed at the location of the wind turbine, is the rated power of the wind turbine, is the power coefficient, is the air density at the location of the wind turbine, is the swept area of ​​the wind turbine blades, is the cut-in wind speed at the location of the wind turbine, is the rated wind speed at the location of the wind turbine, is the cut-out wind speed at the location of the wind turbine.

[0029] It can be seen from the above relationship formula that the wind speed at different wind turbine locations will cause the wind turbines to generate different powers. Considering that it is easier to obtain wind speed data than power data, in order to reduce the impact of data quality on power prediction and the difficulty of data collection, wind speed data is used to analyze and count the correlation between the wind turbines in the wind turbine cluster, so as to determine the target wind turbine for power prediction.

[0030] When calculating the correlation between wind turbines in a wind turbine cluster, in order to fully consider the mutual influence between the entire cluster and improve the comprehensiveness of the correlation analysis results, the present invention uses a Copula function to perform correlation analysis calculations. A Copula function is a function that connects the marginal distributions of various uncertainty parameters and fits their joint distribution form. The n-dimensional Copula function can be expressed as:

[0031] in, is the Copula density function, is the correlation parameter of the Copula function, ~U(0,1), is the uncertainty parameter The marginal distribution function of .

[0032] Since the multidimensional Copula function has a large amount of calculation and low efficiency, the D-Vine Copula function is used to calculate the multidimensional data. The D-Vine Copula function can be expressed as:

[0033] in, is the D-Vine Copula function value of the uncertainty parameter corresponding to the wind speed at the location of the wind turbine and the wind speed at the location of the wind turbine, is the conditional Copula probability density function between the i-th and i+j-th variables under the condition that the uncertainty parameters from the i+1th to the i+j-1th are known, for The marginal probability density function of is the wind speed at the location of the i-th wind turbine, is the correlation parameter of the Copula function.

[0034] The D-Vine Copula function value is composed of several types of Copula functions, specifically five types: normal Copula function, t-Copula function, Gumbel-Copula function, Clayton-Copula function, and Frank-Copula function. Their probability distribution function expressions are as follows: Normal Copula function:

[0035] in is the normal Copula function, The correlation coefficient is The distribution function of the two-dimensional normal distribution is, is the inverse function of the standard normal distribution function of the wind speed at the location of the i-th wind turbine, is the inverse function of the standard normal distribution function of the wind speed at the location of the j-th wind turbine, is the wind speed at the location of the j-th wind turbine, is the wind speed at the location of the i-th wind turbine.

[0036] t-Copula function:

[0037] in, is the t-Copula function, The correlation coefficient is And the degrees of freedom are The distribution function of the two-dimensional Gaussian distribution, The degrees of freedom are The inverse function of the t distribution of the wind speed at the location of the i-th wind turbine is, The degrees of freedom are The inverse function of the t distribution of the wind speed at the location of the j-th wind turbine.

[0038] Gumbel-Copula function:

[0039] in, is the Gumbel-Copula function.

[0040] Clayton-Copula function:

[0041] That is the Clayton-Copula function.

[0042] Frank-Copula function:

[0043] in, is the Frank-Copula function.

[0044] Before using the Copula function to calculate, and It needs to obey the uniform distribution between [0,1]. The empirical distribution can be used to convert the samples of the uncertainty parameter into uniform variables between [0,1] using the empirical distribution formula. The empirical distribution formula is as follows:

[0045] in, is the empirical distribution, n is the sample size, d is the number of variables, is the rank of the sample (the variable Sample Sort from small to large, no larger than The number of samples is called rank).

[0046] After converting each variable into a uniformly distributed variable in the [0,1] interval, the relevant parameters of each type of Copula function are calculated by maximum likelihood estimation. Finally, the Akaike Information Criterion (AIC) is used to determine the optimal Copula function between the variables. The AIC calculation formula is:

[0047] in, is the number of parameters of the copula function, is the maximum likelihood function value of the copula function. The smaller the AIC, the better the model fit, and the optimal copula function can be selected.

[0048] After obtaining the optimal copula function, calculate the Spearman rank correlation coefficient, Kendall rank correlation coefficient, upper tail correlation coefficient, and lower tail correlation coefficient of the function. The calculation formula is as follows: The calculation formula of Spearman rank correlation coefficient is as follows:

[0049] in, is the Spearman rank correlation coefficient, is the Copula function value.

[0050] The calculation formula of Kendall's rank correlation coefficient is as follows:

[0051] in, is the Kendall rank correlation coefficient, is the Copula function value.

[0052] The upper tail correlation coefficient is calculated as follows:

[0053] in, is the upper tail correlation coefficient, is the inverse function value of the marginal distribution function of the wind speed at the location of the j-th wind turbine, is the inverse function value of the marginal distribution function of the wind speed at the location of the i-th wind turbine, is the quantile.

[0054] The formula for calculating the lower tail correlation coefficient is as follows:

[0055] in, is the lower tail correlation coefficient.

[0056] The following example illustrates this in detail. For example, we first collected relevant data from four wind turbines in a wind farm. Some of the data is shown in Table 1: Table 1:

[0057] The wind turbine data collection period was from 10:49:16 on February 28, 2023, to 13:59:16 on March 22, 2023, with a time step of 5 minutes. A total of 6,374 data sets were divided into sample value training set data and test set data in an 8:2 ratio. The first 80% was used as the training set data for correlation analysis and neural network model training, and the latter was used as the test set to verify the effectiveness of the present invention. The D-Vine Copula function was fitted to the wind speed data of each wind turbine training set, and the optimal Copula function and related parameters were finally obtained as shown in Table 2 below: Table 2:

[0058] After obtaining the Copula function between each wind turbine, the Spearman rank correlation coefficient, Kendall rank correlation coefficient, upper tail correlation coefficient, and lower tail correlation coefficient can be calculated according to the rank correlation coefficient calculation formula as shown in the following table: Table 3:

[0059] After obtaining the Spearman rank correlation coefficient, Kendall rank correlation coefficient, upper tail correlation coefficient and lower tail correlation coefficient, in order to reasonably distribute the proportion of these four correlation coefficients in the correlation, the Shapley value method is used to calculate their weight coefficients.

[0060] First calculate the contribution value of each combination:

[0061] in, is the contribution value of the Spearman rank correlation coefficient, is the contribution value of Kendall rank correlation coefficient, is the contribution value of the upper tail correlation coefficient, is the contribution value of the lower tail correlation coefficient, is the contribution value of the combination of Spearman rank correlation coefficient and Kendall rank correlation coefficient, is the contribution value of the combination of Spearman rank correlation coefficient and upper tail correlation coefficient, is the contribution value of the combination of Spearman rank correlation coefficient and lower tail correlation coefficient, is the contribution value of the combination of Kendall rank correlation coefficient and upper tail correlation coefficient, is the contribution value of the combination of Kendall rank correlation coefficient and lower tail correlation coefficient, is the contribution value of the combination of the upper tail correlation coefficient and the lower tail correlation coefficient, is the contribution value of the combination of Spearman rank correlation coefficient, Kendall rank correlation coefficient and upper tail correlation coefficient, is the contribution value of the combination of Spearman rank correlation coefficient, Kendall rank correlation coefficient and lower tail correlation coefficient, is the contribution value of the combination of Spearman rank correlation coefficient, upper tail correlation coefficient and lower tail correlation coefficient, is the contribution value of the combination of Kendall rank correlation coefficient, upper tail correlation coefficient and lower tail correlation coefficient, is the contribution value of the combination of Spearman rank correlation coefficient, Kendall rank correlation coefficient, upper tail correlation coefficient and lower tail correlation coefficient, is the Spearman rank correlation coefficient of the optimal Copula function value, is the Kendall rank correlation coefficient of the optimal Copula function value, is the upper tail correlation coefficient of the optimal Copula function value, is the lower tail correlation coefficient of the optimal Copula function value.

[0062] Then, based on the contribution value, the contribution weight of the Spearman rank correlation coefficient, the contribution weight of the Kendall rank correlation coefficient, the contribution weight of the upper tail correlation coefficient, and the contribution weight of the lower tail correlation coefficient are calculated using the following formula:

[0063] in, is the contribution weight of the Spearman rank correlation coefficient, is the contribution weight of the Kendall rank correlation coefficient, is the contribution weight of the upper tail correlation coefficient, is the contribution weight of the lower tail correlation coefficient.

[0064] Finally, the contribution weight coefficients of the Spearman rank correlation coefficient, the Kendall rank correlation coefficient, the upper tail correlation coefficient, and the lower tail correlation coefficient are calculated using the following formula:

[0065] in, is the contribution weight of the Spearman rank correlation coefficient, is the contribution weight of the Kendall rank correlation coefficient, is the contribution weight of the upper tail correlation coefficient, is the contribution weight of the lower tail correlation coefficient, is the contribution weight coefficient of the Spearman rank correlation coefficient, is the contribution weight coefficient of the Kendall rank correlation coefficient, is the weight coefficient of the contribution of the upper tail correlation coefficient, is the contribution weight coefficient of the lower tail correlation coefficient, It is the contribution value of the combination of Spearman rank correlation coefficient, Kendall rank correlation coefficient, upper tail correlation coefficient and lower tail correlation coefficient.

[0066] By multiplying the above four correlation coefficients with the corresponding weight coefficients and then adding them together, the correlation coefficient between every two wind turbines can be calculated:

[0067] in, is the correlation coefficient between the i-th wind turbine and the j-th wind turbine (correlation coefficient between every two wind turbines), is the Spearman rank correlation coefficient of the optimal copula function formed by the i-th wind turbine and the j-th wind turbine, is the Kendall rank correlation coefficient of the optimal copula function formed by the i-th wind turbine and the j-th wind turbine, is the upper tail correlation coefficient of the optimal copula function formed by the i-th wind turbine and the j-th wind turbine, is the lower tail correlation coefficient of the optimal copula function formed by the i-th wind turbine and the j-th wind turbine.

[0068] By calculating the above expression, the correlation coefficient between every two wind turbines calculated in Table 3 can be obtained, and the results are shown in Table 4.

[0069] Table 4:

[0070] S1012: The processor calculates the correlation coefficients between each wind turbine and the remaining wind turbines based on the correlation coefficients between each two wind turbines, and takes the average value of the calculated correlation coefficients, which is the total correlation coefficient of the wind turbine.

[0071] Specifically, the processor calculates the total correlation coefficient of the wind turbine according to the following formula:

[0072] in, is the correlation coefficient between the i-th wind turbine and the j-th wind turbine, is the total correlation coefficient of the i-th wind turbine, n is the number of wind turbines in the wind turbine cluster, and i is not equal to j.

[0073] S1013: The processor compares the total correlation coefficients of each wind turbine to obtain a maximum total correlation coefficient.

[0074] S1014: The processor determines the wind turbine corresponding to the largest total correlation coefficient as the target wind turbine.

[0075] The following example illustrates that the correlation coefficient between each two wind turbines in Table 3 can be obtained according to the expression of the total correlation coefficient: the total correlation coefficient of the first wind turbine is =0.4352, the total correlation coefficient of the second wind turbine is 0.5417, the total correlation coefficient of the third wind turbine is 0.5466, and the total correlation coefficient of the fourth wind turbine is 0.5396. Therefore, the third wind turbine has the largest correlation coefficient, so the third wind turbine is determined to be the target wind turbine.

[0076] This approach quantifies the synergistic impact of a single wind turbine on the prediction accuracy by calculating the contribution of all possible wind turbine combinations to the prediction. This approach objectively reflects the actual impact of each wind turbine's correlation with other wind turbines on the prediction of the wind turbine cluster, improving the accuracy of target wind turbine selection.

[0077] S102: The processor obtains the wind speeds of multiple wind turbines from time tx to time t and the power of a target wind turbine from time tx to time t.

[0078] Specifically, a wind speed sensor and a power sensor are installed on the target wind turbine. The processor obtains the wind speed of the target wind turbine through the wind speed sensor and obtains the power of the target wind turbine through the power sensor.

[0079] S103, the processor inputs the wind speeds of the multiple wind turbines from time tx to time t into the trained first neural network model to obtain the wind speeds of the remaining wind turbines and the target wind turbine at time t+1.

[0080] Specifically, the training process of the first neural network model is as follows: (1) Obtain the wind speed sample value of the target wind turbine.

[0081] (2) According to the wind speed sample value of the target wind turbine, the k-th order residual of the wind speed sample value and all k-order modal components of the wind speed sample value are obtained.

[0082] More specifically, the wind speed data is decomposed using the ICEEMDAN method. This method implements iterative decomposition by adding special noise containing Gaussian white noise to the data multiple times, thereby obtaining the k-th order residual and the k-th order modal component of the wind speed sample value. The process is as follows: S1, decompose the g-th Gaussian white noise through empirical mode decomposition to obtain the first-order component of the g-th Gaussian white noise.

[0083] S2, the g-th mixed signal is calculated based on the first-order component of the g-th Gaussian white noise and the wind speed sample value. The calculation formula of the g-th mixed signal is as follows:

[0084] in, is the g-th mixed signal, is the wind speed sample value, is a random parameter, is the first-order component of the g-th Gaussian white noise, is the g-th Gaussian white noise.

[0085] S3, calculate the first-order residual of the wind speed sample value based on the g-th mixed signal. The calculation formula of the first-order residual of the wind speed sample value is as follows:

[0086] in, is the g-th mixed signal, is the first-order residual of the wind speed sample value, is the local mean of the g-th mixed signal.

[0087] S4, calculate the first-order modal component of the wind speed sample value based on the wind speed sample value and the first-order residual of the wind speed sample value. The calculation formula of the first-order modal component of the wind speed sample value is as follows:

[0088] in, is the wind speed sample value, is the first-order residual of the wind speed sample value, is the first-order modal component of the wind speed sample value.

[0089] S5, performing empirical mode decomposition on the g-th Gaussian white noise to obtain the second-order component of the g-th Gaussian white noise.

[0090] S6. Calculate the second-order residual of the wind speed sample value based on the first-order residual of the wind speed sample value and the second-order component of the g-th Gaussian white noise. The calculation formula of the second-order residual of the wind speed sample value is as follows:

[0091] in, is the second-order residual of the wind speed sample value, is the first-order residual of the wind speed sample value, is a random parameter, is the second-order component of the g-th Gaussian white noise, is the g-th Gaussian white noise.

[0092] S7, execute the above steps S1 to S6 to obtain the k-th order residual of the wind speed sample value and the k-th order modal component of the wind speed sample value. The calculation formulas for the k-th order residual of the wind speed sample value and the k-th order modal component of the wind speed sample value are as follows:

[0093] in, is the k-th order residual of the wind speed sample value, is the k-1th order residual of the wind speed sample value, is the random parameter added in the K-1th round, is the K-1th order component of the gth Gaussian white noise, is the kth-order modal component of the wind speed sample value.

[0094] (3) All k-order modal components and k-order residuals of the wind speed sample values ​​are input into the original first neural network model for training to obtain the first neural network model.

[0095] Specifically, the first neural network model is a convolutional-long short-term memory hybrid network (CNN-LSTM) model. The convolutional neural network (CNN) model is a feedforward neural network model that includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. The structural formula of a CNN is as follows:

[0096] in, is the output of the convolutional layer, is the output of the pooling layer, is the output of the fully connected layer, is the matrix composed of the k-th order residual and the k-th order modal component of the wind speed sample value (the matrix composed of the input data), is the convolutional layer weight matrix, is the weight matrix of the fully connected layer, is the bias vector of the convolutional layer, is the bias vector of the pooling layer, is the bias vector of the fully connected layer, is the activation function of the convolutional layer, is the pooling method of the pooling layer, is the activation function of the fully connected layer.

[0097] The Long Short-Term Memory (LSTM) model is a special recurrent network model with significant advantages for processing time series signals. The core of the LSTM lies in its gating mechanism, including the forget gate, input gate, and output gate. These gating structures enable the network to learn when to forget old information, when to accept new information, and when to output information. The LSTM structural formula is as follows:

[0098] in, is the output value of the forget gate at time t, is the Sigmoid function, is the forget gate weight coefficient 1 at the tth moment, is the forget gate weight coefficient 2 at the tth moment, is the forget gate bias coefficient, is the output of the CNN neural network, It is the output of the LSTM neural network at the previous moment.

[0099] The input gate determines the input x (t) How much has entered the current kind.

[0100]

[0101]

[0102]

[0103] in, is the input gate; and is the input gate weight, 、 is the bias coefficient, is the information updated at time t, Tanh is the hyperbolic tangent activation function, and is the temporary unit weight, is the temporary unit memory state at time t.

[0104] The output gate determines the output hidden information content.

[0105]

[0106]

[0107] in, is the output gate, and is the output gate weight, is the bias coefficient, is the output at time t.

[0108] This setup solves the problem of lagging prediction results when using wind speed samples (historical wind speeds) from time tx to t for the target wind turbine. Using the ICEEMDAN method to decompose historical wind speeds allows the original first neural network model to better learn the patterns within the data, improving prediction accuracy. Iterative decomposition is achieved by repeatedly adding special noise containing Gaussian white noise to the data, and the reconstruction rate is improved by averaging the modal components obtained at each decomposition order, thereby ensuring the integrity of the reconstructed signal.

[0109] S104, the processor inputs the power of the target wind turbine from time tx to time t into the trained second neural network model to obtain the power of the target wind turbine at time t+1.

[0110] Specifically, the second neural network model is also a convolution-long short-term memory hybrid network model. The training process of the second neural network model is as follows: (1) Obtain the power sample value of the target wind turbine.

[0111] (2) According to the power sample value of the target wind turbine, the k-th order residual of the power sample value and all k-order modal components of the power sample value are obtained. The specific steps are consistent with the decomposition steps of the wind speed in S103.

[0112] (3) All k-order modal components and k-th order residuals of the power sample values ​​are input into the original second neural network model for training to obtain the second neural network model.

[0113] In the training process of the first and second neural networks, only the root mean square error (RMSE) was used as the error evaluation criterion, while in the evaluation of the prediction results, the root mean square error (RMSE), normalized root mean square error (NRMSE), mean absolute error (MAE), normalized mean absolute error (NMAE) and coefficient of determination (R) were used. 2 There are five types. Their respective calculation expressions are:

[0114]

[0115]

[0116]

[0117]

[0118] Where n is the number of data, is the actual value, is the predicted value, is the true mean value.

[0119] The following example illustrates that the input format of the hybrid neural network model is [1,1,1], the convolution kernel dimension of the convolution layer is 2, the number of convolution kernels is 10, the number of hidden units in the long short-term memory network model is 128, the data discard ratio of the forget gate is 0.4, the optimization algorithm is the Adam algorithm, and the L2 regularization factor is 1×10 -4 , the initial learning rate is 0.01, the learning rate reduction period is 1 epoch, the learning rate reduction factor is 0.75, the time series length is 72, and the minimum batch size is 64.

[0120] Figure 2 and Figure 3 The power and wind speed prediction diagram of the target wind turbine is shown in Figure 2. The predicted power error obtained by the neural network is RMSE 33.1058, NRMSE 0.02244, MAE 18.4561, and NMAE 0.01251. It is 0.9699.

[0121] The predicted wind speed errors are RMSE 0.39429, NRMSE 0.034893, MAE 0.2965, and NMAE 0.026239. It is 0.94924.

[0122] S105 , the processor obtains the power of the wind turbine cluster at time t+1 according to the wind speeds of the remaining wind turbines and the target wind turbine at time t+1 and the power of the target wind turbine at time t+1.

[0123] Specifically, the processor calculates the wind speed coefficients of the locations of the multiple wind turbines at the time t+1 according to the wind speeds of the remaining wind turbines and the target wind turbine at the time t+1.

[0124] More specifically, the calculation formula for the wind speed coefficient at the locations of the multiple wind turbines at time t+1 is as follows:

[0125] in, is the wind speed coefficient at the location of the i-th wind turbine at time t+1, is the wind speed at the location of the i-th wind turbine at time t+1, is the wind speed of the target wind turbine at time t+1.

[0126] The processor calculates the power of the wind turbine cluster at time t+1 according to the wind speed coefficients at the locations of the multiple wind turbines at time t+1 and the power of the target wind turbine at time t+1.

[0127] More specifically, the calculation formula for the power of the wind turbine cluster at time t+1 is as follows:

[0128] in, is the wind speed coefficient at the location of the i-th wind turbine at time t+1, is the power of the wind turbine cluster at time t+1, is the power of the target wind turbine at the t+1th moment.

[0129] By dividing the predicted wind speed of the target wind turbine by the predicted wind speed of the remaining wind turbines, the wind speed coefficient of the wind turbine is obtained, and the predicted power of the target wind turbine is expanded or reduced according to the wind speed coefficient to obtain the predicted power of the wind turbine. Finally, the predicted power of all wind turbines is added together to obtain the predicted total power of the wind turbine cluster. The prediction results without using the wind speed coefficient and with using the wind speed coefficient are as follows Figure 4 and attached Figure 5 By amplifying and reducing the standard wind turbine prediction data through the coefficient and then adding them together, the prediction error of the wind turbine cluster can be effectively reduced. The comparison of the prediction error before and after the wind speed coefficient is used is shown in Table 5: Table 5:

[0130] An embodiment of the present application provides a power prediction method for a cluster of wind turbines, in which a processor first performs correlation modeling on the wind speeds of each wind turbine based on a D-Vine Copula model to calculate the required correlation coefficient. Secondly, the Shapley value method is used to allocate and integrate the weights of the correlation coefficients, and the average of the correlation coefficients between each wind turbine and the remaining wind turbines in the cluster is obtained. By determining a target wind turbine, the wind speeds of the target wind turbine and the remaining wind turbines from time tx to time t are input into a trained first neural network model to obtain the wind speeds of the target wind turbine and the remaining wind turbines at time t+1, the power of the target wind turbine from time tx to time t is input into a trained second neural network model to obtain the power of the target wind turbine at time t+1, and the power of the wind turbine cluster at time t+1 is obtained based on the wind speeds at the locations of multiple wind turbines at time t+1, the wind speed of the target wind turbine at time t+1, and the power of the target wind turbine at time t+1. This method amplifies or reduces the predicted power of the target wind turbine (standard wind turbine) at the t+1th moment according to the magnitude relationship between the predicted wind speeds at the t+1th moment, effectively reducing the difference between the target wind turbine and the remaining wind turbines, reducing the prediction error of the power of the wind turbine cluster, and improving the prediction accuracy.

[0131] Based on the above method embodiment, the embodiment of the present application also provides a power prediction device for a wind turbine cluster, such as Figure 6 As shown, Figure 6 This is a schematic diagram of the structure of a power prediction device for a wind turbine cluster provided in an embodiment of the present application. The device includes: a determination module 41, an acquisition module 42, and a data processing module 43. The functions of each module are as follows: Determination module 41 is configured to calculate, based on the correlation coefficients between every two wind turbines, the correlation coefficients between each wind turbine and the remaining wind turbines, and to obtain an average value thereof, which is the total correlation coefficient of the wind turbine; compare the total correlation coefficients of each wind turbine to obtain a maximum total correlation coefficient; and determine the wind turbine corresponding to the maximum total correlation coefficient as the target wind turbine; An acquisition module 42 is configured to acquire the wind speeds of multiple wind turbines from time tx to time t and the power of a target wind turbine from time tx to time t; The data processing module 43 is used to input the wind speeds of the remaining wind turbines and the target wind turbine from time tx to time t into the trained first neural network model to obtain the wind speeds of the remaining wind turbines and the target wind turbine at time t+1; input the power of the target wind turbine from time tx to time t into the trained second neural network model to obtain the power of the target wind turbine at time t+1; and obtain the power of the wind turbine cluster at time t+1 based on the wind speeds at the locations of the multiple wind turbines at time t+1, the wind speed of the target wind turbine at time t+1, and the power of the target wind turbine at time t+1.

[0132] Preferably, the data processing module 42 is specifically used to calculate the total correlation coefficient, which is expressed as:

[0133] in, is the correlation coefficient between the i-th wind turbine and the j-th wind turbine, is the total correlation coefficient of the i-th wind turbine, n is the number of wind turbines in the wind turbine cluster, and i is not equal to j.

[0134] Preferably, the data processing module 42 is specifically configured to calculate the wind speed coefficients of the locations of the plurality of wind turbines at the time t+1 according to the wind speeds of the remaining wind turbines and the target wind turbine at the time t+1; The processor calculates the power of the wind turbine cluster at time t+1 according to the wind speed coefficients at the locations of the multiple wind turbines at time t+1 and the power of the target wind turbine at time t+1.

[0135] Preferably, the data processing module 42 is specifically configured to calculate the wind speed coefficients at the locations of the multiple wind turbines at time t+1 according to the following formula:

[0136] in, is the wind speed coefficient at the location of the i-th wind turbine at time t+1, is the wind speed at the location of the i-th wind turbine at time t+1, is the wind speed of the target wind turbine at the t+1th moment.

[0137] Preferably, the data processing module 42 is specifically configured to calculate the power of the wind turbine cluster at time t+1 according to the following formula:

[0138] in, is the power of the wind turbine cluster at time t+1, is the wind speed coefficient at the location of the i-th wind turbine at time t+1, is the power of the target wind turbine at the t+1th moment.

[0139] The power prediction device for a wind turbine cluster provided in the embodiment of the present application has the same technical features as the power prediction method for a wind turbine cluster provided in the above embodiment, and therefore can solve the same technical problems and achieve the same technical effects.

[0140] The present application also provides a computing device. Figure 7 As shown, this figure is a schematic diagram of a computing device provided by an embodiment of the present application, wherein the computing device 400 includes a bus 401, a processor 402, a communication interface 403, and a memory 404. The processor 402, the memory 404, and the communication interface 403 communicate with each other via the bus 401.

[0141] The bus 401 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0142] The processor 402 may be any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0143] Communication interface 403 is used for external communication. Memory 404 may include volatile memory, such as random access memory (RAM). Memory 404 may also include non-volatile memory, such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid state drive (SSD).

[0144] The memory 404 stores executable codes, and the processor 402 executes the executable codes to perform the aforementioned power prediction method for the wind turbine cluster.

[0145] Embodiments of the present application also provide a computer-readable storage medium. The computer-readable storage medium can be any available medium capable of storing data on a computing device, or a data storage device such as a data center that contains one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, hard disk, or magnetic tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive). The computer-readable storage medium includes instructions that instruct the computing device to execute the above method.

[0146] The present application also provides a computer program product comprising one or more computer instructions that, when loaded and executed on a computing device, fully or partially generate the process or function described in the present application.

[0147] The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer or data center to another website, computer or data center via wired (e.g., coaxial cable, optical fiber) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0148] When the computer program product is executed by a computer, the computer performs any of the aforementioned methods for predicting power for a wind turbine cluster. The computer program product may be a software installation package, which can be downloaded and executed on a computer when any of the aforementioned methods for predicting power for a wind turbine cluster is needed.

[0149] The descriptions of the processes or structures corresponding to the above figures have different emphases. For parts that are not described in detail in a certain process or structure, please refer to the relevant descriptions of other processes or structures.

[0150] The above description is only a specific implementation method of the present application, but the protection scope of the present application is not limited thereto. Any changes or replacements within the technical scope disclosed in the present application should be included in the protection scope of the present application.

Claims

1. A power prediction method for a wind turbine cluster, characterized in that: The method comprises: The processor calculates, based on the correlation coefficients between every two wind turbines, the correlation coefficients between each wind turbine and the remaining wind turbines, and takes an average value thereof, which is the total correlation coefficient of the wind turbine; compares the total correlation coefficients of each wind turbine to obtain a maximum total correlation coefficient; and determines the wind turbine corresponding to the maximum total correlation coefficient as the target wind turbine; The processor obtains the wind speed of the plurality of wind turbines from time tx to time t and the power of the target wind turbine from time tx to time t; The processor inputs the wind speeds of the multiple wind turbines from time tx to time t into the trained first neural network model to obtain the wind speeds of the remaining wind turbines and the target wind turbine at time t+1; The processor inputs the power of the target wind turbine generator from time tx to time t into the trained second neural network model to obtain the power of the target wind turbine generator at time t+1; The processor obtains the power of the wind turbine cluster at time t+1 according to the wind speeds of the remaining wind turbines and the target wind turbine at time t+1 and the power of the target wind turbine at time t+1.

2. The power prediction method for a wind turbine cluster according to claim 1, characterized in that: The processor calculates the correlation coefficients between each wind turbine and the remaining wind turbines based on the correlation coefficients between each two wind turbines, and takes the average value thereof, which is the total correlation coefficient of the wind turbine, including: The calculation expression of the total correlation coefficient is: in, is the correlation coefficient between the i-th wind turbine and the j-th wind turbine, is the total correlation coefficient of the i-th wind turbine, n is the number of wind turbines in the wind turbine cluster, and i is not equal to j.

3. The power prediction method for a wind turbine cluster according to claim 1, characterized in that: The step of obtaining, by the processor, the power of the wind turbine cluster at time t+1 according to the wind speeds of the remaining wind turbines and the target wind turbine at time t+1 and the power of the target wind turbine at time t+1 includes: The processor calculates the wind speed coefficients of the locations of the plurality of wind turbines at the time t+1 according to the wind speeds of the remaining wind turbines and the target wind turbine at the time t+1; The processor calculates the power of the wind turbine cluster at time t+1 according to the wind speed coefficients at the locations of the multiple wind turbines at time t+1 and the power of the target wind turbine at time t+1.

4. The power prediction method for a wind turbine cluster according to claim 3, characterized in that: The step of calculating, by the processor, wind speed coefficients at locations of the plurality of wind turbines at time t+1 based on the wind speeds of the remaining wind turbines and the target wind turbine at time t+1 includes: The calculation formula for the wind speed coefficient at the locations of multiple wind turbines at time t+1 is as follows: in, is the wind speed coefficient at the location of the i-th wind turbine at time t+1, is the wind speed at the location of the i-th wind turbine at time t+1, is the wind speed of the target wind turbine at the t+1th moment.

5. The power prediction method for a wind turbine cluster according to claim 3, characterized in that: The processor calculates the power of the wind turbine cluster at time t+1 based on the wind speed coefficients at the locations of the multiple wind turbines at time t+1 and the power of the target wind turbine at time t+1, comprising: The calculation formula of the wind turbine cluster power at time t+1 is as follows: in, is the power of the wind turbine cluster at time t+1, is the wind speed coefficient at the location of the i-th wind turbine at time t+1, is the power of the target wind turbine at the t+1th moment.

6. A power prediction device for a wind turbine cluster, characterized in that: The device comprises: a determination module configured to calculate, based on the correlation coefficients between every two wind turbines, the correlation coefficients between each wind turbine and the remaining wind turbines, and to obtain an average value thereof, which is the total correlation coefficient of the wind turbine; compare the total correlation coefficients of each wind turbine to obtain a maximum total correlation coefficient; and determine the wind turbine corresponding to the maximum total correlation coefficient as the target wind turbine; An acquisition module, configured to acquire the wind speeds of multiple wind turbines from time tx to time t and the power of a target wind turbine from time tx to time t; A data processing module is used to input the wind speeds of the remaining wind turbines and the target wind turbine from time tx to time t into a trained first neural network model to obtain the wind speeds of the remaining wind turbines and the target wind turbine at time t+1; input the power of the target wind turbine from time tx to time t into a trained second neural network model to obtain the power of the target wind turbine at time t+1; and obtain the power of the wind turbine cluster at time t+1 based on the wind speeds at the locations of the multiple wind turbines at time t+1, the wind speed of the target wind turbine at time t+1, and the power of the target wind turbine at time t+1.

7. The device according to claim 6, characterized in that The data processing module is specifically used to calculate the total correlation coefficient, which is expressed as: in, is the correlation coefficient between the i-th wind turbine and the j-th wind turbine, is the total correlation coefficient of the i-th wind turbine, n is the number of wind turbines in the wind turbine cluster, and i is not equal to j.

8. A computing device, characterized in that including memory and processor; One or more computer programs are stored in the memory, and the one or more computer programs include instructions; when the instructions are executed by the processor, the computing device executes the method according to any one of claims 1 to 5.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium is used to store a computer program, and the computer program is used to execute the method according to any one of claims 1 to 5.

10. A computer program product, characterized in that The computer program product comprises one or more computer instructions. When the computer instructions are executed by a computer, the computer performs the method according to any one of claims 1 to 5.

Images