Short-term wind power prediction method

Abstract

The application discloses a short-term wind power prediction method, steps comprising: S1, collecting historical wind power data, and preprocessing the data, S2, improved frost and ice optimization algorithm, extracting the multi-frequency domain features of wind power, S3, constructing an LSTM model and improving the fishing algorithm, S4, determining the optimal learning rate and penalty factor of the LSTM model based on S3, obtaining an ECFOA-LSTM network model, and S5, obtaining the wind power prediction value based on the IMF component set obtained in S2. The application improves the frost and ice optimization algorithm and the fishing algorithm, determines the optimal learning rate and penalty factor of the model through the improved fishing algorithm, and finally obtains the ECFOA-LSTM network model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wind power prediction, in particular to a short-term wind power prediction method. BACKGROUND

[0002] Wind power plays an irreplaceable role in addressing environmental problems and energy crises, but its strong non-stationarity and intermittency affect the smooth operation of the power grid, power generation efficiency and power quality. Accurate wind power prediction technology is crucial for stable operation and economic dispatch of power systems, and helps wind farms develop scientific and reasonable control strategies to improve wind energy utilization. From the perspective of using models, prediction methods can be mainly divided into physical models, statistical models and combined models. In recent years, combined prediction models have gradually replaced single prediction models due to their high prediction accuracy and have become the mainstream research direction. Combined prediction models apply intelligent optimization algorithms to the field of wind power prediction, optimize the parameters or structure of wind speed prediction models, and further improve the prediction accuracy.

[0003] The existing short-term wind power prediction technology has bottlenecks: the variational mode decomposition (VMD) lacks self-adaptive ability, the signal source signal-to-noise ratio is low, the frequency domain is mixed, and it is easy to over-filter or not completely filter, resulting in inaccurate estimation of the number of signal sources; when separating the signal source, there are problems such as error filtering of the source signal, inaccurate estimation of the number of signal sources, and large time-frequency domain distribution; the fitting ability of the LSTM model depends on the hyperparameters, and the traditional parameter tuning is low in efficiency and poor in accuracy; the real-time monitoring and prediction method based on sensor data is high in cost, limited by the accuracy and reliability of the equipment, and low in economy, so its popularization prospect is limited.

[0004] In summary, the existing technology in the field of short-term wind power prediction mainly has problems such as low prediction accuracy and poor economy, which makes it difficult to effectively support the further development of wind power generation. SUMMARY

[0005] The purpose of the present application is to provide a short-term wind power prediction method to solve the problem of excessive error in the prediction of wind power generation by the LSTM prediction model in the prior art.

[0006] The technical solution adopted by the present application is a short-term wind power prediction method, comprising the following steps:

[0007] Step S1, collect historical wind power data and preprocess the data;

[0008] Step S2, improve the frost and ice optimization algorithm to extract the multi-frequency domain features of wind power;

[0009] Step S3, construct an LSTM model and improve the fishing algorithm;

[0010] Step S4, based on S3, the best learning rate and penalty factor of the LSTM model are determined, and the ECFOA-LSTM network model is obtained;

[0011] Step S5, based on the IMF component set obtained in S2, the wind power prediction value is obtained.

[0012] The specific steps of S2 are as follows:

[0013] S21, the frost and ice optimization algorithm is improved, and the cosine transformation of the frost and ice factor is improved into positive and cosine double transformation, specifically as follows:

[0014] ;

[0015] Wherein, rand represents a random number in (0~1) obeying normal distribution, Riter represents the current iteration number of the frost and ice optimization algorithm, Max_Riter represents the maximum iteration number of the frost and ice optimization algorithm, W represents the parameter of the algorithm, and the value of the application is 5, RimeFactor represents the frost and ice factor, and determines the performance of the frost and ice optimization algorithm;

[0016] S22, the penalty coefficient and the mode number in the ERIME optimization variational mode decomposition are optimized, and the formula is as follows:

[0017] ;

[0018] Wherein, E represents the decision coefficient of the update mode, Rimepop represents the updated penalty coefficient and mode combination, best_rime represents the current best, that is, the penalty coefficient and mode combination with the minimum sample entropy, represents the upper limit of the value of the penalty coefficient and the mode number, represents the lower limit of the value of the penalty coefficient and the mode number, rand and represent random numbers in (0~1) obeying normal distribution;

[0019] S23, taking the minimum sample entropy as the objective function, the wind power data on the training set and the test set are subjected to variational mode decomposition, the intrinsic mode function component set of the wind power is obtained, each IMF component represents local characteristics in different frequency domains, and the mode number K and the penalty coefficient are constantly optimized through the minimum sample entropy to find the best IMF component, and the smaller the sample entropy, the better the IMF component, and the specific formula is as follows:

[0020] ;

[0021] Wherein, SE represents sample entropy, K represents mode number, represents the penalty coefficient, arg min represents the minimization operation, An index representing the sample entropy.

[0022] The specific steps of the S3 improved fishing algorithm are as follows:

[0023] S31, initialize the population size, the formula is as follows:

[0024] ;

[0025] in, Indicates the The group learning rate and penalty factor are The position of the dimensional search space, Indicates the learning rate and penalty factor in The upper bound of the dimensional search space, Indicates the learning rate and penalty factor in the first The lower bound of the dimensional search space, rand represents a random number in (0~1) that follows a normal distribution;

[0026] S32, the exploration phase is divided into independent search mode and group fishing mode. The capture rate is calculated as follows:

[0027] ;

[0028] in, Indicates the capture rate, EFs indicates the current evaluation number, and MaxEFs indicates the maximum evaluation number;

[0029] S33 uses the differential idea to improve the independent search mode of the fishing algorithm and updates the learning rate and penalty factor of the current position. The specific formula is as follows:

[0030] ;

[0031] ;

[0032] Where Exp is the empirical value parameter, and its value range is (-1~1). Indicates the The worst fitness value after the position update (i.e. the largest error, the larger the worse), Indicates the The best fitness value after the second position change is the minimum root mean square error RMSE, represents the fitness value of the learning rate and penalty factor, Indicates the The position after iterations The fitness value of the corresponding learning rate and penalty factor, Indicates the exploration range, represents the Euclidean distance between the learning rate and the penalty factor, EFs represents the current number of evaluations, and MaxEFs represents the maximum number of evaluations. represents the number of iterations of learning rate and penalty factor, Indicates the The first iteration The group learning rate and penalty factor are The position in the dimensional search space, Indicates the The first iteration The group learning rate and penalty factor are The position in the dimensional search space, Indicates the The position after iterations The learning rate and penalty factor at The position in the dimensional search space, is a random number between (0 and 1) that follows a normal distribution; s represents a d-dimensional random unit vector, which is used to determine the main direction and distance of travel; represents a random number in the range (0~2) that follows a normal distribution. represents a random number in the range (1-3) that follows a normal distribution, and rand represents a random number in the range (0-1) that follows a normal distribution;

[0033] S34, in the simulated group fishing mode, select 3~4 groups of learning rates and penalty factors in sequence to form subgroups , subgroup The discrete Laplace operator is obtained through discrete Laplace transform. The specific formula is as follows:

[0034] ;

[0035] ;

[0036] Where, Indicates subgroup Surrounding target points, Indicates the Subgroups at iteration The The group learning rate and penalty factor are The position in the dimensional search space, Indicates the Subgroups at iteration The The group learning rate and penalty factor are Position in the dimensional search space; Indicates the speed at which fishermen approach the center, with a value range of (0~1). For the moving offset, the value range is (-1~1), decreases gradually with the increase of EFs, EFs represents the current evaluation number, MaxEFs represents the maximum evaluation number, laplace represents a discrete Laplace operator, A represents a matrix for randomly arranging the population, represents Subgroup in the secondary iteration The position of the learning rate and the penalty factor in the subgroup, mean represents the mean value of the parameters, that is The center position of the subgroup;

[0037] S35, in the simulation of the fishing algorithm in the large mining stage, the improved diagonal form is used for updating, and the formula is as follows:

[0038] ;

[0039] ;

[0040] In the formula, Var represents the variance, EFs represents the current evaluation number, MaxEFs represents the maximum evaluation number, represents the position of the first after the first The position of the learning rate and the penalty factor in the subgroup, is the global optimal position, and GD is a Gaussian distribution function, represents a random number obeying a normal distribution in (1~3), which is used to distribute the learning rate and the penalty factor in three ranges; represents the average value matrix of each dimension of the center of the learning rate and the penalty factor, diag (A) represents that the values on the diagonal line after randomly arranging the population are taken as the update condition of this round of iteration, mean represents the mean value of the parameters, and rand represents a random number obeying a normal distribution in (0~1).

[0041] The specific steps of the S5 are as follows:

[0042] S51, the first 70% of the IMF component set is input into the ECFOA-LSTM network model as a training set, and the model is trained;

[0043] S52, the remaining 30% of the IMF component set is input into the trained ECFOA-LSTM network model as a prediction set, and the prediction value corresponding to the IMF component is obtained;

[0044] S53, the prediction values corresponding to all IMF components in the test set are added, and the sum is the wind power prediction value.

[0045] The beneficial effects of the present application are:

[0046] The application breaks through the wind power generation prediction bottleneck of a single LSTM prediction model, and has high prediction accuracy and good economy compared with the existing single LSTM prediction model, and effectively supports reliable prediction of wind power generation.

[0047] The application optimizes VMD based on ERIME, and significantly improves the signal decomposition effect.

[0048] The application adopts ECFOA to optimize the LSTM model, and has better model performance.

[0049] The application adopts a random improved frost optimization algorithm, and enhances the randomness of the algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced below, and obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0051] Figure 1 is a step flow chart of the application.

[0052] Figure 2 is a comparison chart of the prediction results of each model.

[0053] Figure 3 (a) is a comparison chart of the effect of the frost optimization algorithm before and after improvement on the test function F1.

[0054] Figure 3 (b) is a comparison chart of the effect of the frost optimization algorithm before and after improvement on the test function F2.

[0055] Figure 3 (c) is a comparison chart of the effect of the frost optimization algorithm before and after improvement on the test function F3.

[0056] Figure 3 (d) is a comparison chart of the effect of the frost optimization algorithm before and after improvement on the test function F4.

[0057] Figure 4 is a comparison chart of the effect of the frost optimization algorithm before and after improvement on the optimization of VMD sample entropy value.

[0058] Figure 5 (a) is a comparison chart of the effect of the fishing algorithm before and after improvement on the test function F1.

[0059] Figure 5 (b) is a comparison chart of the effect of the fishing algorithm before and after improvement on the test function F2.

[0060] Figure 5 (c) is a comparison chart of the effect of the fishing algorithm before and after improvement on the test function F3.

[0061] Fig. 5 (d) is a comparison chart of the effect of the fish algorithm before and after improvement on the test function F4. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the protection scope of the present application. EMBODIMENT

[0063] The embodiments of the present application provide a short-term wind power prediction method, and a flowchart is shown in Fig. 1. Figure 1 The steps include:

[0064] In step S1, historical wind power data is collected, and the data is preprocessed to construct a short-term wind power prediction data set, and then the short-term wind power prediction data set is divided into a training set and a test set in a quantity ratio of 7:3. The short-term wind power prediction data set includes power generation power, wind speed, wind direction and temperature information, and the sampling frequency is 1 hour (24 data points in a day). Based on the actual minimum and maximum power values of the historical data, the original power data is mapped to the interval [0, 1] by using the minimum-maximum normalization method.

[0065] In step S2, the multi-frequency domain features of the wind power are extracted by using an improved frost and ice optimization algorithm (ERIME), and the specific steps are as follows:

[0066] In S21, the frost and ice optimization algorithm is improved, and the cosine transformation of the frost and ice factor is improved into positive and cosine double transformation to enhance the performance of the algorithm to escape from the local optimal solution and expand the search area of the algorithm, and the specific steps are as follows:

[0067] ;

[0068] wherein, rand represents a random number in (0~1) obeying normal distribution, Riter represents the current iteration number of the frost and ice optimization algorithm, Max_Riter represents the maximum iteration number of the frost and ice optimization algorithm, W represents the parameter of the algorithm, and the value of the present application is 5, and RimeFactor represents the frost and ice factor, which determines the performance of the frost and ice optimization algorithm.

[0069] S22, the penalty coefficient and the mode number in the ERIME optimized variational mode decomposition (VMD) are optimized to improve the accuracy of signal decomposition. In the prior art, the decision coefficient of the frost and ice optimization algorithm update mode is usually determined according to the current iteration number and the maximum iteration number, and the present application selects a random number in 0~1 as the decision coefficient of the update mode. Two independent random numbers rand and , the formula is as follows:

[0070] ;

[0071] Wherein, E represents the decision coefficient of the update mode, Rimepop represents the updated penalty coefficient and mode combination, best_rime represents the current best, that is, the penalty coefficient and mode combination with the minimum sample entropy, represents the upper limit of the penalty coefficient and the mode number, represents the lower limit of the penalty coefficient and the mode number, rand and represents a random number in (0~1) obeying normal distribution;

[0072] The comparison of ERIME and standard RIME on the retest function is shown in Figure 3 (the maximum iteration number is 1000, and the population number is 30), wherein a~d are the functions F1 (Shifted Sphere Function), F2 (Shifted Schwefel's Problem 1.2), F3 (Shifted Rotated High Conditioned Elliptic Function), and F4 (Shifted Schwefel's Problem 1.2 with Noise in Fitness Function) solved by the two algorithms. The left graph is the space surface of the corresponding function, and the right graph is the iteration convergence accuracy of ERIME and standard RIME. The four functions have unimodal convergence, complex variable dependence, rotation inseparability and noise robustness. Through the four functions, the comprehensive performance of the algorithm in ideal conditions, complex coupling and real noise scenes can be evaluated, and key benchmarks for improving algorithm design are provided. As can be seen from Figure 3, the convergence speed of ERIME is faster and the convergence accuracy is higher with the increase of iteration number.

[0073] S23, taking the minimum sample entropy as the objective function, the wind power data on the training set and the test set are subjected to variational mode decomposition to obtain the intrinsic mode function (IMF) component set of the wind power, each IMF component represents the local characteristics of different frequency domains, and the mode number K and the penalty coefficient To find the best IMF component, the smaller the sample entropy, the better the IMF component, and the specific formula is as follows:

[0074]

[0075] Wherein, SE represents sample entropy, K represents modal number, represents penalty coefficient, arg min represents minimum operation, represents the index of sample entropy.

[0076] The ERIME and RIME optimization VMD results are as shown in Figure 2 It can be seen from Figure 2 that, compared with RIME, the ERIME of the application has better optimization capability, and maintains better capability in 20 iterations, and the sample entropy is smaller than the solution result of RIME.

[0077] Step S3, construct an LSTM model, and use an improved fishing algorithm (ECFOA) to find the optimal learning rate and penalty factor to improve the convergence speed and generalization performance of the model, avoid the learning rate and penalty factor being too large or too small, leading to gradient explosion or gradient disappearance, and thus leading to distortion of predicted data, thereby leading to unusable prediction results. The LSTM model is a known model in the prior art, including an input layer, an LSTM layer and an output layer, the input layer is used to receive a first data set, the LSTM layer is used to solve long-term dependence in the data set, and the output layer is used to predict short-term wind power. The specific steps of the improved fishing algorithm are as follows:

[0078] S31, initialize the population size, and the formula is as follows:

[0079]

[0080] Wherein, represents the position of the th group of learning rate and penalty factor in the th dimension search space, represents the upper bound of the learning rate and penalty factor in the dimension search space, represents the lower bound of the learning rate and penalty factor in the th dimension search space, and rand represents a random number obeying normal distribution in (0~1).

[0081] The fishing optimization algorithm is divided into an exploration stage and a development stage. In the exploration stage, the fishing optimization algorithm performs global search, and when entering the development stage, the fishing optimization algorithm performs local search.

[0082] ​​S32, the exploration phase is divided into independent search mode and group fishing mode, the conversion between independent search mode and group fishing mode is based on the capture rate The calculation formula for capture rate is as follows:

[0083] ;

[0084] in, Indicates the capture rate, EFs indicates the current evaluation number, and MaxEFs indicates the maximum evaluation number.

[0085] When the capture rate When it approaches 1, fishermen adopt independent search mode. When it approaches 0, fishermen adopt a group fishing mode and use random numbers To simulate, When using independent search mode, When fishing, group fishing mode is adopted.

[0086] S33 uses the differential idea to improve the independent search mode of the fishing algorithm and updates the learning rate and penalty factor of the current position. The specific formula is as follows:

[0087] ;

[0088] ;

[0089] Where Exp is the empirical value parameter, and its value range is (-1,1). Indicates the The worst fitness value after the position update (i.e. the largest error, the larger the worse), Indicates the The best fitness value after the second position change is the minimum root mean square error RMSE, represents the fitness value of the learning rate and penalty factor, Indicates the The position after iterations The fitness value of the corresponding learning rate and penalty factor, Indicates the exploration range, represents the Euclidean distance between the learning rate and the penalty factor, EFs represents the current number of evaluations, and MaxEFs represents the maximum number of evaluations. represents the number of iterations of learning rate and penalty factor, Indicates the The first iteration The group learning rate and penalty factor are The position in the dimensional search space, Indicates the The first iteration The group learning rate and penalty factor are The position in the dimensional search space, Indicates the The position after iterations The learning rate and penalty factor at The position in the dimensional search space, is a random number between (0 and 1) that follows a normal distribution; s represents a d-dimensional random unit vector, which is used to determine the main direction and distance of travel; represents a random number in the range (0~2) that follows a normal distribution. represents a random number in the range (1 to 3) that follows a normal distribution, and rand represents a random number in the range (0 to 1) that follows a normal distribution.

[0090] S34, in the simulated group fishing mode, select 3~4 groups of learning rates and penalty factors in sequence to form subgroups , subgroup The discrete Laplace operator is obtained through discrete Laplace transform, and then a new position is obtained, which enables the algorithm to escape from the local optimum and expand the search range. The specific formula is as follows:

[0091] ;

[0092] Where, Indicates subgroup Surrounding target points, Indicates the Subgroups at iteration The The group learning rate and penalty factor are The position in the dimensional search space, Indicates the Subgroups at iteration The The group learning rate and penalty factor are Position in the dimensional search space; Indicates the speed at which fishermen approach the center, with a value range of (0~1). is the offset of movement, the value range is (-1~1), As EFs increases, it gradually decreases; EFs represents the current number of evaluations, MaxEFs represents the maximum number of evaluations, laplace represents the discrete Laplace operator, and A represents the matrix that randomly arranges the population. express Subgroups at iteration The position of the learning rate and penalty factor in the mean represents the mean of the parameters, that is, The center position of the group.

[0093] S35, in the simulation of the fishing algorithm in the large mining stage, the improved diagonal form is used for updating, so as to enhance the memory ability of the algorithm and speed up the convergence of the algorithm, and the formula is as follows:

[0094] ;

[0095] In the formula, Var represents variance, EFs represents the current evaluation number, MaxEFs represents the maximum evaluation number, represents the first iteration, the first group of learning rate and penalty factor position, is the global optimal position, and GD is a Gaussian distribution function, represents a random number in (1~3) subject to normal distribution, which is used to distribute the learning rate and the penalty factor in three ranges; represents the average value matrix of each dimension of the learning rate and the penalty factor center, diag (A) represents that the values on the diagonal line after randomly arranging the population are taken as the update condition of this round of iteration, mean represents the mean value of the parameters, and rand represents a random number in (0~1) subject to normal distribution.

[0096] Under the experimental setting of the maximum iteration number 1000 and the population size 30, the performance of the ECFOA of the application and the standard CFOA on the test function is shown in Figure 5, and four typical functions are compared in the figure, wherein the left graph of a~d shows the spatial surface form of each function, and the right graph presents the iteration convergence curves of the two algorithms. The experimental results show that, with the increase of the iteration number, the ECFOA exhibits faster convergence speed and higher convergence accuracy compared with the standard CFOA. These test functions simulate ideal conditions, variable coupling, nonlinear correlation and noise interference, and systematically verify the comprehensive optimization ability of the algorithm in a multi-dimensional complex environment, which provides a key benchmark basis for algorithm improvement.

[0097] Step S4, based on the improved fishing algorithm of S3, the optimal learning rate and penalty factor of the LSTM model are determined through continuous optimization, and an ECFOA-LSTM network model is obtained. The learning rate and penalty factor optimized by the ECFOA can more efficiently determine the optimal parameter combination of the ECFOA-LSTM model in the data set compared with artificial parameter adjustment, and the parameters optimized by the improved fishing algorithm of the application can more accurately capture the mapping relationship between the input and the output compared with the standard fishing algorithm.

[0098] Step S5, based on the IMF component set obtained in S2, the wind power prediction value is obtained, and the specific steps are as follows:

[0099] S51, input the first 70% of the IMF component set as a training set into the ECFOA-LSTM network model, and train the model;

[0100] S52, input the remaining 30% of the IMF component set as a prediction set into the trained ECFOA-LSTM network model, and obtain the predicted value corresponding to the IMF component;

[0101] S53, add the predicted values corresponding to all IMF components in the test set, and the sum is the wind power prediction value.

[0102] The ice and frost optimization algorithm is improved, and is combined with variational mode decomposition, so that the original data can be more accurately and efficiently extracted and decomposed and preprocessed, and noise interference can be effectively removed. In addition, the improved catch fish algorithm can deeply optimize the hyperparameters of the LSTM model, so that the network structure of the LSTM model is more reasonable, the learning ability is stronger, the adaptability and processing capacity of the model to complex data are improved, and the prediction accuracy and stability are also improved. Compared with a single model or a traditional combined model, the model has better performance and generalization ability in time series prediction and other tasks.

[0103] Experimental verification

[0104] The prediction effect of the present application and the existing technology of back propagation neural network (Back Propagation Neural Network, BPNN), support vector machine (Support Vector Machine, SVM), long short-term memory network (Long Short-Term Memory, LSTM), catch fish optimization algorithm-long short-term memory network (Catch Fish Optimization Algorithm-Long Short-Term Memory, CFOA-LSTM), ice and frost optimization algorithm-variational mode decomposition-catch fish optimization algorithm-long short-term memory network (Robust Ice-cold Metaheuristic Algorithm-Variational Mode Decomposition-Catch Fish Optimization Algorithm-Long Short-Term Memory, RIME-VMD-CFOA-LSTM) is as follows Figure 2As shown in the comparison results shown in Table 1, it can be found that the root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) of the present application are all smaller than those of the single LSTM and CFOA-LSTM, and therefore, the model of the present application performs better than other models. Taking MAPE as an example, compared with BPNN, the MAPE of the present application is reduced by 43.37%; compared with SVM, the MAPE of the present application is reduced by 31.58%; compared with LSTM, the MAPE of the present application is reduced by 3.92%; compared with CFOA-LSTM, the MAPE of the present application is reduced by 1.43%; and compared with RIME-VMD-CFOA-LSTM, the MAPE of the present application is reduced by 0.96%.

[0105] Table 1 Comparison of indicators of various models

[0106]

[0107] The ERIME of the present application and the RIME of the prior art are compared on the CEC2005 (Congress on Evolutionary Computation 2005 Benchmark Functions) test function set, and the results are shown in Figure 3. As can be seen from Figure 3, the convergence speed of the present application is faster, and the optimization accuracy is higher. The ECFOA of the present application and the CFOA of the prior art are also compared on the CEC2005 test function set, and the results are shown in Figure 4. As can be seen from Figure 4, the convergence speed of the present application is faster, and the optimization accuracy is higher. Figure 4 Figure 4

[0108] Each of the embodiments in the specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment mainly explains the difference from other embodiments. Especially, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the related parts can be referred to the part of the method embodiment.

[0109] The above only describes the preferred embodiments of the present application, and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application is included in the protection scope of the present application.​​

Claims

1. A short-term wind power prediction method, characterized in that the steps include: Step S1, collecting historical wind power data and preprocessing the data; Step S2: improving the frost and ice optimization algorithm to extract multi-frequency domain characteristics of wind power; Step S3, building an LSTM model and improving the fishing algorithm; Step S4, based on S3, determining the optimal learning rate and penalty factor of the LSTM model to obtain the ECFOA-LSTM network model; Step S5, obtaining a wind power forecast value based on the IMF component set obtained in S2; The specific steps of S2 are as follows: S21, improve the frost optimization algorithm, and improve the cosine transform of the frost factor RimeFactor to a double sine and cosine transform, as follows: Where rand represents a random number between 0 and 1 that follows a normal distribution, Riter represents the current iteration number of the frost optimization algorithm, Max_Riter represents the maximum iteration number of the frost optimization algorithm, W represents the algorithm parameter, and its value is 5. RimeFactor represents the frost factor, which determines the performance of the frost optimization algorithm. S22, the penalty coefficient and mode number in variational mode decomposition are optimized by ERIME, and the formula is as follows: E=rand Where E represents the determination coefficient of the update method, Rimepop represents the updated penalty coefficient and mode combination, best_rime represents the current best penalty coefficient and mode combination with the smallest sample entropy, ub represents the upper limit of the penalty coefficient and mode number, lb represents the lower limit of the penalty coefficient and mode number, rand and rand′ represent random numbers between 0 and 1 that follow the normal distribution; S23, with the minimum sample entropy as the objective function, performs variational modal decomposition on the wind power data in the training set and the test set to obtain the intrinsic mode function component set of the wind power. Each IMF component represents the local characteristics of different frequency domains. The modal number K and the penalty coefficient α1 are continuously optimized through the minimum sample entropy to find the optimal IMF component. The smaller the sample entropy, the better the IMF component. The specific formula is as follows: Among them, SE represents sample entropy, K represents the number of modes, α1 represents the penalty coefficient, argmin represents the minimization operation, and v represents the index of sample entropy; The specific steps of the S3 improved fishing algorithm are as follows: S31, initialize the population size, the formula is as follows: Fisher ij =(ub j -lb j )×rand+lb j Among them, Fisher ij Indicates the position of the i-th group of learning rate and penalty factor in the j-dimensional search space, ub j Represents the upper bound of the learning rate and penalty factor in the j-dimensional search space, lb j represents the lower bound of the learning rate and penalty factor in the j-th dimension search space, and rand represents a random number from 0 to 1 that follows a normal distribution; S32, the exploration phase is divided into independent search mode and group fishing mode. The capture rate is calculated as follows: Among them, α2 represents the capture rate, EFs represents the current evaluation number, and MaxEFs represents the maximum evaluation number; S33 uses the differential idea to improve the independent search mode of the fishing algorithm and updates the learning rate and penalty factor of the current position. The specific formula is as follows: Where Exp is the empirical value parameter, ranging from -1 to 1, fit max Indicates the worst fitness value after the tth position update, fit min Indicates the best fitness value after the t-th position change, that is, the minimum root mean square error RMSE, fit i Represents the fitness value of the learning rate and penalty factor, fit pos Represents the fitness value of the learning rate and penalty factor corresponding to the position pos after the Tth iteration, R represents the exploration range, Dis represents the Euclidean distance between the learning rate and the penalty factor, EFs represents the current number of evaluations, MaxEFs represents the maximum number of evaluations, and T represents the number of iterations of the learning rate and penalty factor. represents the position of the i-th group of learning rate and penalty factor in the j-th dimension search space at the T-th iteration, Indicates the position of the i-th group of learning rate and penalty factor in the j-th dimension search space at the T+1th iteration, Represents the position of the learning rate and penalty factor at position pos after the Tth iteration in the j-dimensional search space, r s is a random number between 0 and 1 that follows a normal distribution; s represents a d-dimensional random unit vector, which is used to determine the main direction and distance of travel; rr1 represents a random number between 0 and 2 that follows a normal distribution, rr2 represents a random number between 1 and 3 that follows a normal distribution, and rand represents a random number between 0 and 1 that follows a normal distribution; S34, in the simulated group fishing mode, 3 to 4 groups of learning rates and penalty factors are sequentially selected to form a subgroup c. The discrete Laplace operator of the subgroup c is obtained by discrete Laplace transform. The specific formula is as follows: laplace=A(m+1,n)+A(m-1,n)+A(m,n+1)+A(m,n-1)-4A(m,n) In the formula, Centre c represents the target point surrounded by subgroup c, Indicates the position of the e-th group of learning rate and penalty factor in the j-th dimension search space at the T+1th iteration, Indicates the position of the e-th group learning rate and penalty factor in the j-th dimension search space in the T-th iteration; r2 represents the speed at which fishermen approach the center, ranging from 0 to 1, r3 is the offset of the movement, ranging from -1 to 1, and r3 gradually decreases with the increase of EFs; EFs represents the current number of evaluations, MaxEFs represents the maximum number of evaluations, laplace represents the discrete Plasma operator, and A represents the matrix that randomly arranges the population. Represents the position of the learning rate and penalty factor in subgroup c at T iterations, and mean represents the mean of the parameters, that is, the center position of group c; S35, in the simulated large-scale mining phase of the fishing algorithm, uses the improved diagonal form for updating, and the formula is as follows: Where σ represents the variance, EFs represents the current number of evaluations, MaxEFs represents the maximum number of evaluations, and Fisher i T+1 represents the position of the i-th group of learning rate and penalty factor after the T+1th iteration, Gbest is the global optimal position, GD is the Gaussian distribution function, rr2 represents a random number from 1 to 3 that obeys the normal distribution, which is used to distribute the learning rate and penalty factor in three ranges; Fisher represents the average value matrix of each dimension of the learning rate and penalty factor center, diag(A) represents the update condition of this round of iteration after randomly permuting the population and taking the value on the diagonal, mean represents the mean of the parameter, and rand represents a random number from 0 to 1 that obeys the normal distribution.

2. A short-term wind power prediction method according to claim 1, characterized in that: The specific steps of S5 are as follows: S51, input the first 70% of the IMF component set as a training set into the ECFOA-LSTM network model to train the model; S52, inputting the remaining 30% of the IMF components of the IMF component set into the trained ECFOA-LSTM network model as a prediction set to obtain the prediction values ​​corresponding to the IMF components; S53, adding up the predicted values ​​corresponding to all IMF components of the test set, and the sum is the wind power prediction value.

Images