Short-term photovoltaic power prediction method for optimizing kernel extreme learning machine based on improved dung beetle algorithm

The improved dung beetle algorithm optimized KELM method enhances photovoltaic power prediction accuracy by addressing weather variability and heuristic algorithm limitations, achieving faster convergence and improved prediction performance.

CN119067247APending Publication Date: 2024-12-03CHINA THREE GORGES UNIV
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410922054.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-10
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing photovoltaic power prediction models face challenges due to varying weather conditions, poor parameter selection, and heuristic algorithms prone to local optimality, affecting prediction accuracy and convergence.

Method used

A short-term photovoltaic power prediction method using an improved dung beetle algorithm optimized kernel extreme learning machine (KELM) that includes data preprocessing, variational mode decomposition, and optimized parameter selection through IDBO to enhance convergence and accuracy.

Benefits of technology

The method improves prediction accuracy by addressing environmental influences, data normalization, and optimizing model parameters, leading to faster convergence and better prediction results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067247A_ABST
    Figure CN119067247A_ABST
Patent Text Reader

Abstract

The invention discloses a short-term photovoltaic power prediction method based on an improved dung beetle algorithm optimized kernel extreme learning machine, and the method comprises the steps: firstly obtaining the historical power data of a photovoltaic power station and the meteorological factor data corresponding to a photovoltaic array region, and carrying out the processing of the data; secondly, decomposing the power time sequence data through a variational mode decomposition method VMD, and forming a data set with the meteorological factor data; thirdly, constructing a kernel extreme learning machine KELM model; and finally, optimizing KELM parameters of a kernel extreme learning machine by using an improved dung beetle algorithm IDBO, and predicting the power. By improving the dung beetle algorithm, the defect that the dung beetle algorithm is prone to falling into local optimum can be overcome, the convergence speed is increased, and the photovoltaic power prediction precision is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of photovoltaic power generation output power prediction, and in particular relates to a short-term photovoltaic power prediction method based on an improved dung beetle algorithm optimized kernel extreme learning machine. Background Art

[0002] With the rapid development of photovoltaic power generation technology in the power industry, photovoltaic power generation is affected by environmental factors, which makes photovoltaic power generation intermittent and has greater volatility and randomness. The large-scale grid connection of photovoltaic power stations brings challenges to the power grid, and also tests the safety and economic operation of the power system. By accurately predicting the short-term output power of photovoltaic power generation systems, it is beneficial for the power grid to better formulate and adjust scheduling plans, improve the safety and anti-interference of the power system, which is crucial to improving the safety and stability of the power system. Through the research on photovoltaic power prediction at home and abroad,

[0003] There are still some problems with the artificial intelligence algorithms in the existing technologies in photovoltaic power generation prediction. For example, the photovoltaic power generation varies greatly under different weather conditions, and the prediction effect is not good; the photovoltaic power prediction model has problems with parameter selection, which affects the prediction accuracy; the heuristic algorithm is prone to falling into local optimality, which ultimately affects convergence and finally affects the prediction accuracy. Summary of the invention

[0004] In order to overcome the large differences in photovoltaic power generation under different weather conditions, the poor prediction effect, the difficulty in selecting parameters for photovoltaic power prediction models, the heuristic algorithm is prone to fall into local optimality, which ultimately affects convergence and finally affects the prediction accuracy. The present invention provides a short-term photovoltaic power prediction method based on an improved dung beetle algorithm to optimize a kernel extreme learning machine. The method can make up for the disadvantage that the dung beetle algorithm is prone to fall into local optimality by improving the dung beetle algorithm, accelerate the convergence speed, and improve the accuracy of photovoltaic power prediction.

[0005] The technical solution adopted by the present invention is:

[0006] The short-term photovoltaic power prediction method based on the improved dung beetle algorithm and the optimized nuclear extreme learning machine includes the following steps:

[0007] Step 1: Obtain historical photovoltaic power generation data of the photovoltaic power station and corresponding meteorological factor data of the photovoltaic array area, and process the data;

[0008] Step 2: Decompose the photovoltaic power data by using the variational mode decomposition method VMD, and combine it with the meteorological factor data to form a data set;

[0009] Step 3: Build the kernel extreme learning machine KELM model;

[0010] Step 4: Use the improved dung beetle algorithm IDBO to optimize the kernel extreme learning machine KELM model parameters and predict the short-term photovoltaic power.

[0011] In step 1, the collection time period of the photovoltaic power generation data measured by the photovoltaic power station is set to 7:00 am to 18:00 pm, and the collection interval is 5 minutes; the variables included in the data at each moment include the total horizontal irradiance, scattered horizontal irradiance, temperature, relative humidity, and wind speed meteorological factor data from 7:00 am to 18:00 pm.

[0012] In step 1, data processing includes the following steps:

[0013] S1.1: Clean the abnormal data and make corrections using formula (1):

[0014] P = ηSI[1-β(T c +T0)] (1);

[0015] In formula (1), P is the photovoltaic power generation, η is the conversion efficiency, S is the light receiving area of ​​the photovoltaic array, I is the irradiance, and T c is the component temperature, β is the temperature rise coefficient, and T0 is a constant;

[0016] S1.2: Normalize the data:

[0017]

[0018] In formula (2), x′ is the value of the sample after normalization, x is the value of the initial sample, and x max is the maximum value of the column data, x min The minimum value of the column data.

[0019] In step 2, the constrained variational model of the variational mode decomposition method VMD is:

[0020]

[0021] In formula (3), K is the total number of modal classifications, is the partial derivative operation, δ(t) is the unit pulse function, u k (t) is the modal component, w k is the center frequency, k is the serial number of the current modal component, e is the natural logarithm, f(t) is the original sequence, and j is the imaginary unit;

[0022] By using formula (3), the photovoltaic power generation is decomposed to obtain k IMF components. Combined with the meteorological factor data, the data set input into the kernel extreme learning machine KELM model is:

[0023]

[0024] In formula (4), x t1 、x t2 ,...x tm They represent the values ​​of total horizontal irradiance, diffuse horizontal irradiance, temperature, relative humidity, and wind speed at time t, respectively, m = 5; Represents the value of the i-th IMF component at time t, where i is between [1, k].

[0025] In step 3, the expression of the kernel extreme learning machine KELM model is:

[0026]

[0027] In formula (5), y(x) is the output value, H is the hidden layer output matrix, and H T is the transposed matrix of H, C is the regularization coefficient, I is the identity matrix, T is the target output matrix, x is the input vector, x1, x M are training samples, K(x,x1), K(x,x M ) correspond to x and x1, x and x respectively. M Gaussian kernel function;

[0028] The Gaussian kernel function is shown in formula (6):

[0029]

[0030] In formula (6), K(x i ,x j ) represents x i With x j Gaussian kernel function, σ is the kernel parameter, x i 、x j Input vector for the experiment.

[0031] In step 4, the improved dung beetle algorithm IDBO is used to optimize the kernel extreme learning machine KELM model parameters to predict short-term photovoltaic power. The model parameters include the kernel parameter σ and the regularization coefficient C, and the following steps are included:

[0032] S4.1: Set initial values: population size n, maximum number of iterations T max , population dimension dim, kernel parameter σ and regularization coefficient C;

[0033] S4.2: The fitness function of the population is the mean square error function of the KELM model, and its expression is:

[0034]

[0035] In formula (7), f(σ,C) is the fitness function, σ is the kernel parameter, C is the regularization coefficient, N is the total number of samples, i is the current sample number, yi is the i-th predicted value, is the i-th true value;

[0036] S4.3: The population is initialized using Chebyshev chaotic mapping, and the iteration method is as follows:

[0037] x n+1 =cos(karccosx n ) (8);

[0038] In formula (8), k is the order, x n ∈[-1,1],x n is the nth generation population; x n+1 is the n+1 generation population;

[0039] S4.4: Introduce the golden sine strategy to improve the position update of the dung beetle rolling behavior:

[0040] x i (t+1)=x i (t)×|sin(R1)|+R2×sin(R1)×|c1×X b -c2×x i (t)| (9);

[0041]

[0042] In the formula, x i (t) represents the current position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the i-th dung beetle at the t+1th iteration, R1 is a random number in [0,2π], R2 is a random number in [0,π], X b represents the global optimal position; c1 and c2 are the coefficients containing the golden section, is the golden ratio number;

[0043] S4.5: When the dung beetle's ball rolling behavior is blocked, it performs dancing behavior and updates its position:

[0044] x i (t+1)=x i (t)+tan(θ)|x i (t)-x i (t-1)| (11);

[0045] In formula (11), x i (t) represents the position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the i-th dung beetle at the t+1th iteration, x i(t-1) represents the position of the i-th dung beetle at the t-1th iteration, θ is the deflection angle, θ∈[0,π];

[0046] S4.6: Select the boundaries of the area where dung beetles breed and lay eggs:

[0047]

[0048] In formula (12), X * is the current optimal position, Ub and Lb are the upper and lower boundaries of the optimization problem of the dung beetle breeding and egg-laying area, respectively. * and Lb * are the upper and lower boundaries of the dung beetle breeding and egg-laying area; R = 1-t / T max , t is the current iteration number, T max is the maximum number of iterations;

[0049] S4.7: Introduce the spiral search strategy of the whale optimization algorithm to update the position of the hatching ball:

[0050] B i (t+1)=X * +b1×(B i (t)-Lb * )×e hl ×cos(2πl)+b2×(B i (t)-Ub * )×e hl ×cos(2πl)(13);

[0051] In formula (13), X * represents the current optimal position, b1 and b2 represent random vectors of size 1×D; D represents the dimension of the optimization problem; Ub * and Lb * The upper and lower boundaries of the dung beetle breeding and egg-laying area; B i (t+1) represents the position information of the t+1th iteration of the i-th hatching ball; B i (t) represents the position information of the t-th iteration of the i-th hatching ball; e is the natural logarithm, h is a constant; l is a number between [-1,1];

[0052] S4.8: Boundaries of the foraging area of ​​hatched dung beetles:

[0053]

[0054] In formula (14), X b represents the global optimal position, Lb b and Ub b They represent the lower and upper bounds of the optimal foraging area of ​​the dung beetle, R = 1-t / Tmax , Lb and Ub represent the lower and upper bounds of the foraging area of ​​the small dung beetle, respectively;

[0055] S4.9: Introducing adaptive weights to update the foraging position of dung beetles:

[0056] x i (t+1)=x i (t)+C1×k1×(x i (t)-Lb b )+C2×k2×(x i (t)-Ub b ) (15);

[0057] In formula (15), t is the number of iterations, T max is the maximum number of iterations, x i (t) represents the position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the t+1th iteration of the i-th dung beetle, C1 is a random number that follows a normal distribution, C2 is a random number between (0,1), and Lb b and Ub b They represent the lower and upper bounds of the optimal foraging area of ​​the small dung beetle, respectively;

[0058] S4.10: Update the location of the dung beetle in its fight for food:

[0059] x i (t+1)=X b +S×g×(|x i (t)-X * |+|x i (t)-X b |) (16);

[0060] In formula (16), X b represents the best position for competing for food; S represents a constant value; g is a random vector of size 1×D and obeys a normal distribution; X b represents the global optimal position; x i (t) represents the position of the i-th dung beetle at the t-th iteration; x i (t+1) represents the position of the i-th dung beetle at the t+1th iteration;

[0061] S4.11: Determine whether the algorithm has reached the maximum number of iterations. If so, proceed to step S4.11, otherwise return to S4.3;

[0062] S4.12: The IDBO algorithm ends and outputs the optimized kernel parameter g and kernel regularization coefficient C;

[0063] S4.13: Input the data set into KELM to obtain k prediction sequences, sum and reconstruct the k prediction sequences and perform denormalization to obtain the photovoltaic power prediction value.

[0064] The present invention is a short-term photovoltaic power prediction method based on an improved dung beetle algorithm to optimize a nuclear extreme learning machine, and the technical effects are as follows:

[0065] 1) In step 1 of the present invention, the influence of environmental factors on photovoltaic power generation is taken into account, the abnormal data is cleaned and corrected, and the data is normalized, which is conducive to improving the convergence speed of the model and the accuracy of the prediction results;

[0066] 2) In step 2 of the present invention, the power signal is decomposed by variational mode decomposition to obtain several sub-components with strong regularity, which can fully explore the intrinsic connection between photovoltaic power sequences, and combine the correlation between environmental factors and photovoltaic power sequences, which is conducive to the model's learning of photovoltaic power and improving the accuracy of prediction;

[0067] 3) In step 3 of the present invention, the kernel extreme learning machine uses kernel mapping instead of random mapping on the basis of the extreme learning machine, which can effectively improve the problem of decreased generalization and stability caused by random assignment of hidden layer neurons, and is conducive to improving the accuracy of prediction;

[0068] 4) In step 4 of the present invention, in order to solve the problems of low convergence accuracy and easy falling into local optimal solution of the dung beetle algorithm, the population is initialized by introducing Chebyshev chaotic mapping, so that the distribution of the population in the search space is more uniform, the golden sine strategy is introduced to improve the position update of the dung beetle rolling ball behavior, the spiral search strategy of the whale optimization algorithm is introduced to update the position of the hatching ball, and the adaptive weight is introduced to update the foraging position of the small dung beetle to improve the dung beetle algorithm, which is beneficial to improve the accuracy of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0070] Figure 1 It is a flow chart of the present invention.

[0071] Figure 2 Flowchart for improving the dung beetle algorithm.

[0072] Figure 3 A fitness comparison chart between the improved dung beetle algorithm and the dung beetle algorithm.

[0073] Figure 4 This is a diagram of the prediction results of the present invention and different models. DETAILED DESCRIPTION

[0074] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0075] The short-term photovoltaic power prediction method based on the improved dung beetle algorithm and the optimized nuclear extreme learning machine includes the following steps:

[0076] Step S1: Obtain historical power data of the photovoltaic power station and meteorological factor data corresponding to the photovoltaic array area, and process the data;

[0077] Step S2: Decompose the photovoltaic power data by using the variational mode decomposition method VMD, and form a data set with the meteorological factor data;

[0078] Step S3: construct a kernel extreme learning machine KELM model;

[0079] Step S4: Use the improved dung beetle algorithm IDBO to optimize the kernel extreme learning machine KELM model parameters and predict the short-term photovoltaic power.

[0080] In step S1, data from a photovoltaic power station in a certain place in March, June, September and December of a certain year are selected. The data include measured photovoltaic power generation data and meteorological factor data corresponding to the photovoltaic array area. The collection time period is set to 7:00 am to 18:00 pm, and the collection interval is 5 minutes. There are 16,226 data in total, each of which includes total horizontal irradiance, scattered horizontal irradiance, temperature, relative humidity, wind speed and photovoltaic power generation, which constitute the measured sample data set for photovoltaic power prediction. The meteorological factor data and photovoltaic power generation data from 10:00 to 10:30 on March 12 are selected, as shown in Table 1.

[0081] Table 1 Meteorological factors and photovoltaic power generation data from 10:00 to 10:30 on March 12

[0082]

[0083]

[0084] Step S101, clean the abnormal data and make corrections through formula (1):

[0085] P = ηSI[1-β(T c +T0)] (1);

[0086] In formula (1), P is the photovoltaic power generation, η is the conversion efficiency, S is the light receiving area of ​​the photovoltaic array, I is the irradiance, and T c is the component temperature, β is the temperature rise coefficient, and T0 is a constant;

[0087] Step S102, normalize the data:

[0088]

[0089] In formula (2), x′ is the value of the sample after normalization, x is the value of the initial sample, and x max is the maximum value of the column data, x min The minimum value of the column data;

[0090] Step S2, using the variational mode decomposition method VMD to decompose the photovoltaic power generation series data. The constrained variational model of the VMD decomposition algorithm is shown in formula (3). The number of decompositions k=6 is set to obtain 6 IMF components. Combined with the meteorological factor data, a data set is formed. The data set is divided into a training set and a test set in a ratio of 8:2.

[0091]

[0092] In formula (3), K is the total number of modal classifications, is the partial derivative operation, δ(t) is the unit pulse function, u k (t) is the modal component, w k is the center frequency, k is the serial number of the current modal component, e is the natural logarithm, f(t) is the original sequence, and j is the imaginary unit;

[0093] By using formula (3), the photovoltaic power generation is decomposed into k = 6 IMF components. Combined with the meteorological factor data, the data set input into the kernel extreme learning machine model is:

[0094]

[0095] In formula (4), x t1 、x t2 ,...x tm They represent the total horizontal irradiance, diffuse horizontal irradiance, temperature, relative humidity, and wind speed at time t, respectively. m = 5. represents the value of the i-th IMF component at time t, where i is between [1, k];

[0096] Step S3: Construct the KELM model:

[0097]

[0098] In formula (5), y(x) is the output value, H is the hidden layer output matrix, and H T is the transposed matrix of H, C is the regularization coefficient, I is the identity matrix, T is the target output matrix, x is the input vector, x1, x M are the test input vectors, K(x,x1), K(x,x M ) correspond to x and x1, x and x respectively. M Gaussian kernel function;

[0099] The Gaussian kernel function is shown in formula (6):

[0100]

[0101] In formula (6), K(x i ,x j ) represents x i With x j Gaussian kernel function, σ is the kernel parameter, x i 、x j Input vector for the experiment.

[0102] Step S4, using the improved dung beetle algorithm IDBO to optimize the kernel extreme learning machine KELM parameters, the parameters are the kernel parameter σ and the regularization coefficient C, including:

[0103] Step S401, set initial values: population size n = 30, maximum number of iterations T max =15, population dimension dim=2, kernel parameter σ=30 and regularization coefficient C=4;

[0104] Step S402: The fitness function of the population is the mean square error function of the KELM model, and its expression is:

[0105]

[0106] In formula (7), f(σ,C) is the fitness function, σ is the kernel parameter, C is the regularization coefficient, N is the total number of samples, i is the current sample number, y i is the i-th predicted value, is the i-th true value;

[0107] Step S403, using Chebyshev chaotic mapping to initialize the population, the iteration method is as follows:

[0108] x n+1 =cos(karccosx n ) (8);

[0109] In formula (8), k = 4, x n ∈[-1,1],x n is the nth generation population, x n+1 is the n+1 generation population;

[0110] Step S404, introduce the golden sine strategy to improve the position update of the dung beetle rolling behavior:

[0111] x i (t+1)=x i (t)×|sin(R1)|+R2×sin(R1)×|c1×Xb -c2×x i (t)| (9);

[0112]

[0113] In the formula, x i (t) represents the current position of the dung beetle, x i (t+1) represents the position of the dung beetle in the next iteration, R1 is a random number in [0,2π], R2 is a random number in [0,π], and X b represents the global optimal position, c1 and c2 are the coefficients containing the golden section, is the golden ratio number;

[0114] Step S405, when the dung beetle's ball rolling behavior is blocked, it performs dancing behavior and updates its position:

[0115] In formula (9), x i (t) represents the current position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the ith dung beetle at the t+1th iteration, R1 is a random number in [0,2π], R2 is a random number in [0,π], and in formula (10), c1 and c2 are coefficients containing the golden section. is the golden ratio number;

[0116] Step S405, when the dung beetle's ball rolling behavior is blocked, it performs dancing behavior and updates its position:

[0117] x i (t+1)=x i (t)+tan(θ)|x i (t)-x i (t-1)|(11);

[0118] In formula (11), x i (t) represents the current position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the i-th dung beetle at the t+1th iteration, x i (t-1) represents the position of the i-th dung beetle at the t-1th iteration, θ is the deflection angle, θ∈[0,π];

[0119] Step S406, selecting the boundary of the area where dung beetles breed and lay eggs:

[0120]

[0121] In formula (12), X * is the current optimal position, Ub and Lb are the upper and lower boundaries of the optimization problem of the dung beetle breeding and egg-laying area, Ub * and Lb* are the upper and lower boundaries of the dung beetle breeding and egg-laying area, R = 1-t / T max , t is the current iteration number, T max is the maximum number of iterations;

[0122] Step S407, introduce the spiral search strategy of the whale optimization algorithm to update the position of the hatching ball:

[0123] B i (t+1)=X * +b1×(B i (t)-Lb * )×e hl ×cos(2πl)+b2×(B i (t)-Ub * )×e hl ×cos(2πl)(13);

[0124] In formula (13), X * represents the current best position, b1 and b2 represent random vectors of size 1×D, D represents the dimension of the optimization problem, D=2, Ub * and Lb * B is the upper and lower boundaries of the dung beetle breeding and egg-laying area. i (t+1) represents the position information of the ith hatching ball at the t+1th iteration, B i (t) represents the position information of the t-th iteration of the i-th hatching ball, e is the natural logarithm, h is a constant, and l is a number between [-1,1];

[0125] Step S408, the boundary of the foraging area of ​​the hatched dung beetle:

[0126]

[0127] In formula (14), X b represents the global optimal position, Lb b and Ub b They represent the lower and upper bounds of the optimal foraging area of ​​the dung beetle, R = 1-t / T max , Lb and Ub represent the lower and upper bounds of the foraging area of ​​the small dung beetle, respectively;

[0128] Step S409, introducing adaptive weights to update the foraging position of the dung beetle:

[0129] x i (t+1)=x i (t)+C1×k1×(x i (t)-Lb b )+C2×k2×(x i (t)-Ubb ) (15);

[0130] In formula (15), t is the number of iterations, T max is the maximum number of iterations, x i (t) represents the position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the t+1th iteration of the i-th dung beetle, C1 is a random number that follows a normal distribution, C2 is a random number between (0,1), and Lb b and Ub b They represent the lower and upper bounds of the optimal foraging area of ​​the small dung beetle, respectively;

[0131] Step S410, updating the position of the dung beetle in the behavior of competing for food:

[0132] x i (t+1)=X b +S×g×(|x i (t)-X * |+|x i (t)-X b |) (16);

[0133] In formula (16), X b represents the best position for competing for food, S represents a constant value, g is a random vector of size 1×D and obeys a normal distribution, D=2, X b represents the global optimal position, x i (t) represents the position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the i-th dung beetle at the t+1th iteration;

[0134] Step S411, determine whether the algorithm has reached the maximum number of iterations, if so, proceed to step S412, otherwise return to S403;

[0135] Step S412, the IDBO algorithm ends and outputs the optimized kernel parameter g=3.72 and the kernel regularization coefficient C=28.83;

[0136] Step S413, input the data set into KELM to obtain 6 prediction sequences, sum the 6 prediction sequences and perform inverse normalization to obtain the photovoltaic power prediction value.

[0137] In order to verify the effectiveness of the improved dung beetle algorithm optimized kernel extreme learning machine (VMD-IDBO-KELM) model proposed in the present invention, comparative experiments were carried out with the kernel extreme learning machine (VMD-KELM) and the dung beetle algorithm optimized extreme learning machine (VMD-DBO-KELM). To ensure the reliability of the experiment, the experiments were carried out under the same conditions and environment.

[0138] In the present invention, the mean absolute error (MAE), root mean square error (RMSE), and coefficient of determination (R 2 ) is used as the evaluation index of the model. The experimental results are shown in Table 2. By comparison, it can be seen that the prediction accuracy of the model of the present invention is higher.

[0139] Table 2 Prediction errors of different models

[0140]

[0141]

[0142] In the present invention, Figure 3 A fitness comparison chart of the improved dung beetle algorithm and the dung beetle algorithm is drawn. Figure 3 It can be seen that the improved dung beetle algorithm converges faster than the dung beetle algorithm and can effectively escape from the local optimal solution.

[0143] In the present invention, Figure 4 The prediction results of the present invention and different models are plotted. Figure 4 It can be seen that the present invention has better prediction accuracy.

Claims

1. A short-term photovoltaic power prediction method based on an improved dung beetle algorithm to optimize the nuclear extreme learning machine, characterized by The following steps are involved: Step 1: Obtain historical photovoltaic power generation data of the photovoltaic power station and corresponding meteorological factor data of the photovoltaic array area, and process the data; Step 2: Decompose the photovoltaic power data by using the variational mode decomposition method VMD, and combine it with the meteorological factor data to form a data set; Step 3: Build the kernel extreme learning machine KELM model; Step 4: Use the improved dung beetle algorithm IDBO to optimize the kernel extreme learning machine KELM model parameters and predict the short-term photovoltaic power.

2. The short-term photovoltaic power prediction method based on the improved dung beetle algorithm optimized kernel extreme learning machine according to claim 1 is characterized by: In step 1, the collection time period of the photovoltaic power generation data measured by the photovoltaic power station is set to 7:00 am to 18:00 pm, and the collection interval is 5 minutes; the variables included in the data at each moment include the total horizontal irradiance, scattered horizontal irradiance, temperature, relative humidity, and wind speed meteorological factor data from 7:00 am to 18:00 pm.

3. The short-term photovoltaic power prediction method based on the improved dung beetle algorithm optimized kernel extreme learning machine according to claim 2 is characterized by: In step 1, data processing includes the following steps: S1.1: Clean the abnormal data and make corrections using formula (1): P=ηSI[1-β(T c +T0)] (1); In formula (1), P is the photovoltaic power generation, η is the conversion efficiency, S is the light receiving area of ​​the photovoltaic array, I is the irradiance, and T c is the component temperature, β is the temperature rise coefficient, and T0 is a constant; S1.2: Normalize the data: In formula (2), x′ is the value of the sample after normalization, x is the value of the initial sample, and x max is the maximum value of the column data, x min The minimum value of the column data.

4. The short-term photovoltaic power prediction method based on the improved dung beetle algorithm optimized kernel extreme learning machine according to claim 1 is characterized by: In step 2, the constrained variational model of the variational mode decomposition method VMD is: In formula (3), K is the total number of modal classifications, is the partial derivative operation, δ(t) is the unit pulse function, u k (t) is the modal component, w k is the center frequency, k is the serial number of the current modal component, e is the natural logarithm, f(t) is the original sequence, and j is the imaginary unit; By using formula (3), the photovoltaic power generation is decomposed to obtain k IMF components. Combined with the meteorological factor data, the data set input into the kernel extreme learning machine KELM model is: In formula (4), x t1 、x t2 ,...x tm They represent the values ​​of total horizontal irradiance, diffuse horizontal irradiance, temperature, relative humidity, and wind speed at time t, respectively, m = 5; Represents the value of the i-th IMF component at time t, where i is between [1, k].

5. According to claim 1, the short-term photovoltaic power prediction method based on the improved dung beetle algorithm optimized kernel extreme learning machine, Features: In step 3, the expression of the kernel extreme learning machine KELM model is: In formula (5), y(x) is the output value, H is the hidden layer output matrix, and H T is the transposed matrix of H, C is the regularization coefficient, I is the identity matrix, T is the target output matrix, x is the input vector, x1, x M is the training sample, K(x,x1), K(x,x M ) correspond to x and x1, x and x respectively. M Gaussian kernel function; The Gaussian kernel function is shown in formula (6): In formula (6), K(x i ,x j ) represents x i With x j Gaussian kernel function, σ is the kernel parameter, x i 、x j Input vector for the experiment.

6. According to claim 5, the short-term photovoltaic power prediction method based on the improved dung beetle algorithm optimized kernel extreme learning machine, Features: In step 4, the improved dung beetle algorithm IDBO is used to optimize the kernel extreme learning machine KELM model parameters to predict short-term photovoltaic power. The model parameters include the kernel parameter σ and the regularization coefficient C, and the following steps are included: S4.1: Set initial values: population size n, maximum number of iterations T max , population dimension dim, kernel parameter σ and regularization coefficient C; S4.2: The fitness function of the population is the mean square error function of the KELM model, and its expression is: In formula (7), f(σ,C) is the fitness function, σ is the kernel parameter, C is the regularization coefficient, N is the total number of samples, i is the current sample number, y i is the i-th predicted value, is the i-th true value; S4.3: The population is initialized using Chebyshev chaotic mapping, and the iteration method is as follows: x n+1 =cos(carccosx n ) (8); In formula (8), k is the order, x n ∈[-1,1],x n is the nth generation population; x n+1 is the n+1 generation population; S4.4: Introduce the golden sine strategy to improve the position update of the dung beetle rolling behavior: x i (t+1)=x i (t)×|sin(R1)|+R2×sin(R1)×|c1×X b -c2×x i (t)| (9); In the formula, x i (t) represents the current position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the ith dung beetle at the t+1th iteration, R1 is a random number in [0,2π], R2 is a random number in [0,π], and X b represents the global optimal position; c1 and c2 are the coefficients containing the golden section, is the golden ratio number; S4.5: When the dung beetle's rolling behavior is blocked, it performs dancing behavior and updates its position: x i (t+1)=x i (t)+tan(θ)|x i (t)-x i (t-1)| (11)? In formula (11), x i (t) represents the position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the i-th dung beetle at the t+1th iteration, x i (t-1) represents the position of the i-th dung beetle at the t-1th iteration, θ is the deflection angle, θ∈[0,π]; S4.6: Select the boundaries of the area where dung beetles breed and lay eggs: In formula (12), X * is the current optimal position, Ub and Lb are the upper and lower boundaries of the optimization problem of the dung beetle breeding and egg-laying area, respectively. * and Lb * are the upper and lower boundaries of the dung beetle breeding and egg-laying area; R = 1-t / T max , t is the current iteration number, T max is the maximum number of iterations; S4.7: Introduce the spiral search strategy of the whale optimization algorithm to update the position of the hatching ball: B i (t+1)=X * +b1×(B i (t)-Lb * )×e hl ×cos(2πl)+b2×(B i (t)-Ub * )×e hl ×cos(2πl)(13); In formula (13), X * represents the current optimal position, b1 and b2 represent random vectors of size 1×D; D represents the dimension of the optimization problem; Ub * and Lb * The upper and lower boundaries of the dung beetle breeding and egg-laying area; B i (t+1) represents the position information of the t+1th iteration of the i-th hatching ball; B i (t) represents the position information of the t-th iteration of the i-th hatching ball; e is the natural logarithm, h is a constant; l is a number between [-1,1]; S4.8: Boundaries of the foraging area of ​​hatched dung beetles: In formula (14), X b represents the global optimal position, Lb b and Ub b They represent the lower and upper bounds of the optimal foraging area of ​​the dung beetle, R = 1-t / T max , Lb and Ub represent the lower and upper bounds of the foraging area of ​​the small dung beetle, respectively; S4.9: Introducing adaptive weights to update the foraging position of dung beetles: x i (t+1)=x i (t)+C1×k1×(x i (t)-Lb b )+C2×k2×(x i (t)-Ub b ) (15); In formula (15), t is the number of iterations, T max is the maximum number of iterations, x i (t) represents the position of the i-th dung beetle at the t-th iteration, x i (t+1) represents the position of the t+1th iteration of the i-th dung beetle, C1 is a random number that follows a normal distribution, C2 is a random number between (0,1), and Lb b and Ub b They represent the lower and upper bounds of the optimal foraging area of ​​the small dung beetle, respectively; S4.10: Update the location of the dung beetle in its fight for food: x i (t+1)=X b +S×g×(|x i (t)-X * |+|x i (t)-X b |) (16); In formula (16), X b represents the best position for competing for food; S represents a constant value; g is a random vector of size 1×D and obeys a normal distribution; X b represents the global optimal position; x i (t) represents the position of the i-th dung beetle at the t-th iteration; x i (t+1) represents the position of the i-th dung beetle at the t+1th iteration; S4.11: Determine whether the algorithm has reached the maximum number of iterations. If so, proceed to step S4.11, otherwise return to S4.3; S4.12: The IDBO algorithm ends and outputs the optimized kernel parameter g and kernel regularization coefficient C; S4.13: Input the data set into KELM to obtain k prediction sequences, sum and reconstruct the k prediction sequences and perform denormalization to obtain the photovoltaic power prediction value.

Citation Information

Cited By

  • GH4169 milling process parameter optimization method based on RFR-IDBO algorithm

    CN119647016A