A Photovoltaic Power Generation Prediction Method and System Considering Wavelet Decomposition and Multi-Objective Optimization

Abstract

This invention provides a photovoltaic (PV) power generation prediction method and system considering wavelet decomposition and multi-objective optimization, belonging to the field of PV power generation prediction technology. The method includes: obtaining an initial PV power generation scenario set; denoising the initial PV power generation scenario set to reconstruct the PV power generation scenario set; using a combined prediction system to predict the reconstructed PV power generation scenario set, selecting a single prediction system as a prediction subsystem of the combined system, with each prediction subsystem outputting a prediction result based on the reconstructed PV power generation scenario set; determining the weight parameters of each prediction subsystem; and using the obtained weight parameters of the prediction subsystems and the prediction results of each prediction subsystem on the reconstructed PV power generation scenario set, weightedly outputting the final prediction result. This method can improve the stability and accuracy of PV power generation prediction and effectively assist the integrated energy system in rational optimization and scheduling.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic power generation prediction technology, and particularly relates to a photovoltaic power generation prediction method and system that considers wavelet decomposition and multi-objective optimization algorithms. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] my country's energy structure has long been dominated by fossil fuels such as coal and oil, a situation that is expected to persist for a considerable period. However, fossil fuel reserves are finite and non-renewable, their utilization is inefficient, and their production and use generate substantial pollution emissions, causing irreversible environmental damage. Therefore, the conflict between energy supply and environmental protection is becoming increasingly prominent.

[0004] As the utilization of renewable energy sources such as solar and wind power continues to increase, the penetration rate of renewable energy in integrated energy systems is constantly rising, and photovoltaic power generation technology is gradually becoming a research focus for thermoelectric power in various countries. However, photovoltaic development is characterized by unevenness, randomness, and volatility, which can easily cause impacts on the power grid when connected to it.

[0005] Therefore, forecasting photovoltaic (PV) power generation has become an indispensable task for maintaining power system stability and optimizing the dispatch of integrated energy systems. Accurately predicting PV power output based on historical data allows power dispatching departments to adjust dispatching plans in a timely manner, improving the economy and stability of grid operation, reducing curtailment of solar power, and is of great significance for promoting the development and utilization of solar energy. Existing PV forecasting methods mainly employ neural networks; however, single neural network algorithms inevitably suffer from problems such as low prediction accuracy and inaccurate results. Summary of the Invention

[0006] To overcome the shortcomings of the prior art, this invention provides a photovoltaic power generation prediction method and system that considers wavelet decomposition and multi-objective optimization, which can improve the stability and accuracy of photovoltaic power generation prediction and effectively assist the integrated energy system in reasonable optimization scheduling.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0008] In a first aspect, the present invention discloses a photovoltaic power generation prediction method considering wavelet decomposition and multi-objective optimization, comprising:

[0009] Obtain the initial set of photovoltaic power generation scenarios;

[0010] Denoise the initial photovoltaic power generation scene set to reconstruct the photovoltaic power generation scene set;

[0011] A combined prediction system is used to predict the reconstructed photovoltaic power generation scenario set. A single prediction system is selected as the prediction subsystem of the combined system, and each prediction subsystem outputs the prediction result based on the reconstructed photovoltaic power generation scenario set.

[0012] Determine the weight parameters for each prediction subsystem;

[0013] The final prediction result is output by weighting the obtained weight parameters of the prediction subsystem and the prediction results of each prediction subsystem on the reconstructed photovoltaic power generation scenario set.

[0014] As a further technical solution, wavelet threshold denoising method is selected to reduce high-frequency noise in the initial scene set of photovoltaic power generation to achieve denoising processing.

[0015] Secondly, this invention discloses a photovoltaic power generation prediction system considering wavelet decomposition and multi-objective optimization, comprising:

[0016] The photovoltaic power generation scenario set reconstruction module is configured to: obtain the initial photovoltaic power generation scenario set;

[0017] Denoise the initial photovoltaic power generation scene set to reconstruct the photovoltaic power generation scene set;

[0018] The subsystem prediction module is configured to: use a combined prediction system to predict the reconstructed photovoltaic power generation scenario set, select a single prediction system as the prediction subsystem of the combined system, and each prediction subsystem outputs prediction results based on the reconstructed photovoltaic power generation scenario set;

[0019] The final prediction module is configured to determine the weight parameters of each prediction subsystem.

[0020] The final prediction result is output by weighting the obtained weight parameters of the prediction subsystem and the prediction results of each prediction subsystem on the reconstructed photovoltaic power generation scenario set.

[0021] The above one or more technical solutions have the following beneficial effects:

[0022] 1. This invention considers using wavelet decomposition technology to optimize the initial scene set of photovoltaic power generation, reduce high-frequency noise in the scene set, eliminate inferior scenes, and use a genetic algorithm to automatically optimize the main parameters of the wavelet decomposition mechanism to enhance the applicability of the system.

[0023] 2. A combined prediction system is used to predict the reconstructed scene set, which overcomes the shortcomings of a single prediction system, significantly reduces prediction error, and improves prediction accuracy and stability.

[0024] 3. The multi-objective gray wolf optimization algorithm is used to determine the weight parameters of individual prediction systems in the combined prediction system, which overcomes the errors caused by subjective factors and makes the final prediction results more effective.

[0025] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0026] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0027] Figure 1 This is a flowchart of the prediction method described in Embodiment 1 of the present invention;

[0028] Figure 2 This is a flowchart of the wavelet decomposition algorithm described in Embodiment 1 of the present invention;

[0029] Figure 3 This is a flowchart of the solution process for the multi-objective gray wolf optimization algorithm described in Embodiment 1 of the present invention. Detailed Implementation

[0030] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0031] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0032] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0033] Example 1

[0034] This embodiment discloses a photovoltaic power generation prediction method that considers wavelet decomposition and multi-objective optimization, as shown in the attached figure. Figure 1 As shown, it includes the following steps:

[0035] S1: By consulting the historical database of photovoltaic power plants, obtain the initial scenario set of photovoltaic power generation, including parameters such as historical ambient temperature, humidity, average irradiance, light intensity, and illuminance, as well as the photovoltaic power generation power of the corresponding time period, as the test set of the prediction system.

[0036] S2: Select the wavelet threshold denoising method to reduce high-frequency noise in the initial photovoltaic power generation scene set, that is, to remove bad data points, reduce the impact of extreme environmental factors on photovoltaic power generation prediction, reconstruct the scene set, and use a genetic algorithm to automatically optimize the main parameters of the wavelet threshold denoising method, including the number of decomposition layers k and the threshold α of each layer, to reduce the randomness and volatility of the initial scene.

[0037] Step 101: Initialize the genetic algorithm parameters, including population size, number of iterations, stratification range, threshold range, and correlation coefficient.

[0038] Step 102: Select a wavelet and determine the wavelet decomposition level N, and perform N-level wavelet decomposition calculation on the signal.

[0039] Step 103: For each high-frequency coefficient (three directions) from layer 1 to layer N, select a threshold for quantization processing. The quantization processing method adopts soft threshold quantization.

[0040] Step 104: Select a combination of indicators including mean absolute error, mean absolute percentage error, root mean square error, consistency level and correlation coefficient as the objective function of the genetic algorithm, perform iterative optimization on the wavelet decomposition, and finally select the best set of decomposition layer number k, threshold α for each layer and wavelet decomposition results.

[0041] Step 105: Based on the low-frequency coefficients of the Nth layer of wavelet decomposition and the high-frequency coefficients of the 1st to Nth layers after quantization, perform wavelet reconstruction of the signal to obtain the typical daily photovoltaic power generation value G after removing bad pixels. real (t).

[0042] Among them, the population size is the number of data selected for optimization in one step, the stratification range and threshold range are the ranges of decision variables, and the overall idea of ​​wavelet decomposition optimization is to decompose the original typical photovoltaic power generation data into layers, and then remove the extreme cases in the photovoltaic power generation data, so that the overall photovoltaic power generation dataset has less volatility and the relationship between photovoltaic power generation values ​​and environmental parameters is more accurate.

[0043] S3: To better obtain and utilize the characteristics of typical daily photovoltaic power generation scenarios, a combined prediction system is used to predict the reconstructed scenario set. Referring to four common prediction metrics (mean absolute error, root mean square error, squared absolute percentage error, and coefficient of determination), four single prediction systems with superior prediction results are selected as subsystems of the combined system, including recurrent neural networks, deep convolutional neural networks, long short-term memory neural networks, and backpropagation neural networks. Each prediction subsystem outputs a prediction result {Y} based on the reconstructed scenario set. i (t), i∈[1,4]}.

[0044] S4: Use the weight parameters of each prediction subsystem as decision variables {α} i ,i∈[1,4]}, the root mean square error and square absolute percentage error between the combined predicted value of the prediction system and the actual photovoltaic power generation value are used as objective functions, and the weight parameters of each prediction subsystem are determined by the multi-objective improved gray wolf algorithm.

[0045] The objective function for solving the weight parameter set of each prediction subsystem using the multi-objective gray wolf optimization algorithm is as follows:

[0046]

[0047] In the formula: F1 is the squared absolute percentage error between the combined predicted value of the prediction system and the actual photovoltaic power generation; F2 is the root mean square error between the combined predicted value of the prediction system and the actual photovoltaic power generation; α i Y represents the weighting parameter of the predicted value of the i-th prediction subsystem; i G(t) represents the predicted photovoltaic power generation value of the i-th prediction subsystem at time t; G(t) represents the actual photovoltaic power generation value at time t.

[0048] In this multi-objective gray wolf optimization algorithm, an "observation" strategy is added. After the gray wolf population completes its position update in each iteration, it randomly observes the surrounding positions. An external storage archive set P is added to store the optimal solution during the algorithm's solution process.

[0049] As attached Figure 3 As shown, the process of solving for the weight parameter set of each prediction subsystem using the multi-objective gray wolf optimization algorithm is as follows:

[0050] Step 201: Initialize algorithm parameters: Gray wolf population position X (i.e., weight parameter {α}) i The population size G is given by the expression G(i∈[1,4]). N Number of iterations N G Weight upper and lower limits H d and L d wait;

[0051] Step 202: Initialize the external storage Archive set P, and let the iteration count t = 1;

[0052] Step 203: Calculate the fitness function for each individual in the population, find the current global optimum, and save the position of the optimum individual as M;

[0053] Step 204: Use the roulette wheel strategy to select the three alpha wolves of the current population;

[0054] Step 205: Update the position of the gray wolves in the population using the position update formula;

[0055] Step 206: Use the "observation" strategy to observe the vicinity of each gray wolf's location. If the new location is better, move to that location.

[0056] Step 207: Calculate and compare the individual historical best position and the population best position for each gray wolf;

[0057] Step 208: Archive the optimal position of each gray wolf, compare it with the Pareto optimal solution and M, and update the optimal individual position M;

[0058] Step 209: Compare the new non-dominated solution with each non-dominated solution in the Archive set, and update the outer storage Archive set P;

[0059] Step 210: Determine whether the external storage Archive set P is full. If yes, proceed to step 211; otherwise, proceed to step 212.

[0060] Step 211: Use a non-dominated sorting strategy to exclude a set of weight coefficients;

[0061] Step 212: Add the new results to the external storage Archive set P;

[0062] Step 213: Determine if the number of iterations has reached the upper limit. If yes, proceed to step 215; otherwise, proceed to step 214.

[0063] Step 214: Iteration count t = t + 1, return to step 203;

[0064] Step 215: Output the external storage Archive set P, and select the optimal set of weight coefficients {α} from P. i ,i∈[1,4]}.

[0065] S5: Combine the obtained weight parameters of the prediction subsystem with the prediction models of each prediction subsystem to obtain the combined prediction model.

[0066] This invention employs wavelet threshold denoising technology to optimize the original photovoltaic power generation scene set, effectively reducing the impact of data noise. Four single-indicator prediction models with good performance are selected as prediction subsystems. The weights of the subsystems are determined through a multi-objective gray wolf optimization algorithm, overcoming the shortcomings of individual prediction systems, improving the stability and accuracy of photovoltaic power generation prediction, effectively assisting the integrated energy system in reasonable optimization and scheduling, reducing system energy consumption, and improving the economic efficiency of system operation.

[0067] Example 2

[0068] The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0069] Example 3

[0070] The purpose of this embodiment is to provide a computer-readable storage medium.

[0071] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above method.

[0072] Example 4

[0073] The purpose of this embodiment is to provide a photovoltaic power generation prediction system that considers wavelet decomposition and multi-objective optimization, including:

[0074] The photovoltaic power generation scenario set reconstruction module is configured to: obtain the initial photovoltaic power generation scenario set;

[0075] Denoise the initial photovoltaic power generation scene set to reconstruct the photovoltaic power generation scene set;

[0076] The subsystem prediction module is configured to: use a combined prediction system to predict the reconstructed photovoltaic power generation scenario set, select a single prediction system as the prediction subsystem of the combined system, and each prediction subsystem outputs prediction results based on the reconstructed photovoltaic power generation scenario set;

[0077] The final prediction module is configured to determine the weight parameters of each prediction subsystem.

[0078] The final prediction result is output by weighting the obtained weight parameters of the prediction subsystem and the prediction results of each prediction subsystem on the reconstructed photovoltaic power generation scenario set.

[0079] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0080] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0081] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A photovoltaic power generation prediction method considering wavelet decomposition and multi-objective optimization, characterized in that, include: Obtain the initial set of photovoltaic power generation scenarios; Denoise the initial photovoltaic power generation scene set to reconstruct the photovoltaic power generation scene set; A combined prediction system is used to predict the reconstructed photovoltaic power generation scenario set. A single prediction system is selected as the prediction subsystem of the combined system, and each prediction subsystem outputs the prediction result based on the reconstructed photovoltaic power generation scenario set. Referring to four common prediction metrics, namely mean absolute error, root mean square error, squared absolute percentage error, and coefficient of determination, four single prediction systems with better prediction results were selected as subsystems of the combined system, including recurrent neural networks, deep convolutional neural networks, long short-term memory neural networks, and backpropagation neural networks. Determine the weight parameters for each prediction subsystem; The weight parameters of each prediction subsystem are used as decision variables. The root mean square error and square absolute percentage error between the combined predicted value of the prediction system and the actual photovoltaic power generation value are used as objective functions, and the weight parameters of each prediction subsystem are determined by the multi-objective improved gray wolf algorithm. The objective function for determining the weight parameter set of each prediction subsystem using the multi-objective gray wolf optimization algorithm is as follows: In the formula: The squared absolute percentage error between the combined predicted value of the prediction system and the actual photovoltaic power generation value; The root mean square error between the combined predicted value of the prediction system and the actual photovoltaic power generation value; For the first Weight parameters for the predicted values ​​of each prediction subsystem; For the first Each prediction subsystem in Forecasted photovoltaic power generation at any given time; for The actual power output of photovoltaic power generation at any given time; The final prediction result is output by weighting the obtained weight parameters of the prediction subsystem and the prediction results of each prediction subsystem on the reconstructed photovoltaic power generation scenario set.

2. The photovoltaic power generation prediction method considering wavelet decomposition and multi-objective optimization as described in claim 1, characterized in that, The initial photovoltaic power generation scenario set includes historical ambient temperature, humidity, average irradiance, light intensity, illuminance parameters, and photovoltaic power generation power for the corresponding time period.

3. The photovoltaic power generation prediction method considering wavelet decomposition and multi-objective optimization as described in claim 1, characterized in that, The initial photovoltaic power generation scene set is denoised to reconstruct the photovoltaic power generation scene set, specifically as follows: The wavelet thresholding method is chosen to reduce high-frequency noise in the initial photovoltaic power generation scenario set, that is, to remove bad data points, reduce the impact of extreme environmental factors on photovoltaic power generation prediction, and reconstruct the scenario set.

4. The photovoltaic power generation prediction method considering wavelet decomposition and multi-objective optimization as described in claim 3, characterized in that, The parameters of the wavelet thresholding denoising method are automatically optimized using a genetic algorithm, including the number of decomposition layers k and the threshold for each layer. This reduces the randomness and volatility of the initial scenario.

5. A photovoltaic power generation prediction system considering wavelet decomposition and multi-objective optimization, characterized in that, include: The photovoltaic power generation scenario set reconstruction module is configured to: obtain the initial photovoltaic power generation scenario set; Denoise the initial photovoltaic power generation scene set to reconstruct the photovoltaic power generation scene set; The subsystem prediction module is configured to: use a combined prediction system to predict the reconstructed photovoltaic power generation scenario set, select a single prediction system as the prediction subsystem of the combined system, and each prediction subsystem outputs prediction results based on the reconstructed photovoltaic power generation scenario set; Referring to four common prediction metrics, namely mean absolute error, root mean square error, squared absolute percentage error, and coefficient of determination, four single prediction systems with better prediction results were selected as subsystems of the combined system, including recurrent neural networks, deep convolutional neural networks, long short-term memory neural networks, and backpropagation neural networks. The final prediction module is configured to determine the weight parameters of each prediction subsystem. The weight parameters of each prediction subsystem are used as decision variables. The root mean square error and square absolute percentage error between the combined predicted value of the prediction system and the actual photovoltaic power generation value are used as objective functions, and the weight parameters of each prediction subsystem are determined by the multi-objective improved gray wolf algorithm. The objective function for determining the weight parameter set of each prediction subsystem using the multi-objective gray wolf optimization algorithm is as follows: In the formula: The squared absolute percentage error between the combined predicted value of the prediction system and the actual photovoltaic power generation value; The root mean square error between the combined predicted value of the prediction system and the actual photovoltaic power generation value; For the first Weight parameters for the predicted values ​​of each prediction subsystem; For the first Each prediction subsystem in Forecasted photovoltaic power generation at any given time; for The actual power output of photovoltaic power generation at any given time; The final prediction result is output by weighting the obtained weight parameters of the prediction subsystem and the prediction results of each prediction subsystem on the reconstructed photovoltaic power generation scenario set.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1-4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it performs the steps of the method as described in any one of claims 1-4.

Images