A photovoltaic output power prediction method and system based on neural network

Through the combination of convolutional neural network and multi-layer perceptron, similar daily clustering and prediction of photovoltaic power generation power are carried out, and combined with meteorological data analysis, the problems of low prediction accuracy of photovoltaic power generation and incomplete weather impact analysis in the existing technology are solved. It realizes efficient and accurate photovoltaic output power prediction.

CN115936177BActive Publication Date: 2024-12-06UNIV OF JINAN
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202211345699.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-10-31
Publication Date
2024-12-06
Estimated Expiration
2042-10-31

AI Technical Summary

Technical Problem

The prediction accuracy of the existing photovoltaic power output power is not high. Traditional neural networks are prone to fall into local extreme points, have slow convergence speed, and cannot accurately predict the photovoltaic output power under different weather types, resulting in safe operation of the power grid and renewable energy consumption. Restricted.

Method used

Convolutional neural network is used to perform similar daily clustering, screen out data with the same weather conditions, and use a multi-layer perceptron to predict the photovoltaic power generation power, combine historical meteorological data to analyze meteorological impact factors, and train prediction models under different weather types. , evaluate model performance through MAE, RMSE and MAPE indicators.

Benefits of technology

Improve the accuracy and efficiency of photovoltaic prediction, overcome the problems of local extreme value and convergence speed of traditional neural networks, make up for the insufficient overfitting of multi-layer perception machines, realize accurate prediction of photovoltaic output power under different weather types, and reduce the Prediction error and uncertainty.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115936177B_ABST
    Figure CN115936177B_ABST
Patent Text Reader

Abstract

The present invention discloses a photovoltaic output power prediction method and system based on a neural network, and relates to the field of photovoltaic power generation technology. A photovoltaic power prediction model under different weather types is formed by combining a convolutional neural network and a multi-layer perceptron; and historical meteorological data is input into the photovoltaic power prediction model to obtain a photovoltaic output power prediction result. The present invention uses a convolutional neural network to divide the original data into multiple subsets, and brings together data with the same weather conditions, and uses the characteristics that the photovoltaic power of similar days has a strong correlation to improve the prediction accuracy of the model. The present invention performs a regression analysis on meteorological influencing factors, identifies and optimizes the input variables of the photovoltaic power prediction model, and uses MAE, MAPE, and RMSE three indicators to evaluate model performance and analyze errors, thereby reflecting the comprehensive performance of the model, effectively analyzing the impact of meteorological influencing factors on photovoltaic power, and further improving the prediction accuracy of the model.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation, and in particular to a photovoltaic output power prediction method and system based on a neural network. Background Art

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

[0003] Photovoltaic power generation has the inherent disadvantages of intermittency and randomness. With the access of a large number of photovoltaic power generation to the power grid, it is bound to bring severe challenges to the safe and stable operation of the power system and the guarantee of power quality. Therefore, the prediction of photovoltaic power generation output power has become one of the effective ways to solve this problem. The inventors found that the prediction accuracy of the prior art is not high, there are many prediction errors, and the model training time is long. Traditional neural networks also have the problem of being prone to local extreme points and slow convergence. At the same time, when analyzing the impact of different weather types on photovoltaic power generation in the existing methods, it is impossible to achieve a comprehensive analysis of weather influencing factors, so it is impossible to accurately predict the photovoltaic output power under different weather types. In addition, due to the strong randomness and volatility of photovoltaic output, it will bring challenges to the safe operation and peak-shaving scheduling of the power grid, which will inevitably introduce prediction errors, increase the uncertainty of the prediction results, and thus greatly limit the absorption of renewable energy. Summary of the invention

[0004] In view of the shortcomings of the prior art, the purpose of the present invention is to provide a photovoltaic output power prediction method and system based on neural network, which uses CNN convolutional neural network to cluster similar days and screen out data with the same weather conditions; then uses a multilayer perceptron (MLP) to predict photovoltaic power generation, which makes up for the deficiency of traditional multilayer perceptron that is prone to overfitting and effectively improves the photovoltaic prediction accuracy and efficiency.

[0005] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0006] The first aspect of the present invention provides a photovoltaic output power prediction method based on a neural network, comprising the following steps:

[0007] Obtain historical meteorological data and satellite cloud images of photovoltaic power plants;

[0008] Use convolutional neural networks to classify satellite cloud images;

[0009] Select the pre-trained photovoltaic power prediction model under different weather types according to the classification results;

[0010] The historical meteorological data is used as a model input to obtain a photovoltaic output power prediction result.

[0011] Furthermore, the historical meteorological data and satellite cloud images need to be preprocessed before classification. The preprocessing steps include: filtering the data, deleting abnormal data and filling in missing data, standardizing the filtered data, and setting the data format.

[0012] Furthermore, the weather types include sunny, cloudy and overcast; the historical meteorological data include: total radiation, direct radiation, diffuse radiation, atmospheric temperature, ambient temperature, air pressure, and ambient humidity.

[0013] Furthermore, historical meteorological data are used to conduct correlation analysis of meteorological influencing factors to determine the degree of influence of meteorological influencing factors on photovoltaic power generation.

[0014] Furthermore, the training process of the photovoltaic power prediction model under different weather types is: using historical meteorological data under different weather types and corresponding photovoltaic output power historical data to train photovoltaic power prediction models under different weather types respectively.

[0015] Furthermore, data with the same weather conditions are selected to train a multi-layer perceptron model to learn the mapping relationship between meteorological factor component signals and corresponding photovoltaic power component signals.

[0016] Furthermore, the pre-trained photovoltaic power prediction model under different weather types needs to be evaluated, and the evaluation indicators are mean absolute error (MAE), root mean square error (RMSE) and mean absolute percentage error (MAPE).

[0017] A second aspect of the present invention provides a photovoltaic output power prediction system based on a convolutional neural network, comprising:

[0018] A data acquisition module, configured to acquire historical meteorological data and satellite cloud images of the photovoltaic power station;

[0019] A data classification module, configured to classify satellite cloud images using a convolutional neural network;

[0020] A model selection module is configured to select a pre-trained photovoltaic power prediction model under different weather types according to the classification result;

[0021] The power prediction module is configured to use the historical meteorological data as a model input to obtain a photovoltaic output power prediction result.

[0022] The third aspect of the present invention provides a medium having a program stored thereon, which, when executed by a processor, implements the steps in the photovoltaic output power prediction method based on a neural network as described in the first aspect of the present invention.

[0023] The fourth aspect of the present invention provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the photovoltaic output power prediction method based on a neural network as described in the first aspect of the present invention are implemented.

[0024] One or more of the above technical solutions have the following beneficial effects:

[0025] The present invention selects a new type of neural network, a multi-layer perceptron, combined with the similar day clustering results of a convolutional neural network to predict photovoltaic power generation, effectively improving the photovoltaic prediction accuracy and efficiency. While overcoming the problems of traditional neural networks that are prone to falling into local extreme points and slow convergence speed, the present invention also makes up for the deficiency of traditional multi-layer perceptrons that are prone to overfitting.

[0026] The solution described in the present invention uses a convolutional neural network to divide historical meteorological data into multiple subsets, brings together data with the same weather conditions, and uses the fact that photovoltaic power on similar days has a strong correlation to predict photovoltaic output power, thereby improving the prediction accuracy of the model.

[0027] The scheme described in the present invention performs regression analysis on the generated power and meteorological influencing factors such as irradiance, ambient temperature, relative humidity, wind force, etc., identifies and optimizes the input variables of the photovoltaic power prediction model, and uses MAE, MAPE, and RMSE3 indicators to evaluate the model performance and analyze the error, thereby reflecting the comprehensive performance of the model, effectively analyzing the impact of meteorological influencing factors on photovoltaic power, and improving the prediction accuracy.

[0028] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0030] Figure 1 This is a schematic diagram of correlation analysis of meteorological influencing factors in Example 1 of the present invention;

[0031] Figure 2 This is a flow chart of a photovoltaic output power prediction method in Embodiment 1 of the present invention;

[0032] Figure 3 This is a comparison diagram of evaluation index errors between the photovoltaic output power prediction method of the present invention and the photovoltaic output power prediction method of the prior art in the first embodiment of the present invention;

[0033] FIG4( a ) is a schematic diagram of prediction results of the photovoltaic output power prediction method in spring in Embodiment 1 of the present invention;

[0034] FIG4( b ) is a schematic diagram of prediction results of the photovoltaic output power prediction method in summer in Embodiment 1 of the present invention;

[0035] FIG4( c ) is a schematic diagram of the prediction results of the photovoltaic output power prediction method in autumn in the first embodiment of the present invention.

[0036] FIG4( d ) is a schematic diagram of the prediction results of the photovoltaic output power prediction method in winter in the first embodiment of the present invention. DETAILED DESCRIPTION

[0037] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present application belongs.

[0038] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "include" and / or "include" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or their combinations;

[0039] Embodiment 1:

[0040] Embodiment 1 of the present invention provides a photovoltaic output power prediction method based on a neural network, comprising the following steps:

[0041] Obtaining original solar power (PV) time series data, namely, historical meteorological data and satellite cloud images of photovoltaic power stations, and performing dimensionality reduction on the historical meteorological data;

[0042] Use convolutional neural networks to classify satellite cloud images;

[0043] A pre-trained photovoltaic power prediction model under different weather types is selected according to the classification result; the weather types include sunny, cloudy and overcast; the historical meteorological data include: total radiation, direct radiation, diffuse radiation, atmospheric temperature, ambient temperature, air pressure, ambient humidity.

[0044] The training process of the photovoltaic power prediction model under different weather types is: using historical meteorological data under different weather types and corresponding photovoltaic output power historical data to train the photovoltaic power prediction model under different weather types respectively.

[0045] The historical meteorological data is used as a model input to obtain a photovoltaic output power prediction result.

[0046] As a further technical solution, historical meteorological data and satellite cloud images need to be preprocessed before classification. The preprocessing steps include: filtering the data, deleting abnormal data and filling in missing data, standardizing the filtered data, and setting the data format.

[0047] As a further technical solution, historical meteorological data is used to conduct correlation analysis of meteorological influencing factors to determine the degree of influence of meteorological influencing factors on photovoltaic power generation.

[0048] Specifically, linear regression was used to analyze the correlation of meteorological influencing factors. The specific results are as follows: Figure 1 As shown, R is the correlation coefficient. It can be seen that total radiation, direct radiation, and temperature are relatively important meteorological factors affecting photovoltaic power generation, and have a greater effect on photovoltaic power generation.

[0049] like Figure 2 As shown, a flow chart of the photovoltaic output power prediction method of the present invention is shown, and the method includes:

[0050] Step 1: Select weather data based on satellite cloud images, and then use CNN neural network to classify the weather. According to the satellite cloud images, the weather is divided into three categories, namely rainy, cloudy, and sunny.

[0051] Step 2: Select data with the same weather conditions to train the multi-layer perceptron model to learn the mapping relationship between the meteorological factor component signal and the corresponding photovoltaic power component signal, where the meteorological factor is the model input and the photovoltaic power is the model output. In this embodiment, the meteorological factors are total radiation, direct radiation and temperature. The multi-layer perceptron model includes an input layer, a hidden layer and an output layer.

[0052] As a further technical solution, continuous data of the same number of days are selected to train a multi-layer perceptron model to learn the mapping relationship between the meteorological factor component signal and the corresponding photovoltaic power component signal, and continuous data of the same number of days are selected to train a long short-term memory network (Long Short-Term Memory, LSTM) for comparison, in order to verify the superiority of the technical solution of the present invention.

[0053] Step 3: Use the trained model to predict the test set. The input is total radiation, direct radiation and temperature, and the output is photovoltaic power. The first 90% of the data is the training set, and the last 10% of the data is the test set. The number of small batch samples is selected as 10, and the discard rate is set to 0.5.

[0054] Step 4: Use MAE, MAPE, and RMSE to evaluate model performance and analyze errors. The analysis results are as follows: Figure 3 As shown, the comprehensive performance of the model is reflected.

[0055] As a further technical solution, the evaluation indicators of the prediction model are Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and Mean Absolute Percent Error (MAPE). Mean absolute error is the average value of absolute error, which can better reflect the actual situation of the prediction value error. Root mean square error is the square root of mean square error. The root mean square deviation represents the sample standard deviation of the difference between the predicted value and the observed value. The smaller the mean absolute percentage error is in theory, the better the prediction model fitting effect is and the better the accuracy is. Specific formula:

[0056]

[0057]

[0058]

[0059] in, is the predicted power, y i is the actual power and n is the number of samples.

[0060] according to Figure 2 It can be seen that sunny days have the highest probability, so this embodiment selects sunny days as the detection weather for the hybrid model MAC, and uses MAC to predict the photovoltaic power generation under sunny conditions in the four quarters. MAC is used to first extract the dates of sunny days in the quarter, and then the original data of the photovoltaic power plant under the sunny type is used for prediction; MLP and LSTM are used to predict the original data of continuous dates.

[0061] According to the correlation analysis of meteorological influencing factors, total radiation, direct radiation, and temperature are relatively important meteorological influencing factors of photovoltaic power generation, and have a greater effect on photovoltaic power generation. Therefore, the multi-layer perceptron selects total radiation, direct radiation, and temperature as inputs and compares them with the LSTM prediction results. The comparison results are shown in Figure 4 (a)-(d). It can be clearly seen that the technical solution of the present invention is more accurate and reasonable than the LSTM prediction results.

[0062] Embodiment 2:

[0063] Embodiment 2 of the present invention provides a photovoltaic output power prediction system based on a convolutional neural network, comprising:

[0064] A data acquisition module, configured to acquire historical meteorological data and satellite cloud images of the photovoltaic power station;

[0065] A data classification module, configured to classify satellite cloud images using a convolutional neural network;

[0066] A model selection module is configured to select a pre-trained photovoltaic power prediction model under different weather types according to the classification result;

[0067] The power prediction module is configured to use the historical meteorological data as a model input to obtain a photovoltaic output power prediction result.

[0068] Embodiment three:

[0069] Embodiment 3 of the present invention provides a medium on which a program is stored. When the program is executed by a processor, the steps in the photovoltaic output power prediction method based on neural network as described in Embodiment 1 of the present invention are implemented.

[0070] Embodiment 4:

[0071] Embodiment 4 of the present invention provides a device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the photovoltaic output power prediction method based on neural network as described in Embodiment 1 of the present invention are implemented.

[0072] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part 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 to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0073] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0074] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it 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 on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A photovoltaic output power prediction method based on neural network, characterized in that: The following steps are involved: Obtain historical meteorological data and satellite cloud images of photovoltaic power plants; Use convolutional neural networks to classify satellite cloud images; Select the pre-trained photovoltaic power prediction model under different weather types according to the classification results; Use historical meteorological data to conduct correlation analysis of meteorological influencing factors to determine the impact of meteorological influencing factors on photovoltaic power generation; The training process of the photovoltaic power prediction model under different weather types is as follows: using historical meteorological data under different weather types and corresponding photovoltaic output power historical data to respectively train photovoltaic power prediction models under different weather types; selecting data with the same weather conditions to train a multi-layer perceptron model to learn the mapping relationship between meteorological factor component signals and corresponding photovoltaic power component signals; the meteorological factors are total radiation, direct radiation and temperature; The historical meteorological data is used as a model input to obtain a photovoltaic output power prediction result.

2. The photovoltaic output power prediction method based on neural network according to claim 1, characterized in that: Before classification, historical meteorological data and satellite cloud images need to be preprocessed. The preprocessing steps include: filtering the data, deleting abnormal data and filling in missing data, standardizing the filtered data, and setting the data format.

3. The photovoltaic output power prediction method based on neural network according to claim 1, characterized in that: The weather types include sunny, cloudy and overcast; the historical meteorological data include: total radiation, direct radiation, diffuse radiation, atmospheric temperature, ambient temperature, air pressure, and ambient humidity.

4. The photovoltaic output power prediction method based on neural network according to claim 1, characterized in that: The pre-trained photovoltaic power prediction model under different weather types needs to be evaluated, and the evaluation indicators are mean absolute error, root mean square error and mean absolute percentage error.

5. A photovoltaic output power prediction system based on convolutional neural network, characterized in that: include: A data acquisition module, configured to acquire historical meteorological data and satellite cloud images of the photovoltaic power station; A data classification module, configured to classify satellite cloud images using a convolutional neural network; A model selection module is configured to select a pre-trained photovoltaic power prediction model under different weather types according to the classification result; Use historical meteorological data to conduct correlation analysis of meteorological influencing factors to determine the impact of meteorological influencing factors on photovoltaic power generation; The training process of the photovoltaic power prediction model under different weather types is as follows: using historical meteorological data under different weather types and corresponding photovoltaic output power historical data to respectively train photovoltaic power prediction models under different weather types; selecting data with the same weather conditions to train a multi-layer perceptron model to learn the mapping relationship between meteorological factor component signals and corresponding photovoltaic power component signals; the meteorological factors are total radiation, direct radiation and temperature; The power prediction module is configured to use the historical meteorological data as a model input to obtain a photovoltaic output power prediction result.

6. A computer-readable storage medium, characterized in that: A plurality of instructions are stored therein, and the instructions are suitable for being loaded by a processor of a terminal device and executing the photovoltaic output power prediction method based on a neural network as described in any one of claims 1-4.

7. A terminal device, characterized in that: It includes a processor and a computer-readable storage medium, the processor is used to implement each instruction; the computer-readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded by the processor and executing the photovoltaic output power prediction method based on neural network described in any one of claims 1-4.

Citation Information

Patent Citations

  • Photovoltaic power combination prediction method and system based on multi-source data fusion

    CN113128793A