Photovoltaic power station set time power prediction method based on network model

Abstract

The invention discloses a photovoltaic power station set time power prediction method based on a network model, and the method comprises the following steps: 1, carrying out the statistics of historical data of factors influencing the set time power of a photovoltaic power station according to a set period and a sequence before and after historical time, deleting wrong data, and carrying out the prediction of the set time power of the photovoltaic power station; and finally, obtaining sample data of photovoltaic power prediction according to a normalization method. 2, using an ant colony algorithm to assist the BP neural network model to perform saliency operation on the factors influencing the set power of the photovoltaic power station, using the BP neural network model to perform deep learning on weighted historical data of various factors, and performing forward propagation from an input layer to an output layer; the method can well help and support formulation and macroscopic regulation and control of a long-term plan of the power grid.

Description

Technical Field

[0001] The invention relates to a method for predicting power of a photovoltaic power station at a set time based on a network model. Background Art

[0002] In the past decade, due to the fuel crisis of traditional energy, the country has vigorously advocated the development of new energy. Subsequently, my country's photovoltaic industry has been vigorously developed like mushrooms after rain. However, the photovoltaic industry is greatly affected by the natural external environment and has great uncertainty, especially the unpredictability of long-term power, which brings great difficulties to the macro-control of the power grid.

[0003] At present, the research in this field mainly focuses on the short-term and medium-term power forecasting of photovoltaic power plants. This kind of forecasting work only meets the micro-control of the power grid and cannot provide help for the long-term planning and macro-control of this work. The traditional photovoltaic power plant set time power forecast only relies on experience or simple linear algorithms to make rough estimates and forecasts, which lacks accuracy and reliable scientific basis and cannot meet the needs of macro-precision control of the power grid.

[0004] In order to solve the above problems, a scientific and effective photovoltaic power station setting time power prediction method can help the power grid management department to carry out better macro-control and overall planning, improve the availability and reliability of photovoltaic power generation in the power grid system, reduce the destructiveness of photovoltaic power generation in the power grid system, and ensure the planned, reliable and smooth operation of the power grid.

[0005] Nowadays, information technology has achieved rapid development in artificial intelligence. The rapid development of information technology, and the deep learning algorithm therein, has provided abundant impetus for the development of artificial intelligence. This type of algorithm is particularly good at solving prediction problems, and the accuracy can basically meet the real needs. In order to further improve the accuracy, the present invention further refines the factors affecting the prediction using the significance algorithm, improves the participation of factors with high contribution to the prediction, and improves the effectiveness of problem solving to a greater extent. Therefore, the photovoltaic power station set time power prediction based on significance factor optimization and deep learning network model can provide good help and support for the formulation of long-term power grid plans and macro-control. Summary of the invention

[0006] The purpose of the present invention is to provide a method for predicting the power of a photovoltaic power station at a set time based on a network model, which can provide good assistance and support for the formulation of long-term power grid plans and macro-control.

[0007] In order to solve the above problems, the technical solution of the present invention is: A method for predicting power of a photovoltaic power station at a set time based on a network model, the method comprising the following steps: Step 1: In the order of historical time, the historical data of factors affecting the power of the photovoltaic power station at the set time are counted according to the set period, and the wrong data are deleted. Finally, the sample data of photovoltaic power prediction is obtained according to the normalization method; Step 2: Use the ant colony algorithm to assist the BP neural network model in performing significant operations on the factors that affect the power setting of the photovoltaic power station, and use the BP neural network model to conduct deep learning on the weighted historical data of various factors and perform forward propagation from the input layer to the output layer; Step 3: Compare the intermediate prediction results output by the output layer with the corresponding historical power data to perform back propagation from the output layer to the input layer. At the same time, the weights of the factors mentioned above are iteratively optimized using the gradient descent method to achieve the expected effect of focusing on weighting the significant factors. Step 4: Repeat steps 2 and 3 to perform iterative learning, so that the prediction results are closer to the corresponding historical power data, and a deep learning model with an error smaller than the expected one is obtained; Step 5: According to the current natural environment, the time factor is input into the deep learning model trained in step 5 to predict the power of the photovoltaic power station at the set time.

[0008] The beneficial effects of the present invention are as follows: by inputting the original data into the deep learning model after the optimization of the significance factor, and then calculating it through the ant colony algorithm and the BP neural network model, it is possible to effectively extract effective significance features from the historical data, and train a new prediction model, and complete the power prediction results of the photovoltaic power station setting time through the current data, effectively solving the problem of inaccurate data in the current grid setting time planning and macro-control of photovoltaic power generation. The prediction results basically meet the current technical requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The present invention will be further described below in conjunction with the accompanying drawings: Figure 1 It is a schematic diagram of the structure of the present invention. DETAILED DESCRIPTION

[0010] The following example of the historical impact of a photovoltaic power station in Yunnan and the prediction of the power data of the photovoltaic power station at the set time is used to illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content of this specification. The various details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0011] like Figure 1As shown, a method for predicting the set time power of a photovoltaic power station based on significant factor optimization and a deep learning network model includes the following steps: Step 1: In the order of historical time, the historical data of factors affecting the power of the photovoltaic power station at the set time are counted according to the set period, and the wrong data is deleted; Factors that affect the set power of a photovoltaic power station include: light radiation intensity (direct light radiation, scattered light radiation, total light radiation), atmospheric temperature, atmospheric humidity, cloud height, sun position, air pressure, wind speed and direction at a specified height, and illumination time.

[0012] Step 2: Use the ant colony algorithm to assist the BP neural network model in performing significant operations on the factors that affect the power setting of the photovoltaic power station. Use the BP neural network model to conduct deep learning on the weighted historical data of various factors and perform forward propagation from the input layer to the output layer. The data processing process is as follows: S1: Set a set of parameter weight parameters W and weight change parameters B for the data of each factor, and the two form a parameter vector X={w, b}.

[0013] S2: Set That is, the value of the i-th parameter vector in the k-th optimization iteration, is the change in the i-th parameter in the k-th optimization iteration, is the local optimization objective of the i-th parameter in the k-th optimization iteration, is the global optimization objective of the i-th parameter in the k-th optimization iteration.

[0014] S3: The above parameters are continuously optimized with the iteration of the neural network to highlight the significant factors and to obtain the optimal weight scheme. The optimization formula of the above parameters is as follows: Among them, c1 and c2 are non-negative acceleration factors of BP neural network, r1 and r2 are BP neural network debugging parameters in the (0,1) interval, and ω is the inertia weight in BP neural network.

[0015] The current natural environment short-term data is input into the short-term new energy output prediction model to predict the current new energy output data.

[0016] Step: 3: Compare the intermediate prediction results output by the output layer with the corresponding historical power data for back propagation from the output layer to the input layer. At the same time, the weights of the aforementioned factors are iteratively optimized using the gradient descent method to achieve the expected effect of focusing on weighting the significant factors.

[0017] Step 4: Repeat steps 2 and 3 to perform iterative learning so that the prediction results are closer to the corresponding historical power data and a deep learning model with an error smaller than the expected value is obtained.

[0018] Step 5: According to the current natural environment, the time factor is input into the deep learning model trained in step 4 to predict the power of the photovoltaic power station at the set time.

[0019] In the implementation process, the relevant data of long-term power and related influencing power in a certain area from 2018 to 2022 are collected as training data, and the relevant data of long-term power and related influencing power in the area in 2023 are used as the test data source. Through the above five steps, the prediction model of long-term power is obtained through deep learning network training and learning based on significant factor optimization, and then tested in the test data source. The results show that a higher prediction result is obtained.

[0020] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for predicting the power of a photovoltaic power station at a set time based on a network model, characterized in that: The method comprises the following steps: Step 1: In the order of historical time, the historical data of factors affecting the power of the photovoltaic power station at the set time are counted according to the set period, and the wrong data are deleted. Finally, the sample data of photovoltaic power prediction is obtained according to the normalization method; Step 2: Use the ant colony algorithm to assist the BP neural network model in performing significant operations on the factors that affect the power setting of the photovoltaic power station, and use the BP neural network model to conduct deep learning on the weighted historical data of various factors and perform forward propagation from the input layer to the output layer; Step 3: Compare the intermediate prediction results output by the output layer with the corresponding historical power data to perform back propagation from the output layer to the input layer. At the same time, the weights of the factors mentioned above are iteratively optimized using the gradient descent method to achieve the expected effect of focusing on weighting the significant factors. Step 4: Repeat steps 2 and 3 to perform iterative learning, so that the prediction results are closer to the corresponding historical power data, and a deep learning model with an error smaller than the expected one is obtained; Step 5: According to the current natural environment, the time factor is input into the deep learning model trained in step 4 to predict the power of the photovoltaic power station at the set time.

2. According to claim 1, a photovoltaic power station set time power prediction method based on a network model is characterized in that: In step 1, factors affecting the set power of the photovoltaic power station include: light radiation intensity, atmospheric temperature, atmospheric humidity, cloud height, sun position, air pressure, wind speed and direction at a specified height, and light duration.

3. According to claim 1, a photovoltaic power station set time power prediction method based on a network model is characterized in that: In step 2, the data processing method includes: S1: Set a set of parameter weight parameters w and weight change parameters b for the data of each factor, and the two form a parameter vector X = {w, b}; S2: Set is the value of the i-th parameter vector in the k-th optimization iteration, is the change in the i-th parameter in the k-th optimization iteration, is the local optimization objective of the i-th parameter in the k-th optimization iteration, is the global optimization objective of the i-th parameter in the k-th optimization iteration; S3: The above parameters are continuously optimized with the iteration of the neural network, highlighting the significance factors and obtaining the optimal weight scheme. The optimization formula is: Among them, c1 and c2 are non-negative acceleration factors of the BP neural network, r1 and r2 are BP neural network debugging parameters belonging to the (0, 1) interval, and ω is the inertia weight in the BP neural network.

Images