Photovoltaic power prediction regulation and control method, device, equipment and storage medium

By identifying meteorological data on similar days, the problems of low accuracy and low reliability of photovoltaic power prediction in the prior art are solved, and the accuracy of prediction results and the regulation effect of photovoltaic systems are improved.

CN118889375APending Publication Date: 2024-11-01GUANGDONG POWER GRID CO LTD +1
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410900593.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-05
Publication Date
2024-11-01

AI Technical Summary

Technical Problem

The existing photovoltaic power prediction technology has problems such as decreasing accuracy and low reliability of prediction results. It is mainly due to the direct use of annual data for prediction, which makes it difficult to dig the relationship between model prediction accuracy and meteorological characteristics.

Method used

By obtaining the meteorological data of the target prediction day, identifying the historical days similar to it, and using the meteorological data of the similar days for prediction, building a neural network model for training, generating photovoltaic power prediction results and control strategies, and avoiding the direct use of irrelevant Meteorological characteristics data.

Benefits of technology

It improves the accuracy and reliability of photovoltaic power prediction, reduces secondary prediction errors, and achieves more accurate and reliable regulation of photovoltaic systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118889375A_ABST
    Figure CN118889375A_ABST
Patent Text Reader

Abstract

The invention discloses a photovoltaic power prediction regulation and control method, device and equipment and a storage medium, and the method comprises the steps: obtaining the meteorological data of a target prediction day, recognizing a similar day corresponding to the target prediction day according to a preset similar day data set and the meteorological data of the target prediction day, and obtaining the meteorological data corresponding to the similar day; taking the similar day and the corresponding meteorological data as the input of a preset photovoltaic power prediction model, and predicting to obtain a photovoltaic power prediction result; and generating a control strategy of the photovoltaic system according to the photovoltaic power prediction result, thereby regulating and controlling the photovoltaic system through the control strategy. Wherein the preset photovoltaic power prediction model is obtained by taking the historical date and the meteorological data thereof as the input of a neural network model, taking the corresponding historical photovoltaic data as the output of the neural network model, and training the neural network 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 prediction, and in particular to a photovoltaic power prediction and control method, device, equipment and storage medium. Background Art

[0002] In the context of the clean and low-carbon energy revolution, the proportion of renewable energy represented by photovoltaics in the power system continues to rise. However, photovoltaic output is greatly affected by meteorological factors, day and night cycles and other factors, and has strong volatility and intermittency. After a high proportion of photovoltaics is connected to the power system, it poses a greater risk to the safety and stability of the power grid and increases the difficulty of dispatching and operation. Therefore, photovoltaic power prediction technology has become an important means to reduce the adverse effects of randomness and volatility of photovoltaic output.

[0003] At present, there are two main types of photovoltaic power prediction technology methods: physical prediction method and statistical prediction method. The advantage of the physical prediction method is that the model is simple, but it has disadvantages such as low prediction accuracy, complex parameter selection, and difficulty in obtaining complete and accurate weather forecast information in actual production. The statistical prediction method uses a large amount of historical data and statistical models to learn the mapping laws of things, thereby achieving prediction. However, the statistical prediction method relies on a large amount of historical data, and it is difficult to dig out the mutual connection between meteorological characteristics, which affects the prediction accuracy. Most of the existing technologies directly use the data input model throughout the year for prediction. Due to the large differences in meteorological data under different weather types, directly inputting data into the model will inevitably lead to a decrease in the prediction accuracy of the model and the problem of low reliability of the prediction results. Summary of the invention

[0004] The present invention provides a photovoltaic power prediction and control method, device, equipment and storage medium to solve the technical problems in the prior art that the photovoltaic power prediction accuracy decreases and the prediction result reliability is low.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a photovoltaic power prediction and control method, comprising:

[0006] Acquire the meteorological data of the target forecast day, and identify the similar day corresponding to the target forecast day according to the preset similar day data set and the meteorological data of the target forecast day, and acquire the meteorological data corresponding to the similar day; wherein the preset similar day data set includes each historical date and its meteorological data;

[0007] The similar day and its corresponding meteorological data are used as inputs of a preset photovoltaic power prediction model to predict a photovoltaic power prediction result;

[0008] Generating a control strategy for the photovoltaic system according to the photovoltaic power prediction result, so as to regulate the photovoltaic system through the control strategy;

[0009] The preset photovoltaic power prediction model is obtained by taking the historical date and its meteorological data as the input of the neural network model, taking the corresponding historical photovoltaic data as the output of the neural network model, and training the neural network model.

[0010] As a preferred solution, the step of obtaining the meteorological data of the target forecast day, identifying a similar day corresponding to the target forecast day based on a preset similar day data set and the meteorological data of the target forecast day, and obtaining the meteorological data corresponding to the similar day specifically includes:

[0011] Acquire meteorological data for the target forecast day; wherein the meteorological data includes: irradiance, ambient temperature and relative humidity;

[0012] Meteorological information corresponding to the meteorological data of the target forecast day is identified in a preset similar day data set, and a similar day corresponding to the target forecast day is identified based on the meteorological information, thereby obtaining the meteorological data corresponding to the similar day.

[0013] As a preferred solution, the method for constructing the preset similar day data set includes:

[0014] Historical meteorological data are obtained, and a meteorological characteristic vector is generated according to an average value and a maximum value in the historical meteorological data, and then a similar day data set is constructed according to the meteorological characteristic vector.

[0015] As a preferred solution, the method of identifying meteorological information corresponding to the meteorological data of the target forecast day in the preset similar day data set, and identifying a similar day corresponding to the target forecast day based on the meteorological information, thereby obtaining meteorological data corresponding to the similar day, specifically includes:

[0016] Set the correlation coefficient between the target forecast day and similar days;

[0017] According to the grey correlation coefficient and cosine similarity corresponding to the correlation coefficient, a comprehensive similarity index between the target forecast day and the similar day is constructed;

[0018] Meteorological information corresponding to the meteorological data of the target forecast day is identified in the preset similar day data set, and the meteorological information is screened according to the comprehensive similarity index to obtain the meteorological information of the similar day corresponding to the target forecast day, and the meteorological information is used as the meteorological data corresponding to the similar day.

[0019] As a preferred solution, before using the similar day and its corresponding meteorological data as inputs of a preset photovoltaic power prediction model, the method further includes:

[0020] Calculate the correlation coefficient according to the meteorological data corresponding to the similar day;

[0021] Performing correlation analysis on the meteorological data according to the correlation coefficient to obtain a strong correlation feature having a linear correlation with the photovoltaic output and a weak correlation feature having no linear correlation with the photovoltaic output; wherein both the strong correlation feature and the weak correlation feature belong to features in the meteorological data;

[0022] Multi-dimensional feature crossover is performed on the weakly correlated features so that a plurality of features in the weakly correlated features are cross-multiplied to obtain cross-feature data.

[0023] As a preferred solution, the similar day and its corresponding meteorological data are used as inputs of a preset photovoltaic power prediction model to predict the photovoltaic power prediction result, specifically including:

[0024] The cross-feature data and the strongly correlated features are used as inputs of a preset photovoltaic power prediction model, so that the preset photovoltaic power prediction model operates on the cross-feature data and the strongly correlated features, and outputs a photovoltaic power prediction result; wherein the preset photovoltaic power prediction model is a prediction model of a FC-HBiLSTM network.

[0025] As a preferred solution, the training method of the preset photovoltaic power prediction model specifically includes:

[0026] Acquire historical original photovoltaic data, and perform data preprocessing on the historical original photovoltaic data, so as to detect and clean abnormal data in the historical original photovoltaic data, and obtain historical photovoltaic data;

[0027] Acquire historical dates and corresponding historical meteorological data, and perform correlation analysis on the historical meteorological data according to the correlation coefficient to obtain historical strong correlation features that have linear correlation with photovoltaic output, and historical weak correlation features that do not have linear correlation with photovoltaic output;

[0028] Performing multi-dimensional feature crossover on the historical weakly correlated features, so as to cross-multiply multiple features in the historical weakly correlated features, thereby obtaining historical cross-feature data;

[0029] The historical cross-feature data and the historical strong correlation features are used as the input of the preset photovoltaic power prediction model, the corresponding historical photovoltaic data are used as the output of the neural network model, and the initial FC-HBiLSTM network prediction model is trained according to the comprehensive similarity index to obtain the preset photovoltaic power prediction model.

[0030] Accordingly, the present invention also provides a photovoltaic power prediction and control device, comprising: an acquisition module, a prediction module and a control module;

[0031] The acquisition module is used to acquire the meteorological data of the target forecast day, and identify the similar day corresponding to the target forecast day according to the preset similar day data set and the meteorological data of the target forecast day, and acquire the meteorological data corresponding to the similar day; wherein the preset similar day data set includes each historical date and its meteorological data;

[0032] The prediction module is used to use the similar day and its corresponding meteorological data as input of a preset photovoltaic power prediction model to predict and obtain a photovoltaic power prediction result;

[0033] The control module is used to generate a control strategy for the photovoltaic system according to the photovoltaic power prediction result, so as to control the photovoltaic system through the control strategy;

[0034] The preset photovoltaic power prediction model is obtained by taking the historical date and its meteorological data as the input of the neural network model, taking the corresponding historical photovoltaic data as the output of the neural network model, and training the neural network model.

[0035] Correspondingly, the present invention also provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the photovoltaic power prediction and control method as described in any one of the above.

[0036] Accordingly, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the photovoltaic power prediction and control methods described above.

[0037] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0038] The technical solution of the present invention obtains the meteorological data of the target prediction day, so that the similar day corresponding to the target prediction day and the corresponding meteorological data can be identified according to the preset similar day data set and the meteorological data of the target prediction day, so that the corresponding similar day can be used to obtain the meteorological data, avoiding the problem of low prediction accuracy caused by direct prediction using a large amount of meteorological characteristic data irrelevant to the target prediction day. At the same time, by identifying the similar day corresponding to the target prediction day, it is possible to avoid directly using the predicted data as the input of the model, thereby avoiding the error caused by the photovoltaic power prediction further amplifying the result of the secondary prediction, and by using the similar day and its corresponding meteorological data as the input of the preset photovoltaic power prediction model, the photovoltaic power prediction result of the corresponding target prediction day is predicted, and then the control strategy of the photovoltaic system is generated to regulate the photovoltaic system, thereby improving the reliability and accuracy of the prediction result, so that the photovoltaic system can be regulated more accurately and reliably. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 : A flow chart of the steps of a photovoltaic power prediction and control method provided by an embodiment of the present invention;

[0040] Figure 2 : is a Pearson correlation coefficient heat map of an embodiment of the present invention;

[0041] Figure 3 : is a multi-dimensional feature cross-cutting principle diagram of an embodiment of the present invention;

[0042] Figure 4 : is a schematic diagram of new correlation after feature crossover according to an embodiment of the present invention;

[0043] Figure 5 : is a CNN structure diagram of an embodiment of the present invention;

[0044] Figure 6 : is a schematic diagram of the BiLSTM structure of an embodiment of the present invention;

[0045] Figure 7 :It is a structural diagram of the FC-HBiLSTM model of an embodiment of the present invention;

[0046] Figure 8 : This is a result graph of similar days selected for sunny, cloudy and rainy days according to an embodiment of the present invention;

[0047] Fig. 9 : This is a comparison and prediction result diagram of each model for sunny, cloudy and rainy days according to an embodiment of the present invention;

[0048] Fig.10 : A structural diagram of a photovoltaic power prediction and control device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] Embodiment 1

[0051] Please refer to Figure 1 , a photovoltaic power prediction and control method provided by an embodiment of the present invention includes the following steps S101-S103:

[0052] Step S101: Obtain the meteorological data of the target prediction day, and identify the similar day corresponding to the target prediction day based on the preset similar day data set and the meteorological data of the target prediction day, and obtain the meteorological data corresponding to the similar day; wherein the preset similar day data set includes each historical date and its meteorological data.

[0053] As a preferred solution of this embodiment, the step of obtaining the meteorological data of the target forecast day, identifying a similar day corresponding to the target forecast day according to a preset similar day data set and the meteorological data of the target forecast day, and obtaining the meteorological data corresponding to the similar day specifically includes:

[0054] Obtain meteorological data for a target forecast day; wherein the meteorological data includes: irradiance, ambient temperature and relative humidity; identify meteorological information corresponding to the meteorological data for the target forecast day in a preset similar day data set, and identify a similar day corresponding to the target forecast day based on the meteorological information, thereby obtaining the meteorological data corresponding to the similar day.

[0055] In this embodiment, the irradiance, ambient temperature and relative humidity of the target prediction day are obtained, so that the meteorological information corresponding to the irradiance, ambient temperature and relative humidity of the target prediction day can be accurately and quickly identified in the preset similar day data set, and the date corresponding to the meteorological information can be identified through the meteorological information as a similar day corresponding to the target prediction day.

[0056] In this embodiment, by identifying meteorological information, similar days corresponding to the target prediction day can be accurately determined. At the same time, since similar days are historical data sets in the preset similar day data sets, the data in similar days are real and existing historical data, so that by subsequently predicting photovoltaic power based on meteorological data such as irradiance, ambient temperature and relative humidity of similar days, the error amplification problem caused by continuous use of prediction results for secondary prediction can be reduced.

[0057] As a preferred solution of this embodiment, the method for constructing the preset similar day data set includes:

[0058] Historical meteorological data are obtained, and a meteorological characteristic vector is generated according to an average value and a maximum value in the historical meteorological data, and then a similar day data set is constructed according to the meteorological characteristic vector.

[0059] In this embodiment, by constructing feature vectors of meteorological data for typical days (including but not limited to sunny days, cloudy days and rainy days), and adopting a method combining grey association theory with cosine similarity, a similar day data set including but not limited to sunny days, cloudy days and rainy days can be constructed.

[0060] As a preferred solution of this embodiment, the step of identifying meteorological information corresponding to the meteorological data of the target forecast day in the preset similar day data set, and identifying a similar day corresponding to the target forecast day based on the meteorological information, thereby obtaining meteorological data corresponding to the similar day, specifically includes:

[0061] A correlation coefficient between the target forecast day and similar days is set; a comprehensive similarity index between the target forecast day and similar days is constructed based on the grey correlation coefficient and cosine similarity corresponding to the correlation coefficient; meteorological information corresponding to the meteorological data of the target forecast day is identified in a preset similar day data set, and the meteorological information is screened based on the comprehensive similarity index to obtain the meteorological information of the similar day corresponding to the target forecast day, and the meteorological information is used as the meteorological data corresponding to the similar day.

[0062] In this embodiment, a comprehensive similarity index can be constructed by integrating the grey correlation degree and the cosine similarity. First, the meteorological feature vector X can be constructed by using the average and maximum values ​​of multiple meteorological features in a day. i , where X i =[x i,1 ,x i,2 ,x i,3 ,x i,4 ,x i,5, x i,6 ]Where: x i,1 ,x i,2 ,x i,3 ,x i,4 ,x i,5 ,x i,6 represents the maximum irradiance, average irradiance, maximum ambient temperature, average ambient temperature, maximum relative humidity, and average relative humidity of the i-th day. Then, the target forecast day X is calculated respectively. h Similar to X g Correlation coefficient of the nth component of meteorological data: In the formula, ζ i (n) is the correlation coefficient; Δ=|x h (n)-x g (n)|; r is the resolution coefficient, preferably, r is 0.5. The grey relational coefficient is defined as: Where N is the total number of correlation coefficients of each meteorological data component. Cosine similarity is defined as:

[0063] In this embodiment, in order to avoid the limitation of a single selection method, the numerical similarity and change trend similarity between the predicted day and the similar day sequence are comprehensively considered, and a comprehensive similarity index S is introduced. i , characterizes the overall similarity, which is defined as:

[0064] S i =αR i +(1-α)D i

[0065] Where: α is the adaptive weight coefficient. When the change trend is obvious, the value of α approaches 0, otherwise it approaches 1.

[0066] In this embodiment, by combining similarity indicators, the grey correlation degree and the cosine similarity are integrated, and similar days are selected to screen the meteorological information, so that the meteorological data corresponding to the similar days can be obtained, the correlation between the meteorological data between the similar days and the target forecast day is improved, and the accuracy of the meteorological data on the similar days is improved.

[0067] Step S102: using the similar day and its corresponding meteorological data as input of a preset photovoltaic power prediction model to predict a photovoltaic power prediction result; wherein the preset photovoltaic power prediction model is obtained by using the historical date and its meteorological data as input of a neural network model, using the corresponding historical photovoltaic data as output of the neural network model, and training the neural network model.

[0068] As a preferred solution of this embodiment, before using the similar day and its corresponding meteorological data as input of a preset photovoltaic power prediction model, the method further includes:

[0069] According to the meteorological data corresponding to the similar day, a correlation coefficient is calculated; according to the correlation coefficient, a correlation analysis is performed on the meteorological data to obtain a strong correlation feature with a linear correlation to the photovoltaic output, and a weak correlation feature with no linear correlation to the photovoltaic output; wherein the strong correlation feature and the weak correlation feature both belong to features in the meteorological data; a multi-dimensional feature crossover is performed on the weak correlation feature so that multiple features in the weak correlation feature are cross-multiplied to obtain cross feature data.

[0070] In this embodiment, the correlation coefficient is calculated by using the meteorological data corresponding to similar days. Preferably, the Pearson correlation coefficient can be used for analysis. The Pearson correlation coefficient is:

[0071]

[0072] Where P xy is the Pearson correlation coefficient; x i is the ith value of the similar day feature value; y i is the i-th predicted value; are the average values ​​of the two features.

[0073] In this embodiment, in the actual photovoltaic power prediction work, there is usually a strong linear correlation between irradiance and photovoltaic power, but there is no obvious linear correlation with other features such as wind speed, air pressure and other meteorological features, so a multi-dimensional feature crossover mechanism is introduced to process the weakly correlated feature data. In this embodiment, the feature crossover mechanism multiplies multiple features or performs Cartesian product operations to increase the nonlinear learning ability of the model, helping the model capture the nonlinear relationship between features, thereby improving the generalization ability and prediction accuracy of the model.

[0074] In actual photovoltaic power prediction, for a vector Y containing d features, 1 ,y 2 ,…y d ], the commonly used prediction expression is:

[0075]

[0076] Where: p is the predicted value, b is the offset term, is the weighted sum of d features; ω i Represents the weight of each feature.

[0077] If the predicted value p and the eigenvalue y 1 shows a strong correlation, but with the eigenvalue y 2 The correlation is weak, while the eigenvalue y 1 and the eigenvalue y 2 There is a certain correlation, from which we can know that the predicted value p and the characteristic value y 2 The relationship between 1 To express its deep connection, the formula of cross-characteristics:

[0078] p'=b'+ω 1 'y 1 +ω 3 'y 3

[0079] y3 =y 1 y 2

[0080] Among them, p' is the predicted value after feature crossover; ω i ′、ω 2 ′、ω 3 ′ is the weight corresponding to each feature after feature crossover; y 3 for y 1 ,y 2 The new eigenvalue generated after feature crossover; b′ is the offset term after crossover.

[0081] As a preferred solution of this embodiment, the similar day and its corresponding meteorological data are used as inputs of a preset photovoltaic power prediction model to predict the photovoltaic power prediction result, specifically including:

[0082] The cross-feature data and the strongly correlated features are used as inputs of a preset photovoltaic power prediction model, so that the preset photovoltaic power prediction model operates on the cross-feature data and the strongly correlated features, and outputs a photovoltaic power prediction result; wherein the preset photovoltaic power prediction model is a prediction model of the FC-HBi LSTM network.

[0083] In this embodiment, the cross-feature data and the strong correlation feature are used as the input of the preset photovoltaic power prediction model, so that the preset photovoltaic power prediction model outputs the photovoltaic power prediction result corresponding to the cross-feature data and the strong correlation feature. At the same time, the FC-HBi LSTM network prediction model is used for prediction, and the features that are weakly correlated on the surface but indirectly present deep nonlinear correlations through other features are selected according to the Pearson correlation analysis to cross, thereby enhancing the feature expression ability. Finally, the FC-HBi LSTM network is used as the prediction model, which can add an attention mechanism module on the basis of utilizing the efficient feature extraction ability of CNN and the ability of Bi LSTM to fully consider past and future information, so that the model can better capture the key information in the input sequence. It can be understood that the FC-HB i LSTM network is a network model fused by the CNN network and the BiLSTM. After the CNN extracts the features, the feature map is fused with the output of the Bi LSTM at the feature level, thereby fusion The FC-HBi LSTM network.

[0084] As a preferred solution of this embodiment, the training method of the preset photovoltaic power prediction model specifically includes:

[0085] The historical original photovoltaic data is obtained, and data preprocessing is performed on the historical original photovoltaic data, so as to detect and clean abnormal data in the historical original photovoltaic data to obtain historical photovoltaic data; the historical date and its corresponding historical meteorological data are obtained, and the historical meteorological data is subjected to correlation analysis according to the correlation coefficient to obtain historical strong correlation features with linear correlation to photovoltaic output, and historical weak correlation features with no linear correlation to photovoltaic output; multi-dimensional feature crossover is performed on the historical weak correlation features, so that multiple features in the historical weak correlation features are cross-multiplied to obtain historical cross feature data; the historical cross feature data and the historical strong correlation features are used as inputs of a preset photovoltaic power prediction model, the corresponding historical photovoltaic data are used as outputs of a neural network model, and the initial FC-HBi LSTM network prediction model is trained according to the comprehensive similarity index to obtain a preset photovoltaic power prediction model.

[0086] In this embodiment, historical raw photovoltaic data is obtained and data preprocessing is performed on the raw photovoltaic data. Preferably, the Raida criterion can be used to detect and clean abnormal data. Since photovoltaic data has strong continuity, the removed abnormal values ​​are replaced by the k nearest neighbor method, that is, the average value of the k adjacent data on both sides of the abnormal value is used for replacement.

[0087] In this embodiment, historical dates and their corresponding historical meteorological data are obtained, and correlation analysis is performed on the historical meteorological data according to the correlation coefficient. Features that are weakly correlated on the surface but indirectly present deep nonlinear correlations through other features are selected for cross-correlation to enhance the feature expression capability, thereby obtaining historical strong correlation features that have linear correlation with photovoltaic output and historical weak correlation features that do not have linear correlation with photovoltaic output. Finally, the historical cross-feature data and the historical strong correlation features are used as inputs to a preset photovoltaic power prediction model, and the corresponding historical photovoltaic data are used as outputs of a neural network model. Based on comprehensive similar day indicators, the amount of historical data required for the model is reduced, and the power prediction model is simplified to train a preset photovoltaic power prediction model.

[0088] Step S103: generating a control strategy for the photovoltaic system according to the photovoltaic power prediction result, so as to regulate the photovoltaic system through the control strategy.

[0089] In this embodiment, the control strategy of the corresponding photovoltaic system is generated according to the photovoltaic power prediction results, so that the photovoltaic system is regulated by the control strategy. By determining the control target, that is, determining the control target according to business needs, such as maximizing energy output, optimizing system efficiency, meeting grid demand, etc., the control strategy is generated so that the control strategy is designed according to the prediction results and the control target, including but not limited to adjusting the tilt angle of the photovoltaic panel, cleaning the photovoltaic panel, controlling battery charging and discharging, adjusting the inverter working mode, etc. Further, the control strategy can also be optimized using an optimization algorithm (such as genetic algorithm, particle swarm optimization, etc.) to improve system performance, and the control strategy is tested in a simulated environment to evaluate its impact on the performance of the photovoltaic system, and finally the control strategy is implemented, and the performance of the photovoltaic system is monitored in real time. According to the monitoring results and environmental changes, the control strategy is dynamically adjusted.

[0090] Furthermore, this embodiment uses the full-year data of a photovoltaic station to conduct simulation experiments. The results show that the average absolute error and root mean square error of the method proposed in this embodiment are reduced by 3.596% and 3.659% respectively, and the determination coefficient is increased by 19.2%, which verifies the feasibility and superiority of the prediction and control method proposed in this embodiment.

[0091] The selected meteorological parameters include irradiance, temperature, relative humidity, air pressure, and wind speed. A heat map is used to plot the correlation coefficients of various meteorological characteristics on photovoltaic power generation under the Pearson correlation coefficient, such as Figure 2 As shown. Figure 2 It can be seen that photovoltaic power is positively correlated with radiation, temperature and wind speed, and negatively correlated with air pressure and relative humidity. Among them, the feature that has the greatest impact on photovoltaic power generation is total radiation, and the Pearson correlation coefficient between the two is as high as 0.82, which is a very strong correlation; the correlation with temperature is second, and its correlation coefficient is 0.41, which is a medium correlation; it is very weakly correlated with air pressure, relative humidity and wind speed.

[0092] Figure 3 It is a multi-dimensional feature cross-principle diagram. Through multiplication or Cartesian product operation, a feature with weak correlation is converted into a new nonlinear feature. This not only avoids the waste of feature data with weak correlation, but also increases the nonlinear learning ability and generalization ability of the model, thereby improving the prediction accuracy.

[0093] The feature values ​​with weak correlation are crossed to obtain a new feature value, which not only contains the two independent features after crossing, but also contains the deep connection between the predicted value and the weakly correlated feature. Figure 2It can be seen that the photovoltaic power generation power is strongly correlated with the total radiation, but only moderately correlated with the ambient temperature. However, the ambient temperature and the total radiation also show a moderate correlation level. Therefore, when predicting the photovoltaic power generation power, this paper does not ignore the ambient temperature feature, which has a weak correlation with the power. Instead, it crosses the ambient temperature feature with the radiation feature to form a cross meteorological feature, such as Figure 4 This method not only improves the nonlinear learning ability of the model, but also avoids the occurrence of a large amount of data waste.

[0094] Figure 5 The structure diagram of CNN is shown in Figure 2. The parameter sharing and local sensing characteristics of convolutional neural networks reduce the number of parameters that need to be optimized and can efficiently extract input data features, thereby improving the training efficiency and scalability of the prediction model. CNN is mainly composed of convolutional layers and pooling layers, such as Figure 5 As shown in the figure, the convolution kernel in the convolution layer extracts features from the data, and the pooling layer is used to compress features and generate important feature information to improve the generalization ability of the model, so it has significant advantages for time series prediction.

[0095] The LSTM network well describes the sequence data with temporal and spatial correlation and realizes prediction based on event sequence. However, in the photovoltaic power prediction, there is a bidirectional information flow between photovoltaic data, and LSTM cannot mine the information of future moments, so the Bi LSTM network is introduced. The Bi LSTM structure is shown in Figure 6. Bi LSTM combines forward and backward information, fully learns the positive and reverse laws between historical data, and can better capture the long-term dependencies in sequence data.

[0096] In order to explore the influence of a single module in the proposed model FC-HBi LSTM (M1) on the prediction results and the effectiveness of each module, HBi LSTM (M2), CNN-Bi LSTM (M3) and Bi LSTM (M4) were constructed as comparative models for ablation experiments. The three weather types of the first date (sunny), the second date (cloudy) and the third date (rainy) were predicted respectively. The ablation experiment results are shown as follows: Figure 8 As shown (the small pictures from top to bottom are the first date, the second date and the third date), Figure 8 Each graph in the figure represents the selection results of similar days with three different weather types, and each color of each small graph represents the power curve of a certain day.

[0097] Fig. 9 : is a comparison prediction result diagram of each model for sunny, cloudy and rainy days in this embodiment, specifically, is a comparison prediction result diagram of each model for sunny, cloudy and rainy days in this embodiment ( Fig. 9The small figures in the figure are the first date, the second date and the third date from top to bottom), which is equivalent to the ablation comparison results of the three weather types.

[0098] Implementing the above embodiments has the following effects:

[0099] The technical solution of the present invention obtains the meteorological data of the target prediction day, so that the similar day corresponding to the target prediction day and the corresponding meteorological data can be identified according to the preset similar day data set and the meteorological data of the target prediction day, so that the corresponding similar day can be used to obtain the meteorological data, avoiding the problem of low prediction accuracy caused by direct prediction using a large amount of meteorological characteristic data irrelevant to the target prediction day. At the same time, by identifying the similar day corresponding to the target prediction day, it is possible to avoid directly using the predicted data as the input of the model, thereby avoiding the error caused by the photovoltaic power prediction further amplifying the result of the secondary prediction, and by using the similar day and its corresponding meteorological data as the input of the preset photovoltaic power prediction model, the photovoltaic power prediction result of the corresponding target prediction day is predicted, and then the control strategy of the photovoltaic system is generated to regulate the photovoltaic system, thereby improving the reliability and accuracy of the prediction result, so that the photovoltaic system can be regulated more accurately and reliably.

[0100] Embodiment 2

[0101] See also Fig.10 , which is a photovoltaic power prediction and control device provided by the present invention, including: an acquisition module 201, a prediction module 202 and a control module 203;

[0102] The acquisition module 201 is used to acquire the meteorological data of the target forecast day, and identify the similar day corresponding to the target forecast day according to the preset similar day data set and the meteorological data of the target forecast day, and acquire the meteorological data corresponding to the similar day; wherein the preset similar day data set includes each historical date and its meteorological data;

[0103] The prediction module 202 is used to use the similar day and its corresponding meteorological data as input of a preset photovoltaic power prediction model to predict and obtain a photovoltaic power prediction result;

[0104] The control module 203 is used to generate a control strategy for the photovoltaic system according to the photovoltaic power prediction result, so as to control the photovoltaic system through the control strategy;

[0105] The preset photovoltaic power prediction model is obtained by taking the historical date and its meteorological data as the input of the neural network model, taking the corresponding historical photovoltaic data as the output of the neural network model, and training the neural network model.

[0106] As a preferred solution, the step of obtaining the meteorological data of the target forecast day, identifying a similar day corresponding to the target forecast day based on a preset similar day data set and the meteorological data of the target forecast day, and obtaining the meteorological data corresponding to the similar day specifically includes:

[0107] Acquire meteorological data for the target forecast day; wherein the meteorological data includes: irradiance, ambient temperature and relative humidity;

[0108] Meteorological information corresponding to the meteorological data of the target forecast day is identified in a preset similar day data set, and a similar day corresponding to the target forecast day is identified based on the meteorological information, thereby obtaining the meteorological data corresponding to the similar day.

[0109] As a preferred solution, the method for constructing the preset similar day data set includes:

[0110] Historical meteorological data are obtained, and a meteorological characteristic vector is generated according to an average value and a maximum value in the historical meteorological data, and then a similar day data set is constructed according to the meteorological characteristic vector.

[0111] As a preferred solution, the method of identifying meteorological information corresponding to the meteorological data of the target forecast day in the preset similar day data set, and identifying a similar day corresponding to the target forecast day based on the meteorological information, thereby obtaining meteorological data corresponding to the similar day, specifically includes:

[0112] Set the correlation coefficient between the target forecast day and similar days;

[0113] According to the grey correlation coefficient and cosine similarity corresponding to the correlation coefficient, a comprehensive similarity index between the target forecast day and the similar day is constructed;

[0114] Meteorological information corresponding to the meteorological data of the target forecast day is identified in the preset similar day data set, and the meteorological information is screened according to the comprehensive similarity index to obtain the meteorological information of the similar day corresponding to the target forecast day, and the meteorological information is used as the meteorological data corresponding to the similar day.

[0115] As a preferred solution, before using the similar day and its corresponding meteorological data as inputs of a preset photovoltaic power prediction model, the method further includes:

[0116] Calculate the correlation coefficient according to the meteorological data corresponding to the similar day;

[0117] Performing correlation analysis on the meteorological data according to the correlation coefficient to obtain a strong correlation feature having a linear correlation with the photovoltaic output and a weak correlation feature having no linear correlation with the photovoltaic output; wherein both the strong correlation feature and the weak correlation feature belong to features in the meteorological data;

[0118] Multi-dimensional feature crossover is performed on the weakly correlated features so that a plurality of features in the weakly correlated features are cross-multiplied to obtain cross-feature data.

[0119] As a preferred solution, the similar day and its corresponding meteorological data are used as inputs of a preset photovoltaic power prediction model to predict the photovoltaic power prediction result, specifically including:

[0120] The cross-feature data and the strongly correlated features are used as inputs of a preset photovoltaic power prediction model, so that the preset photovoltaic power prediction model operates on the cross-feature data and the strongly correlated features, and outputs a photovoltaic power prediction result; wherein the preset photovoltaic power prediction model is a prediction model of a FC-HBiLSTM network.

[0121] As a preferred solution, the training method of the preset photovoltaic power prediction model specifically includes:

[0122] Acquire historical original photovoltaic data, and perform data preprocessing on the historical original photovoltaic data, so as to detect and clean abnormal data in the historical original photovoltaic data, and obtain historical photovoltaic data;

[0123] Acquire historical dates and corresponding historical meteorological data, and perform correlation analysis on the historical meteorological data according to the correlation coefficient to obtain historical strong correlation features that have linear correlation with photovoltaic output, and historical weak correlation features that do not have linear correlation with photovoltaic output;

[0124] Performing multi-dimensional feature crossover on the historical weakly correlated features, so as to cross-multiply multiple features in the historical weakly correlated features, thereby obtaining historical cross-feature data;

[0125] The historical cross-feature data and the historical strong correlation features are used as the input of the preset photovoltaic power prediction model, the corresponding historical photovoltaic data are used as the output of the neural network model, and the initial FC-HBi LSTM network prediction model is trained according to the comprehensive similarity index to obtain the preset photovoltaic power prediction model.

[0126] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0127] Implementing the above embodiments has the following effects:

[0128] The technical solution of the present invention obtains the meteorological data of the target prediction day, so that the similar day corresponding to the target prediction day and the corresponding meteorological data can be identified according to the preset similar day data set and the meteorological data of the target prediction day, so that the corresponding similar day can be used to obtain the meteorological data, avoiding the problem of low prediction accuracy caused by direct prediction using a large amount of meteorological characteristic data irrelevant to the target prediction day. At the same time, by identifying the similar day corresponding to the target prediction day, it is possible to avoid directly using the predicted data as the input of the model, thereby avoiding the error caused by the photovoltaic power prediction further amplifying the result of the secondary prediction, and by using the similar day and its corresponding meteorological data as the input of the preset photovoltaic power prediction model, the photovoltaic power prediction result of the corresponding target prediction day is predicted, and then the control strategy of the photovoltaic system is generated to regulate the photovoltaic system, thereby improving the reliability and accuracy of the prediction result, so that the photovoltaic system can be regulated more accurately and reliably.

[0129] Embodiment 3

[0130] Correspondingly, the present invention also provides a terminal device, comprising: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the photovoltaic power prediction and control method as described in any one of the above embodiments is implemented.

[0131] The terminal device of this embodiment includes: a processor, a memory, and a computer program and a computer instruction stored in the memory and executable on the processor. When the processor executes the computer program, each step in the above-mentioned embodiment 1 is implemented, for example: Figure 1 Alternatively, when the processor executes the computer program, the functions of each module / unit in the above device embodiment, such as the prediction module 202, are implemented.

[0132] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of completing specific functions, which are used to describe the execution process of the computer program in the terminal device. For example, the prediction module 202 is used to use the similar day and its corresponding meteorological data as inputs of a preset photovoltaic power prediction model to predict photovoltaic power prediction results.

[0133] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the schematic diagram is merely an example of a terminal device and does not constitute a limitation on the terminal device. The terminal device may include more or fewer components than shown in the diagram, or may combine certain components, or different components. For example, the terminal device may also include an input / output device, a network access device, a bus, etc.

[0134] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and various interfaces and lines are used to connect various parts of the entire terminal device.

[0135] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the terminal device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile terminal, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Med ia Card, SMC), a secure digital (SecureDi gita l, SD) card, a flash card (F l ash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0136] Wherein, if the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-On l yMemory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0137] Embodiment 4

[0138] Accordingly, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the photovoltaic power prediction and control method described in any one of the above embodiments.

[0139] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A photovoltaic power prediction and control method, characterized in that: include: Acquire the meteorological data of the target forecast day, and identify the similar day corresponding to the target forecast day according to the preset similar day data set and the meteorological data of the target forecast day, and acquire the meteorological data corresponding to the similar day; wherein the preset similar day data set includes each historical date and its meteorological data; The similar day and its corresponding meteorological data are used as inputs of a preset photovoltaic power prediction model to predict a photovoltaic power prediction result; Generating a control strategy for the photovoltaic system according to the photovoltaic power prediction result, so as to regulate the photovoltaic system through the control strategy; The preset photovoltaic power prediction model is obtained by taking the historical date and its meteorological data as the input of the neural network model, taking the corresponding historical photovoltaic data as the output of the neural network model, and training the neural network model.

2. A photovoltaic power prediction and control method according to claim 1, characterized in that: The step of obtaining the meteorological data of the target forecast day, identifying a similar day corresponding to the target forecast day according to a preset similar day data set and the meteorological data of the target forecast day, and obtaining the meteorological data corresponding to the similar day specifically includes: Acquire meteorological data for the target forecast day; wherein the meteorological data includes: irradiance, ambient temperature and relative humidity; Meteorological information corresponding to the meteorological data of the target forecast day is identified in a preset similar day data set, and a similar day corresponding to the target forecast day is identified based on the meteorological information, thereby obtaining the meteorological data corresponding to the similar day.

3. A photovoltaic power prediction and control method as claimed in claim 2, characterized in that: The method for constructing the preset similar day data set includes: Historical meteorological data are obtained, and a meteorological characteristic vector is generated according to an average value and a maximum value in the historical meteorological data, and then a similar day data set is constructed according to the meteorological characteristic vector.

4. A photovoltaic power prediction and control method as claimed in claim 2, characterized in that: The step of identifying meteorological information corresponding to the meteorological data of the target forecast day in the preset similar day data set, and identifying a similar day corresponding to the target forecast day based on the meteorological information, thereby obtaining meteorological data corresponding to the similar day, specifically includes: Set the correlation coefficient between the target forecast day and similar days; According to the grey correlation coefficient and cosine similarity corresponding to the correlation coefficient, a comprehensive similarity index between the target forecast day and the similar day is constructed; Meteorological information corresponding to the meteorological data of the target forecast day is identified in the preset similar day data set, and the meteorological information is screened according to the comprehensive similarity index to obtain the meteorological information of the similar day corresponding to the target forecast day, and the meteorological information is used as the meteorological data corresponding to the similar day.

5. A photovoltaic power prediction and control method as claimed in claim 4, characterized in that: Before using the similar day and its corresponding meteorological data as inputs of a preset photovoltaic power prediction model, the method further includes: Calculate the correlation coefficient according to the meteorological data corresponding to the similar day; Performing correlation analysis on the meteorological data according to the correlation coefficient to obtain a strong correlation feature having a linear correlation with the photovoltaic output and a weak correlation feature having no linear correlation with the photovoltaic output; wherein both the strong correlation feature and the weak correlation feature belong to features in the meteorological data; Multi-dimensional feature crossover is performed on the weakly correlated features, so that multiple features in the weakly correlated features are cross-multiplied, thereby obtaining cross-feature data.

6. A photovoltaic power prediction and control method as claimed in claim 5, characterized in that: The similar day and its corresponding meteorological data are used as inputs of a preset photovoltaic power prediction model to predict the photovoltaic power prediction result, specifically including: The cross-feature data and the strongly correlated features are used as inputs of a preset photovoltaic power prediction model, so that the preset photovoltaic power prediction model operates on the cross-feature data and the strongly correlated features, and outputs a photovoltaic power prediction result; wherein the preset photovoltaic power prediction model is a prediction model of a FC-HBiLSTM network.

7. A photovoltaic power prediction and control method as claimed in claim 6, characterized in that: The training method of the preset photovoltaic power prediction model specifically includes: Acquire historical original photovoltaic data, and perform data preprocessing on the historical original photovoltaic data, so as to detect and clean abnormal data in the historical original photovoltaic data, and obtain historical photovoltaic data; Acquire historical dates and corresponding historical meteorological data, and perform correlation analysis on the historical meteorological data according to the correlation coefficient to obtain historical strong correlation features that have linear correlation with photovoltaic output, and historical weak correlation features that do not have linear correlation with photovoltaic output; Performing multi-dimensional feature crossover on the historical weakly correlated features, so as to cross-multiply multiple features in the historical weakly correlated features, thereby obtaining historical cross-feature data; The historical cross-feature data and the historical strong correlation features are used as the input of the preset photovoltaic power prediction model, the corresponding historical photovoltaic data are used as the output of the neural network model, and the initial FC-HBiLSTM network prediction model is trained according to the comprehensive similarity index to obtain the preset photovoltaic power prediction model.

8. A photovoltaic power prediction and control device, characterized in that: include: Acquisition module, prediction module and regulation module; The acquisition module is used to acquire the meteorological data of the target forecast day, and identify the similar day corresponding to the target forecast day according to the preset similar day data set and the meteorological data of the target forecast day, and acquire the meteorological data corresponding to the similar day; wherein the preset similar day data set includes each historical date and its meteorological data; The prediction module is used to use the similar day and its corresponding meteorological data as input of a preset photovoltaic power prediction model to predict and obtain a photovoltaic power prediction result; The control module is used to generate a control strategy for the photovoltaic system according to the photovoltaic power prediction result, so as to control the photovoltaic system through the control strategy; The preset photovoltaic power prediction model is obtained by taking the historical date and its meteorological data as the input of the neural network model, taking the corresponding historical photovoltaic data as the output of the neural network model, and training the neural network model.

9. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the photovoltaic power prediction and control method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the photovoltaic power prediction and control method according to any one of claims 1 to 7.

Citation Information

Cited By

  • Energy efficiency prediction method and device based on solar photovoltaic panel and medium

    CN120454037A

  • Intelligent optical power prediction method and platform

    CN120638337A

  • Intelligent Optical Power Prediction Method and Platform

    CN120638337B