Method and system for medium- and long-term prediction of photovoltaic power station power generation

Abstract

The present invention provides a medium- and long-term prediction method and system for photovoltaic power generation, and relates to the technical field of photovoltaic power generation prediction. The present invention first collects photovoltaic power generation historical data including relevant factors affecting photovoltaic power generation according to weather categories, and sorts and screens the relevant factors; then clusters the screened relevant factors to obtain K clustering results, and divides a day into K similar time periods based on the clustering results; then trains a Bi‑LSTM neural network using the relevant factors and photovoltaic power generation historical data corresponding to each similar time period, thereby obtaining K medium- and long-term photovoltaic power generation prediction models; finally, performs medium- and long-term photovoltaic power generation prediction based on the K medium- and long-term photovoltaic power generation prediction models. The present invention realizes high-precision prediction of renewable energy, thereby reducing resource waste in photovoltaic power station operation and maintenance, and reducing power generation losses.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power generation prediction, and in particular to a medium- and long-term prediction method and system for photovoltaic power station power generation. Background Art

[0002] Renewable energy will become the only choice to replace fossil energy in the future due to its abundant resources, cleanness and renewability. Forecasting renewable energy can not only prevent the waste of power resources caused by various reasons such as overcapacity of power generation and mismatch between power source and power grid development speed, but also further promote the healthy development of the renewable energy industry. Studies have shown that reliable medium- and long-term forecasts of renewable energy can help the operation of relevant institutions and improve the utilization efficiency of renewable energy.

[0003] At present, renewable energy forecasting can be mainly divided into short-term renewable energy forecasting and long-term renewable energy forecasting, while there are fewer studies on medium- and long-term forecasting. Short-term renewable energy forecasting mainly focuses on ultra-short-term (several hours) forecasting and short-term forecasting (several days), but this can no longer meet the needs of social development; and when using traditional linear models for long-term forecasting of renewable energy, most models are basically difficult to solve multi-variable or multi-input problems, and long-term forecasting has the problem of "long-term cycle dependence". These factors affect the accuracy of long-term forecasting of renewable energy.

[0004] It can be seen from this that there is currently no high-precision medium- and long-term forecasting technology for renewable energy in the existing technology. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the deficiencies of the prior art, the present invention provides a method and system for medium- and long-term prediction of power generation of a photovoltaic power station, which solves the problem of low accuracy in medium- and long-term prediction of renewable energy in the prior art.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] In a first aspect, the present invention first proposes a method for medium- and long-term prediction of power generation of a photovoltaic power station, the method comprising:

[0010] Acquire photovoltaic power generation historical data by weather category, wherein the photovoltaic power generation historical data includes relevant factors affecting photovoltaic power generation and photovoltaic power generation historical data, and screen the relevant factors;

[0011] Clustering the screened related factors to obtain K clustering results, and dividing a day into K similar time periods based on the clustering results;

[0012] The Bi-LSTM neural network is trained using the relevant factors and historical data of photovoltaic power generation corresponding to each similar period to obtain K medium- and long-term photovoltaic power generation prediction models;

[0013] A medium- and long-term photovoltaic power generation prediction is performed based on the K medium- and long-term photovoltaic power generation prediction models.

[0014] Preferably, the weather categories include sunny days and cloudy days; the relevant factors include: light incident angle, radiation angle, light intensity, temperature, humidity, cloud coverage, average wind speed, air pressure, rainfall, and snowfall.

[0015] Preferably, the screening of the relevant factors includes:

[0016] The information gains of the relevant factors are calculated and sorted in descending order, and a preset number of the relevant factors at the top are selected.

[0017] Preferably, clustering the screened related factors to obtain K clustering results includes:

[0018] The K-Means clustering algorithm is used to cluster the screened related factors to obtain K clustering results.

[0019] Preferably, the classification using the K-Means clustering algorithm includes:

[0020] The optimal number of clusters K for clustering is determined by calculating the silhouette coefficient and the Davidson-Boulding index to obtain K clustering results.

[0021] In a second aspect, the present invention further proposes a medium- and long-term prediction system for power generation of a photovoltaic power station, the system comprising:

[0022] A related factor screening module is used to obtain photovoltaic power generation historical data, the photovoltaic power generation historical data includes related factors affecting photovoltaic power generation and photovoltaic power generation historical data, and screen the related factors;

[0023] A cluster analysis module, used for clustering the relevant factors after screening to obtain K clustering results, and dividing a day into K similar time periods based on the clustering results;

[0024] The model training module is used to train the Bi-LSTM neural network using the relevant factors and historical data of photovoltaic power generation corresponding to each similar period to obtain K medium- and long-term photovoltaic power generation prediction models;

[0025] The prediction and planning module is used to perform medium- and long-term prediction of photovoltaic power generation based on the K medium- and long-term photovoltaic power generation prediction models.

[0026] Preferably, the weather categories include sunny days and cloudy days; the relevant factors include: light incident angle, radiation angle, light intensity, temperature, humidity, cloud coverage, average wind speed, air pressure, rainfall, and snowfall.

[0027] Preferably, the screening of the relevant factors includes:

[0028] The information gains of the relevant factors are calculated and sorted in descending order, and a preset number of the relevant factors at the top are selected.

[0029] Preferably, clustering the screened related factors to obtain K clustering results includes:

[0030] The K-Means clustering algorithm is used to cluster the screened related factors to obtain K clustering results.

[0031] Preferably, the classification using the K-Means clustering algorithm includes:

[0032] The optimal number of clusters K for clustering is determined by calculating the silhouette coefficient and the Davidson-Boulding index to obtain K clustering results.

[0033] (III) Beneficial effects

[0034] The present invention provides a method and system for medium- and long-term prediction of power generation of a photovoltaic power station. Compared with the prior art, it has the following beneficial effects:

[0035] 1. The method for medium- and long-term prediction of photovoltaic power generation of the present invention obtains photovoltaic power generation historical data including relevant factors and photovoltaic power generation according to weather categories, then screens out relevant factors affecting photovoltaic power generation, clusters the screened relevant factors to obtain K clustering results, and divides a day into K similar time periods based on the clustering results; then trains the Bi-LSTM neural network using the relevant factors and photovoltaic power generation historical data corresponding to each similar time period to obtain K medium- and long-term photovoltaic power generation prediction models; finally, medium- and long-term prediction of photovoltaic power generation is performed based on the K medium- and long-term photovoltaic power generation prediction models. The present invention divides each day into similar time periods according to the relevant factors that have a greater impact on photovoltaic power generation, and then constructs medium- and long-term prediction models of renewable energy for different similar time periods each day, thereby improving the accuracy of medium- and long-term prediction of photovoltaic power generation, thereby reducing the waste of resources in the operation and maintenance of photovoltaic power stations and reducing power generation losses.

[0036] 2. The present invention uses the information gain method to rank and screen the importance of factors affecting photovoltaic power generation, and then uses the K-means clustering algorithm and introduces the silhouette coefficient and DBI index to divide different similar time periods. It can take into account the time dependence in the data and effectively improve the accuracy of the prediction model.

[0037] 3. The present invention uses the Bi-LSTM neural network as a medium- and long-term prediction model for photovoltaic power generation. Under the Bi-LSTM neural network structure, both previous and future information can be used in the output layer, while avoiding the occurrence of gradient vanishing and gradient exploding in the network, thereby improving the accuracy of the prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 This is a flow chart of the method for medium- and long-term prediction of power generation of a photovoltaic power station according to the present invention;

[0040] Figure 2 This is a diagram showing an embodiment of a method for mid- to long-term prediction of power generation of a photovoltaic power station according to an embodiment of the present invention;

[0041] Figure 3 1 is a structural diagram of a medium- and long-term prediction system for power generation of a photovoltaic power station in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are 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.

[0043] The embodiments of the present application provide a method and system for medium- and long-term prediction of power generation of a photovoltaic power station, thereby solving the problem of low accuracy in medium- and long-term prediction of renewable energy in the prior art, achieving high-precision prediction of renewable energy, thereby reducing resource waste in operation and maintenance of photovoltaic power stations and reducing power generation losses.

[0044] The technical solution in the embodiment of the present application is to solve the above technical problems, and the overall idea is as follows:

[0045] In order to solve the problem of low accuracy in the medium- and long-term prediction of renewable energy in the prior art, the present invention first collects historical data of photovoltaic power generation including relevant factors affecting photovoltaic power generation according to weather categories, and sorts and screens the relevant factors; then clusters the screened relevant factors to obtain K clustering results, and divides a day into K similar time periods based on the clustering results; then trains the Bi-LSTM neural network using the relevant factors and photovoltaic power generation historical data corresponding to each similar time period, thereby obtaining K medium- and long-term photovoltaic power generation prediction models; finally, performs medium- and long-term photovoltaic power generation prediction based on the K medium- and long-term photovoltaic power generation prediction models. The present invention realizes high-precision prediction of renewable energy, thereby reducing resource waste in photovoltaic power station operation and maintenance and reducing power generation losses.

[0046] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0047] Embodiment 1:

[0048] In the first aspect, the present invention first proposes a method for predicting the mid- and long-term power generation of a photovoltaic power station, see Figure 1 , the method comprising:

[0049] S1. Obtaining historical data of photovoltaic power generation by weather category, wherein the historical data of photovoltaic power generation includes relevant factors affecting photovoltaic power generation and historical data of photovoltaic power generation, and screening the relevant factors;

[0050] S2. Clustering the screened related factors to obtain K clustering results, and dividing a day into K similar time periods based on the clustering results;

[0051] S3, using the relevant factors corresponding to each similar period and the historical data of photovoltaic power generation to train the Bi-LSTM neural network, and obtain K medium- and long-term photovoltaic power generation prediction models;

[0052] S4. Perform a medium- and long-term photovoltaic power generation prediction based on the K medium- and long-term photovoltaic power generation prediction models.

[0053] It can be seen that the method for medium- and long-term prediction of power generation of a photovoltaic power station in this embodiment, after screening out the relevant factors affecting photovoltaic power generation, clusters the screened relevant factors to obtain K clustering results, and divides a day into K similar time periods based on the clustering results; uses the relevant factors corresponding to each similar time period and the historical data of photovoltaic power generation to train the Bi-LSTM neural network to obtain K medium- and long-term photovoltaic power generation prediction models; finally, medium- and long-term prediction of photovoltaic power generation is performed based on the K medium- and long-term photovoltaic power generation prediction models. The present invention divides each day into similar time periods according to the relevant factors that have a greater impact on photovoltaic power generation, and then constructs medium- and long-term prediction models of renewable energy for different similar time periods each day, thereby improving the accuracy of medium- and long-term prediction of photovoltaic power generation, thereby reducing the waste of resources in the operation and maintenance of photovoltaic power stations and reducing power generation losses.

[0054] The following is combined with Figure 1-2 , and explanations of the specific steps S1-S4 are provided to explain in detail the implementation process of an embodiment of the present invention. In this embodiment, the medium- and long-term prediction method for the power generation of a photovoltaic power station is mainly implemented by the following steps:

[0055] S1. Obtain historical data of photovoltaic power generation according to weather categories, wherein the historical data of photovoltaic power generation includes relevant factors affecting photovoltaic power generation and historical data of photovoltaic power generation, and screen the relevant factors.

[0056] The photovoltaic power generation historical data includes photovoltaic power generation and relevant factors affecting photovoltaic power generation. After the photovoltaic power generation historical data is obtained, the relevant factors affecting photovoltaic power generation in the photovoltaic power generation historical data are screened.

[0057] (1) Obtaining original photovoltaic power generation historical data related to photovoltaic power generation by weather category, and preprocessing the original photovoltaic power generation historical data. Weather categories include sunny days, cloudy days, etc. This embodiment collects data by weather category, which can solve the problems caused by weather to a certain extent. For example, in the case of sunny days, the photovoltaic power generation historical data is collected; in the case of cloudy days, the photovoltaic power generation historical data is collected, and subsequent analysis is performed on this basis.

[0058] Collect historical data on photovoltaic power generation related to solar energy resources and photovoltaic power generation in various regions, including photovoltaic data and related factors affecting photovoltaic power generation, such as: light incident angle, radiation angle, light intensity, temperature, humidity, cloud coverage, average wind speed, air pressure, rainfall, snowfall and other data. When collecting data on related factors, the corresponding values ​​of the related factors and the time points corresponding to the values ​​are used to represent them. For example, for temperature, it can be expressed as (30℃, 14:00), and the time period corresponding to the corresponding related factors can be obtained after subsequent clustering. Then, through preprocessing operations such as data cleaning and data transformation, the missing values ​​and dirty data of the original historical data are processed to obtain the preprocessed historical data of photovoltaic power generation.

[0059] (2) Sort and screen the relevant factors that affect photovoltaic power generation in the historical data of photovoltaic power generation.

[0060] For relevant factors, such as light incident angle, radiation angle, light intensity, temperature, humidity, cloud coverage, average wind speed, air pressure, rainfall, snowfall, etc., the information gain of these relevant factors is calculated, and then the information gain method is used to screen out the relevant factors that have a greater impact on solar energy resources and photovoltaic power generation for the next step of analysis. Specifically, the calculation formula is as follows:

[0061]

[0062]

[0063] IG(X,Y)=H(X)-H(X|Y)

[0064] Among them, variable x and variable y represent the known related factors affecting photovoltaic power generation and photovoltaic power generation respectively. Variable x has n possible values, which are x1, x2, ..., x n , the probability of selecting each relevant factor is p(x1), p(x2), ..., p(x n ). At this time, the entropy of x is H(X). Through the calculation of H(X), the larger the value, the greater its information contribution, and the more information can be obtained through it.

[0065] At the same time, the information entropy H(X|Y) of X obtained through Y can be calculated, and the information gain IG(X,Y) brought by the relevant factor X affecting photovoltaic power generation to the photovoltaic power generation Y can be further calculated. Then, the influencing factors are screened according to how much information each relevant factor can contribute to the photovoltaic power generation. The more information a certain influencing factor X brings, the more important this influencing factor is. Then, the relevant factors are sorted from high to low according to their importance, and then screened, and finally a preset number (a certain number is retained according to actual needs) of relevant factors with greater influence on solar energy resources and photovoltaic power generation are retained.

[0066] S2. Clustering the screened related factors to obtain K clustering results, and dividing a day into K similar time periods based on the clustering results.

[0067] After calculating the information gain in S1, a certain number of relevant factors that have a greater impact on photovoltaic power generation can be screened out, and then these relevant factors are clustered using the k-means clustering algorithm. The clustering mainly involves the following steps:

[0068] 1) Based on the selected relevant factors (such as light intensity, temperature, humidity and other indicators), the k-means clustering algorithm is used to perform cluster analysis on each time period of the day, and the analysis results are evaluated to select the optimal k value.

[0069] During clustering, we first assume an initial K value for clustering, and then determine the optimal number of clusters K by calculating the silhouette coefficient and the Davidson-Bouldin Index (DBI), namely the DBI index, to obtain K clustering results, and add labels to the clustering results.

[0070] For example, suppose that the similar time periods of a day are divided into "morning: 5-8 o'clock; 8-11 o'clock in the morning; 11-14 o'clock in the afternoon; 14-17 o'clock; evening: 17-20 o'clock; first half of the night 20-00 o'clock; second half of the night: 0-5 o'clock", that is, seven categories, and k=7. After that, the input related factors (such as light intensity, temperature, humidity) data are clustered, and then the DBI index is used to evaluate and adjust the k value. When evaluating, the smaller the DBI value, the lower the degree of dispersion, that is, the better the k value is. After continuous adjustment and optimization, a clustering result with relatively high accuracy is finally obtained, that is, the optimal K value. Finally, based on the clustering results, the basic characteristics of the light intensity, temperature, and humidity data corresponding to each type of similar time period are analyzed for subsequent predictions.

[0071] 2) Combine the relevant factors corresponding to the similar time periods of each cluster (for example, the relevant factors corresponding to a similar time period mainly include light intensity, temperature, humidity, etc.), use visualization tools to perform visual analysis on each time period, and summarize the data characteristics of the relevant factors.

[0072] S3. Use the relevant factors and historical data of photovoltaic power generation corresponding to each similar period to train the Bi-LSTM neural network to obtain K medium- and long-term photovoltaic power generation prediction models.

[0073] The Bi-LSTM neural network is as follows:

[0074] f t =σ(W f ·[ht-1 , x t ]+b f )

[0075] i t =a(W i ·[h t-1 , x t ]+b i )

[0076]

[0077]

[0078] o t =σ(W o ·[h t-1 , x t ]+b o )

[0079] h t =o t ⊙tanh(C t )

[0080]

[0081] Among them, x t represents the input of the neuron at time t, that is, based on the K-means clustering in the previous step, the relevant factor data sets of each similar period under different weather categories are input into the Bi-LSTM model for training; h t represents the output at time t, C t represents the current neuron state, W is the weight matrix, b is the bias, σ is the Sigmoid function, ⊙ represents the element-wise product, represents the output of the last forward layer, represents the output of the last backward layer, H t is the final output of the network, that is, the predicted value of photovoltaic power generation. Specifically:

[0082] ① Data input. The relevant factor data screened in S1 above is used as the input of Bi-LSTM neural network training to predict the photovoltaic power generation. Specifically, the classified photovoltaic power generation history data is set and arranged in time series, and then the first 75% of the time series data is used as the training set, and the last 25% is used as the test set.

[0083]

[0084]

[0085] Wherein, train represents the training set, and test represents the test set. Each column represents a relevant factor affecting photovoltaic power generation selected in the above step S1. There are n variables in total, and each row represents a piece of data.

[0086] ② Normalize the data. Use the following formula to normalize the data.

[0087]

[0088] Among them, p i,j Indicates the original value, min indicates p i,j The minimum value of the column, max means p i,j The maximum value of the column, p′ i,j If it is the standardized data, train′ represents the standardized training set and test′ represents the standardized test set.

[0089] ③Train the prediction model. Set the number of Bi-LSTM hidden layers and the number of neurons in the neural network, the number of input and output layers and the number of neurons in each layer, the regularization parameter, the dropout parameter, and select the Adam gradient descent algorithm to optimize the network. According to the clustering results, input the training set and train to obtain the prediction model.

[0090] ④ Set the time series of the Bi-LSTM neural network, specify the training options, and use the observed value at the previous moment to predict the predicted value at the next moment.

[0091] ⑤ Test and evaluate the model obtained above. Input the test set into the prediction model to obtain the prediction results. At the same time, use the root mean square error RMSE, mean absolute error MAE and mean absolute error percentage MAPE to evaluate the prediction results. The smaller their values ​​are, the higher the prediction accuracy is. Specifically, the calculation formula is as follows:

[0092]

[0093]

[0094]

[0095] Among them, M is the total number of data samples, y m is the true value of the historical data of photovoltaic power generation, is the predicted value of the historical data of photovoltaic power generation.

[0096] ⑥ Model optimization. Based on the prediction results, adjust the parameters of the prediction model obtained in step ③ above until the renewable energy prediction model with the highest prediction accuracy is obtained. Specifically, compare the predicted value with the real data, and continuously correct the model parameters according to the degree of proximity between the predicted value and the real value, and finally select the model with the highest accuracy (that is, the predicted value is closest to the real value) as the optimal model. When the network performance meets the expected accuracy (that is, the error between the predicted value and the real value meets the preset requirements), the output is the medium- and long-term photovoltaic power generation prediction model corresponding to the label type of similar time period.

[0097] ⑦ Repeat the above steps continuously, and finally obtain K (that is, the number of optimal clustering results, that is, the number of similar time periods divided into a day) medium- and long-term photovoltaic power generation prediction models.

[0098] S4. Perform a medium- and long-term photovoltaic power generation prediction based on the K medium- and long-term photovoltaic power generation prediction models.

[0099] Based on the K optimal medium- and long-term photovoltaic power generation prediction models finally obtained above, the medium- and long-term photovoltaic power generation of relevant institutions is predicted. For example, by selecting the medium- and long-term photovoltaic power generation prediction model corresponding to the similar period at noon in a day (noon: 11:00-14:00), and inputting the relevant factor data corresponding to the model, the predicted photovoltaic power generation at the similar period at noon can be obtained.

[0100] Finally, based on the prediction results, the changing trends of photovoltaic power generation in various regions are analyzed to provide a reference for the formulation of photovoltaic power generation planning and related scheduling in various regions.

[0101] At this point, the entire process of the method for medium- and long-term prediction of power generation of a photovoltaic power station in this embodiment is completed.

[0102] Embodiment 2:

[0103] In a second aspect, the present invention also provides a medium- and long-term prediction system for photovoltaic power station power generation, see Figure 3 The system comprises:

[0104] A related factor screening module is used to obtain photovoltaic power generation historical data, the photovoltaic power generation historical data includes related factors affecting photovoltaic power generation and photovoltaic power generation historical data, and screen the related factors;

[0105] A cluster analysis module, used for clustering the relevant factors after screening to obtain K clustering results, and dividing a day into K similar time periods based on the clustering results;

[0106] The model training module is used to train the Bi-LSTM neural network using the relevant factors and historical data of photovoltaic power generation corresponding to each similar period to obtain K medium- and long-term photovoltaic power generation prediction models;

[0107] The prediction and planning module is used to perform medium- and long-term prediction of photovoltaic power generation based on the K medium- and long-term photovoltaic power generation prediction models.

[0108] Optionally, the weather categories include sunny and cloudy days; the relevant factors include: light incident angle, radiation angle, light intensity, temperature, humidity, cloud coverage, average wind speed, air pressure, rainfall, and snowfall.

[0109] Optionally, the screening of the relevant factors includes:

[0110] The information gains of the relevant factors are calculated and sorted in descending order, and a preset number of the relevant factors at the top are selected.

[0111] Optionally, clustering the screened related factors to obtain K clustering results includes:

[0112] The K-Means clustering algorithm is used to cluster the screened related factors to obtain K clustering results.

[0113] Optionally, the classification using the K-Means clustering algorithm includes:

[0114] The optimal number of clusters K for clustering is determined by calculating the silhouette coefficient and DBI index to obtain K clustering results.

[0115] It can be understood that the medium- and long-term prediction system for power generation of a photovoltaic power station provided in an embodiment of the present invention corresponds to the medium- and long-term prediction method for power generation of a photovoltaic power station mentioned above, and the explanations, examples, beneficial effects and other parts of the relevant contents can refer to the corresponding contents in the medium- and long-term prediction method for power generation of a photovoltaic power station, and will not be repeated here.

[0116] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0117] 1. The method for medium- and long-term prediction of photovoltaic power generation of the present invention obtains photovoltaic power generation historical data including relevant factors and photovoltaic power generation according to weather categories, then screens out relevant factors affecting photovoltaic power generation, clusters the screened relevant factors to obtain K clustering results, and divides a day into K similar time periods based on the clustering results; then trains the Bi-LSTM neural network using the relevant factors and photovoltaic power generation historical data corresponding to each similar time period to obtain K medium- and long-term photovoltaic power generation prediction models; finally, medium- and long-term prediction of photovoltaic power generation is performed based on the K medium- and long-term photovoltaic power generation prediction models. The present invention divides each day into similar time periods according to the relevant factors that have a greater impact on photovoltaic power generation, and then constructs medium- and long-term prediction models of renewable energy for different similar time periods each day, thereby improving the accuracy of medium- and long-term prediction of photovoltaic power generation, thereby reducing the waste of resources in the operation and maintenance of photovoltaic power stations and reducing power generation losses.

[0118] 2. The present invention uses the information gain method to rank and screen the importance of factors affecting photovoltaic power generation, and then uses the K-means clustering algorithm and introduces the silhouette coefficient and DBI index to divide different similar time periods. It can take into account the time dependence in the data and effectively improve the accuracy of the prediction model.

[0119] 3. The present invention uses the Bi-LSTM neural network as a medium- and long-term prediction model for photovoltaic power generation. Under the Bi-LSTM neural network structure, both previous and future information can be used in the output layer, while avoiding the occurrence of gradient vanishing and gradient exploding in the network, thereby improving the accuracy of the prediction.

[0120] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0121] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for medium- and long-term prediction of power generation of a photovoltaic power station, characterized in that: The method comprises: Acquire historical data of photovoltaic power generation by weather category, wherein the historical data of photovoltaic power generation includes relevant factors affecting photovoltaic power generation and photovoltaic power generation, and screen the relevant factors; Clustering the filtered related factors according to weather categories to obtain K clustering results, and dividing a day into K similar time periods based on the clustering results; The Bi-LSTM neural network is trained using the relevant factors and historical data of photovoltaic power generation corresponding to each similar period to obtain K medium- and long-term photovoltaic power generation prediction models; Performing a medium- and long-term photovoltaic power generation prediction based on the K medium- and long-term photovoltaic power generation prediction models; The photovoltaic power generation historical data is obtained by weather category classification, and the photovoltaic power generation historical data includes relevant factors affecting photovoltaic power generation and photovoltaic power generation, and the relevant factors are screened, including: The original photovoltaic power generation historical data related to photovoltaic power generation is obtained by weather category classification, and these original photovoltaic power generation historical data are preprocessed; the photovoltaic power generation historical data related to solar energy resources, photovoltaic power generation, etc. in various regions are collected, which includes photovoltaic data and related factors affecting photovoltaic power generation; when collecting related factor data, the numerical value corresponding to the related factor and the time point corresponding to the numerical value are used to represent it; the related factors affecting photovoltaic power generation in the photovoltaic power generation historical data are sorted and screened; for the related factors, the information gain of the related factors is calculated, and they are sorted in descending order, and the preset number of related factors ranked in front are selected, and the related factors with greater influence on solar energy resources and photovoltaic power generation are screened out by the information gain method for the next step of analysis.

2. The method according to claim 1, characterized in that The weather categories include sunny days and cloudy days; the relevant factors include: light incident angle, radiation angle, light intensity, temperature, humidity, cloud coverage, average wind speed, air pressure, rainfall, and snowfall.

3. The method according to claim 1, characterized in that The clustering of the screened related factors to obtain K clustering results includes: The K-Means clustering algorithm is used to cluster the screened related factors to obtain K clustering results.

4. The method according to claim 3, characterized in that The classification using K-Means clustering algorithm includes: The optimal number of clusters K for clustering is determined by calculating the silhouette coefficient and the Davidson-Boulding index to obtain K clustering results.

5. A medium- and long-term prediction system for photovoltaic power station power generation, characterized in that: The system comprises: A related factor screening module is used to obtain photovoltaic power generation history data by weather category, wherein the photovoltaic power generation history data includes related factors affecting photovoltaic power generation and photovoltaic power generation, and to screen the related factors; A cluster analysis module, used for clustering the relevant factors after screening to obtain K clustering results, and dividing a day into K similar time periods based on the clustering results; The model training module is used to train the Bi-LSTM neural network using the relevant factors and historical data of photovoltaic power generation corresponding to each similar period to obtain K medium- and long-term photovoltaic power generation prediction models; A prediction and planning module, used for performing medium- and long-term prediction of photovoltaic power generation based on the K medium- and long-term photovoltaic power generation prediction models; The relevant factor screening module is specifically used for: The original photovoltaic power generation historical data related to photovoltaic power generation is obtained by weather category classification, and these original photovoltaic power generation historical data are preprocessed; the photovoltaic power generation historical data related to solar energy resources, photovoltaic power generation, etc. in various regions are collected, which includes photovoltaic data and related factors affecting photovoltaic power generation; when collecting related factor data, the numerical value corresponding to the related factor and the time point corresponding to the numerical value are used to represent it; the related factors affecting photovoltaic power generation in the photovoltaic power generation historical data are sorted and screened; for the related factors, the information gain of the related factors is calculated, and they are sorted in descending order, and the preset number of related factors ranked in front are selected, and the related factors with greater influence on solar energy resources and photovoltaic power generation are screened out by the information gain method for the next step of analysis.

6. The system according to claim 5, characterized in that The weather categories include sunny days and cloudy days; the relevant factors include: light incident angle, radiation angle, light intensity, temperature, humidity, cloud coverage, average wind speed, air pressure, rainfall, and snowfall.

7. The system according to claim 5, characterized in that The clustering of the screened related factors to obtain K clustering results includes: The K-Means clustering algorithm is used to cluster the screened related factors to obtain K clustering results.

8. The system according to claim 7, characterized in that The classification using K-Means clustering algorithm includes: The optimal number of clusters K for clustering is determined by calculating the silhouette coefficient and the Davidson-Boulding index to obtain K clustering results.

Images