Photovoltaic power generation prediction method, device, equipment and medium

Abstract

The invention relates to a photovoltaic power generation prediction method, device, equipment and medium, and the method comprises the steps: obtaining the meteorological data of a region where a to-be-predicted solar photovoltaic power station is located; inputting the meteorological data into a photovoltaic power generation prediction model to obtain the generating capacity of the solar photovoltaic power station; wherein the photovoltaic power generation prediction model is obtained by the following steps: collecting historical meteorological data and generating capacity data of a solar photovoltaic power station to be predicted, extracting season information and humidity level information from the historical meteorological data, and performing feature division on the historical meteorological data to obtain a plurality of groups of feature data sets; taking the plurality of groups of feature data sets as inputs of a plurality of machine learning models, training the plurality of machine learning models, and after training is completed, selecting the machine learning model with the optimal performance index and the most appropriate feature data set; and taking the machine learning model with the optimal performance index as a photovoltaic power generation prediction model. The prediction precision and efficiency of the photovoltaic generating capacity can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of photovoltaic power generation prediction, in particular to a photovoltaic power generation prediction method, device, equipment and medium. BACKGROUND

[0002] With the growth of population and technological progress, energy demand is increasing, and there is an urgent need to shift from traditional fossil fuels to renewable energy systems. As the main form of renewable energy, solar photovoltaic (PV) systems, although sustainable, have fluctuating power generation, making stable power supply a major problem. Therefore, accurate prediction of solar photovoltaic power generation is crucial to improve the competitiveness of photovoltaic power plants. Solar power generation is affected by various meteorological factors (such as humidity, seasonal changes, etc.), and how to effectively integrate these factors into machine learning models to improve prediction accuracy is a major challenge in current research. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a photovoltaic power generation prediction method, device, equipment and medium, which can improve the prediction accuracy and efficiency of photovoltaic power generation.

[0004] The technical solution adopted by the present application to solve its technical problem is to provide a photovoltaic power generation prediction method, comprising the following steps:

[0005] Obtain meteorological data of the area where the solar photovoltaic power plant to be predicted is located;

[0006] Input the meteorological data into a photovoltaic power generation prediction model to obtain the power generation of the solar photovoltaic power plant;

[0007] Wherein, the photovoltaic power generation prediction model is obtained by the following way:

[0008] Collect historical meteorological data and power generation data of the solar photovoltaic power plant to be predicted;

[0009] Extract seasonal information and humidity level information from the historical meteorological data, and preprocess the historical meteorological data and power generation data;

[0010] Perform exploratory data analysis on the preprocessed historical meteorological data and power generation data;

[0011] According to the results of the exploratory data analysis, the historical meteorological data is divided into feature data sets;

[0012] The several groups of feature data sets are respectively taken as inputs of several machine learning models, the several machine learning models are trained, and after the training is completed, the machine learning model with the optimal performance index and the most suitable feature data set are selected; and the machine learning model with the optimal performance index is taken as the photovoltaic power generation prediction model.

[0013] The meteorological data at least include one of the following data: sunshine, radiation, temperature, humidity, air pressure and wind speed.

[0014] The seasonal information and humidity level information are extracted from the historical meteorological data, specifically: the data from March 1 to May 31 is marked as spring, the data from June 1 to August 31 is marked as summer, the data from September 1 to November 30 is marked as autumn, and the data from December 1 to February 29 is marked as winter; the humidity level of the data with humidity above 60 is marked as humid, and the humidity level of the data with humidity less than 60 is marked as dry.

[0015] The several groups of feature data sets include: a first feature data set, a second feature data set, a third feature data set, a fourth feature data set, a fifth feature data set and a sixth feature data set, wherein the first feature data set only contains time, the second feature data set only contains meteorological data, the third feature data set combines time and meteorological data, the fourth feature data set adds seasonal information on the basis of the third feature data set, the fifth feature data set adds humidity level information on the basis of the third feature data set, and the sixth feature data set adds seasonal information on the basis of the fifth feature data set.

[0016] The several machine learning models include a support vector regression model, an extreme gradient boosting regression model and a HIST gradient boosting regression model.

[0017] The machine learning model with the optimal performance index is selected, specifically: a plurality of performance indexes of the several machine learning models after the training is completed are calculated, the plurality of performance indexes of each machine learning model are weighted and summed, and the machine learning model with the highest weighted sum is selected as the machine learning model with the optimal performance index.

[0018] Before the machine learning model with the optimal performance index is taken as the photovoltaic power generation prediction model, the method further includes:

[0019] The machine learning model with the optimal performance index is subjected to uncertainty analysis based on p factors and q factors.

[0020] The technical solution adopted by the application to solve the technical problems is: a photovoltaic power generation prediction device is provided, which includes:

[0021] An acquisition module is configured to acquire meteorological data of a region where a solar photovoltaic power station is located;

[0022] A prediction module is configured to input the meteorological data into a photovoltaic power generation prediction model to obtain power generation of the solar photovoltaic power station.

[0023] The photovoltaic power generation prediction model is obtained by the following method:

[0024] Historical meteorological data and power generation data of the solar photovoltaic power station to be predicted are collected.

[0025] Seasonal information and humidity level information are extracted from the historical meteorological data, and the historical meteorological data and power generation data are preprocessed.

[0026] Exploratory data analysis is performed on the preprocessed historical meteorological data and power generation data.

[0027] According to the result of the exploratory data analysis, the historical meteorological data is divided into features to obtain a plurality of feature data sets.

[0028] The plurality of feature data sets are respectively input into a plurality of machine learning models, the plurality of machine learning models are trained, and after the training is completed, a machine learning model with the optimal performance index and the most suitable feature data set are selected; the machine learning model with the optimal performance index is used as the photovoltaic power generation prediction model.

[0029] The technical solution adopted by the present application to solve its technical problems is to provide an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the photovoltaic power generation prediction method.

[0030] The technical solution adopted by the present application to solve its technical problems is to provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the steps of the photovoltaic power generation prediction method.

[0031] Advantages

[0032] Compared with the prior art, the present application has the following advantages and positive effects: the present application uses multiple performance indicators to select the best model and meaningful feature set, the selected best model has wide applicability, can effectively evaluate the performance of large-scale solar photovoltaic power generation facilities, and considers actual factors in the feature set, so that even if environmental conditions change, the model of the present application can make accurate predictions, improving the prediction accuracy of photovoltaic power generation. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 is a flow chart of the photovoltaic power generation prediction method of the first embodiment of the present application;

[0034] Figure 2 is a flow chart of the construction of the photovoltaic power generation prediction model in the first embodiment of the present application;

[0035] Figure 3 is a schematic diagram of the support vector regression model in the first embodiment of the present application;

[0036] Figure 4 is a flow chart of the selection of the machine learning model with the optimal performance index in the first embodiment of the present application. DETAILED DESCRIPTION

[0037] The present application will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present application and not used to limit the scope of the present application. In addition, it should be understood that after reading the content taught by the present application, those skilled in the art can make various modifications or changes to the present application, and these equivalent forms also fall within the scope defined by the appended claims of the present application.

[0038] The first embodiment of the present application relates to a photovoltaic power generation prediction method, as shown in Figure 1 , comprising the following steps:

[0039] Step 1, obtaining meteorological data of the area where the solar photovoltaic power station to be predicted is located;

[0040] Step 2, inputting the meteorological data into a photovoltaic power generation prediction model to obtain the power generation of the solar photovoltaic power station.

[0041] As shown in Figure 2 , the photovoltaic power generation prediction model is obtained by the following way:

[0042] (1) Collecting historical meteorological data and power generation data of the solar photovoltaic power station to be predicted.

[0043] In the present embodiment, the historical meteorological data and power generation data of the solar photovoltaic power station to be predicted are collected. The above data will be used later to train a machine learning model for accurate power generation prediction. Among them, the historical meteorological data of the solar photovoltaic power station to be predicted includes:

[0044] Time arranged in chronological order, including month, day, hour and year, which is used to generate seasonal information, and the generated seasonal information can improve the efficiency of the prediction system.

[0045] Meteorological data includes sunshine, radiation, temperature, humidity, air pressure and wind speed, etc. which describe the atmospheric characteristics that significantly affect the power generation of the solar photovoltaic power station.

[0046] (2) Extract seasonal information and humidity level information from the historical weather data and pre-process the historical weather data and power generation data.

[0047] Before addressing the data cleaning and scaling issues, the present embodiment extracts seasonal information and humidity level information from the time and weather data. The present embodiment delineates four seasons: spring (March 1st to May 31st), summer (June 1st to August 31st), autumn (September 1st to November 30th) and winter (December 1st to February 28th, February 29th in a leap year), so the data from March 1st to May 31st is labeled as spring, the data from June 1st to August 31st is labeled as summer, the data from September 1st to November 30th is labeled as autumn, and the data from December 1st to February 29th is labeled as winter. In addition, the humidity level information in the present embodiment is divided into dry or humid; if the humidity is 60 or higher, the humidity level is labeled as humid, and if the humidity is less than 60, the humidity level is labeled as dry.

[0048] After that, in order to ensure high-quality data before applying the ML model, a series of steps should be performed, including cleaning, labeling and scaling, which include: deleting null and duplicate values; converting non-numeric variables into numeric representations; normalizing the data, which requires scaling the features to a standard range, usually [0, 1] or [-1, 1], to ensure that each feature contributes equally to the model performance.

[0049] Z-score normalization, also known as standardization, can be used in the present embodiment to convert data to a mean of 0 and a standard deviation of 1. The Z-score normalization equation is given by:

[0050]

[0051] where x i is the original value of feature i in the dataset; x i,std is the value of the Z-score, is the mean of the dataset, σ i is the standard deviation of feature i in the dataset.

[0052] (3) Exploratory data analysis on the pre-processed historical weather data and power generation data.

[0053] Exploratory data analysis (EDA) is crucial in revealing broad patterns in the data and identifying outliers and unexpected data elements. It is a crucial step in any data analysis process. In the present embodiment, the EDA tree technique will be used.

[0054] EDA often uses histograms to display the distribution of a single continuous variable. This is achieved by dividing the data into intervals and computing the frequency of observations in each interval. Histograms are valuable for identifying central tendencies, detecting outliers, and understanding the shape of the data distribution (e.g., whether it is skewed or normal). The frequency Frequency in a histogram can be expressed as:

[0055]

[0056] where NOOB represents the number of observations in Bin and TNO represents the total number of observations.

[0057] EDA also uses scatter plots to display data points in a Cartesian plane, illustrating the relationship between two continuous variables. This technique helps identify patterns, trends, clusters, and outliers in the data.

[0058] (4) Feature division is performed on the historical meteorological data according to the results of the exploratory data analysis, obtaining several groups of feature data sets.

[0059] The main purpose of this step is to explore and analyze the impact of time and meteorological data on the performance of the solar photovoltaic prediction system. Therefore, this embodiment can be divided into six groups of feature data sets, in which:

[0060] The first feature data set S1 only contains the original time-ordered data, i.e., S1 = {M, D, H, Y}, where M, D, H, and Y represent month, day, hour, and year, respectively.

[0061] The second feature data set S2 only contains the original meteorological data, i.e., S2 = {R, S, W, P, T, Hu}, where R, S, W, P, T, and Hu are radiation, sunshine, wind speed, air pressure, temperature, and humidity, respectively.

[0062] The third feature data set S3 combines the original time series and meteorological data, i.e., S3 = S1 ∪ S2 = {M, D, H, Y, R, S, W, P, T, Hu}.

[0063] The fourth feature data set S4 represents the third feature data set S3 with added seasonal information, i.e., S4 = {M, D, H, Y, R, S, W, P, T, Hu, Se}, where Se represents seasonal information, taking values 1 to 4 representing autumn, winter, spring, and summer.

[0064] The fifth feature data set S5 represents the third feature data set S3 with added humidity level information, i.e., S5 = {M, D, H, Y, R, S, W, P, T, Hu, HL}, where HL represents the humidity level, taking values 0 or 1 representing dry and wet states, respectively.

[0065] The sixth feature dataset S6 represents the fifth feature dataset S5 to which seasonal information is added, i.e. S6 = {M, D, H, Y, R, S, W, P, T, Hu, HL, Se}.

[0066] (5) The plurality of feature datasets are respectively taken as inputs of a plurality of machine learning models, the plurality of machine learning models are trained, and after the training is completed, the machine learning model with the optimal performance index and the most suitable feature dataset are selected.

[0067] The machine learning models in this embodiment include a support vector regression model (SVR), an extreme gradient boosting regression model (XGBoost), and a HIST gradient boosting regression model (HistGradientBoosting). The machine learning model with the optimal performance index and the most suitable feature dataset are determined by comparing the three trained machine learning models.

[0068] SVR, XGBoost and HistGradientBoosting are well-known machine learning models that have proven their ability to handle data and complex relationships in non-linearities. They can also effectively manage large datasets. In fact, predicting the power generation of a solar photovoltaic power plant is highly non-linear and depends on many factors such as time, solar radiation and many other meteorological variables. The SVR model contains many kernel types that can effectively handle non-linear relationships by transforming the input data into a higher-dimensional space in which linear relationships can be captured. They have sufficient performance in handling high-dimensional feature spaces and are not susceptible to overfitting, especially under strict regularization parameters. XGBoost is an ensemble learning that integrates weak learners to build a strong predictive model. It can support structured data such as weather conditions (temperature, humidity, wind speed) and time-related features (time of day, day of week, season), which form the main components of solar power generation prediction. Its ability to handle complex interactions between features helps improve prediction accuracy. In addition, its ability to handle missing data and outliers makes it more suitable for real-world datasets that may have many inconsistencies and gaps. HistGradientBoosting is similar to XGBoost and is based on a histogram-based algorithm that is faster when handling large datasets. In addition, HistGradientBoosting also provides insights into feature importance, which helps to understand the relationship between weather conditions, time-related factors and solar power generation.

[0069] The goal of SVR is to estimate a function that captures the relationship between inputs and target variables while maintaining an acceptable tolerance of error. The goal of SVR is not to minimize the total error, but to fit the data within a specific range (see Figure 3). Data points falling within this range are considered acceptable, and the model strives to include as many data points within this range as possible while keeping the model as flat as possible. SVR includes the following parameters:

[0070] Kernel: This parameter specifies the type of kernel function used to analyze the data. Common options include sigmoid, polynomial (poly), radial basis function (RBF), and linear. The RBF kernel is often the preferred choice for non-linear data.

[0071] Regularization parameter (C): This parameter controls the trade-off between minimizing model complexity and achieving low training error. Higher C values can lead to overfitting by reducing training error, while lower values can increase the model's complexity.

[0072] ∈: This represents the tolerance range for errors that are not penalized (see Figure 3 ). Smaller epsilon values indicate that the model is more sensitive, as it allows for smaller error magnitudes.

[0073] Kernel coefficient (gamma): Smaller epsilon values indicate that the model is more sensitive, as it allows for smaller error magnitudes.

[0074] XGBoost regression is a powerful machine learning technique that excels at predicting continuous target variables. It employs gradient boosting of decision trees, iteratively building a set of trees to achieve higher prediction accuracy. Known for its regularization capabilities and proficiency, XGBoost is the preferred choice for regression tasks when dealing with large datasets. By adjusting its hyperparameters, its performance can be significantly improved, including:

[0075] Booster: This parameter specifies the type of boosting model to use.

[0076] n estimators : This parameter specifies the number of boosting rounds or trees to build during training. While higher numbers can improve model performance, they can also increase the likelihood of overfitting and raise computational costs.

[0077] Learning rate: Also known as eta, this parameter helps avoid overfitting by reducing the step size. Lower values typically result in better performance, although they often require more boosting rounds.

[0078] Maximum depth: This parameter determines the maximum depth of the trees. Deeper trees can capture more complex patterns; however, they also have a higher risk of overfitting

[0079] Minimum child weight: This parameter sets the minimum sum of instance weights required for a child node. Higher values require more samples to build child nodes, which helps prevent overfitting.

[0080] Subsample: This parameter represents the percentage of samples used in the growth of each tree. It helps prevent overfitting, and the optimal value is typically in the range of 0.5 to 1.0.

[0081] Colsample Bytree: This parameter represents the features used in the construction of each tree. It helps reduce overfitting and can take values ​​between 0 and 1. When set to 0.5, only 50% of the features are randomly sampled for each tree. The default value of 1.0 means that all available features are used for each tree.

[0082] Gamma: This parameter adjusts the minimum loss reduction required to further split the tree. By reducing the number of splits, a higher gamma value makes the algorithm more conservative and may help avoid overfitting. Possible values ​​range from 0 (allowing more splits) to positive floating-point numbers (increasing the split threshold to prevent overfitting).

[0083] Lambda (L2 regularization): This parameter helps prevent overfitting by regularizing the weights of the model.

[0084] Alpha (L1 regularization): This regularization technique adds a penalty based on the absolute value of the weights, which promotes the sparsity of the model.

[0085] Non-negative floating-point values ​​can be used for alpha and lambda (where 0 indicates no regularization). To avoid overfitting, lambda penalizes the sum of squared weights, while alpha penalizes the sum of absolute weights. The higher the values ​​of these parameters, the stronger the regularization.

[0086] HGBoost is a high-level adaptation of gradient boosting that leverages histogram-based methods to efficiently process data and binary continuous features, thereby reducing training time. This method is designed for superior computational speed and scalability on large datasets. Unlike XGBoost, which primarily relies on exact and approximation techniques, HGBoost mainly uses histogram-based augmentation, and its hyperparameters are as follows:

[0087] learning_rate: This parameter adjusts the contribution of each tree; larger values ​​require more trees to achieve the desired performance. It typically varies between 0.01 and 0.3.

[0088] n_estimators: This parameter represents the number of enhancement stages (trees). Performance can usually be improved by increasing the number of trees, typically between 100 and 1000.

[0089] `max_depth`: This parameter represents the maximum depth of the tree and adjusts the complexity of the model. It typically varies between 3 and 10.

[0090] min_samples_split: This parameter specifies the minimum number of samples required to split a node; helps to avoid overfitting. It usually varies between 20 to 20.

[0091] max_bins: This parameter represents the number of bins used for discretizing continuous features; this affects computation and detail. It usually ranges between 255 to 1024.

[0092] min_samples_leaf: This parameter controls model complexity by setting the minimum number of samples required to be at a leaf node. It usually varies between 2 to 10.

[0093] As shown in Figure 4 , the present embodiment trains the above three machine learning models using the above six feature datasets respectively, after training, calculates the performance indicators such as R2, MAE, MAPE, RMSE and NRMSE of each trained machine learning model, and performs weighted summation on multiple performance indicators of each machine learning model, and selects the machine learning model with the highest weighted summation as the machine learning model with the optimal performance indicators. At the same time, according to the change of the performance indicators of the machine learning model under different feature datasets, the most suitable feature dataset is determined.

[0094] (6) Perform uncertainty analysis on the machine learning model with optimal performance indicators based on p-factor and q-factor.

[0095] The process of understanding and measuring uncertainty in a system or model is called uncertainty analysis. This requires identifying, describing and assessing the sources of uncertainty, and checking how they affect the output of the model or system. The present embodiment performs uncertainty analysis on the machine learning model with optimal performance indicators based on p-factor and q-factor.

[0096] The p-factor is the proportion of predictions that fall within a certain error range. By comparing projected values with observed values, it determines the proportion of values within a tolerable range. This indicator quantifies how much the model captures the variability of the observed data. A high p-factor indicates that a larger proportion of observed values are consistent with the model's predictions. The calculation method is as follows:

[0097]

[0098] where NPAR represents the number of predictions within the acceptable range, and TNAO is the total number of available observations.

[0099] The q-factor represents the proportion of prediction errors. It measures the ratio of predicted values to actual values within a specified error range. It assesses the match between predicted values and observed values, focusing on the overall ratio of predicted values to observed values. It evaluates the overall performance of the model in capturing observed phenomena related to predicted values, providing insights into accuracy and reliability. The calculation method is as follows:

[0100]

[0101] where NPEM (normalized prediction error measure) represents the sum of predicted values adjusted according to errors or deviations from observed values. TNP (total normalized prediction) represents the sum of observed values normalized for comparison.

[0102] The p-factor focuses on the percentage of observed values falling within the acceptable range of predicted values. In contrast, the q-factor is the proportion of predicted values deviating from actual observed values within the acceptable range. When the machine learning model with optimal performance indicators passes the uncertainty analysis, it can be used as a photovoltaic power generation prediction model.

[0103] It is not difficult to find that the present application uses performance indicators such as R2, MAE, MAPE, RMSE and NRMSE to select the best model and a meaningful feature set. After that, through uncertainty analysis to evaluate the degree of uncertainty of prediction and error propagation of the model, the best model selected in this way has wide applicability, can effectively evaluate the performance of large-scale solar photovoltaic power generation facilities, and considers actual factors in the feature set, even if environmental conditions change, the model of the present application can make accurate predictions, improving the prediction accuracy of photovoltaic power generation.

[0104] The second embodiment of the present application relates to a photovoltaic power generation prediction device, comprising:

[0105] The acquisition module is configured to acquire meteorological data of a region where a solar photovoltaic power station to be predicted is located.

[0106] The prediction module is configured to input the meteorological data into a photovoltaic power generation prediction model to obtain the power generation of the solar photovoltaic power station.

[0107] The photovoltaic power generation prediction model is obtained by the following method:

[0108] Collect historical meteorological data and power generation data of the solar photovoltaic power station to be predicted.

[0109] Extract seasonal information and humidity level information from the historical meteorological data, and preprocess the historical meteorological data and power generation data.

[0110] Perform exploratory data analysis on the preprocessed historical meteorological data and power generation data.

[0111] characterizing the historical meteorological data according to the results of the exploratory data analysis, to obtain several groups of feature data sets;

[0112] training several machine learning models respectively taking the several groups of feature data sets as inputs, and selecting a machine learning model with optimal performance indicators and the most suitable feature data set after the training is completed; and taking the machine learning model with optimal performance indicators as the photovoltaic power generation prediction model.

[0113] The meteorological data at least includes one of the following data: sunshine, radiation, temperature, humidity, air pressure and wind speed.

[0114] The seasonal information and humidity level information are extracted from the historical meteorological data, specifically: data from March 1st to May 31st is marked as spring, data from June 1st to August 31st is marked as summer, data from September 1st to November 30th is marked as autumn, and data from December 1st to February 29th is marked as winter; the humidity level of data with humidity above 60 is marked as humid, and the humidity level of data with humidity less than 60 is marked as dry.

[0115] The several groups of feature data sets include: a first feature data set, a second feature data set, a third feature data set, a fourth feature data set, a fifth feature data set and a sixth feature data set, wherein the first feature data set only contains time, the second feature data set only contains meteorological data, the third feature data set combines time and meteorological data, the fourth feature data set adds seasonal information on the basis of the third feature data set, the fifth feature data set adds humidity level information on the basis of the third feature data set, and the sixth feature data set adds seasonal information on the basis of the fifth feature data set.

[0116] The several machine learning models include a support vector regression model, an extreme gradient boosting regression model and a HIST gradient boosting regression model.

[0117] The machine learning model with optimal performance indicators is selected, specifically: the multiple performance indicators of the several machine learning models after training are calculated, the multiple performance indicators of each machine learning model are weighted and summed, and the machine learning model with the highest weighted sum is selected as the machine learning model with optimal performance indicators.

[0118] Before the machine learning model with optimal performance indicators is taken as the photovoltaic power generation prediction model, the method further includes:

[0119] performing uncertainty analysis on the machine learning model with optimal performance indicators based on p-factor and q-factor.

[0120] A third embodiment of the present application relates to an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the photovoltaic power generation prediction method of the first embodiment when executing the computer program.

[0121] A fourth embodiment of the present application relates to a computer readable storage medium having stored thereon a computer program, the computer program implementing the steps of the photovoltaic power generation prediction method of the first embodiment when executed by a processor.

[0122] Those skilled in the art will appreciate that embodiments of the present application can be supplied as methods, systems, or computer program products. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, magnetic disks and optical storage) embodying computer program code thereon.

[0123] The present application is described with reference to the flowchart and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the present application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 The flowchart and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the present application. Figure 1 The flowchart and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the present application.

[0124] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowchart and / or block diagram block or blocks. Figure 1 The flowchart and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the present application. Figure 1 The flowchart and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the present application.

[0125] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart and / or block diagram block or blocks. Figure 1 The flowchart and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the present application.Figure 1 the steps of the functions specified in the one or more blocks.

[0126] The above description is merely that of a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, and all of them should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A photovoltaic power generation prediction method characterized by, The method comprises the following steps: acquiring meteorological data of a region where a solar photovoltaic power station to be predicted is located; inputting the meteorological data into a photovoltaic power generation prediction model to obtain power generation of the solar photovoltaic power station; wherein the photovoltaic power generation prediction model is obtained by the following way: collecting historical meteorological data and power generation data of the solar photovoltaic power station to be predicted; extracting seasonal information and humidity level information from the historical meteorological data, and preprocessing the historical meteorological data and the power generation data; performing exploratory data analysis on the preprocessed historical meteorological data and power generation data; performing feature division on the historical meteorological data according to the result of the exploratory data analysis to obtain a plurality of feature data sets; training a plurality of machine learning models by taking the plurality of feature data sets as inputs of the plurality of machine learning models respectively, and selecting a machine learning model with optimal performance indicators and a most suitable feature data set after the training is completed; taking the machine learning model with optimal performance indicators as the photovoltaic power generation prediction model.

2. The photovoltaic power generation prediction method according to claim 1, characterized by, The meteorological data at least includes one of the following data: sunshine, radiation, temperature, humidity, air pressure and wind speed.

3. The photovoltaic power generation prediction method according to claim 1, characterized by, The seasonal information and humidity level information are extracted from the historical meteorological data in the following way: data from March 1 to May 31 is marked as spring, data from June 1 to August 31 is marked as summer, data from September 1 to November 30 is marked as autumn, and data from December 1 to February 29 is marked as winter; the humidity level of data with humidity above 60 is marked as humid, and the humidity level of data with humidity less than 60 is marked as dry.

4. The photovoltaic power generation prediction method according to claim 1, characterized by, The plurality of feature data sets comprises a first feature data set, a second feature data set, a third feature data set, a fourth feature data set, a fifth feature data set and a sixth feature data set, wherein the first feature data set only contains time, the second feature data set only contains meteorological data, the third feature data set combines time and meteorological data, the fourth feature data set adds seasonal information on the basis of the third feature data set, the fifth feature data set adds humidity level information on the basis of the third feature data set, and the sixth feature data set adds seasonal information on the basis of the fifth feature data set.

5. The photovoltaic power generation prediction method according to claim 1, characterized by, The plurality of machine learning models comprises a support vector regression model, an extreme gradient boosting regression model and a HIST gradient boosting regression model.

6. The photovoltaic power generation prediction method according to claim 1, characterized by, The machine learning model with optimal performance indicators is selected in the following way: a plurality of performance indicators of the plurality of machine learning models after the training is completed are calculated, the plurality of performance indicators of each machine learning model are weighted and summed, and a machine learning model with the highest weighted sum is selected as the machine learning model with optimal performance indicators.

7. The photovoltaic power generation prediction method according to claim 1, characterized by, Before the machine learning model with optimal performance indicators is taken as the photovoltaic power generation prediction model, the following step is further included: performing uncertainty analysis on the machine learning model with optimal performance indicators based on p factor and q factor.

8. A photovoltaic power generation prediction device characterized by comprising: The method comprises the following steps: an acquisition module configured to acquire meteorological data of a region where a solar photovoltaic power station to be predicted is located; The prediction module is configured to input the meteorological data into a photovoltaic power generation prediction model to obtain a power generation amount of the solar photovoltaic power station. The photovoltaic power generation prediction model is obtained by the following method: collecting historical meteorological data and power generation amount data of a solar photovoltaic power station to be predicted; extracting seasonal information and humidity level information from the historical meteorological data, and preprocessing the historical meteorological data and the power generation amount data; performing exploratory data analysis on the preprocessed historical meteorological data and the power generation amount data; performing feature division on the historical meteorological data according to a result of the exploratory data analysis to obtain a plurality of feature data sets; training a plurality of machine learning models by taking the plurality of feature data sets as inputs of the plurality of machine learning models respectively, and selecting a machine learning model with an optimal performance index and a most suitable feature data set after the training is completed; taking the machine learning model with the optimal performance index as the photovoltaic power generation prediction model.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the steps of the photovoltaic power generation prediction method according to any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the photovoltaic power generation prediction method according to any one of claims 1-7.

Images