Photovoltaic power prediction method, device, equipment, system, medium and chip

By using the weighted gray least squares support vector machine model and gray correlation, combined with the characteristic data of weather type, a prediction model is trained, which solves the error problem of existing photovoltaic power prediction on a short time scale, and realizes the photovoltaic power Accurate prediction.

CN118940908APending Publication Date: 2024-11-12BEIJING SMARTCHIP MICROELECTRONICS TECHNOLOGY CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411114478.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The existing photovoltaic power prediction methods have large prediction errors on a short time scale, and the ground radiation data obtained by satellite remote sensing technology have a large deviation from the ground observation data, making it difficult to achieve accurate prediction of photovoltaic power generation and cannot meet the accuracy and accuracy of the power system. Time scale requirements.

Method used

By obtaining the weather prediction type and meteorological characteristics time series data of the predicted day, using the weighted gray least squares support vector machine model, combining the gray correlation degree and correlation coefficient, a prediction model corresponding to the weather type is trained to predict the photovoltaic power of the predicted day. , adjust the model to improve prediction accuracy.

Benefits of technology

Accurate prediction of photovoltaic power for multiple important power consumption moments within a day with a relatively close time is achieved, which reduces prediction errors and improves the performance and stability of the prediction model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940908A_ABST
    Figure CN118940908A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of photovoltaic, in particular to a photovoltaic power prediction method, device, equipment, system, medium and chip, and the method comprises the steps: obtaining the weather prediction type and meteorological characteristic time sequence data of a prediction day, and enabling the meteorological characteristic time sequence data to comprise the time sequence data of a target meteorological characteristic strongly related to the photovoltaic power; acquiring a prediction model corresponding to the weather prediction type from a pre-trained prediction model set as a target prediction model; and inputting the meteorological characteristic time sequence data of the prediction day into the target prediction model, executing the target prediction model, and obtaining photovoltaic power time sequence prediction data of the prediction day output by the target prediction model. According to the technical scheme, short-term photovoltaic power prediction can be accurately carried out, and the method is mainly used for photovoltaic power short-term prediction.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of photovoltaic technology, and in particular to a photovoltaic power prediction method, device, equipment, system, medium and chip. Background Art

[0002] With the rapid development of economy, human beings are facing the problems of energy shortage and environmental pollution. Compared with other new energy sources, solar energy has the advantages of wide distribution, high utilization rate and no pollution, and is one of the most important clean energy sources in human society. Photovoltaic power generation converts solar energy into electrical energy, which can be consumed locally or connected to the power grid and sent to distant places. However, photovoltaic power generation is affected by meteorological factors, and its power generation is intermittent and volatile, which brings great challenges to grid connection and local consumption. In order to reduce its impact on the power grid and ensure the stability of the power grid, it is necessary to accurately predict the power generation of the photovoltaic system.

[0003] One existing photovoltaic power prediction method is to establish a prediction model based on historical meteorological data and historical photovoltaic power generation data to predict photovoltaic power generation. However, this prediction method is only suitable for predictions on a longer time scale with little change in power generation. For predictions on a shorter time scale with large changes in power generation, the prediction error is large. Another photovoltaic power prediction method is to predict photovoltaic power based on satellite cloud images and ground monitoring real-time calculation of solar radiation methods. However, the ground radiation data obtained by satellite remote sensing technology has a large deviation from the radiation data observed on the ground, making it difficult to achieve accurate prediction of photovoltaic power generation. It can be seen that the existing photovoltaic power prediction methods are not adaptable enough and cannot meet the current power system's requirements for the accuracy and time scale of photovoltaic power prediction. Summary of the invention

[0004] In order to solve the problems in the related art, the embodiments of the present disclosure provide a photovoltaic power prediction method, device, equipment, system, medium and chip.

[0005] In a first aspect, an embodiment of the present disclosure provides a photovoltaic power prediction method, comprising:

[0006] Acquire the weather forecast type and meteorological characteristic time series data of the forecast day, wherein the meteorological characteristic time series data includes time series data of target meteorological characteristics that are strongly correlated with photovoltaic power;

[0007] Obtaining a prediction model corresponding to the weather prediction type from a set of pre-trained prediction models as a target prediction model;

[0008] The meteorological characteristic time series data of the forecast day is input into the target forecast model, and the target forecast model is executed to obtain the photovoltaic power time series forecast data of the forecast day output by the target forecast model.

[0009] In a possible implementation, the method further includes:

[0010] Acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day;

[0011] Dividing the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day;

[0012] For each weather type, obtaining a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type;

[0013] The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

[0014] In a possible implementation manner, obtaining, from the sample data corresponding to the weather type, a plurality of sample data with the highest correlation with the weather type as the training sample data corresponding to the weather type includes:

[0015] The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0016]

[0017] Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), the minimum absolute difference, maxy(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j) the maximum absolute difference;

[0018] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0019] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0020]

[0021] in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector;

[0022] The linear regression function of the weighted grey least squares support vector machine model is:

[0023]

[0024] Among them, α i is the Lagrange multiplier, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0025] In a possible implementation manner, the acquiring a plurality of sample data includes:

[0026] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0027] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0028] In a possible implementation manner, the step of acquiring the meteorological characteristic time series processing data and the photovoltaic power time series processing data of a plurality of sample days includes:

[0029] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0030] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0031] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0032] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0033] In a possible implementation manner, the obtaining of initial meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days includes:

[0034] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0035] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0036] In a possible implementation, the method further includes:

[0037] When the error between the photovoltaic power time series prediction data on the forecast day and the photovoltaic power time series data on the forecast day exceeds a predetermined error, the meteorological characteristic time series data and the photovoltaic power time series data on the forecast day are determined as new sample data corresponding to the weather type on the forecast day;

[0038] Based on the new sample data corresponding to the weather type of the forecast day, the forecast model corresponding to the weather type of the forecast day is adjusted.

[0039] In a second aspect, an embodiment of the present disclosure provides a method for training a prediction model for photovoltaic power prediction, comprising:

[0040] Acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day;

[0041] Dividing the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day;

[0042] For each weather type, sample data corresponding to the weather type is used to train a prediction model corresponding to the weather type.

[0043] In a possible implementation, the using the sample data corresponding to the weather type to train a prediction model corresponding to the weather type includes:

[0044] From the sample data corresponding to the weather type, obtain a plurality of sample data with the highest correlation with the weather type as training sample data corresponding to the weather type;

[0045] The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

[0046] In a possible implementation manner, obtaining, from the sample data corresponding to the weather type, a plurality of sample data with the highest correlation with the weather type as the training sample data corresponding to the weather type includes:

[0047] The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0048]

[0049] Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), the minimum absolute difference, max|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j) the maximum absolute difference;

[0050] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0051] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0052]

[0053] in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector;

[0054] The linear regression function of the weighted grey least squares support vector machine model is:

[0055]

[0056] Among them, α i is the Lagrange multiplier, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0057] In a possible implementation manner, the acquiring a plurality of sample data includes:

[0058] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0059] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0060] In a possible implementation manner, the step of acquiring the meteorological characteristic time series processing data and the photovoltaic power time series processing data of a plurality of sample days includes:

[0061] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0062] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0063] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0064] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0065] In a possible implementation manner, the obtaining of initial meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days includes:

[0066] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0067] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0068] In a possible implementation, the method further includes:

[0069] When the error between the photovoltaic power time series prediction data of the prediction day predicted by the prediction model and the photovoltaic power time series data of the prediction day exceeds a predetermined error, the meteorological characteristic time series data and the photovoltaic power time series data of the prediction day are determined as new sample data corresponding to the weather type of the prediction day;

[0070] Based on the new sample data corresponding to the weather type of the forecast day, the forecast model corresponding to the weather type of the forecast day is adjusted.

[0071] In a third aspect, an embodiment of the present disclosure provides a photovoltaic power prediction device, comprising:

[0072] A data acquisition module is configured to acquire a weather forecast type and meteorological characteristic time series data of a forecast day, wherein the meteorological characteristic time series data includes time series data of target meteorological characteristics that are strongly correlated with photovoltaic power;

[0073] A model acquisition module is configured to acquire a prediction model corresponding to the weather prediction type from a pre-trained prediction model set as a target prediction model;

[0074] The prediction module is configured to input the meteorological characteristic time series data of the prediction day into the target prediction model, execute the target prediction model, and obtain the photovoltaic power time series prediction data of the prediction day output by the target prediction model.

[0075] In a possible implementation, the device further includes:

[0076] The model training module is configured to obtain multiple sample data, the sample data including meteorological characteristic time series data and photovoltaic power time series data of the sample day; divide the multiple sample data into sample data corresponding to different weather types according to the weather type of the sample day; for each weather type, obtain multiple sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type; use the training sample data corresponding to the weather type to train a prediction model corresponding to the weather type.

[0077] In a possible implementation, the model training module obtains a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type, and the portion of the sample data for training corresponding to the weather type is configured as follows:

[0078] The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0079]

[0080] Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), the minimum absolute difference, max|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j) the maximum absolute difference;

[0081] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0082] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0083]

[0084] in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector;

[0085] The linear regression function of the weighted grey least squares support vector machine model is:

[0086]

[0087] Among them, α i is the Lagrange multiplier, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0088] In a possible implementation, the part of acquiring multiple sample data in the model training module is configured as follows:

[0089] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0090] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0091] In a possible implementation manner, the part of the model training module that obtains the meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows:

[0092] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0093] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0094] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0095] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0096] In a possible implementation manner, the part of the model training module that obtains the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows:

[0097] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0098] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0099] In a possible implementation, the device further includes:

[0100] The model adjustment module is configured to determine the meteorological characteristic time series data and photovoltaic power time series data of the forecast day as new sample data corresponding to the weather type of the forecast day when the error between the photovoltaic power time series prediction data of the forecast day and the photovoltaic power time series data of the forecast day exceeds a predetermined error; and adjust the prediction model corresponding to the weather type of the forecast day based on the new sample data corresponding to the weather type of the forecast day.

[0101] In a fourth aspect, an embodiment of the present disclosure provides a training device for a prediction model of photovoltaic power prediction, comprising:

[0102] A sample acquisition module is configured to acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day;

[0103] A sample division module is configured to divide the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day;

[0104] The training module is configured to train, for each weather type, a prediction model corresponding to the weather type using sample data corresponding to the weather type.

[0105] In a possible implementation, the training module is configured as follows:

[0106] For each weather type, obtaining a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type;

[0107] The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

[0108] In a possible implementation manner, the training module obtains a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type, and the portion of the sample data for training corresponding to the weather type is configured as follows:

[0109] The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0110]

[0111] Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), the minimum absolute difference, max|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j) the maximum absolute difference;

[0112] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0113] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0114]

[0115] in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector;

[0116] The linear regression function of the weighted grey least squares support vector machine model is:

[0117]

[0118] Among them, α i is the Lagrange multiplier, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0119] In a possible implementation, the sample acquisition module is configured as follows:

[0120] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0121] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0122] In a possible implementation manner, the part of the sample acquisition module that acquires the meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows:

[0123] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0124] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0125] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0126] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0127] In a possible implementation manner, the part of the sample acquisition module that acquires the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows:

[0128] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0129] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0130] In a possible implementation, the device further includes:

[0131] A new sample acquisition module is configured to determine the meteorological characteristic time series data and the photovoltaic power time series data of the forecast day as new sample data corresponding to the weather type of the forecast day when the error between the photovoltaic power time series prediction data of the forecast day predicted by the forecast model and the photovoltaic power time series data of the forecast day exceeds a predetermined error;

[0132] The adjustment module is configured to adjust the prediction model corresponding to the weather type of the forecast day based on the new sample data corresponding to the weather type of the forecast day.

[0133] In a fifth aspect, an embodiment of the present disclosure provides an electronic device, comprising a memory and a processor, wherein the memory is used to store one or more computer instructions, and wherein the one or more computer instructions are executed by the processor to implement a method as described in any one of the first aspect or the second aspect.

[0134] In a sixth aspect, an embodiment of the present disclosure provides a system, comprising the above-mentioned electronic device.

[0135] In a seventh aspect, an embodiment of the present disclosure provides a readable storage medium on which computer instructions are stored, and the computer instructions implement the above-mentioned method steps when executed by a processor.

[0136] In an eighth aspect, a chip is provided in an embodiment of the present disclosure, and the chip includes the device as described in any one of the third aspect or the fourth aspect above.

[0137] According to the technical solution provided by the embodiment of the present disclosure, the weather forecast type and meteorological characteristic time series data of the forecast day can be obtained, and the meteorological characteristic time series data is the time series data of the target meteorological characteristics that are strongly correlated with the photovoltaic power; the prediction model corresponding to the weather forecast type is obtained from a set of pre-trained prediction models as the target prediction model; the meteorological characteristic time series data of the forecast day is input into the target prediction model, and the target prediction model is executed to obtain the photovoltaic power time series prediction data of the forecast day output by the target prediction model; in this way, the photovoltaic power time series prediction data of the forecast day with the corresponding weather type is predicted using multiple prediction models trained respectively according to the weather type, so that the photovoltaic power at multiple important electricity consumption moments in the forecast day with a relatively close time can be accurately predicted.

[0138] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0139] Other features, objectives and advantages of the present disclosure will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings. In the accompanying drawings:

[0140] Figure 1 A flow chart of a photovoltaic power prediction method according to an embodiment of the present disclosure is shown.

[0141] Figure 2 A flow chart showing a method for training a prediction model for photovoltaic power prediction according to an embodiment of the present disclosure.

[0142] Figure 3 A structural block diagram of a photovoltaic power prediction device according to an embodiment of the present disclosure is shown.

[0143] Figure 4 A structural block diagram of a training device for a prediction model for photovoltaic power prediction according to an embodiment of the present disclosure is shown.

[0144] Figure 5 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0145] Figure 6 A schematic diagram showing the structure of a computer system suitable for implementing the method of the embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0146] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement them. In addition, for the sake of clarity, parts not related to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0147] In the present disclosure, it should be understood that terms such as "include" or "have" are intended to indicate the presence of features, numbers, steps, behaviors, components, parts, or a combination thereof disclosed in the present specification, and are not intended to exclude the possibility that one or more other features, numbers, steps, behaviors, components, parts, or a combination thereof exist or are added.

[0148] It should also be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0149] Figure 1 FIG. 1 is a flow chart showing a photovoltaic power prediction method according to an embodiment of the present disclosure. Figure 1 As shown, the photovoltaic power prediction method includes the following steps S101-S103:

[0150] In step S101, the weather forecast type and the forecasted meteorological characteristic time series data of the forecast day are obtained, wherein the meteorological characteristic time series data is the time series data of the target meteorological characteristic that is strongly correlated with the photovoltaic power;

[0151] In step S102, a prediction model corresponding to the weather prediction type is obtained from a set of pre-trained prediction models as a target prediction model;

[0152] In step S103, the predicted meteorological characteristic time series data is input into the target prediction model, and the target prediction model is executed to obtain the photovoltaic power time series prediction data of the predicted day output by the target prediction model.

[0153] In one possible implementation, the photovoltaic power prediction method is applicable to electronic devices that can perform photovoltaic power prediction. The electronic devices may be various electronic devices in the distribution network system, such as FTU (Feeder Terminal Unit, distribution switch monitoring terminal), DTU (Distribution Terminal Unit, switchgear terminal equipment), intelligent fusion terminal, etc., or the electronic device may be a device on a distribution master station or a distribution substation.

[0154] In one possible implementation, the photovoltaic power prediction method is mainly used for short-term prediction of photovoltaic power. The short-term prediction here refers to the photovoltaic power time series prediction data for the corresponding prediction day. The prediction day can be the day closest to the current time. Since photovoltaic power generation is intermittent and periodic, it does not generate electricity at night. In order to improve the prediction accuracy, this implementation usually chooses to predict the photovoltaic power from 6:00 to 19:00 on the next day (that is, the prediction day) after sunset, such as 20:00-24:00. The photovoltaic power time series prediction data for the prediction day can be the time series data formed by the photovoltaic power at several important electricity consumption moments within 6:00-19:00 on the prediction day.

[0155] In one possible implementation, when performing photovoltaic power prediction, the weather forecast type and predicted meteorological feature time series data of the prediction day can be obtained through the meteorological forecast information of the prediction day. For example, the weather forecast type can be sunny, rainy, cloudy and other weather types. There are many kinds of meteorological characteristics, such as humidity, temperature, solar radiation intensity, wind, rainfall, weather type, etc. The target meteorological feature that is strongly correlated with photovoltaic power can be some meteorological features that are strongly correlated with photovoltaic power selected from these meteorological features. The strong correlation here is also called high correlation, that is, when the time series data of the target meteorological feature changes, the corresponding photovoltaic power time series data is very likely to change. The meteorological feature time series data is a sequence data composed of the feature data of these target meteorological features at the corresponding time.

[0156] In one possible implementation, multiple prediction models can be pre-trained, each prediction model corresponds to a weather type, and the prediction model corresponding to the weather type can be trained based on the sample data corresponding to the weather type, and is used to predict the photovoltaic power time series prediction data of the prediction day with the weather type. That is, the prediction model corresponding to the weather type is used to determine the photovoltaic power time series prediction data of the prediction day based on the meteorological characteristic time series data of the prediction day with the weather type.

[0157] In a possible implementation, if the weather forecast type for the predicted day is sunny, the prediction model corresponding to the sunny day can be first obtained from a pre-trained prediction model set, and then the meteorological characteristic data at multiple times between 6:00 and 19:00 on the predicted day, that is, the meteorological characteristic time series data, are input into the prediction model corresponding to the sunny day, and the prediction model corresponding to the sunny day is executed to obtain the photovoltaic power at multiple times between 6:00 and 19:00 on the predicted day, that is, the photovoltaic power time series data, output by the prediction model corresponding to the sunny day.

[0158] This implementation can obtain the weather forecast type and meteorological characteristic time series data of the forecast day, wherein the meteorological characteristic time series data is the time series data of the target meteorological characteristics that are strongly correlated with the photovoltaic power; the prediction model corresponding to the weather forecast type is obtained from a set of pre-trained prediction models as the target prediction model; the meteorological characteristic time series data of the forecast day is input into the target prediction model, and the target prediction model is executed to obtain the photovoltaic power time series prediction data of the forecast day output by the target prediction model; in this way, multiple prediction models trained according to weather types are used to predict the photovoltaic power time series prediction data of the forecast day with the corresponding weather type, so that the photovoltaic power at multiple important electricity consumption moments in the forecast day with a relatively close time can be accurately predicted.

[0159] In a possible implementation, the method may further include the following steps:

[0160] Acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day;

[0161] Dividing the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day;

[0162] For each weather type, obtaining a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type;

[0163] The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

[0164] In this embodiment, for example, photovoltaic power data and target meteorological characteristic data for the whole year from April 30, 2022 to April 29, 2023 can be collected, and the sampling interval is 5 minutes. In this way, photovoltaic power data and target meteorological characteristic data at multiple important moments from 6:00 to 19:00 every day from April 30, 2022 to April 29, 2023 can be obtained. The photovoltaic power data and target meteorological characteristic data at multiple important moments from 6:00 to 19:00 every day are also the meteorological characteristic time series data and photovoltaic power time series data of the sample day.

[0165] In this embodiment, after obtaining multiple sample data, the multiple sample data can be divided into sample data corresponding to different weather types according to the weather type of the sample day. For example, the sample data corresponding to sunny days form sample set A, the sample data corresponding to rainy days form sample set B, and the sample data corresponding to cloudy days form sample set C. In this way, the multiple sample data can be divided into three sample sets.

[0166] In this implementation, in order to make the performance of the trained prediction model better, for each weather type, multiple sample data with the highest correlation with the weather type can be selected from the corresponding sample set as the training sample data corresponding to the weather type, and the training sample data corresponding to the weather type is used to train the prediction model corresponding to the weather type. For example, for sunny days, M (M is an integer greater than 1) sample data with the highest correlation with sunny days can be selected from sample set A as the training sample data corresponding to sunny days, and the prediction model for sunny days can be trained using the training sample data corresponding to sunny days.

[0167] In this embodiment, a plurality of sample data with the highest correlation with each weather type can be selected from the sample data corresponding to each weather type to train a prediction model corresponding to the weather type, so that a prediction model with accurate prediction results can be trained.

[0168] In a possible implementation manner, obtaining, from the sample data corresponding to the weather type, a plurality of sample data with the highest correlation with the weather type as the training sample data corresponding to the weather type includes:

[0169] The grey correlation between the target day corresponding to the weather type and each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0170]

[0171] Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), that is, min|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k)|, max|y(j)-x i(j)|When j takes the value 1, 2, ... k, y(j) and x i (j); that is, max|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k) the maximum value among |

[0172] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0173] In this embodiment, the grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type can be calculated to obtain the grey correlation between each sample day in the sample data corresponding to the weather type and the target day. The sample data of the multiple sample days with the highest grey correlation can be used as the multiple sample data with the highest correlation with the weather type. In this way, the training sample data corresponding to the weather type can be obtained.

[0174] In this embodiment, the target day may be the first forecast day with the weather type. For example, at the beginning of the photovoltaic power forecast, the weather type of the first forecast day is first obtained. Assuming it is a sunny day, the gray correlation between the target meteorological characteristics of each sample day in the sample set of the sunny day and the target meteorological characteristics of the first forecast day can be calculated based on the sample set of the sunny day and the target meteorological characteristics of the first forecast day. Based on the gray correlation, the sample data of the sample day with a high correlation with the first forecast day is obtained as the training sample data of the sunny day, and the prediction model of the sunny day is obtained by training, and the photovoltaic power time series prediction data of the first forecast day is predicted by the prediction model. Assuming that the weather type of the second forecast day is still sunny, the previously trained prediction model of the sunny day is used to predict the photovoltaic power time series prediction data of the second forecast day; of course, if the weather type of the second forecast day is rainy, the training sample data of the rainy day can also be obtained, and the prediction model of the rainy day can be obtained by training, and the photovoltaic power time series prediction data of the second forecast day can be predicted by the prediction model.

[0175] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0176]

[0177] in, To convert the input vector x in the i-th training sample data iA nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector;

[0178] Refers to minimize;

[0179] The linear regression function of the weighted grey least squares support vector machine model is:

[0180]

[0181] Among them, α i is the Lagrange multiplier, α i ≥0, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0182] In this embodiment, the existing least squares support vector machine (LSSVM) is a kind of machine learning method, which is very good at dealing with the photovoltaic power prediction problem of small sample nonlinear regression, and has been widely used in the field of photovoltaic power prediction. Its basic idea is: through a pre-selected nonlinear mapping, the input is mapped to a high-dimensional feature space, and the optimal decision function is constructed in this space. The model requires all training data to participate in each prediction during the training process, and the prediction accuracy is optimal only when the error variable conforms to the Gaussian distribution, thereby limiting the robustness of the LSSVM model. In order to solve the above problems, this embodiment adopts a weighted grey least squares support vector machine model (WG-LSSVM) to predict photovoltaic power generation, and its basic principles are as follows:

[0183] Assume that the training sample data corresponding to the weather type is the sample set In this sample set A, x i represents the meteorological characteristic time series data in the i-th sample data, that is, the input vector of the model, y iis the photovoltaic power time series data in the i-th sample data, and m represents the total number of samples in the sample set A. When performing nonlinear regression estimation on the sample points in a given sample set, the main method of SVM is to map all the sample points to a high-dimensional feature space through a nonlinear function, and complete linear regression in this high-dimensional feature space. The regression function at this time is f(x) = w·φ(x) + b; the WG-LSSVM model uses the square error loss function instead of the insensitive loss function e in the standard SVM. i , and give each insensitive loss function e i Assign a corresponding weight v i , using equality constraints to replace inequality constraints, the optimization problem is converted to:

[0184]

[0185] Where C represents the correction factor, C>0. After the Lagrangian function is introduced, the corresponding linear regression function is:

[0186]

[0187] From the above process, we can see that the WG-LSSVM method has only one parameter C to be selected, and only needs to solve the Lagrange equations to get α i and b, which has the advantages of being fast, simple, and stable.

[0188] In a possible implementation manner, the acquiring a plurality of sample data includes:

[0189] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0190] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0191] In this embodiment, the processing data of the target meteorological characteristics corresponding to multiple moments of a sample day (i.e., meteorological characteristic time series processing data) and the processing data of photovoltaic power corresponding to multiple moments (i.e., photovoltaic power time series processing data) can be obtained as an initial sample data. The processing here refers to processing the data into the specifications required for model training.

[0192] In this implementation, the validity and accuracy of the sample data are the premise and basis for scientifically establishing a system prediction model. Although the initial sample data obtained is accurate data that can objectively reflect the development trend of the photovoltaic power generation system, the initial sample data reflects the historical status of this stage. Therefore, in order to ensure a more reasonable prediction of photovoltaic power, this implementation proposes an average weakening buffer operator, which is used to pre-process the initial sample data to weaken the development trend of the photovoltaic power generation system, thereby feeding back the qualitative analysis results to the future development trend of the photovoltaic power generation system, and avoiding the prediction results of the subsequent training prediction model from being inconsistent with the qualitative analysis conclusions.

[0193] The following is an example of the photovoltaic power time series processing data Y = (y(1), y(2), ... y(q)) of a sample day. The average weakening buffer operator D is used to weaken the sequence Y, that is,

[0194] YD=(y(1)d,y(2)d,…y(q)d);

[0195]

[0196] Among them, q is the total amount of photovoltaic power time series processing data Y, and YD is the buffer sequence, that is, the weakened photovoltaic power time series data.

[0197] In this embodiment, after the meteorological characteristic time series processing data and photovoltaic power time series processing data of the multiple sample days are weakened using the average weakening buffer operator, the weakened meteorological characteristic time series data and photovoltaic power time series data of the multiple sample days can be obtained, and the weakened meteorological characteristic time series data and photovoltaic power time series data of the multiple sample days form a plurality of sample data.

[0198] In a possible implementation manner, the step of acquiring the meteorological characteristic time series processing data and the photovoltaic power time series processing data of a plurality of sample days includes:

[0199] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0200] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0201] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0202] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0203] In this embodiment, when training the model, the selection of model input variables is very important. Too many input variables will lead to excessive calculation costs and reduce the operating efficiency of the model. Therefore, this embodiment can determine the target meteorological features with a high degree of correlation with photovoltaic power from the initial meteorological features by calculating the correlation coefficient. For example, the Pearson correlation coefficient and / or the Spearson rank correlation coefficient of various initial meteorological features and photovoltaic power can be calculated. The Pearson correlation coefficient is often used to analyze the strength of the linear relationship between variables (i.e., various meteorological features and photovoltaic power), while the Spearson rank correlation coefficient only measures the monotonic relationship between variables, does not consider the influence of specific values, and is often used to measure the nonlinear relationship between variables. Therefore, preferably, the two correlation coefficients can be comprehensively considered, and the meteorological features with a high degree of correlation can be selected as the target meteorological features. For example, the weighted average of the two correlation coefficients can be calculated to obtain a comprehensive correlation coefficient. If the comprehensive correlation coefficient of the initial meteorological feature is greater than a predetermined threshold, the initial meteorological feature whose correlation coefficient meets the predetermined conditions is the target meteorological feature.

[0204] In this embodiment, after determining the target meteorological characteristics, the time series processing data of the target meteorological characteristics of each sample day in the plurality of processed sample data may be obtained, that is, the time series processing data of the meteorological characteristics in the above sample data.

[0205] In a possible implementation manner, the obtaining of initial meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days includes:

[0206] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0207] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0208] In this embodiment, the extreme values ​​that appear in the outlier sample and the processing of the outliers will directly affect the accuracy of subsequent predictions. This embodiment can use the physical discrimination method to analyze the rules of each initial meteorological feature and photovoltaic power to find outliers. For example, if the photovoltaic power must be greater than 0, then the photovoltaic power less than 0 can be determined as an outlier, and the outlier can be replaced by 0 or replaced by the average value of the photovoltaic power at two adjacent moments; the wind direction data should be between 0-360 degrees, and the wind direction data that is not in this range can be determined as an outlier, and the outlier can be replaced by 0 or replaced by the average value of the wind direction data at two adjacent moments. Moreover, the values ​​of the initial meteorological characteristics and photovoltaic power are subject to normal distribution, and the 3σ principle can be used to identify and process outliers. The 3σ principle is also called the Laida criterion. It first assumes that a set of data contains only random errors, calculates and processes them to obtain the standard deviation, and determines an interval according to a certain probability. It is believed that any error exceeding this interval is not a random error but a gross error. The value containing the gross error is determined as an outlier, and the outlier can be replaced by 0 or replaced by the average value of its two adjacent moments.

[0209] In this embodiment, the dimensions of the initial meteorological characteristic time series raw data and the photovoltaic power time series raw data are different. The large difference in dimensions and quantity sets will bring great difficulties to data analysis. In order to eliminate the impact of dimensions, this embodiment can normalize the initial meteorological characteristic time series raw data and the photovoltaic power time series raw data. For example, the initial meteorological characteristic time series raw data and the photovoltaic power time series raw data are mapped to [0,1]. The normalization processing here can be maximum and minimum value normalization, mean variance normalization, etc. Taking maximum and minimum value normalization as an example, the normalization formula of the photovoltaic power time series raw data is as follows:

[0210]

[0211] Among them, x is the value of photovoltaic power before normalization, x' is the value of photovoltaic power after normalization, and x min is the minimum photovoltaic power value in all photovoltaic power time series raw data, x max It is the maximum value of photovoltaic power in all photovoltaic power time series raw data.

[0212] In a possible implementation, the method may further include the following steps:

[0213] When the error between the photovoltaic power time series prediction data on the forecast day and the photovoltaic power time series data on the forecast day exceeds a predetermined error, the meteorological characteristic time series data and the photovoltaic power time series data on the forecast day are determined as new sample data corresponding to the weather type on the forecast day;

[0214] Based on the new sample data corresponding to the weather type of the forecast day, the forecast model corresponding to the weather type of the forecast day is adjusted.

[0215] In this embodiment, after the prediction day has passed, the photovoltaic power time series data of the prediction day can be obtained, and the error between the photovoltaic power time series prediction data of the prediction day and the photovoltaic power time series data of the prediction day can be calculated, for example, the RMSE (Root Mean Squard Error), MAE (Mean Absolute Error), and R2 (R-Square, determination coefficient) between the two are calculated; if the error exceeds the predetermined error, it means that the photovoltaic power time series prediction data predicted by the prediction model is inaccurate. At this time, the meteorological characteristic time series data and the photovoltaic power time series data of the prediction day can be determined as new sample data corresponding to the weather type of the prediction day, and the new sample data can be used to adjust the prediction model corresponding to the weather type of the prediction day; in this way, the prediction model is continuously adjusted and optimized according to the prediction results, and the obtained prediction model will be more accurate over time.

[0216] Figure 2 A flow chart showing a method for training a prediction model for photovoltaic power prediction according to an embodiment of the present disclosure is shown. Figure 2 As shown, the training method of the prediction model includes the following steps S201-S203:

[0217] In step S201, a plurality of sample data are obtained, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day;

[0218] In step S202, the plurality of sample data are divided into sample data corresponding to different weather types according to the weather type of the sample day;

[0219] In step S203, for each weather type, sample data corresponding to the weather type is used to train a prediction model corresponding to the weather type.

[0220] In a possible implementation, the prediction model training method is applicable to an electronic device capable of executing prediction model training.

[0221] For example, the photovoltaic power data and target meteorological characteristic data for the whole year from April 30, 2022 to April 29, 2023 can be collected, with a sampling interval of 5 minutes. In this way, the photovoltaic power data and target meteorological characteristic data at multiple important moments from 6:00 to 19:00 every day from April 30, 2022 to April 29, 2023 can be obtained. The photovoltaic power data and target meteorological characteristic data at multiple important moments from 6:00 to 19:00 every day are also the meteorological characteristic time series data and photovoltaic power time series data of the sample day.

[0222] In a possible implementation, after obtaining multiple sample data, the multiple sample data can be divided into sample data corresponding to different weather types according to the weather type of the sample day. For example, the sample data corresponding to sunny days form sample set A, the sample data corresponding to rainy days form sample set B, and the sample data corresponding to cloudy days form sample set C. In this way, the multiple sample data can be divided into three sample sets.

[0223] In a possible implementation, a prediction model corresponding to sunny days can be obtained by training using sample set A corresponding to sunny days, a prediction model corresponding to rainy days can be obtained by training using sample set B corresponding to rainy days, and a prediction model corresponding to cloudy days can be obtained by training using sample set B corresponding to cloudy days. The prediction model corresponding to the corresponding weather type is used to predict the photovoltaic power time series prediction data of the prediction day with the weather type, that is, the prediction model corresponding to the weather type is used to determine the photovoltaic power time series prediction data of the prediction day based on the meteorological characteristic time series data of the prediction day with the weather type. For example, if the weather forecast type of the prediction day is obtained as sunny days, the prediction model corresponding to sunny days can be obtained from the pre-trained prediction model set, and then the meteorological characteristic data at multiple times within 6:00-19:00 on the prediction day, that is, the meteorological characteristic time series data, are input into the prediction model corresponding to the sunny day, and the prediction model corresponding to the sunny day is executed to obtain the photovoltaic power at multiple times within 6:00-19:00 on the prediction day, that is, the photovoltaic power time series data output by the prediction model corresponding to the sunny day.

[0224] This implementation can predict the photovoltaic power time series prediction data of the prediction day with the corresponding weather type according to multiple prediction models trained respectively for the weather type, and can accurately predict the photovoltaic power at multiple important electricity consumption moments in the prediction day with a relatively close time.

[0225] In a possible implementation, the using the sample data corresponding to the weather type to train a prediction model corresponding to the weather type includes:

[0226] From the sample data corresponding to the weather type, obtain a plurality of sample data with the highest correlation with the weather type as training sample data corresponding to the weather type;

[0227] The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

[0228] In this implementation, in order to improve the performance of the trained prediction model, for each weather type, multiple sample data with the highest correlation with the weather type can be selected from the corresponding sample set as the training sample data corresponding to the weather type, and the training sample data corresponding to the weather type is used to train the prediction model corresponding to the weather type. For example, for sunny days, N (N is an integer greater than 1) sample data with the highest correlation with sunny days can be selected from sample set A as the training sample data corresponding to sunny days, and the prediction model for sunny days can be trained using the training sample data corresponding to sunny days.

[0229] In this embodiment, a plurality of sample data with the highest correlation with each weather type can be selected from the sample data corresponding to each weather type to train a prediction model corresponding to the weather type, so that a prediction model with accurate prediction results can be trained.

[0230] In a possible implementation manner, obtaining, from the sample data corresponding to the weather type, a plurality of sample data with the highest correlation with the weather type as the training sample data corresponding to the weather type includes:

[0231] The grey correlation between the target day corresponding to the weather type and each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0232]

[0233] Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i(j), that is, min|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k)|, max|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j); that is, max|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k) the maximum value among |

[0234] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0235] In this embodiment, the grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type can be calculated to obtain the grey correlation between each sample day in the sample data corresponding to the weather type and the target day. The sample data of the multiple sample days with the highest grey correlation can be used as the multiple sample data with the highest correlation with the weather type. In this way, the training sample data corresponding to the weather type can be obtained.

[0236] In this embodiment, the target day may be the first forecast day with the weather type. For example, at the beginning of the photovoltaic power forecast, the weather type of the first forecast day is first obtained. Assuming it is a sunny day, the gray correlation between the target meteorological characteristics of each sample day in the sample set of the sunny day and the target meteorological characteristics of the first forecast day can be calculated based on the sample set of the sunny day and the target meteorological characteristics of the first forecast day. Based on the gray correlation, the sample data of the sample day with a high correlation with the first forecast day is obtained as the training sample data of the sunny day, and the prediction model of the sunny day is obtained by training, and the photovoltaic power time series prediction data of the first forecast day is predicted by the prediction model. Assuming that the weather type of the second forecast day is still sunny, the previously trained prediction model of the sunny day is used to predict the photovoltaic power time series prediction data of the second forecast day; of course, if the weather type of the second forecast day is rainy, the training sample data of the rainy day can also be obtained, and the prediction model of the rainy day can be obtained by training, and the photovoltaic power time series prediction data of the second forecast day can be predicted by the prediction model.

[0237] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0238]

[0239] in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector;

[0240] Refers to minimize;

[0241] The linear regression function of the weighted grey least squares support vector machine model is:

[0242]

[0243] Among them, α i is the Lagrange multiplier, α i ≥0, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0244] In this embodiment, the existing least squares support vector machine (LSSVM) is a kind of machine learning method, which is very good at dealing with the photovoltaic power prediction problem of small sample nonlinear regression, and has been widely used in the field of photovoltaic power prediction. Its basic idea is: through a pre-selected nonlinear mapping, the input is mapped to a high-dimensional feature space, and the optimal decision function is constructed in this space. The model requires all training data to participate in each prediction during the training process, and the prediction accuracy is optimal only when the error variable conforms to the Gaussian distribution, thereby limiting the robustness of the LSSVM model. In order to solve the above problems, this embodiment adopts a weighted grey least squares support vector machine model (WG-LSSVM) to predict photovoltaic power generation, and its basic principles are as follows:

[0245] Assume that the training sample data corresponding to the weather type is the sample set In this sample set A, x i represents the meteorological characteristic time series data in the i-th sample data, that is, the input vector of the model, y i is the photovoltaic power time series data in the i-th sample data, and m represents the total number of samples in the sample set A. When performing nonlinear regression estimation on the sample points in a given sample set, the main method of SVM is to map all the sample points to a high-dimensional feature space through a nonlinear function, and complete linear regression in this high-dimensional feature space. The regression function at this time is f(x) = w·φ(x) + b; the WG-LSSVM model uses the square error loss function instead of the insensitive loss function e in the standard SVM. i , and give each insensitive loss function e i Assign a corresponding weight v i , using equality constraints to replace inequality constraints, the optimization problem is converted to:

[0246]

[0247] Where C represents the correction factor, C>0. After the Lagrangian function is introduced, the corresponding linear regression function is:

[0248]

[0249] From the above process, we can see that the WG-LSSVM method has only one parameter C to be selected, and only needs to solve the Lagrange equations to get α i and b, which has the advantages of being fast, simple, and stable.

[0250] In a possible implementation manner, the acquiring a plurality of sample data includes:

[0251] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0252] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0253] In this embodiment, the processing data of the target meteorological characteristics corresponding to multiple moments of a sample day (i.e., meteorological characteristic time series processing data) and the processing data of photovoltaic power corresponding to multiple moments (i.e., photovoltaic power time series processing data) can be obtained as an initial sample data. The processing here refers to processing the data into the specifications required for model training.

[0254] In this implementation, the validity and accuracy of the sample data are the premise and basis for scientifically establishing a system prediction model. Although the initial sample data obtained is accurate data that can objectively reflect the development trend of the photovoltaic power generation system, the initial sample data reflects the historical status of this stage. Therefore, in order to ensure a more reasonable prediction of photovoltaic power, this implementation proposes an average weakening buffer operator, which is used to pre-process the initial sample data to weaken the development trend of the photovoltaic power generation system, thereby feeding back the qualitative analysis results to the future development trend of the photovoltaic power generation system, and avoiding the prediction results of the subsequent training prediction model from being inconsistent with the qualitative analysis conclusions.

[0255] The following is an example of the photovoltaic power time series processing data Y = (y(1), y(2), ... y(q)) of a sample day. The average weakening buffer operator D is used to weaken the sequence Y, that is,

[0256] YD=(y(1)d,y(2)d,…y(q)d);

[0257]

[0258] Among them, q is the total amount of photovoltaic power time series processing data Y, and YD is the buffer sequence, that is, the weakened photovoltaic power time series data.

[0259] In this embodiment, after the meteorological characteristic time series processing data and photovoltaic power time series processing data of the multiple sample days are weakened using the average weakening buffer operator, the weakened meteorological characteristic time series data and photovoltaic power time series data of the multiple sample days can be obtained, and the weakened meteorological characteristic time series data and photovoltaic power time series data of the multiple sample days form a plurality of sample data.

[0260] In a possible implementation manner, the step of acquiring the meteorological characteristic time series processing data and the photovoltaic power time series processing data of a plurality of sample days includes:

[0261] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0262] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0263] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0264] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0265] In this embodiment, when training the model, the selection of model input variables is very important. Too many input variables will lead to excessive calculation costs and reduce the operating efficiency of the model. Therefore, this embodiment can determine the target meteorological features with a high degree of correlation with photovoltaic power from the initial meteorological features by calculating the correlation coefficient. For example, the Pearson correlation coefficient and / or the Spearson rank correlation coefficient of various initial meteorological features and photovoltaic power can be calculated. The Pearson correlation coefficient is often used to analyze the strength of the linear relationship between variables (i.e., various meteorological features and photovoltaic power), while the Spearson rank correlation coefficient only measures the monotonic relationship between variables, does not consider the influence of specific values, and is often used to measure the nonlinear relationship between variables. Therefore, preferably, the two correlation coefficients can be comprehensively considered, and the meteorological features with a high degree of correlation can be selected as the target meteorological features. For example, the weighted average of the two correlation coefficients can be calculated to obtain a comprehensive correlation coefficient. If the comprehensive correlation coefficient of the initial meteorological feature is greater than a predetermined threshold, the initial meteorological feature whose correlation coefficient meets the predetermined conditions is the target meteorological feature.

[0266] In this embodiment, after determining the target meteorological characteristics, the time series processing data of the target meteorological characteristics of each sample day in the plurality of processed sample data may be obtained, that is, the time series processing data of the meteorological characteristics in the above sample data.

[0267] In a possible implementation manner, the obtaining of initial meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days includes:

[0268] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0269] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0270] In this embodiment, the extreme values ​​that appear in the outlier sample and the processing of the outliers will directly affect the accuracy of subsequent predictions. This embodiment can use the physical discrimination method to analyze the rules of each initial meteorological feature and photovoltaic power to find outliers. For example, if the photovoltaic power must be greater than 0, then the photovoltaic power less than 0 can be determined as an outlier, and the outlier can be replaced by 0 or replaced by the average value of the photovoltaic power at two adjacent moments; the wind direction data should be between 0-360 degrees, and the wind direction data that is not in this range can be determined as an outlier, and the outlier can be replaced by 0 or replaced by the average value of the wind direction data at two adjacent moments. Moreover, the values ​​of the initial meteorological characteristics and photovoltaic power are subject to normal distribution, and the 3σ principle can be used to identify and process outliers. The 3σ principle is also called the Laida criterion. It first assumes that a set of data contains only random errors, calculates and processes them to obtain the standard deviation, and determines an interval according to a certain probability. It is believed that any error exceeding this interval is not a random error but a gross error. The value containing the gross error is determined as an outlier, and the outlier can be replaced by 0 or replaced by the average value of its two adjacent moments.

[0271] In this embodiment, the dimensions of the initial meteorological characteristic time series raw data and the photovoltaic power time series raw data are different. The large difference in dimensions and quantity sets will bring great difficulties to data analysis. In order to eliminate the impact of dimensions, this embodiment can normalize the initial meteorological characteristic time series raw data and the photovoltaic power time series raw data. For example, the initial meteorological characteristic time series raw data and the photovoltaic power time series raw data are mapped to [0,1]. The normalization processing here can be maximum and minimum value normalization, mean variance normalization, etc. Taking maximum and minimum value normalization as an example, the normalization formula of the photovoltaic power time series raw data is as follows:

[0272]

[0273] Among them, x is the value of photovoltaic power before normalization, x' is the value of photovoltaic power after normalization, and x min is the minimum photovoltaic power value in all photovoltaic power time series raw data, x max It is the maximum value of photovoltaic power in all photovoltaic power time series raw data.

[0274] In a possible implementation, the method may further include the following steps:

[0275] When the error between the photovoltaic power time series prediction data of the prediction day predicted by the prediction model and the photovoltaic power time series data of the prediction day exceeds a predetermined error, the meteorological characteristic time series data and the photovoltaic power time series data of the prediction day are determined as new sample data corresponding to the weather type of the prediction day;

[0276] Based on the new sample data corresponding to the weather type of the forecast day, the forecast model corresponding to the weather type of the forecast day is adjusted.

[0277] In this embodiment, after the prediction day has passed, the photovoltaic power time series data of the prediction day can be obtained, and the error between the photovoltaic power time series prediction data of the prediction day and the photovoltaic power time series data of the prediction day can be calculated, for example, the RMSE (Root Mean Squard Error), MAE (Mean Absolute Error), and R2 (R-Square, determination coefficient) between the two are calculated; if the error exceeds the predetermined error, it means that the photovoltaic power time series prediction data predicted by the prediction model is inaccurate. At this time, the meteorological characteristic time series data and the photovoltaic power time series data of the prediction day can be determined as new sample data corresponding to the weather type of the prediction day, and the new sample data can be used to adjust the prediction model corresponding to the weather type of the prediction day; in this way, the prediction model is continuously adjusted and optimized according to the prediction results, and the obtained prediction model will be more accurate over time.

[0278] The present disclosure also provides a photovoltaic power prediction device, Figure 3 The structure block diagram of the photovoltaic power prediction device according to the embodiment of the present disclosure is shown, and the device can be implemented as part or all of the electronic device through software, hardware or a combination of both. Figure 3 As shown, the photovoltaic power prediction device comprises:

[0279] The data acquisition module 301 is configured to acquire the weather forecast type and meteorological characteristic time series data of the forecast day, wherein the meteorological characteristic time series data includes the time series data of the target meteorological characteristic that is strongly related to the photovoltaic power;

[0280] The model acquisition module 302 is configured to acquire a prediction model corresponding to the weather prediction type from a pre-trained prediction model set as a target prediction model;

[0281] The prediction module 303 is configured to input the meteorological characteristic time series data of the prediction day into the target prediction model, execute the target prediction model, and obtain the photovoltaic power time series prediction data of the prediction day output by the target prediction model.

[0282] In a possible implementation, the device further includes:

[0283] The model training module is configured to obtain multiple sample data, the sample data including meteorological characteristic time series data and photovoltaic power time series data of the sample day; divide the multiple sample data into sample data corresponding to different weather types according to the weather type of the sample day; for each weather type, obtain multiple sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type; use the training sample data corresponding to the weather type to train a prediction model corresponding to the weather type.

[0284] In a possible implementation, the model training module obtains a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type, and the portion of the sample data for training corresponding to the weather type is configured as follows:

[0285] The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0286]

[0287] Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), that is, min|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k)|, max|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j); that is, max|y(j)-x i (j)| is|y(1)-x i(1)|、|y(2)-x i (2)|……|y(k)-x i (k) the maximum value among |

[0288] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0289] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0290]

[0291] in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector;

[0292] Refers to minimize;

[0293] The linear regression function of the weighted grey least squares support vector machine model is:

[0294]

[0295] Among them, α i is the Lagrange multiplier, α i ≥0, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0296] In a possible implementation, the part of acquiring multiple sample data in the model training module is configured as follows:

[0297] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0298] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0299] In a possible implementation manner, the part of the model training module that obtains the meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows:

[0300] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0301] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0302] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0303] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0304] In a possible implementation manner, the part of the model training module that obtains the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows:

[0305] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0306] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0307] In a possible implementation, the device further includes:

[0308] The model adjustment module is configured to determine the meteorological characteristic time series data and photovoltaic power time series data of the forecast day as new sample data corresponding to the weather type of the forecast day when the error between the photovoltaic power time series prediction data of the forecast day and the photovoltaic power time series data of the forecast day exceeds a predetermined error; and adjust the prediction model corresponding to the weather type of the forecast day based on the new sample data corresponding to the weather type of the forecast day.

[0309] The present disclosure also provides a training device for a prediction model for photovoltaic power prediction. Figure 4 The structure block diagram of the training device of the prediction model of photovoltaic power prediction according to the embodiment of the present disclosure is shown, and the device can be implemented as part or all of the electronic device through software, hardware or a combination of both. Figure 4 As shown, the training device comprises:

[0310] The sample acquisition module 401 is configured to acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of the sample day;

[0311] The sample division module 402 is configured to divide the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day;

[0312] The training module 403 is configured to train, for each weather type, a prediction model corresponding to the weather type using sample data corresponding to the weather type.

[0313] In a possible implementation, the training module is configured as follows:

[0314] For each weather type, obtaining a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type;

[0315] The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

[0316] In a possible implementation manner, the training module obtains a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type, and the portion of the sample data for training corresponding to the weather type is configured as follows:

[0317] The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0318]

[0319] Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x iy(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), that is, min|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k)|, max|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j); that is, max|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k) the maximum value among |

[0320] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0321] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0322]

[0323] in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector;

[0324] Refers to minimize;

[0325] The linear regression function of the weighted grey least squares support vector machine model is:

[0326]

[0327] Among them, α i is the Lagrange multiplier, α i ≥0, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0328] In a possible implementation, the sample acquisition module is configured as follows:

[0329] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0330] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0331] In a possible implementation manner, the part of the sample acquisition module that acquires the meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows:

[0332] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0333] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0334] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0335] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0336] In a possible implementation manner, the part of the sample acquisition module that acquires the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows:

[0337] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0338] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0339] In a possible implementation, the device further includes:

[0340] A new sample acquisition module is configured to determine the meteorological characteristic time series data and the photovoltaic power time series data of the forecast day as new sample data corresponding to the weather type of the forecast day when the error between the photovoltaic power time series prediction data of the forecast day predicted by the forecast model and the photovoltaic power time series data of the forecast day exceeds a predetermined error;

[0341] The adjustment module is configured to adjust the prediction model corresponding to the weather type of the forecast day based on the new sample data corresponding to the weather type of the forecast day.

[0342] The technical terms and technical features mentioned in the implementation manner of the present device are the same as or similar to those mentioned in the implementation manner of the above method. For the explanation and description of the technical terms and technical features involved in the present device, reference may be made to the explanation and description of the implementation manner of the above method, and they will not be repeated here.

[0343] The present disclosure also discloses an electronic device, Figure 5 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown.

[0344] like Figure 5 As shown, the electronic device 500 includes a memory 501 and a processor 502, wherein the memory 501 is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor 502 to implement the method according to an embodiment of the present disclosure.

[0345] In one possible implementation, the one or more computer instructions are executed by the processor 502 to implement a photovoltaic power prediction method according to an embodiment of the present disclosure, the photovoltaic power prediction method comprising:

[0346] Acquire the weather forecast type and meteorological characteristic time series data of the forecast day, wherein the meteorological characteristic time series data includes time series data of target meteorological characteristics that are strongly correlated with photovoltaic power;

[0347] Obtaining a prediction model corresponding to the weather prediction type from a set of pre-trained prediction models as a target prediction model;

[0348] The meteorological characteristic time series data of the forecast day is input into the target forecast model, and the target forecast model is executed to obtain the photovoltaic power time series forecast data of the forecast day output by the target forecast model.

[0349] In a possible implementation, the method further includes:

[0350] Acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day;

[0351] Dividing the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day;

[0352] For each weather type, obtaining a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type;

[0353] The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

[0354] In a possible implementation manner, obtaining, from the sample data corresponding to the weather type, a plurality of sample data with the highest correlation with the weather type as the training sample data corresponding to the weather type includes:

[0355] The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0356]

[0357] Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and xi (j), that is, min|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k)|, max|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j); that is, max|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k) the maximum value among |

[0358] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0359] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0360]

[0361] in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector;

[0362] Refers to minimize;

[0363] The linear regression function of the weighted grey least squares support vector machine model is:

[0364]

[0365] Among them, α i is the Lagrange multiplier, α i ≥0, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0366] In a possible implementation manner, the acquiring a plurality of sample data includes:

[0367] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0368] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0369] In a possible implementation manner, the step of acquiring the meteorological characteristic time series processing data and the photovoltaic power time series processing data of a plurality of sample days includes:

[0370] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0371] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0372] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0373] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0374] In a possible implementation manner, the obtaining of initial meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days includes:

[0375] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0376] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0377] In a possible implementation, the method further includes:

[0378] When the error between the photovoltaic power time series prediction data on the forecast day and the photovoltaic power time series data on the forecast day exceeds a predetermined error, the meteorological characteristic time series data and the photovoltaic power time series data on the forecast day are determined as new sample data corresponding to the weather type on the forecast day;

[0379] Based on the new sample data corresponding to the weather type of the forecast day, the forecast model corresponding to the weather type of the forecast day is adjusted.

[0380] In a possible implementation, the one or more computer instructions are executed by the processor 502 to implement a training method for a prediction model for photovoltaic power prediction according to an embodiment of the present disclosure, and the training method for the prediction model includes:

[0381] Acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day;

[0382] Dividing the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day;

[0383] For each weather type, sample data corresponding to the weather type is used to train a prediction model corresponding to the weather type.

[0384] In a possible implementation, the using the sample data corresponding to the weather type to train a prediction model corresponding to the weather type includes:

[0385] From the sample data corresponding to the weather type, obtain a plurality of sample data with the highest correlation with the weather type as training sample data corresponding to the weather type;

[0386] The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

[0387] In a possible implementation manner, obtaining, from the sample data corresponding to the weather type, a plurality of sample data with the highest correlation with the weather type as the training sample data corresponding to the weather type includes:

[0388] The grey correlation between the target day corresponding to the weather type and each sample day in the sample data corresponding to the weather type is calculated according to the following formula:

[0389]

[0390] Among them, r iis the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), that is, min|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k)|, max|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j); that is, max|y(j)-x i (j)| is|y(1)-x i (1)|、|y(2)-x i (2)|……|y(k)-x i (k) the maximum value among |

[0391] The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

[0392] In a possible implementation, the prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is:

[0393]

[0394] in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction modeli The corresponding output vector;

[0395] Refers to minimize;

[0396] The linear regression function of the weighted grey least squares support vector machine model is:

[0397]

[0398] Among them, α i is the Lagrange multiplier, α i ≥0, is a nonlinear function that maps the input x to a high-dimensional feature space.

[0399] In a possible implementation manner, the acquiring a plurality of sample data includes:

[0400] Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day;

[0401] The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

[0402] In a possible implementation manner, the step of acquiring the meteorological characteristic time series processing data and the photovoltaic power time series processing data of a plurality of sample days includes:

[0403] Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0404] Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power;

[0405] determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature;

[0406] The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

[0407] In a possible implementation manner, the obtaining of initial meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days includes:

[0408] Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day;

[0409] The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

[0410] In a possible implementation, the method further includes:

[0411] When the error between the photovoltaic power time series prediction data of the prediction day predicted by the prediction model and the photovoltaic power time series data of the prediction day exceeds a predetermined error, the meteorological characteristic time series data and the photovoltaic power time series data of the prediction day are determined as new sample data corresponding to the weather type of the prediction day;

[0412] Based on the new sample data corresponding to the weather type of the forecast day, the forecast model corresponding to the weather type of the forecast day is adjusted.

[0413] The present disclosure also discloses a distribution network system, which includes the above Figure 5 The electronic equipment shown. The distribution network system includes the main distribution station, the sub-distribution station (usually located in the substation, optional), the remote distribution terminal (such as FTU, DTU, etc.) and the communication network. The main distribution station is located in the city dispatching center, the sub-distribution station is deployed in the substation, the sub-station is responsible for communicating with the DTU / TTU and other power terminal equipment in the area under its jurisdiction, and the main station is responsible for communicating with each sub-station. Figure 5 The electronic device shown can be any electronic device in the distribution network system. The electronic device can accurately and efficiently predict the photovoltaic power generation power. Based on the prediction ability of the electronic device, the distribution network system can realize effective photovoltaic equipment management, photovoltaic power generation grid-connected regulation, and efficient linkage of source, grid, load and storage, etc., so that the efficiency and benefits of distribution network management are greatly improved.

[0414] The disclosed embodiments also provide a chip, which may be any chip that can implement a photovoltaic power prediction device, or, the chip may also be any chip that can implement a training device for the above-mentioned prediction model, and the device may be implemented as part or all of the chip through software, hardware, or a combination of both.

[0415] Figure 6A schematic diagram showing the structure of a computer system suitable for implementing the method of the embodiment of the present disclosure is shown.

[0416] like Figure 6 As shown, the computer system 600 includes a processing unit 601, which can perform various processes in the above-mentioned embodiments according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the computer system 600 are also stored. The processing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0417] The following components are connected to the I / O interface 605: an input part 606 including a keyboard, a mouse, etc.; an output part 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage part 608 including a hard disk, etc.; and a communication part 609 including a network interface card such as a LAN card, a modem, etc. The communication part 609 performs communication processing via a network such as the Internet. The drive 610 is also connected to the I / O interface 605 as needed. Removable media 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., are installed on the drive 610 as needed, so that the computer program read therefrom is installed into the storage part 608 as needed. Among them, the processing unit 601 can be implemented as a processing unit such as a CPU, a GPU, a TPU, an FPGA, an NPU, etc.

[0418] In particular, according to an embodiment of the present disclosure, the method described above can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes computer instructions, and the computer instructions are executed by a processor to implement the method steps described above. In such an embodiment, the computer program product can be downloaded and installed from a network through the communication part 609, and / or installed from a removable medium 611.

[0419] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0420] The units or modules involved in the embodiments described in the present disclosure may be implemented by software or by programmable hardware. The units or modules described may also be set in a processor, and the names of these units or modules do not constitute limitations on the units or modules themselves in some cases.

[0421] As another aspect, the present disclosure further provides a computer-readable storage medium, which may be a computer-readable storage medium included in the electronic device or computer system in the above embodiment; or a computer-readable storage medium that exists independently and is not assembled into a device. The computer-readable storage medium stores one or more programs, and the programs are used by one or more processors to execute the method described in the present disclosure.

[0422] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present disclosure is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other.

Claims

1. A photovoltaic power prediction method, characterized in that: include: Acquire the weather forecast type and meteorological characteristic time series data of the forecast day, wherein the meteorological characteristic time series data includes time series data of target meteorological characteristics that are strongly correlated with photovoltaic power; Obtaining a prediction model corresponding to the weather prediction type from a set of pre-trained prediction models as a target prediction model; The meteorological characteristic time series data of the forecast day is input into the target forecast model, and the target forecast model is executed to obtain the photovoltaic power time series forecast data of the forecast day output by the target forecast model.

2. The method according to claim 1, characterized in that The method further comprises: Acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day; Dividing the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day; For each weather type, obtaining a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type; The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

3. The method according to claim 2, characterized in that The step of acquiring, from the sample data corresponding to the weather type, a plurality of sample data with the highest correlation with the weather type as training sample data corresponding to the weather type includes: The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula: Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), the minimum absolute difference, maxy(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j) the maximum absolute difference; The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

4. The method according to claim 3, characterized in that The prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is: in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector; The linear regression function of the weighted grey least squares support vector machine model is: Among them, α i is the Lagrange multiplier, is a nonlinear function that maps the input x to a high-dimensional feature space.

5. The method according to claim 2, characterized in that: The obtaining of a plurality of sample data comprises: Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day; The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

6. The method according to claim 5, characterized in that The step of obtaining the meteorological characteristic time series processing data and the photovoltaic power time series processing data of a plurality of sample days includes: Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day; Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power; determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature; The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

7. The method according to claim 6, characterized in that The step of obtaining initial meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days includes: Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day; The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

8. The method according to claim 2, characterized in that: The method further comprises: When the error between the photovoltaic power time series prediction data on the forecast day and the photovoltaic power time series data on the forecast day exceeds a predetermined error, the meteorological characteristic time series data and the photovoltaic power time series data on the forecast day are determined as new sample data corresponding to the weather type on the forecast day; Based on the new sample data corresponding to the weather type of the forecast day, the forecast model corresponding to the weather type of the forecast day is adjusted.

9. A method for training a prediction model for photovoltaic power prediction, characterized in that: include: Acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day; Dividing the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day; For each weather type, sample data corresponding to the weather type is used to train a prediction model corresponding to the weather type.

10. The method according to claim 9, characterized in that The using the sample data corresponding to the weather type to train a prediction model corresponding to the weather type includes: From the sample data corresponding to the weather type, obtain a plurality of sample data with the highest correlation with the weather type as training sample data corresponding to the weather type; The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

11. The method according to claim 10, characterized in that The step of acquiring, from the sample data corresponding to the weather type, a plurality of sample data with the highest correlation with the weather type as training sample data corresponding to the weather type includes: The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula: Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), the minimum absolute difference, maxy(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j) the maximum absolute difference; The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

12. The method according to claim 11, characterized in that The prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is: in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector; The linear regression function of the weighted grey least squares support vector machine model is: Among them, α i is the Lagrange multiplier, is a nonlinear function that maps the input x to a high-dimensional feature space.

13. The method according to claim 9 or 10, characterized in that: The obtaining of a plurality of sample data comprises: Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day; The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

14. The method according to claim 13, characterized in that The step of obtaining the meteorological characteristic time series processing data and the photovoltaic power time series processing data of a plurality of sample days includes: Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day; Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power; determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature; The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

15. The method according to claim 14, characterized in that The step of obtaining initial meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days includes: Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day; The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

16. The method according to claim 9 or 10, characterized in that The method further comprises: When the error between the photovoltaic power time series prediction data of the prediction day predicted by the prediction model and the photovoltaic power time series data of the prediction day exceeds a predetermined error, the meteorological characteristic time series data and the photovoltaic power time series data of the prediction day are determined as new sample data corresponding to the weather type of the prediction day; Based on the new sample data corresponding to the weather type of the forecast day, the forecast model corresponding to the weather type of the forecast day is adjusted.

17. A photovoltaic power prediction device, characterized in that: include: A data acquisition module is configured to acquire a weather forecast type and meteorological characteristic time series data of a forecast day, wherein the meteorological characteristic time series data includes time series data of target meteorological characteristics that are strongly correlated with photovoltaic power; A model acquisition module is configured to acquire a prediction model corresponding to the weather prediction type from a pre-trained prediction model set as a target prediction model; The prediction module is configured to input the meteorological characteristic time series data of the prediction day into the target prediction model, execute the target prediction model, and obtain the photovoltaic power time series prediction data of the prediction day output by the target prediction model.

18. The device according to claim 17, characterized in that The device also includes: The model training module is configured to obtain multiple sample data, the sample data including meteorological characteristic time series data and photovoltaic power time series data of the sample day; divide the multiple sample data into sample data corresponding to different weather types according to the weather type of the sample day; for each weather type, obtain multiple sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type; use the training sample data corresponding to the weather type to train a prediction model corresponding to the weather type.

19. The device according to claim 18, characterized in that The model training module obtains a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type, and the portion of the sample data for training corresponding to the weather type is configured as follows: Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), the minimum absolute difference, max|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j) the maximum absolute difference; The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

20. The device according to claim 18, characterized in that The prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is: in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector; The linear regression function of the weighted grey least squares support vector machine model is: Among them, α i is the Lagrange multiplier, is a nonlinear function that maps the input x to a high-dimensional feature space.

21. The device according to claim 18, characterized in that The part of acquiring multiple sample data in the model training module is configured as follows: Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day; The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

22. The device according to claim 21, characterized in that The part of the model training module that obtains the meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows: Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day; Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power; determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature; The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

23. The device according to claim 22, characterized in that The part of the model training module that obtains the initial meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days is configured as follows: Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day; The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

24. The device according to claim 18, characterized in that The device also includes: The model adjustment module is configured to determine the meteorological characteristic time series data and photovoltaic power time series data of the forecast day as new sample data corresponding to the weather type of the forecast day when the error between the photovoltaic power time series prediction data of the forecast day and the photovoltaic power time series data of the forecast day exceeds a predetermined error; and adjust the prediction model corresponding to the weather type of the forecast day based on the new sample data corresponding to the weather type of the forecast day.

25. A training device for a prediction model for photovoltaic power prediction, characterized in that: include: A sample acquisition module is configured to acquire a plurality of sample data, wherein the sample data includes meteorological characteristic time series data and photovoltaic power time series data of a sample day; A sample division module is configured to divide the plurality of sample data into sample data corresponding to different weather types according to the weather type of the sample day; The training module is configured to train, for each weather type, a prediction model corresponding to the weather type using sample data corresponding to the weather type.

26. The device according to claim 25, characterized in that The training module is configured as follows: For each weather type, obtaining a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type as training sample data corresponding to the weather type; The prediction model corresponding to the weather type is trained using the training sample data corresponding to the weather type.

27. The device according to claim 25, characterized in that The training module obtains a plurality of sample data with the highest correlation with the weather type from the sample data corresponding to the weather type, and the portion of the sample data for training corresponding to the weather type is configured as follows: The grey correlation between the target day corresponding to the weather type and the target meteorological characteristics of each sample day in the sample data corresponding to the weather type is calculated according to the following formula: Among them, r i is the grey correlation degree of the i-th sample day; N is the total number of types of target meteorological features in the meteorological feature time series data; n is the total number of sample days; ρ is the preset resolution coefficient; y(k) is the characteristic value of the k-th target meteorological feature of the target day corresponding to the weather type, x i y(k) is the characteristic value of the kth target meteorological feature on the ith sample day in the sample data corresponding to the weather type; y(j) is the characteristic value of the jth target meteorological feature on the target day corresponding to the weather type, x i (j) is the characteristic value of the jth target meteorological characteristic on the i-th sample day in the sample data corresponding to the weather type, min|y(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j), the minimum absolute difference, maxy(j)-x i (j)|When j takes the value 1, 2, ... k, y(j) and x i (j) the maximum absolute difference; The sample data of multiple sample days with the highest grey correlation degree are used as training sample data corresponding to the weather type.

28. The device according to claim 27, characterized in that The prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is: The prediction model is a weighted grey least squares support vector machine model, and the optimization problem of the weighted grey least squares support vector machine model is: in, To convert the input vector x in the i-th training sample data i A nonlinear function mapped to a high-dimensional feature space, w is the weight vector, b is the bias value, C represents the correction factor, C>0, and the insensitive loss function e of the i-th training sample data is i Corresponding to a weight v i , m is the number of training sample data corresponding to the weather type, y i The input vector x output by the prediction model i The corresponding output vector; The linear regression function of the weighted grey least squares support vector machine model is: Among them, α i is the Lagrange multiplier, is a nonlinear function that maps the input x to a high-dimensional feature space.

29. The device according to claim 25 or 26, characterized in that The sample acquisition module is configured as follows: Acquire meteorological characteristic time series processing data and photovoltaic power time series processing data of multiple sample days, wherein the meteorological characteristic time series processing data includes processing data of target meteorological characteristics corresponding to multiple moments of the sample day, and the photovoltaic power time series processing data includes processing data of photovoltaic power corresponding to multiple moments of the sample day; The meteorological characteristic time series processing data and the photovoltaic power time series processing data of the multiple sample days are weakened by using an average weakening buffer operator to obtain the multiple sample data.

30. The device according to claim 29, characterized in that The part of the sample acquisition module that acquires the meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows: Acquire time series processing data of initial meteorological characteristics and time series processing data of photovoltaic power for multiple sample days; the time series processing data of initial meteorological characteristics includes processing data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day; Based on the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of the plurality of sample days, calculating the correlation coefficient between various initial meteorological characteristics in the initial meteorological characteristic time series processing data and the photovoltaic power; determining the initial meteorological feature whose correlation coefficient meets the predetermined condition as the target meteorological feature; The processed data of the target meteorological characteristics corresponding to the multiple moments of each sample day are determined as the meteorological characteristic time series processed data of each sample day.

31. The device according to claim 30, characterized in that The part of the sample acquisition module that acquires the initial meteorological characteristic time series processing data and the photovoltaic power time series processing data of multiple sample days is configured as follows: Acquire initial meteorological characteristic time series raw data and photovoltaic power time series raw data of multiple sample days; the initial meteorological characteristic time series raw data includes raw data of various initial meteorological characteristics related to photovoltaic power corresponding to multiple moments of the sample day; The initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days are subjected to outlier processing and normalization processing to obtain the initial meteorological characteristic time series original data and the photovoltaic power time series original data of the multiple sample days.

32. The device according to claim 25 or 26, characterized in that The device also includes: A new sample acquisition module is configured to determine the meteorological characteristic time series data and the photovoltaic power time series data of the forecast day as new sample data corresponding to the weather type of the forecast day when the error between the photovoltaic power time series prediction data of the forecast day predicted by the forecast model and the photovoltaic power time series data of the forecast day exceeds a predetermined error; The adjustment module is configured to adjust the prediction model corresponding to the weather type of the forecast day based on the new sample data corresponding to the weather type of the forecast day.

33. An electronic device, characterized in that: The method comprises a memory and a processor, wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the method according to any one of claims 1 to 16.

34. A power distribution system, characterized in that: An electronic device comprising the electronic device described in claim 33.

35. A readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed by a processor, the method according to any one of claims 1 to 16 is implemented.

36. A chip, characterized in that: The chip comprises the device according to any one of claims 17-32.

Citation Information

Cited By

  • Photovoltaic power station abnormity monitoring method and system based on large model data analysis

    CN120433459A