Photovoltaic power prediction method and system based on multi-source data fusion and feature dimension reduction
By integrating multi-source data and feature dimensionality reduction methods in photovoltaic power prediction, the problem of failing to make full use of historical photovoltaic data and overfitting in traditional methods is solved, achieving higher prediction accuracy and lower computing requirements.
Patent Information
- Application Number
- CN202411374387.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-29
- Publication Date
- 2025-02-28
AI Technical Summary
The traditional photovoltaic power prediction method is too one-sided in the data set structure, fails to fully consider the impact of historical photovoltaic data, and does not perform feature dimensionality reduction, resulting in an increase in model calculation volume and overfitting, affecting the prediction accuracy.
A photovoltaic power prediction method based on multi-source data fusion and characteristic dimensionality reduction is proposed. By extracting high autocorrelation historical photovoltaic data and meteorological environment factors, cross-correlation analysis of input characteristics and output characteristics is carried out, meteorological environment factors and historical photovoltaic data are decomposed as inherent mode function components at different frequencies, and data dimensions are reduced through nuclear principal component dimensionality reduction technology. Finally, a two-way long and short-term memory network model is used for training and prediction.
It effectively avoids overfitting, reduces model complexity and computing requirements, improves prediction accuracy and operating efficiency, and enhances the generalization ability of the model, significantly improving the accuracy of photovoltaic power prediction.
Smart Images
Figure CN119543094A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of photovoltaic power prediction, and in particular to a photovoltaic power prediction method and system based on multi-source data fusion and feature dimension reduction. Background Art
[0002] The accuracy of photovoltaic power prediction affects the photovoltaic absorption capacity, the formulation of dispatch plans for new power systems, and the operating stability of the entire power system.
[0003] The traditional method is somewhat one-sided in the construction of the data set. It only considers the impact of meteorological and environmental factors on photovoltaics, but does not consider that historical photovoltaics will also have an impact on predicted photovoltaics. In addition, the traditional method often decomposes the features without performing dimensionality reduction processing, which will lead to a sharp increase in the amount of calculation of the prediction model, overfitting of the model, and limited prediction accuracy. Summary of the invention
[0004] In view of the above-mentioned problems, the present invention is proposed.
[0005] Therefore, the technical problem solved by the present invention is: how to provide a photovoltaic power prediction method that can both decompose and reduce dimensions.
[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0007] In a first aspect, an embodiment of the present invention provides a photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, comprising:
[0008] Extract historical photovoltaic data with high autocorrelation and construct the initial data set in combination with meteorological environmental factors;
[0009] Perform cross-correlation analysis on the input features and output features of the initial data set, extract highly correlated features, and construct a candidate data set;
[0010] The meteorological environmental factors and historical photovoltaic data in the initial data set are decomposed into intrinsic mode function components at different frequencies, and the final input features are obtained by dimensionality reduction. The final data set is constructed by combining the output photovoltaic power.
[0011] Perform training and prediction on the final dataset.
[0012] As a preferred solution of photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, the following is the solution:
[0013] The extracting of high autocorrelation historical photovoltaic data comprises:
[0014] The autocorrelation function method is used to extract historical photovoltaic data with a correlation greater than 0.85 with the predicted day.
[0015] As a preferred solution of photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, the following is the solution:
[0016] The cross-correlation analysis of input features and output features of the initial data set includes:
[0017] The input features in the initial data set include meteorological environmental factors and historical photovoltaic data, and the output features in the initial data set include predicted photovoltaic power.
[0018] As a preferred solution of photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, the following is the solution:
[0019] The cross-correlation analysis of input features and output features of the initial data set further includes:
[0020] The maximum mutual information coefficient method is used to perform cross-correlation analysis on the input features and output features of the initial data set, and features with a cross-correlation greater than 0.9 are extracted and combined with the output features as candidate data sets.
[0021] As a preferred solution of photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, the following is the solution:
[0022] Decomposing the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies includes:
[0023] The empirical wavelet decomposition method is used to decompose the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies.
[0024] As a preferred solution of photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, the following is the solution:
[0025] The final input features obtained by dimension reduction include:
[0026] The kernel principal component dimensionality reduction technique is used to reduce the dimensionality, map the intrinsic mode function components at different frequencies to the high-dimensional space, and reduce the data dimension through principal component analysis to retain the most representative features.
[0027] As a preferred solution of photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, the following is the solution:
[0028] The training and prediction of the final data set includes:
[0029] The bidirectional long short-term memory network model is used to train and predict the final dataset.
[0030] In a second aspect, an embodiment of the present invention provides a photovoltaic power prediction system based on multi-source data fusion and feature dimension reduction, comprising:
[0031] Initialization module, used to extract historical photovoltaic data with high autocorrelation and construct the initial data set in combination with meteorological environmental factors;
[0032] The candidate construction module is used to perform cross-correlation analysis on the input features and output features of the initial data set, extract highly correlated features, and construct a candidate data set;
[0033] The data set construction module is used to decompose the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies, reduce the dimension to obtain the final input features, and construct the final data set in combination with the output photovoltaic power;
[0034] The training prediction module is used to train and predict the final data set.
[0035] In a third aspect, an embodiment of the present invention provides a computing device, including:
[0036] Memory and processor;
[0037] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the one or more programs are executed by the one or more processors, the one or more processors implement the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction as described in any embodiment of the present invention.
[0038] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction.
[0039] The beneficial effects of the present invention are as follows: the system of the present invention extracts features related to photovoltaic power prediction to ensure that the data set covers key environmental factors and historical data; through the kernel principal component dimensionality reduction technology, irrelevant features are removed, the data structure is optimized, and the model complexity is reduced; feature selection is combined with dimensionality reduction to effectively avoid overfitting and make the model more stable on unseen data; the input feature dimension is reduced, the computing requirements during training and prediction are reduced, and the operating efficiency is improved; through in-depth analysis of environmental factors and historical data, combined with a bidirectional long short-term memory network, the prediction accuracy is significantly improved; meteorological environmental factors are taken into account to enable the model to better adapt to different climatic conditions; while improving the comprehensiveness of the input features, the irrelevant feature dimensions can be reduced, thereby improving the generalization ability of the model; the amount of calculation can be effectively reduced and the prediction accuracy of the model can be improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.
[0041] Figure 1 It is an overall flow chart of the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction described in the first embodiment of the present invention;
[0042] Figure 2 is a graph showing the calculation results of the photovoltaic power autocorrelation coefficient in a simulation example of the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction described in the second embodiment of the present invention;
[0043] Figure 3 It is a local enlarged diagram of the calculation result of the photovoltaic power autocorrelation coefficient in the simulation example of the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction described in the second embodiment of the present invention;
[0044] Figure 4 is a diagram of the calculation results of the maximum mutual correlation coefficient in a simulation example of the photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction described in the second embodiment of the present invention;
[0045] Figure 5 It is a characteristic component diagram under the decomposition technology in the simulation example of the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction described in the second embodiment of the present invention;
[0046] Figure 6 It is a feature contribution graph under the dimensionality reduction technology in a simulation example of a photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction described in the second embodiment of the present invention;
[0047] Figure 7 is a comparison result diagram of features before and after dimensionality reduction in a simulation example of a photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction described in a second embodiment of the present invention;
[0048] Figure 8 is a comparison diagram of model prediction results in a simulation example of a photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction according to a second embodiment of the present invention;
[0049] Fig. 9 It is a comparison diagram of model prediction errors in a simulation example of a photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction described in the second embodiment of the present invention. DETAILED DESCRIPTION
[0050] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.
[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.
[0052] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.
[0053] Example 1
[0054] Reference Figure 1 , which is the first embodiment of the present invention, provides a photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, including:
[0055] S1: Extract historical photovoltaic data with high autocorrelation and construct the initial data set in combination with meteorological environmental factors;
[0056] In the embodiment of the present application, extracting historical photovoltaic data with high autocorrelation includes:
[0057] The autocorrelation function method is used to extract historical photovoltaic data with a correlation greater than 0.85 with the predicted day.
[0058] Specifically, the autocorrelation function method is used to analyze the autocorrelation between the predicted daily photovoltaic and the historical photovoltaic, and the photovoltaic with high autocorrelation is extracted as the candidate input features. The maximum mutual information coefficient method is used to determine the highly correlated environmental factors and candidate input features that affect the photovoltaic power, and construct the initial data set.
[0059] S2: Perform cross-correlation analysis on the input features and output features of the initial data set, extract highly correlated features, and construct a candidate data set;
[0060] In the embodiment of the present application, performing cross-correlation analysis of input features and output features on the initial data set includes:
[0061] The input features in the initial data set include meteorological environmental factors and historical photovoltaic data, and the output features in the initial data set include predicted photovoltaic power;
[0062] The maximum mutual information coefficient method is used to perform cross-correlation analysis on the input features and output features of the initial data set, and features with a cross-correlation greater than 0.9 are extracted and combined with the output features as candidate data sets.
[0063] S3: Decompose the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies, reduce the dimension to obtain the final input features, and combine the output photovoltaic power to construct the final data set;
[0064] In the embodiment of the present application, decomposing the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies includes:
[0065] The empirical wavelet decomposition method is used to decompose the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies.
[0066] The final input features obtained by dimensionality reduction include:
[0067] The kernel principal component dimensionality reduction technique is used to reduce the dimensionality, map the intrinsic mode function components at different frequencies to the high-dimensional space, and reduce the data dimension through principal component analysis to retain the most representative features.
[0068] It should be noted that the use of empirical wavelet decomposition method can better extract the key factors of input features; the use of kernel principal component dimensionality reduction technology can ensure that the key factors are extracted while reducing the dimension of input features, thereby reducing the computational complexity of the model.
[0069] S4: Train and predict on the final dataset.
[0070] In the embodiment of the present application, training and predicting the final data set includes:
[0071] The bidirectional long short-term memory network model is used to train and predict the final dataset.
[0072] The above is a schematic scheme of the photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction of this embodiment. It should be noted that the technical scheme of the photovoltaic power prediction system based on multi-source data fusion and feature dimensionality reduction and the technical scheme of the photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction mentioned above belong to the same concept. The details of the technical scheme of the photovoltaic power prediction system based on multi-source data fusion and feature dimensionality reduction in this embodiment that are not described in detail can all be referred to the description of the technical scheme of the photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction mentioned above.
[0073] The photovoltaic power prediction system based on multi-source data fusion and feature dimension reduction in this embodiment includes:
[0074] Initialization module, used to extract historical photovoltaic data with high autocorrelation and construct the initial data set in combination with meteorological environmental factors;
[0075] The candidate construction module is used to perform cross-correlation analysis on the input features and output features of the initial data set, extract highly correlated features, and construct a candidate data set;
[0076] The data set construction module is used to decompose the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies, reduce the dimension to obtain the final input features, and construct the final data set in combination with the output photovoltaic power;
[0077] The training prediction module is used to train and predict the final data set.
[0078] This embodiment further provides a computing device, which is applicable to the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, including:
[0079] Memory and processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction proposed in the above embodiment.
[0080] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction proposed in the above embodiment is implemented.
[0081] The storage medium proposed in this embodiment and the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction proposed in the above embodiment belong to the same inventive concept. The technical details not described in detail in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.
[0082] Example 2
[0083] Reference Figure 2-Figure 9 , which is an embodiment of the present invention, provides a photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.
[0084] Depend on Figure 2It can be seen that by analyzing the photovoltaic power sequence using the autocorrelation function method, the closer the photovoltaic power is to the predicted date and time, the larger the autocorrelation coefficient is. Conversely, the farther away from the predicted date and time, the smaller the autocorrelation coefficient is and the smaller the correlation is. The autocorrelation of the predicted simultaneous points is generally greater than the autocorrelation coefficient of the photovoltaic power at non-simultaneous points.
[0085] Depend on Figure 3 It can be seen that there are 8 time points with autocorrelation greater than 0.85, namely 5 time points on the same day, 5 time points one day ago, and the same time point two days ago, as shown in the circle.
[0086] Depend on Figure 4 It can be seen that: the figure shows the mutual correlation between 18 input features and output features in the initial data set, and the features with a mutual correlation greater than 0.9 are selected as candidate input features, that is, 8 features are selected, including total radiation, direct radiation, scattered radiation, photovoltaic power at a time point 1 lag behind the current day, photovoltaic power at the same time point 2 days ago, photovoltaic power at the same time point +1 1 day ago, photovoltaic power at the same time point -1 1 day ago, and photovoltaic power at the same time point 1 day ago.
[0087] Depend on Figure 5 , 6 From 7, we can see that the 8 features are decomposed into 7, 7, 8, 7, 8, 7, 7 components respectively, totaling 58 component features. If the feature contribution rate is greater than 95%, one feature is selected as the final input feature of the data set.
[0088] Depend on Figure 8 , 9 It can be seen that the model provided by this method is closer to the true value in the curve fitting method. The prediction effect of the comparison model 2 is second, and the prediction effect of the comparison model 1 is the worst. This is because there are too many input features and the model is overfitting.
[0089] It can be seen from Table 1 that the prediction accuracy of the model provided by this method is 97.87%, the prediction accuracy of the comparison model 1 is 60.22%, and the prediction accuracy of the comparison model 2 is 93.35%, which proves that the model prediction effect provided by this method is the best.
[0090] Table 1 Model prediction performance evaluation index of the present invention
[0091]
[0092] Table 2 Model types and meanings of the present invention
[0093]
[0094]
[0095] In summary, after verification by actual photovoltaic power data, the model proposed in the present invention has high prediction accuracy.
[0096] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.
Claims
1. A photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction, characterized in that: include: Extract historical photovoltaic data with high autocorrelation and construct the initial data set in combination with meteorological environmental factors; Perform cross-correlation analysis on the input features and output features of the initial data set, extract highly correlated features, and construct a candidate data set; The meteorological environmental factors and historical photovoltaic data in the initial data set are decomposed into intrinsic mode function components at different frequencies, and the final input features are obtained by dimensionality reduction. The final data set is constructed by combining the output photovoltaic power. Perform training and prediction on the final dataset.
2. The photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction according to claim 1, characterized in that: The extracting of high autocorrelation historical photovoltaic data comprises: The autocorrelation function method is used to extract historical photovoltaic data with a correlation greater than 0.85 with the predicted day.
3. The photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction as claimed in claim 2, characterized in that: The cross-correlation analysis of input features and output features of the initial data set includes: The input features in the initial data set include meteorological environmental factors and historical photovoltaic data, and the output features in the initial data set include predicted photovoltaic power.
4. The photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction as claimed in claim 3, characterized in that: The cross-correlation analysis of input features and output features of the initial data set further includes: The maximum mutual information coefficient method is used to perform cross-correlation analysis on the input features and output features of the initial data set, and features with a cross-correlation greater than 0.9 are extracted and combined with the output features as candidate data sets.
5. The photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction according to claim 4, characterized in that: Decomposing the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies includes: The empirical wavelet decomposition method is used to decompose the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies.
6. The photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction according to claim 5, characterized in that: The final input features obtained by dimension reduction include: The kernel principal component dimensionality reduction technique is used to reduce the dimensionality, map the intrinsic mode function components at different frequencies to the high-dimensional space, and reduce the data dimension through principal component analysis to retain the most representative features.
7. The photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction according to claim 6, characterized in that: The training and prediction of the final data set includes: The bidirectional long short-term memory network model is used to train and predict the final dataset.
8. A system using the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction as described in any one of claims 1 to 7, characterized in that: include: Initialization module, used to extract historical photovoltaic data with high autocorrelation and construct the initial data set in combination with meteorological environmental factors; The candidate construction module is used to perform cross-correlation analysis on the input features and output features of the initial data set, extract highly correlated features, and construct a candidate data set; The data set construction module is used to decompose the meteorological environmental factors and historical photovoltaic data in the initial data set into intrinsic mode function components at different frequencies, reduce the dimension to obtain the final input features, and construct the final data set in combination with the output photovoltaic power; The training prediction module is used to train and predict the final data set.
9. A computing device comprising: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the photovoltaic power prediction method based on multi-source data fusion and feature dimensionality reduction as described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the photovoltaic power prediction method based on multi-source data fusion and feature dimension reduction as described in any one of claims 1 to 7.