Photovoltaic power generation capacity prediction method and device and storage medium

By combining the volatility-weighted LASSO model and the PSO-SVM model, the problem of low accuracy in photovoltaic power generation prediction is solved, achieving higher prediction accuracy and adaptability, especially in handling long-term data and data volatility.

CN118940032APending Publication Date: 2024-11-12HUNAN HENGDIAN INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410968082.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

In the existing photovoltaic power generation prediction methods, the LSTM algorithm has the problem of low accuracy in medium-term data applications, which is difficult to effectively solve long-term data problems.

Method used

Using a combination of volatility-weighted LASSO model and PSO-SVM prediction model, the volatility-weighted LASSO model is trained in real time through feature reconstruction and data cleaning, and the pre-trained PSO-SVM model is called for power generation prediction, which improves the prediction accuracy.

Benefits of technology

It improves the accuracy of photovoltaic power generation prediction, can better consider the timing changes and volatility of the data, and enhances the robustness and adaptability of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940032A_ABST
    Figure CN118940032A_ABST
Patent Text Reader

Abstract

The invention relates to the field of data processing, and discloses a photovoltaic generating capacity prediction method and device, and a storage medium. The method comprises the following steps: when photovoltaic power generation data is obtained, generating a feature data set according to the photovoltaic power generation data; performing feature reconstruction on the feature data set to obtain a feature matrix set; according to the feature matrix set, training a volatility weighted LASSO model in real time, calling the volatility weighted LASSO model, and predicting a future prediction value set corresponding to the feature matrix set; and calling a pre-trained PSO-SVM prediction model, and executing a generating capacity prediction operation according to the future prediction value set to obtain a generating capacity prediction value. In the embodiment of the invention, the introduction of the volatility weighting technology helps to better consider the time sequence change and volatility of the data in the model training process. The problem that the LASSO model independently deals with long-term data and the data volatility is limited is solved. And thus, the accuracy of photovoltaic generating capacity prediction is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a photovoltaic power generation prediction method, device and storage medium. Background Art

[0002] With the continuous increase of renewable energy, photovoltaic power generation plays an increasingly important role in regional power grids. Photovoltaic power generation forecasting is crucial to regional power grid operations, helping to optimize power dispatching, ensure supply and demand balance, and improve system reliability and economy.

[0003] At present, some progress has been made in the research of photovoltaic power generation prediction. For example, the medium- and long-term photovoltaic power generation of the Long Short-Term Memory (LSTM) neural network was used. The experimental data set used the long-term historical power generation of the photovoltaic power station and the light intensity around the power station. Due to the limited memory capacity of the cell unit of the LSTM algorithm, it has a good application in medium-term data and is difficult to solve the problem of long-term data. This leads to low accuracy in photovoltaic power generation prediction. Summary of the invention

[0004] The main purpose of the present invention is to solve the technical problem of low accuracy in photovoltaic power generation prediction.

[0005] A first aspect of the present invention provides a method for predicting photovoltaic power generation. The method for predicting photovoltaic power generation comprises: When the photovoltaic power generation data is obtained, a feature data set is generated according to the photovoltaic power generation data; Performing feature reconstruction on the feature data set to obtain a feature matrix set; According to the feature matrix set, a volatility-weighted LASSO model is trained in real time, and the volatility-weighted LASSO model is called to predict a future prediction value set corresponding to the feature matrix set; The pre-trained PSO-SVM prediction model is called, and a power generation prediction operation is performed according to the future prediction value set to obtain a power generation prediction value.

[0006] Optionally, in a first implementation of the first aspect of the present invention, the step of training the volatility-weighted LASSO model in real time according to the feature matrix set includes: Calculating the volatility weighted data corresponding to each feature matrix in the feature matrix set to obtain a volatility weighted data set; generating a plurality of sub-models corresponding to the feature matrix set according to the volatility weighted data set; A plurality of the sub-models are integrated into the volatility-weighted LASSO model.

[0007] Optionally, in a second implementation of the first aspect of the present invention, the step of generating a plurality of sub-models corresponding to the feature matrix set according to the volatility weighted data set includes: According to the volatility weighted data set, a sub-model corresponding to each feature matrix in the feature matrix set is generated to obtain a plurality of sub-models corresponding to the feature matrix set.

[0008] Optionally, in a third implementation manner of the first aspect of the present invention, the step of generating, according to the volatility weighted data set, a sub-model corresponding to each feature matrix in the feature matrix set comprises: According to each volatility-weighted data in the volatility-weighted data set, a pre-trained lasso prediction model is adjusted to generate a sub-model corresponding to each feature matrix in the feature matrix set.

[0009] Optionally, in a fourth implementation of the first aspect of the present invention, before the step of collecting photovoltaic power generation data to obtain a characteristic data set, the method further includes: Train the initial PSO-SVM prediction model; The particle swarm optimization (PSO) algorithm is called to optimize the initial PSO-SVM prediction model to obtain the PSO-SVM prediction model.

[0010] Optionally, in a fifth implementation of the first aspect of the present invention, the step of training a volatility-weighted LASSO model in real time according to the feature matrix set, and calling the volatility-weighted LASSO model to predict a future prediction value set corresponding to the feature matrix set includes: The volatility weighted LASSO model is called to perform a future value prediction operation on each feature data in the feature matrix set to obtain the future prediction value set.

[0011] Optionally, in a sixth implementation of the first aspect of the present invention, the step of collecting photovoltaic power generation data to obtain a characteristic data set includes: Collecting photovoltaic power generation data, performing data cleaning on the photovoltaic power generation data, and obtaining target photovoltaic power generation data; Perform feature extraction on the target photovoltaic power generation data to obtain the feature data set.

[0012] Optionally, in a seventh implementation manner of the first aspect of the present invention, when the photovoltaic power generation data is obtained, before the step of generating a feature data set according to the photovoltaic power generation data, the method further includes: Collect temperature data, air pressure data, humidity data, wind speed data, surface temperature data, and visibility data for a preset time period to obtain photovoltaic power generation data.

[0013] A second aspect of the present invention provides a photovoltaic power generation prediction device, comprising: a memory and at least one processor, wherein instructions are stored in the memory, and the memory and the at least one processor are interconnected via lines; the at least one processor calls the instructions in the memory so that the photovoltaic power generation prediction device executes the above-mentioned photovoltaic power generation prediction method.

[0014] A third aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned photovoltaic power generation prediction method.

[0015] In an embodiment of the present invention, when photovoltaic power generation data is obtained, a feature data set is generated according to the photovoltaic power generation data; feature reconstruction is performed on the feature data set to obtain a feature matrix set; according to the feature matrix set, a volatility weighted LASSO model is trained in real time, and the volatility weighted LASSO model is called to predict the future prediction value set corresponding to the feature matrix set; the pre-trained PSO-SVM prediction model is called, and the power generation prediction operation is performed according to the future prediction value set to obtain the power generation prediction value. The introduction of volatility weighting technology helps to better consider the time series changes and volatility of data during the model training process. It solves the problem that the LASSO model has limited ability to deal with long-term data and data volatility alone. This improves the accuracy of photovoltaic power generation prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of an embodiment of a method for predicting photovoltaic power generation in an embodiment of the present invention; Figure 2 A schematic diagram of a specific embodiment of step 101 of the photovoltaic power generation prediction method in an embodiment of the present invention; Figure 3 A schematic diagram of a specific embodiment of step 103 of the photovoltaic power generation prediction method in an embodiment of the present invention; Figure 4 This is another specific embodiment schematic diagram of step 103 of the photovoltaic power generation prediction method in an embodiment of the present invention; Figure 5 Schematic diagram of an embodiment of a photovoltaic power generation prediction device in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] The embodiments of the present invention provide a photovoltaic power generation prediction method, device and storage medium.

[0018] The embodiments disclosed in the present invention will be described in more detail below with reference to the accompanying drawings. Although certain embodiments disclosed in the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments disclosed in the present invention are only for exemplary purposes and are not intended to limit the scope of protection disclosed in the present invention.

[0019] In the description of the embodiments disclosed in the present invention, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0020] The professional terms involved in the prediction method of photovoltaic power generation include but are not limited to: Feature data set: A set of key information extracted from raw photovoltaic power generation data for model training and prediction. It has a direct or indirect relationship with power generation.

[0021] Feature reconstruction: transform or process the original feature data set to generate a feature matrix set that is more suitable for model training or prediction.

[0022] Volatility-weighted LASSO model: A forecasting model that combines LASSO (Least Absolute Shrinkage and Selection Operator) regression with volatility weighting. LASSO regression is used to prevent overfitting and automatically perform feature selection, while volatility weighting may be used to adjust model weights based on the volatility of the data.

[0023] PSO-SVM prediction model: A prediction model that combines the Particle Swarm Optimization (PSO) algorithm and the Support Vector Machine (SVM). The PSO algorithm is used to optimize the parameters of the SVM to improve the prediction accuracy.

[0024] Particle Swarm Optimization (PSO) algorithm: An optimization algorithm based on swarm intelligence that simulates the behavior of bird flocks foraging. During the optimization process, each particle represents a potential solution, and the optimal solution is found by iteratively updating the position and velocity of the particle.

[0025] Support Vector Machine (SVM): A supervised learning method used for classification and regression analysis. SVM achieves classification or regression by finding a hyperplane that maximizes the margin between different classes.

[0026] Data cleaning: During the data preprocessing stage, data is checked, corrected, and deleted to remove noise, correct errors, and handle missing values, thereby improving data quality.

[0027] Feature extraction: Extracting feature information useful for the prediction target from the raw data. In the photovoltaic power generation prediction, this includes but is not limited to weather conditions, time information, historical power generation, etc.

[0028] Photovoltaic power generation data: various data related to photovoltaic power generation, such as power generation, weather conditions (temperature, air pressure, humidity, wind speed, etc.), geographic location information, etc.

[0029] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , an embodiment of the photovoltaic power generation prediction method in the embodiment of the present invention includes: 101. When photovoltaic power generation data is obtained, a feature data set is generated according to the photovoltaic power generation data; Specifically, there are many sources that affect photovoltaic power generation, including human factors (neglecting cleaning, unreasonable design, etc.), equipment factors (photovoltaic installation method, intersection line loss, inverter efficiency, etc.), and environmental meteorological factors (temperature, weather, air pressure, precipitation, etc.). The values ​​of six meteorological factors, namely temperature, air pressure, humidity, wind speed, surface temperature, and visibility, were collected from 8:00 to 18:00 every day at the hour, and these values ​​were then used to form a 66-dimensional feature as the input of the combined model. The formula corresponding to the weather sample set D is:

[0030] in, is a 66-dimensional weather feature vector, is the number of samples.

[0031] The output is the photovoltaic daily power generation of the area corresponding to the feature, and the corresponding mathematical formula is as follows:

[0032] in, For sample The corresponding target value is a one-dimensional vector; m is the number of samples.

[0033] Relevant data of photovoltaic power stations, such as power generation, light intensity, temperature, etc., are collected in real time or regularly through sensors or other data sources.

[0034] Preprocess the collected data to remove invalid values, abnormal values, duplicate values, etc. to ensure the accuracy and consistency of the data.

[0035] Optionally, air temperature data, air pressure data, humidity data, wind speed data, surface temperature data, and visibility data are collected during a preset time period to obtain photovoltaic power generation data.

[0036] Features useful for power generation prediction, such as historical power generation, light intensity, temperature, weather type, etc., are extracted from the cleaned data to form a feature data set.

[0037] Specifically, refer to Figure 2 , Figure 2 This is a schematic diagram of a specific embodiment of step 101 of the method for predicting photovoltaic power generation in an embodiment of the present invention. Step 101 also includes the following specific implementation methods: 1011. Collect photovoltaic power generation data, and perform data cleaning on the photovoltaic power generation data to obtain target photovoltaic power generation data; 1012. Perform feature extraction on the target photovoltaic power generation data to obtain the feature data set.

[0038] Specifically, due to the complexity and diversity of the acquired data, it is not possible to completely ensure that the data acquired by the weather station fully meets the modeling conditions, so the acquired data needs to be cleaned. The data cleaning rules are as follows: ①Fill or discard the obtained empty values.

[0039] ② De-duplicate the obtained duplicate data.

[0040] ③Clean the data with incorrect range.

[0041] ④ Carry out necessary verification on data that does not conform to the norm to ensure the reliability of the data.

[0042] In steps 1011-1012, after data cleaning, feature extraction is performed on the target photovoltaic power generation data, from which key information useful for photovoltaic power generation prediction can be extracted, including time, weather conditions, geographical location, photovoltaic panel performance and other factors, thereby improving the generalization ability and prediction accuracy of the prediction model.

[0043] 102. Perform feature reconstruction on the feature data set to obtain a feature matrix set; Specifically, each feature in the feature data set is standardized to eliminate the dimensional differences between different features. According to the prediction requirements, the features are transformed, such as using polynomial features, interaction terms, etc., to increase the complexity and prediction ability of the model. The processed feature data is organized into a feature matrix set in chronological order or other logical order, and each feature matrix corresponds to data within a period of time.

[0044] 103. According to the feature matrix set, a volatility-weighted LASSO model is trained in real time, and the volatility-weighted LASSO model is called to predict a future prediction value set corresponding to the feature matrix set; Specifically, the volatility of each feature matrix in the feature matrix set is analyzed, and different weights are assigned to each feature matrix according to the volatility. For each weighted feature matrix, a sub-model is trained using the LASSO (Least Absolute Shrinkage and Selection Operator) algorithm, and the complexity and overfitting of the model are controlled by adjusting the penalty term parameters of the LASSO. All sub-models are integrated into a volatility-weighted LASSO model, and weighted average or voting is performed according to the weights of each sub-model to improve the overall prediction performance.

[0045] Optionally, the volatility-weighted LASSO model is called to perform a future value prediction operation on each feature data in the feature matrix set to obtain the future prediction value set. By calling the volatility-weighted LASSO model to perform a future value prediction operation on each feature data in the feature matrix set, a fine-grained prediction of each feature data can be achieved. Minor changes in the data can be captured, thereby improving the accuracy of the overall prediction.

[0046] Optional, Figure 3 This is another specific embodiment schematic diagram of step 103 of the photovoltaic power generation prediction method in an embodiment of the present invention. Step 103 also includes the following specific implementation methods: 1031. Calculate the volatility weighted data corresponding to each feature matrix in the feature matrix set to obtain a volatility weighted data set; 1032. Generate a plurality of sub-models corresponding to the feature matrix set according to the volatility weighted data set; Optionally, a sub-model corresponding to each feature matrix in the feature matrix set is generated according to the volatility weighted data set, so as to obtain a plurality of sub-models corresponding to the feature matrix set. By respectively establishing a sub-model for each feature matrix, it is possible to perform special optimization for the specific attributes and volatility of each feature matrix, thereby improving the prediction accuracy of the model on the feature matrix.

[0047] Optionally, according to each volatility-weighted data in the volatility-weighted data set, a pre-trained lasso prediction model is adjusted to generate a sub-model corresponding to each feature matrix in the feature matrix set. By adjusting the lasso prediction model for each volatility-weighted data, a sub-model that better conforms to the characteristics of the data can be generated, thereby improving the prediction accuracy of the model on the specific data.

[0048] 1033. Integrate the multiple sub-models into the volatility-weighted LASSO model.

[0049] Specifically, based on the lasso method, volatility weighting is incorporated into the lasso method as the weight of the model, that is, volatility is used as a component of the penalty term to adjust the degree of shrinkage of the coefficients in the model. The principle is that when the data volatility is high, the penalty for the coefficient should be stricter, while when the volatility is low, it can be relatively loose. This can ensure that the model is more robust in the face of data fluctuations, while maintaining sensitivity to the data when volatility is low. The specific steps are as follows.

[0050] Step 1: Estimating the volatility of each feature of the matrix. This embodiment uses the GARCH (1,1) model to estimate the volatility of each independent variable of the matrix. The implementation process is as follows: Step 1: Model Setting. Volatility The calculation formula is as follows:

[0051] in, is a constant term, representing the base level of volatility; and are model parameters that measure past volatility and the impact of past volatility on current volatility, respectively; is the past error term; is the conditional variance, which represents the volatility at time t; is the mean; is the observed value of the time series at time t; are independent and identically distributed random error terms, usually assumed to follow a standard normal distribution.

[0052] Step 2: Parameter estimation. This embodiment uses the gradient descent method to estimate , and .

[0053] Step 3: Volatility prediction. Use the parameters obtained in step 2 to make predictions in the following formula:

[0054] Step 4: Loop through steps 1 to 3 to obtain the volatility corresponding to each independent variable in the matrix.

[0055] Step 2: Construct the penalty term of the lasso algorithm. It is added to the penalty term of the lasso algorithm in the form of a reciprocal, as follows:

[0056]

[0057] in, is the observed value; is the eigenvector, β is the coefficient vector, is the regularization parameter, p is the number of features, and m is the number of samples.

[0058] Step 4: Solve the objective function. Use the coordinate descent method to solve the function in step 3 and get the adjusted coefficient β. Then build a volatility weighted lasso model. And predict the future value pre.

[0059] Step 5: For "Feature Sequence Reconstruction" The matrix contained in the collection is looped from step 1 to step 4 to obtain multiple lasso models, and the prediction results of each model are obtained.

[0060] The form of multiple lasso models (multiple sub-models) is as follows:

[0061] in, is a collection of volatility-weighted lasso models; To reconstruct the matrix The volatility-weighted lasso model constructed.

[0062] The future prediction values ​​obtained by each model are in the following form: in, A collection of prediction results for each volatility-weighted lasso model; For Model The predicted result vector.

[0063] In steps 1031-1033, by calculating the volatility weighted data for each feature matrix in the feature matrix set and generating multiple sub-models accordingly, the dynamic changes in the photovoltaic power generation data can be better captured. This enables the model to more accurately reflect the volatility characteristics in the data, thereby improving the model's adaptability to the prediction of photovoltaic power generation at different times and under different conditions.

[0064] Furthermore, in steps 1031-1033, multiple sub-models are integrated into a volatility-weighted LASSO model, which actually utilizes the idea of ​​ensemble learning. Ensemble learning usually achieves better generalization performance than a single learner by combining the prediction results of multiple learners, thereby improving the overall prediction accuracy.

[0065] Optional, Figure 4This is another specific embodiment schematic diagram of step 103 of the photovoltaic power generation prediction method in an embodiment of the present invention. Step 103 also includes the following specific implementation methods: 1034. Calculate the volatility weighted data corresponding to each feature matrix in the feature matrix set to obtain a volatility weighted data set; 1035. Generate a plurality of sub-models corresponding to the feature matrix set according to the volatility weighted data set; 1036. Integrate the plurality of sub-models into the volatility-weighted LASSO model.

[0066] In steps 1034-1036, by calculating the volatility weighted data of each feature matrix in the feature matrix set, the volatility characteristics of photovoltaic power generation data in different time periods or under different conditions can be more accurately reflected. The weighted processing enables the model to pay more attention to those data points that have a greater impact on the prediction results during training, thereby improving the accuracy of the prediction. The concept of "real-time training" is emphasized, which means that the model can dynamically adjust its own parameters and structure according to the latest photovoltaic power generation data to quickly adapt to the real-time changes of the photovoltaic power generation system.

[0067] 104. Calling a pre-trained PSO-SVM prediction model, performing a power generation prediction operation according to the future prediction value set, and obtaining a power generation prediction value.

[0068] Specifically, the feature matrix is ​​input, and the latest feature matrix is ​​input into the trained volatility weighted LASSO model. The model outputs the corresponding future prediction value set according to the input feature matrix, and these prediction value sets represent the power generation prediction results in the future period of time.

[0069] Specifically, a pre-trained PSO-SVM (Particle Swarm Optimization-Support Vector Machine) prediction model is loaded. The future prediction value set output by the volatility-weighted LASSO model is used as the input of the PSO-SVM model. The PSO-SVM model further performs power generation prediction based on the input future prediction value set to obtain the final power generation prediction value.

[0070] The scikit-learn library is used to implement SVM modeling. The kernel function is the Gaussian kernel function (RBF). The PSO algorithm is used to optimize the penalty coefficient c and parameter g of the RBF kernel function in the SVM model to find the optimal c and g. Then, the model trained under the optimal parameters is used to predict the power generation. The prediction result is recorded as .

[0071] Specifically, before step 101, the following specific implementation methods are also included: 201. Train the initial PSO-SVM prediction model; 202. Calling a particle swarm optimization (PSO) algorithm to optimize the initial PSO-SVM prediction model to obtain the PSO-SVM prediction model.

[0072] Specifically, the PSO algorithm is used to optimize the parameters in the RBF kernel function. The calculation steps are: ①Initialize c, g values ​​and change rates.

[0073] c. The formula for initializing the g value is as follows: Where: : the maximum value of c; : minimum value of c; : the maximum value of g; : minimum value of g; i: number of particles; r: random number, range (0, 1).

[0074] c.g change rate initialization:

[0075] ② Calculation of initial fitness value: The RMSE of the cross-validation method is selected as the fitness function, and the calculation formula is:

[0076] In the formula, is the true value, is the predicted value, and n is the number of predicted samples.

[0077] ③ Initialize the optimal values ​​of c, g and the global optimal value: The formula for calculating the optimal initial value is:

[0078] Search of .

[0079] ④Update value and rate of change: The calculation formula for updating the c and g values ​​in the kth iteration is:

[0080] The calculation formula for the change rate of c and g in the kth iteration is:

[0081] In the formula, is the inertia weight of iteration k times, K is the maximum number of iterations, and k is the current number of iterations.

[0082] At the same time, in order to prevent the results from being out of range, the restrictions are as follows:

[0083] In the formula, is the maximum value of the change value; is the minimum value of the change.

[0084] ⑤Update the optimal value and the global optimal value: The calculation formula for the updated optimal value and the global optimal value is:

[0085] ⑥ Determine that the termination conditions are met; When the fitness value reaches the requirement or the number of iterations is reached, the iteration ends and the result is output; otherwise, it returns to execution ③.

[0086] In an embodiment of the present invention, when photovoltaic power generation data is obtained, a feature data set is generated according to the photovoltaic power generation data; feature reconstruction is performed on the feature data set to obtain a feature matrix set; according to the feature matrix set, a volatility weighted LASSO model is trained in real time, and the volatility weighted LASSO model is called to predict the future prediction value set corresponding to the feature matrix set; the pre-trained PSO-SVM prediction model is called, and the power generation prediction operation is performed according to the future prediction value set to obtain the power generation prediction value. The introduction of volatility weighting technology helps to better consider the time series changes and volatility of data during the model training process. It solves the problem that the LASSO model has limited ability to deal with long-term data and data volatility alone. This improves the accuracy of photovoltaic power generation prediction.

[0087] Figure 5 : is a schematic diagram of the structure of a photovoltaic power generation prediction device provided by an embodiment of the present invention. The photovoltaic power generation prediction device 500 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 510 (for example, one or more processors) and a memory 520, and one or more storage media 530 (for example, one or more mass storage devices) storing application programs 533 or data 532. Among them, the memory 520 and the storage medium 530 can be short-term storage or permanent storage. The program stored in the storage medium 530 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the photovoltaic power generation prediction device 500. Furthermore, the processor 510 can be configured to communicate with the storage medium 530 to execute a series of instruction operations in the storage medium 530 on the photovoltaic power generation prediction device 500.

[0088] The photovoltaic power generation prediction device 500 may also include one or more power supplies 540, one or more wired or wireless network interfaces 550, one or more input and output interfaces 560, and / or one or more operating systems 531, such as Windows Serve, Mac OS X, Unix, Linux, Free BSD, etc. It will be understood by those skilled in the art that Figure 5 The structure of the photovoltaic power generation prediction device shown does not constitute a limitation on the photovoltaic power generation prediction device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0089] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the photovoltaic power generation prediction method.

[0090] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0091] In addition, although each operation is described in a specific order, this should be understood as requiring such operation to be performed in the specific order shown or in a sequential order, or requiring that all illustrated operations should be performed to obtain desired results. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present disclosure. Some features described in the context of a separate embodiment can also be implemented in a single implementation in combination. On the contrary, the various features described in the context of a single implementation can also be implemented in multiple implementations individually or in any suitable sub-combination mode.

[0092] Although the subject matter has been described in language specific to structural features and / or methodological logical actions, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing the claims.

Claims

1. A method for predicting photovoltaic power generation, characterized in that: The photovoltaic power generation prediction method comprises: When the photovoltaic power generation data is obtained, a feature data set is generated according to the photovoltaic power generation data; Performing feature reconstruction on the feature data set to obtain a feature matrix set; According to the feature matrix set, a volatility-weighted LASSO model is trained in real time, and the volatility-weighted LASSO model is called to predict a future prediction value set corresponding to the feature matrix set; The pre-trained PSO-SVM prediction model is called, and a power generation prediction operation is performed according to the future prediction value set to obtain a power generation prediction value.

2. The photovoltaic power generation prediction method according to claim 1, characterized in that: The step of training the volatility weighted LASSO model in real time according to the feature matrix set comprises: Calculating the volatility weighted data corresponding to each feature matrix in the feature matrix set to obtain a volatility weighted data set; Generating a plurality of sub-models corresponding to the feature matrix set according to the volatility weighted data set; A plurality of the sub-models are integrated into the volatility-weighted LASSO model.

3. The photovoltaic power generation prediction method according to claim 2, characterized in that: The step of generating a plurality of sub-models corresponding to the feature matrix set according to the volatility weighted data set comprises: According to the volatility weighted data set, a sub-model corresponding to each feature matrix in the feature matrix set is generated to obtain a plurality of sub-models corresponding to the feature matrix set.

4. The photovoltaic power generation prediction method according to claim 3, characterized in that: The step of generating a sub-model corresponding to each feature matrix in the feature matrix set according to the volatility weighted data set comprises: According to each volatility-weighted data in the volatility-weighted data set, a pre-trained lasso prediction model is adjusted to generate a sub-model corresponding to each feature matrix in the feature matrix set.

5. The photovoltaic power generation prediction method according to claim 1, characterized in that: Before the step of collecting photovoltaic power generation data to obtain a characteristic data set, the method further includes: Train the initial PSO-SVM prediction model; The particle swarm optimization (PSO) algorithm is called to optimize the initial PSO-SVM prediction model to obtain the PSO-SVM prediction model.

6. The photovoltaic power generation prediction method according to claim 1, characterized in that: The step of predicting the future prediction value set corresponding to the feature matrix set according to calling the volatility weighted LASSO model comprises: The volatility weighted LASSO model is called to perform a future value prediction operation on each feature data in the feature matrix set to obtain the future prediction value set.

7. The photovoltaic power generation prediction method according to claim 1, characterized in that: The step of collecting photovoltaic power generation data to obtain a characteristic data set includes: Collecting photovoltaic power generation data, performing data cleaning on the photovoltaic power generation data, and obtaining target photovoltaic power generation data; Perform feature extraction on the target photovoltaic power generation data to obtain the feature data set.

8. The photovoltaic power generation prediction method according to claim 1, characterized in that: When the photovoltaic power generation data is obtained, before the step of generating a feature data set according to the photovoltaic power generation data, the method further includes: Collect temperature data, air pressure data, humidity data, wind speed data, surface temperature data, and visibility data for a preset time period to obtain photovoltaic power generation data.

9. A photovoltaic power generation prediction device, characterized in that: The photovoltaic power generation prediction device comprises: a memory and at least one processor, the memory stores instructions, and the memory and the at least one processor are interconnected via a line; The at least one processor calls the instructions in the memory to enable the photovoltaic power generation prediction device to execute the photovoltaic power generation prediction method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting photovoltaic power generation according to any one of claims 1 to 8 is implemented.