A photovoltaic power generation power prediction method and system based on improved FVIM-XGBoost-A

Abstract

The invention discloses a photovoltaic power prediction method and system based on improved FVIM-XGBoost-A, the method comprising: performing missing value processing on photovoltaic environment data according to a preset ARIMA-X model, normalizing and smoothing the data, adding a disturbance mechanism to the FVIM algorithm, further avoiding local optimum, using the global optimal solution output by the improved FVIM algorithm, obtaining the optimal hyperparameter combination, and constructing the tree structure of XGBoost; proposing an ARSV-Loss loss function in the XGBoost model and using an adaptive multi-level regularization term, so that the XGBoost model can be adjusted more complexly in the search space, which is helpful to jump out of the local optimal solution and avoid overfitting, integrating the improved FVIM algorithm and the XGBoost model and adding an attention mechanism to form an improved FVIM-XGBoost-A model. The FVIM-XGBoost-A model can improve the accuracy of photovoltaic power prediction.

Description

Technical Field

[0001] The present invention belongs to the technical field of renewable energy power generation, and in particular relates to a photovoltaic power generation power prediction method and system based on improved FVIM-XGBoost-A. Background Art

[0002] At present, the photovoltaic power generation industry is rapidly rising, and the installed capacity of photovoltaic power generation continues to grow. However, the problem that comes with it is the volatility and uncertainty of photovoltaic power generation. Photovoltaic power generation is affected by many factors. These factors may include changes in light intensity, temperature fluctuations, shadows, and aging of equipment. The main influencing factors include the following aspects: First, changes in weather conditions have a significant impact on photovoltaic power generation. For example, cloudy days, rainy days, or extreme weather will greatly reduce the power generation of photovoltaic systems; secondly, the aging and pollution of photovoltaic modules (such as dust, bird droppings, etc.) will lead to reduced power generation efficiency, thereby increasing the uncertainty of power prediction; in addition, the differences in light conditions in different regions also make photovoltaic power prediction more regional and complex; finally, insufficient monitoring and maintenance may cause system failures, thereby affecting power generation performance. Therefore, it is necessary to accurately predict photovoltaic power generation in order to improve power generation efficiency and balance the supply and demand of the power grid.

[0003] However, traditional forecasting methods are usually unable to cope with the complex volatility and nonlinear characteristics caused by meteorological factors and equipment conditions in the photovoltaic power generation process, and the forecasting accuracy is limited. Summary of the invention

[0004] The present invention provides a photovoltaic power generation power prediction method and system based on improved FVIM-XGBoost-A, which are used to solve the technical problem of limited prediction accuracy caused by complex volatility and nonlinear characteristics caused by meteorological factors and equipment conditions in the photovoltaic power generation process.

[0005] In a first aspect, the present invention provides a photovoltaic power generation prediction method based on an improved FVIM-XGBoost-A, comprising:

[0006] Acquire photovoltaic power generation data and photovoltaic environment data, and preprocess the photovoltaic environment data according to a preset ARIMA-X model to obtain target photovoltaic environment data;

[0007] According to the preset perturbation mechanism, the FVIM algorithm is globally optimized to obtain the global optimal solution. , and obtain the improved FVIM algorithm;

[0008] Construct an XGBoost model, integrate the improved FVIM algorithm and the XGBoost model, and add an attention mechanism to obtain an improved FVIM-XGBoost-A model;

[0009] Inputting the photovoltaic power generation data and the target photovoltaic environment data into the improved FVIM-XGBoost-A model, iteratively training the improved FVIM-XGBoost-A model, and obtaining a photovoltaic power generation power prediction model;

[0010] The acquired real-time photovoltaic environment data is input into the photovoltaic power generation prediction model, and the photovoltaic power generation prediction model outputs the photovoltaic power generation prediction result.

[0011] In a second aspect, the present invention provides a photovoltaic power generation prediction system based on an improved FVIM-XGBoost-A, comprising:

[0012] An acquisition module is configured to acquire photovoltaic power generation data and photovoltaic environment data, and preprocess the photovoltaic environment data according to a preset ARIMA-X model to obtain target photovoltaic environment data;

[0013] The optimization module is configured to perform global optimization on the FVIM algorithm according to the preset perturbation mechanism to obtain the global optimal solution. , and obtain the improved FVIM algorithm;

[0014] A construction module is configured to construct an XGBoost model, integrate the improved FVIM algorithm and the XGBoost model, and add an attention mechanism to obtain an improved FVIM-XGBoost-A model;

[0015] A training module, configured to input the photovoltaic power generation data and the target photovoltaic environment data into the improved FVIM-XGBoost-A model, iteratively train the improved FVIM-XGBoost-A model, and obtain a photovoltaic power generation power prediction model;

[0016] The output module is configured to input the acquired real-time photovoltaic environment data into the photovoltaic power generation prediction model, and the photovoltaic power generation prediction model outputs the photovoltaic power generation prediction result.

[0017] In a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the photovoltaic power generation prediction method based on the improved FVIM-XGBoost-A of any embodiment of the present invention.

[0018] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon. When the program instructions are executed by a processor, the processor executes the steps of the photovoltaic power generation prediction method based on the improved FVIM-XGBoost-A of any embodiment of the present invention.

[0019] The photovoltaic power generation prediction method and system based on the improved FVIM-XGBoost-A of this application processes the missing values ​​of photovoltaic environmental data according to the preset ARIMA-X model, normalizes and smoothes the data, and adds the disturbance mechanism to the FVIM algorithm to further avoid local optimality and use the global optimal solution output by the improved FVIM algorithm. After obtaining the optimal hyperparameter combination, the tree structure of XGBoost is constructed; the ARSV-Loss loss function is proposed in the XGBoost model and the adaptive multi-level regularization term is used, so that the XGBoost model can make more complex adjustments in the search space, which helps to jump out of the local optimal solution and avoid overfitting. The improved FVIM algorithm and the XGBoost model are integrated and the attention mechanism is added to form an improved FVIM-XGBoost-A model, which can make a more accurate prediction of photovoltaic power generation, thereby improving the accuracy of photovoltaic power generation prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 A flowchart of a photovoltaic power generation prediction method based on improved FVIM-XGBoost-A provided in one embodiment of the present invention;

[0022] Figure 2 A structural block diagram of a photovoltaic power generation prediction system based on improved FVIM-XGBoost-A provided by one embodiment of the present invention;

[0023] Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] See also Figure 1 , which shows a flow chart of a photovoltaic power prediction method based on improved FVIM-XGBoost-A of the present application.

[0026] like Figure 1 As shown, the photovoltaic power generation prediction method based on the improved FVIM-XGBoost-A specifically includes the following steps:

[0027] Step S101 , obtaining photovoltaic power generation data and photovoltaic environment data, and preprocessing the photovoltaic environment data according to a preset ARIMA-X model to obtain target photovoltaic environment data.

[0028] In this step, the photovoltaic environment data is processed for missing values ​​according to the preset ARIMA-X model, and the expression is:

[0029] ,

[0030] ,

[0031] ,

[0032] In the formula, For the The sequence value of a time point, is the autoregressive part, is the order of the autoregressive term, is the order of the moving average term, is the coefficient of each lag term, For the The time series value at each time point, is the coefficient of the error term, For the The residual at each time point, is the moving average part, is the exogenous variable part, is the number of exogenous variables, is the lag order, is the coefficient of the exogenous variable, For the The exogenous variable The value at a time point, is the interaction term, is the number of interaction terms, is the coefficient of the interaction term, is the value of the exogenous variable, is the lagged value of the time series, is a nonlinear term, is the number of exogenous variables that introduce nonlinear terms, is the coefficient of the linear term, For the exogenous variable, is the dynamic error term, is the random error of the standard normal distribution, is the conditional variance at the current time point, is a constant term, is the lagged square coefficient of the error term, is the lagged coefficient of the past conditional variance, is the contribution of the square of the error term at the previous moment to the conditional variance at the current moment, is the contribution of the conditional variance of the previous moment to the current conditional variance, For the The conditional variance at each time point.

[0033] Due to the huge amount of photovoltaic data and the complexity of the data, the data weight coefficient is added when the data is normalized. Different weights are assigned according to the importance of the features. The changes in other time steps and the mutual influence between global features are also considered. The expression is:

[0034] ,

[0035] In the formula, is the normalized data value, is the dynamic weight, is the original value of the current data point, For time The mean value of the data at time is the nonlinear adjustment factor, is the mixing adjustment factor, is the number of historical time steps, is the weight of the data point in the historical time step, For time The original value of the moment, For time The mean value of the data at time is the maximum value of the current feature, is the minimum value of the current feature, To control the nonlinear adjustment factor of the maximum and minimum values, is the regulating factor, For time The maximum value at a moment, For time The minimum value of the moment;

[0036] Photovoltaic power generation data may sometimes experience sudden fluctuations due to external factors such as the environment and equipment. Data smoothing can better extract trend information and reduce the impact of short-term noise, thereby improving the prediction accuracy of the model. The introduction of multi-scale weights allows the smoothing process to consider changes in different time scales at the same time. The expression is:

[0037] ,

[0038] In the formula, For time The multi-scale weighted moving average of time, is the time scale quantity, For the Time point The weight of the scale, For time The original data value at the moment, For time The original data value at the moment, For the The nonlinear adjustment factor of the time scale is For the The weighted nonlinear exponent of the time scale, For the The nonlinear exponent of the time scale, is the size of the sliding window.

[0039] Step S102: perform global optimization on the FVIM algorithm according to the preset perturbation mechanism to obtain the global optimal solution. , and the improved FVIM algorithm is obtained.

[0040] In this step, the first vector is randomly initialized , the second vector , the third vector and the fourth vector , set the initial parameter value: the initial value of the parameter that controls the update amplitude of the vector , Bayesian optimization guide parameter initial value , initial weight value , initial noise intensity and chaotic variables , initialize the Bayesian optimized proxy model and train the proxy model with the initial vector;

[0041] For each vector , using the Bayesian optimization model to calculate the expected improvement of each vector , the expression is:

[0042] ,

[0043] ,

[0044] In the formula, is the mean prediction of the Bayesian optimization model for the current vector position, is the current global optimal value, is the cumulative distribution function of the standard normal distribution, is the variance prediction of the current vector position by the Bayesian optimization model, is the probability density function, Standardize variables for normal distribution;

[0045] According to the current number of iterations, the parameters in the FVIM algorithm are dynamically adjusted. The expression is:

[0046] ,

[0047] In the formula, For iteration The control vector update amplitude parameter at this time, is the initial value of the parameter that controls the update amplitude of the vector, is the maximum number of iterations, is the number of iterations, For iteration The Bayesian optimization bootstrap parameters at this time are Bootstrap parameter initial values ​​for Bayesian optimization, For iteration The weight value at this time, is the initial weight value, For iteration The noise intensity at that time, is the initial noise intensity;

[0048] For the first vector , the second vector , the third vector and the fourth vector To update the position, the expression is:

[0049] ,

[0050] ,

[0051] ,

[0052] ,

[0053] In the formula, , , , The first vector , the second vector , the third vector and the fourth vector Updated location, , , , are all random numbers between [0,1]. For the The local optimal solution of vectors, , , , are the expected improvements for each vector calculated using the Bayesian optimization model, , , , are the chaotic variables corresponding to each vector, , , , are the noise terms of each vector at different time steps, For the The average position of individuals;

[0054] Repeat the iteration until the termination condition is met and output the current global optimal solution .

[0055] Step S103, constructing an XGBoost model, integrating the improved FVIM algorithm and the XGBoost model, and adding an attention mechanism to obtain an improved FVIM-XGBoost-A model.

[0056] In this step, in order to improve the performance and generalization ability of the model, a new loss function (ARSV-Loss) is proposed and the regularization term is improved. The specific content is: Taking into account the complex and changeable weather, which leads to the existence of outliers in the data or significant fluctuations in the power generation data, in order to improve the accuracy of the algorithm, the ARSV-Loss loss function is proposed, and the adaptive multi-level regularization term is used to enable the model to make more complex adjustments in the search space, which helps to jump out of the local optimal solution and avoid overfitting. The specific formula is:

[0057] ,

[0058] ,

[0059] In the formula, For the data set The actual photovoltaic power generation value of samples, For the The photovoltaic power generation value of samples, For the The gradient of samples, For the The second-order derivative of samples;

[0060] ,

[0061] ,

[0062] In the formula, is the loss function, is the time step The actual value of are model parameters, is the time step The predicted value of is the length of the time series data, is the regularization term, is the threshold parameter, is the weight of the smoothing factor, is the smoothness parameter, For the model parameters, is the number of parameters, is the weight of the time series error term, is a parameter that controls the strength of the sum of squared weights, is a parameter that controls the strength of the absolute value and strength of the weights, is a parameter that controls the strength of the logarithmic regularization, is the control parameter of the high-order regularization term;

[0063] Perform a second-order Taylor expansion on the optimization objective function in the XGBoost model and update the leaf node weights. The expression is:

[0064] ,

[0065] ,

[0066] In the formula, For the The approximate form of the objective function in the round iteration is, For the Number of iterations For samples The output value of is the regularization term, To control the penalty intensity of the absolute value of the leaf node weight, To control the strength of the sum of squared weights of leaf nodes, is a high-order regularization parameter, For leaf nodes The index set of the sample set, For the The optimal weight of leaf nodes, is the total number of samples.

[0067] Output the final prediction model, the expression is:

[0068] ,

[0069] In the formula, For the The final predicted output value of samples, is the weight coefficient of each tree, is the total number of trees constructed, For the Tree pair samples The predicted output.

[0070] It should be noted that the attention mechanism is added when integrating the improved FVIM algorithm and the XGBoost model to enhance the ability of the improved FVIM-XGBoost-A model to focus on important features, improve the overall prediction accuracy and improve the performance of the FVIM-XGBoost-A model. Specifically:

[0071] Define the attention score calculation, the expression is:

[0072] ,

[0073] In the formula, For the The attention score of each feature, is the hyperbolic tangent function, , are all trainable weight matrices of the attention mechanism, For the eigenvalues, is the bias term, is the adaptive frequency parameter, is the control parameter of Gaussian noise, is the standard normal distribution, is the parameter that controls the intensity of chaotic disturbance. is the frequency control parameter;

[0074] Calculate the attention weight, the expression is:

[0075] ,

[0076] In the formula, For the The attention weight of each feature, For the The attention score of each feature, is a nonlinear activation function, is the total number of features, is an exponential function;

[0077] Calculate the weighted eigenvector, the expression is:

[0078] ,

[0079] In the formula, is the coefficient that controls the influence of the sine combination term, is the coefficient that controls the influence of the sine combination term, is the coefficient that controls the influence of the logarithmic combination term.

[0080] Step S104: input the photovoltaic power generation data and the target photovoltaic environment data into the improved FVIM-XGBoost-A model, iteratively train the improved FVIM-XGBoost-A model, and obtain a photovoltaic power generation power prediction model.

[0081] Step S105, inputting the acquired real-time photovoltaic environment data into the photovoltaic power generation prediction model, and the photovoltaic power generation prediction model outputs the photovoltaic power generation prediction result.

[0082] In summary, the method of this application processes the missing values ​​of photovoltaic environmental data according to the preset ARIMA-X model, normalizes and smoothes the data, adds the disturbance mechanism to the FVIM algorithm, further avoids local optimality, and uses the global optimal solution output by the improved FVIM algorithm. After obtaining the optimal hyperparameter combination, the tree structure of XGBoost is constructed; the ARSV-Loss loss function is proposed in the XGBoost model and the adaptive multi-level regularization term is used, so that the XGBoost model can make more complex adjustments in the search space, which helps to jump out of the local optimal solution and avoid overfitting. The improved FVIM algorithm and the XGBoost model are integrated and the attention mechanism is added to form an improved FVIM-XGBoost-A model, which can make a more accurate prediction of photovoltaic power generation, thereby improving the accuracy of photovoltaic power generation prediction.

[0083] See also Figure 2 , which shows a structural block diagram of a photovoltaic power generation prediction system based on improved FVIM-XGBoost-A of the present application.

[0084] like Figure 2 As shown, the photovoltaic power generation prediction system 200 includes an acquisition module 210 , an optimization module 220 , a construction module 230 , a training module 240 and an output module 250 .

[0085] The acquisition module 210 is configured to acquire photovoltaic power generation data and photovoltaic environment data, and pre-process the photovoltaic environment data according to the preset ARIMA-X model to obtain the target photovoltaic environment data; the optimization module 220 is configured to perform global optimization on the FVIM algorithm according to the preset disturbance mechanism to obtain the global optimal solution. , and obtain an improved FVIM algorithm; a construction module 230 is configured to construct an XGBoost model, integrate the improved FVIM algorithm and the XGBoost model, and add an attention mechanism to obtain an improved FVIM-XGBoost-A model; a training module 240 is configured to input the photovoltaic power generation data and the target photovoltaic environment data into the improved FVIM-XGBoost-A model, and iteratively train the improved FVIM-XGBoost-A model to obtain a photovoltaic power generation power prediction model; an output module 250 is configured to input the acquired real-time photovoltaic environment data into the photovoltaic power generation power prediction model, and the photovoltaic power generation power prediction model outputs a photovoltaic power generation power prediction result.

[0086] It should be understood that Figure 2 Modules and references documented in Figure 1 Therefore, the operations and features described above for the method and the corresponding technical effects are also applicable to Figure 2 The modules in it will not be described in detail here.

[0087] In some other embodiments, the embodiments of the present invention further provide a computer-readable storage medium having a computer program stored thereon, and when the program instructions are executed by a processor, the processor executes the photovoltaic power generation power prediction method based on the improved FVIM-XGBoost-A in any of the above method embodiments;

[0088] As an implementation mode, the computer-readable storage medium of the present invention stores computer-executable instructions, and the computer-executable instructions are configured as follows:

[0089] Acquire photovoltaic power generation data and photovoltaic environment data, and pre-process the photovoltaic environment data according to a preset ARIMA-X model to obtain target photovoltaic environment data;

[0090] According to the preset perturbation mechanism, the FVIM algorithm is globally optimized to obtain the global optimal solution. , and obtain the improved FVIM algorithm;

[0091] Construct an XGBoost model, integrate the improved FVIM algorithm and the XGBoost model, and add an attention mechanism to obtain an improved FVIM-XGBoost-A model;

[0092] Inputting the photovoltaic power generation data and the target photovoltaic environment data into the improved FVIM-XGBoost-A model, iteratively training the improved FVIM-XGBoost-A model, and obtaining a photovoltaic power generation power prediction model;

[0093] The acquired real-time photovoltaic environment data is input into the photovoltaic power generation prediction model, and the photovoltaic power generation prediction model outputs the photovoltaic power generation prediction result.

[0094] The computer-readable storage medium may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application required for at least one function; the storage data area may store data created based on the use of the photovoltaic power generation prediction system based on the improved FVIM-XGBoost-A, etc. In addition, the computer-readable storage medium may include a high-speed random access memory, and may also include a memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the photovoltaic power generation prediction system based on the improved FVIM-XGBoost-A through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0095] Figure 3 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, the device includes: a processor 310 and a memory 320. The electronic device may also include: an input device 330 and an output device 340. The processor 310, the memory 320, the input device 330 and the output device 340 may be connected via a bus or other means. Figure 3In the example, the bus connection is used. The memory 320 is the above-mentioned computer-readable storage medium. The processor 310 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 320, that is, the photovoltaic power generation prediction method based on the improved FVIM-XGBoost-A of the above-mentioned method embodiment is realized. The input device 330 can receive input digital or character information, and generate key signal input related to user settings and function control of the photovoltaic power generation prediction system based on the improved FVIM-XGBoost-A. The output device 340 may include a display device such as a display screen.

[0096] The electronic device can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.

[0097] As an implementation mode, the electronic device is applied to a photovoltaic power generation prediction system based on the improved FVIM-XGBoost-A, and is used for a client, and includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can:

[0098] Acquire photovoltaic power generation data and photovoltaic environment data, and preprocess the photovoltaic environment data according to a preset ARIMA-X model to obtain target photovoltaic environment data;

[0099] According to the preset perturbation mechanism, the FVIM algorithm is globally optimized to obtain the global optimal solution. , and obtain the improved FVIM algorithm;

[0100] Construct an XGBoost model, integrate the improved FVIM algorithm and the XGBoost model, and add an attention mechanism to obtain an improved FVIM-XGBoost-A model;

[0101] Inputting the photovoltaic power generation data and the target photovoltaic environment data into the improved FVIM-XGBoost-A model, iteratively training the improved FVIM-XGBoost-A model, and obtaining a photovoltaic power generation power prediction model;

[0102] The acquired real-time photovoltaic environment data is input into the photovoltaic power generation prediction model, and the photovoltaic power generation prediction model outputs the photovoltaic power generation prediction result.

[0103] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0104] Finally, 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 aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A photovoltaic power generation prediction method based on improved FVIM-XGBoost-A, characterized in that: include: Acquire photovoltaic power generation data and photovoltaic environment data, and pre-process the photovoltaic environment data according to a preset ARIMA-X model to obtain target photovoltaic environment data; According to the preset perturbation mechanism, the FVIM algorithm is globally optimized to obtain the global optimal solution. , and obtain the improved FVIM algorithm; Constructing an XGBoost model, integrating the improved FVIM algorithm and the XGBoost model, and adding an attention mechanism to obtain an improved FVIM-XGBoost-A model, wherein the integration of the improved FVIM algorithm and the XGBoost model, and adding an attention mechanism to obtain an improved FVIM-XGBoost-A model includes: Define the attention score calculation, the expression is: ， In the formula, For the The attention score of each feature, is the hyperbolic tangent function, , are all trainable weight matrices of the attention mechanism, For the eigenvalues, is the bias term, is the adaptive frequency parameter, is the control parameter of Gaussian noise, is the standard normal distribution, is the parameter that controls the intensity of chaotic disturbance. is the frequency control parameter; Calculate the attention weight, the expression is: ， In the formula, For the The attention weight of each feature, For the The attention score of each feature, is a nonlinear activation function, is the total number of features, is an exponential function; Calculate the weighted eigenvector, the expression is: ， In the formula, is the weighted eigenvector, is the coefficient that controls the influence of the sine combination term, is the coefficient for controlling the influence of logarithmic combination terms; Inputting the photovoltaic power generation data and the target photovoltaic environment data into the improved FVIM-XGBoost-A model, iteratively training the improved FVIM-XGBoost-A model, and obtaining a photovoltaic power generation power prediction model; The acquired real-time photovoltaic environment data is input into the photovoltaic power generation prediction model, and the photovoltaic power generation prediction model outputs the photovoltaic power generation prediction result.

2. The photovoltaic power generation prediction method based on improved FVIM-XGBoost-A according to claim 1 is characterized in that: The photovoltaic environment data is preprocessed according to the preset ARIMA-X model to obtain the target photovoltaic environment data, including: The photovoltaic environment data is processed for missing values ​​according to the preset ARIMA-X model, and the expression is: ， ， ， In the formula, For the The sequence value of a time point, is the autoregressive part, is the order of the autoregressive term, is the order of the moving average term, is the coefficient of each lag term, For the The time series value at each time point, is the coefficient of the error term, For the The residual at each time point, is the moving average part, is the exogenous variable part, is the number of exogenous variables, is the lag order, is the coefficient of the exogenous variable, For the The exogenous variable The value at a time point, is the interaction term, is the number of interaction terms, is the coefficient of the interaction term, is the value of the exogenous variable, is the lagged value of the time series, is a nonlinear term, is the number of exogenous variables that introduce nonlinear terms, is the coefficient of the linear term, For the exogenous variable, is the dynamic error term, is the random error of the standard normal distribution, is the conditional variance at the current time point, is a constant term, is the lagged square coefficient of the error term, is the lagged coefficient of the past conditional variance, is the contribution of the square of the error term at the previous moment to the conditional variance at the current moment, is the contribution of the conditional variance of the previous moment to the current conditional variance, For the The conditional variance at each time point.

3. The photovoltaic power prediction method based on improved FVIM-XGBoost-A according to claim 1 is characterized in that: The FVIM algorithm is globally optimized according to the preset perturbation mechanism to obtain the global optimal solution. include: Randomly initialize the first vector , the second vector , the third vector and the fourth vector , set the initial parameter value: the initial value of the parameter that controls the update amplitude of the vector , Bayesian optimization guide parameter initial value , initial weight value , initial noise intensity and chaotic variables , initialize the Bayesian optimized proxy model and train the proxy model with the initial vector; For each vector , using the Bayesian optimization model to calculate the expected improvement of each vector , the expression is: ， ， In the formula, is the mean prediction of the Bayesian optimization model for the current vector position, is the current global optimal value, is the cumulative distribution function of the standard normal distribution, is the variance prediction of the current vector position by the Bayesian optimization model, is the probability density function, Standardize variables for normal distribution; According to the current number of iterations, the parameters in the FVIM algorithm are dynamically adjusted. The expression is: ， In the formula, For iteration The control vector update amplitude parameter at this time, is the initial value of the parameter that controls the update amplitude of the vector, is the maximum number of iterations, is the number of iterations, For iteration The Bayesian optimization bootstrap parameters at this time are Bootstrap parameter initial values ​​for Bayesian optimization, For iteration The weight value at this time, is the initial weight value, For iteration The noise intensity at that time, is the initial noise intensity; For the first vector , the second vector , the third vector and the fourth vector To update the position, the expression is: ， ， ， ， In the formula, , , , The first vector , the second vector , the third vector and the fourth vector Updated location, , , , are all random numbers between [0,1]. is the local optimal solution of the vector, , , , are the expected improvements for each vector calculated using the Bayesian optimization model, , , , are the chaotic variables corresponding to each vector, , , , are the noise terms of each vector at different time steps, For the The average position of individuals; Repeat the iteration until the termination condition is met and output the current global optimal solution .

4. The photovoltaic power generation prediction method based on improved FVIM-XGBoost-A according to claim 1 is characterized in that: The XGBoost model includes the ARSV-Loss loss function, and the expression of the adaptive multi-level regularization term in the ARSV-Loss loss function is: ， ， In the formula, For the data set The actual photovoltaic power generation value of samples, For the The photovoltaic power generation value of samples, For the The gradient of samples, For the The second-order derivative of samples; ， ， In the formula, is the loss function, is the actual value of the time step, are model parameters, is the time step The predicted value of is the length of the time series data, is the regularization term, is the threshold parameter, is the weight of the smoothing factor, is the smoothness parameter, is the model parameter, is the number of parameters, is the weight of the time series error term, is a parameter that controls the strength of the sum of squared weights, is a parameter that controls the strength of the absolute value and strength of the weights, is a parameter that controls the strength of the logarithmic regularization, is the control parameter of the high-order regularization term; Perform a second-order Taylor expansion on the optimization objective function in the XGBoost model and update the leaf node weights. The expression is: ， ， In the formula, For the The approximate form of the objective function in the round iteration is, For the Number of iterations For samples The output value of is the regularization term, To control the penalty intensity of the absolute value of the leaf node weight, To control the strength of the sum of squared weights of leaf nodes, is a high-order regularization parameter, is the index set of the sample set of leaf nodes, For the The optimal weight of leaf nodes, is the total number of samples.

5. A photovoltaic power generation prediction system based on improved FVIM-XGBoost-A, characterized in that: include: An acquisition module is configured to acquire photovoltaic power generation data and photovoltaic environment data, and preprocess the photovoltaic environment data according to a preset ARIMA-X model to obtain target photovoltaic environment data; The optimization module is configured to perform global optimization on the FVIM algorithm according to the preset perturbation mechanism to obtain the global optimal solution. , and obtain the improved FVIM algorithm; A construction module is configured to construct an XGBoost model, integrate the improved FVIM algorithm and the XGBoost model, and add an attention mechanism to obtain an improved FVIM-XGBoost-A model, wherein the integration of the improved FVIM algorithm and the XGBoost model, and adding an attention mechanism to obtain an improved FVIM-XGBoost-A model includes: Define the attention score calculation, the expression is: ， In the formula, For the The attention score of each feature, is the hyperbolic tangent function, , are all trainable weight matrices of the attention mechanism, For the eigenvalues, is the bias term, is the adaptive frequency parameter, is the control parameter of Gaussian noise, is the standard normal distribution, is the parameter that controls the intensity of chaotic disturbance. is the frequency control parameter; Calculate the attention weight, the expression is: ， In the formula, For the The attention weight of each feature, For the The attention score of each feature, is a nonlinear activation function, is the total number of features, is an exponential function; Calculate the weighted eigenvector, the expression is: ， In the formula, is the weighted eigenvector, is the coefficient that controls the influence of the sine combination term, is the coefficient for controlling the influence of logarithmic combination terms; A training module, configured to input the photovoltaic power generation data and the target photovoltaic environment data into the improved FVIM-XGBoost-A model, iteratively train the improved FVIM-XGBoost-A model, and obtain a photovoltaic power generation power prediction model; The output module is configured to input the acquired real-time photovoltaic environment data into the photovoltaic power generation prediction model, and the photovoltaic power generation prediction model outputs the photovoltaic power generation prediction result.

6. An electronic device, characterized in that: include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images