Short-term photovoltaic and wind turbine power prediction method and system based on nuclear extreme learning machine

Abstract

The present invention discloses a short-term photovoltaic and wind turbine power prediction method and system based on a nuclear extreme learning machine, which relates to the field of photovoltaic power prediction technology, including: collecting historical power output data and historical meteorological data of photovoltaics and wind turbines; defining a time series window, extracting historical power output data features and historical meteorological data features; constructing a nuclear extreme learning machine model based on a radial basis function to perform short-term power prediction. The short-term photovoltaic and wind turbine power prediction method based on a nuclear extreme learning machine provided by the present invention improves the accuracy and adaptability of the prediction model. It can more accurately capture and learn the complex nonlinear relationship between power output and meteorological conditions, thereby improving the accuracy of the prediction. It enables the operation and maintenance team to respond in time and optimize maintenance and overhaul plans. It optimizes the efficiency of energy production and optimizes the load distribution of the power grid, while ensuring the stability of the power grid, improving economic benefits.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power prediction, and in particular to a short-term photovoltaic and wind turbine power prediction method and system based on a nuclear extreme learning machine. Background Art

[0002] Photovoltaic and wind power generation are important renewable energy sources. Their power output is affected by meteorological conditions such as solar irradiance, wind speed, temperature and other factors, and has significant uncertainty and volatility. Traditional power forecasting technologies, such as statistical time series models and simple machine learning methods, can predict energy output to a certain extent, but these methods often assume that data has a linear relationship and are difficult to accurately capture dynamic changes in complex environments. In addition, these traditional methods usually rely on manual experience for the selection and optimization of model parameters, and lack effective mechanisms to automatically adjust the model to adapt to real-time changing environments and data characteristics.

[0003] In addition, the energy management strategies in the existing technologies are usually static, that is, fixed charging and discharging cycles and maintenance plans are preset, which not only increases the energy waste caused by untimely response, but also increases the risk of equipment failure due to failure to adjust the equipment operation status in time to adapt to sudden energy fluctuations. Therefore, the existing prediction and management systems have obvious deficiencies in flexibility and adaptability, which limits the application efficiency and reliability of photovoltaic and wind power generation technologies in complex environments. 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 that the existing power prediction methods are insufficient in capturing nonlinear meteorological factors and dynamic energy output fluctuations, and the optimization problem of how to dynamically adjust the energy management strategy.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a short-term photovoltaic and wind turbine power prediction method based on a nuclear extreme learning machine, comprising:

[0007] Collect historical power output data of photovoltaic and wind turbines and historical meteorological data;

[0008] Define the time series window and extract the historical power output data features and historical meteorological data features;

[0009] A kernel extreme learning machine model is constructed based on radial basis function for short-term power prediction.

[0010] As a preferred solution of the short-term photovoltaic and wind turbine power prediction method based on nuclear extreme learning machine described in the present invention, the historical meteorological data includes solar irradiance, wind speed, temperature, humidity and air pressure.

[0011] As a preferred solution of the short-term photovoltaic and wind turbine power prediction method based on the kernel extreme learning machine described in the present invention, wherein: the extraction of historical power output data features includes setting a window size W for each prediction time point t, calculating the mean and standard deviation of the power in each window, and determining the power skewness and power kurtosis based on the mean and standard deviation of the power;

[0012] Construct a multivariate linear regression model to predict the trend of data in the short term;

[0013] The average value is expressed as,

[0014] ,

[0015] in, represents the average value of data within the time window, represents the length of the time window, Indicates the current time point. Represents an index, which is used to traverse the time points within the window. Indicates time point Observed values ​​of data;

[0016] The standard deviation is expressed as,

[0017] ,

[0018] in, represents the standard deviation of the data within the time window, Indicates time point The data observation value, represents a defined mean value;

[0019] The power skewness and power kurtosis are determined as,

[0020] ,

[0021] Among them, Skew represents skewness, which measures the asymmetry of data distribution, and Kurt represents kurtosis, which measures the sharpness of data distribution;

[0022] The multivariate linear regression model is expressed as,

[0023] ,

[0024] in, represents the power output prediction result, represents the regression model parameters, Indicates time The skewness value of Indicates time The kurtosis value of represents the error term.

[0025] As a preferred solution of the short-term photovoltaic and wind turbine power prediction method based on the kernel extreme learning machine described in the present invention, wherein: the historical meteorological data feature is to calculate the moving average and moving standard deviation of each historical meteorological data;

[0026] The STL decomposition technique is used to extract the trend component and seasonal component from the moving average series of each meteorological parameter;

[0027] The moving average is expressed as,

[0028] ,

[0029] in, Indicates time The moving average of Indicates time point Observed values ​​of meteorological parameters;

[0030] The moving standard deviation is expressed as,

[0031] ,

[0032] in, Indicates time The moving standard deviation of

[0033] The STL decomposition technique is expressed as,

[0034] ,

[0035] ,

[0036] ,

[0037] ,

[0038] in, represents the deseasonalized data, represents the trend component, represents the initial seasonal component, represents the seasonalized data, represents the seasonal component, represents the bandwidth parameter of the Loess method in trend fitting, represents the bandwidth parameter of the Loess method in seasonal fitting, and Represents the adjustment factor.

[0039] As a preferred solution of the short-term photovoltaic and wind turbine power prediction method based on nuclear extreme learning machine described in the present invention, the power output prediction result is used , the trend component of each historical meteorological data and seasonal components Construct a feature matrix, apply data normalization to process the feature matrix, and input it into the kernel extreme learning machine model;

[0040] The feature matrix is ​​expressed as,

[0041] ,

[0042] in, Indicates meteorological parameters In time The trend component of Indicates meteorological parameters In time The seasonal component of Indicates Meteorological parameters, Represents the total number of meteorological parameters considered.

[0043] As a preferred solution of the short-term photovoltaic and wind turbine power prediction method based on the kernel extreme learning machine described in the present invention, the construction of the kernel extreme learning machine model based on the radial basis function includes adaptively initializing the kernel parameters and automatically adjusting the width parameters of the radial basis function kernel according to the variance of the data, which is expressed as:

[0044] ,

[0045] in, Represents the width parameter that controls the spatial expansion of the kernel function. represents the variance of the input feature matrix, represents a constant;

[0046] According to different data characteristics, a multi-scale RBF kernel function is constructed, which is expressed as:

[0047] ,

[0048] in, Represents data points and The kernel function value between Indicates The weight of the scale, Indicates The width parameter of each scale, Indicates The adjustment parameters of the feature scale;

[0049] The regularization parameter is introduced to balance the complexity of the model and the training efficiency, expressed as,

[0050] ,

[0051] in, represents the loss function, represents the target value vector, represents the kernel matrix calculated based on the multi-scale RBF kernel, represents the output weight, represents the regularization parameter;

[0052] The loss function is minimized by solving the regularized linear equations as:

[0053] ,

[0054] in, represents the identity matrix, with dimensions same, is the transpose of the kernel matrix;

[0055] Introducing the firefly algorithm to optimize kernel parameters and weight coefficient , expressed as,

[0056] ,

[0057] ,

[0058] in, Firefly No. The current value of the kernel width parameter, Indicates the updated firefly No. The kernel width parameter value, represents the attraction coefficient used to update the kernel width parameter, The best firefly among all fireflies No. kernel width parameter, represents the parameters for generating random perturbations, Indicates The upper and lower bounds of the kernel width parameter, Firefly No. The current value of the weight, Indicates the updated firefly No. Weight value, represents the attraction coefficient used to update the weights, The best firefly among all fireflies No. weights, Indicates The upper and lower bounds of the weights;

[0059] Use the optimized kernel parameters and weight coefficients to update the extreme learning machine model, expressed as,

[0060] ,

[0061] in, represents an updated data point;

[0062] The short-term power prediction is expressed as:

[0063] .

[0064] As a preferred solution of the short-term photovoltaic and wind turbine power prediction method based on the kernel extreme learning machine described in the present invention, the short-term power prediction includes determining the high output threshold as , determine the low output threshold as ;

[0065] like , judged as high power output, adjust the energy output management of photovoltaic or wind power generation systems by increasing the charging rate of the battery storage system to utilize the additional energy, inspect and maintain wind turbine blades or photovoltaic panels based on the predicted high power output, prepare to adjust the grid load and optimize power distribution;

[0066] like , judged as low power output, reduce the charging rate of the battery storage system and increase the discharge operation, balance the supply and demand relationship of the power grid, postpone non-critical equipment maintenance activities, adjust the power purchase plan, give priority to the use of energy storage equipment to meet basic needs, and adjust the battery charging and discharging strategy and maintenance plan according to the actual output.

[0067] Another object of the present invention is to provide a short-term photovoltaic and wind turbine power prediction system based on a nuclear extreme learning machine, which can solve the problems of insufficient accuracy and difficulty in parameter optimization in existing prediction technologies by constructing an intelligent prediction system based on a nuclear extreme learning machine and a firefly algorithm, while improving the adaptability and response efficiency of energy management.

[0068] To solve the above technical problems, the present invention provides the following technical solutions: a short-term photovoltaic and wind turbine power prediction system based on a nuclear extreme learning machine, comprising: a data acquisition module, a feature extraction module and a power prediction module; the data acquisition module is used to collect historical power output data and historical meteorological data of photovoltaics and wind turbines; the feature extraction module is used to define a time series window and extract historical power output data features and historical meteorological data features; the power prediction module is used to construct a nuclear extreme learning machine model based on a radial basis function to perform short-term power prediction.

[0069] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the short-term photovoltaic and wind turbine power prediction method based on a kernel extreme learning machine are implemented as described above.

[0070] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the short-term photovoltaic and wind turbine power prediction method based on a kernel extreme learning machine as described above.

[0071] Beneficial effects of the invention: The short-term photovoltaic and wind turbine power prediction method based on the kernel extreme learning machine provided by the invention improves the accuracy and adaptability of the prediction model by introducing the kernel extreme learning machine and the firefly algorithm. By dynamically adjusting the kernel function parameters, the model can more accurately capture and learn the complex nonlinear relationship between power output and meteorological conditions, thereby improving the accuracy of the prediction.

[0072] It helps to identify power fluctuations that may affect system stability and safety in advance, allowing the operation and maintenance team to respond in time and optimize maintenance and overhaul plans.

[0073] The combination of predictive capabilities and dynamic energy management strategies optimizes the efficiency of energy production and optimizes the load distribution of the power grid, thereby improving economic benefits while ensuring power grid stability. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] 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 work.

[0075] Figure 1 An overall flow chart of a short-term photovoltaic and wind turbine power prediction method based on a kernel extreme learning machine provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0076] 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.

[0077] 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.

[0078] Example 1, reference Figure 1 , which is an embodiment of the present invention, provides a short-term photovoltaic and wind turbine power prediction method based on a kernel extreme learning machine, comprising:

[0079] Collect historical power output data of photovoltaic and wind turbines and historical meteorological data;

[0080] Define the time series window and extract the historical power output data features and historical meteorological data features;

[0081] A kernel extreme learning machine model is constructed based on radial basis function for short-term power prediction.

[0082] The historical meteorological data include solar irradiance, wind speed, temperature, humidity and air pressure.

[0083] The extracting of historical power output data features includes setting a window size W for each prediction time point t, calculating the mean value and standard deviation of the power in each window, and determining the power skewness and power kurtosis based on the mean value and standard deviation of the power;

[0084] Construct a multivariate linear regression model to predict the trend of data in the short term;

[0085] The average value is expressed as,

[0086] ,

[0087] in, represents the average value of data within the time window, represents the length of the time window, Indicates the current time point. Represents an index, which is used to traverse the time points within the window. Indicates time point Observed values ​​of data;

[0088] The standard deviation is expressed as,

[0089] ,

[0090] in, represents the standard deviation of the data within the time window, Indicates time point The data observation value, represents a defined mean value;

[0091] The power skewness and power kurtosis are determined as,

[0092] ,

[0093] Among them, Skew represents skewness, which measures the asymmetry of data distribution, and Kurt represents kurtosis, which measures the sharpness of data distribution;

[0094] The multivariate linear regression model is expressed as,

[0095] ,

[0096] in, represents the power output prediction result, represents the regression model parameters, Indicates time The skewness value of Indicates time The kurtosis value of represents the error term.

[0097] The historical meteorological data feature is to calculate a moving average and a moving standard deviation for each type of historical meteorological data;

[0098] The STL decomposition technique is used to extract the trend component and seasonal component from the moving average series of each meteorological parameter;

[0099] The moving average is expressed as,

[0100] ,

[0101] in, Indicates time The moving average of Indicates time point Observed values ​​of meteorological parameters;

[0102] The moving standard deviation is expressed as,

[0103] ,

[0104] in, Indicates time The moving standard deviation of

[0105] The STL decomposition technique is expressed as,

[0106] ,

[0107] ,

[0108] ,

[0109] ,

[0110] in, represents the deseasonalized data, represents the trend component, represents the initial seasonal component, represents the seasonalized data, represents the seasonal component, represents the bandwidth parameter of the Loess method in trend fitting, represents the bandwidth parameter of the Loess method in seasonal fitting, and Represents the adjustment factor.

[0111] The LOESS method is to smooth by fitting a low-order polynomial on a local data segment. The steps are as follows: select a data point and select a window around it (that is, select a segment of data).

[0112] A polynomial is fitted within this window using weighted least squares. The weights are controlled by a smoothing parameter (bandwidth), with the more distant points having smaller weights.

[0113] Repeat the above steps for each point in the window to obtain a smoothed curve.

[0114] Use power output to predict results , the trend component of each historical meteorological data and seasonal components Construct a feature matrix, apply data normalization to process the feature matrix, and input it into the kernel extreme learning machine model;

[0115] The feature matrix is ​​expressed as,

[0116] ,

[0117] in, Indicates meteorological parameters In time The trend component of Indicates meteorological parameters In time The seasonal component of Indicates Meteorological parameters, Represents the total number of meteorological parameters considered.

[0118] The construction of the kernel extreme learning machine model based on the radial basis function includes adaptively initializing the kernel parameters and automatically adjusting the width parameter of the radial basis function kernel according to the variance of the data, which is expressed as:

[0119] ,

[0120] in, Represents the width parameter that controls the spatial expansion of the kernel function. represents the variance of the input feature matrix, represents a constant;

[0121] According to different data characteristics, a multi-scale RBF kernel function is constructed, which is expressed as:

[0122] ,

[0123] in, Represents data points and The kernel function value between Indicates The weight of the scale, Indicates The width parameter of each scale, Indicates The adjustment parameters of the feature scale;

[0124] The regularization parameter is introduced to balance the complexity of the model and the training efficiency, expressed as,

[0125] ,

[0126] in, represents the loss function, represents the target value vector, represents the kernel matrix calculated based on the multi-scale RBF kernel, represents the output weight, represents the regularization parameter;

[0127] The first term of the loss function is the data error term, which represents the error between the model prediction value and the actual target value. is a regularization term, which represents the penalty for model complexity.

[0128] The loss function is minimized by solving the regularized linear equations as:

[0129] ,

[0130] in, represents the identity matrix, with dimensions same, is the transpose of the kernel matrix;

[0131] The identity matrix I is a diagonal matrix where the elements on the diagonal are 1 and the remaining elements are 0. Its dimension is the same as the kernel matrix K. For the m×m identity matrix I, it is expressed as:

[0132] ,

[0133] By introducing regularization parameters and solving regularized linear equations, the model can not only obtain good fitting results on the training data, but also effectively control the complexity of the model, prevent overfitting, and improve the generalization ability of the model on new data. The regularization term imposes a penalty on the model parameter ω, so that the model parameter is not too large, thereby smoothing the model and enhancing robustness.

[0134] Introducing the firefly algorithm to optimize kernel parameters and weight coefficient , expressed as,

[0135] ,

[0136] ,

[0137] in, Firefly No. The current value of the kernel width parameter, Indicates the updated firefly No. The kernel width parameter value, represents the attraction coefficient used to update the kernel width parameter, The best firefly among all fireflies No. kernel width parameter, represents the parameters for generating random perturbations, Indicates The upper and lower bounds of the kernel width parameter, Firefly No. The current value of the weight, Indicates the updated firefly No. Weight value, represents the attraction coefficient used to update the weights, The best firefly among all fireflies No. weights, Indicates The upper and lower bounds of the weights;

[0138] Use the optimized kernel parameters and weight coefficients to update the extreme learning machine model, expressed as,

[0139] ,

[0140] in, represents an updated data point;

[0141] The short-term power prediction is expressed as:

[0142] .

[0143] The short-term power prediction includes determining a high output threshold as , determine the low output threshold as ;

[0144] like , judged as high power output, adjust the energy output management of photovoltaic or wind power generation systems by increasing the charging rate of the battery storage system to utilize the additional energy, inspect and maintain wind turbine blades or photovoltaic panels based on the predicted high power output, prepare to adjust the grid load and optimize power distribution;

[0145] like , judged as low power output, reduce the charging rate of the battery storage system and increase the discharge operation, balance the supply and demand relationship of the power grid, postpone non-critical equipment maintenance activities, adjust the power purchase plan, give priority to the use of energy storage equipment to meet basic needs, and adjust the battery charging and discharging strategy and maintenance plan according to the actual output.

[0146] Embodiment 2 is an embodiment of the present invention, which provides a short-term photovoltaic and wind turbine power prediction system based on a kernel extreme learning machine, including:

[0147] Data acquisition module, feature extraction module and power prediction module;

[0148] The data acquisition module is used to collect historical power output data of photovoltaic and wind turbines and historical meteorological data;

[0149] The feature extraction module is used to define a time series window and extract historical power output data features and historical meteorological data features;

[0150] The power prediction module is used to construct a kernel extreme learning machine model based on radial basis functions to perform short-term power prediction.

[0151] Embodiment 3, an embodiment of the present invention, is different from the first two embodiments in that:

[0152] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0153] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0154] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0155] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0156] Example 4 is an embodiment of the present invention, which provides a short-term photovoltaic and wind turbine power prediction method based on a nuclear extreme learning machine. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0157] A real-world dataset containing historical power output and relevant meteorological parameters (wind speed, solar irradiance, temperature, etc.) is selected. Assume that the dataset contains 1,000 data points, divided into a training set (800 data points) and a test set (200 data points). The data is cleaned and normalized to ensure the consistency and accuracy of the model input.

[0158] Traditional methods use multiple linear regression MLR, support vector machine SVM, neural network NN and random forest RF models.

[0159] Multiple linear regression (MLR) directly uses the mean and standard deviation of historical power output and meteorological data as features.

[0160] The support vector machine (SVM) uses the mean and standard deviation of historical power output and meteorological data as features.

[0161] The neural network NN uses the mean and standard deviation of historical power output and meteorological data as features.

[0162] Random Forest RF uses the mean and standard deviation of historical power output and meteorological data as features.

[0163] The method of the present invention uses a feature construction process, calculates the moving average and standard deviation, and applies STL decomposition to extract trend and seasonal components.

[0164] All models were trained. Mean square error (MSE), mean absolute error (MAE), and coefficient of determination (R²) were used as performance indicators to evaluate the prediction accuracy and stability of the models.

[0165] Under the same conditions, all models are predicted using the same test dataset, and the performance data of each model is collected and recorded. The experimental results are shown in Table 1.

[0166] Table 1 Comparison of experimental results

[0167] ,

[0168] Traditional methods only use simple statistical features in the feature construction process, without considering the complex patterns and nonlinear relationships in the data. The method of the present invention captures the long-term trend and cyclical changes in the data by calculating the moving average and standard deviation and applying STL decomposition to extract the trend and seasonal components, so that the model can perform power prediction more accurately.

[0169] Traditional methods are relatively simple in model optimization and do not introduce adaptive optimization mechanisms. The method of the present invention optimizes kernel parameters and weights through the firefly algorithm and adaptively adjusts model parameters, so that the model can better adapt to different data characteristics and improve the prediction accuracy and stability of the model.

[0170] In traditional methods, the nonlinear processing ability of the model is weak and it is difficult to capture the complex relationships in the data. The method of the present invention introduces a multi-scale RBF kernel function, which enables the model to process the complex nonlinear relationships in the data and improves the model's fitting ability and prediction accuracy.

[0171] Through the above experiments and analysis, it can be seen that the superior performance of the method of the present invention in short-term power forecasting lies not only in its prediction accuracy and stability, but also in its accurate capture of complex data relationships through feature construction and model optimization. Compared with traditional methods, the method of the present invention significantly improves energy utilization efficiency and system stability in practical applications by optimizing energy management and equipment maintenance strategies.

[0172] 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 short-term photovoltaic and wind turbine power prediction method based on a nuclear extreme learning machine, characterized in that: include: Collect historical power output data of photovoltaic and wind turbines and historical meteorological data; Define the time series window and extract the historical power output data features and historical meteorological data features; A kernel extreme learning machine model is constructed based on radial basis functions for short-term power prediction; The extracting of historical power output data features includes setting a window size W for each prediction time point t, calculating the mean value and standard deviation of the power in each window, and determining the power skewness and power kurtosis based on the mean value and standard deviation of the power; Construct a multivariate linear regression model to predict the trend of data in the short term; The average value is expressed as, Among them, μ represents the average value of the data in the time window, W represents the length of the time window, t represents the current time point, i represents the index used to traverse the time points in the window, and x i Represents the data observation value at time point i; The standard deviation is expressed as, Among them, σ represents the standard deviation of the data in the time window, x i represents the data observation value at time point i, and μ represents the defined mean value; The power skewness and power kurtosis are determined as, Among them, Skew represents skewness, which measures the asymmetry of data distribution, and Kurt represents kurtosis, which measures the sharpness of data distribution; The multivariate linear regression model is expressed as, Y t =α0+α1Skew t +α2Kurt t +∈ t Among them, Y t represents the power output prediction result, α0, α1, α2 represent the regression model parameters, Skew t represents the skewness value at time t, Kurt t represents the kurtosis value at time t, ∈ t represents the error term.

2. The short-term photovoltaic and wind turbine power prediction method based on nuclear extreme learning machine according to claim 1, characterized in that: The historical meteorological data include solar irradiance, wind speed, temperature, humidity and air pressure.

3. The short-term photovoltaic and wind turbine power prediction method based on nuclear extreme learning machine according to claim 2, characterized in that: The historical meteorological data feature is to calculate a moving average and a moving standard deviation for each type of historical meteorological data; The STL decomposition technique is used to extract the trend component and seasonal component from the moving average series of each meteorological parameter; The moving average is expressed as, Among them, M μ,t (z) represents the moving average of time t, z i represents the observed value of meteorological parameters at time point i; The moving standard deviation is expressed as, Among them, M σ,t (z) represents the moving standard deviation at time t; The STL decomposition technique is expressed as, D t =M μ,t (z)-S init T t =Loess(D t ,span=η×(1+α·M σ,t (z))) D′ t =M μ,t (z)-T t S t =Loess(D′ t ,span=θ×(1+βM σ,t (z))) Among them, D t represents the deseasonalized data, T t represents the trend component, S init represents the initial seasonal component, D′ t represents the seasonalized data, S t represents the seasonal component, η represents the bandwidth parameter of the Loess method in trend fitting, θ represents the bandwidth parameter of the Loess method in seasonal fitting, and α and β represent adjustment factors.

4. The short-term photovoltaic and wind turbine power prediction method based on nuclear extreme learning machine according to claim 3 is characterized in that: Use the power output to predict the result Y t , the trend component T of each historical meteorological data t and seasonal component S t Construct a feature matrix, apply data normalization to process the feature matrix, and input it into the kernel extreme learning machine model; The feature matrix is ​​expressed as, Feature Matrix=[Y t ,T t (z1),S t (z1),…,T t (z n 0,S t (z n )] Among them, T t (z i ) represents the meteorological parameter z i The trend component at time t, S t (z i ) represents the meteorological parameter z i The seasonal component at time t, z i represents the i-th meteorological parameter, and n represents the total number of meteorological parameters considered.

5. The short-term photovoltaic and wind turbine power prediction method based on nuclear extreme learning machine according to claim 4, characterized in that: The construction of the kernel extreme learning machine model based on the radial basis function includes adaptively initializing the kernel parameters and automatically adjusting the width parameter of the radial basis function kernel according to the variance of the data, which is expressed as: Among them, γ represents the width parameter that controls the spatial expansion of the kernel function, Var(Feature Matrix) represents the variance of the input feature matrix, and ∈ represents a constant; According to different data characteristics, a multi-scale RBF kernel function is constructed, which is expressed as: Among them, K(x i ,x j ) represents the data point x i and x j The kernel function value between k represents the weight of the kth scale, γ k represents the width parameter of the kth scale, represents the adjustment parameter of the kth feature scale; The regularization parameter is introduced to balance the complexity of the model and the training efficiency, expressed as, in, represents the loss function, Y represents the target value vector, K represents the kernel matrix calculated based on the multi-scale RBF kernel, ω represents the output weight, and λ represents the regularization parameter; The loss function is minimized by solving the regularized linear equations as: ω=(K T K+λI) -1 K T Y Among them, I represents the identity matrix, which has the same dimension as K. T is the transpose of the kernel matrix; Introducing the firefly algorithm to optimize the kernel parameter γ k and weight coefficient α k , expressed as, Among them, γ k,i represents the current value of the k-th kernel width parameter of firefly i, represents the updated value of the k-th kernel width parameter of firefly i, β γ represents the attraction coefficient used to update the kernel width parameter, represents the kth kernel width parameter of the best performing firefly j among all fireflies, ξ represents the parameter for generating random perturbations, γ k,max ,γ k,min represents the upper and lower bounds of the kth kernel width parameter, α k,i represents the current value of the kth weight of firefly i, represents the kth weight value of firefly i after the update, β α represents the attraction coefficient used to update the weights, represents the kth weight of the best performing firefly j among all fireflies, α k,max ,α k,min represents the upper and lower bounds of the kth weight; Use the optimized kernel parameters and weight coefficients to update the extreme learning machine model, expressed as, Among them, x new represents an updated data point; The short-term power prediction is expressed as: y new =K new (x new ,x i )ω。 6. The short-term photovoltaic and wind turbine power prediction method based on nuclear extreme learning machine according to claim 5, characterized in that: The short-term power prediction comprises determining a high output threshold as μ+σ and determining a low output threshold as μ-σ; If y new >μ+σ, it is judged as high power output, and the energy output management of photovoltaic or wind power generation system is adjusted by increasing the charging rate of battery storage system to utilize the additional energy. According to the predicted high power output, the wind turbine blades or photovoltaic panels are inspected and maintained, and the grid load is prepared to be adjusted and the power distribution is optimized; If y new <μ-σ, it is judged as low power output, the charging rate of the battery storage system is reduced, and the discharge operation is increased to balance the supply and demand relationship of the power grid, postpone non-critical equipment maintenance activities, adjust the power purchase plan, give priority to the use of energy storage equipment to meet basic needs, and adjust the battery charging and discharging strategy and maintenance plan according to the actual output.

7. A system using the short-term photovoltaic and wind turbine power prediction method based on a nuclear extreme learning machine as described in any one of claims 1 to 6, characterized in that: include: Data acquisition module, feature extraction module and power prediction module; The data acquisition module is used to collect historical power output data of photovoltaic and wind turbines and historical meteorological data; The feature extraction module is used to define a time series window and extract historical power output data features and historical meteorological data features; The power prediction module is used to construct a kernel extreme learning machine model based on radial basis functions to perform short-term power prediction.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the short-term photovoltaic and wind turbine power prediction method based on the kernel extreme learning machine described in any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the short-term photovoltaic and wind turbine power prediction method based on a kernel extreme learning machine according to any one of claims 1 to 6 are implemented.

Images