Power load prediction method based on multi-source heterogeneous data feature fusion, system equipment and medium

Abstract

The invention discloses a power load prediction method based on multi-source heterogeneous data feature fusion, system equipment and a medium, and relates to the technical field of power system load prediction, and the method comprises the steps: obtaining a power load prediction data set, and carrying out the preprocessing; performing input feature processing on the static data, the graph data and the time series data based on a multi-layer perceptron, a graph convolutional network and a gating loop unit network, and fusing the processed features based on a feature splicing method to obtain comprehensive feature information; based on full-connection neural network construction, taking the comprehensive features as input, inputting training set data into a model, and performing iterative optimization according to a model effect to adjust model parameters and hyper-parameters; and based on the parameter optimization adjustment model, using the evaluated optimal model to predict the power load in the future time period. The method is superior to a traditional method in precision and stability, can better adapt to complex and changeable load prediction scenes, and has wide application prospects and remarkable actual value.

Description

Technical Field

[0001] The present invention relates to the technical field of power system load forecasting, and specifically to a power load forecasting method, system equipment and medium based on multi-source heterogeneous data feature fusion. Background Art

[0002] Load forecasting is of great significance in the planning and construction of power systems, the determination of power market boundary conditions, and the balance of power supply and demand. First, in terms of power system planning and construction, accurate load forecasting can help power grid companies and power generation companies formulate more scientific and reasonable power generation and transmission and distribution plans, thereby optimizing resource allocation and improving system operation efficiency. In terms of determining power market boundary conditions, load forecasting provides market operators with key information on electricity demand, helping them determine the balance point of power market supply and demand, thereby optimizing the design of market mechanisms and ensuring the healthy operation of the market. In terms of the balance of power supply and demand, through load forecasting, all participating entities can identify the risk of supply and demand imbalance in advance, take corresponding regulatory measures, and ensure the stability and reliability of the power system.

[0003] The existing load forecasting models still have some problems and challenges in practical applications: Incomplete input features: The current load forecasting models are less selective in feature selection, which results in limited feature information that the model can learn. Due to the incomplete features, the load forecasting performance in different scenarios varies greatly, making it difficult to adapt to the changing market environment. The correspondence between input features and prediction algorithms is not clear enough: For different data structures in the power grid and power market, the existing prediction algorithms lack certain selection criteria. Different data types and structures have different requirements for prediction models, and the lack of unified standards limits the applicability and prediction accuracy of the models. Insufficient combined prediction capabilities based on prediction results: In load forecasting, the combined prediction of output results often does not have sufficient ability to extract feature information, especially when predicting load peaks and valleys. Existing models are difficult to effectively capture the extreme values ​​of load changes, affecting the accuracy and reliability of predictions. 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 it is difficult for existing models to effectively capture the extreme value of load changes, which affects the accuracy and reliability of prediction.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for power load forecasting based on multi-source heterogeneous data feature fusion, comprising:

[0007] Obtain the power load forecast data set for preprocessing, and divide the entire data set after the preprocessing is completed;

[0008] Based on multi-layer perceptron, graph convolutional network and gated recurrent unit network, the input features of static data, graph data and time series data are processed respectively, and the processed features are fused based on the feature splicing method to obtain comprehensive feature information;

[0009] Based on the fully connected neural network, a prediction model with comprehensive features as input is constructed and the loss function and optimization algorithm are defined. The training set data is input into the model and the model parameters and hyperparameters are adjusted according to the model effect through iterative optimization.

[0010] The prediction accuracy of the model is evaluated based on the test set, the model is adjusted based on parameter optimization, and the evaluated optimal model is used to predict the power load in the future time period.

[0011] As a preferred solution of the power load forecasting method based on multi-source heterogeneous data feature fusion described in the present invention, wherein: the obtaining of the power load forecasting data set for preprocessing includes obtaining the power load forecasting data set, which is divided into static multivariate data, graph data, and time series data according to type;

[0012] The static multivariate data includes date type and holiday name; the graph data includes grid structure information; the time series data includes load history data, load forecast data, weather history data and weather forecast data;

[0013] Perform data mapping, map character information into numerical information, and map grid structure information into matrix information; perform data cleaning, and fill in missing data through forward and backward filling and linear filling; perform outlier detection through statistical and machine learning methods, and handle outliers through deletion and replacement methods; perform data normalization through statistical methods, and scale all input data in the training set to the range of 0-1.

[0014] As a preferred solution of the power load forecasting method based on multi-source heterogeneous data feature fusion described in the present invention, wherein: the input feature processing includes: a static data processing unit, which processes static data through a multi-layer perceptron MLP, and the unit includes two fully connected layers, and extracts static features through a ReLU activation function; a graph data processing unit, which uses a graph convolutional network GCN to process power grid structure data, and the GCNConv layer is used to aggregate neighbor node information and extract the relationship between nodes. The input is a node feature matrix and edge connection information, and a graph feature representation is obtained through two layers of GCN convolution;

[0015] Time series data processing unit GRU, use GRU to process time series data, the GRU layer learns the dynamic characteristics of the time series, obtains the final hidden state through the GRU network, and extracts the time series features through the fully connected layer fc.

[0016] As a preferred solution of the power load forecasting method based on multi-source heterogeneous data feature fusion described in the present invention, wherein: the comprehensive feature information obtained by fusing the processed features includes: static features , graph features , Time Series Data Characteristics ; Using feature extraction units for static, graph and time series data, various types of data are input into corresponding units during forward propagation and corresponding feature representations are obtained;

[0017] Feature concatenation, concatenates the three types of features along the feature dimension to form a comprehensive feature representation, and the formula is expressed as:

[0018] The splicing formula is:

[0019] Comprehensive feature dimension, comprehensive features after splicing The dimension is , then the comprehensive feature vector contains the information of all input features; Generate a feature fusion layer, use a fully connected layer, and obtain the fused feature representation by linear transformation and nonlinear activation of the concatenated features; The prediction output layer is used to represent the final output layer, and the fused features are mapped to the prediction target space to generate the load prediction results.

[0020] As a preferred solution of the power load forecasting method based on multi-source heterogeneous data feature fusion described in the present invention, wherein: the prediction model based on the fully connected neural network with comprehensive features as input and the loss function and optimization algorithm are defined, including: After being processed by the feature fusion layer, it is input into a fully connected neural network to generate the prediction output; assuming that the dimension of the input feature is ,Adopt a multi-layer fully connected layer structure, and finally output the predicted value;

[0021] Select the MSE loss function and Adam optimizer, and set the learning rate; perform forward propagation, input the input features into the model, and generate predicted values; for each new sample , input it into the trained model for forward propagation, and calculate the output of the model The formula is: , in, represents the forward propagation process of the model, Represents the current parameters of the model;

[0022] Calculate the MSE loss between the predicted value and the true load value; each time a new sample is received After that, the model performs forward propagation to obtain the predicted value , and calculate the loss function , the commonly used loss function is the mean square error MSE; assuming the predicted value is , the target value is , loss function Defined as:

[0023] To calculate the gradient of the loss with respect to the output, back propagation first starts from the output layer to calculate the gradient; assuming that the output layer of the neural network is a single predicted value, the gradient of the loss function with respect to the output The formula is:

[0024] According to the chain rule, we then calculate the gradient of each layer of network parameters; assuming that the network contains multiple hidden layers, we calculate the gradient of each layer using the chain rule; Assume It is The weighted input of the layer, It is The activation value of the layer and the weight are , the bias is , the activation function is ;

[0025] The gradient formula of the output layer is expressed as: , in, represents the derivative of the activation function;

[0026] The gradient of the hidden layer, for each layer From the last layer Calculate forward: , in, Indicates The error term of the layer, Indicates Backpropagation and optimization of the layer's weight matrix, calculating gradients through backpropagation, and using the optimizer to update the model parameters;

[0027] Calculate the gradient of each layer's weights and biases, and get the error term , the gradient of weights and biases of each layer can be calculated, and the gradient formula of weights is expressed as:

[0028] Calculate the MSE loss between the predicted value and the true load value; optimize the model through back-propagation, calculate the gradient through back-propagation, and update the model parameters using the optimizer; calculate the loss against the model parameters Gradient ,make ;

[0029] The mean formula of the gradient is expressed as:

[0030] The mean formula of the squared gradient is expressed as:

[0031] The deviation correction formula is expressed as:

[0032] The parameter update formula is expressed as: , in, represents the current gradient, represents the learning rate, and represents the attenuation rate, Represents a small value, represents the moving average of the gradient, represents the mean gradient of the previous time step, represents the mean of the squared gradient, represents the bias-corrected first-order moment estimate, represents the bias-corrected second-order moment estimate, represents the second moment estimate for the previous time step.

[0033] As a preferred solution of the power load forecasting method based on multi-source heterogeneous data feature fusion described in the present invention, wherein: the adjusting model parameters and hyperparameters includes, if the model is found to be underfitting and overfitting on the validation set, optimizing by adjusting the model parameters and hyperparameters; selecting an initial hyperparameter combination, training the model using the training set and recording the validation set loss;

[0034] Based on the evaluated hyperparameter combination, a proxy model is constructed to predict the performance of other hyperparameter combinations and select the next sampling point. The acquisition function is the expected improvement El, and the calculation formula is expressed as: , in, represents the predicted value of the surrogate model, Represents the current best hyperparameters; E represents the expected value for future improvements; update the proxy model, train the model using the new hyperparameter combination, update the validation set loss, incorporate the results into the proxy model, and repeat the iteration until the termination condition is met.

[0035] As a preferred solution of the power load forecasting method based on multi-source heterogeneous data feature fusion described in the present invention, wherein: the use of the evaluated optimal model to forecast the power load in the future time period includes defining the safety upper limit parameter of the parameter power load , Risk threshold of system overload , the threshold for judging whether to start the backup power supply ;

[0036] Low load situation, if the load is predicted Less than parameter , indicating that the current load is low; the system starts energy-saving mode to reduce the load output in non-critical areas; dispatches non-critical load equipment, optimizes energy use, and reduces the overall load;

[0037] Medium load situation, if the load is predicted In the parameters and During this period, the system load is normal, but close to the upper limit, and it maintains stable operation; the system continues to maintain normal load operation, but pays attention to real-time load fluctuations; if the power grid has redundancy, the backup power supply is started to prepare for future load fluctuations;

[0038] High load conditions, if the load is predicted Greater than parameter , indicating that the load is close to or exceeds the carrying capacity of the power system, and starting load dispatch; starting the backup power supply and allocating the excess load to the backup equipment or area; implementing the load reduction strategy, dispatching non-critical areas and equipment to reduce the load; alarming and notifying relevant personnel that manual intervention is required;

[0039] If the load is predicted Exceeding Parameters , judge that the system is facing overload, start the emergency backup power supply and trigger the system protection mechanism, implement load reduction measures for the entire network, start the emergency load cutting procedure to avoid system collapse; trigger the alarm system to provide real-time decision support and visual feedback to the operator;

[0040] When overload and high load situations occur, forecast parameter adjustments are made.

[0041] As a preferred solution of the power load forecasting method based on multi-source heterogeneous data feature fusion described in the present invention, the prediction parameter adjustment includes error optimization and hyperparameter adjustment according to the prediction results and the actual output of the system. Through error feedback, the system dynamically optimizes the prediction model to maintain good prediction accuracy. If the error Less than the preset threshold , indicating that the model is accurate and the current rule continues to be executed; if the error Increases and is less than the threshold , no need to adjust parameters, re-predict, and continue normal operation;

[0042] If the error Larger and greater than the threshold , indicating that the model prediction is inaccurate, and there may be underfitting or overfitting; adjust the learning rate and batch size hyperparameters through online optimization, and use dynamic learning rate adjustment rules.

[0043] As a preferred solution of the power load forecasting system based on multi-source heterogeneous data feature fusion described in the present invention, it includes: a data acquisition module, a data preprocessing module, a power load forecasting model training module, and a power load forecasting model prediction module;

[0044] Data acquisition module, which collects various relevant data from multi-source heterogeneous data, including weather data, historical load data, and power grid structure;

[0045] Data preprocessing module, including non-numerical feature numerical mapping module, data missing and anomaly detection module, and data normalization module; cleans the collected data, removes noise and outliers, and normalizes data of different scales;

[0046] The power load forecasting model training module includes an input data feature processing module and a feature fusion module; the input data feature processing module includes feature extraction units for static, graph and time series data; the system uses multiple algorithms to extract meaningful features from pre-processed data for different data features, and uses the fused feature data to train the forecasting model;

[0047] The power load forecasting model prediction module, including the data import module, the model prediction module and the model prediction result visualization module, applies the trained model to the actual data for load forecasting. The system will evaluate and analyze the prediction results to provide a basis and support for relevant decisions.

[0048] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a step of a power load forecasting method based on multi-source heterogeneous data feature fusion.

[0049] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a power load forecasting method based on multi-source heterogeneous data feature fusion.

[0050] Beneficial effects of the present invention: The power load forecasting method based on multi-source heterogeneous data feature fusion provided by the present invention adopts multi-source heterogeneous data feature fusion technology, which can simultaneously sample influencing factors of multiple dimensions, thereby extracting more valuable information. Using the most appropriate algorithm for feature extraction for different types of data can maximize the capture of potential laws in various types of data and improve the accuracy and reliability of prediction. Multi-source data feature fusion is not only a simple data splicing, but a deep fusion at the feature level. By fusing data features from different sources, new and more representative features are formed. This fusion method can make full use of the correlation and complementarity between different data sources, so that the model can more comprehensively understand and capture the complex factors that affect load changes. The load forecasting algorithm based on multi-source heterogeneous data feature fusion is superior to traditional methods in both accuracy and stability, can better adapt to complex and changeable load forecasting scenarios, and has broad application prospects and significant practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] 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.

[0052] Figure 1 An overall flow chart of a method for power load forecasting based on multi-source heterogeneous data feature fusion provided for the first embodiment of the present invention.

[0053] Figure 2 A schematic diagram of module connections of a method for power load forecasting based on multi-source heterogeneous data feature fusion provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0054] 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.

[0055] Example 1, reference Figure 1, which is an embodiment of the present invention, provides a method for power load forecasting based on multi-source heterogeneous data feature fusion, comprising:

[0056] S1: Obtain the power load forecast data set for preprocessing, and divide the entire data set after the preprocessing is completed.

[0057] Furthermore, the obtaining of the power load prediction data set for preprocessing includes obtaining the power load prediction data set, which is divided into static multivariate data, graph data, and time series data by type.

[0058] Furthermore, the static multivariate data includes date type and holiday name; the graph data includes grid structure information; and the time series data includes load history data, load forecast data, weather history data and weather forecast data.

[0059] The Pearson correlation analysis method and the physical characteristics of load data are used to select key indicators required for load forecasting, including static feature data such as date type, holiday name, etc., graphical data such as power grid structure and transmission capacity, and time series data such as historical load data, historical weather data, and predicted weather data.

[0060] Furthermore, data mapping is performed to map character information into numerical information and grid structure information into matrix information. Data cleaning is performed to fill in missing data through forward and backward filling and linear filling. Outlier detection is performed through statistical and machine learning methods, and outliers are processed through deletion and replacement methods. Data normalization is performed through statistical methods to scale all input data in the training set to the range of 0-1.

[0061] Furthermore, the data preprocessing module cleans the collected data and removes noise and outliers to ensure data quality. Next, the data of different scales are normalized to make the data conform to a unified standard, which is convenient for subsequent feature extraction and model training. Data preprocessing is a key step to ensure data quality and consistency, and has an important impact on the final prediction results.

[0062] Collect information on the grid structure and line transmission level, weather data and load data in the predicted area, and build a complete load forecasting data set for model training and verification.

[0063] Based on the data cleaning method, missing values ​​in the data set are identified, and corresponding data such as date type and holiday name are filled in. Linear filling or forward filling is performed on indispensable missing data such as load data and weather data, and abnormal data is modified.

[0064] Based on statistical methods, the data is normalized and the feature data is scaled to the range of 0 to 1, eliminating the impact of the differences in the order of magnitude and dimension of different features on model training, thereby improving the model prediction accuracy. The normalization method expression is as follows: , In the formula, Represents the original data, represents the normalized data, represents the minimum value of the original data set, Indicates the maximum value of the original data set.

[0065] It should be noted that the power load forecasting system includes the following key modules: provincial / regional power load forecasting related characteristic data collection module, data preprocessing module, power load forecasting model training module, and power load forecasting model prediction module. In the data acquisition module, the system collects various relevant data from multi-source heterogeneous data, including weather data, historical load data, power grid structure, etc. Data acquisition is the foundation of the entire system. Through the comprehensive collection of multi-source data, it provides rich information resources for subsequent data processing and analysis.

[0066] S2: Based on the multi-layer perceptron, graph convolutional network and gated recurrent unit network, the input features of static data, graph data and time series data are processed respectively, and the processed features are fused based on the feature splicing method to obtain comprehensive feature information.

[0067] Furthermore, in the feature fusion and model training module, the system uses multiple algorithms to extract meaningful features from the preprocessed data. Then, these fused feature data are used to train the prediction model, and traditional machine learning algorithms or deep learning algorithms can be selected. In this step, the close combination of feature fusion and model training helps to improve the prediction accuracy and generalization ability of the model.

[0068] Furthermore, the input feature processing includes: a static data processing module MLP, which processes static data through MLP. The module contains two fully connected layers and extracts static features through the ReLU activation function; a graph data processing module GCN, which uses GCN to process power grid structure data. The GCNConv layer is used to aggregate neighbor node information and extract the relationship between nodes. The input is the node feature matrix and edge connection information, and the graph feature representation is obtained through two layers of GCN convolution.

[0069] Furthermore, the time series data processing module GRU uses GRU to process time series data. The GRU layer learns the dynamic features of the time series, obtains the final hidden state through the GRU network, and extracts the time series features through the fully connected layer (fc) to update the gate . Input time series data ,in is the number of time steps.

[0070] The single-step update rule of GRU is as follows: , Reset Gate

[0071] , Candidate hidden states

[0072] , Hide status updates

[0073] , in, , , is the weight matrix of GRU.

[0074] The method based on feature splicing fuses the processed features to obtain comprehensive feature information including static features. , graph features , Time Series Data Characteristics ; Use feature extraction modules for static, graph and time series data, input various types of data into corresponding modules during forward propagation and obtain corresponding feature representations.

[0075] Feature concatenation, concatenates the three types of features along the feature dimension to form a comprehensive feature representation, and the formula is expressed as: , The splicing formula is:

[0076] Comprehensive feature dimension, comprehensive features after splicing The dimension is , then the comprehensive feature vector contains the information of all input features.

[0077] The feature fusion layer uses a fully connected layer to obtain the fused feature representation by linear transformation and nonlinear activation of the concatenated features. The prediction output layer represents the final output layer, which maps the fused features to the prediction target space and generates the load prediction results.

[0078] After linear transformation, the obtained feature vectors of different dimensions are transformed into feature matrices with the same number of rows, and then the matrices are spliced ​​by column to obtain the fused features as the input of the subsequent network; after continuous iteration and model training of the fully connected neural network, the optimal parameters are obtained to generate a prediction model. The single-step prediction value of the load forecast is obtained, and then the multi-step prediction method is used to perform multi-step load forecasting.

[0079] Furthermore, the trained optimal model is used to predict the power load data for the next day and the next week, and different indicators are used to evaluate the power load forecasting model to assess the accuracy of the forecast.

[0080] Furthermore, the formulas for each indicator are: , , , in, Represents the number of samples of predicted values ​​and true values, represents the predicted value, represents the true value;

[0081] It should be noted that the model prediction module applies the trained model to actual data for load forecasting. The system will evaluate and analyze the forecast results to ensure the accuracy and reliability of the forecast. The model forecast is the final output of the entire system. Through the forecast of actual data, the system can provide accurate load forecast results and provide a basis and support for relevant decisions.

[0082] S3: Based on a fully connected neural network, a prediction model with comprehensive features as input is constructed and the loss function and optimization algorithm are defined. The training set data is input into the model and the model parameters and hyperparameters are iteratively optimized and adjusted according to the model effect. The fitting effect of the model is evaluated based on the validation data set.

[0083] Furthermore, the prediction model based on the fully connected neural network with comprehensive features as input and the loss function and optimization algorithm are defined, including: After being processed by the feature fusion layer, it is input into a fully connected neural network to generate the prediction output; assuming that the dimension of the input feature is ,A multi-layer fully connected layer structure can be adopted and the predicted value can be output at last.

[0084] Furthermore, the network structure is defined, assuming that the fully connected network contains two hidden layers and the activation function uses ReLU.

[0085] Furthermore, the model formula uses Represents the input layer: , The hidden layer formula is expressed as: , in, , .

[0086] The output layer formula is expressed as: , in, , Indicates the dimension of the final output for each parameter .

[0087] Select the MSE loss function and Adam optimizer, and set the learning rate; forward propagation, input the input features into the model to generate predicted values;

[0088] Calculate the MSE loss between the predicted value and the true load value; each time a new sample is received After that, the model performs forward propagation to obtain the predicted value , and calculate the loss function , the commonly used loss function is the mean square error MSE. Assume that the predicted value is , the target value is , loss function Defined as:

[0089] To calculate the gradient of the loss with respect to the output, backpropagation starts from the output layer to calculate the gradient. Assuming that the output layer of the neural network is a single predicted value (regression problem), the gradient of the loss function (with respect to the output )for:

[0090] This gradient indicates the magnitude of the prediction error and reflects the difference between the predicted value and the actual value.

[0091] According to the chain rule, we can calculate the gradient of each layer of network parameters. Assuming that the network contains multiple hidden layers, we can calculate the gradient of each layer by the chain rule. It is The weighted input of the layer, It is The activation value of the layer and the weight are , the bias is , the activation function is but:

[0092] Gradient of the output layer: , in, represents the derivative of the activation function (for the ReLU activation function, if , otherwise 0).

[0093] The gradient of the hidden layer, for each layer From the last layer Calculate forward: , in, Indicates The error term of the layer, Indicates The weight matrices of the layers are back-propagated and optimized, gradients are calculated through back-propagation, and the optimizer is used to update the model parameters.

[0094] Calculate the gradient of each layer's weights and biases, and get the error term , the gradient of weights and biases of each layer can be calculated, and the gradient formula of weights is expressed as: , in, represents the output of the previous layer, Represents the error term of the current layer. The gradient formula of the bias is expressed as:

[0095] Parameter updating is a key step in online learning. After each gradient is calculated, the model parameters are updated using an optimization algorithm. The most commonly used optimization algorithms include Gradient Descent and its variants, such as Adam and RMSProp.

[0096] Use ordinary gradient descent to update the parameters. The update rule of ordinary gradient descent is as follows: , , in, Represents the learning rate, which controls the step size of each parameter update. and Indicates the current time step The weights and biases. and Represents the gradients of the weights and biases of this layer.

[0097] in, The error term of the current layer is further expressed as follows:

[0098] Furthermore, the mean formula of the squared gradient is expressed as:

[0099] Furthermore, the deviation correction formula is expressed as:

[0100] The parameter update formula is expressed as: ,

[0101] in, represents the current gradient, represents the learning rate, and represents the attenuation rate, Represents a small value, represents the moving average of the gradient, represents the mean of the squared gradient; Represents the model parameters after gradient update.

[0102] Furthermore, the adjusting of model parameters and hyperparameters includes, if the model is found to be underfitting or overfitting on the validation set, optimizing by adjusting the model parameters and hyperparameters; selecting an initial hyperparameter combination, training the model using the training set and recording the validation set loss.

[0103] The adjusting of model parameters and hyperparameters includes, if the model is found to be underfitting or overfitting on the validation set, optimizing by adjusting the model parameters and hyperparameters; selecting an initial hyperparameter combination, training the model using the training set and recording the validation set loss;

[0104] Based on the evaluated hyperparameter combination, a proxy model is constructed to predict the performance of other hyperparameter combinations and select the next sampling point. The acquisition function is the expected improvement El, and the calculation formula is expressed as: , in, represents the predicted value of the surrogate model, Represents the current best hyperparameters; E represents the expected value for future improvements; update the proxy model, train the model using the new hyperparameter combination, update the validation set loss, incorporate the results into the proxy model, and repeat the iteration until the termination condition is met.

[0105] It should be noted that the model is trained using a new hyperparameter combination, the validation set loss is updated, the results are incorporated into the proxy model, and the iterations are repeated until the termination condition is met.

[0106] S4: If there is no underfitting or overfitting phenomenon, the prediction accuracy of the model is evaluated based on the test set, the model is adjusted based on parameter optimization, and the evaluated optimal model is used to predict the power load in the future time period.

[0107] Furthermore, the use of the evaluated optimal model to predict the power load in the future time period includes the optimization goal of feedback adjustment of the predicted value to maximize the acquisition function formula expressed as:

[0108] By maximizing the acquisition function, the Bayesian optimization method gradually approaches the adjustment of the global optimal hyperparameter combination model hyperparameters; a judgment threshold for the difference in loss between the validation set and the test set is set (like ), used to determine whether there is overfitting:

[0109] If the validation set loss is high, increase the number of hidden layer nodes; if the validation set loss is low but the test set loss is high, reduce the number of hidden layer nodes.

[0110] If underfitting occurs, increase the number of hidden layer nodes to improve the expressiveness of the model.

[0111] It can be adjusted using the incremental proportional method or the fixed incremental method. The incremental proportional method formula is expressed as: , Fixed increment method: , The reduction ratio method formula is expressed as: , The fixed reduction method formula is expressed as:

[0112] The use of the evaluated optimal model to predict the power load in the future time period includes defining a safety upper limit parameter of the parameter power load. , Risk threshold of system overload , the threshold for judging whether to start the backup power supply .

[0113] Low load situation, if the load is predicted Less than parameter , indicating that the current load is low; the system starts energy-saving mode to reduce the load output in non-critical areas; dispatches non-critical load equipment, optimizes energy use, and reduces the overall load.

[0114] Medium load situation, if the load is predicted In the parameters and During this time, the system load is normal but close to the upper limit, maintaining stable operation; the system continues to maintain normal load operation, but pays attention to real-time load fluctuations; if the power grid has redundancy, start the backup power supply to prepare for future load fluctuations.

[0115] High load conditions, if the load is predicted Greater than parameter , indicating that the load is close to or exceeds the carrying capacity of the power system, and load scheduling is initiated; the backup power supply is started and the excess load is allocated to the backup equipment or area; the load reduction strategy is implemented, and non-critical areas and equipment are dispatched to reduce the load; an alarm is sounded and relevant personnel are notified that manual intervention is required.

[0116] If the load is predicted Exceeding Parameters , judge that the system is facing overload, start the emergency backup power supply and trigger the system protection mechanism, implement load reduction measures for the entire network, start the emergency load cutting procedure to avoid system collapse; trigger the alarm system to provide real-time decision support and visual feedback to the operator;

[0117] When overload and high load situations occur, forecast parameter adjustments are made and the model is adjusted based on parameter optimization.

[0118] The parameter optimization and adjustment model includes error optimization and hyperparameter adjustment according to the prediction results and the actual output of the system. Through error feedback, the system dynamically optimizes the prediction model to maintain good prediction accuracy. Less than the preset threshold , indicating that the model is accurate and the current rule continues to be executed; if the error Increases and is less than the threshold , no need to adjust parameters, re-predict and continue normal operation.

[0119] If the error Larger and greater than the threshold , indicating that the model prediction is inaccurate, and there may be underfitting or overfitting; adjust the learning rate and batch size hyperparameters through online optimization, and use dynamic learning rate to adjust the model.

[0120] It should be noted that the model is updated, the adjusted number of hidden layer nodes is applied to the model structure, the current model structure and parameters are determined as the final model, and it is used for the actual prediction task.

[0121] Example 2: The following is an embodiment of the present invention, which provides a method for power load forecasting based on multi-source heterogeneous data feature fusion. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0122] Dataset division: After preprocessing, the dataset is divided into training set, validation set and test set, with the ratio of 70%, 15% and 15% respectively. This division ratio ensures the balance between model training and validation. The experiment uses a multi-layer perceptron (MLP) to process static data and extract static feature representations; uses a graph convolutional network (GCN) to process power grid structure data to obtain the feature relationship between nodes; and uses a gated recurrent unit (GRU) to process time series data to capture time-dependent features. The features after the above processing are static features, graph features and time series features. The three types of features are spliced ​​to form a comprehensive integrated feature representation as the input of the subsequent fully connected neural network model.

[0123] With comprehensive features as input, a fully connected neural network (FCNN) is constructed for load forecasting. The model is designed as a two-layer hidden layer structure, and ReLU is selected as the activation function. During the training process, the mean square error (MSE) is used as the loss function, and the Adam optimizer is used for parameter optimization. During training, iterative optimization is used to adjust the model parameters and hyperparameters, including the number of hidden layer nodes, learning rate, etc., so that the model achieves the best fit effect on the validation set.

[0124] The fitting effect of the model is evaluated on the validation set. If the validation set loss is high, the model's expressiveness is increased by increasing the number of hidden layer nodes. If the validation set loss is low but the test set loss is high, the model complexity is reduced by reducing the number of hidden layer nodes. After ensuring that the model is neither overfitting nor underfitting, the prediction accuracy of the model is evaluated on the test set to determine the stability of the model. Finally, the model with the optimal parameter configuration is used to predict the load data for the next week to verify the effectiveness of this method in practical applications.

[0125] During the training process, the mean square error (MSE) is used as the loss function, and the Adam optimizer is used for parameter optimization. During training, the training set data is input into the model, and the parameters and hyperparameters of the model are adjusted through multiple iterations to minimize the loss function.

[0126] By continuously adjusting the model's hyperparameters (such as the number of hidden layer nodes, learning rate, etc.), we ensure that the model achieves the best fit on the validation set. If the validation set loss is high, we will try to increase the number of hidden layer nodes to enhance the model's expressiveness; if the validation set loss is low but the test set loss is high, we will reduce the number of hidden layer nodes to avoid overfitting due to the model being too complex.

[0127] After the model training is completed, the validation set is used to evaluate the model's fitting effect. If overfitting or underfitting occurs on the validation set, the model is optimized by adjusting hyperparameters (such as the number of hidden layer nodes and learning rate). In order to ensure that the model has good generalization ability on unknown data, the loss and prediction accuracy on the test set must also meet certain standards.

[0128] The stability and prediction accuracy of the model are further verified by evaluating it on the test set. Finally, the prediction error on the test set is calculated to confirm the applicability of the model.

[0129] After the parameters and hyperparameters of the model are optimized, it is applied to the power load forecast for the next week. Using the input data for the next week, the model gives the corresponding forecast results, which are compared with the actual load data to verify the effectiveness and accuracy of the model in the actual forecast.

[0130] The power load forecasting method based on multi-layer model fusion of the present invention shows excellent accuracy and stability in load forecasting. The prediction errors of the training set and the validation set are both less than 3%. During the training process, the model gradually improves its adaptability on the validation set by dynamically adjusting the number of hidden layer nodes. In addition, the prediction error on the test set is also basically maintained within 3%, indicating that the model has a good generalization effect in actual scenarios and has a certain robustness.

[0131] The experimental data shows that the error of future predicted load is between 2.4% and 2.7%, which is slightly lower than the training and validation stages, verifying that this method still has high accuracy in future data prediction. Compared with the traditional single model prediction method, the present invention significantly improves the sensitivity to load fluctuations by separately processing static, graph structure and time series features. For example, when the wind speed and humidity change, the predicted load error remains at a low level, showing the strong robustness of this method to meteorological factors. Traditional methods often have high errors when faced with dynamically changing meteorological factors, while this method effectively captures the dynamic relationship of the time series through the GRU network, and can still maintain a high prediction accuracy when meteorological factors change.

[0132] Embodiment 3, as Figure 2 As shown, an embodiment of the present invention provides a power load forecasting system based on multi-source heterogeneous data feature fusion, including a data acquisition module 100, a data preprocessing module 200, a power load forecasting model training module 300, and a power load forecasting model prediction module 400.

[0133] The data collection module 100 collects various relevant data from multi-source heterogeneous data, including weather data, historical load data, and power grid structure.

[0134] The data preprocessing module 200 includes a non-numerical feature numerical mapping module 201, a data missing and anomaly detection module 202, and a data normalization module 203; the collected data is cleaned, noise and outliers are removed, and data of different scales are normalized.

[0135] The power load prediction model training module 300 includes an input data feature processing module 301 and a feature fusion module 302; the input data feature processing module includes feature extraction units for static, graph and time series data; the system uses multiple algorithms to extract meaningful features from preprocessed data among different data features, and uses the fused feature data to train the prediction model.

[0136] The power load forecasting model prediction module 400, including a data import module 401, a model prediction module 402 and a model prediction result visualization module 403, applies the trained model to actual data for load forecasting. The system will evaluate and analyze the prediction results to provide a basis and support for relevant decisions.

[0137] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it 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 for 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 of each embodiment 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. Various media that can store program codes.

[0138] 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.

[0139] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk case (magnetic devices), 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.

[0140] 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 logical 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. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0141] 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 power load forecasting method based on multi-source heterogeneous data feature fusion, characterized in that: include: Obtain the power load forecast data set for preprocessing, and divide the entire data set after the preprocessing is completed; Based on multi-layer perceptron, graph convolutional network and gated recurrent unit network, the input features of static data, graph data and time series data are processed respectively, and the processed features are fused based on the feature splicing method to obtain comprehensive feature information; Based on the fully connected neural network, a prediction model with comprehensive features as input is constructed and the loss function and optimization algorithm are defined. The training set data is input into the model and the model parameters and hyperparameters are adjusted according to the model effect through iterative optimization. The prediction accuracy of the model is evaluated based on the test set, the model is adjusted based on parameter optimization, and the evaluated optimal model is used to predict the power load in the future time period.

2. The method for power load forecasting based on multi-source heterogeneous data feature fusion according to claim 1 is characterized in that: The obtaining of the power load prediction data set for preprocessing includes obtaining the power load prediction data set, which is divided into static multivariate data, graph data, and time series data according to the type; The static multivariate data includes date type and holiday name; the graph data includes grid structure information; the time series data includes load history data, load forecast data, weather history data and weather forecast data; Perform data mapping, map character information into numerical information, and map grid structure information into matrix information; perform data cleaning, and fill in missing data through forward and backward filling and linear filling; perform outlier detection through statistical and machine learning methods, and handle outliers through deletion and replacement methods; perform data normalization through statistical methods, and scale all input data in the training set to the range of 0-1.

3. The method for power load forecasting based on multi-source heterogeneous data feature fusion according to claim 1 or 2, characterized in that: The input feature processing includes: a static data processing unit, which processes static data through a multi-layer perceptron MLP, and the unit includes two fully connected layers, and extracts static features through a ReLU activation function; a graph data processing unit, which uses a graph convolutional network GCN to process power grid structure data, and the GCNConv layer is used to aggregate neighbor node information and extract the relationship between nodes. The input is a node feature matrix and edge connection information, and a graph feature representation is obtained through two layers of GCN convolution; The time series data processing unit uses the gated recurrent unit network GRU to process time series data. The GRU layer learns the dynamic characteristics of the time series, obtains the final hidden state through the GRU network, and extracts the time series features through the fully connected layer fc.

4. The method for power load forecasting based on multi-source heterogeneous data feature fusion according to claim 1, characterized in that: The comprehensive feature information obtained by fusing the processed features includes: static features , graph features , Time Series Data Characteristics ; Using feature extraction units for static, graph and time series data, various types of data are input into corresponding units during forward propagation and corresponding feature representations are obtained; Feature concatenation, concatenates the three types of features along the feature dimension to form a comprehensive feature representation, and the formula is expressed as: ； The splicing formula is: ； Comprehensive feature dimension, comprehensive features after splicing The dimension is , then the comprehensive feature vector contains the information of all input features; Generate a feature fusion layer, use a fully connected layer, and obtain the fused feature representation by linear transformation and nonlinear activation of the concatenated features; The prediction output layer is used to represent the final output layer, and the fused features are mapped to the prediction target space to generate the load prediction results.

5. The method for power load forecasting based on multi-source heterogeneous data feature fusion according to claim 4 is characterized in that: The prediction model based on the fully connected neural network with comprehensive features as input includes: After being processed by the feature fusion layer, it is input into a fully connected neural network to generate the prediction output; assuming that the dimension of the input feature is ,Adopt a multi-layer fully connected layer structure, and finally output the predicted value; Select the MSE loss function and Adam optimizer, and set the learning rate; perform forward propagation, input the input features into the model, and generate predicted values; For each new sample , input it into the trained model for forward propagation, and calculate the output of the model The formula is: ； in, represents the forward propagation process of the model, Represents the current parameters of the model.

6. The method for power load forecasting based on multi-source heterogeneous data feature fusion according to any one of claims 1, 2 or 5, characterized in that: The defining loss function and optimization algorithm include calculating the MSE loss between the predicted value and the actual load value; Each time a new sample is received After that, the model performs forward propagation to obtain the predicted value , and calculate the loss function , the loss function is the mean square error MSE; assuming the predicted value is , the target value is , loss function Defined as: ； To calculate the gradient of the loss with respect to the output, back propagation first starts from the output layer to calculate the gradient; assuming that the output layer of the neural network is a single predicted value, the gradient of the loss function with respect to the output The formula is: ； According to the chain rule, we then calculate the gradient of each layer of network parameters; assuming that the network contains multiple hidden layers, we calculate the gradient of each layer using the chain rule; Assume It is The weighted input of the layer, It is The activation value of the layer and the weight are , the bias is , the activation function is ; The gradient formula of the output layer is expressed as: ； in, represents the derivative of the activation function; The gradient of the hidden layer, for each layer From the last layer Calculate forward: ； in, Indicates The error term of the layer, Indicates Backpropagation and optimization of the layer's weight matrix, calculating gradients through backpropagation, and using the optimizer to update the model parameters; Calculate the gradient of each layer's weights and biases, and get the error term , the gradient of weights and biases of each layer can be calculated, and the gradient formula of weights is expressed as: ； Calculate the MSE loss between the predicted value and the true load value; optimize the model through back-propagation, calculate the gradient through back-propagation, and update the model parameters using the optimizer; calculate the loss against the model parameters Gradient ,make ; The mean formula of the gradient is expressed as: ； The mean formula of the squared gradient is expressed as: ； The deviation correction formula is expressed as: ； The parameter update formula is expressed as: ； in, represents the current gradient, represents the learning rate, and represents the attenuation rate, Represents a small value, represents the moving average of the gradient, represents the mean gradient of the previous time step, represents the mean of the squared gradient, represents the bias-corrected first-order moment estimate, represents the bias-corrected second-order moment estimate, represents the second moment estimate for the previous time step.

7. The method for power load forecasting based on multi-source heterogeneous data feature fusion according to any one of claims 1, 4 or 5, characterized in that: The adjusting of model parameters and hyperparameters includes, if the model is found to be underfitting or overfitting on the validation set, optimizing by adjusting the model parameters and hyperparameters; selecting an initial hyperparameter combination, training the model using the training set and recording the validation set loss; Based on the evaluated hyperparameter combination, a proxy model is constructed to predict the performance of other hyperparameter combinations and select the next sampling point. The acquisition function is the expected improvement El, and the calculation formula is expressed as: ； in, represents the predicted value of the surrogate model, Represents the current best hyperparameters; E represents the expected value for future improvements; update the proxy model, train the model using the new hyperparameter combination, update the validation set loss, incorporate the results into the proxy model, and repeat the iteration until the termination condition is met.

8. The method for power load forecasting based on multi-source heterogeneous data feature fusion according to any one of claims 1, 2, 4 or 5, characterized in that: The use of the evaluated optimal model to predict the power load in the future time period includes defining a safety upper limit parameter of the parameter power load. , Risk threshold of system overload , the threshold for judging whether to start the backup power supply ; Low load situation, if the load is predicted Less than parameter , indicating that the current load is low; the system starts energy-saving mode to reduce the load output in non-critical areas; dispatches non-critical load equipment, optimizes energy use, and reduces the overall load; Medium load situation, if the load is predicted In the parameters and The system load is normal, but close to the upper limit, and maintains stable operation; the system continues to maintain normal load operation, but pays attention to real-time load fluctuations; If the grid is redundant, start the backup power supply to prepare for future load fluctuations; High load conditions, if the load is predicted Greater than parameter , indicating that the load is close to or exceeds the carrying capacity of the power system, and load dispatch is initiated; Start backup power and distribute excess load to backup equipment or areas; implement load reduction strategies and dispatch non-critical areas and equipment to reduce load; Alarm and notify relevant personnel, manual intervention is required; If the load is predicted Exceeding Parameters , judge that the system is facing overload, start the emergency backup power supply and trigger the system protection mechanism, implement load reduction measures for the entire network, start the emergency load cutting procedure to avoid system collapse; trigger the alarm system to provide real-time decision support and visual feedback to the operator; When overload and high load situations occur, forecast parameter adjustments are made and the model is adjusted based on parameter optimization.

9. The method for power load forecasting based on multi-source heterogeneous data feature fusion according to claim 8, characterized in that: The parameter optimization and adjustment model includes error optimization and hyperparameter adjustment according to the prediction results and the actual output of the system. Through error feedback, the system dynamically optimizes the prediction model to maintain good prediction accuracy. Less than the preset threshold , indicating that the model is accurate and the current rule continues to be executed; if the error Increases and is less than the threshold , no need to adjust parameters, re-predict, and continue normal operation; If the error Larger and greater than the threshold , indicating that the model prediction is inaccurate, and there may be underfitting or overfitting; adjust the learning rate and batch size hyperparameters through online optimization, and use dynamic learning rate to adjust the model.

10. A system for power load forecasting method based on multi-source heterogeneous data feature fusion, characterized by: It comprises a data acquisition module (100), a data preprocessing module (200), a power load prediction model training module (300), and a power load prediction model prediction module (400); A data collection module (100) collects various relevant data from multi-source heterogeneous data, including weather data, historical load data, and power grid structure; The data preprocessing module (200) includes a non-numerical feature numerical mapping module (201), a data missing and anomaly detection module (202), and a data normalization module (203); the collected data is cleaned, noise and abnormal values ​​are removed, and data of different scales are normalized; The power load prediction model training module (300) comprises an input data feature processing module (301) and a feature fusion module (302); the input data feature processing module comprises a feature extraction unit for static, graph and time series data; the system uses a plurality of algorithms to extract meaningful features from the pre-processed data for different data features, and uses the fused feature data to train the prediction model; The power load forecasting model prediction module (400) includes a data import module (401), a model prediction module (402) and a model prediction result visualization module (403), which applies the trained model to actual data for load forecasting. The system will evaluate and analyze the prediction results to provide a basis and support for relevant decisions.

11. 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 power load forecasting method based on multi-source heterogeneous data feature fusion are implemented.

12. 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 a power load forecasting method based on multi-source heterogeneous data feature fusion are implemented.

Images