Research On Short-Term Electric Load Forecasting Model Based On Deep Learning

With the development of society,the importance of electrical energy is becoming more and more important.The detailed analysis of the data generated during the operation of the power grid system and related data is becoming a prerequisite to ensure stable operation.Transmission grid.Assembling and using a number of intelligent monitoring equipment,the power grid system has collected more data than ever before.It is an urgent need to adjust the electrical plan according to the prediction results of the electrical load prediction model to prevent damage caused by inaccurate predictions.In this paper,we propose a deep learning-based short-term power load prediction model that considers the problem of low precision by considering the factors that influence the shortness of the prediction step size of unilateral incomplete information acquisition according to the deep learning method.The main findings of the existing power load prediction model are as follows.（1）Considering the various characteristics of the power load series affected by various external factors and the short prediction step length of the existing load model,A short-term power load forecasting model based on convolutional neural network-bidirectional long-short-term memory network（CNN＿BILSTM）is proposed.First,we use convolutional neural networks to extract and reconstruct various input features and process power load data using bidirectional long-term and short-term memory networks.Next,according to the above network,we establish a prediction model with multiple factors,multiple steps and updates,and case analysis to make sure that the model can maintain good prediction accuracy even in the long prediction steps.（2）In the current power load prediction model,the short-term power load prediction method（VMD-XGBoost）is proposed based on the highest slope boost algorithm,which requires improvement of prediction accuracy,aiming at the problem of insufficient information acquisition for prediction.Has become.First we decompose the original preprocessed power load sequence using the variable modal method.The following sets up the XGBoost prediction model of the decomposed sequence.Finally,the prediction results of each sequence are added to get the final load prediction value.Case analysis confirms that the proposed method has higher prediction accuracy and lower computational complexity.（3）In multi-component prediction,it aims at the low efficiency of the conventional power load prediction model and the need for optimization of the prediction model,which is a short-term model according to the variable mode decomposition sample entropy time line network（VMD-SA）.-TCN)is being proposed.Power load prediction model.First,the power load sequence is decomposed to calculate the entropy values of each component,and then the subsequences with close entropy values are merged.Next,a TCN prediction model is established so that a new sequence is predicted after recombination.Finally,each sequence is predicted.The results obtained are superimposed on the linearity to give the final prediction.Case analysis improves the predictive accuracy of model predictions.