Research On Smart Grid Load Optimization Prediction Based On Deep Learning

Power system plays an important role in the development of national economy,maintaining the supply and demand balance of the power grid is the key condition for ensuring the stable operation of the power system.Due to the current ultra-large capacity energy storage technology has not been compromised.Therefore,a corresponding power generation dispatch plan must be implemented to ensure stable operation of the power system.Effective short-term forecasting of power load is the basis for the formulation of the dispatch plan,accurate short-term load forecasting can make the scheduling plan more accurate,and it can reduce the loss of power due to inaccurate prediction.To make short-term load forecasting more accurate,this paper first improves the maximum Lyapounov exponent method in the classical prediction method,by constructing an autoregressive error model,it effectively solves the subjective factors existing in the traditional maximum Lyapounov exponential method,such as the problem of embedding dimension.By comparing the prediction results of the traditional maximum Lyapounov exponent method with the improved maximum Lyapounov exponent method,it can be concluded that the prediction results using the improved Lyapounov method are better,the error range has been effectively reduced.However,due to the high accuracy requirements of the forecasting accuracy of the grids recent dispatching plan,the improved maximum Lyapounov exponent method predicts that the fitting accuracy still falls short of the specified conditions.Therefore,this paper proposes a short-term load forecasting optimization method based on deep learning.This method builds a deep learning neural network architecture,adopt self-learning and optimization methods,analyze and mine the characteristic information implied in the power data,the resulting short-term load forecasting is more accurate,the error range is smaller,and the robustness is better.By comparing the improved maximum Lyapounov exponential model with the results predicted by the deep learning prediction model,It can be concluded that:the short-term load forecasting effect based on the deep learning model is significantly better than the improved Lyapounov prediction method,The error range is significantly reduced,Individual maximum error is eliminated,Accuracy is significantly improved and robustness is better.At last,this article is based on the MATLAB2017 experimental platform,the proposed model was simulated using historical data provided by Guizhou Power Grid Corporation,the validity and feasibility of the model were verified,it can be provided reference for relevant departments of the grid company.