Research Of Short-term Power Load Forecasting Based On PSO-LSTM Algorithm

Power load is an important part of power system operation and planning,and its accurate analysis and prediction is of great significance.In view of the low accuracy and low automation of short-term power load forecasting,this paper designs and implements a power load forecasting model based on LSTM algorithm,uses particle swarm optimization algorithm to optimize the model parameters,and designs a power load forecasting platform which integrates data management,model training and results display to make the whole forecasting process more convenient and intelligent.The main work and research results of this paper are as follows:(1)Collect and collate the relevant literature on power load forecasting at home and abroad,clarify the significance of power load forecasting research,understand the relevant research status,and analyze the internal variation law of power load data and possible external factors.(2)Statistical data of true power load and possible influencing factors for the whole year of 2018 in Spain,and correlation analysis between power load data and influencing factors data,to find out the influencing factors with large correlation coefficient,and with the power load data as input of the model in order to improve the accuracy of the model.(3)Build power load forecasting models using RNN and LSTM neural networks respectively,and design unified evaluation criteria.Train the models using Spanish power load data and factor data for the whole year of 2018.Use the trained models to predict power load data for the next week respectively.The comparison results show that LSTM neural networks have long-term memory for long-term data which can make the predictions more accurate.(4)By introducing the parameter optimization method,the LSTM model is optimized by using the particle swarm optimization algorithm.The original parameters set artificially according to experience are changed to the particle swarm algorithm to search for the optimal parameters automatically and iteratively.The power load data of the next week is also predicted by using the model optimized by the particle swarm algorithm.The results show that the optimized model is superior to the non-particle swarm algorithm.It has higher prediction accuracy before transformation.(5)Build an electric load forecasting platform.The front end uses C# language to design the interface,and SQL Server database is used to realize data management.The platform mainly consists of three modules,the data management module is mainly used to manage the power load data and related influencing factors data used in the experiment;the model training module mainly calls Python algorithm model file and trains the model with training data;the result display module is used to display the model prediction results and model accuracy in the form of tables and polylines.In this paper,from data,algorithm,optimization to platform design,a set of intelligent power load forecasting system is realized.The research results have a higher prediction accuracy,which has a strong practical value for the power management department to carry out local or large-scale short-term power load forecasting,power dispatching and other work.