Research On Intelligent Forecasting Method Of Gas Load Based On LSTM

Gas load prediction is an important basis for the formulation of gas pipe network planning,which is of great significance for improving the efficiency of gas enterprises,maintaining the safe and stable operation of the system,rationally dispatching gas and ensuring urban life.Due to complex factors affecting gas load,inaccurate data,technical defects,improper data processing and other reasons,the accuracy and reliability of load prediction is not strong,therefore,the study of intelligent gas load prediction method has gradually become the focus of enterprise researchers.Through the investigation of a gas group,it is found that the main method of load forecasting at present is to obtain the required gas load in the future by referring to the gas load data of the same period or recent period in history and adjusting it in combination with manual experience,but the prediction accuracy is always low.Therefore,it has become a crucial task to establish intelligent prediction model of gas load based on data mining,design intelligent prediction system of gas load,and timely and accurately predict daily load data in the future period.In order to solve the above problems,based on the actual requirements of load data forecasting business,this thesis designs an intelligent forecasting model of gas load based on data mining,and develops and implements an intelligent forecasting system.Firstly,the business requirements of gas load prediction are fully studied,the load data sources are given,key prediction technologies are studied,and the overall framework of intelligent load prediction model based on data mining is designed.The key content of the research is determined to be the characteristic reduction of load impact factors and the multi-intelligent comprehensive prediction method of gas load.Secondly,the load prediction data model is designed from the overall and detailed structure,and the characteristics and impact factors of the obtained gas load data are analyzed.Besides,data preprocessing such as missing value,outlier value,factor quantization and normalization are completed,so as to provide high-quality complete data for the research of load data prediction methods.Thirdly,the feature dimension reduction method was studied,and PCA extraction algorithm was selected to extract the principal components of the impact factors,reduce the complexity of data,and improve the computational efficiency and result accuracy of the prediction algorithm.Then,a gas load prediction method based on multi-intelligent comprehensive algorithm is proposed.The algorithm set is selected according to the characteristics of the load data,based on LSTM network algorithm,BP neural network,support vector machine and grey theory algorithm are used to complete the load data prediction,and the multi-intelligent comprehensive prediction method is obtained by combining four single algorithms.According to the results of each algorithm,the results of multi-intelligence synthesis are predicted.Finally,under the background of load data forecasting business,the intelligent forecasting system of gas load is designed and realized combined with the forecasting model,and the practical value of the method proposed in this thesis is tested through practice.The research results show that the intelligent gas load forecasting method can predict the gas consumption in the specified time period with low error,effectively solve the problem of load data forecasting,improve the work efficiency,reduce the negative impact caused by the subjective misjudgment of decision makers,such as supply and marketing imbalance,gas storage shortage and so on,and has a high guiding significance.