Study On Chengdu Residents’ Electricity Consumption Forecast Based On Machine Learning

As one of the important energy consumption,electric power is an important support to protect social activities.However,with the development of human society,the problem of energy shortage has become increasingly prominent.Therefore,the management and optimization of the power system has been included in the important national strategic measures,which requires the reasonable planning of the regional distribution network to further ensure the reliability of the end-user electricity consumption and reduce the per capita outage time.The scientific and reasonable prediction of the medium and long-term electricity consumption can provide data reference for the distribution planning of distribution network,and provide data support for China’s Demand Side Management.Sichuan Province,as a large power production and output Province in China,has more than 80%of hydropower generation.Therefore,this essay takes the residents’ monthly electricity consumption of Chengdu as the research object.Based on the monthly electricity consumption data from 2010 to 2018,three different methods are used to predict the medium and long-term electricity consumption,and explore the optimal model for medium and long-term electricity consumption.Through literature review,this essay first studies and summarizes the existing electricity consumption prediction method,summarizes the characteristics and classification of electricity consumption prediction,defines the concept of medium and long-term electricity consumption prediction,and determines the main influence factors of residential electricity consumption,and then collects corresponding data.Secondly,five factors with the highest correlation degree with monthly electricity consumption are determined by grey correlation analysis method,and then included in grey prediction model,BP neural network model and LSTM neural network model.Through comparative analysis of these three methods,the most suitable model for medium and long-term power forecasting is found.Thirdly,the optimal model is selected to forecast the monthly electricity consumption of Chengdu in 2019,and the forecast results are analyzed by descriptive statistics.Finally,this essay studies the characteristics of power system in Sichuan Province from both sides of supply and demand,and puts forward some suggestions on Demand Side Management of Sichuan Province Based on the forecast data.