Power Demand Analysis And Mid-long Term Prediction Of Qingyuan City Based On ANN-ARIMA Model

Electricity is a kind of energy using electric energy as power,which has played an important part in socio-economic and social development.Unlike disposable energy such as coal and petroleum,electric,it is impossible to store the electric energy on a large scale due to its own characteristics.If the power supply is surplus,the electric energy will be wasted.And insufficient power supply will result in negative effect on the social production and people's lives.Thence,a reasonable analysis and precise forecast for electric energy demand.In the view of researching,he forecast for electric energy demand has two main kind,electricity consumption and electricity loading,in this paper we focus on the forecast for electricity consumption.In the view of time span,the forecast for electricity demand can be divided into short-term forecasts,medium-term forecasts and long-term forecasts.Short-term forecasts generally refers to forecasts on daily basis,medium-term forecasts generally refer to forecasts on a monthly basis,and long-term forecasts generally refer to forecasts on an annual basis.The forecasting methods used commonly are follows: load density method,elastic coefficient method,time series method,artificial intelligence method.Based on the features of mid-term and long-term electricity consumption in Qingyuan,this paper use time series method and artificial intelligence method respectively.In the medium-term power demand prediction,this paper forecast the electricity consumption of Qingyuan for the first half of 2020 using seasonal index method,ARIMA model and the improved ARIMA respectively due to monthly data is seasonal.It can be obvious that the improved ARIMA has a best performance in prediction,follow by multiplicative seasonal index method,the additive seasonal index method,and the ARIMA model.In the long-term power demand prediction,this paper forecast the annual electricity consumption of Qingyuan in 2018 and 2019 using the moving average method,exponential smoothing method,and the ANN model based on Adaboost.Compared to exponential smoothing method,improved ANN model shows better performance,and moving average method has a performance slightly inferior to exponential smoothing method.The main work of this article are as follows:(1)Aiming at the difficulty of forecasting mid-term electricity consumption data,we gave full play to the advantages of traditional time series forecasting models,combining additive seasonal analysis and ARIMA forecasting models.Based on these,more accurate forecasting results are obtained.(2)Aiming at the problem of insufficient long-term electricity consumption data and insufficient information,we have fully considered the impact of economic factors on long-term electricity consumption,and applied the enhanced ANN model based on Adaboost to the field of long-term electricity consumption forecasting,which demonstrated superiority forecast performance.