Load forecasting analysis for regional and industry power systems-based on ARIMA model and LSTM model

Abstract

The forecasting and analysis for the power system load profoundly affect the security of electricity for production and daily life. Based on the load data of 15- minute interval in a regional power grid in China, this paper forecasts the load of 15 - minute interval in the next 10 days, the maximum values of daily load in the next 3 months, and the maximum and minimum values of daily load in the next 3 months for each industry in the region on the basis of the ARIMA model and the LSTM model respectively, and then compares the forecasting effects of the two models. The conclusions show that (1) The LSTM model has a higher forecasting accuracy than that of the ARIMA model in general. (2) Daily electricity load levels are higher for large industry and commerce and lower for non-general industry and general industry. (3) The sudden changes of electricity load in each industry occur mainly on holidays and days with rainy weather. And the magnitude of the sudden change in the load of large industries is the largest. Finally, innovative suggestions are made for different industries in the context of the “Double Carbon Plan”.

Keywords

• load forecasting
• load abrupt change
• arima model
• lstm model