Electrical Load Forecasting Using Time Series Analysis
Shilpa. G. N, G. S. Sheshadri
Abstract
This paper presents electrical load forecasting analysis and forecasted results based on identification of stochastic time series models for short term. Three predictive models namely, the autoregressive moving average (ARMA), autoregressive integrated moving average (ARIMA) and autoregressive integrated moving average model with exogenous variables (ARIMAX) are proposed. The mean absolute percentage errors (MAPE) of these models are computed and compared. Forecasting results show that ARIMA and ARIMAX Models performance is better ensured, thereby improving the forecasting accuracy significantly compared to ARMA Model. Further, it is shown that ARIMAX Model slightly outperforms ARIMA Model. The proposed methodology has been applied, on Karnataka State Demand Data-2019 for short term electrical demand prediction. This approach of time series modeling can accurately predict the practical power system hourly demand considering into account public holidays, weekdays and weekends.