Medium-term load forecasting and error distribution for power system flexibility analysis

Abstract

Load forecasting is used in many arenas of power system planning and analysis including expansion planning, load switching planning, and power system flexibility analysis etc. Medium-term load forecasting is an important category of electric load forecasting that covers a time span of up to one year ahead. It suits outage and maintenance planning, load switching operation as well as power system flexibility studies. In this analysis four forecasting models were investigated in to study the feasibility of using them in the power system flexibility analysis. The forecasting methods studied in this research are namely, the multiple linear regression, the nonlinear regression (gaussian process), the auto regressive integrated moving average (ARIMA) model and the neural network. In order to analyse the flexibility of the power system, the feasibility of using those models were checked using the mean absolute percentage error (MAPE). The artificial neural network model was the most feasible method which had the low MAPE value to medium-term load forecasting among the models which were tested.