Long-term Electric Peak Load Forecasting of Tehran Regional Electric Company using a Combinatorial Artificial Intelligence Approach

Forecasting the long-term electrical yearly peak load is pivotal in power system expansion planning. The accuracy of the forecasting method holds immense significance in preempting economic losses and budgetary issues arising from unwarranted or inadequate investments. Although conventional techniques like time-series methodologies such as Auto-Regressive Integrated Moving Average (ARIMA) are extensively employed for long-term electrical peak load and energy demand forecasts, their limitations in dealing with inefficiencies, nonlinearity, and seasonality trends present considerable challenges. This paper proposes a novel approach that leverages the ARIMA method, incorporating Support Vector Regression (SVR) and the Genetic Algorithm (GA) technique. This approach aims to forecast the long-term yearly peak load of the Tehran Regional Electric Company (TREC), Iran’s largest regional electric company. The SVR algorithm parameters are fine-tuned to minimize forecasting errors using a combination of GA and the ARIMA method. The resulting optimized forecasting approach, ARIMA-GA-SVR, is applied in a real-life case study network within TREC. Comparative analysis with existing forecasting methods is conducted. The ARIMA-GA-SVR approach is a reliable and accurate forecasting solution based on established error criteria and simulation outcomes.