Forecasting of Türkiye's net electricity consumption with metaheuristic algorithms

Abstract

This study advances the literature by integrating and benchmarking five state-of-the-art metaheuristic algorithms to forecast Türkiye's net electricity demand using linear and exponential models: artificial ecosystem-based optimization (AEO), grey wolf optimizer (GWO), particle swarm optimization (PSO), artificial bee colony (ABC), and Harris Hawks optimization (HHO). While metaheuristic optimization methods have been utilized in energy forecasting, this study distinguishes itself by employing the novel AEO algorithm, which has demonstrated superior performance to traditional methods in similar domains, thereby contributing a fresh perspective to electricity demand forecasting. All algorithms were trained using data from 1980 to 2009, incorporating population, gross domestic product (GDP), installed power, and gross generation variables, and tested with data from 2010 to 2019. Statistical metrics (R², MAPE, MBE, rRMSE, and MAE) were used to evaluate algorithm performance. This study projects an annual growth rate in net electricity consumption ranging from 2.14 % to 2.59 %, with cumulative increases by 2050 ranging from 92.63 % to 120.75 %. These findings underscore the importance of proactive energy investment planning to mitigate potential economic challenges arising from significant increases in electricity consumption.