Efficiency of Ensemble Learning Algorithms in the Analysis of Effects of Covid-19 Pandemic on Electricity Consumption in Turkey

Abstract

The COVID-19 pandemic associated with the lockdown measures caused an extraordinary impact on the economies of all countries in the world. Under lockdown, dramatic reductions in industry and services resulted in electricity demand dropping to Sunday levels, though higher domestic use yielded a relatively small partial offset. In this study, we analyzed overall electricity consumption in Turkey starting from pre-COVID days until now to illustrate the pandemic's effects on consumption. For this purpose, we built an ensemble machine learning model for the analysis. Findings revealed that the proposed boosting (AdaBoost) ensemble algorithm (RMSE: 41848.7, MAE: 18574.3, R2 :0.89) is a significant contributory factor in the analysis of data related to electricity consumption. Results also show that boosting (AdaBoost) ensemble learning algorithm is more preferable in the use of energy-related data than the bagging (random forest) and single-based algorithms (deep neural networks).