Comparative analysis of deep learning techniques based COVID-19 impact assessment on electricity consumption in distribution network

Abstract

Energy is a fundamental human need for several activities. Energy can be impacted by several factors ranging from technical to social and environmental. The impact of COVID-19 outbreak on the energy sector is enormous with serious global socioeconomic disruptions affecting all economic sectors, including tourism, industry, higher education, and the electricity industry. Based on the unstructured data obtained from Eko Electricity Distribution Company this paper proposes three deep learning (DL) models namely: Long Short-Term Memory (LSTM), Simple Recurrent Neural Network (SimpleRNN), and Gated Recurrent Unit (GRU) were used to analyse the effect of COVID-19 pandemic on energy consumption and predict future energy consumption in various district in Lagos, Nigeria. The models were evaluated using the following performance metrics namely: Mean Absolute Percentage Error (MAPE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). On overall, the lowest MAPE, MAE, RMSE, and MSE of 0.120, 71.073, 93.981, and 8832.466 were obtained for LSTM in Orile, SRNN in Ijora, and GRU in Ijora, respectively. Generally, the GRU performed better in predicting energy consumption in most of the districts of the case study than the LSTM and SimpleRNN. Hence, GRU model can be considered the optimal model for energy consumption prediction in the case study. The importance of having this model is that it can help the government and other stakeholders in economic planning of electricity distribution networks.