Geo-Temporal Analysis of Electricity Consumption Growth in Togo using Deep Learning

Abstract

Electrical energy plays a vital role in daily life. The developing world is making good progress in improving access to electricity toward UN sustainable Goal number 7 through better energy planning. Understanding energy consumption growth remains a fundamental aspect of energy planning. This study aimed to investigate the dynamic of electricity consumption growth in Togo and to build the most suitable deep learning model for predicting the yearly electricity consumption of the country's different regions. The analysis uses a data-driven approach to determining the relationship between average consumption and electrification rate. On the other hand, yearly electricity forecasting was modeled using multiple layer perceptron (MLP), convolutional neural network (CNN), and long short-term memory (LSTM) models, and evaluated using the mean absolute error metric. As a result of this research, the average consumption decreases as the electrification rate increases. It was also found that the multivariate and multi-steps LSTM model is the most suitable model for predicting yearly electricity consumption. However, the model accuracy can be increased when combining it with the CNN model. The former shows that low consumers can be aggregated on the same transformer because the average consumption is decreasing. The latter revealed the potential of deep learning techniques in predicting yearly residential electricity consumption. The CEET can use the LSTM model to right-size the residential electricity supply.