Machine learning-assisted innovative charging strategy for e-mobility in rural communities operated by redundant energy on solar PV mini-grids

Green transportation using solar energy with nearly zero emissions is of global importance to address the challenges of modern energy access for the transport sector, greenhouse gas emissions and global warming. In the Global South and in most off-grid areas, solar PV mini-grids are being used to provide energy access. However, there is redundant energy from these mini-grid systems during peak sunshine hours, which could be used for further profitable activities. E-mobility is a key use case that could be incorporated into the operation of mini-grids to minimise redundant energy, improve system performance, and increase mini-grid profitability. In this study, a model of a Machine Learning (ML)-based control system incorporating Internet of Things (IoT) for e-tricycle charging is proposed to optimise the use of energy from mini-grids for green transportation. In a case study, three ML models, namely Artificial Neural Network, Extreme Gradient Boosting, K-Nearest Neighbour and Random Forest, were trained on data acquired from three mini-grids to predict redundant energy for efficient electric vehicle (EV) charging. The results revealed that on average, the three communities had redundant energy in the ranges of 56.98–119.86 kWh, 74.39–311.87 kWh, and 57.03–274.66 kWh per day. Having validated the ML models, all the models could predict redundant energy successfully.