Abstract
The demand for electricity in most industrial and commercial cities of Zimbabwe is on the rise due to both industrial expansion and high rate of urbanisation. Sole reliance on fossil fuels for electricity generation is no longer sustainable, given the associated carbon dioxide emissions that pollute the eco-system, and thus contributing to global warming. To address this, the concept of distributed generation introduces Renewable Energy (RE)-based Microgrids (MGs) as a potential solution to meet the growing electricity needs. However, the inherent unpredictability of RE resources poses a significant challenge, potentially leading to interruptions or unreliable power supply. This research proposes an innovative solution by incorporating Internet of Things (IoT) devices into the MG system. The MG setup includes a solar farm, wind farm, battery storage system (BSS), diesel generator, and residential load. The key contribution lies in the implementation of an optimal scheduling algorithm aimed at enhancing the efficiency of MG systems. Furthermore, a comparative analysis of three machine learning algorithms—Artificial Neural Network, Random Forest, and Extreme Gradient Boosting—was conducted in order to identify the most accurate predictor for determining the optimal power sources to meet load demand at any given time. The findings highlight a marked improvement in electricity availability through the integration of IoTs in the microgrid. In terms of predictive accuracy, the Extreme Gradient Boosting machine learning algorithm demonstrated superior performance compared to the Artificial Neural Network and Random Forest algorithms. This innovative approach not only addresses the current electricity demands but also sets the stage for sustainable and efficient power generation in commercial cities of Zimbabwe, paving the way for a greener and more reliable energy future.