Performance prediction and optimization of a hybrid renewable-energy-based multigeneration system using machine learning

This paper proposes an innovative multi-generation hybrid renewable energy system (HRES). The proposed HRES is optimized using the recently developed equilibrium optimizer algorithm and the efficacy of this algorithm is investigated by comparing the optimized sizing with results obtained from more traditional optimization algorithms, i.e., gray wolf optimizer, lightning search algorithm, and artificial bee colony. The results indicate that using the equilibrium optimizer algorithm outperforms conventional optimization algorithms in minimizing the levelized cost of electricity to $0.83 per kWh. This proves applicability of this modern optimization algorithm in improving HRES, which is a novelty of this research. Moreover, machine learning is implemented to predict the exergy efficiency of the proposed HRES, which provides insight into the performance and sustainability of the system. A thermodynamic database is established to train the machine learning model using secondary data. Using the multi-layer perceptron class, the exergy efficiency of the system is predicted; the trained model is able to deliver a prediction with an R-Squared value of 0.98. Thus, the appropriateness of machine learning algorithms in predicting the performance of HRES is verified. This research will, therefore, pave the way towards expanding real-world application of HRES's. This study provides insights into the optimization and economic and environmental benefits of hybrid renewable energy systems and inform policymakers on how to incentivize their implementation. As the study is performed for a specific location, exploring the feasibility of implementing the proposed system in other locations is recommended for future research.