Optimal Sizing Techniques for Hybrid Photovoltaic Systems Using Artificial Neural Networks (ANNs): A Review Paper

The sizing process in hybrid photovoltaic (PV) systems’ design is critical to ensure proper operation, reasonable costing, and meeting the load demand. Moreover, the system configuration is dependent on the load demand and type, storage requirements, and other factors. The complexity of these systems requires advanced techniques that can yield the optimal size and configuration. The aim for the sizing is to satisfy the load demand and minimize the use of conventional sources such as diesel generators or grid-supply. Various artificial neural network (ANN)-based techniques and predictive algorithms have been employed in research studies such as clonal selection algorithm, genetic algorithm, particle swarm optimization, tabu search algorithm, and multi-objective self-adaptive differential evolution. In this paper, a review is carried out to evaluate the current sizing techniques of hybrid PV systems in remote areas. The review focuses on two components: (1) the sizing techniques and (2) the implementation of ANN in the process. The general trend for the process includes data collection, input parameter identification, and model training and testing. However, the system design process is dependent on the load demand of the location and resource availability such as solar irradiance and wind speed. Moreover, other aspects were considered such as economic viability and environmental impact. The available resources were estimated differently in the literature; in some studies, historical data and weather patterns are based on metrological data and in other studies, it is based on satellite imported data. The incorporation of energy storage systems such as batteries and thermal storage units was also considered. The optimization algorithms were found to effectively help determine the optimal sizing of the hybrid system to achieve maximum efficiency and minimal costs. These models were found to improve the accuracy, reduce calculation time, and improve the cost-effectiveness of the systems.