Learning search algorithm to solve real-world optimization problems and parameter extract of photovoltaic models

Solar energy is widely acknowledged as a promising and abundant source of clean electricity. Unfortunately, the efficiency of converting solar energy into electricity using photovoltaic (PV) systems is not yet satisfactory due to technical limitations. To improve this, it is essential to develop an accurate model that incorporates well-estimated parameters. However, the parameter identification process in the PV model is challenging due to its nonlinear and multi-modal characteristics. In this study, we propose a novel metaheuristic algorithm called the learning search algorithm (LSA) to address the parameter estimation problem in solar PV models. LSA utilizes historical experience and social information to guide the search process, thus enhancing global exploitation capability. Additionally, it improves the learning ability of the population through teaching and active learning activities based on optimal individuals, which enhances local development capability. The algorithm also incorporates a dynamic self-adaptive control factor to balance global exploration and local development capabilities. Experimental results demonstrate that our proposed LSA outperforms other comparison algorithms in terms of accuracy, convergence rate, and stability in parameter identification of PV models. Statistical tests confirm the superior efficiency and effectiveness of the LSA in parameter estimation. Moreover, our algorithm demonstrates competitive performance in solving real-world optimization problems with constraints. Overall, our study contributes to the improvement of solar energy conversion efficiency through the development of an accurate parameter estimation model using the LSA.