Backtracking search algorithm with competitive learning for identification of unknown parameters of photovoltaic systems

Metaheuristic algorithms have been successfully used to parameter identification of photovoltaic systems. However, this still faces the following two challenges. Firstly, most of the applied algorithms are complex and need some extra control parameters except the essential population size and stopping criterion, which is against their applications in photovoltaic systems with different characteristics. Secondly, how to obtain model parameters with higher accuracy and reliability has been a very valuable topic. To address the two challenges, this paper presents a new optimization method called backtracking search algorithm with competitive learning (CBSA) for parameter identification of photovoltaic systems. The remarkable features of CBSA are that it has a very simple structure and only needs the essential parameters. The core idea of CBSA is to increase the chance of backtracking search algorithm (BSA) to jump out of the local optimum by the designed competitive learning mechanism. In CBSA, the population is first divided into two subpopulations by built competitive mechanism. Then each subpopulation is optimized by the different search operators with multiple learning strategies. In order to test the performance of CBSA, CBSA is first employed to solve five challenging engineering design optimization problems and then is used to estimate the unknown parameters of three photovoltaic models. Experimental results show the solutions offered by CBSA outperform those of the compared algorithms including BSA, two recently proposed variants of BSA and other some state-of-the-art algorithms on nearly all test problems, which proves the effectiveness of the improved strategies. (c) 2020 Elsevier Ltd. All rights reserved.