Parameter identification of photovoltaic models based on the landscape-aware online algorithm selection framework

Abstract A photovoltaic (PV) power generation system is a facility that converts solar energy into electrical energy, in which PV cells serve as key components whose control parameters are susceptible to environmental factors. Therefore, fast and accurate extraction of PV cell model parameters is crucial for the simulation, evaluation, and control of PV systems. In recent years, the use of metaheuristic optimization algorithms for parameter extraction of PV models has received significant attention. However, most existing methods are designed based on empirical knowledge with limited consideration of the characteristics of the problem itself, which often leads to issues such as high randomness and a tendency to fall into local optima during the search process. To address these problems, this paper proposes a Landscape-aware Online Algorithm Selection (LaOAS) framework for solving PV model parameter identification optimization problems. This framework integrates a landscape classification technique, an algorithm switching mechanism, and a dictionary-based context memory method. Using LSHADE-EpSin and LSHADE-cnEpSin as the baseline algorithms, we instantiate LaOAS to obtain the LaOAS-LSHADES algorithm. Comprehensive experimental results on a benchmark function set demonstrate that LaOAS-LSHADES achieves excellent performance across various test problems. Finally, the proposed method is applied to different types of PV model parameter identification optimization problems, further validating its robustness, efficiency, and scalability. The above research shows that LaOAS provides a flexible and efficient framework for solving PV model parameter identification problems.