Parameter extraction of solar photovoltaic models via quadratic interpolation learning differential evolution

Abstract

The parameter extraction problem of solar photovoltaic (PV) models is a highly nonlinear multimodal optimization problem. In this paper, quadratic interpolation learning differential evolution (QILDE) is proposed to solve it. Differential evolution (DE) is a preeminent metaheuristic algorithm with good exploration. However, its exploitation is poor, resulting in low searching precision when applied to the problem. To overcome this deficiency, in QILDE, quadratic interpolation (QI) is embedded in the crossover operation of DE to construct a QI learning-backup crossover operation to enhance the performance of DE. The mutation scheme of DE is primarily responsible for exploring the new search space while QI is mainly in charge of exploiting the local solution space around the best individual, which, therefore, can achieve a good trade-off between exploitation and exploration. QILDE is applied to six different PV cases. The experimental results demonstrate that QI coupled with the mutation scheme DE/best/2 can obtain superior results in solving the parameter extraction problem of PV models. Besides, compared with other advanced algorithms, QILDE shows highly competitive performance in terms of solution quality, extraction accuracy, robust stability, convergence property, computational time, and statistical significance. In addition, the current–voltage characteristics provided by QILDE agree well with the measured data for different PV models under different operating conditions.