An efficient hybrid algorithm based on particle swarm optimisation and teaching‐learning‐based optimisation for parameter estimation of photovoltaic models

Abstract In recent years, many meta‐heuristic algorithms have been investigated to estimate the parameters of photovoltaic (PV) models. However, the accuracy of the estimated parameters still needs to be concerned, especially for some complex PV models with many unknown parameters. In order to estimate the unknown parameters of the PV models more precisely and reliably, an efficient hybrid algorithm based on particle swarm optimisation and teaching‐learning‐based optimisation (PSOTLBO) is proposed in this paper. In PSOTLBO, inspired by the learner phase of teaching‐learning‐based optimisation (TLBO), an improved learner phase is designed and introduced into the basic PSO to enhance the global search ability and the ability to get rid of local optimum. The improved learner phase divides the population into four groups according to three values, which are the average fitness values of the overall population, the population in the first half of the fitness ranking and the population in the second half of the fitness ranking. Typically, each group has its particular movement pattern concentrating on exploration or exploitation respectively to improve the search efficiency of the algorithm. Furthermore, to deal with individuals beyond the boundary, a new designed probabilistic rebound strategy is introduced, which increases the diversity of population and avoids population aggregation at the search boundary. Then, the proposed PSOTLBO is applied to estimate the parameters of the single diode model, double diode model and PV module model. The comparative results between PSOTLBO and other 14 advanced algorithms show that the average root mean square error values of different PV models obtained by PSOTLBO are 9.86021878E−04, 9.82630511E−04, 2.42507487E−03, 1.72981371E−03, and 1.66006031E−02, respectively, which indicate that PSOTLBO can provide more accurate and stable parameter estimation results than other compared algorithms. Furthermore, the convergence experimental results demonstrate that PSOTLBO has outstanding performance in convergence speed and stability.