Enhanced single-diode model parameter extraction method for photovoltaic cells and modules based on integrating genetic algorithm, particle swarm optimization, and comparative objective functions

Accurate modeling of the operational behavior of photovoltaic systems is crucial to optimizing and predicting system performance. One of the well-established and widely used modeling techniques is the single-diode equivalent circuit that delivers a sufficiently accurate description of the electric behavior of both photovoltaic cells and modules under various operational conditions. The single-diode model uses five parameters to reproduce the I-V curve for specific operational conditions. However, these five parameters must be extracted from measured or simulated I-V curves. This paper proposes a novel, accurate, and fast method for extracting the single-diode model's five parameters from measured I-V curves based on a genetic algorithm combined with particle swarm optimization to find the optimal controlling parameters of the genetic algorithm. This approach results in a significant performance improvement in accuracy and convergence speed. The paper also proposes a concept for determining the optimum number of current-voltage data points in the I-V curve, enabling an optimum trade-off between a sufficiently high accuracy and computational costs. Finally, the effect of different objective function formulations on the result has been investigated by comparing the usage of three different objective functions: the implicit form of the single-diode model, the Lambert W-function-based formulation of the explicit single-diode model, and a system of equations based on least square fitting. From the results, it could be concluded that the implicit formulation of the single-diode model delivered the best results compared to the two other formulations. Performance evaluations showed significantly lower error values than recent literature, with mean percent errors of 0.038%, 0.34%, and 0.87% received for the investigated monocrystalline cell, poly-crystalline module, and amorphous module, respectively. The computational cost was reduced by more than 60% after determining the optimum number of I-V points per curve, which was in the range of 20-30 points for each measured curve.