Improving Parameter Extraction in Photovoltaic Models: The Role of Initialization Methods in Particle Swarm

Abstract

This study investigates the effect of initialization strategies on the performance of Particle Swarm Optimization (PSO) for parameter extraction in photovoltaic (PV) models, specifically the Single Diode Model (SDM) and the Double Diode Model (DDM). Two initialization methods, Uniform Random Sampling Initialization (URSI) and Latin Hypercube Sampling (LHS), were compared to evaluate their impact on accuracy, stability, and computational efficiency. For the SDM, LHS reduced the mean RMSE from 1.7798×10⁻³ to 1.7127×10⁻³ (a 3.8% decrease) and the standard deviation by 19.7%, while maintaining a comparable computational time of 0.3988 s compared to 0.3948 s. In the DDM, LHS achieved a mean RMSE of 7.9489×10⁻⁴, representing a 2.3% reduction relative to 8.1348×10⁻⁴, and decreased the standard deviation by 50.4% from 1.2176×10⁻⁴ to 6.0390×10⁻⁵, with nearly identical execution times. Overall, the results indicate that LHS significantly enhances the reliability and robustness of PSO by improving convergence stability and parameter accuracy under various operating conditions. These findings highlight the critical role of efficient initialization strategies in metaheuristic optimization for accurate and consistent PV system modelling.