Parameter estimation of photovoltaic (PV) models is crucial for the theoretical analysis and performance evaluation of PV cells and modules, with the objective of enhancing their efficiency and reliability, thereby supporting the long-term sustainability of solar energy systems. Nevertheless, the nonlinear and multimodal characteristics of PV models make the task of accurate parameter estimation challenging. This paper proposes an improved differential evolution algorithm, named opposition-based parent selection differential evolution (OBPSDE), to enhance the reliability and robustness of PV parameter estimation. The method integrates a parent-selection mechanism with an opposition-based learning strategy to exploit both solution quality and population diversity during the search process. The proposed method is evaluated using measured data from several PV cells and modules (RTC France, PVM752GaAs, PWP201, and STP6-120/36) for parameter estimation of the double-diode model (DDM). Its performance is compared with standard DE, DE variants, and four metaheuristic algorithms using statistical metrics including root mean square error (RMSE), individual absolute error (IAE), and mean absolute error (MAE). The results indicate that OBPSDE achieves stable performance, competitive computational cost, and improved convergence behavior, with RMSE values of 6.93726 × 10^–4 for RTC France, 5.89070 × 10^–5 for PVM752GaAs, 1.93772 × 10^–3 for PWP201, and 1.39519 × 10^–2 for STP6-120/36. Additionally, the improved parameter estimation accuracy may support more reliable performance prediction and analysis of PV systems, contributing to effective PV system modeling and diagnostic applications.