A Nonlinear L-SHADE Variant for Photovoltaic Parameter Estimation and Solar Power Generation Forecasting

Solar energy has attracted increasing attention due to its clean, renewable, and cost-effective characteristics. Among solar-related optimization tasks, photovoltaic (PV) model parameter estimation and solar power generation forecasting are two important problems for improving PV system modeling accuracy, performance evaluation, and operational management. However, both tasks involve nonlinear characteristics, coupled variables, and uncertainty caused by environmental variations, which make them challenging for conventional optimization methods. To address these issues, this paper proposes a nonlinear L-SHADE variant, termed NL-SHADE, in which a sigmoid-based nonlinear population reduction strategy is introduced to replace the linear population reduction mechanism in L-SHADE. The proposed design aims to provide a more flexible balance between exploration and exploitation while preserving the simplicity of the original framework. The performance of NL-SHADE is evaluated on six benchmark PV parameter estimation problems and 14 solar power generation forecasting datasets. For PV parameter estimation, NL-SHADE achieves highly competitive results, and the summarized Wilcoxon rank-sum results over the six benchmark problems confirm its overall advantages over most compared algorithms while maintaining strong competitiveness against DPDE and L-SHADE. For solar power generation forecasting, NL-SHADE ranks first among all compared methods according to the Friedman ranking analysis. Moreover, in 42 pairwise Wilcoxon comparisons against L-SHADE, NL-SHADE achieves 10 wins, 30 ties, and only 2 losses. These results indicate that the proposed nonlinear population reduction mechanism improves the search behavior of L-SHADE and enhances its robustness across different solar energy optimization tasks. Overall, NL-SHADE provides an effective and general optimization framework for both PV model parameter estimation and solar power generation forecasting, showing promising practical value for PV system analysis, forecasting, and operation-related decision support.