@article{PENG2026, 
author = {Tian PENG and Hao NI and Xinyu ZHANG and Qianlong LIU and Chu ZHANG},
title = {Virtual simulation-based experimental teaching design for photovoltaic model parameter identification in new power systems},
year = {2026},
journal = {Experimental Technology and Management},
volume = {43},
number = {7},
pages = {147-156},
keywords = {photovoltaic cell model, fitness–distance balance mechanism, uniform initialization strategy, instructional design},
url = {https://www.sciopen.com/article/10.16791/j.cnki.sjg.2026.07.017},
doi = {10.16791/j.cnki.sjg.2026.07.017},
abstract = {ObjectiveAs new power systems characterized by high renewable energy penetration continue to evolve, accurately and efficiently identifying photovoltaic (PV) model parameters has become critical for performance prediction, optimal operation, and intelligent maintenance of PV power plants. Classical models, including the single-diode model (SDM), double-diode model (DDM), and module-level models, contain multiple nonlinear parameters that resist precise estimation. Conventional parameter identification methods frequently exhibit slow convergence, premature stagnation, and poor robustness. Meanwhile, in engineering education, the inherent complexity of nonlinear modeling and metaheuristic optimization poses considerable challenges for students seeking to master the full identification workflow. Against this backdrop, the present study proposes an improved parameter identification algorithm offering enhanced accuracy and stability, alongside a modular virtual simulation-based experimental teaching platform that bridges algorithm research and educational practice in the context of new power systems.MethodsAn improved supply–demand optimization algorithm that incorporates a fitness–distance balance (FDB) mechanism and a half-uniform initialization strategy, designated FDB-SDO, is developed. The FDB strategy assesses candidate solutions by jointly weighing normalized fitness values and distance information relative to the current best solution, thereby maintaining an effective balance between global exploration and local exploitation. The half-uniform initialization strategy broadens population diversity during the early iterations, mitigating the excessive aggregation that purely random initialization can produce. Mutation reinforcement and opposition-based learning are additionally embedded to counteract premature convergence and bolster global search capability. FDB-SDO is applied to identify the parameters of SDM, DDM, and two representative PV module models. Root mean square error (RMSE) serves as the objective function, quantifying the fitting accuracy between simulated and experimentally measured I–U data. Performance is verified through comparisons with several well-established metaheuristic algorithms under identical parameter settings. Concurrently, a modular virtual simulation-based experimental teaching platform is constructed that decomposes the identification process into discrete modules spanning model cognition, objective function construction, algorithm implementation, multi-strategy improvement, convergence analysis, and result evaluation.ResultsExperimental results demonstrate that FDB-SDO attains competitive or superior RMSE values while requiring fewer objective function evaluations than the comparison algorithms. For both SDM and DDM, the proposed method produces lower or equivalent fitting errors at reduced computational cost, reflecting improved efficiency. In the more demanding DDM scenario involving seven parameters, FDB-SDO exhibits stronger robustness and more precise estimation of diode-related parameters. For module-level models, the algorithm sustains stable convergence and consistent identification accuracy. Statistical analysis across multiple independent runs yields smaller standard deviations, confirming enhanced stability. The I–U and P–U curves reconstructed from the identified parameters closely reproduce the experimental measurements, validating the reliability of the proposed approach. From an educational standpoint, the modular platform allows students to visualize convergence trajectories, benchmark algorithm performance, and examine how individual strategies affect identification accuracy.ConclusionsFDB-SDO effectively accelerates convergence, elevates identification accuracy, and strengthens robustness in PV model parameter extraction. Through the integration of fitness–distance guidance and diversity-preserving initialization, the algorithm mitigates premature convergence and reinforces global search capability. The accompanying modular virtual simulation teaching design converts a complex optimization problem into structured, progressive learning tasks that connect theoretical modeling with engineering application. This framework offers a scalable instructional paradigm for nurturing innovative talent in renewable energy engineering within the context of new power systems.}
}