Two-Stage Parameter Identification Method for Electromagnetic Transient Simulation Models of Grid-Connected Photovoltaic Systems
Parameter identification for electromagnetic transient (EMT) models of grid-connected photovoltaic (PV) systems suffers from weak identifiability of controller parameters when environmental, electrical, and controller parameters are optimized simultaneously. This paper proposes a two-stage identification framework that partitions parameters by physical meaning into an environmental/electrical set and a controller set. For the environmental/electrical set, a Sobol global sensitivity analysis based on variance decomposition screens key parameters. For the controller set, a dynamic response feature clustering method combined with an unsupervised screening strategy using an inter-cluster mean difference index reduces the parameter space. Differentiated fitness functions are constructed for each stage, and an improved quantum dung beetle optimization (IQDBO) algorithm incorporating quantum angle encoding and a stagnation perturbation mechanism performs the identification sequentially. Case studies demonstrate that the proposed method compresses the search space and improves controller parameter identifiability. Compared with particle swarm optimization (PSO) and grey wolf optimizer (GWO), the IQDBO-based method achieves superior identification accuracy and convergence stability. Environmental and electrical parameter identification errors remain below 1%, while controller parameter errors remain below 3%. The framework addresses the weak identifiability bottleneck in unified optimization and provides a practical pathway for EMT model calibration in PV grid-connected systems. Future work will extend the method to multiple operating conditions and noisy field data, and develop accelerated computation strategies.