关键词: 光伏阵列, 数学建模, 参数辨识, 特征提取, 健康评估, 故障诊断

Abstract:

As an important part of a PV generation system，the healthy operation and maintenance of PV arrays has become increasingly critical with the worsening shortage of fossil energy. The research statuses of modeling methods，parameter identification，feature extraction，and health evaluation of PV arrays are systematically reviewed. Three modeling approaches are compared： equivalent circuit-based models offer mechanistic interpretability but suffer from low accuracy； numerical simulation-based models are suitable for complex operating conditions and offer relatively high precision； and neural network-based models，as black-box models，make fault mechanism interpretation difficult. Parameter identification techniques are categorized into analytical methods and intelligent optimization algorithms. The former lacks accuracy in model parameter identification，and the latter uses analytical initial values to constrain the search space to achieve optimization with metaheuristic algorithms. The proposed hybrid method optimizes the iteration process using analytical initial values，balancing computational speed and accuracy. For feature extraction，the representational differences among statistical features，signal decomposition，and deep learning are analyzed，highlighting the role of I-V curve standardization in suppressing environmental noise. To clarify the logical relationship between health assessment and fault diagnosis，based on the PV performance ratio and a five-level health status classification system，an efficient and accurate strategy for PV arrays' status monitoring and intelligent operation and maintenance is formulated. The established technical framework integrates "mechanism modeling → parameter identification→feature extraction→status evaluation"，providing comprehensive methodological support for intelligent operation and maintenance of PV arrays.

Key words: PV array, mathematical modeling, parameter identification, feature extraction, health evaluation, fault diagnosis