Parameter identification of the photovoltaic (PV) model is essential to research in the PV field. A two-stage method of model parameter identification based on maximum power matching (MPM) and improved flow direction algorithm (IFDA) is proposed. The two-stage method, i.e., MPM-based rough extraction and IFDA-based precise identification. The quality of I–V data dramatically impacts the PV model’s parameter identification accuracy. At first, the measured I–V curves are preprocessed. The process includes outlier removal, curve fitting, and sparsification of I–V data. Then the MPM is used for the rough extraction of model parameters from the preprocessed I–V data. Finally, the rough extraction results are used as the initial values of the iterations in the precise identification using IFDA. The root mean square error (RMSE) between the measured and calculated currents is used as the fitness function of IFDA. The experimental section compares sixteen methods. IFDA presents the highest accuracy with the smallest RMSE at 0.0024 A. Five methods with outstanding performance are selected for 1000 repetitive experiments. The IFDA is the most stable with RMSE in the range of 0.002 A to 0.003 A.

光伏（PV）模型的参数识别对于光伏领域的研究至关重要。提出了一种基于最大功率匹配（MPM）和改进流向算法（IFDA）的两阶段模型参数辨识方法。两阶段方法，即基于MPM的粗提取和基于IFDA的精确识别。I–V 数据的质量极大地影响了 PV 模型的参数识别精度。首先，对测得的 I-V 曲线进行预处理。该过程包括异常值去除、曲线拟合和 I-V 数据的稀疏化。然后 MPM 用于从预处理的 I-V 数据中粗略提取模型参数。最后，粗略的提取结果被用作迭代的初始值，用于使用IFDA进行精确识别。测量电流和计算电流之间的均方根误差 (RMSE) 用作 IFDA 的适应度函数。实验部分比较了十六种方法。IFDA 给出了最高的精度，最小的 RMSE 为 0.0024 A。选择了 5 种性能突出的方法进行 1000 次重复实验。IFDA 最稳定，RMSE 在 0.002 A 到 0.003 A 范围内。