基于真实同步波形测量数据的逆变型资源次周期动态数据驱动建模

As inverter-based resources (IBRs) become increasingly integrated into modern power systems, there is a growing need for adept modeling techniques that can accurately capture their dynamic response to disturbances. In this paper, three novel data-driven methods are proposed to model the dynamic response of IBRs during sub-cycle disturbances. We use real-world voltage and current waveform data from waveform measurement units (WMUs) installed at IBRs at a test site in California. The proposed methods are designed based on long short-term memory (LSTM) networks. They vary in terms of the architecture of the LSTM networks and the feature extraction characteristics. Furthermore, two methods are proposed to reuse or adjust a model from one IBR to capture the response of another IBR based on the analysis of time-synchronized waveform measurements at two IBRs. The results demonstrate that despite the model being constructed based on a different IBR, the proposed methods exhibit high accuracy. Experimental results demonstrate and validate the high accuracy of the models and provide a comparison with recent data-driven methods in the literature.