Research on Data-Driven Fault Location for Inverter Based Resources Interfaced Power Systems

发布时间:2024-10-28浏览次数:10

Speaker:  Xinchen Zou

Time:      14:00Oct. 31

LocationSIST 1A200

Host:      Yu Liu

Abstract:

Inverter based resources (IBRs) are widely adopted to integrate large scale renewables into modern power systems. However, due to the influence of inverter controls, the characteristics of IBRs during transients is quite different from those from synchronous generators (SGs). Therefore, data-driven methods are applied to transients related scenarios in power systems, such as fault location, which have the potential to extract features within transient waveforms and to effectively fit the mapping from the input waveforms to the output targets.

In fact, data-driven fault location methods encounter difficulties. Existing data-driven approaches need plenty of high quality fault data which are quite limited in practical lines. This talk introduces a data-driven transmission line fault location method using small dataset. Transfer learning is applied to modify the pre-trained weights of the neural network using a small dataset, including steps such as freeze-training and fine-tuning to get an accurate result.

In the meanwhile, when analyzing IBRs’ performances during power system transients, the control strategies of IBRs are typically required to enable various applications However, in practice, control strategies are often encapsulated as black boxes by manufacturers. In this talk, a physics-informed data-driven classification method is introduced to identify the control strategy of IBRs during transients. The proposed method only requires single-end voltage and current measurements at the IBR side during transients such as faults.

The accurate analytical fault analysis model of IBRs is an important basis for fault location of power system. However, factors of dynamic characteristics of the control system make the accurate analytical models hardly be guaranteed. To address this issue, this talk gives an data-driven resource modelling method utilizing Bi-LSTM network to fit the complex relationship between voltage and current during fault .

Bio:

Mr. Xinchen Zou joined PSPAL in September 2022. He is currently a master student in PSPAL (starting from September 2022). He received the B.S. degree of Computer Science from Harbin Institute of Technology, in Summer 2022. His research interests include data driven fault diagnosis of power system and power electronic systems.