CBAM-CNN Based Transient Overvoltage Preventive Control Considering Piecewise Linear Control Sensitivity
Zhe Zhang, Boyu Qin, Tao Ding, Xin Gao, Yixing Zhang
Abstract
The occurrence of commutation failure (CF) faults in line-commutated converter-based HVDC systems may cause transient overvoltage issues, leading to tripping accidents in large-scale renewable energy plants. To ensure the voltage safety and stability under CF faults, this paper proposes a transient overvoltage preventive control strategy based on convolutional block attention module (CBAM) and convolutional neural network (CNN). First, the CBAM and CNN are integrated to predict the overvoltage peak values of AC buses in the DC vicinity, exploring key response characteristics related to voltage dynamic process. Second, based on the CBAM-CNN model, the control sensitivity index with piecewise linear representation is defined to assess the suppression effect of various control measures on transient overvoltage levels. Moreover, considering the regulating characteristics of controllable resources, the optimization control problem containing actual operational constraints of the power grid is established, and the optimal preventive control strategy is formulated to guarantee the safety and stability of power systems. Finally, the proposed overvoltage prediction model and piecewise linear control sensitivity index are verified under typical fault scenarios in a simplified simulation system of local area power grids, which provide effective guidance for the determination of transient overvoltage preventive control strategy.