Open Access

ARTICLE

Transient Voltage Control for AC-DC Hybrid Power System Based on ISAO-CNN-BiGRU

Xueting Cheng¹, Rui Xu^2,*, Liming Bo¹, Cheng Liu², Huiping Zheng¹, Zhichong Cao²

1 State Grid Shanxi Electric Power Company Electric Power Science Research Institute, Taiyuan, 030001, China
2 Key Laboratory of Modern Power System Simulation and Control & Renewable Energy Technology, Ministry of Education (Northeast Electric Power University), Jilin, 132012, China

* Corresponding Author: Rui Xu. Email:

(This article belongs to the Special Issue: Advances in Renewable Energy Systems: Integrating Machine Learning for Enhanced Efficiency and Optimization)

Energy Engineering 2026, 123(4), 10 https://doi.org/10.32604/ee.2025.072350

Received 25 August 2025; Accepted 29 September 2025; Issue published 27 March 2026

Abstract

To address the issue of transient low-voltage instability in AC-DC hybrid power systems following large disturbances, conventional voltage assessment and control strategies typically adopt a sequential “assess-then-act” paradigm, which struggles to simultaneously meet the requirements for both high accuracy and rapid response. This paper proposes a transient voltage assessment and control method based on a hybrid neural network incorporated with an improved snow ablation optimization (ISAO) algorithm. The core innovation of the proposed method lies in constructing an intelligent “physics-informed and neural network-integrated” framework, which achieves the integration of stability assessment and control strategy generation. Firstly, to construct a highly correlated input set, response characteristics reflecting the system’s voltage stable/unstable states are screened. Simultaneously, the transient voltage severity index (TVSI) is introduced as a comprehensive metric to quantify the system’s post-disturbance transient voltage performance. Furthermore, the load bus voltage sensitivity index (LVSI) is defined as the ratio of the voltage change magnitude at a load node (or bus) to the change in the system-level TVSI, thereby pinpointing the response characteristics of critical load nodes. Secondly, both the transient voltage stability assessment result and its corresponding under-voltage load shedding (UVLS) control amount are jointly utilized as the outputs of the response-driven model. Subsequently, the snow ablation optimization (SAO) algorithm is enhanced using a good point set strategy and a Gaussian mutation strategy. This improved algorithm is then employed to optimize the key hyperparameters of the hybrid neural network. Finally, the superiority of the proposed method is validated on a modified CEPRI-36 system and an actual power grid case. Comparisons with various artificial intelligence methods demonstrate its significant advantages in model speed and accuracy. Additionally, when compared to traditional emergency control schemes and UVLS strategies, the proposed method exhibits exceptional rapidness and real-time capability in control decision-making.

---

Keywords

---