Open Access

ARTICLE

Consider the Transient Stability Multi-Classification Evaluation Model for the Grid Connection of New Energy

Nan Li^1,2,*, Yinan Wang², Liang Huang³, Yabin Zhu⁴, Guangyao Zhang⁵

1 Key Laboratory of Modern Power System Simulation and Control & Renewable Energy Technology, Ministry of Education, Northeast Electric Power University, Jilin, 132012, China
2 School of Electrical Engineering, Northeast Electric Power University, Jilin, 132012, China
3 State Grid Jilin Electric Power Co., Ltd., Siping Power Supply Company, Siping, 136000, China
4 Xianyang Power Supply Company, State Grid Shaanxi Electric Power Company, Xianyang, 712000, China
5 Extra High Voltage Company, State Grid Shandong Electric Power Company, Jinan, 250000, China

* Corresponding Author: Nan Li. Email:

Energy Engineering 2026, 123(11), 19 https://doi.org/10.32604/ee.2025.073712

Received 24 September 2025; Accepted 24 December 2025; Issue published 24 September 2026

Abstract

The large-scale integration of new energy into power systems significantly elevates the risk of instability. To achieve an accurate assessment of power system transient stability, a multi-classification assessment model based on an improved TCN-ResNeXt is proposed. The core of this model lies in a dual-branch structure, which enables the extraction and interactive fusion of dynamic temporal features and spatial features at multiple scales. By integrating the Triplet Attention mechanism, the model enhances focus on key features across the three dimensions of channel, space, and time—effectively boosting the assessment performance of the transient stability multi-classification model. To address the sample imbalance issue, a solution based on loss function improvement is proposed. This solution uses the entropy of the sample’s posterior probability as a discriminant criterion to screen out boundary samples. It incorporates boundary sample loss into the loss function, increasing cost sensitivity to hard-to-classify boundary samples, and thereby improving the assessment accuracy of the multi-classification model for hard-to-classify samples. The effectiveness of the proposed assessment model validated on the CSEE-DAS system, which features a high proportion of new energy penetration. Furthermore, the model is applied to the traditional IEEE 140-bus system, demonstrating its strong generalization capability. Additionally, in noise resistance performance tests, the proposed model outperforms other comparative assessment models.

---

Keywords

---