Full Program »
Explainable Artificial Intelligence For Power System Security Assessment: A Case Study On Short-Term Voltage Stability
The intelligent stability assessment methods such as neural network show outstanding advantages in efficiency. However, the lack of interpretability makes it hard to be understood by power system operators and seriously affects its practicality. Taking short-term voltage security assessment as an example, this paper enhances the interpretability of neural network-based intelligent security assessment method in three aspects: physical meaning, concise rules and local hyperplane. Firstly, the security region boundary is approximated by the intelligent assessment model to clarify the physical meaning of the model. Meanwhile, the approximate mathematical expression of the security region can be obtained. Secondly, a model simplification method is proposed to obtain globally explainable security rules. Thirdly, based on the boundary expression, the local hyperplane of security boundary is constructed to provide local explainability. Test results on WSCC 9-bus system verify the effectiveness of the proposed method.