Power System Transient Stability Assessment via Probabilistic Distribution Prediction Enhancement of Maximum Lyapunov Exponent

The transient stability assessment (TSA) is the basis for secure operation of power systems especially when unexpected faults occur. Naive maximum Lyapunov exponent (MLE) is restricted by accuracy and timeliness while pure artificial intelligence (AI) techniques are hindered by mechanism interpretability. In this paper an intelligent augmentation scheme is proposed for TSA, in which firstly the distribution of power angle MLE is estimated by diffusion kernel density estimation (DKDE) and the kernels of probabilistic distribution functions are predicted by a gated recurrent unit (GRU) network. Then a neighborhood approximation estimation algorithm is put forward for transient stability discrimination referring to adjacent samples with similar MLE distributions. The TSA precision and interpretability of AI approaches and traditional methods could be integrated. Case study verifies the efficacy of the proposed scheme with 97.44% TSA accuracy.