Online static security assessment for cascading failure using stacked De-noising Auto-encoder

Abstract

The phenomenon of cascading failures (CF) has a great research potential for wind integrated power systems (WIPS) because of stochastic nature of wind. CF may cause unwanted power system blackout if not handled effectively. To evaluate the impact of N-k contingencies which causes CF, this paper proposes multi-layer stacked de-noising auto-encoder (SDAE) based online static security assessment (SSA) using risk based feature set construction for WIPS. A feature set is constructed taking into account the branch overloading probability and risk along with other features. Online SSA is performed after SDAE training is done successfully. Multi-level feature extraction is performed by multi-layer SDAE after executing the pre-screening process for redundant and irrelevant input features using mutual information theory. The proposed method is executed on modified New England 39 bus power system which is further extended for larger power system of IEEE 118 bus power system. Training time and root mean square error (RMSE) of the proposed method are compared with those of long short term memory (LSTM) and auto-encoder (AE). The effectiveness of the proposed method is also demonstrated by comparing the results with ORNL-PSerc-Alaska (OPA) model.