Adversarial Feature Learning of Online Monitoring Data for Operational Risk Assessment in Distribution Networks

With the deployment of online monitoring systems in distribution networks, massive amounts of data collected through them contain rich information on the operating states of the networks. By leveraging the data, an unsupervised approach based on bidirectional generative adversarial networks (BiGANs) is proposed for operational risk assessment in distribution networks in this paper. The approach includes two stages: (1) adversarial feature learning. The most representative features are extracted from the online monitoring data and a statistical index N
_φ
is calculated for the features, during which we make no assumptions or simplifications on the real data; (2) operational risk assessment. The confidence level 1 - α for the population mean of the standardized N
_φ
is combined with the operational risk levels which are divided into emergency, high risk, preventive, and normal, and the p value for each data point is calculated and compared with the pre-defined interval of α/2 to determine the risk levels. The proposed approach is capable of discovering the latent structure of the real data and providing more accurate assessment result. The synthetic data is employed to illustrate the selection of parameters involved in the proposed approach. Case studies on the real-world online monitoring data validate the effectiveness and advantages of the proposed approach in risk assessment.