Three-phase imbalance prediction: A hazard-based method

Three-phase power systems are widely used in highpower applications and they are designed to operate in a near-balanced fashion at a nominal frequency for optimal performance. Although a slight voltage imbalance is tolerant in real-life scenarios, a significant three-phase imbalance has detrimental effects for power delivery and even damages many key components in a power system. Therefore, how to prevent or reduce the three-phase imbalance has become a critical issue in smart grids. In this paper, we attempt to address this challenge by introducing a hazard-based model to predict the three-phase imbalance based on historical status data. Firstly, we model the three-phase imbalance as a probability event and thus formalize the three-phase imbalance prediction problem as a time-dependent classification problem. Secondly, we introduce the feature extraction method and present the hazard-based prediction model. Experimental results on a realworld dataset demonstrate the effectiveness of the proposed prediction model.