Predicting Typhoon-Induced Transmission Line Outages with Coordination of Static and Dynamic Data

The prediction of typhoon-induced transmission line outages is essential to improve the resilience of urban power systems. This paper proposes a novel data-driven prediction model to promote the accuracy by quantifying the cumulative influence of dynamic data and mitigating the data imbalance. In the model, the static data and the dynamic data compose the disaster-causing feature vector as model input. Then, the denoising adaptive synthetic (ADASYN) sampling algorithm is proposed to select target samples purposely and generate minority samples adaptively to balance the dataset. Also, the discriminative model guarantees the consistency of the data distribution. Thereby, the dual path model is proposed to quantify the stable impact of static data and cumulative impact of dynamic data based on the feedforward neural network and gated recurrent unit (GRU), respectively, and fuse the extracted features with the multi-head attention mechanism to predict the category of the number of line outages. Based on the real dataset, this paper compares the performance of the denoising ADASYN algorithm and dual path model with benchmarking algorithms. The experiment results indicate that the proposed method witnesses a better accuracy in predicting typhoon-induced transmission line outages.