Time sequence information-based transformer for the judgement on the state of power dispatching

摘要

The key to avoiding power system abnormality, which often causes serious safety accidents, is finding abnormal data from massive data. The current abnormal detection model of the power dispatching system has low accuracy and low efficiency. Therefore, this paper proposes a new deep transformer model based on time-sequence features to automatically detect abnormal data. The model eliminates useless features in redundant data through the self-attention mechanism, extracts and analyses time-series features, for accurately detecting the abnormal. We evaluate the proposed model on the local dataset. The average detection accuracy of the model is 87.24% which has reached or even exceeded the accuracy of manual detection.