Profiling Pumped Storage Power Station via Multi-Sequence Joint Regression

Suggesting personalized tags to the Pumped storage hydropower plants (PSHPs) towards purchase requirements forecasting plays a key role in achieving the smart power grids. However, current tag suggestion solutions only take single sequence into consideration, and predict single label for PSHPs, resulting in suboptimal forecasting accuracy. In this paper, we propose a novel Multi-Sequence Joint Regression (MSJR) model towards the task of PSHP tagging. In particular, MSJR exploits multi-sequence as input for collaborative perception purpose, and a multi-label regression module is built in the MSJR framework to predict tags describing the purchase requirements of PSHPs. Our encouraging experimental results on a real-world dataset, crawled from the ERP system of the State Grid Xin Yuan, validate the superiority of the our MSJR over several existing tagging suggestion methods.