Long-Term Load Forecasting of CCHP System Based on Rolling Training Mechanism for STS

The Combined Cooling, Heating and Power(CCHP) system, due to its high energy utilization rate and its ability to effectively reduce environmental pollution, has attracted worldwide attention as energy issues become increasingly severe. Accurate long-term load forecasting is of great significance for the safe and stable operation of cogeneration systems and cost savings. In this paper, an rolling training mechanism-based Sequence to Sequence(STS) algorithm is proposed to improve the long-term prediction of cold loads of a Shanghai Energy Station. The proposed method is compared with the Gate Recurrent Unit(GRU) and ordinary Sequence-to-Sequence (STS) under different predictive time horizon and outperforms with substantial superiority in terms of prediction accuracy and effectiveness in mid-to-long-term forecasting.