YANGZong-yi,XIANGLing-pei,XIANGYu,et al.Dish-LSTM and DP?based Optimization Strategy for Water Supply Pump Scheduling[J].China Water & Wastewater,2025,41(22):39-44.
Dish-LSTM and DP?based Optimization Strategy for Water Supply Pump Scheduling
China Water & Wastewater[ISSN:1000-4062/CN:12-1073/TU]
volume:
第41卷
Number:
第22期
Page:
39-44
Column:
Date of publication:
2025-11-17
- Keywords:
- LSTM model; hybrid prediction model of water demand and pressure drop; dynamic programming algorithm; pumping station scheduling
- Abstract:
- To address issues such as energy waste caused by traditional manual scheduling of water supply networks and excessive pressure settings at pumping station outlets that exacerbate pipeline leakage, a combined forecasting model (Dish-LSTM) integrating long short-term memory (LSTM) networks with the distribution shift (Dish) paradigm is proposed. By using node pressure drops instead of node pressures as input and output features, the model significantly improves prediction accuracy and efficiency. Additionally, an optimized pressure control strategy for pumping stations is designed based on the prediction of pressure drops at the most unfavorable points in the water supply network. This strategy uses predicted and control values of pressure drops to inversely guide the future operating pressure of pumping stations. Combined with the dynamic programming (DP) algorithm, an optimized scheduling model for pumping stations is constructed to generate scheduling schemes that balance energy consumption and water supply safety. Simulation tests conducted on a large waterworks in a central urban area of Chongqing demonstrate that this strategy reduces the average pipeline pressure by 41.2 kPa and decreases pumping station energy consumption by approximately 10.57%.
Last Update:
2025-11-17