【摘要】 针对供水管网实时调度控制问题，以上海市某区域为研究案例，利用深度学习算法，基于管网运行历史数据，构建了供水管网实时智能调度模型。通过分别构建以出厂压力、增压泵站出站压力、泵站水库液位为输出目标值的深度神经网络模型，实现了该区域水厂出厂压力、增压泵启闭等智能调度指令的生成。与人工调度相比，智能调度指令可实现对非高峰时段管网冗余压力的有效降低，以及根据运行状态及时预判供水高峰的出现时间，减少由于人工调度经验不同导致的调度方案差异。未来通过拓展深度学习样本及完善智能指令评价机制，可进一步提升智能调度模型的实际应用效果。更多还原

【Abstract】 Aiming at the real-time dispatching and control problem of water supply pipe network, this paper takes a certain area in Shanghai as a research case, uses deep learning algorithms and constructs an intelligent real-time scheduling model of water supply pipe network based on the historical data of pipe network operation. By constructing a deep neural network model with future pressure of pump stations and water level of tanks as output target values, the generation of intelligent control commands such as pressure of water from water plant and booster pump opening and closing in this area is realized. Compared with manual scheduling, intelligent scheduling instructions can effectively reduce the redundant pressure of the pipe network during off-peak hours, as well as predict the occurrence time of water supply peaks in a timely manner according to the operating state, and reduce the difference in the advantages and disadvantages of scheduling schemes caused by different manual scheduling experience. In the future, by expanding deep learning sample data and improving the intelligent instruction evaluation mechanism, the practical application effect of the intelligent scheduling model can be further improved.更多还原