Distributed Predictive Control For Water Supply Network Systems

Municipal water supply network is prone to induce leakage and pipe explosion,which is usually difficult to be quickly located and repaired.Combined with the research status at home and abroad,this paper extracted the features of topological structure and its working principle from the perspective of system science,abstracting the complex water supply network into a hydraulic weighted directed graph.The networks will be simplified into a "tree" or "co-tree" main network layout which is hierarchically,by using Dijkstra algorithm to calculate the minimum dissipated power weight path and its path frequency from water sources to all nodes.A boundary criterion has been proposed to check the belongings of those nodes which are close to the border area and the final DMA partition will be obtained.There are several common characteristics for water supply network,such as high dimensional,nonlinear and a very high degree of coupling,and we can establish discrete-time state space models in all DMAs.Utilizing interconnecting variables to indicate influence between neighboring sectors,a large-scale online optimization problem will be converted into several miniature distributed optimization problems,thus computational scale and complexity can be effectively reduced.Then we propose a water demand prediction method based on time series model and bring in a series of items,like Safety storage term,smoothness of the control actions and power consumption cost,into objective function.Serial and parallel schemes have been compared and optimal control sequence will be obtained by updating Lagrangian multiplier through iterations.Finally,we conducting simulations in InfoWorks WS software and the case illustrates the effectiveness of our proposed method.