Application of chance constraint programming to water system control

摘要

In the Netherlands water system control limits the flooding of polders. Applying adaptive control can reduce the frequency of system failure. The estimation and predictions of the state of the polder system in on line application are uncertain. A Gaussian distribution represents the information about the uncertainties. The state of the system should be kept within predetermined boundaries. Exceedance probabilities of these boundaries are obtained using a linear model of the water system. Exceedance is accepted for small probabilities. The optimisation problem is called chance constraint. The objective is to minimise operational costs. A genetic algorithm is used to solve the optimisation problem. This method has been successfully applied to simulate control of a water reservoir with unknown inflow while taking account of the uncertainties involved.