MODEL PREDICTIVE CONTROL FOR REAL TIME OPERATION OF HYDRAULIC STRUCTURES FOR DRAINING THE OPERATIONAL AREA OF THE DUTCH WATER AUTHORITY NOORDERZIJLVEST

摘要

A pilot project in 2012 for the Dutch regional water authority Noorderzijlvest has shown that the application of Model Predictive Control (MPC) can increase the safety level of the water system during flood events by an anticipatory pre-release of water. Furthermore, energy costs of pumps can be reduced by making tactical use of the water storage and shifting pump activities during normal operating conditions to off-peak hours. In this paper, the extension of the pilot to a real time decision support system is presented. It supports the daily operation of 34 aggregated structures both in wet and dry periods by providing optimal control settings through the application of MPC. We explain the improved prediction model that is accurate and fast enough for optimization purposes, and how it is integrated in the operational flood early warning system. Besides the prediction model, the weights of the individual objective function terms are an important element of MPC, since they shape the overall control objective. We developed special features in the forecasting system to permit the operators to adjust the objective function with respect to seasonal changes in order to evaluate different control strategies.