Data assimilation for real-time estimation of hydraulic states and unmeasured perturbations in a 1D hydrodynamic model. Application to water management problems and comparison of Kalman filter and sequential Monte Carlo approaches

Scarcity of water resource and increasing competition for its use promoted recently the development of advanced control algorithms and SCADA technologies for the automatic management of open-surface hydraulic systems. In order to control hydraulic devices on irrigation canals or rivers, detailed information on the hydraulic state of the system must be available. This is particularly true when the control algorithms are based on Linear Quadratic Gaussian (LQG) or Predictive Control (PC) approaches using full state space models. Usually, the only known quantities are water levels, measured in limited locations. In this case, the design of an observer is a very useful tool for reconstructing unmeasured data, such as discharges or water levels at other locations, unknown perturbations, such as inflows or outflows, and model parameters. Several approaches are able to provide such observers. The paper illustrate the use of Kalman Filter or sequential Monte Carlo State Observer on these water management problems.