Estimation Of Offtake Discharge And Cross-Device Parameters Using Data Assimilation For The Automatic Control Of An Irrigation Canal

The automatic control of open surface hydraulic systems such as Rivers (with dams and/or hydropower plants) or irrigation canals (with gated cross-devices) almost always use hydraulic models. These models can be used in different manners, either just to test and validate controllers prior to implementation, to tune the controller parameters off-line, or used on-line in real-time. The control algorithm calculates the control action variables u, using measured variables z obtained from the real system, in order to achieve some objectives for some controlled variables y. These models have always limited precision due to unknown or wrong: parameters, input variables and internal states. Among the parameters we find cross-device discharge and bed friction coefficients. Among input variables we find the inflow or outflow discharges entering into the river, or taken by users from the canal. Indeed, they are rarely measured, or in the best cases with a limited precision. This is a problem since the tuning of the control parameters of the feedback loops depends a lot on the dynamics of the system and therefore on the previous listed parameters. A feedforward control component, very useful for this class of delayed systems, could benefit from the knowledge of the input variables. In this paper we will show how data assimilation technics can reconstruct these unknown parameters and variables. We will also focus on the required number and locations of the measurements, to be able to reconstruct this correctly. We will study the best or required configurations allowing to use this information, detect and isolate some problems, correct the model, and reconstruct the wrong or unknown variables, inputs or parameters of the model. The framework we will use for this study is the Kalman filtering one. We will see that this framework is very powerful to solve the above described problems