Fault Tolerant Model Predictive Control Applied to a Simulated Hydroelectric System

This paper proposes a joint data-driven and model-based approach oriented to the design of a fault tolerant controller for regulating the speed of a Francis turbine included in a hydroelectric system developed in the Matlab and Simulink environments. The nonlinear characteristics of hydraulic turbine and the inelastic water hammer effects were considered to simulate the comprehensive behaviour of this dynamic process. The data-driven strategy has been suggested for deriving in a straightforward way a prediction filter able to reconstruct the fault affecting the hydroelectric system. The fault tolerant controller development requires the knowledge of the dynamic model of the monitored system, which is achieved by means of a model predictive control scheme that compensates the fault. These features of the work represent an important point when plug-and-play implementations are considered for a viable application of an effective fault tolerant methodology. In particular, by means of this methodology, the fault tolerance properties are achieved by using an active approach. It is assumed that the fault considered in this work affects the electric servomotor used as a governor. The performances obtained are compared to those of a passive solution already implemented by the authors for the same hydroelectric system.