Application of model based predictive control to a pumped storage hydroelectric plant
This thesis describes the development of a Predictive Control to SISO and multivariable linear and nonlinear models of Dinorwig pumped storage hydroelectric power station. The results show that Generalised Predictive Control (GPC) offers significantly better performance across the plant's operating range when compared with classic PI controllers. The GPC controller produces a faster response when the station is operating with a single unit while preserving stability as the operating conditions change when multiple units are on-line. Inclusion of constraints in the GPC controller yields a fast, well-damped response in the common case when only a single Unit is in operation, without compromising stability when multiple Units are on-line. Simulation has also shown that improved power delivery is obtained when the plant is operated in frequency control mode. In the final part of the work a Mixed Logical Dynamical (MLD) predictive control was developed and applied to a MIMO nonlinear elastic model of Dinorwig. The results show that MLD predictive control is faster and less sensitive than the constrained GPC. The MLD predictive control can also be integrated with high-level plant functions.