Simulation studies of a GPC controller for a hydroelectric plant

In this paper, the performance of constrained Generalized Predictive Control (CGPC) is investigated on a comprehensive non-linear multivariable simulation of the Dinorwig pumped-storage hydroelectric power station. The responses of the system are compared with those of the classic PI controller, as currently implemented on the plant. The results show that CGPC offers significantly better performance over the whole operating envelope. Fixed-parameter CGPC produces a faster primary response and better steady-state accuracy when the station is operating with a single unit, while also preserving stability when the operating conditions change as multiple units come on-line. It can accommodate large amplitude ramp power changes in combination with feed-forward control. It is less sensitive than PI to changes in hydraulic head and gives better power-tracking when units operate in frequency-control mode. Finally, the computational time of the CGPC algorithm is investigated.