A new reinforcement learning based automatic generation controller for hydro-thermal power systems

Abstract

Recently, we proposed a reinforcement learning (RL) based approach for designing an automatic generation controller for a two-area power system (Ahamed, T.P.I. et al., Electric Power Systems Research, vol.63, p.9-26, 2002), where we demonstrated the efficacy of the approach on an identical, simple, two-area model. This paper aims to demonstrate an alternative RL-AGC design which is simpler. Its effectiveness is demonstrated by considering a hydro-thermal system whose dynamics are more complicated than the system considered previously.