The　traditional　hydroelectric　unit　control　system　mostly　adopts　proportional　integral　derivative　(PID)　control　strategy,　which　has　simple　control　algorithm　structure　and　is　easy　to　understand　and　realize　by　the　operation　and　maintenance　personnel　of　the　power　station.　However,　its　control　objective　is　single,　and　the　dynamic　process　and　internal　state　information　of　the　system　water　machine　electricity　coupling　are　less　considered,　especially　the　lack　of　the　ability　to　constrain　the　system　state　variables　In　order　to　meet　the　increasingly　complex　operating　environment　in　the　regulation　system　of　hydropower　units　high-quality　control　needs.　Therefore,　a　step　intelligent　model　predictive　control　method　for　synchronous　generator　excitation　system　based　on　tree　algorithm　is　proposed.　The　step　control　and　intelligent　optimization　algorithm　are　introduced　into　the　rolling　optimization　of　model　predictive　control,　which　effectively　reduces　the　dimension　of　nonlinear　programming　solution　in　the　rolling　optimization　and　improves　the　feasibility　of　the　control　algorithm.　In　order　to　ensure　the　closed-loop　stability　of　the　control　system,　the　concept　of　＇quasi　infinite　time　domain　optimization＇　is　introduced　into　the　design　process　of　the　model　predictive　control　of　the　excitation　system　of　synchronous　generator.　The　terminal　penalty　term　and　the　terminal　inequality　domain　are　added　in　the　rolling　optimization　link,　and　the　＇virtual＇　local　linear　feedback　controller　outside　the　prediction　domain　is　designed.　Compared　with　the　traditional　excitation　control　method　in　single　machine　and　multi　machine　systems,　it　is　proved　that　the　predictive　control　method　has　stronger　ability　of　generator　terminal　voltage　maintenance　and　electromechanical　oscillation　damping　than　the　conventional　excitation　control　method.　©　2021　IOP　Conference　Series:　Earth　and　Environmental　Science.