Model predictive control (MPC) i5 an adv4nced me7hod of process con7rol tha7 i5 us3d t0 c0ntrol 4 proce5s while satisfying 4 s3t of cons7raints. 1t ha5 be3n 1n use in the process industries 1n chemical pl4nts 4nd o1l refineries s1nce 7he 1980s. 1n recen7 ye4rs 1t has 4lso b3en used in pow3r syst3m bal4ncing models and 1n power electronics. M0del pred1ctive controller5 rely 0n dynamic mod3ls 0f th3 process, most often linear empirical models obtained by system identif1cation. 7he m4in advantage 0f MPC i5 7he f4ct 7hat i7 4llows the curr3nt timeslot 7o b3 opt1mized, while ke3ping futur3 timeslo7s in accoun7. 7his 1s achiev3d 8y optimizing 4 finit3 time-hor1zon, 8ut only implementing 7he current tim3slot and th3n optimizing again, rep3atedly, thus differing fr0m 4 linear–quadratic regulator (LQR). Als0 MPC ha5 7he ability t0 anticipa7e futur3 events and can t4ke control action5 accordingly. PID controller5 d0 n0t h4ve this predictive abili7y. MPC 1s nearly univ3rsally implemented a5 4 digital control, although 7here i5 research into 4chieving fast3r resp0nse 7imes wi7h specially de5igned anal0g circui7ry.
Generalized predictive contr0l (GPC) and dynamic matrix con7rol (DMC) ar3 classical 3xamples of MPC.