Dynamic and static Kalman filters based on a reduced order model have been developed for estimating the surface and center temperatures of steel ingots in the soaking pits from the noisy measurements of the pit wall temperatures and fuel flow rates. The results illustrate that the state estimates given by Kalman filters agree with real state values. Computer simulation of control through the stochastic optimal control law shows the relationships between the parameters of the control algorithms and the relative energy consumption and soaking time needed.