Abstract Model predictive controller (MPC) is a successful in control of the building's environment. The inherent part of the MPC is a system's model which is the bottleneck of this control approach. Data collected for the system identification very often comes from the closed-loop operation which often leads to a failure of classical identification approaches. A new algorithm is presented here when a persistent excitation (PE) condition is incorporated into a control criterion. The controller ensures sufficiently excited data, which makes the system re-identification possible. The algorithm uses the maximization of the minimal eigenvalue of the information matrix to achieve a richer identification data. As a result, the data used for re-identification are sufficiently excited and the estimate of the system parameters is better.