Abstract: For hybrid systems which incorporate both dynamic and discrete event models in process industries, the models of material balance would be changed because of frequent scheme switch. The redundancy degree of the whole sensor network is time variant so that the traditional data reconciliation methods could hardly be used in practical processes. In this paper a new approach of data reconciliation for hybrid systems is proposed. For those nodes where the discrete scheduling events might happen, scheduling-equations are established and added to the models. So the new data reconciliation model with parameters of random scheduling-equations could be built. In this way, the model’s redundancy is improved, which enhances the solvability of the data reconciliation problem. Then its optimal solution could be obtained by using a reconciliation algorithm with uncertain models. Some simulations of the simplified model for a refinery are given and a comparison is made between the proposed approach and previous researches. The simulations results demonstrate the efficiency and robustness of the proposed method.