An implementation scheme of MPC (Model Predictive Control) system combined with model error detection function based on plug-in technology is proposed. Based on an MPC model error detection method that uses additional small amplitude test signals, the model detection function including online data acquisition, test signal adding and MPC operation monitoring is implemented to extend the batch-wise adapting MPC system so as to integrate production control, model identification and detection. The model detection results can distinguish the mismatched models so that the users can reidentify them. This kind of system helps assure MPC's control performance without additional controller operation and maintenance cost, and also contribute to guarantee economical, efficient and safe operation of industrial productions.Index Terms--Model error detection, model predictive control, plug-in.