In Cyber Physical System (CPS), there are many reasons for the uncertain data generated in applications, such as sensor errors, electronic noise, etc. Therefore, it is necessary to deal with uncertain data specifically. This paper expands the complex event query language to support the processing of uncertain data, and gives a model of event probability. Aiming at the characteristics of data stream in CPS applications, an effective parallel complex event processing method for uncertain data stream is proposed. This method can efficiently process the raw data stream generated by CPS by means of combining matching tree and cache table. The experiments show that this method can process big uncertain data streams in CPS and have acceptable performance and scalability.