With the increase of event flow generated from sensor kind electronic devices in IOT(Internet of things) and increasing demand of matching accuracy/confidence of more complex events, uncertain complex event processing is becoming more and more been concerned. A large number of RFID or sensor monitoring data exist in precision agriculture, but current hardware and wireless communication techniques cannot support 100% confident data. One stream processing engine which can process uncertain data in precision agriculture is needed. In this paper, a new type of complex event processing engine PUCEP(Provenance uncertain complex event processing) was proposed, in which probability flow theory and data provenance management theory were added based on the existing flow processing engine SASE. Sufficient approximate lineage query algorithm is used to calculate the probability of an event in order to improve the efficiency of probability calculation of large amount of data and the pattern matching was carried out by using the two fork tree and NFA. This optimized method can not only calculate the probability of outputs of compound events but also improve the matching efficiency of uncertain complex events, thereby reducing the computation cost and response time to a realistic degree. The experiment uses sensor data acquired from an agricultural greenhouse and shows that this method is efficient in processing complex events over probabilistic event streams.