In the petrochemical process, a risk monitoring and early warning system for hazardous chemicals is designed based on CPS technology to solve the problems of monitoring, identification and prediction of major hazard sources of hazardous chemicals. The system is divided into three layers: physical layer, network layer and decision layer. At the physical layer, the front⁃end hazardous chemicals data collection, data transmission and local storage are completed, and the data are classified and transmitted to the cloud server in real time. The data passes through the network layer and is identified by the decision⁃making layer for major hazard sources and early warning for hidden dangers of hazardous chemicals. Finally, the whole data is displayed. Genetic algorithm and neural network algorithm are used in the decision⁃making level for risk assessment of major hazard sources. Simulation experiments show that this method can accurately complete the risk assessment of major hazard sources of hazardous chemicals. In the laboratory, the remote monitoring and early warning system for hazardous chemicals based on CPS is proved to have the capability of hazard source identification and risk hazard prediction through actual verification.