Multi-agent based early-warning for monitoring of major hazard installations

Abstract

If the status information of major hazard installations is not measured, the hazards can not be controlled. Distributed and remote monitoring and control are important to manage the distributed hazard installations and to keep people and properties from dangers. In the paper, the multi-agent group (MAG) model for monitoring and control of hazard installations is proposed. The monitoring and control tasks are decomposed into 5 Agents, among which the Feature Extraction Agent (FEA) and the Early-warning Analysis Agent (EAA) perform the early-warning function of hazard installation monitoring. The structures, main functions and working processes of the FEA and the EAA are analyzed. The wavelet based feature extraction method and the Kalman filter based prediction method are adopted. Based on multi-agent technologies, the early-warning system of hazard installation is more adaptive, flexible, and expandable.