(369g) Model-Based Hazard Monitoring and Process Safety Control Using Model Predictive Control

Safety is critical in the process industries due to the severe consequences for both lives and property. There are many causes of these incidents disasters such as human error, hazardous materials release and manufacturing defects. Techniques such as hazards and operability analysis (HAZOP), fault trees and what-if scenarios are performed to evaluate the safety of a process. These probability risk assessment (PRA) methods combine expert opinions with statistical data to provide quantitative measures of risk. The limitation of PRA is the difficulty of the consideration of dynamic physical models during the process operation. In this work, we consider the model-based hazard monitoring and safety control to improve both accident prevention and dynamic risk assessment. Our approach has two fundamental ingredients: (1) the use of state-space models and state variables to capture the dynamics of process hazard; (2) the model predictive control actions which are computed by formulating and solving a dynamic optimization problem on-line that takes advantage of a dynamic process model while accounting for process constraints. The work addresses model predictive control design for compressed air energy storage system with the purpose of achieving safety control and energy savings. The compressed air energy storage system with salt cavern air storage includes thermal energy exchange process during air compression and air expansion, thermal energy storage process of the compressed air storage and the power grid control process. The modelling and control of the CAES system are motivated by shifting electricity from peak periods to off-peak periods. Finally, the controller performance is assessed by numerical simulations.