NMPC based Temperature Control in Fed-batch Reactor to Avoid Thermal Runaway

Abstract

Numerous fatal accidents occurred in the recent past caused by thermal reactor runaway, despite the fact, that the phenomenon of thermal runaway is well-known. However, accidents caused by runaway can be foreseen by using correct model and proper reactor runaway criteria, hence the operators can intervene in time in the system to prevent undesired events. Model Predictive Control (MPC) methodology is proposed to avoid reactor runaway during the optimal operation of fed-batch reactors. Fed-batch reactors are applied to carry out highly exothermic reactions safely, where the reactor and the sequence of process steps in normal operation are designed model-based. However, some of the parameters of reactor system (e.g. heat transfer coefficient) can change slightly over time or operators can make mistakes which can lead to trigger a runaway reaction. MPC can support the safe production by keeping the reactor controllable zone during the whole operation, in which also reactor runaway criteria can be implemented to predict the development of thermal runaway. Since the heat removal is limited the temperature is controlled by the inlet rate of feeding reactant and the cooling capacity is operating almost at maximum. To avoid thermal runaway, the feed rate of reactant is constrained by runaway criterion (namely Modified Dynamic Condition) which increases the safety of the process operation. Modified Dynamic Condition is implemented in NMPC algorithm as a penalty term, and also the criterion is used to define the minimal length of the considered prediction horizon based on process safety time. Moreover, the variable feed rate has an impact on economic of the operation since with decreasing the reactant concentration the feed rate increases to keep the reactor temperature close the optimal. Reactor efficiency is increased while the whole operation stays in controllable zone. The objective is to maximize productivity under the whole region of parameter uncertainty.