An Intelligent Assistant for Obtaining the Optimal Policy during Operation Transients in a HRSG

This paper shows how to use Markov Decision Processes to guide the actions of the operation personnel during a transient in a heat recovery steam generator (HRSG). The critical events shown correspond to situations where a strong participation of human operators is required. The aim is to help the steam generation control system to set the plant on a safe operation state. The proposed system also takes into account that either the actuators or control valves in the loop can fail. The possible recommended actions are of parallel nature so that different people can command different control devices to control the situation. A reward function specify the desired and non-desired operation states.