In recent years the idea of artificial intelligence has been focused around the concept of rational agent. An agent is an (software or hardware) entity that can receive signals from the environment and act upon that environment through output signals, trying to carry out an appropriate task. Seldom agents are considered as stand-alone systems; on the contrary, their main strength can be found in the interaction with other agents, constituting the so-called multiagent system. In the present work, a multiagent system was chosen as a control system of a single-shaft heavy-duty gas turbine in the multi input multi output mode. The shaft rotational speed (power frequency) and stack temperature (related to the overall gas turbine efficiency) represent the controlled variables; on the other hand, the fuel mass flow (VCE) and the variable inlet guide vanes (VIGV) have been chosen as manipulating variables. The results show that the multiagent approach to the control problem effectively counteracts the load reduction (including the load rejection condition) with limited overshoot in the controlled variables (as other control algorithms do) while showing a good level of adaptivity, readiness, precision, robustness, and stability.
A neuro-fuzzy controller for grid-connected heavy-duty gas turbine power plants
