(293a) Neural Network Controller for Regulation of a Water-Cooled Fuel Cell Stack

Fuel cells are readily available for use as backup generators for residential applications in the 5 kW range. The advantages these generators have over conventional gasoline or natural gas powered generators include improved effciency, easy maintenance due to the absence of moving parts, and reduced noise during operation. The idea of using a fuel cell for dynamic power demand is relatively new and this application requires the development of a control system that keeps the system at optimal conditions during power demand uctuationss. In a previous paper, we developed a mathematical model for a novel water-cooled fuel-cell stack that provided power up to 5 kW for a residential application. In this paper, a backpropagation neural network is designed to control the temperature of a 5 kW hydrogen fuel cell stack. The power demand is obtained experimentally from a 3-bedroom house. The controller is initially trained to recognize the pattern of dynamic power demand from a consumption source. The power demand is the driving force in current production from the fuel cell, which in turn increases the temperature of the fuel cell stack. This temperature change is controlled by changing the flow rate of cooling water through the fuel cell stack. The results show that the neural network controller has excellent performance in maintaining the stack temperature at the desired set-point despite signicant fluctuations in power demand.