Control of a DC-DC Boost Converter for Fuel-Cell-Powered Marine Applications

Abstract

Economic factors together with protection laws and policies pertaining to marine pollution drive research for improved power generation. Fuel cells, being fuel efficient and environmentally friendly, could provide a desirable option and suitable alternative to conventional propulsion systems based on fossil fuels or even nuclear fission. Fuel cells are becoming fast a mature technology and employed in many various other areas. Flexibility of special purpose watercraft, power autonomy and modularity can all benefit from the use of fuel cells. Specifically, proton exchange membrane fuel cells are considered among the most promising options for marine propulsion applications. Switching converters are the common interface intermitted between fuel cells and the load in order to provide a stable regulated voltage. DC-DC converters have been widely used since the advent of semiconductors. These devices are typically adopted to accomplish voltage regulation tasks for a multitude of applications: from renewable energy power-plants to military, medical and transportation systems. Nonetheless voltage regulators exhibit the need for consistent closed- and open-loop control. Most common approaches are PID controllers, sliding mode controllers and artificial neural networks that are considered in this work. An artificial neural network (ANN) is an adaptive, often nonlinear system that learns to perform a functional mapping from data. In our approach, a typical example of a fuel cell, a power converter outfitted with an ANN controller, and a resistive load configuration is investigated. Simulation studies are crucial in power electronics to essentially predict the behavior of the device before any hardware implementation. General requirements, design specification together with control strategies can be iteratively tested using computer simulations. This paper shows the simulation results of the full system behavior, as described above, under dynamic conditions. Initially, an open-loop simulation of the system is performed. Next, an appropriately trained ANN is incorporated to the switching model of the DC-DC converter to perform simulations for validation. Conversely, during design and calibration of the ANN controller, instead of the switching model of the DC-DC converter, a trained ANN equivalent is employed.