Fuel Cells and Mobile Robots Alex Wilhelm, Dr. Jon Pharoah, Dr. Brian Surgenor 1 Due to their scalability, new applications for fuel cells are being investigated all the time. Some see them replacing batteries in certain portable applications, offering extended run times with simple refuelling rather than the need for a lengthy recharge. Most portable fuel cell development is focusing on mainstream uses such as cell phones and laptops, but there are other battery-powered devices where fuel cells could offer benefits. While not yet a household item, mobile robots can already clean your carpet and mow your lawn, and are used in science, search and rescue and military exploration. Research and experimentation was done this summer at Queen s University in Kingston, Ontario, Canada under the guidance of Dr. Brian Surgenor and Dr. Jon Pharoah to investigate the feasibility of fuel cell powered mobile robots. Funded in part by a National Science and Engineering Research Council (NSERC) undergraduate award, a fourth year Mechanical Engineering student, Alexander Wilhelm, spent the summer converting an existing teaching robot to fuel cell power. Used in the Mechanical department s mechatronics course, the existing robots are based on a BASIC Stamp, a common microprocessor for hobbyists, and two DC motors, with room for various sensors and accessories. They are powered using 12V 2.2Ah lead acid batteries (26.4Wh), which run the robots for about four hours. Testing determined that the robot needs an average of 6W, with a peak draw of 14W. Analysis suggested that an 8W fuel cell stack would handle regular demand well and a series of super capacitors could jump in to supply peak power when needed, thus reducing the size and cost of the fuel cell required by using a hybrid arrangement. 1 E-mail 0aw2@qlink.queensu.ca, pharoah@me.queensu.ca, surgenor@me.queensu.ca respectively
While much work is being done to improve direct methanol fuel cells, no stacks of suitable size are yet commercially available and methanol management remains an issue. Thus, despite the advantageous power density of methanol and the ability to store it easily as a liquid, hydrogen proved to be the only fuel that could be implemented at present. It of course is not easily stored; here too, there is very little that is commercially available at the scale proposed. While small metal hydride tanks can store a fair amount of hydrogen per volume, they are relatively heavy. Compressed gas tanks are lighter, but bulkier, and the necessary fittings and regulator to deal with the high pressures have not been optimized for weight or size. Also, these tanks need to be filled by a gas supplier. Ultimately, a small hydride tank holding 20 standard litres from GfE was selected, which can be refilled in about 15 minutes from a standard cylinder using an adapter. A pressure regulator made by Beswick Engineering that is designed for metal hydrides was used to supply the fuel cell from the hydride, however, it turned out that although the hydride tanks are filled at up to 435psi (30bar), they only release at up to 7psi (0.5bar), making the regulation down to the suggested 3psi rather trivial. Although hydrogen PEM cells are more prevalent, there are still few readily available. The chosen 10W stack from BCS Fuel Cells Inc. won out with its ability to use ambient air rather than pure pressurized oxygen and be self-humidified from water it generates. It also had the highest available voltage, ranging from 10V at no load to about 5V at 2A, for a nominal power density of 10mW/cm 2. Furthermore, it can be operated in dead-ended mode with only an occasional purge, which is necessary to remove excess water that has reached the anode side. This simplifies the balance of plant greatly. Its fairly high weight and larger than necessary power output were drawbacks. In order to supply the robot with a constant voltage, a LM2588 dc/dc converter evaluation board from National Semiconductor capable of boosting voltages from as low as 3.5V up to a consistent 13.5V level with over 90% efficiency was used. A voltage sensor, a current sensor, and the robot s microprocessor control the purging of the stack. A straight line fit is used to approximate the stack s voltage-current curve; when values
drop too far below, the microprocessor activates a small solenoid valve and hydrogen is blown through the stack. Thus, control of the power system is achieved using the robot s existing processing capability. Power sharing between the stack and the capacitors is passive and merely based on voltage differences. Initially, drawing high currents before the stack had had time to warm up would cause a breakdown in voltage with no recovery, as no strong convection current yet existed to drive air through the open cathode channels. The addition of a small CPU fan to draw air up through the stack removed this limitation, providing more oxygen with only a small parasitic loss of about 0.18W and also increasing peak performance significantly up to 5A at 5V, for a density of 25mW/cm 2. The robot ran numerous successful trials, achieving a continuous run-time of up to two hours with one fill. Unlike the battery, which would normally require a ten-hour recharge after being fully depleted, a new hydrogen tank or a 15min. refill is enough to get the robot going again. Though fuel cell performance drops somewhat towards the end of the tank, this is compensated for by the voltage converter. Nevertheless, the performance of the metal hydride was poorer than expected, as its small size limited the amount of hydrogen that could be drawn at a time due to its excessive cooling. A custom made heat sink gave some improvement, but the metal hydride still brings with it some of the problems normally associated with batteries; recharge time and power degradation. The project showed some of the problems involved in powering robots with fuel cells and also hinted at some of the pluses, though power densities of the prototype system were slightly below the lead acid battery system it replaced and one order of magnitude below a possible lithium ion system. One of the big advantages of fuel cells and a difference to batteries is that they decouple energy storage from power production. This makes it easy to provide more energy (in the form of fuel) as needed, and as long as fuel is supplied, the power available stays the same. In effect, a fuel cell gives similar benefits as a internal combustion engine, but it is also quieter, more efficient, non-polluting, and more easily scaled-down.
However, it is more complex than a battery and its 30-50% fuel efficiency can t compare to the over 90% cycle efficiencies of batteries if a source of electricity is available anyway. If a mobile robot currently uses lithium batteries, weight is probably a significant factor, and cost a lesser issue. For such a situation, fuel cells, probably in tandem with capacitors or a small battery, will become attractive in the future once component weights are optimized. If long run times are important and power requirements are relatively low say, for some sort of security surveillance robot a fuel cell becomes even more attractive. On the other hand, if a robot currently uses a lead acid battery, cost is probably more important than weight. Though fuel cells already compete well compared to lead acid batteries on a weight basis, a prediction of $5/W for portable fuel cells (Dyer) still remains an order of magnitude higher than that of the battery, not to mention the increased fuel costs.
Figure 1: Side view of MechBot showing space vacated by battery Figure 2: MechBot with fuel cell system
Figure 3: Top view of MechBot
Table 1: Fuel cell system specifications for MechBot Component Details Fuel Cell 10W-12W 10V-6V air-breathing selfhumidifying, BCS Fuel Cells Inc. Stack Fuel MHS20 Metal Hydride Cylinder, 20 std. L. Storage volume, GfE Hydralloy E60/0 Power LM2588 Simple Switcher 5A Flyback Conversion Regulator Evaluation circuit, National Semiconductor Pressure PRD2-2N1-1-A Ultra Miniature two stage Regulator pressure regulator, Beswick Engineering Co. Hybrid 4 PowerStor Aerogel Capacitors, B1030- Power BR5685, 6.8F each Fan Model KDE1204PFV1, 4cm x 4cm x1 cm, Solenoid Valve rated at 12V. 0.5W, SUNON V 2 Digital Solenoid Valve, model V2201PV5S80, 5V, 0.5W, Parker/Pneutronics References Dyer, C., Fuel Cells for Portable Applications, 2002, J. of Power Sources, 106, pp. 31-34 About Us The Queen's-RMC Fuel Cell Research Centre (FCRC) is Canada's leading universitybased research and development organization in partnership with industry dedicated to advancing the knowledge base for addressing the key technology challenges to the adoption of fuel cell applications. We can be found at www.fcrc.ca. The robot is now being used in our fuel cell elective and our mechatronics elective (final year courses in mechanical engineering).