Guide to Lightweight Electric/Hybrid Vehicles: Possible energy storage systems

Home | Articles | Forum | Glossary | Books

1. Electronic battery

Electric vehicles are at a historical turning point - the point where technology permits the performance of electric vehicles to exceed the performance of thermal engines. Currently quality battery technology is expensive and heavy. This favors hybrids with small peaking batteries - typically 10 Ah at 300 V, with a capability of 70 kW for 2 minutes. New battery geometries have been developed for this application in the form of high performance D cells with 6 mm thick end caps and M6 terminals. A single string of cells handles 100 amps for 10 seconds. The heat is transferred to the end caps then removed by forced air cooling. It is vital to maintain an even cell state of charge in long strings. This solution has two drawbacks at present: (1) Cost - a 300 V 6.5 Ah stack costs more than $10 000 in January 98; (2) reliability - with only one string a single high resistance cell disables the battery.

Using the best available nickel-cadmium cells, it is possible to build a reliable peak power stack and use electronic control to maintain equal currents in strings. A later section will consider the next generation, an all electric hybrid with aluminum battery and alkaline fuel cell; this last important energy storage system is also discussed in depth in Section 4. A review of different battery types and performances is given in Section 5.

2. Battery performance: existing systems

All battery technologies can offer some solution to the peak power problem but there is only one parameter which ultimately matters, and this is the internal resistance of the cell. This is much more related to cell geometry than cell chemistry, as we shall discover. When AA, C, D and F cells were originally designed, nobody was thinking of discharging them at hundreds of amps, so it is hardly surprising that they are not ideal for the purpose. This problem will become even more extreme as power density improves:


D cell characteristics

The first volume hybrid electrics in the market came from Toyota (Prius) and Nissan, Fig. 1.

Toyota uses a 288 V string of NiMh D cells to give a peak power of 21 kW. The battery is mounted in horizontal strings of 6 cells in a matrix weighing 75 kg and is made by Matsushita Panasonic EV Energy. Nissan has worked with Sony who have developed the lithium ion battery, of AEA Technology (UK), into a high performance D cell, of 3.6 V 18 Ah, with a peak power of 1700 W/ kg. (This cell is 250 mm long × 32 mm OD and includes an electronic regulator.) Both solutions use advanced thermal management and both deliver the performance but at high cost. The task in hand is to reduce costs by 50%.

2.1 BATTERY PERFORMANCE: ALTERNATIVE APPROACH Question: Which application gets the best peak power battery performance at present? Answers:

Portable power tools and model airplanes; chemistry: nickel-cadmium; cell size: RR/C 1-2 ampere hour; peak discharge: 30° C; maximum current: 32 amps limited by connections (tags),

Toyota Prius battery pack Principal specifications

Characteristics of nickel metal-hydride batteries for HEVs

Fig. 1 Production hybrid battery technologies compared:

(a) Toyota Prius; (b) Sony/Nissan. Example of battery pack system for HEVs ( Toyota spec.)



Fig. 2. The performance issues of these cells are complex. Amazingly there are 14 different types of sealed Ni-Cad cells in four main product groups: (a) standard, (b) high discharge current, (c) fast charge and (d) high temperature.

It is the double-sintered fast charge cells that are used in model airplanes and in the world championships the Sanyo Cadnica is specified using a 1.7 Ah cell. A Tamiya connector can deliver up to 32 amps. Battery packs of 7.2 V six packs are usually fitted. In the power tool area, Black and Decker have the Versapack, a three cell string. Two or four strings are used in series with larger cordless power tools. One problem is that sealed Ni-Cad cells cannot be connected in parallel. At the end of the charge cycle, the cell voltage falls, causing rapid charge/discharge between parallel cells. On the face of it, the packaging problem is daunting, with more than 2000 cells in a pack.


--------p31 Sony/Nissan battery pack Specifications of prototype cell Rated capacity 22 Ah (4.0 V) Cell shape Cylindrical Dimensions, mm D50 x L250 Weight 1.2 kg

Fig. 2 Typical Ni-Cad packages and capacities.


The results discussed here are based on Sanyo products but the same trends are seen in Varta, Panasonic and Saft cells. See Fig. 2.

Why does internal resistance matter so much? This is because at 25° C discharge rate, the voltage drop on a 1.2 V cell is as tabulated below:

Voltage drop on a 1.2 V cell at 25° C discharge rate

RR C Sub D D

112.5 mV 160 mV 250 mV 280 mV

The 1.2 V cell is thus no longer a 1.2 V cell but closer to 1 V at 20° C. Why does the 1 Ah cell win? It is because it is short and fat - the others are long and thin. Cell geometry is the decisive factor for low internal resistance.


To test the performance of individual cells, Polaron built a string of six and charged/discharged at 24° C; that is 24 amps on 1 Ah cells. After five cycles, discharge time increased to 130 seconds, and temperature rise was about 10° C. It ran for several hundred cycles with virtually no change in characteristics. This is very severe compared to the true operating conditions, where the cells will have to supply 24 amps for perhaps 10 seconds under real world conditions, Fig. 3.

Fig. 3 Discharge test rig.


To achieve 70 kW for 2 minutes will require ten strings of 260 cells at l Ah. This is economical but not optimal. Three strings of 3.3 Ah would be optimal. This corresponds to using short D cells.

The use of l Ah packages does have some benefits. Spot welded connections can handle the current without special packaging. The key problem is that of automatic assembly with so many cells. This is easily accomplished using welding robots. One technique is to use two parallel plates which each can hold two strings of 260 cells. The cells are sealed with O-rings so that the center of each cell is oil cooled. The connections have to be in air, because the gas seal should not be immersed in oil as the seal may be damaged - and the oil contaminated with potassium hydroxide.

At a link current of 25 A nickel-plated steel links seem to be quite adequate. A second interesting packaging concept would be to create a ten pack version of the Versapack concept. These could be mounted in horizontal strings and air cooled for cell equalization.


Given ten parallel strings for 70 kW peak power, there is only one way to ensure equal currents in each string - active regulation. This seems a very expensive proposition but it fits in well with the structure of modern inverter drives. The present trend is to use a 300 V battery with a boost chopper and increase the battery voltage to give a DC bus of 600 V, as used in industrial drives.

Normally one would parallel a number of transistors to give a current rating of 300 amps. Three times 100 amps would be optimal as this is a complete 3-phase pack of IGBTs. In this case it is necessary to use smaller packs of 30 amps where each leg has its own independent current regulator.

This would not be an attractive proposition if it were not for the fact that at this current the required control circuitry is available economically. Ten such circuits may be connected to a common DC bus. This arrangement ensures excellent current sharing in both charge and discharge and prevents the situation where strings charge/discharge into one another. What we need now is some improvement in battery packaging. To operate three strings in parallel would be optimal from a cost versus reliability standpoint, as one would use the separate 100 amp phase legs in a six pack to control the current in the individual strings, Fig. 4.


It has been shown that a matrix of small rechargeable cells can be made to give large peak powers on a repeated basis with excellent life performance. The new D cell designs in NiMh and lithium ion are very expensive and a cheaper alternative is to use l Ah Ni-Cad cells with ultra-low internal resistance. Using this technique it would be possible to buy the cells for a 21 kW/10 second pack for about $2000 in 1999. To this the packaging and control cost must be added. However, at present this is still significantly cheaper than the use of custom battery packages.

Cell geometry is the decisive factor in achieving low internal resistance and there is much room for improvement on existing cell packages. Short cells with minimum distance from foil to terminal give best results. The use of multiple strings of cells in parallel, with active current sharing, improves reliability and reduces cost since the currents in individual packages are modest compared to single strings. Temperature control of the strings helps to maintain even state of charge at high charge/discharge rates and keeps cells cool, while extending cell life.

Fig. 4 Cell current sharing: typical EV drive (top); current loop control of PWM chopper (center); multiple chopper implementation (below).

Fig. 5 Exploded view of aluminum/air bipolar battery (courtesy Eltech Systems).

The development of advanced battery chemistries with increased power and energy density will place even greater demands on cell packaging in the future and a new family of optimum proportions needs to be designed for the job.

3 Status of the aluminum battery

In 1997, patents were filed in Finland for a new aluminum secondary battery. The inventor was Rainer Partanen of Europositron Corporation who claims major improvements in power density and energy density for the new cell based on a 1.5 V EMF. The author is interested in this problem because it represents one of the last major barriers to be overcome before the widespread introduction of electric and hybrid vehicles. In recent years, significant effort has been directed at improving secondary battery performance and this effort is beginning to bear fruit. We can now see advanced lead acid, nickel metal hydride and Lithium Ion products out in the market place with performance of up to 100 Wh/kg and 200 W/kg.

Market requirements fall into two distinct categories: (a) small peak power batteries of 500 Wh (2 kWh for hybrids) and (b) 30-100 kWh for pure electric vehicles. Each of the cell types has its own distinctive attributes but none has so far succeeded in making the breakthrough required for mass market EV implementation. The fundamental problem is one of weight. At the factory gate, vehicle cost is almost proportional to mass, as is vehicle accelerative and gradient performance.

Consequently it will take at least 300 Wh/kg and 600 W/kg to achieve the performance/weight ratio for long range electrics we really desire. This would make one type of hybrid particularly attractive - the small fuel cell running continuously together with a large battery.

If we consider a low loss platform for a passenger car at 850 kg with N = 0.3, we have 250 kg for the battery. At 300 Wh/kg we obtain 75 kWh, which would give a range of more than 250 miles after allowing for auxiliary losses. A low loss platform would consume 5 kW at 60 mph + 5 kW for auxiliary losses - total 10 kW - this equates to 7.5 hours at 60 mph = 450 miles steady state.

This level of performance is an order of magnitude better than lead-acid at the current time.

Clearly a new approach to the problem is required.

3.1 WHY ALUMINUM? In simple terms the answer involves (a) abundance, (b) low cost and (c) high energy storage. If we consider the recent developments in batteries they all seem to use materials like nickel which are highly dense and limited in supply. Likewise in fuel cells using platinum catalysts material scarcity is implicit, the annual production being about 80 tons worldwide. Any mass market battery needs to use materials available in abundance. In the bumper year of 1985, 77 million tons of bauxite were mined worldwide; aluminum is one of the most plentiful materials available on Earth. In terms of 1999 costs, aluminum is $2000 per ton so 250 kg would cost $500 - an acceptable sum.

In terms of energy storage, aluminum has one of the highest electrical charge storage per unit weight except for the alkali metals:

Aluminum 0.11 coulombs per gram 2.98 Ah per gram Lithium 0.14 coulombs per gram 3.86 Ah per gram Beryllium 0.22 coulombs per gram 5.94 Ah per gram Zinc 0.03 coulombs per gram 0.82 Ah per gram

Lithium and beryllium are alkali metals and are not suitable for use with liquid electrolytes, due to rapid corrosion, so are normally used with solid electrolytes.

Fig. 6 Conceptual design of filter/precipitator system when integrated with aluminum/oxygen battery (courtesy Eltech Systems).


The first serious attempt to build an aluminum battery was made in 1960 by Solomon Zaromb3 working for the US Philco Company. In Zaromb's concept for an aluminum air cell, the anode was aluminum, partnered with potassium hydroxide, and air was the cathode. This battery could store 15 times the energy of lead-acid, achieving 500 Wh/kg and a plate current density of 1 A/sq.

cm. The main drawback was corrosion in the off-condition, which resulted in the production of a jelly of aluminum hydroxide and the evolution of hydrogen gas. To overcome this problem Zaromb developed polycyclic/aromatic inhibitors and had a space below the cell for the aluminum hydroxide to collect. The chemical reaction is Al + 3H2 O = Al(OH) 3 + 3/2H2

In 1985 another attempt was made by DESPIC4 , using a saline electrolyte. Additions of small quantities of trace elements such as tin, titanium, indium or gallium move the corrosion potential in the negative direction. DESPIC built this cell with wedge-shaped anodes which permitted mechanical recharging, using sea water as the electrolyte in some cases. The battery was developed by ALUPOWER commercially. The battery had limited peak power capability because of conductivity limitations of the electrolyte, but provided substantial watt-hour capacity.

Other attempts have involved aluminum chloride (chloroaluminate) which is a molten salt at room temperature, with chlorine held in a graphite electrode. This attempt in 1988 by Gifford and Palmisano gives limited capacity due to high ohmic resistance of the graphite. Equally significant is work by Gileadi and co-workers who have succeeded in depositing aluminum from organic solvents though the mechanisms of the reactions are not well understood at this time.

Between 1990 and 1995 Dr E. J. Rudd led a team at Eltech Research in Fairport Harbor, Ohio, USA, which built a mechanically recharged aluminum battery for the PNGV program, Fig. 5. It had 280 cells and stored 190 kWh with a peak power of 55 kW, and weighed 195 kg. This battery used a pumped electrolyte system with a separate filter/precipitator to remove the aluminum hydroxide jelly, Fig. 6. Alupower built a 6 kW aluminum-air range-extender system under the same program, Fig. 7.

Fig. 7 A 6 kW range extender by Alupower, with lead-acid main battery.

Fig. 8 Characteristics of D cell (32 × 62 mm) against those of a 1 liter Partanen cell. Aluminum metal in anode/cathode 486 g 180 cm^3 Anion/cation reactant solution 1199 g 820 cm^3 Theoretical maximum energy and current capacity 2100 Wh/liter 1448 Ah/liter 1246 Wh/kg 859 Ah/kg Practical cells in a package should achieve 70-80% of the above values when package mass is included.


The cell invented by Rainer Partanen, Fig. 8, is an attempt to defeat the disadvantages of the aluminum-air cell. It is a secondary battery which uses coated aluminum for the anode and pure aluminum for the cathode. The electrolyte is a mixture of two elements: (a) an anion/ cation solution currently consisting in proportion of 68 g of 25% ammonia water mixed with 208 g of aluminum hydroxide, and made up with water to give 1 liter of solution; (b) a semi organic additive consisting of metal amines.

The exact formulation of the additive is a commercial secret. The inventor claims that this electrolyte achieves a large increase in charge carrier mobility and this results in figures of up to 1246 Wh/kg and 2100 Wh/liter, which have been achieved in many prototype cells that have been constructed. The figures relate to active materials, without casing. It is suggested that the technology is suitable for the construction of plate (wet cell) and foil (sealed) cells, with no limitations on capacity. The test cells have achieved a life of up to 3000 cycles, the main degradation mechanism being corrosion of the coating on the anode during recharging. One remaining hurdle to be overcome is the identification of a better coating material to reduce the corrosion.

The battery has some unusual characteristics in that it operates over a very wide temperature range, -40 to +70° C. This is in stark contrast to most batteries whose low temperature/high temperature performance is poor. The cell voltage is a nominal 1.5 V. Some interesting consequences arise if one assumes that the claims are true. The most significant is packaging. If we take the D cell which is 32 mm diameter × 58 mm long, as used by Panasonic/Toyota in the PRIUS battery pack, a battery with 150 g active mass stores 6.8 Ah and has a peak discharge current of around 100 amps. If we build a D Cell at a value of 1246 Wh/kg, this leads to a figure of 150 Ah. Polaron understand that very high levels of discharge current are possible - the inventor claims up to 20 times more power than existing cells in the market - but finding methods of supporting these currents in such a small space is a major challenge to achieve low terminal resistance, lead-outs and sealing, Fig. 9. It is claimed that the new technology uses environmentally safe materials which are fully recyclable.

Other developments which lend support to this invention are the emergence of ultracapacitors and electrolytic capacitors, both using aluminum electrodes with biological electrolytes. Very significantly, ultracapacitors operate well at low temperatures. In Russia 24 V modules, 150 mm diameter × 600 mm long store 20 000 joules and are used for starting diesel engines at -40° C.


The technical background to the invention is the result of a remarkable discovery in the field of complex electrochemistry and is based on the composition of the solution for electrical analysis and catalysis, releasing the energy potential of aluminum. Patent protection is being applied in three areas:

(l) The first is a solution which, under discharge, generates a reaction on the cathode side causing the energy potential of the aluminum to be released, and by ionization changes the molecular structure from metal to solution. Patent application Fi 954902 PCT/EPO (published).

(2) The second is a solution which, under discharge, generates a decomposition reaction in the chemical reactant mass. This is in crystal form which dissolves into solution and produces electrical potential on the anode side. Patent application Fi 981229 PCT/EPO (registered).

(3) The third component are the electrodes which have a dual role. They are formed of materials which enable them to act concurrently as non-ionized anode and ionized cathode. These electrodes are used in a multi-cell configuration as in existing battery technology. Patent application Fi 981379 PCT/EPO (registered).

The composition of these solutions, and the reactant mass, have the capability of producing an electrical current from the non-ionized anode and aluminum cathode when conductivity (resistance) is placed between them. When discharging, power is generated by the energy of the released aluminum, which reduces to about 35% of its original molecular density. When recharging, the reactant solution returns to its original form in solution and crystal mass and the aluminum atoms are deposited back onto the electrodes.


An aluminum secondary battery looks to be a very promising candidate for the storage of substantial energy. Whether the inventor Rainer Partanen has found the correct technique remains to be demonstrated. Although the claims for peak power and energy density seem very high, Sony have demonstrated 1800 watts per kg in lithium-ion recently and aluminum-air cells achieved 500 Wh/kg in 1964. The author considers aluminum to be a worthy contender for advanced battery construction and clearly this is an area which merits much greater investigation in the future. One point is clear - by making the active aluminum electrode the cathode, the parasitic reaction that is the big drawback of the aluminum-air cell is avoided, because the 1.5 V potential across the cell suppresses the reaction. Two questions that remain to be answered concern the levels of conductivity and mobility that will need to be exceptional to justify the claims made for the Partanen cell, also whether cell packaging will be a significant problem, requiring a new range of packages to be developed.

4 Advanced fuel-cell control systems

This section considers the development of a fuel-cell controller and power converter for a vehicle weighing 2 tons, for operation in an urban environment. The techniques employed can be used with either PEM membrane fuel cells or alkaline units. The main challenge is to re-engineer a high cost system into a volume-manufactured product but this is unlikely to be achieved 'overnight'.

What is required is a new generation of components which are plastic as opposed to metal based.

The power electronics are practical, but need integrated packaging to reduce costs. Equally important is improvement in the fuel-cell stack specifications. This section considers the requirements and performance of a low pressure scheme at the current state of the art and predicts the measures needed to achieve significant cost reduction.

Modern hybrid cars are demonstrating major improvements in fuel consumption (3 liters/100 km) and emissions (ULEV limits) compared to conventional thermal engines. These designs use small peaking batteries which weigh less than 100 kg, for a family sedan, and store perhaps 2 kWh.

Fig. 9 Aluminum/oxygen power system and its characteristics.



Power 2.5 kW Capacity 100 kWh Voltage 120 V nominal

Endurance 40 h at full power Fuel 25 kg aluminum anodes Oxidant 22 kg oxygen at 4000 lb/in^2 Buoyancy Neutral, including aluminum hull section Time to refuel 3 h


Mass 360 kg Battery diameter 470 mm Hull diameter 533 mm System length 2235 mm Non-dimensional performance:

Volumetric energy density 265 Wh/l Gravimetric energy density 265 Wh/kg


A new aluminum battery chemistry has been identified whereby it should be possible to store 50 kWh in a weight of 150 kg in perhaps 3/5 years from now. Nickel-metal hydride needs 500 kg with current technology to achieve 50 kWh. This makes a new type of hybrid an interesting long-term contender - the electric hybrid with a small fuel cell. In this vehicle a 2-5 kW fuel cell would charge the battery continuously. The only time the battery would become discharged would be if one travelled more than 400 km in one day. In this case the battery would be rapidly charged at a service station. Since the battery is light the cost is moderate and because it is not normally deep cycled a long life can be expected. Aluminum test cells have already demonstrated over 3000 deep discharge cycles and operation down to -80°C, as seen in the previous section.

At the present time we need to use larger fuel cells and smaller batteries similar to the hybrids with thermal engines. The vehicle which is going to be the development testbed is the new TX1 London taxi chassis made by LTI International, a division of Manganese Bronze in Coventry, shown in Fig. 10. This vehicle has been chosen because of growing air quality problems in London. The City of Westminster is now an Air Quality Improvement Area. This is mainly due to a large increase in diesel use which has resulted in unacceptable levels of PM10 emissions. Public Transport is a major contributor, with the concentration of large numbers of vehicles in the central zone.

Two types of fuel cell are attractive for use in vehicles - the PEM membrane and the alkaline types, as described in the following section. Both types have undergone a revolution in stack design in the last few years with the result that the stack (Fig. 11) is no longer the major cost item in small systems, it is the fuel-cell controller and the power converter. In this section we shall review the problems to be solved and offer some suggestions as to the likely course of development.

As always the fundamental issue is to convert a high cost technology for mass production civilian use. Current (1998) fuel cells cost $1000 per kW and most of that cost lies in the control system and power conversion. Stacks will cost less than $100 per kW in mass production. The challenge is to reduce the control system cost. It is for this reason that most vehicle fuel-cell manufacturers are opting to supply the stacks, and leave the car industry to manufacture the controller, Fig. 12.

This is an opportunity that Fuel Cell Control Ltd intends to take up by offering control systems commercially.

Fig. 10 TXI London taxi.

Fig. 11 Developed PEM fuel cell: (a) plate; (b) stack; (c) anode; (d) cathode.


Here is a typical specification:


Power: 7.2 kW max Output voltage: 96 V DC no-load 64 V DC full-load Output current: 110 amps Operating temperature: 70°C Fuel: Air 45 cubic meters per hour Pure hydrogen 5 cubic meters per hour Hydrogen storage: Cryogenic - 180°C High pressure 200 bar DC/DC converter 1 Input: 60-100 V DC Output: 0-396 V at 2.45 V per cell, lead-acid - 18 A Current ripple less than 1 part in 10 000 Fuel-cell controller Intelligence: 80 I/O programmable logic controller at 24 V DC Control: Close loop: hydrogen 0-1/10 bar; air 0-45 cubic meters/hour proportional to demand Purge: Dry nitrogen loop

Viable energy storage systems 43 Pumps: (l) Hydrogen 80 watts (2) Air 320 watts (3) KOH 85 watts (4)Water 10 watts Valves: 10 off electropneumatic control Preheat: 2 kW - 312 V DC DC/DC converter 2 Input: 200/400 V DC Output: 27.6 V DC 600 W for control system

----------- Fig. 12 Fuel-cell system specification.

Figure 2.13 shows the cell layout of the two fuel-cell types, alkaline and proton exchange membrane (PEM). In the alkaline type, the electrolyte is a liquid - potassium hydroxide or KOH.

This is the same material used in alkaline batteries. The anode membrane is porous and has eight very small amounts of platinum catalyst (1 g would cover three football fields of plate area). The cathode side has a silver catalyst made by Hoechst. It is possible to use platinum but the cost is much greater.

In the PEM type the electrolyte is a solid, Nafion 115 sheet - a proprietary Du Pont product; there are competitors such as Dow Chemical and Ashai in Japan. The anode is similar to the alkaline type. The cathode membrane has a platinum catalyst and much research has been aimed at reducing the cathode loading which is why many PEM cells use the high pressure approach, since it helps to reduce the amount of catalyst for a given current density. Catalysts are the main cost in stack construction and optimizing their use is a major research area. Other differences between the two types are cooling and source gas purity.

In both systems about 40% of the fuel expended is given out as heat. In the PEM type, water cooling plates are used to remove this heat. In the alkaline type the electrolyte does this job and has the added advantage that it does not freeze at 0°C. Consequently both systems need a liquid cooling system.

Fig. 13 Alkali (left) and PEM cell layouts compared.

In Europe, Esso is already committed to making hydrogen available at vehicle service stations.

Hydrogen may also be used to power aircraft in the future. In America, gasoline is effectively subsidized (see Section 4) which makes it very hard for other fuels to compete. One of the main interests there has been reforming gasoline and methanol to produce hydrogen. If done in the vehicle this produces a hydrogen supply which contains a high concentration of carbon monoxide. PEM systems can be made to tolerate this impurity. To date alkaline stacks need pure hydrogen.

However, the whole business of on-board reforming is undesirable in terms of cost and complexity and is inefficient in terms of fuel consumption compared to using pure hydrogen made at a central facility. There are two main ways hydrogen can be stored: gas or liquid. As a gas it is usually compressed to 200 bar and stored in steel tanks with man-made fiber reinforcement and carbon additives to assist in the absorption. This technique works for large vehicles where bottles can be roof mounted - buses, for example. As a liquid the energy density is three times that of gasoline, being 57 000 BTUs per lb compared with 19 000 BTUs per lb for gasoline. Two gallons would be needed to travel 500 miles in a 3 liter/100 km (80/100 mpg) PNGV specification vehicle. The gas liquefies at -180°C and 20 bar. In modern super-insulated vehicle tanks, hydrogen can be kept liquid for 2 weeks without refrigeration. A 20 watt Sterling cycle refrigerator can keep it liquid indefinitely. This system is suitable for application where space is limited, such as airplanes and cars. Many people believe the compression process uses too much energy. In fact it is the LIND refrigeration cycle, which is used to take hydrogen down to -269°C from -180°C where hydrogen is liquid at atmospheric pressure, that is the heavy consumer of compressor energy.

Polaron believe the technique permits the early use of hydrogen because tank exchange is possible until the investment in on-board refueling is possible. The tank for a car would only be the size of an outboard-engine boat fuel tank. Cryogenic storage is already well established in the natural gas industry where liquid natural gas (methane) at -160°C is used to fuel 1000 bhp heavy duty trucks in Europe and Japan.

Considering again fuel-cell stacks, in either system an anode or cathode plate is 2.5 mm thick, so a pair of plates give a 5 mm build-up. Each pair of plates gives 1 V at no-load and typically 0.66 V at full load. This means that the stack length is around 500 mm, plus manifolds, for a 64 V, 7.2 kW continuous rating stack. It should also be pointed out that stack power doubles, at least, if pure oxygen is used instead of air. This is unlikely, however, as on-board enrichment to 40% oxygen is promised in the near future, as are some significant improvements in stack chemistry, especially in the catalyst area.

The fuel-cell stack is controlled by regulating the hydrogen pressure in the range 0-30 milli-bars.

A recirculation loop permits water vapor to be added as PEM fuel cells work best with wet hydrogen. The air pressure is regulated by changing the blower speed in conjunction with the fuel-cell current demand. This takes 10 seconds to rise in a low pressure system, but may fall rapidly. The DC/DC converter determines the load applied to the fuel cell.


As we can now see in Fig. 14, a fuel cell is a complex system and the key problems are that the feedstock must be kept pure and power consumption minimized in the auxiliaries. There are two main fuel-cell operating strategies: high pressure, 1-3.5 bar, and low pressure, at 1/20 bar The benefit of high pressure systems is fast hydrogen diffusion in the membrane which results in fast response - less than 1 second. Consequently it is possible for the fuel cell to follow the vehicle load profile and operate without a battery. This strategy is spoilt by warm-up issues.

The stack must be at 70°C to deliver rated power. Warm-up can take 15 minutes. Another problem is that the power to supply the compressed services is significant - perhaps 25% of output at peak power.

Low pressure systems have modest auxiliary power needs, perhaps 10% of rated output at full power and proportionally less at low power. The main consumers are the air compressor and the KOH pump. The price is slower response. It typically takes 10 seconds for the fuel cell to ramp up to full power, consequently a peaking battery is needed to provide power during acceleration. This means that generally a smaller fuel cell may be used.

Fuel cells are the opposite of most electrical devices in that peak efficiency occurs at minimum load. In a high pressure system this profile is ideal for a motorway express coach where most time is spent cruising at 20/30% of maximum power. Hence efficiency is good in cruise and less so in continuous urban cycle duty. However, efficiency and emissions are always better than the equivalent thermal engine. Efficiency is 60% at no-load and 40% at full-load, at the current state of development - and the theoretical maximum is 83%.

Fig. 14 Fuel-cell system schematic.

Fig. 15 Fuel-cell control system.


The cost of the fuel-cell controller (Fig. 15) is split up, at present, as follows: 20% each on pumps and compressors; valves and actuators; programmable logic controller; DC/DC converters; sensors and transducers. The challenge is to achieve a 5:1 cost reduction for mass-market viability.

Fig. 16 Hydrogen/air blowers shown to left of drive electronics.

The hydrogen pump (shown left in Fig. 16) is a side channel blower and has to operate at 1/30 bar at 5 cubic meters/hour with wet hydrogen at 70°C, plus slight KOH contamination. The pressure criterion usually results in a choice of blower made by Gast and Rietschle. The standard unit is a 150 mm cube and weighs 5 kg. The operating point is 2800 rpm and power consumption by the pump is around 85 W, with an additional 36 W of copper loss in the motor. Fuel Cell Control Ltd rewound the 2 pole D56 induction motor as a 4 pole 20 V unit. Our second attempt will be an 8 pole design which should reduce the copper loss to about 8 watts. The low voltage is chosen for safety and the unit will be driven with a linear sine wave inverter (shown center in Fig. 16). The dv/dt is kept low to avoid spontaneous ignition, in case hydrogen enters the motor chamber. The windings are potted, to avoid direct electrical contact and reduce the free volume in the motor chamber. All parts in the hydrogen contact area are nickel plated (zinc-copper-nickel). The blower is made of aluminum.

The air pump (shown right in Fig. 16) is about a 300 mm cube and weighs about 15 kg. The motor is a 1/2 hp D63 induction machine and has been rewound as 8 pole 20 V, 15 A, 325 W at 192 Hz (2900 rpm). The inverter is a switching unit to minimize losses, as the ignition risk is lower than with hydrogen. A 30 amp inverter provides good efficiency and the speed of this blower is adjusted with variations in power demand.

For the future, the company are working on a high speed channel blower, to operate at 10 000 rpm, using a brushless DC motor. In star-winding form, at 4000 rpm, it will satisfy 5 cubic meters per hour and at 10 000 rpm 45 cubic meters per hour. Thus a single design could do both jobs and it only weighs about l kg. However, silencing must be carefully considered.

The water and KOH pumps are standard 10 and 85 watt capacity permanent-magnet brushed DC driven pumps running at 3000 rpm on 28 V DC. The pumps are magnetically coupled with Talcum parts to resist aggressive fluids (such as potassium hydroxide).


Fig. 17 Kit of valves.

Selecting suitable valves with such a diverse array of media and operating conditions has not been easy, Fig. 17. In fact the valves themselves are neither expensive nor heavy. The problem area is the actuators and it is intended to redesign these for the next version. Currently there is no safety legislation in place for hydrogen powered vehicles. The onus is on the supplier to demonstrate fitness for purpose and that all reasonable precautions have been taken. It is felt this will change once meaningful experience has been achieved. Clearly, declaring a vehicle to be a class 1 safety area would destroy all economic viability. Consequently, as with gasoline and propane, safe techniques need to be established and demonstrated before regulations are enforced. Some are obvious, such as no fuel or fuel processes to be contained in the passenger compartment. Others require experience such as fuel storage and distribution. Storage in a closed building needs careful consideration.

The valves can be neatly split into two groups, high and low pressure. The high pressure units are standard metal valves with electric solenoid actuators and spring return; they operate with 28 V coils. The larger units were chosen as 2 and 3 way plastic ball valves, using polypropylene bodies and EPDM seals for KOH compatibility plus high temperature operation (70-80°C).

Polaron had a major problem with the actuators. Fail-safe operation with low power consumption was needed. Solenoid valves in larger sizes use the controlled medium as a pilot fluid and consequently do not operate reliably with pressures as low as 1/30 bar. The solenoids are direct acting, with no economy measures or permanent magnet biasing, and thus consume significant power. In the end nitrogen was used as a pilot fluid, with 4 watt pilot valves to control the opening.

This approach works well but the valves use up a lot of space, especially the actuators.

The intention for the future is to design a plate with spring-loaded clutches operating from a common motor drive. This should cut down on the volume and permit a much lower cost solution.

Actuation accounted for 70% of the cost and 70% of the volume of the valves. It was found to be a niche sector market where nobody has a comprehensive system of interchangeable valves, seals and actuators suitable for onerous conditions.


An 80 I/O PLC with interface modules cost $1500 in 1998. Quantity build could halve this price--but still nowhere near the objectives. A Mitsubishi F Series was chosen for development, Fig. 18. Production units are destined to use a custom-engineered microprocessor unit based on a Siemens/Thompson C167 CAN bus processor, which is becoming a standard in the European and American car industry. This unit must represent one of the toughest design challenges. To convert low voltage, heavy current into higher voltage with galvanic isolation, ultra-low current ripple and high efficiency. Many solutions have been analyzed, although this one offers the best combination of characteristics. Figure 2.19 shows electronic system circuits.

Let us consider a square wave phase-shift chopper: at minimum volts input, we need full reinforcement to achieve 396 V output. However, at low load we have a 96 V DC link and perhaps 90 degrees phase shift between A and B. This is not too bad, except that we only draw output current for 50% of the time: this means that the DC link contains 100% current ripple at 2F switching frequency. Since the pulse width should always be 50% plus, the solution is to have three such choppers with 120° phase shift between them. This has the effect of overlapping the converters if the correct measures are taken. Consequently, the supply now only contains 30% ripple worst case at 6F, but when the current is largest we have maximum overlap; see Fig. 19 (bottom).

The first attempt is to build this converter with each stage operating at 3 kHz, with toroidal 0.08 mm silicon steel cores. This is both silent and efficient. The edges are deliberately softened to reduce dv/dt (capacitive ripple). At low frequency this does not cost much in losses, with 10 microsecond edges, but reduces the spikes when the diodes reverse recover. The design is adaptable to different output voltages by rewinding the output transformers and chokes. It is believed 90% efficiency can be achieved with this design at 60 V input, 7.2 kW. A double L/C Filter attenuates the current to the fuel-cell stack to ensure compliance with 0.01% ripple current rating. The reason for this is to prevent poisoning of the fuel-cell catalysts.

The cost of this unit is a problem. The silicon for the main switches are LAPT transistors, at 15 A and 200 V, using 2SA1302 and 2SC3276; the six switches cost $150 in parts (1998 prices). It is intended to have these parts integrated into a high power package. The chokes cost $150 and capacitors $60. To improve the costs higher frequency is needed; time will tell if this can be achieved without sacrificing efficiency and current ripple.

Fig. 18 Mitsubishi 80 I/O F-series PLC with DAC, ADC and RS232 modules.


This is perhaps the hardest area in which to reduce the cost. The devices operate under hostile conditions and are currently made to order. One success has been to reduce the cost of the 2 kW preheater from $1000 to $200 by a complete redesign. In other areas there has been success in combining functions such as the flow meter and the temperature controller in the KOH loop. A complete new family of low cost sensors is needed before production cost targets can be met.


It is early days for vehicle fuel cells and the main challenge is better, lighter, cheaper, more convenient to use parts - preferably plastics. Insulated packaging of semiconductors is the main issue in the DC/DC converters at low voltage and heavy current. This in time will lead to control systems with lower parts count and greater reliability. As we get down the learning curve, and volumes increase, costs will fall. Major improvements in fuel storage, oxygen enrichment and stack materials should lead to continuing increases in current density and hence smaller stacks for the same power output.

Fig. 19 Electronic system circuits: (a) DC/DC converter; (b) phase-shift chopper; (c) wave phase: full reinforcement (left) and 90° shift (right).

5 Waste heat recovery, key element in supercar efficiency

In the longer term vehicles will be electrically propelled using flywheel storage, hydrogen fuel cells or both. These systems potentially offer high cycle efficiencies, and low emissions vital to improving air quality in our big cities.

Bill Clinton's initiative for US family cars to achieve 100 miles per gallon by 2003 is a serious challenge to the engineering industry; see Section 4. Current solutions include lightweight structures, reduced running losses and small engines. However, the author believes that the target can be achieved using conventional vehicle platforms with low drag floor pans and instead attention is focused on high efficiency drivetrains. Small engines give poor acceleration so to achieve acceptable performance hybrid technology is required. The internal combustion engine car achieves 28% efficiency under motorway conditions and half that on an urban cycle. The key problem is how to convert more of that energy into useful work. The account below will investigate two schemes: (a) turbine recovery system and (b) thermoelectric recovery system.

In both schemes the energy produced is converted into electricity. This is because such an arrangement provides a plausible method for matching the power into the electrical drive. It is accepted that all mechanical solutions are also viable in a hybrid vehicle.


Fig. 20 Parallel hybrid drive Wankel engine and brushless DC motor.

Figure 2.20 illustrates a parallel hybrid driveline using a Wankel engine and a brushless DC Motor.

The package can produce 70 kW peak power and 20 kW average. This combination provides excellent acceleration using energy stored in a small flat plate lead-acid battery. Tests to date show this battery still delivers 100% peak power of 45 kW and 80% capacity after 22 000 cycles to 30% depth of discharge. Thermal management is vital to achieving these figures.

The engine operates on a two stroke cycle and produces high quality waste heat at the exhaust.

The temperature of the gas is around 1000°C. This gas contains 72% of the energy in the fuel. If we can convert a third of this energy into electricity we can nearly double the NTG of the vehicle under motorway conditions.

Why is this important? Hybrid solutions are very effective at improving efficiency under urban cycle conditions but make no impact under motorway conditions. Here is a system that can begin to solve this problem. Supposing in the existing scheme we require 20 kW to operate a vehicle at 70 mph on a motorway. The engine produces 55 kW of waste heat. If a third is converted to electricity 18.33 kW is made available and perhaps 15 kW of this is additional mechanical power.

Consequently the engine only has to produce 60% of the mechanical power to give the same output of 20 kW and the electrical system could be reduced to 10 kW. To summarize, the 20 kW requirement could be met by using 12 kW of engine power and 8 kW of waste heat recovery power. Future schemes could possibly improve on these figures.

5.2 THE TURBINE RECOVERY SCHEME In this scheme the concept is to use a small turbine system in association with an electric generator.

Figure 2.21 illustrates the conceptual realization of the idea. Some modeling of the turbine shows that, for reasonable efficiency, speeds of 150 000 rpm are necessary. If such speeds can be achieved the sizes of the components will be tiny. For example, for 10 kW, the rotor of the generator will be 25 mm diameter by 15 mm long. The generator needs to be kept separate from the turbine stages due to the high temperatures involved in the thermodynamic processes. The turbine bearings will be hydrodynamic gas bearings - the back of the turbine rotors and the shaft will be plated with a zig-zag pattern, microns thick, designed to created turbulence at high speeds. Dynamically the system will operate below the first critical speed.

The generator bearings will be angle contact bearings (6 mm) using ceramic balls (RHP-INA) and Kluber Isoflex Super LDS 18 grease. The rear bearing is preloaded and free to slide by means of a crinkle spring. The coupling between generator and motor can be a tongue and fork, permitting easy removal, as the torque is below 1 Nm.

Fig. 21 Block diagram of turbine waste heat.

The generator losses preheat the air entering the first compressor stage. The generator has laminations of 0.2 mm radiometal with powder coat insulation applied to ensure minimum eddy current losses. The rotor has a one-piece tubular magnet, 22 mm diameter and 15 mm long by 3.5 mm thick, of 'one-five' samarium cobalt. This is glued onto a stainless steel shaft of high resistivity. The magnets are retained by a prestressed carbon fiber ring of 1.5 mm wall thickness.

The generator has a 4 pole configuration and the machine winding is designed to give 165 V (RMS) at 150 000 rpm, resulting in a line current of 42 A at 10 kW. An advantage to this method of construction is that the generator may be built and tested separately from the turbine.

The turbine rotors will be only 40 mm outside diameter and machined from aluminum. The rotors are held to the shaft by Loctited nuts and there is a hole down the center to facilitate temperature measurement. One of the nuts contains the fork for the drive coupling.

The design of these stages with their casing and expanders is confidential. The heat exchanger is an air to air unit rated for 30 kW at a temperature of 600°C. The same unit also functions as an exhaust silencer for the engine and special construction techniques are required to resist the high temperature of the exhaust from a Wankel engine (typically 1000°C). Polaron envisage a battery of 216 V nominal varying between 180 V and 255 V. The speed of the turbine may vary over the entire range but meaningful output will only occur between 120 000 and 150 000 rpm.

The turbine, Fig. 22, is started by a transistor bridge connected across the diode bridge and as the compression of inlet air starts, and is expanded, the output turbine takes over supplying rotational power, and the transistor bridge is then used as a switching regulator, to match the generator voltage to the battery voltage. Sensorless timing techniques are possible but it is simpler to use three Hall sensors operating from the rotor field system.

Fig. 22 Turbine recovery system and recuperator power control.

You might ask: Why not let the load line of the turbine generator intersect with that of the battery on an open loop basis? The problem is that in most cases one would not obtain the correct operating point. The turbine power is proportional to speed cubed. One obtains the correct operating point at just one speed for a given power, whereas the battery operates from 1.75 to 2.35 V per cell. Consequently it is necessary to have closed loop control of the power flow from generator to battery. But there is a second reason; this mode of control with the transistor bridge permits the turbine to be used as a brake - power flow is reversible between turbine and battery. This is very useful when negotiating long steep gradients, for example.

Overall it is believed that an efficiency of 30% is achievable with such a process and thus the system can make a major contribution to fuel utilization under motorway conditions.

5.3 THERMOELECTRIC GENERATOR The turbine recuperator technique involves some very high technology mechanics to make the system work. It prompts the questions: Is there any other way of achieving the same objective? Is a solid state solution possible? Thermo-electrical devices were invented in 1821 and are perhaps best known today for the small fridges we have on our cars and boats to cool food and drinks. An array of bismuth telluride chips 40 mm square can produce 60 watts of cooling with a temperature differential of 20°C. If we go back 60 years to the 1930s there were thermopiles which one placed into a fire and the pile provided the current for a vacuum tube radio. It is only very recently that here in the UK a group of engineers started to ask the question 'Why are thermopiles so inefficient?' What happens to the 96% of the energy consumed that does not appear at the output terminals? Why is the output voltage so small - typically microvolts per °C at top temperature? At Southampton University Dr Harold Aspden soon identified the answer to the efficiency question. The energy was being consumed by circulating currents within the device. It was then realized that if a dielectric was placed between the thermopile layers, and the pile was oscillated mechanically, that an AC voltage could be obtained up to 50 times the amplitude of the original DC voltage, Fig. 23. This oscillation has been tested with frequencies from DC to RF and the process holds good across the spectrum. Dr Aspden has concentrated his efforts on producing thermopile arrays for use on the roof of a building, with temperature differentials of 20-40°C.

However, if we return to our waste heat recovery problem we are dealing with top temperatures of 600°C plus and consequently alternative materials will be required and the number of stages in series to produce a given voltage will be reduced. But, with a top temperature of 30°C existing, devices can convert 20 W of power with an efficiency of 25%. It should be emphasized that this work is at an early stage of development at this time.

The thermopile elements suitable are iron and constantin 40% nickel/60% copper (Type J thermocouple material); at 600°C, with mechanical excitation, a voltage of 300-500 mV per stage can be achieved, hence 500 cells in series would produce 216 V DC. The circulating current in each cell is proportional to the temperature difference but the output AC voltage may be controlled by adjusting the amplitude of the mechanical excitation. The most interesting point is that to give 10 kW a suitable unit could be very compact - our calculations suggest about 100 mm cube. We believe the mechanical excitation is best supplied by ultrasonic piezoelectric transducers driven by a HiFi amplifier. The power required is around 200 watts. One interesting point is that the unit offers reversible power flow. How? It can be converted from refrigerator to heater and act as a braking device.

Fig. 23 Aspden thermo-generator and its control system (below).

Top of Page

PREV.   NEXT.   Index HOME