Many flywheels are built with composite rotors because of the higher strength-to-weight ratio of composites versus steel. Unfortunately, the strength advantage is reduced by the additional mechanical complexity of a composite rotor and the safety derating factor necessary for composites. The derating is necessary because the failure behavior for composite materials is more difficult to predict than for steel. Composite rotors also impose a stricter limit on the rotor operating temperature and temperature cycling.

Furthermore, while a composite flywheel of a certain energy storage capacity may have a lighter rotor than a steel flywheel system of the same energy storage capacity, the rotor itself is only a small fraction of the overall flywheel system weight. Since the remainder of the system, such as the stator, containment, and electronics, remains basically the same, the reduction in weight is a very small percentage of the total system weight. This is particularly true for high power flywheels designed for discharges in the range of tens of seconds. Since many flywheel systems are aimed at power quality and hybrid vehicle load-levelling applications that require short, high power discharges, many of them fall into this category. However, because the mass of the electromagnetic rotor and stator is a function of power, for flywheels with a high ratio of power-versus-energy-storage, the electromagnetic portions of the rotor and stator will comprise an even larger portion of the system mass, further reducing the advantage of composite rotors.

In a similar manner, the vacuum and burst containment necessary for composite rotors also offsets their advantages over steel rotors, especially when compared to an integrated design where a large portion of the containment structure is comprised by the motor stator and housing.

Perhaps the most attractive feature of a steel rotor integrated flywheel design is that it is similar in material composition, design, and complexity to a standard electric machine, therefore, it is reasonable to expect that for volume production the manufacturing costs of such a flywheel would be comparable to those of a similar-sized electric machine. The cost of a standard electric machine can be estimated based on its weight, and, using this method, the cost of a steel rotor integrated flywheel design could be much lower than other designs.

The integrated flywheel configuration can achieve comparable specific energy to a composite rotor flywheel system. While a comparison between a composite rotor and steel rotor of the same weight (disregarding all the other components of the flywheel system) might indicate that twice as much energy could be stored in the composite rotor, the additional housing and containment weight for the composite rotor in a conventional or barrel configuration versus a steel rotor in an integrated configuration offsets this advantage to a certain degree.

As an example, consider a composite rotor 100kW and 420W·hr flywheel energy storage system with a rotor of mass 40kg and total system weight of 545kg.1 The rotor of a steel rotor flywheel with the same energy storage available might weigh two times more, but since the rotor weight is a small fraction of the total system weight and the weight of the remainder of the system remains basically the constant, the steel rotor system would have a specific energy only 7% lower. It would be incorrect to say that a well-engineered composite flywheel energy storage system could not have a higher specific energy, but it is argued that a steel rotor flywheel in an integrated configuration could have comparable specific energy at lower complexity and lower cost.

A high power integrated flywheel system with a solid-steel rotor can achieve similar performance to a composite rotor flywheel system, with less complexity, lower manufacturing costs, and lower material costs.

High efficiency, a robust rotor structure, and low rotor losses are the key requirements for a flywheel system’s motor/generator. High efficiency is required so that the flywheel can be an effective energy storage medium. Motor efficiency must be high over the entire speed and power range of operation, in this case 50,000rpm to 100,000rpm, with a power rating of up to 30kW. In addition, the zero power spinning losses of the machine must be very low. A robust rotor structure is necessary for obvious reliability and safety reasons.

The third requirement of low rotor losses is also critical and drives many of the design decisions in a flywheel system. Because high-speed flywheels operate in a vacuum to reduce windage losses, the main paths for heat transfer from the rotor are through radiation and through the bearings (if ball bearings are used). The amount of heat transfer through radiation is small except at high temperatures, and the thermal path through the bearings is minimal; therefore, controlling the rotor losses is critical to prevent overheating of the rotor. Fig. 1 plots the rate of heat loss through radiation versus rotor temperature. This graph was calculated for the prototype rotor, and assumes that the stator temperature remains at 50˚C. From this, it can be seen that even 300W of rotor loss (1% of the 30kW output power) would lead to a steady-state rotor temperature of 320˚C. This thermal limit on rotor losses rules out machines such as induction machines and switched reluctance machines because of their respective conduction and core losses on the rotor. The remaining motors to consider are permanent magnet and synchronous reluctance.

Permanent magnet (PM) motors are currently the most commonly used motors for flywheel systems.2, 3, 4 While they have the advantages of high efficiency and low rotor losses, the presence of PMs on the rotor makes the rotor more temperature sensitive (thus requiring even lower rotor losses) and the mechanical structure of the rotor more complicated, because of the brittleness and low strength of the PMs. The cost of the PMs, especially the high-temperature Sm-Co type, can also be considerable.

Synchronous reluctance (SR) motors are also used in some flywheel systems. They can also have high efficiency, low rotor losses, and low zero torque spinning losses. Unfortunately, it is difficult to construct an SR rotor with a high Ld/Lq ratio while maintaining a robust rotor structure. Examples of SR rotors constructed using several axially bonded sheets of high strength steel as in5 and6 have been made, however, this leads to a moderate Ld/Lq ratio which in turn leads to a moderate power factor. Low power factor increases the required VA rating of the drive and can add significantly to the cost of the system.

All three of these motor types (PM, SR, and homopolar inductor) share the advantage of high efficiency, however, PM rotors tend to be more temperature sensitive, mechanically complex, and costly; and solid rotor SR motors have either complex rotor structures or low power factors. While these issues can be overcome, homopolar inductor motors present an attractive alternative with a low-cost rotor machined from a single piece of steel that is more robust and less temperature sensitive than PM or SR rotors. In addition, a homopolar inductor motor with a slotless stator and six-step drive eliminates stator slot harmonics and maintains low rotor losses while also allowing operation at unity (or any desired) power factor.

These advantages of robust rotor structure and low rotor losses make homopolar inductor motors particularly well-suited for flywheel energy storage applications.

Although not widely used in practice, homopolar inductor motors have been researched for a variety of applications. They are sometimes referred to as “synchronous homopolar motors”,7, 8, 9 or “homopolar motors”,10, 11 but “homopolar inductor motor”12, 13, 14 is also commonly used and will be the term applied in this thesis. The defining feature of these motors is the homopolar d-axis magnetic field created by a field winding,15, 16, 17, 18, 19 by permanent magnets, or by a combination of permanent magnets and windings.20 The principle is the same as in a traditional synchronous generator, with which the homopolar inductor motor has similar terminal characteristics. However, in the case of the homopolar inductor motor, the field winding is fixed to the stator and encircles the rotor rather than being placed on the rotor. The field winding and the magnetizing flux path in the present motor design are shown schematically in Fig. 3. Note that the rotor pole faces on the upper part of the rotor are offset from the pole faces on the lower part (see Figs. 3 and 24).

There are several advantages to having the field winding in the stator. Among these are elimination of slip rings and greatly simplified rotor construction, making it practical to construct the rotor from a single piece of high strength steel. This feature makes homopolar motors very attractive for high-speed operation; a single piece steel rotor is used in the design presented here and in.21, 22, 23 Other homopolar rotor designs include laminations,24 permanent magnets,25 or other non-magnetic structural elements to increase strength and reduce windage losses.26 Additional advantages of having the field winding in the stator include ease in cooling the field winding and increased volume available for this winding. The large volume available for the field winding allows high flux levels to be achieved efficiently, making a slotless stator design feasible.

The slotless stator is an advantage for solid rotor machines because it eliminates slotting induced rotor losses.27 A slotless stator also allows for higher gap flux densities because saturation of the stator teeth is no longer a concern. The design principle is similar to a slotless permanent magnet machine, with the advantage that the magnetizing field can be controlled to keep efficiency high at low and zero torque. A possible disadvantage of the slotless stator is the difficulty in constructing the armature winding, which must be bonded to the smooth inner bore of the stator iron. A relatively simple and effective process was developed to construct the winding.

Flywheels have very demanding requirements for their bearing systems because of the high rotational speeds and operation in a vacuum. Since it is impossible to balance a high-speed rotor so that the center of mass and the bearings’ axis of rotation coincide perfectly, it is often preferable to mount the flywheel rotor on compliant bearings. Compliant bearings allow the rotor to translate radially, allowing the axis of rotation to shift, and thus allowing the rotor to spin about the center of mass. This dramatically reduces the forces on the bearings caused by rotor unbalance.

Compliant bearings can be achieved either by using a magnetic bearing which levitates the rotor with magnetic fields or conventional ball bearings in a compliant mount. Magnetic bearings have the advantage that they can be controlled electronically, and their stiffness and other performance characteristics can be tuned during operation. They are also vacuum compatible and capable of operating over a very large temperature range. One small disadvantage is that magnetic bearings are larger than conventional bearings, but their primary disadvantages are their cost and complexity. Unfortunately, there are several disadvantages inherent to using ball bearings in a flywheel. Although there are standard ball bearings designed for operation at speeds up to 100,000 rpm, their lifetime is reduced by operation in a vacuum. The grease used to lubricate these bearings is not vacuum compatible, and thus it volatizes as the bearings heat up and eventually leads to bearing failure. There are vacuum compatible greases, but they are generally high viscosity or for low temperatures and unsuitable for this application.

PROTOTYPE

The rotor for the machine consists of a single piece of high strength steel. As shown in Fig. 3 and Fig. 26, four poles are cut into both the upper and lower parts of the rotor, with the lower poles rotated 45 degrees with respect to the upper poles. The center portion of the rotor is cylindrical, and the field winding encircles this portion of the rotor. The four upper poles are all the same magnetic polarity (N), and the flux returns down through the backiron to the lower set of poles (S). The machine has 8-poles, and no saliency, i.e. Ld/Lq = 1.

Two different rotors were built, as shown in Fig. 26. The profile for each rotor was analyzed using FEM (finite element method), and a shape optimization was applied to achieve the desired rotor MMF profile. For the sinusoidal rotor, the goal was to design a rotor to achieve a sinusoidal MMF waveform with no harmonics. A minimum and maximum magnetic gap were specified, and a 2D FEM analysis was conducted to determine the MMF waveform for a given rotor profile. The rotor profile was modified iteratively to minimize the harmonics. Fig. 24 shows the sinusoidal rotor profile and the resulting MMF profile.

For the square-cut rotor, FEM analysis was used to adjust the size of the pole-face arc with parameter α (as defined in Fig. 4) so that the rotor would have an Ld/Lq = 1. Since the FEM analysis takes into account the effects of fringing flux, the pole faces were adjusted so that they spanned 42˚ instead of the full 45˚ if fringing flux was ignored.

The two rotor shapes were built with the intent of comparing their performance. The sinusoidal rotor has no rotor MMF harmonics so its associated stator core loss would presumably be smaller. However, it also has a lower Lmf which would then require more field winding excitation to achieve the same flux level. Unfortunately, the square-cut rotor was damaged during a bearing failure before complete experimental results could be obtained, and a comparison was not possible.

The housing consisted of a carbon steel tube that contained the stator stacks and field winding. Also included were the endplates that were made out of stainless steel in Fig. 27.

One of the endcaps was modified to accommodate an infrared thermocouple which was used to measure the temperature of the rotor. All the wiring was connected through vacuum-sealed feedthroughs, which were all located on one endplate. The same endplate also contained a vacuum connection to connect to the vacuum pump.

The stator was made from 0.005” thick laminations, stacked and press-fit into a steel tube that serves as both the back iron and housing for the machine. Only the field winding flux, and not the alternating flux of the armature, travels through the back iron, so core loss in the back iron is not an issue.

The field winding was wound around a bobbin and also pressed into the back iron. Fig. 28 shows the field winding and the split bobbin that it was wound around. The bobbin was made out of aluminum and split into two sections to improve heat transfer to the stator stacks and the housing.

The most challenging part of the prototype construction was the winding of the stator armature. The armature was formed from rectangular Litz wire bonded to the inner bore using thin sheets of FR4 prepreg. FR4 prepreg is the partially cured form of the yellow-green epoxy-fiberglass laminate commonly used as printed circuit board substrate. The type of FR4 employed here was very thin (roughly 63.5μm or 0.0025”). A diagram of the construction assembly is shown in Fig. 5. First, a layer of FR4 was placed against the inner bore, followed by the windings, and then an additional layer of FR4 on the inside of the windings. Then an air bladder was inserted and inflated to 1 atm. (15 psi) to compress the FR4-Litz wire-FR4 assembly against the inner bore. The stator assembly was then baked in an oven to reflow and fully cure the epoxy in the FR4. After baking, the air bladder was removed, and the result was a smooth and solid winding structure bonded tightly to the inner bore of the motor. After the windings are bonded to the inner bore, the endturns are potted using 3M DP-105 epoxy. Photos of the completed armature are shown in Fig. 29 and 30.

A compliant bearing mount was designed and built to allow the rotor to operate at high speeds. The mount consisted of a tolerance ring28 clamped around the outside of the bearing and compressed into a bore. The tolerance ring, (Fig. 6), is a band of spring steel with ridges that flex to provide compliance.

Angular contact ball bearings with ceramic nitride balls were used (Barden model number CZSB101JSSDL). An axial preload is necessary for angular contact bearings to operate properly. In this design, a 20 lbf. preload was provided by an axially loaded wavespring mounted in one of the endplates.

Fig. 7 diagrams the principle components of the inverter, sensing, and control electronics. The 3-phase inverter consisted of three 600V, 200A IGBT half-bridge packages (Powerex PM200DSA060). The IGBTs were driven by opto-isolated gate driver chips, which were driven by logic level signals generated by the Altera FPGA. The inverter was connected to a dc bus fed by a 6-pulse diode rectifier, which was supplied by a 3-phase variac or directly from the line. When discharging the flywheel, a dump resistor (not shown in the diagram) was used to dissipate the excess power from the bus. The field winding for the inverter was powered by a dc-dc converter supplied by a separate 200V power supply. A description of the dc-dc converter can be found in.29

Current sensors were put in place to measure each of the three motor phases ia, ib, and ic; the field winding current if; and the dc bus current into the inverter ibus. Voltage measurements of phase voltages va, vb, vc, and vbus were also available. The analog sensing signals were filtered through an analog filtering box which performed a 3-phase to 2-phase conversion on the armature current and armature voltage measurements. The analog filter box is described in footnote 30.30

The analog outputs of the analog filter box are sampled by A/D converters on the dSPACE DSP card. The dSPACE processor executes the control algorithm, and calculates up- dated ωe-command and if-command. The ωe-command is passed to the Altera FPGA, which uses it to generate the gate drive signals. The if-command is converted into a PWM gate drive signal, which is then passed to the dc-dc converter.

Fig. 8 diagrams the three main processes running in the dSPACE controller. The inverter switching and the generation of a sampling interrupt are both handled by the Altera FPGA, which coordinates the timing with the dSPACE card with control interrupt int0 and sampling interrupt int1.

Sampling interrupt int1 triggers the sampling process sample fcn in the dSPACE card at a rate of 6ωe. Samples of armature currents and voltages are taken by the A/D converters just before inverter switching occurs, so the orientation of the measurements relative to the inverter voltage is known. These samples (which are measured in the stationary frame) are then rotated into the synchronous reference frame, filtered, and then made available to the motor control process io fcn. Since the sampling process sample fcn runs at a rate of 6ωe, it is typically running 4 to 16 times faster than the control process io fcn. Thus, the filtering has only a small effect on the delay of the controller.

The control process io fcn calculates updated commands for ωe and if based on the sampled currents id and iq. This process is triggered by int0 at a fixed rate of 1.5kHz. The interrupt int0 also synchronizes the communication of the ωe command back to the FPGA. The if command calculated in io fcn is passed directly to pwm fcn, the dc-dc converter control process. This process is triggered off an internally generated 50 kHz clock, and in each cycle, pwm fcn samples if, calculates a new duty cycle for the dc-dc converter, and generates the appropriate gate drive signal.

For safety reasons, the high-speed testing of the flywheel system was conducted in a containment chamber mounted in a pit beneath the surface of the floor. The pit was approximately 3’ deep and 4’ by 4’ in area. The motor was mounted on a steel plate bolted directly to the concrete floor of the pit, and a 3” thick steel pipe encircled the flywheel system. Sandbags were placed between the outside wall of the pipe and the concrete wall of the pit. Two 1” thick steel plates and sandbags were placed on top of the assembly with a hydraulic lift and bolted down with eight 1” diameter steel rods. When completely assembled, the top plates are just below the level of the floor. Cabling for the motor is run through a feedthrough at the bottom of the steel pipe and around the top plates. Photos of the motor in the containment pit are shown in Figs. 31-32.