More than 95% of these motors were rated at less than 1hp (745W) and a very large proportion of them are AC motors – renowned for being power hungry and inefficient. Better control techniques, however, can improve the efficiency of AC motors to the degree that, with the right dsp algorithms, they become cheaper to run than more complex and expensive dc motors.
A later ACEEE report estimated some 20% of motors in use in 2006 used electronic control, providing an energy saving relative to 1976 of around 40TWh. That equated to roughly 2% of the electricity consumed by motors. By 2030, the ACEEE report estimates the savings will exceed 100TWh. By then, electric vehicles will have become more prevalent, providing another load on the electricity supply.
Some motors stand to make greater gains than others: in a number of markets, the conversion to more efficient control has already begun. Motors in large household goods once used wire brushes to maintain constant power contact with the motor's stator, incurring relatively large power losses from friction and resistance.
These are gradually switching to permanent magnet motors that use solid state power and feedback devices to replace the brushes, with better pulsewidth modulated voltage control. Industrial motors have tended to fare better in terms of overall efficiency, largely because the payback in reduced energy can be quick – so polyphase induction motors rated at more than 100hp are now common.
The ACEEE regards drive systems as being the 'frontier of energy saving possibilities for semiconductors' – mostly to reduce the variability in supply voltage that can lead not just to lower efficiency, but also lifespan. A number of types of drive control are in action today. The most basic is the volts per hertz technique, generally used on fans and pumps.
It is cheap to implement on a basic 8bit microcontroller and avoids one of the biggest problems of using the simplest control technique for an AC motor of just controlling motor speed by changing the applied frequency. The flux, magnetising current and torque all depend on a built in ratio of a motor's design, expressed in V/Hz. Increasing frequency without a corresponding increase in voltage will boost speed, but reduce the torque.
The way around that is to alter the applied voltage in line with frequency so that torque can be maintained with every change in speed. The relationship between voltage and frequency tends to break down at low speeds. To provide sufficient torque at low speeds, a voltage boost is often needed from the motor drive. But, if the load is also low, then the motor can saturate and potentially overheat.
This is where the more advanced vector control techniques can make their mark. With flux vector control, the speed and torque relationship can be separated, with each governed by a different control loop based on a model of the motor's operation. However, a big problem with ac induction motors is that the speed of the rotor does not match the speed at which the magnetic flux driving it rotates. Instead, the mechanical speed tends to lag – the difference is known as the slip speed.
Early vector control techniques relied on encoders and resolvers to work out where the rotor is at any point, but they cost money and provide a further point of failure, affecting overall reliability. It is possible to estimate the motor position using stator currents and voltages, although the accuracy of this technique is not very good at low speeds without further signal processing.
This is where more sophisticated vector control algorithms come in – they provide a more accurate way of working out where a rotor is. In general, the more processing power you throw at the problem, the more accurate this estimated position will be and the more efficient you can make the motor.
Field oriented control (see fig 1) is the most maths intensive technique used today but provides the best speed and torque control available for ac motors.
It provides dc performance for ac induction motors, so lets designers replace dc motors with often cheaper ac alternatives. The stator in an ac induction motor has three windings that can be controlled independently – stators with single phase windings cannot be controlled as accurately. The stator moves the rotor as the result of the sum of the force from the three phases.
The coils can be driven to produce torque or apply force along the axis of the stator that does not affect rotation. These are, in motor control theory, the quadrature and direct axes, respectively. If you want to generate rotation, you need to maximise quadrature force while keeping direct force to a minimum. Field oriented control translates the electrical readings and model into a direct quadrature coordinate system, normally fixed to the rotor, to ease the calculations for the desired flux.
However, the calculations – typically the Park and Clarke transforms – are mathematically intensive. The Clarke transform takes the three phase currents and uses them estimate currents in a Cartesian coordinate system. These are then taken by the Park transform to provide currents that make sense in the direct quadrature coordinate system. These currents, together with the flux angle, are then used to calculate the electric torque of the motor.
The slip in a sensorless motor can also be estimated from the measured currents. The accuracy of this estimation depends again on the complexity of the mathematics. Much of the effort in advanced motor control is now going into this area as it provides the greatest gains in efficiency – the closer the slip estimate is to reality, the less power is wasted. Traditionally, one of the most common techniques for estimating actual motor speed uses sampled phase voltages and currents fed into an adaptive motor model to estimate the reactive power.
The advantage of this technique over other proposals is that it does not need any knowledge of stator resistance, which is prone to change with temperature. Not having to know the stator resistance therefore makes the model more tolerant to changes in motor conditions. However, the relationship between motor speed and the measured voltages and currents of the individual phases is not simple.
The model reference adaptive system deals with that problem indirectly by calculating the reactive power using two different sets of equations. There are two models in the system. One is a reference model that calculates the reactive power at a point in time without any reference to the mechanical speed of the motor – it relies on the back electromotive force of the motor. A second adaptive model uses an estimate of motor speed.
A closed loop proportional integral (PI) controller drives the adaptive model to produce the same reactive power output as the reference model and to produce a more accurate estimate of the motor's rotational speed. A big problem with this technique is that it does not account for mechanical losses and, as with other approaches, it does not have good accuracy at low speeds. More powerful dsps are gradually making it feasible to use a potentially more accurate system: the extended Kalman filter.
The great advantage of the Kalman filter is that it is good at removing the influence of noise from the model. With a Kalman filter, the estimated variables are corrected with a predictor that relies on a set of state and variance matrices. The matrices themselves are derived from a relatively complex system of equations.
The problem with using the Kalman filter is that it relies on matrix inversions and a large number of matrix multiplications – greatly increasing the number of operations needed per second to derive an accurate correction for the slip speed. As dsp performance improves and chipmakers add hardware coprocessing support for matrix operations, the Kalman filter is likely to become a more widely used technique.
Direct torque control (see fig 2) provides an alternative to field oriented control. Instead of using coordinate transforms and a current based model to estimate flux angle, this technique applies two hysteresis comparators and a look up table of switching states. This results in a relatively simple implementation. Without the PI controller, response time is better but the variation in flux and torque can be large.
One approach used to improve accuracy was to apply fuzzy logic. An alternative is to use space vector pulsewidth modulation, using the measured current and voltage applied to a model of the motor, to derive an output voltage level. This sounds simple, but relies on the ability to solve a system of quadratic equations. Field oriented and direct torque control offer advantages in different circumstances.
Typically, direct torque control wins when transient response is important. Field oriented control tends to offer better efficiency with unpredictable load. As the mathematical understanding of these systems improves – and more hardware support is added – that situation may change.
