The knowledge of the rotor position and the rotor velocity of an electrical machine is typically required to design effective controllers for AC machines. It can be measured with an encoder or a resolver. However, the absence of such sensors is preferred in many low to medium power applications due to reliability or cost considerations. Thus, techniques are necessary to estimate the rotor position and/or speed. This estimated value is then used to control the electrical machine.
The typical approach is to use an observer or estimator based on the dynamic system model to reconstruct the back-EMF vector. This information can then be used to estimate the rotor position, e.g. with a PLL loop. This approach is sufficient for applications that do not require a high torque at low speed, e.g. fans or wind power. However, traction or servo drives typically require maximum toque at zero speed. For machines that have a saliency (at least at high-frequency), a high frequency disturbance can be “injected” into the machine and the response can be used to assess the rotor position.
The quality of a position estimation depends on the estimation technique and the quality of the low and high frequency system models. To achieve high precision estimation, high-fidelity models need typically to be generated for a specific drive system. Moreover, the resulting estimator or observer needs to converge sufficiently fast and independent from the initial error. Also, an observer and controller can be designed and implemented independently form each other without affecting their stability. However, some fine tuning is typically required to achieve the best overall results.