Torque Estimation for Speed Sensorless Induction Motor Control With Virtual Voltage Injection at Extremely Low Speeds

The sensorless control of induction motors suffers from issues such as poor stability and insufficient load capacity in the low synchronous speed region. Traditional signal injection methods are not suitable for induction motors with weak anisotropy. The virtual voltage injection method can solve the instability problem in the low synchronous speed region. However, this method also results in an increase in the d-axis current required for the motor under heavy load, thereby increasing the no-load and light-load losses of the motor. Traditional torque estimation methods cannot accurately estimate torque at extremely low speeds after virtual voltage injection. Therefore, this article builds and trains a load torque estimation network for the sensorless control of induction motors with virtual voltage injection. Then the network is implemented in the control chip through optimized design, achieving edge intelligence. Furthermore, an adaptive d-axis current reference control scheme is designed based on the estimated torque. On the basis of ensuring the stability improvement effect of virtual voltage injection on the control system, the d-axis current reference can adaptively change with the load, thereby improving the motor efficiency. Finally, the feasibility of the proposed method is verified by experiments.