Resumen
A sensorless online temperature estimator is presented in this paper, which estimates the temperature using a novel signal injection strategy. This allows to eliminate the temperature sensors in the machine, as well as their faults, increasing the system reliability. A double dead-time DC signal is injected in the machine, adding a controlled offset in the control drive through the inverter. The proposed strategy eliminates the effect of the dead-time in the injected signal, which is an important drawback in DC injection strategies for resistance estimation. Furthermore, additional hardware is not needed. The strategy has been implemented in an inverter-fed railway traction induction machine. The proposed algorithm has been validated in a real test-bench.