Initial Temperature Determination for Real-Time Thermal Models in Permanent Magnet Synchronous Machines

The applications of real-time thermal models (RTTMs) to sensorless temperature monitoring are limited due to unknown initial temperatures. To address this issue, this article proposes a method to determine initial temperatures for all main components, only relying on one thermal sensor installed in a specific component, ideally in the active winding (AW) or end winding (EW), where the measurement practicality and the signal-to-noise ratio (SNR) in the initial temperature estimation process are all considered. In the proposed method, the measured temperature of the specific component in each sampling step is integrated into the RTTM to predict its next-step temperature, and thus, its temperature deviations between the predicted and measured values contain the initial temperature information of the other critical components. Based on several sampling steps, their initial temperatures can be estimated with high estimation accuracy. Afterward, the estimated initial temperatures can be used to correct the temperature states of the employed RTTM, and thus, the temperature errors caused by the inaccurate initial temperatures can be eliminated in a short time. The proposed method is applied to a well-trained four-order RTTM for an interior permanent magnet synchronous machine (PMSM) and validated experimentally.