Frequency domain temperature model :a new method in on-line temperature estimation for power modules in drives applications

Abstract

The operating temperature of the components within an electronic device has a significant impact on the reliability of a product. In a variable speed drive the power semiconductors in the inverter stage are often operated close to their maximum temperature when the inverter is operating at a low output frequency or during an overload. The temperature of these components must be continuously monitored to prevent them from overheating, but direct measurement of the temperature is only possible if a special test configuration can be used. This is not practicable in a commercial drive and to protect the inverter the temperature of the power semiconductors must be estimated by an on-line thermal model.

The work presented in this thesis describes the development of a novel thermal model that can be implemented using the existing computational resources available in a commercial variable speed drive. The thermal model is based on the transient thermal impedance measured between each device and the internal thermistor in a power module. These form a thermal impedance matrix which can be used to calculate the instantaneous temperature of every device in the inverter. However, with the existing computational resources it is not possible to implement the complete matrix without aliasing. To reduce the risk of aliasing the number of calculations performed during each sample period must be reduced. This is achieved by using a frequency domain model that has been developed to calculate the peak temperature of the hottest devices.

To validate the thermal model it has been implemented in a commercial drive. The drive has been modified to allow the temperature of the power semiconductors in the inverter to be measured using a high speed thermal camera. This allows the temperature estimated by the on-line thermal model to be compared directly with the temperatures measured when the inverter is operating under typical load conditions. Comparisons of the measured and estimated temperatures in several operating conditions are presented. These conditions were chosen to highlight the advantages and disadvantages of the frequency domain model.