Advanced thermal modelling and management techniques to improve power density in next generation power electronics

This thesis sets out a series of new techniques to improve the thermal management of power electronics. The work is motivated by the increasing impetus to design smaller, more energy efficient electronic power systems for a range of applications, notably electric vehicles. Thermal management is an increasingly important tool which can facilitate improvements in power density through better monitoring and control of system temperatures. This thesis seeks to deliver improvements in implementing this strategy.

A review of the state of the art in thermal management is reported, focussing on temperature measurement, thermal characterisation and system modelling techniques. In addition, novel techniques for arbitrary dissipation control and die temperature measurements in semiconductor devices are presented. A novel analysis of the limitations of low-order thermal models is also described. Improvements and applications of these techniques form the basis of this thesis.

The pseudorandom binary sequence (PRBS) technique for system identification is applied throughout the thesis to characterise thermal systems. A mathematical analysis is provided, together with a novel technique to determine the minimum gain which can be identified by PRBS techniques in the presence of noise. A novel improvement to the PRBS technique for typically ten times more noise resilient measurements is then developed based on mathematical mixing of different frequency PRBS signals. In parallel, a novel technique is formulated to estimate the temperature throughout a multiple device system using digital IIR filters and PRBS thermal characterisation, which achieves errors of 3-5% when demonstrated practically. By combining these techniques, a comprehensive temperature estimation and control methodology is implemented for a multiple device system under active cooling. Finally, the expansion of the proposed methodologies to steady-state die temperature estimation is presented with comparable accuracy to surface temperature measurements, increasing the usefulness of the developed techniques in a practical setting.