Wavelet-based prognostic-oriented temperature sensing with sigma-delta ADCs in power applications

Thermal stress is one of many possible failure causes of power electronics systems; thermal cycles are known to produce mechanical fatigue on power electronic devices, thus leading to their failure in time. Temperature swing can be somehow controlled, if some decrease in efficiency can be tolerated by the power system, changing gate driver parameters to heat up the device in low load conditions. This is known as Active Thermal Control (ATC). To implement ATC, the temperature information of devices is needed. Since this sensing is carried out near power conductors, switched at high frequency, strong disturbances will affect the reading. This paper proposes a new signal-conditioning scheme based on discrete wavelet transform (DWT) and compares it to some established filtering techniques to assess which is most suitable for the application. Hints about insulation of measuring equipment will be given, too. Moreover, the possibility offered by the algorithms to store long-Time temperature information (through compression of incoming data) will be described, as large-span observation can help prognostic algorithms in forecasting the remaining useful life of the power apparatus.