Analysis of Power Processing Architectures for Thermoelectric Energy Harvesting

This paper analyzes, from a power processing architecture standpoint, the recovery of thermal energy waste by means of thermoelectric (TE) modules and arrays. The existence, in many industrial scenarios, of stable and often significant temperature gradients enables a number of possibilities for effective processing and recovery of waste heat, which could result in marked economic savings and environmental benefits if adopted on a large scale and systematically embedded into industrial processes. A review of the TE source is provided first, along with an extensive experimental characterization of commercial bismuth-telluride TE cells. Results indicate that maximum power extraction from TE generators can be achieved at a TE efficiency close to optimal, motivating the adoption of maximum power point tracking architectures, traditionally employed in photovoltaic systems, to TE plants as well. Three power processing architectures are then analyzed and compared in terms of their maximum power extraction capabilities and of the efficiency constraint they pose on the power processors. Different from the photovoltaic case, the simple series configuration of TE modules allows to extract most of the available power even in the presence of rather large mismatches among the modules. For even larger mismatch levels, the differential power processing (DPP) concept, already introduced for dc-dc distribution systems and photovoltaic plants, can be successfully adopted to improve power extraction. On the other hand, the module-integrated converter architecture, another well-established solution for photovoltaic sources, is found to be much less indicated for TE generators than the DPP solution. Main conclusions are experimentally validated using a DPP architecture with a two-cell test bed operated at different thermal gradients.