Multiscale Simulation Methods for Thermoelectric Generators

Doktorsavhandling, 2016

Rising energy prices and greater environmental awareness, along with stringent emissions legislation, in the automotive industry make it possible to introduce techniques in the aftertreatment system that have previously been unprofitable. One such technique, studied here, is heat recovery from exhaust gases using thermoelectric generators. The design of thermoelectric modules and heat exchangers for thermoelectric generation relies, to a large extent, on simulation tools. Thermoelectric phenomena are well known, and several researchers have used first principle simulation to solve for thermoelectric generation in thermoelectric pairs and single modules. In order to obtain predictions that agree with measurements, knowledge of not only temperature-dependent material but also internal thermal and electrical contact resistances is required. A method that enables accurate quantification of contact resistances inside thermoelectric generators and which gives detailed insight into how these reduce module performance has been developed within the scope of this research. When implementing these resistances in first principle simulations, excellent agreement between measured and simulated performance has been achieved. First principle simulations allow great insight into thermoelectric performance and provide details, such as local current distribution, that are hard to measure or obtain with other methods and are great, for example, when designing modules. First principle models, on the other hand, are computationally too demanding when used to design heat exchangers that contain a large system of modules. Therefore, a novel framework for characterization and simulation of thermoelectric generator systems that allows for accurate and efficient prediction of electric and thermal performance has been developed in this research. When used in conjunction with CFD analysis, this framework allows for efficient modelling of electrical and thermal performance without relaxing the important two-way coupling of energy transport. This efficiency comes from the fact that the modelling does not require full resolution as first principle simulations do. Therefore it solves the scale separation problem and allows for multiphysics simulation with just a minor increase in computational power. All simulations were validated with experiments on different levels, both for individual modules, small systems of modules, and, finally, engine bench tests were used to validate a full-scale heat exchanger prototype containing a large number of modules and a complex fluid flow.