Zachary Smith and Dr. Matthew Jones, Mechanical Engineering
Introduction
A considerable amount of energy that is produced in a typical power generation process is rejected to the environment as low-grade waste heat. This heat is rejected at a temperature that is difficult to efficiently convert into electricity. One promising method of recovering this waste heat is through the use of thermoelectric generators (TEGs) that operate by means of the Seebeck effect. Effort has previously been placed in predicting the power output of TEGs. With knowledge of the temperature differential across a thermoelectric device and the properties of the electrical components the device is connected to, the electrical power output can be predicted. However, the complex geometry of the internal components of the thermoelectric device makes the heat transfer difficult to model analytically. As a result, the power output is difficult to predict without experimentally testing the system for each individual application.
The goal of this research was to experimentally obtain a model for the heat transfer through a thermoelectric device used in waste heat recovery. This would allow the feasibility of using a thermoelectric device for power generation to be evaluated in an inexpensive and time-efficient manner. An accurate analytical model would make future applications of thermoelectric devices used in power generation easier to analyze in a time-efficient manner.
Methodology
An experimental system was devised in which a TEG could experience a range of heating and cooling environments and the power output could be continuously monitored. A TEG (Custom Thermoelectric) was placed in between a custom built aluminum heat sink (77mm X 77mm) and a steel diffuser plate. Nylon screws were used to hold the system together. An 180W magnetic heater (Kats 1153) was used to heat the steel diffuser plate. Thermocouples were placed on the hot side of the TEG, the cold side of the TEG, and in between the heater and the diffuser plate. Thermal paste was used to minimize contact resistance. The TEG was connected in series with a 4Ω resistor and voltage measurements were taken across the resistor. A fan was placed 4 inches away from the heat sink and wind speed was adjusted by varying the voltage input to the fan. The system was modeled as a one-dimensional thermal circuit. This model assumed that no heat was lost through the insulation and that temperatures were uniform on the surface of the TEG.
Results and Discussion
As expected, the voltage output of the TEG varied approximately linearly with temperature differential across the TEG over the range of temperatures that were used (see figure 1). This voltage can be used to find power output using eq. (1).
These values can be compared to published data by the manufacturers for these temperature ranges to give some validation for this section of the model.
The temperature difference across the diffuser plate and the heat sink was much larger than predicted (see table 1). It was determined that the thermal contact resistance was more significant than anticipated. The thermal contact resistance of the system was on the same order of magnitude as the thermal resistance due to conduction through the diffuser plate and through the heat sink. This resulted in large amounts of error as only an estimate of the actual contact resistance could be determined from the given setup. With no thermal resistance it is anticipated that the temperature difference across the diffuser plate would not exceed a few degrees.
Conclusion
Further work will need to be done to allow for a more complete analytical model of the TEG system. It is recommended that the system is simplified to determine more exact values for the thermal contact resistance as well as the thermal resistance through the TEG itself. This could be accomplished by designing a system in which a thicker diffuser plate and heat sink are used in order to measure more temperatures throughout the system. This will give a clearer idea of where large temperature jumps occur. This will also ensure that contact resistance and not heat loss through the insulation is the major cause of the larger than anticipated temperature jumps. This information could be combined with the rest of the model to more accurately predict the entire performance of the system. It is also recommended that a wider range of temperatures is evaluated to allow for a more complete model over temperature ranges in which the system may be used.