Simulation of thermoelectric heat pumps in nearly zero energy buildings: why do all models seem to be right?

The use of thermoelectric heat pumps for heat, ventilation, and air conditioning in nearly-zero-energy buildings is one of the most promising applications of thermoelectrics. However, simulation works in the literature are predominately based on the simple model, which was proven to exhibit significant deviations from experimental results. Nine modelling techniques have been compared in this work, according to statistical methods based on uncertainty analysis, in terms of predicted coefficient of performance and cooling power. These techniques come from the combination of three simulation models for thermoelectric modules (simple model, improved model, electric analogy) and five methods for implementing the thermoelectric properties. The main conclusion is that there is no statistical difference in the mean values of coefficient of performance and cooling power provided by these modelling techniques under all the scenarios, at 95% level of confidence. However, differences appear in the precision of these results in terms of uncertainty of the confidence intervals. Minimum values of uncertainty are obtained when the thermal resistance ratio approaches 0.1, being ±8% when using temperature-dependent expressions for the thermoelectric properties, ±18% when using Lineykin's method, and ± 25% when using Chen's method. The best combination is that composed of the simple model and temperature-dependent expressions for the thermoelectric properties. Additionally, if low values of resistance ratio are anticipated, empirical expressions from the literature can be used for the thermal resistance of the heat exchangers; for high values, though, experimental tests should be deployed, especially for the heat exchanger on the hot side.