Prediction of the effect of load resistance and heat input on the performance of thermoelectric generator using numerical and artificial neural network models

The load resistance in the thermoelectric generators (TEGs) is crucial for optimizing power output, and managing load resistances and operating conditions are integral elements of TEG system design. In this context, it is very important to predict the performance of TEGs at variable operating conditions. This research addresses an important problem in TEG by predicting the effects of load resistance and heat input on performance using both numerical and ANN models. The three-dimensional finite volume methods applied by employing ANSYS software, and the results were compared with experimental and ANN results in terms of voltage, current, power output and efficiency. In case of ambient temperature values of 17 degrees C, the average absolute errors of ANN and numerical model were calculated as 2.09 % and 4.22 % for voltage output, 0.79 % and 7.73 % for current output, 5.35 % and 12.09 % for power output, 4.14 % and 12.09 % for efficiency, respectively. As a result, it was observed that the ANN model gives better results compared to the numerical model. The highest power output was obtained at load resistance of 5.4 ohm and hot surface temperature of 150 degrees C, with the value of 0.827 W experimentally, 0.905 W numerically, and 0.824 W with ANN models. Besides, to evaluate the effect of ambient temperature, two additional temperatures (10 degrees C and 25 degrees C) were tested and it was found that decreasing the ambient temperature increased the TEG performance. The results showed that significant improvements on power and efficiency performance of the TEG can be achieved with optimised operating conditions.