#### Optimization of segmented thermoelectric power generators using artificial neural network (ANN) and genetic optimization algorithm = 인공 신경망과 유전 최적화 알고리즘을 활용한 세그먼트 열전발전기 최적화

Cited 0 time in Cited 0 time in
• Hit : 74
In recent years, thermoelectric power generators (TEGs) have attracted enormous attention as environment-friendly electric generators utilizing waste heat. However, the efficiency of these power generators is not as high as other commercially available power generators. Most of the thermoelectric material properties that affect the performance of TEGs often vary with temperature and the performance of different thermoelectric materials is limited to their corresponding certain temperature range. To account for large temperature ranges and improve the efficiency of TEGs, the segmentation of different thermoelectric materials has been extensively studied. In designing the optimal segmentation with high-efficiency outputs, one has to consider not only the figure of merit (ZT) which is the most prominent factor for the performance of a single segment TEGs but also the efficiency compatibility factor ($S_e=({\surd(1+ZT)-1)}/{\alpha T}$) of different thermoelectric materials and the interface temperatures have major impacts on enhancing the efficiency of segmented TEGs. In the application of high power outputs, power factor ($\sigma \alpha^2$) is an important criterion as a figure of merit (ZT). In this study, fast inference artificial neural network (ANN) models have been trained based on the data sets generated from COMSOL-Multiphysics (5.2) simulations for different segmentation sequences of thermoelectric materials while varying lengths of each segment, and the external load resistivity. Based on the generated training data sets, ANN models under Bayesian regularizations (BR) optimizer are formulated that accurately predict the power and efficiency of segmented TEGs. Finally, the objective function is formulated using the trained neural network and optimized using a genetic optimization algorithm (GOA) and suggested the segmented TEGs with the most favorable power and efficiency outputs.
Ryu, Seunghwaresearcher유승화researcher
Description
한국과학기술원 :기계공학과,
Publisher
한국과학기술원
Issue Date
2020
Identifier
325007
Language
eng
Description

학위논문(석사) - 한국과학기술원 : 기계공학과, 2020.8,[iv, 41 p. :]

Keywords

Thermoelectric▼aTEG▼aSTEG▼aBayesian regularization▼aANN▼aNeural Network▼aCOMSOL-Multiphysics; 열전기▼a열전발전기▼a세그먼트화 열전발전기▼a베이지안 정규화▼a인공 신경망▼a신경망▼aCOMSOL-Multiphysics

URI
http://hdl.handle.net/10203/284955