Thermoelectricity, which involves direct conversion of heat to electricity and vice versa, is a promising renewable energy alternative. However, low conversion efficiency of thermoelectric (TE) materials has restricted their applications. Enhancing the TE figure of merit ( \(ZT=\frac{{S}^{2} \sigma }{k}T)\) is challenging because of conflicting association between electrical conductivity (σ), thermal conductivity (k), and Seebeck coefficient (S). Conducting polymer-based composite TE materials have attracted significant attention since σ, k, and S can be somewhat independently tuned in engineered nanocomposites and because of their potential in flexible electronics. Carbon nanotube (CNT) and polyaniline (PANI) nanocomposites in particular present huge promise due to controllable π–π interaction. Nevertheless, reaching high ZT has so far remained difficult because of the complex interrelations between various structural and compositional features of the nanocomposites and ZT. To address this problem, we have used machine learning and genetic algorithms (GA) to design parameter sets for optimal results. A database of significant variables such as morphology, functionality, type, and concentration of CNT as well as types of dopants in PANI-CNT nanocomposites was extracted from literature. Modeling of nanocomposite was done by artificial neural networks (ANN). GA was used to find the best collection of solutions for different compositions, and the impact of design variables on σ, k, and S was assessed.
