• • The improved thermal network incorporates nonlinear convective and radiative heat dissipation branches, enabling accurate estimation of TEG module temperature difference under natural convection, where conventional linear models fail due to the nonlinearity of heat dissipation with temperature.
• • Multi-objective parameter identification using NSGA-II successfully extracts thermal parameters that are difficult to calculate theoretically, with the estimation errors of hot-side, cold-side, and heat sink temperatures minimized simultaneously, ensuring high-fidelity temperature difference estimation.
• • Experimental validation under natural convection conditions confirms that the proposed method outperforms traditional equivalent thermal networks, achieving high-precision dynamic estimation of the cold- and hot-side temperature difference, which is essential for online performance evaluation of TEG modules in waste heat recovery systems.
• • The method demonstrates adaptability to forced convection environments by simply replacing the convective thermal resistance expression, providing a versatile framework for TEG module temperature estimation across various cooling conditions, thereby enhancing the reliability of thermoelectric power generation system assessments.