Online monitoring of temperature and thermal stresses is an important way to ensure the safety of power plants considering fatigue and creep damages. The effect of online monitoring is determined by the accuracy and calculating time of monitoring models. In this paper, the improved first-order analytical models of temperature and thermal stresses considering temperature-dependent material properties have been derived by using homotopy analysis method (HAM) and superposition principle. The optimal convergence control parameters are obtained by calculating the mean-square residual errors. The validity and accuracy of the proposed models were proved by results comparisons with finite element method (FEM) and artificial parameter method.