Generally, the accuracy stability of precision motorized spindle unit is influenced by its structural thermal balance behaviors (power matching degree of spindle structural heat generationsdissipations). For the accurate analysis onto this spindle heat transfer behaviors based on temperature detections, this paper describes a method constructed with the numerical simulation technology and a proposed GA-ELM algorithm. Firstly, heat-fluid-solid coupling FE transient model is established to simulate spindle thermal behaviors, its thermal loads / boundary conditions are initially according to empirical calculations. Secondly, based on spindle temperature detections, these initial values are applied onto thermal simulations and corrected by GA, to make simulated spindle temperatures gradually approach detections. Specially, ELM is adopted to estimate the functional relationships from parent population to child population generated by genetic operators with the increasing fitness values, and the trained ELM model is utilized to ensure the GA faster convergence. Eventually, based on the corrected thermal load / boundary condition values and simulation results, the time-varying power matching conditions of spindle structural heat generation -dissipation are analyzed. This study provides a theoretical basis for the optimization and promotion of spindle structural design and coolant control strategy.