The invention discloses a heat conductivity coefficient multi-scale prediction method for hardened portland cement paste. The method disclosed by the invention comprises the following steps of: performing multi-phase and multi-scale division according to the microstructure and the composition of the hardened portland cement paste, enabling each scale to contain a typical phase, then calculating the volume percentages of the different phases, and further calculating a heat conductivity coefficient of the hardened portland cement paste through a generalized self-consistent method according to the volume percentage, the heat conductivity coefficient and other parameters of each phase on each scale. According to the method disclosed by the invention, the influence of the microstructure and the performances of composition materials of the hardened portland cement paste on the macroscopic heat conductivity coefficient of the cement paste is considered, the considered influence factors are comprehensive and reasonable, the form is simple and the use is convenient. A prediction model constructed according to the method can predict the heat conductivity coefficients of the hardened portland cement paste with different water-cement ratios and different types of cement more accurately.