In dynamic thermal management of multi-heat source systems, full-domain thermal analysis and real-time assessment are core elements. However, conventional discrete measurements and reconstruction techniques struggle to capture dynamic evolution of the temperature field in real time. Therefore, a full-domain temperature field reconstruction strategy based on discrete measurements is described. The eigenbasis functions of the temperature field are extracted by singular value decomposition (SVD), and the Gray Wolf Optimization (GWO) is introduced to optimize the layout of discrete sensors. A coefficient matrix for the full-domain temperature field reconstruction is computed by combining the eigenbasis with the discrete measurements. The reliability of this approach is validated based on numerical reconstruction experiments of four heat source systems. The results indicate that after sensor placement optimization, the level of theoretical reconstruction error for the four heat source systems is significantly reduced by at least three orders of magnitude. Moreover, in the temperature field reconstruction of a typical PCB multi-heat source system, comparison with numerical experiment results shows an average error of 0.12 ℃ and a root-mean-square percentage error below 1%. The reliability of the strategy in practical applications is validated, providing a reference for thermal analysis and control of electronic equipment.