In a microwave application device that uses a multi-microwave source array for spatial power synthesis, how to coordinate the feed-in power status information of multiple microwave sources to use the self-organization characteristics of temperature distribution to optimize temperature uniformity is the focus of this research. For this reason, on the one hand, this paper proposes the necessary elements for microwave sources to form an agent and constructs a technical solution. On this basis, the second-order non-fully connected communication topology consensus algorithm based on algebraic graph theory is introduced to cooperate with the power feed state information of multiple microwave sources to ensure that there will be no new hot spots in the process of using self-organization characteristics to optimize temperature distribution. On the other hand, the finite element method is used to construct a numerical calculation model that solves the mixed optimization of integer variables and continuous variables. Then carried out effective calculations to optimize the uniformity of temperature field distribution. The simulation experiment verifies the effectiveness of the microwave source intelligentization scheme. And the numerical calculation results show that the proposed model can improve the uniformity of 24.3%~55.4% and 20.4%~82.9% at each horizontal and vertical section respectively compared with the general heating model. Simultaneously, the thermal energy conversion efficiency can improve to 10.0%~ 43.7%. The above results verify that the proposed multi-source microwave heating temperature uniformity optimization method based on the consistency theory is feasible and efficient.