Accuracy and speed of reconstruction algorithm play an important role in the temperature field measurement for a boiler by acoustic tomography. A dynamic model of a 3D temperature field reconstruction by acoustic tomography is established. A dynamic reconstruction algorithm is proposed considering both the acoustic measurement information and the dynamic evolution information of the temperature field. An objective function is built, which fuses the measurement information, space constraint of the temperature field and the dynamic evolution information. A regularization matrix is established based on the smooth constraint method which reflects the positional relationship between spatially adjacent pixels. A method combining Tikhonov regularization and optimization is adopted to solve the function. The numerical simulations show that the reconstruction speed of the algorithm fusing dynamic evolution information is similar to static reconstruction algorithm including the least square method, the algebraic reconstruction technique and the standard Tikhonov regularization algorithm. The image quality and noise immunity of the algorithm fusing dynamic evolution information are better than the results obtained from the static algorithms. An innovative method with high effectiveness is provided for temperature field reconstruction by acoustic.