The tubular furnace is one of the main production equipment in petrochemical industry, the main functions of which is to heat the liquid oil in multiple branch tubes in the furnace to the target temperature. Since the temperature of each branch furnace tube is affected by the feed composition, the distribution position of the furnace tube and the uneven distribution of the furnace temperature, these factors may result in the deviation of the oil outlet temperature of each branch, and the serious temperature deviation may lead to the coking of the furnace tube and even cause accidents. In order to overcome the problem of unbalanced outlet temperature of each branch tube in the tubular furnace, this paper proposes a temperature control method GSA-MPIDNN, which is based on genetic simulated annealing (GSA) algorithm to optimize multi-input multi-output proportion-integration-differentiation neural network(MPIDNN). The GSA algorithm is used to find out the optimal initial weights of the MPIDNN, to overcome the deficiency of the algorithm by manually setting the initial weights, and to improve the control performance of the MPIDNN controller on the outlet temperature of the tubular furnace. The Matlab software is used to build the mathematical model of GSA-MPIDNN controller and tubular furnace, and the results are compared and analyzed with the traditional methods such as MPIDNN, PID and fuzzy PID, etc. The results show that the convergence time and error of GSA-MPIDNN are better than the traditional methods, which verifies the effectiveness of the method.