Due to the important role of steel industry in economy, using modern control techniques in steel productions to improve the quality of steel and consequently, to improve related industrial products, is of great importance. In this thesis a predictive control method based on neural networks for temperature control of strip in ROT cooling section is described. The equations that model the thermal behavior of strip on the ROT are partial differential equations with nonlinear and time varying boundary conditions. The aim of this thesis is designing a predictive controller to control temperature of strip in ROT. Since the predictive controller has two basic blocks; model of process and optimization block, a Neural Network model is used in this thesis. Furthermore, the optimization block, used in the designed controller is based upon the branch and bound algorithm. The performance of the designed controller has been shown with simulations results.