Abstract: The quenching and cooling of hot-rolling steel is a important step to improve the quality and mechanical properties of steel plates. It is the key to a quenching procedure to control the speed of cooling. Against the inherent shortcoming of the traditional quenching model and for the requirement of expanding steel varieties, specifications and improving the precision of quenching temperature, a temperature forecast model in quenching and control cooling was established by the method of neural networks. Combining this forecast model with the previous regression mathematical model, the real-times control of quenching and control cooling was accomplished. The result shows that the comprehensive model improved greatly the controlling accuracy of quenching and cooling.