There are inevitable defects for direct composition control in distillation processes, and temperature inference control is an alternative means. To make the temperature of an intermediate stage stable is a dominant method in temperature control. However, even with a satisfactory response in the stage temperature, there could still be product composition offsets. To solve this problem, we explore the relative stability of the composition profile pattern in distillation columns and present a modeling method based on the profile pattern. We then establish a dynamic relationship between the controlled temperature and the product composition according to this relative stability and utilize this method in closed-loop control. The temperature set-point is adjusted adaptively according to the state of the profile pattern, and the purity offsets could be decreased or eliminated. Conventional and heat integrated distillation columns are both illustrated in the simulation study. We develop two main types of control schemes for the two processes to investigate the effects of adaptive control methods. These control schemes are discussed to explain how to choose suitable methods for different situations.