The combustion in circulated fluidized-bed (CFB) boiler is a complex and multi-variables coupling process. It is too difficult to build an exact mathematic model to describe the process. Bed temperature control system is an important subsystem of burning system, maintaining normal bed temperature is the key for stable running of boiler, and it is difficult for the norm PID algorithm to control it. According to these, we use the bed temperature control system based BP neural network in DCS. The practice shows that it can improve the stability of bed temperature control system with the application of BP neural network.