A method for alarming water level of boiler drum on nuclear power plant based on BP Neural Network

Drum water level is an important parameter for boilers on both thermal power plant and nuclear power plant. It is hard to measure the level correctly. So it brings some difficulties to the control based on the drum water level, even the alarm. Usually, more than three water gauges are installed for drum water level measurement. And it adopts two-out-of-three strategy for obtaining the final alarm signal in distributed control system (DCS), which is often the false alarm. Without the right alarm, it is to result in very serious disaster on power plant. One approach based on Back-Propagation (BP) Neural Network is proposed in this paper for solving the problem. The measurements from different water gauges are inputted into the BP Neural Network after fuzzy process and the output of the Network represents the type of alarm. Some data of the drum water level from a nuclear power plant is applied with the method of the paper. From the experiments, it can be seen that the alarm accuracy is increased rapidly.