Applying Deep Learning to Solve Alarm Flooding in Digital Nuclear Power Plant Control Rooms

Abstract

As the nuclear industry starts to shift to more digital controls and systems more information is provided to the control room and displayed on computer monitor workstations. This combined with alarm panels reduced to one alarm display creates the problem called alarm flooding, a situation where an overload of information can be caused during a plant disturbance or other abnormal operating condition. This project focused on finding a workable solution to assist operators in handling and understanding alarms during emergency situations. A generic pressurized water reactor simulator was used to collect process and alarm signals in scenarios that introduced common incidents causing expected alarms, as well as malfunctions causing unexpected alarms. Deep neural networks were used to model the collected data. Results showed that the models were able to correctly filter many of the expected alarms, indicating that deep learning has potential to overcome the problem of alarm flooding.