An attention-enabled multi-sensor surveillance framework in proactive relay surveillance in nuclear power plants and power systems

Monitoring power and nuclear systems in time is really important because these systems are very complicated and the electrical processes are always changing. If we do not find problems with the relays away it can cause the system to fail, damage equipment or even put people in danger. To deal with these problems this research suggests a three-part plan that includes a Signal-Driven Feature Refinement Engine, a Spatial Relay-Aware Temporal Patterns Classifier and a Sequential Dense Attention Prediction Network. The Signal-Driven Feature Refinement Engine takes the data from many sensors removes the noise makes sure all the data is consistent and picks the most important features using special techniques that look at how things are related. The Spatial Relay-Aware Temporal Patterns Classifier looks at how different parts of the system’re connected and finds patterns over time and it is very good at predicting the state of the relays in real time getting it right 97.96% of the time. The Sequential Dense Attention Prediction Network uses an architecture that pays attention to the most important things one after another and it is even better at predicting what the relays will do in the future getting it right 98.93% of the time. The results of the experiments show that the plan we proposed is reliable and works well for finding problems with the relays. The whole system is good for monitoring power and nuclear systems predicting what might happen and making smart decisions especially in systems that are very important, for the electrical infrastructure. This is a solution that can be used for many different systems and it provides a lot of detailed information.