The operating environment of secondary equipment in smart substations is complex

and traditional protection strategies struggle to address challenges such as equipment aging

environmental changes

and coupling of multi-source faults. To solve the above problems

an adaptive protection strategy integrating multi-source online monitoring and deep learning-based fault prediction is proposed. A monitoring system covering electrical quantities

mechanical quantities and environmental quantities is constructed

and a recurrent neural network is used to realize fault situation prediction. On this basis

the protection setting values and action logic are dynamically adjusted according to the prediction results. Experimental results show that this strategy improves the fault prediction accuracy to 97.3%

reduces the protection action response time to 0.42 seconds

and controls the setting deviation within 2%

providing effective technical support for the safe and stable operation of secondary equipment in smart substations.