Maintenance Strategy Optimization for Smart Grid Sensor Networks Based on Dynamic Resilience Importance Using LSTM

This study first proposes a Dynamic Resilience Importance (DRI) framework for smart grid sensor networks (SGSN) that innovatively integrates three core contributions to address reliability challenges under network interference. First, we introduce a novel three-dimensional importance measure that uniquely combines loss, maintenance, and predictive dimensions, extending classical importance measure beyond traditional static approaches. Second, we develop an innovative LSTM-based fault prediction model with attention mechanism that processes multi-dimensional sensor data (communication metrics and power quality parameters) to achieve superior prediction accuracy of 78.49-97.68% across 1-24 hour horizons. Third, we propose a dynamic maintenance optimization strategy that adaptively weights predictive and traditional importance measures based on model confidence. Comprehensive evaluation using IEEE 14-bus system demonstrates that DRI achieves the lowest System Reliability Volatility Index (SRVI = 0.0811), representing a 38.5% improvement over conventional approaches. Results validate DRI’s breakthrough capability in maintaining consistent system reliability under sustained network interference, providing a fundamentally new approach for next-generation smart grid security and resilience enhancement.