Adaptive Multi-Modal Fault Prognostics in High-Voltage Transmission Networks using Edge-Optimized Hybrid Ensemble Intelligence
The growing sophistication of contemporary high-voltage transmission systems requires fault monitoring systems that are no longer reactive, but rather predictive-based intelligence. The paper introduces an adaptive multi-modal fault prognostics system incorporating synchronous electrical waveforms and infrared thermography measurements into an edge-optimized hybrid ensemble system. In contrast to the traditional single-modality or simulation-based methods, the suggested system will utilize the concept of cross-modal feature fusion in real time to detect incipient fault signatures before a critical failure happens. Gradient-boosted and highly randomized tree ensembles are utilized with adaptive temporal feature weighting to be sensitive to anomalies at an early stage while being computationally efficient to implement as an embedded system. It is based on edge devices to guarantee the low-latency inference, less bandwidth consumption and supported by withstanding the communication breakages. Experimental validation with hybrid laboratory-field datasets is seen to achieve better prognostic accuracy, performance in cases of environmental variability, and this is also better interpreted with feature-attribution analysis. The suggested architecture provides a scalable and comprehensible predictive maintenance and resilient smart-grid operation solution.