Multi-Algorithm Collaborative Temperature Sensing in Wireless Power Systems: A Data-Driven Early Fault Detection Method

The operation of wireless power systems (WPS) requires temperature monitoring because accurate temperature measurements help to detect overheating which serves as a fault warning. This research introduces an early fault detection system that uses multiple temperature sensing algorithms to analyze data for equipment failures. The Internet of Things (IoT) network gathers thermal measurements from various sensor units which undergo pre-processing through Savitzky-Golay filtering for noise reduction and principal component analysis for feature extraction. The Efficient Decision-tuned Least Squares Support Vector Machine (EDLSSVM) conducts fault detection while the Decision Tree (DT) model produces understandable rules to analyze threshold-crossing patterns. The two models combine their outputs through a consensus process which decreases false positive results and enhances overall system trustworthiness. The hybrid system combines nonlinear modeling with understandable rules to boost classification results which creates a dependable system that detects system faults and performs predictive maintenance in current WPS systems. The proposed consensus mechanism uses adaptive weights based on model confidence and decision consistency, which distinguishes it from traditional ensemble methods. The system automatically resolves conflicts by treating all models as equal. The research used both simulated data and real-time data to conduct their tests under different environmental conditions which included both STC and CEC testing scenarios. The multi-algorithm approach outperforms single-model methods, achieving 0.982 accuracy, 0.96 precision, 0.98 recall, and 0.97 F1-score. The results show that IoT monitoring systems combined with collaborative machine learning enable efficient real-time fault detection and predictive maintenance capabilities for modern WPS systems.