面向复杂用电场景负荷监测的高鲁棒性事件检测方法OA

Event-driven non-intrusive load monitoring relies on event detection to infer the working status of a single load from changes in aggregated measurements.However,existing event detection methods suffer from significant false positives and missed detections in complex electricity usage scenarios,commonly characterized by long switching transients,short switching intervals,and high fluctuations.Therefore,a robust event detection method is proposed.The method first detects events at multiple scales using an improved sliding window combined with trend analysis and then applies adaptive thresholding to mitigate the impact of fluctuations.Furthermore,it separates short-interval events occurring within long switching transients based on outlier detection.Experiments demonstrate that the proposed method consistently outperforms existing approaches across diverse complex scenarios in public datasets.