The event-based non-intrusive load monitoring technique extracts transient power waveform features from the total load data by event detection and then uses them to identify the operating status of internal appliances. However

for complex appliances with long transient processes

the existing methods suffer from inherent defects such as inaccurate localization of event start and end points and incomplete transient power waveform feature extraction

which directly impacts load identification. To this end

a multi-resolution template matching method for non-intrusive load identification is proposed. In the load feature template library construction stage

the feature templates representing different types of transient power modes are selected from the existing load feature samples library by undirected graph clustering and time series averaging methods. In the load status online identification stage

the personalized variable-length sliding window strategy is used to extract load transient power waveform samples from the total power sequence

and the multi-resolution waveform feature template matching method is used to identify the operating status of appliances. Comparison test results on real-world data have shown that our method can extract load feature samples with better consistency and completeness and achieve higher load identification accuracy in different scenarios.