Non-intrusive load identification technology can obtain the usage of various equipment at low cost

and can monitor and analyze power load on line

which is of great significance for load forecasting

demand response and other applications. In view of the diversity of general industrial and commercial users

the variety of loads and the complexity of equipment operation characteristics

anindustrial and commercial load identification scheme based on normalized three threshold event detection and LDA classifier is proposed

Firstly

a unified load event detection framework with adjustable parameters is designed for devices with different energy levels and different start-stop characteristics

which improves the detection accuracy of slow-moving

segmented and oscillating load events. Then an equipment type identification algorithm based on multivariate features and LDA linear discrimination is proposed

which achieves the same identification performance as nonlinear classifiers such as random forest while ensuring the computational efficiency of the edge.