Research on online non-intrusive load identification system based on multi-threaded CUSUM-MLP algorithm

Non-Intrusive Load Monitoring (NILM) has been promoted and many methods have been developed so far, which lead the online identification of loads into the focused research point. This paper proposes an online identification system framework of NILM based on multi-threaded Cumulative Summation-Multilayer Perceptron (CUSUM-MLP) event detection and identification algorithm. It contains a main thread and six sub-threads, and a combination of signal and slot mechanisms to accomplish the online recognition task. Data reception, data packetization and feature extraction are designed to be fulfilled in the main thread. Real-time data presentation, data storage, feature storage and real-time images are performed in four sub-threads. Aiming for the online mode, a sub-thread to update the data is designed. The CUSUM-MLP algorithm is packed as a sub-thread for event detection and load identification. Based on the proposed multi-threaded mechanism embedded with the CUSUM-MLP algorithm, the NILM online recognition system is verified through experiments, and shows high accuracy, good robustness and real-time performance.