Non-intrusive Load Monitoring Based Data-Free Incremental Electrical Appliance Identification

In the realm of modern energy management systems, Non-Intrusive Load Monitoring (NILM) technology provides an innovative solution for power suppliers. It allows for remote monitoring and analysis of users’ electricity consumption patterns without requiring direct access to each device. This technology works by collecting comprehensive energy data at the user’s meter entry point and using algorithms to differentiate and identify the energy consumption characteristics of each device. However, with the increasing number of smart devices used by residential and commercial users, traditional NILM methods face the challenge of dealing with more complex and diverse electricity usage patterns. To address these challenges, we propose a novel incremental NILM framework. This framework has the ability to independently learn new load categories without load data, which is crucial for adapting in real-time to the changing variety of load devices, while also effectively addressing user privacy concerns. We introduce a VI-SCRM for extracting V-I trajectory features, incorporating a reconstruction convolution module. This method reduces the redundancy of V-I trajectory features and eliminates insignificant features, thereby improving algorithm efficiency. Additionally, this approach streamlines the model training process and prevents the model from ‘forgetting’ previously acquired knowledge. By implementing this framework in NILM, the system can continually adjust to new types of devices. Our experimental results confirm the effectiveness of this approach and highlight its potential advantages in practical applications.