Non-intrusive Load Adaptive Identification Method Based on Siamese Network

Abstract

The present invention proposes a non-intrusive load self-adaptive identification method based on a twin network. The method uses the V-I trajectory and active power of the load as the characteristics of the load to be identified, and uses the twin network to determine the similarity of the V-I trajectory of the load. By matching with the feature library, the load number information is obtained, so as to realize load identification. Among them, the user establishes the load serial number information and the actual type of electrical appliance for mapping according to the saved usage time combined with the actual usage situation of the day. Through the dynamic construction of the feature database, the present invention can realize accurate identification of unknown loads. Finally, the validity and versatility of the model are verified in the PLAID dataset and COOLL dataset.

Description

technical field [0001] The present invention relates to the field of non-intrusive load monitoring (NILM), in particular to a twin network-based non-intrusive load self-adaptive identification method. Background technique [0002] Understanding user energy consumption is of great significance to load management. In recent years, non-intrusive load monitoring (Non-intrusive load monitoring, NILM) technology has attracted widespread attention. Traditional intrusive load monitoring requires the installation of acquisition and communication devices at each electrical load to detect the load status, and requires modification of existing electrical appliances or lines, which is difficult and expensive to implement. The non-intrusive load monitoring technology monitors the power bus to analyze the status of each load in the line, which has the advantages of strong versatility and low cost. [0003] Traditional non-intrusive load recognition algorithms are mostly based on classific...

AI Technical Summary

Problems solved by technology

[0003] Traditional non-intrusive load recognition algorithms are mostly based on classification models, which mainly have the following disadvantages: First, the models based on supervised learning methods require a large amount of labeled data to train the models, which is often unsatisfactory in reality; and these methods usually cannot To identify unknown equipment, the load identification model based on the classification method can only identify the learned loads, but cannot identify new loads and unknown loads; third, the model has poor versatility, and the types of loads in different families are not the same. Traditional load identification The method can only be modeled and optimized for specific situations, and its versatility is poor

Images