Improved method of non-intrusive load monitoring based on compressed sensing
Compressed Sensing (CS) has become one of the way to solve the problem of massive monitoring data in smart grid due to its characteristics of low-frequency sampling and simple compression. However, its application in non-intrusive load monitoring (NILM) has not yet been deeply studied. In order to m...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | International Journal of Electrical Power & Energy Systems |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S014206152500287X |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Compressed Sensing (CS) has become one of the way to solve the problem of massive monitoring data in smart grid due to its characteristics of low-frequency sampling and simple compression. However, its application in non-intrusive load monitoring (NILM) has not yet been deeply studied. In order to meet the requirements of dense data acquisition or high-frequency sampling in traditional NILM, NILM based on CS (NILM-CS) and its key elements’ improvement methods have been deeply explored for the first time. Firstly, NILM-CS model is established and its key elements is analyze. Then, event detection method and eigenvalue extraction method for load analysis of NILM-CS are proposed. On this basis, two theorems of ‘sparsity proportion’ were analyzed and Multiple Extraction Direction Pursuit based on Gradient (MEDP-G) algorithm was proposed to improve the reconstruction algorithm. At the same time, the Sparse Cosine Window (SCW) matrix was constructed to improve the measurement matrix, ultimately forming an improved NILM-CS method with MEDP-G algorithm and SCW matrix suitable for NILM. Experiment shows that the NILM-CS and its improved methods proposed in this paper are reasonable. The load decomposition accuracy of NILM-CS is greater than 92 %, and the load identification accuracy is close to 90 %, whose effect is similar to traditional compression method but reduces sampling frequency by 50 %, and is superior to traditional NILM with same-frequency sampling by 1 ∼ 5 percentage points. Even better, improved NILM-CS can further improve load decomposition accuracy by 1.2 ∼ 1.8 percentage points and improve load identification accuracy by 1.3 ∼ 4.8 percentage points compared to NILM-CS, which has a load analysis effect far superior to traditional NILM ways for solving massive data problems. |
|---|---|
| ISSN: | 0142-0615 |