Work

Non-intrusive residential load identification based on load feature matrix and CBAM-BiLSTM algorithm

Year: 2024

Type: article

Abstract: With the increasing demand for the refined management of residential loads, the study of the non-invasive load monitoring (NILM) technologies has attracted much attention in recent years. This paper proposes a novel method of residential load identification based on load feature matrix and improved neural networks. Firstly, it constructs a unified scale bitmap format gray image consisted of multiple load feature matrix including: V-I characteristic curve, 1–16 harmonic currents, 1-cycle steady-state current waveform, maximum and minimum current values, active and reactive power. Secondly, it adopts a convolutional layer to extract image features and performs further feature extraction through a convolutional block attention module (CBAM). Thirdly, the feature matrix is converted and input to a bidirectional long short-term memory (BiLSTM) for training and identification. Furthermore, the identification results are optimized with dynamic time warping (DTW). The effectiveness of the proposed method is verified by the commonly used PLAID database.

Source: Frontiers in Energy Research

DOI: 10.3389/fenrg.2024.1443700

Authors: Shunfu Lin, Bing Zhao, Yinfeng Zhan, Junsu Yu, Xiaoyan Bian, Dongdong Li

Institutions: Shanghai University of Electric Power, State Grid Shandong Electric Power Company (China)

Language: English

Topic: Smart Grid Energy Management

Subfield: Electrical and Electronic Engineering

Field: Engineering

Domain: Physical Sciences

SDGs: Affordable and clean energy, Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $2,795