SEDR-Seq2P: A Lightweight Dilated Residual Sequence-to-Point Network for Multi-Task Industrial NILM

2026-07-30

AI Summary

Industrial non-intrusive load monitoring (NILM) faces significant generalization challenges due to measurement noise and concurrent appliance operation, hindering the direct transfer of models developed for residential settings. To address this, this work proposes SEDR-Seq2P, an enhanced Seq2Point-based approach that integrates a lightweight dilated residual module and Squeeze-and-Excitation (SE) attention mechanism within a unified multi-task framework capable of simultaneously disaggregating multiple industrial loads. Experimental results demonstrate that SEDR-Seq2P outperforms the Seq2Point baseline by reducing mean absolute error (MAE) by 7%, improving the coefficient of determination by 1%, and increasing the matching rate by 0.8%. Moreover, compared to WaveNet, it achieves a 58% reduction in inference latency, substantially improving the trade-off between accuracy and computational efficiency.

Abstract

Industrial NILM remains challenging because measurement noise and widespread concurrent machine operation reduce the generalization of models tuned on residential data. This work adopts a one-to-many, multi-task disaggregation setting, in which a single network estimates multiple industrial machine loads from aggregate power. Under a unified evaluation protocol on IMDELD, we benchmark Seq2Seq, Seq2SubSeq, Seq2Point, GRU, and WaveNet using energy-estimation metrics and the accuracy-delay criterion. While Seq2Point offers a stronger accuracy-delay balance than Seq2Seq/Seq2SubSeq, GRU and WaveNet achieve higher accuracy at markedly higher computational cost. To close this gap, we propose SEDR-Seq2P, a lightweight Seq2Point extension with dilated residual blocks and squeeze-and-excitation attention. Relative to the Seq2Point baseline, SEDR-Seq2P reduces MAE by approximately 7%, improves the coefficient of determination by approximately 1%, and increases the match rate by approximately 0.8%. In addition, compared to WaveNet, SEDR-Seq2P reduces inference latency by approximately 58%, yielding a favorable accuracy-delay trade-off for scalable industrial deployment.

Problem

• Industrial NILM
• measurement noise
• concurrent machine operation
• load disaggregation
• model generalization

Innovation

• dilated residual blocks
• squeeze-and-excitation attention
• Seq2Point
• lightweight network
• multi-task NILM