Prior Flow Variational Autoencoder: A density estimation model for Non-Intrusive Load Monitoring

TL;DR

This paper introduces PFVAE, a novel deep neural network model combining Variational Autoencoders and Normalizing Flows for simultaneous appliance-level power demand estimation in NILM, showing significant accuracy improvements.

Contribution

The paper presents PFVAE, a unified density estimation model for NILM that estimates multiple appliances' power demands simultaneously, unlike previous models that handle appliances separately.

Findings

Achieves up to 81% improvement in normalized disaggregation error.

Achieves up to 86% improvement in signal aggregated error.

Demonstrates competitive performance on real-world dataset.

Abstract

Non-Intrusive Load Monitoring (NILM) is a computational technique to estimate the power loads' appliance-by-appliance from the whole consumption measured by a single meter. In this paper, we propose a conditional density estimation model, based on deep neural networks, that joins a Conditional Variational Autoencoder with a Conditional Invertible Normalizing Flow model to estimate the individual appliance's power demand. The resulting model is called Prior Flow Variational Autoencoder or, for simplicity PFVAE. Thus, instead of having one model per appliance, the resulting model is responsible for estimating the power demand, appliance-by-appliance, at once. We train and evaluate our proposed model in a publicly available dataset composed of power demand measures from a poultry feed factory located in Brazil. The proposed model's quality is evaluated by comparing the obtained normalized disaggregation error (NDE) and signal aggregated error (SAE) with the previous work values on the same dataset. Our proposal achieves highly competitive results, and for six of the eight machines belonging to the dataset, we observe consistent improvements that go from 28% up to 81% in NDE and from 27% up to 86% in SAE.