Deep Online Hierarchical Unsupervised Learning for Pattern Mining from Utility Usage Data

Abstract

Non-intrusive load monitoring (NILM) aims at separating a whole-home energy signal into its appliance components. Such method can be harnessed to provide various services to better manage and control energy consumption (optimal planning and saving). NILM has been traditionally approached from signal processing and electrical engineering perspectives. Recently, machine learning has started to play an important role in NILM. While most work has focused on supervised algorithms, unsupervised approaches can be more interesting and of practical use in real case scenarios. Specifically, they do not require labelled training data to be acquired from individual appliances and the algorithm can be deployed to operate on the measured aggregate data directly. In this paper, we propose a fully unsupervised NILM framework based on Bayesian hierarchical mixture models. In particular, we develop a new method based on Gaussian Latent Dirichlet Allocation (GLDA) in order to extract global components that summarise the energy signal. These components provide a representation of the consumption patterns. Designed to cope with big data, our algorithm, unlike existing NILM ones, does not focus on appliance recognition. To handle this massive data, GLDA works online. Another novelty of this work compared to the existing NILM is that the data involves different utilities (e.g, electricity, water and gas) as well as some sensors measurements. Finally, we propose different evaluation methods to analyse the results which show that our algorithm finds useful patterns.

Deep online hierarchical dynamic unsupervised learning for pattern mining from utility usage data

“…This work is a continuation of our previous work [38,39]. In [38], online Gaussian Latent Dirichlet Allocation (GLDA) is proposed to extract global components that summarise the energy signal.…”

“…In order to process continuous observations, the dirichlet distribution of LDA is replaced by Gaussian one resulting in Gaussian Latent Dirichlet Allocation (GLDA) [38]. We also derive DBN-LDA model where the HMM-part is omitted [39] Algorithm 1 Deep Online Hierarchical Dynamic Unsupervised Pattern mining for energy consumption behaviour 1: Input: raw-data window length, R; preprocessed-data window length, N ; length of time series, T ; number of components, K; number of patterns, P : total number of iterations, C; learning rate parameters, κ and τ 0 ; hyperparameters, α, α 0 , ω and σ. 2: Initialisation: variational parameters:…”

“…In [38], online Gaussian Latent Dirichlet Allocation (GLDA) is proposed to extract global components that summarise the energy signal. These components provide a rep- [38,39] ours resentation of the consumption patterns. The algorithm is applied on the same data-set as in this paper.…”

“…However, in contrast to [38], DBN-LDA-HMM employs deep learning to construct features rather than engineering them using signal processing technique. Similarly, the work by [39] also employs deep learning for features extraction. However, DBN-LDA-HMM also considers temporal dependency.…”

Building a Graph Signal Processing Model Using Dynamic Time Warping for Load Disaggregation

“…Hence, unsupervised methods are attractive. Popular unsupervised NILM methods are based on Dynamic Bayesian network models and specifically Hidden Markov Models (HMM) and their extensions including Factorial HMM, Additive Factorial HMM, and Gaussian Latent Dirichlet Allocation [ 6 , 8 , 11 , 12 , 13 , 24 ]. However, HMM-based NILM methods require good quality data to build the models, suffer from noise caused by unknown appliances, and are ineffective for rarely used appliances or appliances that are never used alone [ 8 , 11 , 25 ].…”