This paper reviews non-intrusive load monitoring (NILM) approaches that employ deep neural networks to disaggregate appliances from low frequency data, i.e., data with sampling rates lower than the AC base frequency. The overall purpose of this review is, firstly, to gain an overview on the state of the research up to November 2020, and secondly, to identify worthwhile open research topics. Accordingly, we first review the many degrees of freedom of these approaches, what has already been done in the literature, and compile the main characteristics of the reviewed publications in an extensive overview table. The second part of the paper discusses selected aspects of the literature and corresponding research gaps. In particular, we do a performance comparison with respect to reported mean absolute error (MAE) and F${}_1$-scores and observe different recurring elements in the best performing approaches, namely data sampling intervals below 10 s, a large field of view, the usage of generative adversarial network (GAN) losses, multi-task learning, and post-processing. Subsequently, multiple input features, multi-task learning, and related research gaps are discussed, the need for comparative studies is highlighted, and finally, missing elements for a successful deployment of NILM approaches based on deep neural networks are pointed out. We conclude the review with an outlook on possible future scenarios.

This paper reviews non-intrusive load monitoring (NILM) approaches that employ deep neural networks to disaggregate appliances from low frequency data, i.e. data with sampling rates lower than the AC base frequency. We first review the many degrees of freedom of these approaches, what has already been done in literature, and compile the main characteristics of the reviewed publications in an extensive overview table. The second part of the paper discusses selected aspects of the literature and corresponding research gaps. In particular, we do a performance comparison with respect to reported MAE and F$_1$-scores and observe different recurring elements in the best performing approaches, namely data sampling intervals below 10\,s, a large field of view, the usage of GAN losses, multi-task learning, and post-processing. Subsequently, multiple input features, multi-task learning and related research gaps are discussed, the need for comparative studies is highlighted, and finally, missing elements for a successful deployment of NILM approaches based on deep neural networks are pointed out. We conclude the review with an outlook on possible future scenarios.

Forecasting or modeling the on-off times of domestic appliances has gained increasing attention in recent years. However, comparing currently published results is difficult due to the many different data-sets and performance measures employed. In this paper, we evaluate the performance of three increasingly sophisticated approaches within a common framework on three data-sets each spanning 2 years. The approaches forecast the future on-off times of the appliances for the next 24 h on an hourly basis, solely based on historic energy consumption data. The appliances investigated are driven by user behavior and consume a significant fraction of the household’s total electrical energy consumption. We find that for all algorithms the average area under curve (AUC) in the receiver operating characteristic (ROC) is in the range between 72% and 73%, i.e. indicating mediocre prediction quality. We conclude that historic consumption data alone is not sufficient for a good quality hourly forecast.

Datasets with measurements of both solar electricity production and domestic electricity consumption separated into the major loads are interesting for research focussing on (i) local optimization of solar energy consumption and (ii) non-intrusive load monitoring. To this end, we publish the iHomeLab RAPT dataset consisting of electrical power traces from five houses in the greater Lucerne region in Switzerland spanning a period from 1.5 up to 3.5 years with a sampling frequency of five minutes. For each house, the electrical energy consumption of the aggregated household and specific appliances such as dishwasher, washing machine, tumble dryer, hot water boiler, or heating pump were metered. Additionally, the data includes electric production data from PV panels for all five houses, and battery power flow measurement data from two houses. Thermal metadata is also provided for the three houses with a heating pump.

Grid-connected photovoltaic (PV) capacity is increasing and is currently estimated to account for 3.0% of worldwide energy generation. One strategy to balance fluctuating PV power is to incentivize self-consumption by shifting certain loads. The potential improvement in the amount of self-consumption is usually estimated using smart meter and PV production data. Smart meter data are usually available only at sampling frequences far below the Nyquist limit. In this paper we investigate how this insufficient sampling rate affects the estimated self-consumption potential of shiftable household appliances (washing machines, tumble dryers and dishwashers). We base our analyses on measured consumption data from 16 households in the UK and corresponding PV data. We found that the simulated results have a marked dependence on the data sampling rate. The amount of self-consumed energy estimated with data sampled every 10 min was overestimated by 30–40% compared to estimations using data with 1 min sampling rate. We therefore recommend to take this factor into account when making predictions on the impact of appliance load shifting on the rate of self-consumption.