Analysis of the energy consumption of household appliances from 15 minute smart meter data and technical data sheets

This work focuses on estimating the electrical energy consumption of household appliances, and it does so with two methods. The first method presented aims to estimate the energy consumption of an appliance from technical sheet data and energy class. The objective is to provide Midori with a model to calculate the energy consumption of an appliance starting from the least possible number of inputs obtained from a customer survey. The appliances for which models have been obtained are fridges, freezers, washing machines, washer-dryers, tumble dryers, dishwashers, televisions, electric water heaters and heat pumps. The models obtained have been tested against unseen appliances to validate the results and are now being used by Midori to provide energy consultancy services to its customers. The second part of this work proposes a model to estimate the electricity consumption of an appliance starting from the household’s aggregate electricity consumption. The model proposed is a bidirectional LSTM that takes 15 minute smart meter electricity data as input. The model is trained on the freely-available REFIT dataset, which is composed of data from 20 houses in the UK collected over a period of almost two years. The testing of the model has been carried out both on other houses in the REFIT dataset itself and on the ECO dataset, another freely-available dataset containing smart meter data of 6 houses in Switzerland collected over a period of 8 months. This work contributes to existing research by providing models that estimate the energy consumption of household appliances according to the latest EU legislation on energy classes. In addition, this work also provides a model that outperforms current models in literature on 15 minute data both for appliance energy consumption estimation and appliance activation detection. Finally, this work also provides an up-to-date literature review on 15-minute energy disaggregation works in literature.