W8LP2 – Predicting Household Electricity Demand Profiles from Measurements (MRes project)
17th July 2017 Matthew Li

Predicting Household Electricity Demand Profiles from Measurements

Matthew Li, Loughborough

Background and Context

Relationships between household characteristics and overall electricity consumption have been widely investigated (Jones & Lomas, 2015; Huebner 2016); however, the extent to which these characteristics influence electricity use at any particular time of the day is less well understood. Work to gain a deeper understanding of the time-varying nature of domestic electricity demand is considered timely given the UK Government target of obtaining 15% of energy from renewable sources by 2020 (DECC, 2013), and in light of the intermittent nature of the UK solar and wind resources on both seasonal and diurnal scales.

The ongoing UK roll-out of smart meters means that automated collection of half-hourly electricity demand data for individual dwellings across the UK will soon be possible. While the availability of such data is undoubtedly of huge potential value, the sheer volume of data is likely to mean that new methods of describing and analysing the shape of daily electricity demand will be required in order to extract meaningful insights and conclusions.

The large-scale collection of household electricity demand data also presents the opportunity to both support and challenge electricity demand modelling methods. Limited access to high-resolution consumption data has tended to mean that load profiling methods have been trained against small data sets exhibiting homogeneity of household types and covering limited time-frames. Smart meter data could potentially be used to augment, or even replace, established time-use-survey-based methods of modelling occupancy when predicting load profiles, while also providing a huge empirical data set for model verification.

Research Hypotheses

This project seeks to address three research hypotheses:

  1. Statistical relationships will exist between household electricity profiles and household characteristics;
  2. The shape of daily household electricity demand will exhibit significant variations between different months and day types;
  3. Electricity demand in UK homes can be predicted using measurements made in lots of homes.


The project aims to investigate the usefulness of monitored demand data in understanding and predicting the time-varying behaviour of domestic electricity demand. The following research methodology is applied:

  1. Literature review to establish:
    (i) Current understanding of the shape of domestic electricity demand;
    (ii) Factors influencing domestic electricity demand;
    (iii) Existing methods employed in prediction of daily electricity load profile shapes.
  2. Identification of a smart-meter-like domestic electricity consumption dataset.
  3. Development of a methodology for describing daily load profiles.
  4. Statistical analyses to identify the influence of household characteristics and seasonsal/day-type variability on the shape of daily demand.
  5. Comparison of results from 4 with similar analyses on a set of synthetically generated load profiles.