Rationale: With the increasing penetration of sensing and measurement technology, communication network technology, as well as cloud computing and big data storage and analytics, traditional energy systems are being digitized. Large amounts of energy production and consumption data are generated, collected and stored. This provides the possibility to implement energy big data mining and analysis. The energy consumption behavior of households can be described in three dimensions, namely time dimension, user dimension and spatial dimension and the existence of real time individual data provide an opportunity of robust analysis. A set of socioeconomic, demographic and the real time energy consumption data can been combined to assess the impact of behavioral insights on energy consumption pat terns. This high level of understanding of current energy usage situation provide knowledge of good and secure investments for bridging the energy efficiency gap and means for verification of expected results of investments.
Use case 3 seeks to examine the role of big data in assessing the impact of behavioural insights in energy consumption. Thus, the following questions are asked:
1. Can a data-driven ML analysis on pre-treatment data efficiently assist the design of a randomized control trial?
2. Can ML data-driven approaches identify heterogeneity in treatment effects that was not specified in a preanalysis plan?
3. Do ML methodologies outperform the traditional econometric energy consumption and production forecasting models?
Design and impact: Real time energy consumption data will be collected at a disaggregated level in the apartment buildings as well as the adjacent EV chargers. Socioeconomic and demographic variables will be collected through public census data and a survey will estimate the existence of specific behavioral biases regarding energy consumption. The goal is to estimate consumer propensities, by using a logistic regression model over the set of socioeconomic, demographic, and behavioral variables. Statistical results will show whether housing type, number of inhabitants, age, and end use behavior are strong predictors for choosing energy efficient appliances and building energy retrofit investments. Based on project s results policy measures will be suggested for increasing energy efficiency designing improved information campaigns by targeting key demographics.
Key Performance Indicators (KPIs):
1. Amount of energy consumption and production data coming from residential users (> 50,000 consumers in both Greece and Sweden).
2. Performance measures for ML methodologies estimating heterogeneous treatment effects (mean square error).
3. Prediction model performance measures (accuracy, recall, adjusted R-square, etc.).
Involved Energy Actors: CWATT
This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 957117. The information contained in this website reflects only the authors’ view. EC is not responsible for any use that may be made of this information.