Statistical Data-Driven Regression Method for Urban Electricity Demand Modelling

Abstract

As the focus of the energy transition within cities worldwide moves towards local communities and neighbourhoods, the need for insights in the dynamics of local electricity demand increases. Detailed local electricity demand information is, however, often not available. This paper proposes a statistical data-driven method to model local electricity demand for mixed urban areas, using a combination of other openly available datasets. Such datasets however are mutually incompatible without further conversion. The proposed method overcomes this problem. Linear regression is used to combine these different datasets, whereby the regression coefficients have the meaning of scaling factors for different types of electricity consumers (households, offices, shops, etc.). The method is calibrated and validated using respectively a training and a test dataset of Dutch municipalities, yielding R-squared values for most consumer types between 61% and 98%. The application of the method for local electricity demand modelling is illustrated for three Dutch municipalities with different consumer compositions.

A Data-Driven Method for Unsupervised Electricity Consumption Characterisation at the District Level and Beyond

“…They detected that the future tendency should integrate data-driven models and simulationbased models, as each of them provides interesting advantages. In Voulis et al (2018a), urban electricity demand modelling was tested for Dutch municipalities, where a combination of multiple data sets (reference electricity demand profiles, local customers composition data and aggregated local annual demand data) were used to train a regression model for local electricity demand prediction with an interesting application for local renewable energy transition plans (Voulis et al, 2018b). Kontokosta and Tull (2017) developed a predictive energy use model at the building, district, and city scales using training data from energy disclosure policies and predictors from the widely available property and zoning information.…”