Probabilistic Day-Ahead Forecasting of Household Electricity Demand

Abstract : Forecasting electricity demand at the local level of a building up to a feeder is increasingly necessary in several applications in the smart-grids context. Actors like aggregators and retailers, and tools like home energy management systems, require such forecasts as input. In this paper, a probabilistic day-ahead forecasting model is proposed to predict hourly electrical demand from individual households. This stochastic model uses smart-meter data and temperature predictions to make quantile forecasts. Performance is evaluated using data from a real-life smart grid demonstration site developed in Évora, Portugal as part of the European project SENSIBLE. The proposed model consistently outperforms a persistence model and provides reliable probabilistic forecasts.
Keywords : Forecast Electricity
Document type :
Conference papers
Complete list of metadatas

Cited literature [7 references]  Display  Hide  Download

https://hal.archives-ouvertes.fr/hal-01518373
Contributor : Magalie Prudon <>
Submitted on : Thursday, May 4, 2017 - 3:21:32 PM
Last modification on : Monday, November 12, 2018 - 11:00:34 AM
Long-term archiving on: Saturday, August 5, 2017 - 1:25:31 PM

File

CIRED_2017_0625.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01518373, version 1

Citation

Alexis Gerossier, Robin Girard, Georges Kariniotakis, Andrea Michiorri. Probabilistic Day-Ahead Forecasting of Household Electricity Demand. CIRED 2017 - 24th International Conference on Electricity Distribution, Jun 2017, Glasgow, United Kingdom. pp.0625. ⟨hal-01518373⟩

Share

Metrics

Record views

446

Files downloads

504