Published May 23, 2019 | Version v1
Miscellaneous Restricted

Short-Term Forecasting of Electricity Demand of Smart Homes and Distribution Grids

Creators

Description

This thesis is devoted to the short-term forecasting of electricity demand of smart homes and distribution grids. The household demand data provided by smart meters is analyzed to characterize the electricity demand at the local scale and compared to this at the regional scale, so as to examine the aggregation effect. This thorough analysis enables the designing of models that forecast the future demand. The models make use of advanced statistical tools and machine-learning techniques. The inputs are selected with special care for their relevancy to the household demand. To be deployed in an operational environment, the models must be replicable: low to no maintenance, adaptability to various situations, and robustness to the lack of data. Several demand forecasting products are developed and compared to actual datasets: probabilistic forecasts at different temporal and spatial resolutions, and daily demand scenarios. Finally, the habits related to a domestic appliance, namely the charging of an electric vehicle battery, are modeled in order to generate forecasting scenarios of the appliance demand. (author)

Abstract (French)

Cette these s'interesse a la prevision a court terme de la demande electrique d'une maison intelligente et des reseaux de distribution. Les donnees mesurees par les compteurs intelligents permettent de caracteriser la demande electrique a l'echelle d'une maison et de la comparer a la demande regionale, pour etudier notamment l'effet de foisonnement. Cette analyse permet de developper des modeles de prevision de cette demande. Ces modeles sont de nature statistique et font usage de methodes d'apprentissage automatique. Un soin particulier est porte a la selection de variables d'entree pertinentes. Afin d'etre deployes dans un environnement operationnel, les modeles doivent faire preuve de replicabilite: fonctionnement autonome, aptitude a s'adapter a de multiple situations, et robustesse face aux donnees erronees. Plusieurs produits de prevision sont developpes et evalues avec plusieurs jeux de donnees: des previsions probabilistes a differentes resolutions, et des scenarios journaliers de la demande. Enfin, les habitudes relatives a un usage electrique particulier, a savoir le chargement d'une batterie de vehicule electrique, sont modelisees pour produire des scenarios predictifs de la demande de cet usage specifique. (auteur)

Files

Restricted

The record is publicly accessible, but files are restricted to users with access.

Additional details

Additional titles

Original title (English)
Prevision a court terme de la demande electrique des maisons intelligentes et des reseaux de distribution

Publishing Information

Imprint Pagination
269 p.
Report number
FRNC-TH--14702

INIS

Country of Publication
France
Country of Input or Organization
France
INIS RN
54081818
Subject category
S24: POWER TRANSMISSION AND DISTRIBUTION;
Resource subtype / Literary indicator
Thesis
Descriptors DEI
ALGORITHMS; BATTERY CHARGING; COMPUTERIZED SIMULATION; ELECTRIC-POWERED VEHICLES; FORECASTING; HOUSEHOLDS; LOAD MANAGEMENT; MULTI-PARAMETER ANALYSIS; OPTIMIZATION; POWER DEMAND; POWER DISTRIBUTION SYSTEMS; POWER METERS; PROBABILISTIC ESTIMATION; RESIDENTIAL SECTOR; SMART GRIDS; TIME RESOLUTION
Descriptors DEC
CALCULATION METHODS; DEMAND; ELECTRIC MEASURING INSTRUMENTS; ELECTRICAL EQUIPMENT; ENERGY SYSTEMS; EQUIPMENT; MANAGEMENT; MATHEMATICAL LOGIC; MEASURING INSTRUMENTS; METERS; POWER SYSTEMS; RESOLUTION; SIMULATION; TIMING PROPERTIES; VEHICLES

Optional Information

Notes
Available from the INIS Liaison Officer for France, see the INIS website for current contact and E-mail addresses