Forecasting the electricity consumption by aggregating specialized experts; a review of the sequential aggregation of specialized experts, with an application to Slovakian and French country-wide one-day-ahead (half-)hourly predictions

Marie Devaine; Yannig Goude; Gilles Stoltz

hal-00484940, version 2

Forecasting the electricity consumption by aggregating specialized experts; a review of the sequential aggregation of specialized experts, with an application to Slovakian and French country-wide one-day-ahead (half-)hourly predictions

Marie Devaine () ¹, Yannig Goude () ², Gilles Stoltz ( Auteur à contacter de préférence ) ^1, 3, 4

(27/03/2011)

Résumé : We consider a variant of the sequential prediction of arbitrary sequences based on experts advice, called prediction with specialized experts. We design aggregation rules, that sequentially combine the forecasts provided by the experts; the latter are specialized and need not output a prediction at all time instances while the aggregation rules have to. We provide first a review of the literature on specialized experts and take a new look at some aggregation rules (obtained as adaptations or extensions of earlier rules). We then consider an application to the sequential short-term (one-day-ahead) forecasting of electricity consumption; to do so, we consider two data sets, a Slovakian one and a French one, respectively concerned with hourly and half-hourly predictions. We introduce and develop a general methodology to perform the stated empirical studies. The introduced aggregation rules demonstrate an improved accuracy on the data sets at hand; the improvements lie in a reduced mean squared error but also in a more robust behavior with respect to large occasional errors.

1 : Département de Mathématiques et Applications (DMA)
CNRS : UMR8553 – Ecole normale supérieure de Paris - ENS Paris
2 : E.D.F. Division Recherche et Développement [Clamart]
EDF
3 : Groupement de Recherche et d'Etudes en Gestion à HEC (GREGH)
GROUPE HEC – CNRS : UMR2959
4 : CLASSIC (INRIA Paris - Rocquencourt)
Ecole normale supérieure de Paris - ENS Paris – INRIA

hal-00484940, version 2
http://hal.archives-ouvertes.fr/hal-00484940
oai:hal.archives-ouvertes.fr:hal-00484940
Contributeur : Gilles Stoltz <>
Soumis le : Dimanche 27 Mars 2011, 14:21:54
Dernière modification le : Vendredi 15 Avril 2011, 15:33:01

Voir la fiche détaillée

Documents associés

PDF :

Exporter

Bibtex EndNote TEI RefWorks

%% hal-00484940, version 2
%% http://hal.archives-ouvertes.fr/hal-00484940
@unpublished{devaine:hal-00484940,
    hal_id = {hal-00484940},
    url = {http://hal.archives-ouvertes.fr/hal-00484940},
    title = {{Forecasting the electricity consumption by aggregating specialized experts; a review of the sequential aggregation of specialized experts, with an application to Slovakian and French country-wide one-day-ahead (half-)hourly predictions}},
    author = {Devaine, Marie and Goude, Yannig and Stoltz, Gilles},
    abstract = {{We consider a variant of the sequential prediction of arbitrary sequences based on experts advice, called prediction with specialized experts. We design aggregation rules, that sequentially combine the forecasts provided by the experts; the latter are specialized and need not output a prediction at all time instances while the aggregation rules have to. We provide first a review of the literature on specialized experts and take a new look at some aggregation rules (obtained as adaptations or extensions of earlier rules). We then consider an application to the sequential short-term (one-day-ahead) forecasting of electricity consumption; to do so, we consider two data sets, a Slovakian one and a French one, respectively concerned with hourly and half-hourly predictions. We introduce and develop a general methodology to perform the stated empirical studies. The introduced aggregation rules demonstrate an improved accuracy on the data sets at hand; the improvements lie in a reduced mean squared error but also in a more robust behavior with respect to large occasional errors.}},
    keywords = {Prediction with experts advice ; Specialized experts ; Individual sequences ; Application to real data},
    language = {Anglais},
    affiliation = {D{\'e}partement de Math{\'e}matiques et Applications - DMA , E.D.F. Division Recherche et D{\'e}veloppement [Clamart] , Groupement de Recherche et d'Etudes en Gestion {\`a} HEC - GREGH , CLASSIC - INRIA Paris - Rocquencourt},
    note = {50 pages },
    year = {2011},
    month = Mar,
    pdf = {http://hal.archives-ouvertes.fr/hal-00484940/PDF/Devaine-Goude-Stoltz.pdf},
}

<?xml version="1.0" encoding="UTF-8"?>
<listBibl xmlns="http://www.tei-c.org/ns/1.0" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:mml="http://www.w3.org/1998/Math/MathML">
<listBibl type="article"><head>Documents sans référence de publication (Preprint)</head>
<biblStruct type="article" xml:id="hal-00484940"><analytic>
<author><persName><forename>Marie</forename><surname>Devaine</surname></persName>
<affiliation>
<orgName type="laboratory">DMA</orgName>
<orgName type="team">Département de Mathématiques et Applications</orgName>
<country>FR</country>
<orgName type="institution">CNRS : UMR8553</orgName>
<orgName type="institution">Ecole normale supérieure de Paris - ENS Paris</orgName>
</affiliation>
</author>
<author><persName><forename>Yannig</forename><surname>Goude</surname></persName>
<affiliation>
<orgName type="team">E.D.F. Division Recherche et Développement [Clamart]</orgName>
<country>FR</country>
<orgName type="institution">EDF</orgName>
</affiliation>
</author>
<author><persName><forename>Gilles</forename><surname>Stoltz</surname></persName>
<affiliation>
<orgName type="laboratory">DMA</orgName>
<orgName type="team">Département de Mathématiques et Applications</orgName>
<country>FR</country>
<orgName type="institution">CNRS : UMR8553</orgName>
<orgName type="institution">Ecole normale supérieure de Paris - ENS Paris</orgName>
</affiliation>
<affiliation>
<orgName type="laboratory">GREGH</orgName>
<orgName type="team">Groupement de Recherche et d'Etudes en Gestion à HEC</orgName>
<country>FR</country>
<orgName type="institution">GROUPE HEC</orgName>
<orgName type="institution">CNRS : UMR2959</orgName>
</affiliation>
<affiliation>
<orgName type="laboratory">INRIA Paris - Rocquencourt</orgName>
<orgName type="team">CLASSIC</orgName>
<country>FR</country>
<orgName type="institution">Ecole normale supérieure de Paris - ENS Paris</orgName>
<orgName type="institution">INRIA</orgName>
</affiliation>
</author>
<title type="main" level="a">Forecasting the electricity consumption by aggregating specialized experts; a review of the sequential aggregation of specialized experts, with an application to Slovakian and French country-wide one-day-ahead (half-)hourly predictions</title></analytic><monogr><imprint/></monogr><idno type="HALid">http://hal.archives-ouvertes.fr/hal-00484940</idno><idno type="HALFile">http://hal.archives-ouvertes.fr/hal-00484940/PDF/Devaine-Goude-Stoltz.pdf</idno></biblStruct>
</listBibl>
</listBibl>