Machine learning based electric load forecasting for short and long-term period

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27240%2F18%3A10240148" target="_blank" >RIV/61989100:27240/18:10240148 - isvavai.cz</a>

Nalezeny alternativní kódy

RIV/61989100:27730/18:10240148

Výsledek na webu

<a href="http://dx.doi.org/10.1109/WF-IoT.2018.8355123" target="_blank" >http://dx.doi.org/10.1109/WF-IoT.2018.8355123</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/WF-IoT.2018.8355123" target="_blank" >10.1109/WF-IoT.2018.8355123</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Machine learning based electric load forecasting for short and long-term period

Popis výsledku v původním jazyce

Electricity is currently the most important energy vector in the domestic sector and industry. Unlike fuels, electricity is hard and expensive to store. This creates the need of precise coupling between generation and demand. In addition, the transmission lines of electric power need to be sized for a given maximum power, and overloading them may result in blackout or electrical accidents. For these reasons, energy consumption forecasting is vital. The time scale for forecasting depends on who is interested in such prediction. Grid operators have to predict the electricity demand for the next day, to program the generation accordingly. Grid designers have to predict energy consumption at the scale of years, to ensure that the infrastructure is sufficient. On the other hand, smart grid controllers with almost instant response time may need a prediction on the order of minutes. We have seen that changing the time scale in electricity load forecasting changes the approach, and that depending on the scale different methods should be used to ensure the highest accuracy with the smallest computational cost. We show here how forecasting accuracy decreases with the increase of time scale due to the impossibility of using of all variables. Several well established computational models were compared on three different regression based criteria and the results revealed that boosting model was able to outperform their competitors in most of the comparisons.

Název v anglickém jazyce

Machine learning based electric load forecasting for short and long-term period

Popis výsledku anglicky

Electricity is currently the most important energy vector in the domestic sector and industry. Unlike fuels, electricity is hard and expensive to store. This creates the need of precise coupling between generation and demand. In addition, the transmission lines of electric power need to be sized for a given maximum power, and overloading them may result in blackout or electrical accidents. For these reasons, energy consumption forecasting is vital. The time scale for forecasting depends on who is interested in such prediction. Grid operators have to predict the electricity demand for the next day, to program the generation accordingly. Grid designers have to predict energy consumption at the scale of years, to ensure that the infrastructure is sufficient. On the other hand, smart grid controllers with almost instant response time may need a prediction on the order of minutes. We have seen that changing the time scale in electricity load forecasting changes the approach, and that depending on the scale different methods should be used to ensure the highest accuracy with the smallest computational cost. We show here how forecasting accuracy decreases with the increase of time scale due to the impossibility of using of all variables. Several well established computational models were compared on three different regression based criteria and the results revealed that boosting model was able to outperform their competitors in most of the comparisons.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2018

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

IEEE World Forum on Internet of Things, WF-IoT 2018 - Proceedings

ISBN

978-1-4673-9944-9

ISSN

—

e-ISSN

neuvedeno

Počet stran výsledku

6

Strana od-do

511-516

Název nakladatele

IEEE

Místo vydání

Piscataway

Místo konání akce

Singapur

Datum konání akce

5. 2. 2018

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—