We help you find the perfect fit.

Swiss Ai Research Overview Platform

28 Research Topics
Reset all filters
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Filter
Reset all filters
Select all Unselect all
Close Show projects Back
Show all filters
71 Application Fields
Reset all filters
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Filter
Reset all filters
Select all Unselect all
Close Show projects Back
Show all filters
34 Institutions
Reset all filters
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Select all Unselect all
Close
Filter
Reset all filters
Select all Unselect all
Close Show projects Back
Show all filters
Machine Learning for Demand Response (ML4DR)

Lay summary

Inhalt und Ziel des Forschungsprojekts

-Entwickeln und Testen von Algorithmen für maschinelles Lernen zur Abschätzung der Bereitschaft von Kunden, Teillastbeschränkungen verschiedener Geräte zu akzeptieren

-Test der Eignung verschiedener Datenquellen für die Prognosen

-Vorbereitung des Designs einer anschließenden Feldstudie zur Untersuchung der Effektivität der Algorithmen im Vergleich zu Vorhersagen auf Basis von Umfragen

Wissenschaftlicher und gesellschaftlicher Kontext des Forschungsprojekts

Im Rahmen dieser Arbeit wird ein Prototyp eines verbesserten Lastmanagement-Algorithmus erstellt, der in anschließenden Feldversuchen getestet werden kann. Ein verbessertes Lastmanagement könnte dazu beitragen, den Netzausbaubedarf zu reduzieren und die Integration größerer Anteile erneuerbarer Energien zu geringeren Kosten zu ermöglichen.

Abstract

In many countries, renewable energy investment is starting to become economically viable without support. This is leading to the disruption of electric utility’ business and potential grid expansion needs in order to accommodate the increasing amount of renewable electricity. Therefore, it is necessary to find out solutions to minimize these concerns. One of the promising solutions to this problem is the increasing usage of demand response (partial load restrictions). To leverage demand response, it is important to understand customer preferences for adjusting electricity demand to match with power supply. A conventional method to engage consumers is to assess their willingness to allow load control using a survey approach and dispatch their flexibility assuming constant preferences. However, static estimates based on surveys are likely to be imprecise, because consumers have limited experience with demand restrictions and because consumer preferences will vary across time. Allowing consumers to override device settings, has been shown to increase acceptance of automation, because consumers do not need to anticipate their load shifting cost. Automatic updates of consumer preferences based on override actions could thus reduce the need for static assessments of customer willingness to allow load control and revolutionize the way that the power system is operated. This research thus plans to develop machine learning algorithms for identifying time-varying customer willingness to accept demand response and test how these could be used to automate the price-based device control based on consumer actions in real-time. The efficiency of different learning algorithms will also be assessed using a simulated test case. The practice partner will provide anonymized smart meter data from their customers that can be used to calibrate the machine learning models in a test case. Subject to the agreement of customers, we will investigate to what extent the predictions of consumer willingness to allow load shifting can be improved by the usage of unconventional data sources, such as calendars, location services, social media profiles etc. This work also provides the design of subsequent field studies to compare the effectiveness between machine learning approaches and survey-based methods (future works).

Last updated:1646210231

Aksornchan Chaianong