Optimizing EV flexibility for spot and mFRR market participation
Authors
Jensen, Andreas Ravnholt ; Smedt, Mikkel Müller ; Pedersen, Raymond Asoklis Kronborg
Term
4. term
Education
Publication year
2025
Submitted on
2025-06-05
Pages
15
Abstract
Dette speciale undersøger, hvordan flåder af elbiler kan aggregeres og planlægges som én fleksibel enhed (FlexOffer) for at deltage i det danske day-ahead spotmarked og den manuelle frekvensgenopretningsreserve (mFRR). Hver elbils fleksibilitet modelleres som enten et Total Energy Constraint FlexOffer eller et afhængighedsbaseret FlexOffer, hvor sidstnævnte bedre fanger sammenhængen mellem tidsperioder og batteriets ladetilstand. Da markederne kræver minimumsbud på 1 MWh, aggregeres individuelle FlexOffers via klyngealgoritmer (agglomerativ og k-means) og tidsmæssig justering; vi introducerer to nye justeringsteknikker, fleksibilitetsalignment og en hurtig variant, fast fleksibilitetsalignment. Herefter anvendes en idealiseret to-trins lineær programmeringsmodel med perfekte prisprognoser, hvor trin 1 afgiver bud i spot- og reservemarkedet, og trin 2 tilpasser driften til faktiske mFRR-aktiveringer. Tre planlægningsstrategier sammenlignes: sekventiel (reserve før spot), fælles ko-optimering og kun-spot. Eksperimenter over en måned med op til 500.000 elbiler viser, at afhængighedsbaserede FlexOffers konsekvent overgår totalenergi-varianten, at fælles ko-optimering giver de største besparelser (op til 92,14% af et teoretisk optimum), og at fast fleksibilitetsalignment markant forbedrer køretid med kun marginalt lavere besparelser. Arbejdet bidrager med hurtigere aggregering, en LP-ramme der ko-optimerer spot, mFRR-kapacitet og aktivering under realistisk føreradfærd, samt dokumenteret skalerbarhed til meget store flåder; resultaterne forudsætter dog perfekte prisforudsigelser.
This thesis investigates how fleets of electric vehicles can be aggregated and scheduled as a single flexible unit (FlexOffer) to participate in Denmark’s day-ahead spot market and the manual frequency restoration reserve (mFRR) market. Each vehicle’s charging flexibility is modeled as either a Total Energy Constraint FlexOffer or a dependency-based FlexOffer, the latter capturing inter-temporal dependencies and state of charge more precisely. Because market minimum bid sizes are 1 MWh, individual FlexOffers are clustered using agglomerative and k-means methods and aligned in time; we introduce two alignment techniques, flexibility alignment and a faster variant, fast flexibility alignment. An idealized two-stage linear programming model with perfect price forecasts then places bids (stage 1) and adjusts operations to actual mFRR activations (stage 2). We compare three scheduling strategies: sequential (reserve before spot), joint co-optimization, and spot-only. One-month experiments with fleets up to 500,000 EVs show that dependency-based FlexOffers consistently outperform total-energy FlexOffers, joint co-optimization delivers the highest savings (up to 92.14% of a theoretical optimum), and fast flexibility alignment substantially reduces runtime with only a slight reduction in savings. The contributions include faster aggregation, an LP framework that co-optimizes spot, mFRR capacity, and activation under realistic driver behavior, and demonstrated scalability to very large fleets, noting the assumption of perfect price forecasts.
[This abstract was generated with the help of AI]
Documents