Dealing with uncertainties in the operational planning of district heating plants
Author
Refsgaard, Jonathan
Term
4. Term
Publication year
2021
Abstract
Specialet undersøger, hvordan et dansk fjernvarmeværk som prisstiller kan beregne optimale bud i flere dereguleringerede elmarkeder med marginalprissætning under usikkerhed for at understøtte sektorkobling i et fremtidigt 100% vedvarende energisystem. Assens Fjernvarme anvendes som case, og fire budmetoder sammenlignes: prisuafhængig budgivning, et enkeltbud afledt heraf med en øvre prisgrænse, den state-of-the-art varmeenheds-udskiftningsmetode anvendt i energiplanlægningsværktøjet energyTRADE, samt en scenariebaseret metode med Sample Average Approximation (SAA). Prognoser og scenarier for vejr og spotpriser genereres med Monte Carlo-simulering og Markovkæder trænet på historiske data for solindstråling, temperatur, vind og priser; reguleringspriser modelleres med Monte Carlo og regression. Metoderne indlejres i en realistisk multi-stage mixed-integer lineær programmeringsmodel med fem dages rullende planlægningshorisont, der simulerer drift og budgivning i sekventielle elmarkeder under usikkerhed. Resultaterne viser, at SAA-metoden beregner bud, som giver både teknisk-økonomisk optimal drift og bedre sektorkobling ved at købe og sælge el efter markedets prissignaler. Desuden viser resultaterne, at deltagelse i flere elmarkeder kan reducere driftsomkostningerne markant, og at suboptimale day-ahead-bud kan afbødes med efterfølgende bud i reguleringsmarkedet.
This thesis examines how a Danish district heating plant acting as a price-taker can compute optimal bids across multiple deregulated electricity markets with marginal pricing under uncertainty to support sector coupling in a future 100% renewable energy system. Using Assens District Heating as an instrumental case, four bidding approaches are compared: price-independent bidding, a single-bid variant with an upper price cap, the state-of-the-art heat unit replacement method used in the operational tool energyTRADE, and a scenario-based method using Sample Average Approximation (SAA). Forecasts and scenarios for weather and spot prices are generated via Monte Carlo simulation and Markov chains trained on historical solar irradiation, temperature, wind, and price data; regulating prices are modeled with Monte Carlo and regression. The methods are embedded in a realistic multi-stage mixed-integer linear programming model with a five-day rolling planning horizon that simulates operation and sequential market participation under uncertainty. Results indicate that the SAA-based method produces bids leading to techno-economic optimal operation and improved sector coupling by aligning electricity purchases and sales with market price signals. The findings also show that participating in multiple electricity markets can significantly lower operating costs, and that suboptimal day-ahead positions can be mitigated through subsequent bids in the regulating market.
[This summary has been generated with the help of AI directly from the project (PDF)]
Documents