An optimization scheduling strategy for hydrogen-based integrated energy systems using multi-agent deep reinforcement learning

[Display omitted] •An optimized scheduling strategy for hydrogen based IES was proposed using MADRL.•Variable temperature PEMFC mechanism is considered to improve energy supply flexibility.•Time series segment is used to improve the integrity of state information.•Decision variable grouping is used...

Celý popis

Uložené v:
Podrobná bibliografia
Vydané v:Energy conversion and management Ročník 326
Hlavní autori: Zhang, Lei, He, Ye, Hatziargyriou, Nikos D., Wu, Hongbin, Han, Pingping
Médium: Journal Article
Jazyk:English
Vydavateľské údaje: Elsevier Ltd 15.02.2025
Predmet:
administrative management
algorithms
case studies
electricity
energy conversion
fuels
heat
hydrogen
Hydrogen energy utilization
Integrated energy systems
issues and policy
Multi agent deep reinforcement learning
Multi-objective optimization
Optimal scheduling
time series analysis
uncertainty
wind
Multi agent deep reinforcement learning
Multi-objective optimization
Hydrogen energy utilization
Integrated energy systems
Optimal scheduling
ISSN:0196-8904
On-line prístup:Získať plný text
Tagy: Pridať tag
Žiadne tagy, Buďte prvý, kto otaguje tento záznam!
Popis
Shrnutí:[Display omitted] •An optimized scheduling strategy for hydrogen based IES was proposed using MADRL.•Variable temperature PEMFC mechanism is considered to improve energy supply flexibility.•Time series segment is used to improve the integrity of state information.•Decision variable grouping is used to improve the scheduling performance of agents. The introduction of hydrogen energy utilization will further highlight the high dimensionality, nonlinearity, and uncertainty of integrated energy systems, thereby increasing the difficulty of their scheduling strategy formulation. In this paper, a method based on multi-agent deep reinforcement learning for optimizing the scheduling of integrated energy systems is proposed. Firstly, a comprehensive energy system model was established, integrating wind, solar, and hydrogen coupling to provide heating, cooling, and electricity. Subsequently, flexible model for proton exchange membrane fuel cells was developed, and its electrical and thermal output characteristics were analyzed. Four scheduling performance evaluation indicators were defined, and a reward function for deep reinforcement learning was constructed based on the Technique for Order of Preference by Similarity to Ideal Solution approach. Combining the strategy of representing time series segment states and decision variable grouping training, the multi-agent twin delayed deep deterministic policy gradients algorithm was improved to establish a scheduling strategy learning model. Finally, a case study analysis is conducted to compare the proposed method with the baseline approach, confirming that the proposed approach significantly enhances the flexibility of scheduling strategies, ensures a balance between low-carbon emissions and economic efficiency, and ultimately improves the overall scheduling performance.
Bibliografia:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:0196-8904
DOI:10.1016/j.enconman.2025.119483