Loading [MathJax]/extensions/tex2jax.js

Design and Dispatch of Decentralized Energy Systems using Artificial Neural Networks

Research output: Chapter in book/report/conference proceedingConference contributionResearchpeer review

Authors

External Research Organisations

  • WETEC Systems GmbH

Details

Original languageEnglish
Title of host publicationConference Proceedings of the 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems
Subtitle of host publicationECOS 2024
Pages2307-2318
Number of pages12
ISBN (electronic)9798331307660
Publication statusPublished - 30 Jun 2024
Event37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems, ECOS 2024 - Rhodes, Greece
Duration: 30 Jun 20245 Jul 2024

Abstract

Decentralized energy systems, pivotal in transitioning towards a sustainable energy future, require intelligent dispatch strategies for the operation of flexible components in order to integrate inflexible renewable energy sources economically. Conventional rule-based dispatch strategies often fail to optimally exploit the capabilities of flexible system components, while optimal dispatch models, based on the assumption of perfect forecasting, tend to overestimate their performance. This paper investigates the effectiveness of artificial neural networks (ANNs) as a dynamic dispatch strategy for distributed energy systems (DES), evaluating their performance in operational scenarios with a predefined system layout and during the system design optimization phase. Our analysis shows that ANN-based dispatch strategies outperform conventional rule-based methods by up to 8.19% in operational efficiency according to training datasets and by 3.19% in validation datasets. However, they fall short of optimal dispatch strategies by 4.52% and 1.59% in training and validation datasets, respectively. When applied to system design optimization, ANN-based strategies outperform rule-based approaches by 5.80-9.19% but underperform against optimal dispatch designs by 10.63%. Crucially, the study highlights that dispatch strategies not only influence overall system costs but also significantly impact the sizing and configuration of individual system components. This underlines the importance of incorporating intelligent dispatch strategies like ANNs early in the design process to ensure a balanced and cost-effective system architecture.

ASJC Scopus subject areas

Sustainable Development Goals

Cite this

Design and Dispatch of Decentralized Energy Systems using Artificial Neural Networks. / Koenemann, Lukas; Bensmann, Astrid; Gerster, Johannes et al.
Conference Proceedings of the 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems: ECOS 2024. 2024. p. 2307-2318.

Research output: Chapter in book/report/conference proceedingConference contributionResearchpeer review

Koenemann, L, Bensmann, A, Gerster, J & Hanke-Rauschenbach, R 2024, Design and Dispatch of Decentralized Energy Systems using Artificial Neural Networks. in Conference Proceedings of the 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems: ECOS 2024. pp. 2307-2318, 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems, ECOS 2024, Rhodes, Greece, 30 Jun 2024. https://doi.org/10.52202/077185-0198
Koenemann, L., Bensmann, A., Gerster, J., & Hanke-Rauschenbach, R. (2024). Design and Dispatch of Decentralized Energy Systems using Artificial Neural Networks. In Conference Proceedings of the 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems: ECOS 2024 (pp. 2307-2318) https://doi.org/10.52202/077185-0198
Koenemann L, Bensmann A, Gerster J, Hanke-Rauschenbach R. Design and Dispatch of Decentralized Energy Systems using Artificial Neural Networks. In Conference Proceedings of the 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems: ECOS 2024. 2024. p. 2307-2318 doi: 10.52202/077185-0198
Koenemann, Lukas ; Bensmann, Astrid ; Gerster, Johannes et al. / Design and Dispatch of Decentralized Energy Systems using Artificial Neural Networks. Conference Proceedings of the 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems: ECOS 2024. 2024. pp. 2307-2318
Download
@inproceedings{b5c158e810884ba68335310cadfc637d,
title = "Design and Dispatch of Decentralized Energy Systems using Artificial Neural Networks",
abstract = "Decentralized energy systems, pivotal in transitioning towards a sustainable energy future, require intelligent dispatch strategies for the operation of flexible components in order to integrate inflexible renewable energy sources economically. Conventional rule-based dispatch strategies often fail to optimally exploit the capabilities of flexible system components, while optimal dispatch models, based on the assumption of perfect forecasting, tend to overestimate their performance. This paper investigates the effectiveness of artificial neural networks (ANNs) as a dynamic dispatch strategy for distributed energy systems (DES), evaluating their performance in operational scenarios with a predefined system layout and during the system design optimization phase. Our analysis shows that ANN-based dispatch strategies outperform conventional rule-based methods by up to 8.19% in operational efficiency according to training datasets and by 3.19% in validation datasets. However, they fall short of optimal dispatch strategies by 4.52% and 1.59% in training and validation datasets, respectively. When applied to system design optimization, ANN-based strategies outperform rule-based approaches by 5.80-9.19% but underperform against optimal dispatch designs by 10.63%. Crucially, the study highlights that dispatch strategies not only influence overall system costs but also significantly impact the sizing and configuration of individual system components. This underlines the importance of incorporating intelligent dispatch strategies like ANNs early in the design process to ensure a balanced and cost-effective system architecture.",
author = "Lukas Koenemann and Astrid Bensmann and Johannes Gerster and Richard Hanke-Rauschenbach",
note = "Publisher Copyright: {\textcopyright} 2024 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems, ECOS 2024. All rights reserved.; 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems, ECOS 2024 ; Conference date: 30-06-2024 Through 05-07-2024",
year = "2024",
month = jun,
day = "30",
doi = "10.52202/077185-0198",
language = "English",
pages = "2307--2318",
booktitle = "Conference Proceedings of the 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems",

}

Download

TY - GEN

T1 - Design and Dispatch of Decentralized Energy Systems using Artificial Neural Networks

AU - Koenemann, Lukas

AU - Bensmann, Astrid

AU - Gerster, Johannes

AU - Hanke-Rauschenbach, Richard

N1 - Publisher Copyright: © 2024 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems, ECOS 2024. All rights reserved.

PY - 2024/6/30

Y1 - 2024/6/30

N2 - Decentralized energy systems, pivotal in transitioning towards a sustainable energy future, require intelligent dispatch strategies for the operation of flexible components in order to integrate inflexible renewable energy sources economically. Conventional rule-based dispatch strategies often fail to optimally exploit the capabilities of flexible system components, while optimal dispatch models, based on the assumption of perfect forecasting, tend to overestimate their performance. This paper investigates the effectiveness of artificial neural networks (ANNs) as a dynamic dispatch strategy for distributed energy systems (DES), evaluating their performance in operational scenarios with a predefined system layout and during the system design optimization phase. Our analysis shows that ANN-based dispatch strategies outperform conventional rule-based methods by up to 8.19% in operational efficiency according to training datasets and by 3.19% in validation datasets. However, they fall short of optimal dispatch strategies by 4.52% and 1.59% in training and validation datasets, respectively. When applied to system design optimization, ANN-based strategies outperform rule-based approaches by 5.80-9.19% but underperform against optimal dispatch designs by 10.63%. Crucially, the study highlights that dispatch strategies not only influence overall system costs but also significantly impact the sizing and configuration of individual system components. This underlines the importance of incorporating intelligent dispatch strategies like ANNs early in the design process to ensure a balanced and cost-effective system architecture.

AB - Decentralized energy systems, pivotal in transitioning towards a sustainable energy future, require intelligent dispatch strategies for the operation of flexible components in order to integrate inflexible renewable energy sources economically. Conventional rule-based dispatch strategies often fail to optimally exploit the capabilities of flexible system components, while optimal dispatch models, based on the assumption of perfect forecasting, tend to overestimate their performance. This paper investigates the effectiveness of artificial neural networks (ANNs) as a dynamic dispatch strategy for distributed energy systems (DES), evaluating their performance in operational scenarios with a predefined system layout and during the system design optimization phase. Our analysis shows that ANN-based dispatch strategies outperform conventional rule-based methods by up to 8.19% in operational efficiency according to training datasets and by 3.19% in validation datasets. However, they fall short of optimal dispatch strategies by 4.52% and 1.59% in training and validation datasets, respectively. When applied to system design optimization, ANN-based strategies outperform rule-based approaches by 5.80-9.19% but underperform against optimal dispatch designs by 10.63%. Crucially, the study highlights that dispatch strategies not only influence overall system costs but also significantly impact the sizing and configuration of individual system components. This underlines the importance of incorporating intelligent dispatch strategies like ANNs early in the design process to ensure a balanced and cost-effective system architecture.

UR - http://www.scopus.com/inward/record.url?scp=85216918890&partnerID=8YFLogxK

U2 - 10.52202/077185-0198

DO - 10.52202/077185-0198

M3 - Conference contribution

AN - SCOPUS:85216918890

SP - 2307

EP - 2318

BT - Conference Proceedings of the 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems

T2 - 37th International Conference on Efficiency, Cost, Optimization, Simulation and Environmental Impact of Energy Systems, ECOS 2024

Y2 - 30 June 2024 through 5 July 2024

ER -

By the same author(s)