THE ROLE OF MLOPS IN ADVANCING GREEN ENERGY: AN EVALUATION OF TECHNOLOGIES AND PRACTICES
DOI:
https://doi.org/10.32689/maup.it.2025.2.12Keywords:
Machine learning, MLOps, Virtual power plant, Distributed energy, Cloud technology, ForecastingAbstract
Machine learning–driven decision making is essential for the efficient operation of cloud-hosted virtual power plants (VPPs) aggregating hundreds to thousands of distributed energy resources (DERs). However, manually deploying and maintaining ML models at scale introduces delays, inconsistency and high operational overhead. In this paper, we survey six widely adopted MLOps frameworks–Kubeflow, Apache Airflow, MLflow, Azure ML, AWS SageMaker and Google Vertex AI– against four criteria critical to VPP environments: industry adoption, feature-set completeness, interoperability with major ML frameworks and cloud platforms, and licensing or cost constraints. Drawing on public documentation, repository activity, case studies and market research, we identify trade-offs between open-source flexibility and managed-service convenience.Our analysis shows that Apache Airflow offers the most mature and extensible pipeline orchestration for on-premise and multi-cloud VPP deployments, while Kubeflow excels in Kubernetes-native contexts. Managed services like SageMaker and Azure ML deliver faster time-to-value for teams lacking dedicated infrastructure expertise but incur higher costs and vendor lock-in. Finally, we provide domain-tailored recommendations for integrating continuous training, evaluation and monitoring into VPP forecasting workflows, demonstrating how MLOps adoption can improve prediction latency and grid responsiveness.The goal of this article is to evaluate and compare leading MLOps frameworks–open-source and managed cloud services– against key criteria (adoption, feature completeness, interoperability, and cost) and to recommend the most suitable solutions for cloud-hosted virtual power plants.Methodology. We selected six MLOps frameworks based on adoption, features, interoperability and cost; extracted data from official docs, repositories and market reports; scored each tool against our criteria; and distilled domain-specific recommendations for cloud-hosted VPPs.Scientific Novelty. This article explores the under-researched intersection of MLOps and virtual power plants (VPPs), addressing the specific challenges of applying automated ML workflows to large-scale, cloud-hosted VPP systems. It provides the first domain-specific comparison of MLOps tools tailored to the operational and forecasting needs of VPPs.Conclusion. MLOps can significantly enhance the performance and scalability of virtual power plants. This study identifies the most suitable tools for VPP use cases, highlighting Apache Airflow and Kubeflow as strong open-source options, while managed services may suit teams with limited infrastructure expertise.
References
“Azure Machine Learning – ML as a Service | Microsoft Azure.” Accessed: Apr. 13, 2024. URL: https://azure.microsoft.com/en-gb/products/machine-learning
“Azure Machine Learning vs Amazon SageMaker: Data Science And Machine Learning Comparison,” 6sense. Accessed: Apr. 13, 2024. URL: https://www.6sense.com/tech/data-science-and-machine-learning/azuremachinelearning-vs-amazonsagemaker
D. Kreuzberger, N. Kühl, and S. Hirschl, “Machine Learning Operations (MLOps): Overview, Definition, and Architecture,” IEEE Access, 2023. vol. 11, pp. 31866–31879, doi: 10.1109/ACCESS.2023.3262138.
E. Peltonen and S. Dias, “LinkEdge: Open-sourced MLOps Integration with IoT Edge,” in Proceedings of the 3rd Eclipse Security, AI, Architecture and Modelling Conference on Cloud to Edge Continuum, Ludwigsburg Germany: ACM, Oct. 2023, pp. 67–76. doi: 10.1145/3624486.3624496.
F. Sattar, A. Husnain, and T. Ghaoud, “Integration of Distributed Energy Resources into a Virtual Power Plant-A Pilot Project in Dubai,” 2023 IEEE/IAS Industrial and Commercial Power System Asia (I&CPS Asia), pp. 526–531, Jul. 2023, doi: 10.1109/ICPSAsia58343.2023.10294691.
G. Robles, A. Capiluppi, J. M. Gonzalez-Barahona, B. Lundell, and J. Gamalielsson, “Development effort estimation in free/open source software from activity in version control systems,” Empir Software Eng, 2022. vol. 27, no. 6, p. 135, doi: 10.1007/s10664-022-10166-x.
H.-M. Chung, S. Maharjan, Y. Zhang, F. Eliassen, and K. Strunz, “Optimal Energy Trading With Demand Responses in Cloud Computing Enabled Virtual Power Plant in Smart Grids,” IEEE Trans. Cloud Comput., 2022. vol. 10, no. 1, pp. 17–30, doi: 10.1109/TCC.2021.3118563.
J. Rodríguez-García, D. Ribó-Pérez, C. Álvarez-Bel, and E. Peñalvo-López, “Novel Conceptual Architecture for the Next-Generation Electricity Markets to Enhance a Large Penetration of Renewable Energy,” Energies, 2019. vol. 12, no. 13, Art. no. 13, doi: 10.3390/en12132605.
K. Salama, J. Kazmierczak, and D. Schut, “Practitioners guide to MLOps: A framework for continuous delivery and automation of machine learning.”.
M. Merenda, C. Porcaro, and D. Iero, “Edge Machine Learning for AI-Enabled IoT Devices: A Review,” Sensors, 2020. vol. 20, no. 9, p. 2533, doi: 10.3390/s20092533.
“Machine Learning Service – Amazon SageMaker – AWS,” Amazon Web Services, Inc. Accessed: Apr. 13, 2024. URL: https://aws.amazon.com/sagemaker/
“MLOps: Continuous delivery and automation pipelines in machine learning. Cloud Architecture Center,” Google Cloud. Accessed: Mar. 16, 2024. URL: https://cloud.google.com/architecture/mlops-continuous-delivery-and-automation- pipelines-in-machine-learning
N. Naval and J. M. Yusta, “Virtual power plant models and electricity markets – A review,” Renewable and Sustainable Energy Reviews, 2021. vol. 149, p. 111393, doi: 10.1016/j.rser.2021.111393.
R. S. Kenett, X. Franch, A. Susi, and N. Galanis, “Adoption of Free Libre Open Source Software (FLOSS): A Risk Management Perspective,” 2014 IEEE 38th Annual Computer Software and Applications Conference, pp. 171–180, Jul. 2014, doi: 10.1109/COMPSAC.2014.25.
R. Subramanya, S. Sierla, and V. Vyatkin, “From DevOps to MLOps: Overview and Application to Electricity Market Forecasting,” Applied Sciences, 2022. vol. 12, no. 19, Art. no. 19, doi: 10.3390/app12199851.
T.-Y. Kim and S.-B. Cho, “Predicting residential energy consumption using CNN-LSTM neural networks,” Energy, 2019. vol. 182, pp. 72–81, Sep. doi: 10.1016/j.energy.2019.05.230.
V. Omelčenko and V. Manokhin, “Optimal Balancing of Wind Parks with Virtual Power Plants,” in Frontiers in Energy Research, Nov. 2021, p. 665295. doi: 10.3389/fenrg.2021.665295.
“Vertex AI,” Google Cloud. Accessed: Apr. 13, 2024. URL: https://cloud.google.com/vertex-ai
W. Shi, J. Cao, Q. Zhang, Y. Li, and L. Xu, “Edge Computing: Vision and Challenges,” IEEE Internet Things J., vol. 3, no. 5, pp. 637–646, Oct. 2016, doi: 10.1109/JIOT.2016.2579198.
