A Formalized Method for Pre-Domain Validation of a Multi-Agent Runtime with a Prospective Application to Power Supply
DOI:
https://doi.org/10.33042/2079-424X.2026.65.3.01Ключові слова:
Multi-Agent System, Artificial Intelligence, Power Supply, Technical Operation, Runtime Environment, Action Validation, DispatchingАнотація
We present a formalized method for pre-domain validation of an LLM-based multi-agent runtime and assess its feasibility through experiments. In the model, an agent proposes an action, and a central controller validates it while maintaining state, the event source, and the journal. Nine adaptive engineering trials were conducted on the Dice Ritual testbed with five persistent players under 13- and 15-category rulesets. The primary endpoint is automatic completion of every round with correct original final scorecards, bonuses, totals, and ranking; automatic player and controller corrections are allowed. Seven attempts produced conditionally verified correct final results, one completed with an incorrect final result, and one stopped before completion. Review of the seven correct results covered 35 scorecards and 475 categories. In one such game, an unrepaired intermediate display error did not change the score of the action actually selected; in another game, a category error caused an unwarranted bonus and an incorrect final total. Original strict verdicts and separate post-game adjudications are retained. The results support the possibility of correct automatic execution under tested conditions, without guaranteeing success in every run. Numerical findings depend on explicit reconstruction assumptions; the exact content of some delivered packets is unavailable. K5 is an auxiliary representation of dice values whose causal benefit was not evaluated. Physical correctness and operational benefit for an actual power system remain outside this assessment.
Посилання
1. Binyamin, S. S., & Ben Slama, S. (2022). Multi-agent systems for resource allocation and scheduling in a smart grid. Sensors, 22(21), 8099. https://doi.org/10.3390/s22218099
2. Costa, G. B., Damiani, J. S., Marchesan, G., Morais, A. P., Bretas, A. S., & Cardoso, G., Jr. (2022). A multi-agent approach to distribution system fault section estimation in smart grid environment. Electric Power Systems Research, 204, 107658. https://doi.org/10.1016/j.epsr.2021.107658
3. Debenedetti, E., Zhang, J., Balunović, M., Beurer-Kellner, L., Fischer, M., & Tramèr, F. (2024). AgentDojo: A dynamic environment to evaluate prompt injection attacks and defenses for LLM agents. Advances in Neural Information Processing Systems, 37, 82895–82920. https://doi.org/10.52202/079017-2636
4. Doshi, A., Hong, Y., Xu, C., Kang, E., Kapravelos, A., & Kästner, C. (2026). Towards verifiably safe tool use for LLM agents. In Proceedings of the IEEE/ACM 48th International Conference on Software Engineering (ICSE-NIER ’26) (pp. 201–205). https://doi.org/10.1145/3786582.3786839
5. Gai, O., Novoseltsev, O., Vorushylo, A., Khomyak, O., & Gai, H. (2024). Features of multi-agent evaluation of of microgrid systems efficiency in parallel operation with the power system according to the reliability criterion. Machinery & Energetics, 15(3), 73–83. https://doi.org/10.31548/machinery/3.2024.73
6. Hong, S., Zhuge, M., Chen, J., Zheng, X., Cheng, Y., Wang, J., … Schmidhuber, J. (2024). MetaGPT: Meta programming for a multi-agent collaborative framework. International Conference on Learning Representations. https://openreview.net/forum?id=VtmBAGCN7o
7. Huang, X., Ruan, W., Huang, W., Jin, G., Dong, Y., Wu, C., … Mustafa, M. A. (2024). A survey of safety and trustworthiness of large language models through the lens of verification and validation. Artificial Intelligence Review, 57(7), 175. https://doi.org/10.1007/s10462-024-10824-0
8. International Electrotechnical Commission. (2013). IEC 62443-3-3:2013 Industrial communication networks—Network and system security—Part 3-3: System security requirements and security levels. Geneva, Switzerland: Author.
9. International Organization for Standardization. (2024). ISO 55000:2024 Asset management—Vocabulary, overview and principles. Geneva, Switzerland: Author.
10. International Organization for Standardization & International Electrotechnical Commission. (2023). ISO/IEC 42001:2023 Information technology—Artificial intelligence—Management system. Geneva, Switzerland: Authors.
11. Izmirlioglu, Y., Pham, L., Son, T. C., & Pontelli, E. (2024). A survey of multi-agent systems for smartgrids. Energies, 17(15), 3620. https://doi.org/10.3390/en17153620
12. Li, G., Hammoud, H. A. A. K., Itani, H., Khizbullin, D., & Ghanem, B. (2023). CAMEL: Communicative agents for “mind” exploration of large language model society. Advances in Neural Information Processing Systems, 36, 51991–52008. https://doi.org/10.52202/075280-2264
13. Li, X., Wang, S., Zeng, S., Wu, Y., & Yang, Y. (2024). A survey on LLM-based multi-agent systems: Workflow, infrastructure, and challenges. Vicinagearth, 1(1), 9. https://doi.org/10.1007/s44336-024-00009-2
14. Liu, X., Yu, H., Zhang, H., Xu, Y., Lei, X., Lai, H., … Tang, J. (2024). AgentBench: Evaluating LLMs as agents. International Conference on Learning Representations. https://openreview.net/forum?id=zAdUB0aCTQ
15. Olorunfemi, B. O., & Nwulu, N. (2023). Multi-agent system implementation in demand response: A literature review and bibliometric evaluation. AIMS Energy, 11(6), 1179–1210. https://doi.org/10.3934/energy.2023054
16. Pérez Hernández, M., Puchkova, A., & Parlikad, A. K. (2024). Multi-agent learning of asset maintenance plans through localised subnetworks. Engineering Applications of Artificial Intelligence, 127, 107362. https://doi.org/10.1016/j.engappai.2023.107362
17. Rokhforoz, P., Montazeri, M., & Fink, O. (2023). Safe multi-agent deep reinforcement learning for joint bidding and maintenance scheduling of generation units. Reliability Engineering & System Safety, 232, 109081. https://doi.org/10.1016/j.ress.2022.109081
18. Russell, S., & Norvig, P. (2021). Artificial intelligence: A modern approach (4th ed., Global ed.). Harlow, England: Pearson.
19. Tabassi, E. (2023). Artificial intelligence risk management framework (AI RMF 1.0) (NIST AI 100-1). Gaithersburg, MD: National Institute of Standards and Technology. https://doi.org/10.6028/NIST.AI.100-1
20. Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., … Wen, J. (2024). A survey on large language model based autonomous agents. Frontiers of Computer Science, 18(6), 186345. https://doi.org/10.1007/s11704-024-40231-1
21. Wooldridge, M. (2009). An introduction to multiagent systems (2nd ed.). Chichester, England: Wiley.
22. Wu, Q., Bansal, G., Zhang, J., Wu, Y., Li, B., Zhu, E., … Wang, C. (2024). AutoGen: Enabling next-gen LLM applications via multi-agent conversations. Conference on Language Modeling (COLM 2024). https://openreview.net/forum?id=WpkVmCo8d0
23. Xiao, Z., Liu, H., Zhou, X., Cong, S., & Ju, Y. (2026). Power ops agent: A knowledge and data dual-driven multi-agent framework for substation operation and maintenance. Expert Systems with Applications, 318, 131965. https://doi.org/10.1016/j.eswa.2026.131965
Завантаження
Опубліковано
Номер
Розділ
Ліцензія
Авторське право (c) 2026 Світлотехніка
Ця робота ліцензується відповідно до ліцензії Creative Commons Attribution-NonCommercial 4.0 International License.
Фактом про прийняття автором умов і вимог, є надсилання автором своїх матеріалів для публікації в журналі ліцензіата.
За цією угодою автор показує, що поданий матеріал:
- не порушує авторські права інших осіб чи організацій;
- раніше не було опубліковано в інших журналах і не було подано для публікації в інших журналах.
Права, які автор передає до Редакційної ради з інженерії освітлення та електроенергії, включають, зокрема, права:
- опублікувати статтю англійською мовою та розповсюдити її друковану версію;
- розповсюджувати електронну версію статті будь-якими електронними засобами (розміщення на офіційному веб-сайті журналу в електронних базах даних, сховищах тощо).
Автор залишає за собою всі права за умовами Creative Commons Attribution License International CC-BY.
У всіх випадках наявність бібліографічного посилання на статтю або гіперпосилання на його електронну копію на офіційному веб-сайті журналу є обов'язковим.
@LepeJournal