Intelligent Supervisory Layer Based on a Language Model for Real-Time Optimization Systems

Aleksandr Karmanov

Abstract


This paper presents a conceptual framework for integrating large language models (LLMs) into hierarchical industrial control architectures based on Real-Time Optimization (RTO) and Model Predictive Control (MPC). The proposed supervisory LLM layer serves as a cognitive interface between economic optimization and predictive regulation, providing context-aware adaptation of setpoints, constraints, and cost function weights. A structured communication protocol is introduced, enabling the agent to receive system measurements, historical time-series data, and external contextual information through a Model Context Protocol (MCP) interface. The LLM interprets these heterogeneous inputs and generates machine-readable recommendations for the RTO layer, which validates them through feasibility and stability checks before application. This ensures that cognitive reasoning remains safely encapsulated within the formal guarantees of hierarchical control. The resulting architecture maintains the mathematical rigor of RTO–MPC while extending its adaptability to unstructured knowledge sources, such as schedules, forecasts, and operational reports. The approach enables semantic adaptation of control parameters without compromising real-time performance and offers potential for implementation in energy management, refrigeration systems, and intelligent building operations.

Full Text:

PDF (Russian)

References


Vaccari M., Pannocchia G. A modifier-adaptation strategy towards offset-free economic MPC //Processes. – 2016. – Т. 5. – №. 1. – С. 2.

Afram A., Janabi-Sharifi F. Theory and applications of HVAC control systems–A review of model predictive control (MPC) //Building and environment. – 2014. – Т. 72. – С. 343-355.

Serale G. et al. Model predictive control (MPC) for enhancing building and HVAC system energy efficiency: Problem formulation, applications and opportunities //Energies. – 2018. – Т. 11. – №. 3. – С. 631.

Darby M. L. et al. RTO: An overview and assessment of current practice //Journal of Process control. – 2011. – Т. 21. – №. 6. – С. 874-884.

Hovgaard T. G. et al. Model predictive control technologies for efficient and flexible power consumption in refrigeration systems //Energy. – 2012. – Т. 44. – №. 1. – С. 105-116.

Bujarbaruah M. et al. Adaptive MPC for iterative tasks //2018 IEEE Conference on Decision and Control (CDC). – IEEE, 2018. – С. 6322-6327.

Wu R., Ai J., Li T. InstructMPC: A Human-LLM-in-the-Loop Framework for Context-Aware Control //arXiv preprint arXiv:2504.05946. – 2025.

Tøndel P., Johansen T. A., Bemporad A. An algorithm for multi-parametric quadratic programming and explicit MPC solutions //Automatica. – 2003. – Т. 39. – №. 3. – С. 489-497.

Huang X. et al. Understanding the planning of LLM agents: A survey //arXiv preprint arXiv:2402.02716. – 2024.

Ahmadi A., Sharif S., Banad Y. M. MCP Bridge: A Lightweight, LLM-Agnostic RESTful Proxy for Model Context Protocol Servers //arXiv preprint arXiv:2504.08999. – 2025.

Adamo F. et al. SCADA/HMI systems in advanced educational courses //IEEE Transactions on Instrumentation and measurement. – 2007. – Т. 56. – №. 1. – С. 4-10.

White J. et al. A prompt pattern catalog to enhance prompt engineering with chatgpt //arXiv preprint arXiv:2302.11382. – 2023.

Jin M. et al. Time-llm: Time series forecasting by reprogramming large language models //arXiv preprint arXiv:2310.01728. – 2023.

Wang X. et al. From news to forecast: Integrating event analysis in llm-based time series forecasting with reflection //Advances in Neural Information Processing Systems. – 2024. – Т. 37. – С. 58118-58153.

Ji Z. et al. Towards mitigating LLM hallucination via self reflection //The 2023 Conference on Empirical Methods in Natural Language Processing. – 2023.


Refbacks

  • There are currently no refbacks.


Abava  Кибербезопасность ИТ-КОНГРЕСС ВМК МГУ 2026 СНЭ

ISSN: 2307-8162