Demand-side thermal modeling-driven control optimization for energy-efficient industrial space cooling

Abstract

The industrial sector represents a major contributor to global energy consumption and carbon emissions, with industrial buildings accounting for approximately one-third of building energy use. The substantial energy demand for cooling industrial spaces, particularly in high-tech manufacturing facilities, highlights the need for strategies to reduce energy consumption. However, the effective implementation of water-side HVAC control strategies is hindered by a lack of robust, accurate cooling load prediction—particularly under small sample sizes or incomplete data, which are common in industrial settings. Therefore, this thesis proposes an integrated solution that combines demand-side HVAC system modeling with water-side system control optimization, achieving energy-efficient industrial space cooling.
Cooling load prediction is the key for control optimization of the whole cooling system. To address the lack of generalizable, physically interpretable models for complex industrial settings, this research develops a parametric, encapsulated modeling framework using Modelica. It simplifies the representation of large-scale production spaces and air-side systems while maintaining prediction accuracy. Key innovations include: (1) the use of the Morris method for sensitivity analysis to identify parameters critical to predicting cooling load, indoor air temperature, and relative humidity; (2) a machine learning sub-model, address missing or incomplete facility data; and (3) genetic algorithm-based calibration to align model predictions with measured data. Practical validation studies in three real industrial plants shows that the framework achieves a cooling load prediction error below 8%, outperforming traditional black-box and grey-box models by over 30%, and showing stronger robustness under the condition of missing or incomplete data.
Building on the proposed modeling framework, this paper introduces a decoupled, step-wise control optimization framework for water-side cooling systems. Unlike MPC, DRL, or hybrid strategies that rely on high computational cost, black-box approximations, or opaque decision processes, the proposed approach leverages physical modeling to separately optimize thermal load prediction, hydraulic behavior, and system-level energy use. This framework dynamically adjusts operations of water-side system and distribution system based on a hierarchical optimization process. A modeling approach based on Modelica similar to that in the modeling framework is adopted as the model for control optimization, including thermal response model, hydraulic model and water-side system model. Thermal response models first determine cooling load of the production space and the corresponding chilled water flow rates and return water temperatures. This is followed by hydraulic optimization and chiller optimization, to minimize pump and chiller energy consumption to identify energy-efficient configurations, including pump/chiller sequencing and chilled water temperature. Simulation results from the practical validation study demonstrate that, compared to rule-based control strategies, this framework reduces pump energy by 9.58%, air-cooled chiller energy by 30.90%, and water-cooled chiller energy by 21.68%.
In the future, further research could explore the integration of this control framework with real-time data from smart building systems to enhance its responsiveness to dynamic operating conditions. Additionally, applying the framework to other climate zones and different types of industrial processes could expand its applicability and refine its predictive capabilities. Investigating the potential for incorporating renewable energy sources, such as photovoltaic systems or thermal storage, into the control strategy could further enhance energy efficiency and contribute to broader sustainability goals in industrial operations.

Date of Award	15 Jul 2026
Original language	English
Awarding Institution	University of Nottingham
Supervisor	Dengfeng Du (Supervisor), Zhiang Zhang (Supervisor) & Rabee Reffat (Supervisor)