New model-based predictive control system uses Neural ODEs to help buildings improve efficiency, with early testing showing energy savings of up to 19%.
SWORDS, Ireland—August 27, 2026 — Trane Technologies (NYSE: TT), a global climate innovator, today announced a major milestone in building automation research: the development of a Model‑Based Predictive Control (MPC) system powered by Neural Ordinary Differential Equations (Neural ODEs) forecasting. The MPC system combines intelligent optimization with advanced predictive modeling to deliver near‑optimal control of building performance, while Neural ODEs provide a powerful, flexible prediction engine that helps unlock new levels of energy efficiency, comfort and cost savings.
Based on research conducted at the BrainBox AI Trane Technologies AI Lab, with support and contributions from leading AI researchers at Mila, Quebec Artificial Intelligence Institute, the concept is akin to modern navigation apps that optimize the most efficient travel routes while balancing user-defined constraints, such as avoiding highways or tolls. MPC with Neural ODEs applies the same principle to building systems. This technique requires significantly less cloud computing energy than traditional Long Short-Term Memory (LSTM) models, supporting more sustainable and efficient AI use.
Operationally, the system can pre-cool a building in the summer or pre-heat it in the winter ahead of peak energy price periods, then minimize system operation when prices are highest, while maintaining predefined comfort parameters. In early testing, the control strategy demonstrated energy savings of up to 19%, exceeding BrainBox AI’s existing control algorithms by more than 3%. While still in the exploratory phase, these findings demonstrate strong potential for commercialization, following further advancement through field trials as a next step.
“MPC with Neural ODEs demonstrates what is possible when cutting‑edge mathematical modeling meets real‑world building challenges,” shared Jean-Simon Venne, President, Founder, Chief Technology Officer of BrainBox AI and Head of the AI Lab. “It allows us to simulate thousands of potential future control strategies in seconds and choose the path that positively impacts costs, emissions and occupant comfort. It marks a giant leap in the advancement of true predictive autonomy for HVAC systems, and we are proud to lead the industry in this new era of smart building control.”
The system uses a physics-informed Neural ODE model to predict thermal behavior. This enables smart buildings to proactively manage building energy systems rather than simply reacting to real-time changes. Unlike traditional reactive control, which waits for sensors to detect changes before taking action, predictive control leverages forecasts to determine an optimal path while simultaneously balancing multiple objectives and constraints.
Key features of MPC with Neural ODEs include:
-
Objective-driven control: Balances multiple objectives, such as lowering energy costs and reducing carbon emissions.
-
Scalable deployment: Requires less training data and fewer computational resources, accelerating onboarding of new buildings.
-
Customizable constraints: Integrates parameters such as comfort ranges, equipment cycling frequency and grid emissions profiles for each building.
-
Real-time re-optimization: Updates optimization at each control cycle, adapting instantly to changing conditions, sensor inputs and forecasts.
-
Transparent insights: Provides clear, user-friendly feedback and reporting to help users and building managers understand what drives optimization choices.
MPC with Neural ODEs is engineered to support a wide range of applications, from demand response to overall energy reduction, by adjusting the objective function and weighting factors such as peak‑demand cost or total energy use. This flexibility enables buildings to intelligently reduce peak demand use and respond dynamically to grid conditions. By incorporating emissions data from energy providers, the approach can also help buildings improve sustainability outcomes alongside operational efficiency.
To learn more about the technical research, read the published findings in “Neural differential equations for temperature control in buildings under demand response programs,” published in Applied Energy and available on ScienceDirect.
# # #
About Trane Technologies
Trane Technologies is a global climate innovator. Through our strategic brands Trane® and Thermo King®, and our portfolio of environmentally responsible products and services, we bring efficient and sustainable climate solutions to buildings, homes and transportation. Visit tranetechnologies.com.
Forward-Looking Statements
This news release includes “forward-looking statements” within the meaning of securities laws, which are statements that are not historical facts, including statements that relate to the Company’s AI innovation initiatives and the anticipated benefits of the BrainBox AI Trane Technologies AI Lab and related technologies. These forward-looking statements are based on our current expectations and are subject to risks and uncertainties, which may cause actual results to differ materially from our current expectations. Factors that could cause such differences can be found in our Form 10-K for the year ended December 31, 2025, as well as our subsequent reports on Form 10-Q and other SEC filings. New risks and uncertainties arise from time to time, and it is impossible for us to predict these events or how they may affect the Company. We assume no obligation to update these forward-looking statements.
Media Contact:
Travis Bullard
+1-919-802-2593
Media@tranetechnologies.com
Investors Contact:
Zachary Nagle
+1-704-990-3913
InvestorRelations@tranetechnologies.com
BrainBox AI Contact:
Liz Culley-Sullo
+1-514-781-3316
l.culley-sullo@brainboxai.com
