Optimizing Energy Consumption of HVAC Systems in Commercial and Industrial Settings is vital

The United States has implemented various policies to enhance energy efficiency and reduce greenhouse gas emissions, including the Energy Policy Act and the Energy Star program. The US Department of Energy (DOE) sets standards and supports research to advance HVAC efficiency. With diverse climate zones requiring versatile solutions, optimizing HVAC systems is crucial for reducing energy consumption. Integrating smart technologies and renewable energy sources like solar and wind power helps the US move towards its sustainability goals.

AI and IoT are transforming HVAC systems by enabling energy optimization through data analysis and real-time adjustments. Key contributions include::

• Dynamic Control Systems: AI and IoT sensors allow HVAC systems to adapt to real-time conditions like occupancy and weather, ensuring optimal performance.
• Smart Monitoring: Continuous monitoring by IoT devices detects inefficiencies and enables timely interventions.
• Energy Consumption Forecasting: AI models predict future energy needs based on historical data, improving planning and reducing wastage.
• Integration with Building Management Systems: AI and IoT integrate HVAC with building management systems, enhancing overall energy efficiency.

GFT Spain’s AI team set to discover the potential of implementing AI-based control for the case of an HVAC system in a commercial building. The object of study is one section of 1160m2 within an office building.  This section is divided in eight different thermal zones and is serviced by one single RTU reheat system with the following characteristics:

• Design Airflow: 33,980 m³/hr
• Cooling Capacity: 356 MBH (30 tons or 105.5 kW)
• Heat Pump: 117 kW (400 MBH) (nominal) heat pump (initially air-source, later replaced with water-source)

The temperature in each thermal zone is fine-tuned by Fan-powered Thermal Units:

• Hydronic Heating Coils: 117 kW or 400 MBH nominal capacity
• Air Distribution (Diffusers) for fine-tuned cooling

The HVAC system provides information about the following variables of interest.

Additionally, data about the outside weather, and occupancy of the building is accessible.

Our methodology consists of integrating a simulation model, which has the capacity of dynamically identifying the correlation between operational variables as conditions dynamically change, with an optimization model, which seeks the optimal control parameters that maintain a target temperature at a minimal energy consumption rate.  The optimization model is conditioned by a cost function that prevents the system from deviating away from the desired target temperature.  We trained these models over a combination of historic and generated dataset resulting in an algorithm which is calibrated to offer optimal HVAC settings over new, unseen conditions.

In use, our solution takes as input parameters the target temperature for each thermal zone, the instantaneous outside weather, and past HVAC operational parameters.  With this, it produces a set of optimal HVAC parameters, including  “AI RTU Supply Air Setpoint” or “AI Thermal Zone Fan Speed” and others.  These optimized parameters are applied to the HVAC system, which results on the desired indoor temperature, at a minimized energy consumption.

Our solution was evaluated against data from July 1st to December 6th of 2023, for the building described before.  During this time, the non-optimized HVAC generated 353 temperature violations, with an average violation of 2.2 F out of range (violations are periods when temperature deviates outside the minimum and maximum set in the system).  The energy consumption for the same period was of 54MW with a mean instantaneous consumption of 3.5 KWh.

Our solution determined that an AI control is capable of reducing the number of violations from 353 to 72 (5X improvement) and a reduction of the average deviation magnate per violation from 2.2°F a 0.45°F (5X).  

The consumed energy can be reduced from 54MW to 46 MW (15% improvement)

The consequences derived from our work are far reaching. 

We demonstrated how to use AI to drastically reduce the energy consumption of HVAC system, which saves money from ongoing operational expenses.  With this we also avoid the release of up to 1 ton of carbon to the atmosphere per MW of energy that is not consumed by an HVAC system.

We also demonstrated that beyond saving economic savings, the AI-based control of HVAC systems can reduce the number of temperature violations, making HVAC systems more adequate to human comfort and productivity.

Finally, our approach can be implemented as a traditional closed-loop implementation, which means that virtually any HVAC systems currently operating can become smarter, and more efficient.

Optimizing the energy consumption of HVAC systems in commercial and industrial settings is not just an operational necessity but a critical component of global sustainability efforts. AI and IoT play a pivotal role in this optimization process, providing proven solutions that ensure HVAC systems are both energy-efficient and cost-effective.

Interested in learning how GFT can change the energy consumption pattern of your organization?  Let’s talk.