Reducing Energy Consumption in Water-Cooled Chiller Plants: How Machine Learning Can Help Energy Managers

As an energy manager of a commercial building, your primary responsibility is to ensure that the building operates in the most energy-efficient way possible. HVAC systems, which include the condenser loop of a water-cooled chiller plant, are among the largest energy consumers in commercial buildings. In fact, in 2019, cooling accounted for 27% of the total energy consumption in commercial buildings in the United States, with HVAC systems responsible for 40% of that usage. However, by leveraging machine learning techniques such as Gaussian processes and deep learning, you can optimize the operation of a condenser loop and other HVAC system components, leading to significant energy and cost savings for the building.

Gaussian processes are a type of machine learning algorithm that can be used to model the behavior of complex systems. In the context of a water-cooled chiller plant, a Gaussian process can be trained to predict the optimal setpoints for the condenser loop based on a variety of factors such as outdoor temperature, humidity, and the cooling load of the building. By continually updating the model with real-time data, the Gaussian process can adapt to changing conditions and identify opportunities to reduce energy consumption while maintaining optimal performance.

Another machine learning technique that can be applied to the optimization of HVAC systems is deep learning. Deep learning algorithms are capable of processing large amounts of data and identifying complex patterns, making them well-suited for applications such as energy management. For example, a deep learning algorithm can be trained to analyze data from the chiller plant and identify patterns or anomalies that indicate opportunities for optimization. The algorithm can also be used to predict future energy consumption and identify areas for improvement.

By incorporating machine learning into the operation of a water-cooled chiller plant, energy managers can achieve significant energy and cost savings. In one case study, a commercial building in New York City was able to reduce its energy consumption by 10% by implementing a machine learning-based optimization system for its HVAC system. Over the course of a year, this led to a cost savings of $57,000.

In addition to the immediate cost savings, the optimization of HVAC systems using machine learning can also have long-term benefits for commercial buildings. By reducing energy consumption and associated costs, building owners can improve the bottom line and invest savings in priorities like deferred maintenance. Additionally, the implementation of sustainable practices such as energy-efficient HVAC systems can help to enhance the building's reputation and even to comply with new regulations, such as New York’s Local Law 97.

When considering the implementation of machine learning-based optimization systems for HVAC systems, it's important to work with experienced professionals who can ensure that the system is properly designed and integrated into the building's existing infrastructure. In addition, ongoing monitoring and maintenance are critical to ensure that the system continues to perform at optimal levels over time.

In conclusion, the optimization of HVAC systems using machine learning techniques such as Gaussian processes and deep learning represents a significant opportunity for energy managers of commercial buildings to reduce energy consumption and associated costs. By implementing a machine learning-based optimization system for the condenser loop of a water-cooled chiller plant, energy managers can achieve significant energy and cost savings while also contributing to a more sustainable future for the building and the surrounding community.

Published 
April 4, 2023
Get more value out of your data with Manifest
Whether you’re looking to reduce costs, increase efficiency, or improve performance, Manifest can help. Book a demo with us to see how Manifest can help you achieve the business outcomes you need.