The goal of this project was to verify the feasibility of employing a model-based approach to central plant operation and diagnostics at U.S. Department of Defense (DoD) facilities, and to quantify the associated benefits. This field demonstration was designed to validate: effectiveness in reducing electricity consumption and associated greenhouse gas (GHG) emissions; user satisfaction; cost-effectiveness and viability of system economics; and validity of model calibration.

Technology Description

In this demonstration, Lawrence Berkeley National Laboratory (LBNL) developed a hybrid data-driven and physics model-based tool for energy efficiency in central cooling plants. Once developed, the PlantInsight technology was implemented at the U.S. Naval Academy (USNA), and the technology performance objectives were evaluated. PlantInsight provides detection and diagnosis of three types of faults: fan cycling, chiller cycling, and poor chiller efficiency. It also provides analysis of optimal condenser water setpoint temperatures to minimize plant energy consumption. A calibrated simulation model is used in the algorithms to identify poor chiller efficiency and optimal condenser water temperature, while the cycling faults are identified using purely data-driven models. In addition, the tool offers visualization for operators to track key parameters such as cooling plant load and chilled water loop temperature. Figure ES-1 contains a diagram of the Modelica model used to conduct the cooling optimization and the architecture of the PlantInsight tool, as implemented for the demonstration. Figure ES-2 contains screen shots of the user interface.

Demonstration Results

Model calibration: To ensure that the models developed to simulate the central plant were representative of the central plant’s actual physical performance, the chiller and tower models were calibrated to measured data from the site. The calibration goal targeted a difference between model-predicted and measured parameters of less than 10% for 90% of data points. This was achieved for ten of ten tower cells for which data were available, and for three of six chillers. The soundness of the calibration process was confirmed; however, calibration was challenged by the limited volume of data representing full-capacity chiller operation, and perhaps by inaccurate water temperature sensor data, or faulty operations underlying the data.

User satisfaction: The demonstration technology was evaluated to determine whether PlantInsight offered equal or improved satisfaction relative to existing operational tools. Survey responses indicated that satisfaction with the capabilities of PlantInsight was equal to or better than that with the preexisting Johnson Controls International (JCI) Metasys system that is used for plant operations.

Energy and GHG emissions savings: The demonstration targeted 10% annual reductions in electricity consumption and associated GHG emissions at the central cooling plant. The results of the savings analysis indicated that daily energy savings greater than 10% are obtainable for approximately six months of the year, mainly during the winter season. However, for the year as a whole, energy savings of approximately 1.5% are obtainable. Since savings were driven by wet bulb temperature (lower), which occur in winter, when total plant consumption is lowest, larger annual savings are possible in drier climates. GHG emissions were quantified using a conversion factor based on references published by the U.S. Environmental Protection Agency. Since the conversion factor was represented as a single constant for the region, the emissions reduction results are the same as those for energy, in terms of percent savings.

System economics: Assessment of system economics based on standard capital budgeting metrics provides a gauge for determining financial feasibility of the demonstration technology. The analysis showed that simple and discounted payback can be met in 1.4 years, well within the five-year target that was established.

Implementation Issues

Future implementation of the technology will require attention to three areas: information technology (IT) security, maintenance and evolution, and scale-up and transition.

IT security: The PlantInsight technology requires unidirectional transfer of cooling plant operational data from the site to the application’s database. The application is hosted on a web server. In the USNA demonstration, port 443 was used for secure communication from the building automation system (BAS) kiosk to PlantInsight. To satisfy DoD IT security requirements, future installations can consider several options. PlantInsight can be integrated within existing accredited applications or could be put through the accreditation process itself. Another option that was explored was to push plant operational data from the installation to a server farm on a secure DoD network, with PlantInsight accessing the data through a virtual private network (VPN) application.

Technology maintenance and evolution: As the demonstration comes to a conclusion, LBNL will work with UNSA IT and Naval District Washington (NDW) to identify options to transfer the tool to a server and location that will comply with security requirements. Although it is not yet used universally throughout the industry, companies such as HOK, JCI, and United Technologies Corporation (UTC) have staff that are familiar with the modeling language (Modelica) upon which the tool is built. They could potentially be contracted to support future model modification and calibration.

Technology scale-up and transition: To make the PlantInsight Tool available to other DoD installations, it will be released through an open source software license. This will enable stand-alone use according to its current design, or adaptation for use within existing installation energy management facility and information systems as described in the considerations of IT security. Several types of documentation have been developed to support these future transition activities, and to support ongoing use at USNA.


Future implementations of the technology will benefit from awareness of the following higher-level lessons that were learned throughout the course of the demonstration. First, operators place strong value on access to tools that provide visibility into how controls impact energy use and cost. This is not as a rule available in today’s commercial analytics technologies that span BASs, meter analytics tools, or equipment-specific FDD tools. As such, heating, ventilation and air conditioning (HVAC) optimization technologies represent advances in the state of today’s available technology, and this is even more true of optimization tools that incorporate physics-based modeling approaches. Environmental Security Technology Certification Program (ESTCP) and USNA have acted as a leader in the demonstration of these cutting-edge solutions, and future implementations will continue to contribute to the state of knowledge of their development and application.

Model-predictive optimization combined with FDD is recognized as a critical aspect of realizing the dynamic low-energy buildings of tomorrow; it can deliver even more impact by expanding the set of parameters included in the optimization, and the number of end-uses that are considered. Although these technologies represent advanced applications, the external infrastructure to support their delivery is mature; cloud hosting and computational scalability are well supported through modern IT solutions. The most significant practical implementation barrier is brittle building data acquisition and communication systems that present chronic challenges to applications that interface with controls data. Finally, we note that the creation and calibration of physics-based models for use in the operational phase of the building life-cycle is highly dependent upon the specific algorithms with which they will be paired. The open, reference implementations that are delivered with PlantInsight are important contributions to the industry’s continued success in leveraging these promising approaches for next-generation building energy efficiency.

  • Chiller ,

  • HVAC Controls ,

  • Optimization ,

  • Fault Detection and Diagnostics (FDD) ,

  • Energy Modeling