Fault detection and diagnostic (FDD) methods are receiving increasing consideration for application in space-conditioning equipment as a method to reduce energy consumption and refrigerant emissions, and to provide more reliable comfort. It is anticipated that utility rebate programs and building energy regulations will promote the use of FDD methods as cost-effective energy efficiency measures leading to increased market acceptance. This project will accelerate market penetration of air conditioner and heat pump FDD technology by developing effective FDD algorithms, developing open-source performance measurement tools, and by formulating a standard procedure for rating different commercial FDD products based on their potential to avoid performance degradation and increased energy consumption.
Objective - Develop fault detection and diagnostic (FDD) methods and measurement tools to ensure air conditioners (ACs) and heat pumps (HPs) perform as designed throughout their lifetime and develop a testing and rating methodology to assess relative merits of different commercial FDD products.
What is the new technical idea? This project will advance measurement science and facilitate implementation of residential FDD methods by pursuing two research tasks. The first task will develop new adaptable FDD algorithms to detect AC and HP faulty operation. Adaptable FDD methods are required to accommodate different equipment installations and aging effects. The datasets generated for fault-free and faulty operation of residential ACs are very well suited for processing by machine learning algorithms. A large base of open-source machine-learning codes is available on the world wide web; one source is the GitHub repository system. The combination of the NIST-developed FDD datalogger’s stream of data with cloud computing resources that utilize machine-learning algorithms will be investigated to develop and demonstrate cloud-based FDD methods for AC equipment. Generalization of these techniques and application of the statistical methods engrained within the FDD algorithms will give U.S. industry opportunities for innovation and will promote faster introduction of this technology into the marketplace. These FDD methods will not only be applicable to residential heat pumps, but also to other systems that operate on the vapor-compression principle.
FDD devices, like any other products marketed based on functionality, must have their performance metrics determined to promote market competition and product improvement. Within the second research task of this project, two performance metrics will be developed for FDD modules: the first, “energetic” metric will capture potential energy savings resulting from detecting/correcting various faults, and the second metric will capture the FDD device’s ability to avoid false alarms.
The main products of the project will be a database of fault-free and faulty performance and an AC/HP FDD Tester/Evaluator. During a test of a commercial FDD method (e.g., an algorithm embedded in a heat pump’s control unit or in a third-party handheld FDD device), the AC/HP FDD Tester/Evaluator will output a set of heat pump parameters that potentially could be used by the FDD module and, based on these parameters, the module will make a diagnosis regarding the “health” of the heat pump system. Calculation of the “energetic” metric will include weather data to weight the robustness of the FDD device’s diagnostic capabilities over an entire year.
What is the research plan? Two complementary tasks have been defined to achieve the goals of this project. The goal of Task 1 is to develop and refine an adaptable (self-training) FDD algorithm suitable for deployment with field-assembled systems. Past efforts devised a novel self-training methodology for developing no-fault performance models of a split heat pump operating in the cooling mode. This methodology was devised using steady-state laboratory measurements taken in previous years and then it was refined using additional laboratory data that included cycling. However, its validation required “real world” data that included system cycling and noisy transients. The first stage of this validation used operating data from the heat pump installed in the NIST net-zero energy residential test facility (NZERTF). The second stage of validation needs data collected from a human-occupied house. With this goal in mind, we formulated an SBIR project to develop a prototype custom data collection logger for remotely monitoring residential split AC and HP systems. This SBIR project produced the hardware necessary to remotely monitor a residential split-system, led to a greater understanding of problems associated with remote monitoring, and positioned us to explore and apply the capabilities of the Internet of Things (IoT) technology to measure, control, and log with little expense. Using this SBIR project as a basis, we have developed a new datalogger that addresses some of the shortcomings of the first generation datalogger. Specifically, the new logger is cheaper, easier to program, uses a full version of Python, has larger memory/storage space, and a more readily available offering of open-source code. We will refine the new datalogger and use it to conduct measurements on field-installed, human-occupied systems to verify the performance of our self-training FDD protocol. We will also explore possibilities of applying artificial-intelligence-based methods for advancing the robustness of the current FDD protocol.
The goal of Task 2 is to develop a test method for rating AC and HP FDD protocols. During an earlier phase of the project, a beta version of a Tester/Evaluator was developed with an embedded Inverse AC Model, which was based on a specific AC unit. The function of the model is to generate no-fault and faulty scenarios for testing/rating FDD methodologies. However, subsequent tests and study demonstrated that different ACs can respond differently and suffer a different degree of performance degradation for identical faults, which increases the complexity of the problem at hand. With this finding, significant elements of this Task 2 are: (1) development of an understanding of how different system types and configurations respond to common faults; (2) development of “game-proof” fault testing scenarios; and (3) development of performance metrics for FDD protocols. In support of this task, we performed an extensive laboratory test program of low-efficiency and high-efficiency ACs to study in more detail their responses to single and multiple simultaneous faults. The test program involved measurements of cooling/heating total capacity, latent capacity, electrical power demand, efficiency, refrigerant temperatures/pressures at key system locations, and included the assessment of fault intensity. The goal is to evaluate the Inverse AC Model’s capability to reproduce the fault-driven responses observed in laboratory testing, and to modify/upgrade the model to obtain the reproducibility, if practical, or devise other concepts for testing FDD methods.
ASHRAE Standard Project Committee 207P “Laboratory Method of Test of Fault Detection and Diagnostics Applied to Commercial Air-Cooled Packaged Systems” was formed in January of 2012. A draft method of testing air-side economizers in packaged rooftop units was released in 2018. The final version should be approved and released in 2019. This committee will become a Standing Standard Project Committee to develop methods of test for even more system faults. This committee has directly used concepts and software developed in Task 1 and 2 to formulate test methods for different types of faults. NIST personnel have actively supported the work of this committee by chairing a sub-committee on airflow faults and contributing to the development of standardized definitions of faults. NIST personnel will continue to support the development of this standard by working on the committee and supplying actual test data where applicable.