Fault diagnosis method and device for commercial building heating ventilation air conditioning system and medium

Abstract

The invention discloses a fault diagnosis method and device for a commercial building heating ventilation air-conditioning system, and a medium, belongs to the technical field of intelligent building operation and maintenance, and is used for solving the limitations in the aspects of data adaptability, generalization ability, dynamic response capture and automatic diagnosis in the fault diagnosis of the existing heating ventilation air-conditioning system. And the intelligent operation and maintenance requirements of modern buildings are difficult to meet. The method comprises the following steps: carrying out hierarchical relationship definition between related equipment entities on basic data in the heating ventilation air-conditioning system to obtain an equipment hierarchical semantic network; performing data simulation processing on the running state of each equipment entity in the heating ventilation air-conditioning system to obtain a dynamic virtual model; constructing an unsupervised fault monitoring model; real-time dynamic fault detection and fault triggering judgment are conducted on the real-time operation data of the heating ventilation air conditioning system, and current fault data are determined; and carrying out hierarchical positioning processing of a fault source under related reconstruction errors on the current fault data, and generating a fault diagnosis report.

Description

Technical Field

[0001] This application relates to the field of intelligent building operation and maintenance technology, and in particular to a fault diagnosis method, equipment and medium for a commercial building heating, ventilation and air conditioning system. Background Technology

[0002] Heating, ventilation, and air conditioning (HVAC) systems are central to building energy consumption and environmental comfort control, and their stable operation directly impacts building energy efficiency, occupant health, and operating costs. However, their complex structure and highly dynamic operating environment, influenced by factors such as season and load, make fault detection and diagnosis (FDD) a key challenge in operation and maintenance. This is especially true for light commercial buildings (typically referring to buildings with fewer than six floors and a floor area of ​​less than 2,500 square feet, such as small offices and clinics), where the potential for energy savings is enormous. Studies have shown that effective FDD can reduce energy consumption by 20-30%.

[0003] Existing fault diagnosis technologies have several limitations, making it difficult to meet the actual needs of building operation and maintenance. First, they are highly dependent on and of poor quality. Traditional fault diagnosis methods rely on labeled datasets, but actual HVAC operation data is mostly unlabeled and raw, containing a large number of missing values, noise, inconsistencies, and time series breaks, making it difficult to directly use for model training. Furthermore, manual cleaning and labeling are time-consuming and labor-intensive, typically limiting them to faults occurring within a specific data collection period. Second, they lack generalization ability. Significant differences in the design, function, and equipment configuration of different buildings lead to varying HVAC system characteristics. Existing methods are often customized for specific buildings, making cross-configuration migration difficult. Digital twin technology also suffers from difficulty in model reuse due to its high dependence on specific building data models and parameters. Third, they lack dynamic system response. HVAC operating states are dynamically changing, and parameter fluctuations are time-correlated. Traditional algorithms such as Principal Component Analysis (PCA) are mainly trained on static or steady-state data, making it difficult to effectively capture transient processes and prone to false alarms or missed alarms. Existing methods often rely on statistical methods to identify "steady-state" periods for training, but real systems fluctuate continuously, making true steady-state data difficult to achieve, resulting in the loss of dynamic information. Fourth, the diagnostic information is vague, making it difficult to achieve automated root cause localization. Existing digital twin systems mostly rely on predefined expert rules or human experience, which can often only detect the existence of faults but cannot accurately locate the source of the fault or assess its severity, thus delaying the timing of handling. PCA-based methods can usually only report abnormal sensors and lack an understanding of the system hierarchy (such as the central system and regional equipment), making it difficult for operators to trace the root cause of faults in multi-level systems.

[0004] In summary, the limitations of existing technologies in terms of data adaptability (handling unlabeled, incomplete, and noisy data), generalization ability (cross-building configuration migration), dynamic response capture (adapting to the time-varying characteristics of the system), and automated diagnosis (accurately locating the root cause and assessing the severity) make it difficult to meet the needs of intelligent operation and maintenance of modern buildings. Summary of the Invention

[0005] This application provides a fault diagnosis method, equipment, and medium for HVAC systems in commercial buildings to solve the following technical problems: In the fault diagnosis of existing HVAC systems, there are limitations in data adaptability, generalization ability, dynamic response capture, and automated diagnosis, which makes it difficult to meet the needs of intelligent operation and maintenance of modern buildings.

[0006] The embodiments of this application adopt the following technical solutions:

[0007] On one hand, this application provides a fault diagnosis method for a commercial building HVAC system, comprising: defining hierarchical relationships between equipment entities in the basic data of the HVAC system to obtain an equipment hierarchical semantic network; performing data simulation processing on the operating status of each equipment entity in the HVAC system to obtain a dynamic virtual model; combining the equipment hierarchical semantic network and the dynamic virtual model to obtain a digital twin ontology framework based on the HVAC system; constructing an unsupervised fault monitoring model based on the preprocessed historical data steady-state training set of the HVAC system; performing real-time dynamic fault detection and fault trigger judgment on the real-time operating data of the HVAC system through the unsupervised fault monitoring model to determine the current fault data; and performing hierarchical localization processing on the current fault data under reconstruction error through the digital twin ontology framework to generate a fault diagnosis report.

[0008] This application addresses core issues in fault diagnosis within light commercial building scenarios, such as high data dependency, weak generalization ability, and lack of dynamic response, by integrating a digital twin ontology framework with an improved unsupervised learning algorithm. This method requires no labeled data, adapts to different building configurations, and achieves fully automated diagnosis from anomaly detection and root cause localization to severity assessment. Compared to traditional PCA methods, this application significantly reduces false alarm rates and greatly improves fault location accuracy.

[0009] In one feasible implementation, the basic data of the HVAC system is used to define the hierarchical relationships between relevant equipment entities to obtain a device-level semantic network. Specifically, this includes: collecting real-time sensor data, equipment specification data, and operation and maintenance history data from the HVAC system; inputting the real-time sensor data, equipment specification data, and operation and maintenance history data to obtain the basic data; and using RDF triples constructed by the semantic modeling standard, and based on the basic data, defining the hierarchical relationships between equipment entities to generate the device-level semantic network with physical topology relationships.

[0010] In one feasible implementation, data simulation processing is performed on the operating status of each device entity in the HVAC system to obtain a dynamic virtual model. Specifically, this includes: generating real-time virtual sensors related to the operating status of each device entity in the HVAC system based on a multilayer perceptron; wherein the real-time virtual sensors can simulate the operating status of physical entities in real time; performing data simulation processing on the operating status of each device entity in the HVAC system using the virtual sensors, and generating the dynamic virtual model based on the parameter change status of the physical devices and virtual in-situ modeling technology used for the HVAC system; wherein the dynamic virtual model is used for a one-to-one mapping with each device entity in the HVAC system.

[0011] In one feasible implementation, an unsupervised fault monitoring model is constructed based on the preprocessed historical steady-state training set of the HVAC system. Specifically, this includes: collecting historical operating datasets from the HVAC system; calculating the distribution of daily minimum values ​​in the historical operating dataset based on the reconstructed values ​​and reconstruction errors from principal component analysis, and obtaining a data cleaning threshold based on the system operating characteristics and skewness of the HVAC system; removing and converging outliers between the historical operating dataset and the data cleaning threshold to generate the historical data steady-state training set; and constructing the unsupervised fault monitoring model based on the historical data steady-state training set and the corresponding data validation set.

[0012] In one feasible implementation, the unsupervised fault monitoring model is used to perform real-time dynamic fault detection on the real-time operating data of the HVAC system. This includes: using a sliding window in the unsupervised fault monitoring model to calculate the average and standard deviation of the reconstruction error of the real-time operating data under time-series conditions, obtaining the average and standard deviation of the equipment operating data based on the equipment operating parameters in the HVAC system; calculating a dynamic fault threshold between the average and standard deviation of the equipment operating data based on a preset adjustable multiplier and the moving variance of the reconstruction error, obtaining an upper fault threshold and a lower fault threshold; wherein the upper fault threshold and the lower fault threshold represent the dynamic fault detection range for the operating data of each piece of equipment in the HVAC system.

[0013] In one feasible implementation, the unsupervised fault monitoring model is used to determine the fault trigger of the real-time operating data of the HVAC system and identify the current fault data. This includes: determining a baseline threshold based on a preset ideal adjustable multiplier and the daily minimum average of the reconstruction error in the real-time operating data; if the average value of the equipment operating data in the current real-time operating data is greater than the baseline threshold, then the fault trigger of the real-time operating data is determined as a Level 1 fault warning; wherein, the Level 1 fault warning is a minor fault warning; if the average value of the equipment operating data in the current real-time operating data is greater than the baseline threshold, and the equipment operating parameters in the current real-time operating data exceed the upper or lower fault threshold, then the fault trigger of the real-time operating data is determined as a Level 2 fault warning; wherein, the Level 2 fault warning is a severe fault warning; and the current real-time operating data that meets the fault dynamic detection range, the Level 1 fault warning, and the Level 2 fault warning is determined as the current fault data.

[0014] In one feasible implementation, the digital twin ontology framework is used to perform hierarchical location processing of the fault source under reconstruction error in the current fault data, generating a fault diagnosis report. Specifically, this includes: calculating the deviation between the reconstruction error in the normal state and the fault state of the current fault data to obtain deviation result data; if the deviation result data is detected, then, based on the corresponding hierarchical relationships of the system levels in the digital twin ontology framework, identifying and judging the central-level and regional-level systems of the HVAC system in the current fault data to obtain equipment-level fault results; wherein, the equipment-level fault results include: regional-level equipment fault results and central-level equipment fault results; based on the equipment-level fault results, and through the equipment mapping relationships and physical sensor data in the digital twin ontology framework, the fault diagnosis report is generated; wherein, the fault diagnosis report includes: the specific location of the fault, the scope of the fault's impact, and the fault operation and maintenance strategy.

[0015] In one feasible implementation, the current fault data is used to identify and judge the central-level system and regional-level equipment of the HVAC system to obtain equipment-level results. Specifically, this includes: extracting key fault variables from the current fault data; if the key fault variable belongs to the reconstruction error anomaly state of regional-level equipment, then through the corresponding subordinate relationship of the system level in the digital twin ontology framework, the key fault variable is analyzed and located for equipment topology relationships to obtain the regional-level equipment fault result; if the key fault variable belongs to the reconstruction error anomaly state of central-level equipment, and the reconstruction error in the same time period exceeds the fault threshold, then the key fault variable is used to identify the fault anomaly propagation of regional-level equipment under the central-level equipment to obtain the central-level equipment fault result.

[0016] Secondly, embodiments of this application also provide a fault diagnosis device for a commercial building HVAC system, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to execute a fault diagnosis method for a commercial building HVAC system as described in any of the above embodiments.

[0017] Thirdly, embodiments of this application also provide a non-volatile computer storage medium, which is a non-volatile computer-readable storage medium storing at least one program, each program including instructions, which, when executed by a terminal, cause the terminal to perform a fault diagnosis method for a commercial building HVAC system as described in any of the above embodiments.

[0018] This application provides a fault diagnosis method, equipment, and medium for HVAC systems in commercial buildings. Compared with the prior art, the embodiments of this application have the following beneficial technical effects:

[0019] 1. System Transparency: By constructing a digital twin ontology framework, real-time monitoring and visualization of the HVAC system are achieved, improving the system's transparency and manageability.

[0020] 2. Real-time fault detection: Real-time dynamic fault detection using unsupervised fault monitoring models can quickly respond to potential system problems and reduce the time and impact of fault occurrence.

[0021] 3. Improve diagnostic accuracy: By combining device-level semantic networks and dynamic virtual models, the source of faults can be located more accurately, thus improving the accuracy of fault diagnosis.

[0022] 4. Reduce downtime: Quick and accurate fault diagnosis and location can reduce system downtime and lower operating costs caused by faults.

[0023] 5. Optimize maintenance plans: Fault monitoring models built on historical data can help predict future faults, thereby optimizing maintenance plans and reducing the frequency of preventative maintenance.

[0024] 6. Enhanced predictive maintenance: By analyzing historical data, the system can predict potential failures, enabling predictive maintenance and reducing the probability of failures.

[0025] 7. Resource utilization efficiency: Through real-time monitoring and fault diagnosis, resources can be utilized more effectively, avoiding unnecessary energy consumption and maintenance costs.

[0026] 8. Simplified operation process: Automated fault diagnosis report generation simplifies the operation process, reduces manual intervention, and improves work efficiency.

[0027] 9. Enhanced security: Timely detection and handling of potential security vulnerabilities to improve system stability and security. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0029] Figure 1 A flowchart of a fault diagnosis method for a commercial building HVAC system is provided as an embodiment of this application;

[0030] Figure 2 This is a structural schematic diagram of a fault diagnosis device for a commercial building HVAC system provided in an embodiment of this application. Detailed Implementation

[0031] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0032] This application provides a fault diagnosis method for HVAC systems in commercial buildings, such as... Figure 1 As shown, the fault diagnosis method for HVAC systems in commercial buildings specifically includes steps S101-S105:

[0033] S101. Define the hierarchical relationship between relevant equipment entities in the basic data of the HVAC system to obtain the equipment hierarchical semantic network.

[0034] Specifically, it is necessary to first collect real-time sensor data, equipment specification data, and historical operation and maintenance data from the HVAC system. Then, this real-time sensor data, equipment specification data, and historical operation and maintenance data are entered into the system to obtain the basic data.

[0035] Furthermore, using RDF triples constructed through semantic modeling standards and based on basic data, hierarchical relationships between device entities are defined to generate a device-level semantic network with physical topology relationships.

[0036] In one embodiment, real-time sensor data (temperature, pressure, valve opening, etc.) is transmitted to the system via the Internet of Things (IoT). Equipment specifications (such as sensor accuracy, pump rated power) and maintenance history (such as filter replacement time) are entered by the DT administrator. Then, based on RDF triples constructed using semantic modeling standards (such as Brick Schema), the hierarchical relationships between equipment entities are defined, forming a machine-readable semantic network. This ontology empowers the AI ​​agent to understand the physical topology, for example, by resolving the hierarchical relationship between "central chiller unit" and "regional VAV terminal" through SPARQL queries, providing a semantic reasoning basis for subsequent fault localization.

[0037] S102. Perform data simulation processing on the operating status of each equipment entity in the HVAC system to obtain a dynamic virtual model.

[0038] Specifically, a multilayer perceptron is needed to generate real-time virtual sensors that reflect the operating status of various equipment entities within the HVAC system. These real-time virtual sensors can mimic the operating status of physical entities in real time.

[0039] Furthermore, virtual sensors are used to simulate the operating status of various equipment entities in the HVAC system. Based on the parameter changes of the physical equipment and virtual in-situ modeling technology used for HVAC systems, a dynamic virtual model is generated. This dynamic virtual model is then used to map one-to-one with each equipment entity in the HVAC system.

[0040] In one embodiment, a dynamic virtual model also needs to be deployed. This can be generated first using Virtual In-Situ Modeling (VIM) technology, specifically by using a Multilayer Perceptron (MLP) to generate real-time virtual sensors (such as a cold water return temperature prediction model), which can mimic the operating state of physical entities in real time (e.g., predicting return water temperature). It corresponds one-to-one with the physical entities, and its accuracy is measured using RMSE (Real-Time Error) and R² (Ratio of Fit), ensuring close resemblance to reality. Simultaneously, when physical equipment parameters change (e.g., water pump replacement), the AI ​​agent will automatically trigger an ontology update (via OFG tools) and retrain the associated virtual model using VIM tools. Finally, a dynamic virtual model is generated to achieve a one-to-one mapping with each equipment entity in the HVAC system. Then, the equipment-level semantic network and the dynamic virtual model are combined to obtain a digital twin ontology framework based on the HVAC system.

[0041] S103. Based on the steady-state training set of historical data of the preprocessed HVAC system, construct an unsupervised fault monitoring model.

[0042] Specifically, the first step is to collect historical operational datasets from the HVAC system.

[0043] Furthermore, based on the reconstructed values ​​and reconstruction errors from principal component analysis, the distribution of daily minimum values ​​in the historical operating dataset is calculated, and the data cleaning threshold is obtained based on the system operating characteristics and skewness of the HVAC system.

[0044] Furthermore, it is necessary to remove and converge outlier data between the historical dataset and the data cleaning threshold to generate a steady-state training set of historical data.

[0045] In one embodiment, to address the incompleteness and noise of the original historical data, an iterative cleaning process is designed and implemented: A training subset is automatically selected based on the daily minimum distribution of the reconstruction error (SPE) of PCA (Principal Component Analysis), and the process is iterated using equation (1-2).

[0046]

[0047] in, The original data vector; The PCA reconstruction value represents the reconstruction error.

[0048] Threshold represents the threshold value; This represents the minimum daily average for all normal days. The minimum standard deviation for each day of all normal dates; m can be determined based on the dynamic characteristics of the building's HVAC system operation and The degree of skewness is adjusted, with m=3 by default.

[0049] On a certain day The abnormal daily data exceeding the threshold are then removed cyclically until convergence to a steady-state dataset. Finally, the number of principal components is selected to ensure that the cumulative variance (CV) is ≥92%, preserving the core dynamic features of the system, and generating a steady-state training set of historical data.

[0050] Furthermore, an unsupervised fault monitoring model is constructed based on the historical steady-state training set and the corresponding data validation set.

[0051] S104. Using an unsupervised fault monitoring model, perform real-time dynamic fault detection and fault trigger judgment on the real-time operating data of the HVAC system to determine the current fault data.

[0052] Specifically, by using the sliding window in the unsupervised fault monitoring model, the average value and standard deviation of the reconstruction error of the real-time operating data under the time series are calculated to obtain the average value and standard deviation of the equipment operating data based on the equipment operating parameters in the HVAC system.

[0053] Furthermore, by combining the preset adjustable multiplier and the moving variance of the reconstruction error, a dynamic fault threshold is calculated between the average value and the standard deviation of the equipment operating data to obtain the upper and lower fault thresholds. These upper and lower fault thresholds represent the dynamic fault detection range for the operating data of each device in the HVAC system.

[0054] In one embodiment, the dynamic characteristics of HVAC (Heating, Ventilation and Air Conditioning) systems vary greatly among different buildings. For light commercial buildings, a fixed-size 15-day "sliding window" needs to be set, sliding sequentially over time to calculate the average SPE (SPEtma) and standard deviation (SPEtmd) of real-time data so that the model can adapt to the dynamic changes of the system. The dynamic threshold is calculated using the following formulas (3) to (5):

[0055]

[0056] Where UL is the upper limit of the fault threshold; LL is the lower limit of the fault threshold; ws represents the sliding window size, set to 15 days; m std It is an adjustable multiplier, set to 3; t is the current time, t≥15.

[0057] Furthermore, a baseline threshold is determined based on a preset ideal adjustable multiplier and the daily minimum average of reconstruction errors in the real-time operating data. If the average value of the current real-time operating data exceeds the baseline threshold, the fault triggering of the real-time operating data is classified as a Level 1 fault warning. A Level 1 fault warning is a minor fault warning. If the average value of the current real-time operating data exceeds the baseline threshold, and any of the current real-time operating parameters exceed either the upper or lower fault threshold, the fault triggering of the real-time operating data is classified as a Level 2 fault warning. A Level 2 fault warning is a severe fault warning.

[0058] In one embodiment, if only dynamic thresholds are used for fault warning, the current real-time operating data will miss a large number of minor faults that exceed UL and LL but have deviated from the normal mode. Therefore, it is necessary to determine the baseline threshold by applying formula (6):

[0059] Where, m thrsh It is relative to m std Ideally, it should be set at 10% to 20% of m. This application selects m. thrsh =0.5; SPE msd This refers to the moving variance of SPE.

[0060] The first-level fault warning is: when the SPE on a certain day... ma (Average equipment operating data) exceeds Threshold SPE If the baseline threshold is reached, a Level 1 fault warning is triggered, which indicates a minor fault. A Level 2 fault warning is triggered when the SPE on a certain day... ma (Average equipment operating data) exceeds Threshold SPE If the baseline threshold is exceeded and the dynamic boundary of UL (upper limit fault threshold) or LL (lower limit fault threshold) is exceeded, a level 2 fault warning, i.e., a serious fault, is triggered.

[0061] Furthermore, the current real-time operating data that meets the requirements of dynamic fault detection range, first-level fault warning, and second-level fault warning is determined as the current fault data.

[0062] S105. Using the digital twin ontology framework, perform hierarchical location processing of the fault source under the relevant reconstruction error in the current fault data, and generate a fault diagnosis report.

[0063] Specifically, the deviation of the reconstruction error between the normal state and the fault state is calculated on the current fault data to obtain the deviation result data.

[0064] Furthermore, if deviation data is detected, the system hierarchical relationships within the digital twin ontology framework are used to identify and determine the central and regional systems of the HVAC system, thereby obtaining equipment-level fault results. These equipment-level fault results include both regional-level and central-level equipment fault results.

[0065] As a feasible implementation method, identifying and judging the central-level system and regional-level equipment of the HVAC system based on the current fault data can include: first, extracting key fault variables from the current fault data. If the key fault variable belongs to the reconstruction error anomaly state of regional-level equipment, then through the corresponding subordinate relationship of the system level in the digital twin ontology framework, the key fault variable is analyzed and located for relevant equipment topology relationships to obtain the regional-level equipment fault result. If the key fault variable belongs to the reconstruction error anomaly state of the central-level equipment, and the reconstruction error in the same time period exceeds the fault threshold, then the key fault variable is identified for fault anomaly propagation to regional-level equipment under the central-level equipment to obtain the central-level equipment fault result.

[0066] In one embodiment, after a fault is detected, i.e., after the current fault data is identified, the mapping relationship between the dynamic virtual model and the physical system in the digital twin model is used to locate the fault source through reconstruction error analysis. The deviation of the reconstruction error between normal and fault states is calculated. When a deviation (deviation result data) is detected, the corresponding subordinate relationship at the system level in the digital twin ontology framework can be used to determine whether the fault occurs in the central-level system or the regional-level system of HVAC. If an abnormal reconstruction error is detected in a regional-level device (such as a VAV terminal), the AI ​​agent automatically executes a SPARQL query to resolve the device topology relationship. If the reconstruction error of the upstream central-level device (such as an AHU) returned by the query exceeds a threshold in the same period, it is determined to be a secondary anomaly caused by the propagation of a fault in the central system; otherwise, it is confirmed as a local device fault. Alternatively, for example, if the key variable belongs to a central device such as an AHU or chiller (such as "air supply pressure" or "total flow"), the fault is determined to be located in the central system (central-level device fault result). If the key variable is "room temperature" and the associated "VAV box" control signal is abnormal, then the fault is identified as a VAV actuator (regional equipment fault result); if only the room data is abnormal and the VAV control is normal, then the fault is identified as a room sensor fault (regional equipment fault result).

[0067] Furthermore, based on the device-level fault results, and through the device mapping relationships in the digital twin ontology framework and physical sensor data, a fault diagnosis report is generated. This report includes: the specific location of the fault, the scope of its impact, and the corresponding maintenance strategy.

[0068] As a feasible implementation, the simulation results of the virtual model and physical sensor data can be integrated based on the device mapping relationship in the digital twin ontology framework to generate a fault diagnosis report, including the fault location, impact range, and suggested maintenance strategies, and update the digital twin database to optimize subsequent diagnostic accuracy. Furthermore, if the virtual model fails during a fault (e.g., residual abnormalities), the VIM tool is triggered to retrain the model (using a new dataset), and the updated model file and accuracy metrics (e.g., RMSE) are synchronized to the database.

[0069] In addition, embodiments of this application also provide a fault diagnosis device for HVAC systems in commercial buildings, such as... Figure 2 As shown, the fault diagnosis equipment 200 for commercial building HVAC systems specifically includes:

[0070] At least one processor 201; and a memory 202 communicatively connected to the at least one processor 201; wherein the memory 202 stores instructions executable by the at least one processor 201 to enable the at least one processor 201 to execute:

[0071] Define the hierarchical relationships between relevant equipment entities in the basic data of the HVAC system to obtain the equipment hierarchical semantic network;

[0072] The operating status of each equipment entity in the HVAC system is simulated to obtain a dynamic virtual model; and the equipment-level semantic network and the dynamic virtual model are combined to obtain a digital twin ontology framework based on the HVAC system.

[0073] An unsupervised fault monitoring model is constructed based on the preprocessed historical data steady-state training set of the HVAC system.

[0074] By using an unsupervised fault monitoring model, real-time dynamic fault detection and fault trigger judgment are performed on the real-time operating data of the HVAC system to determine the current fault data.

[0075] Using a digital twin ontology framework, the current fault data is processed to locate the fault source under reconstruction error at a hierarchical level, generating a fault diagnosis report.

[0076] This application addresses core issues in fault diagnosis within light commercial building scenarios, such as high data dependency, weak generalization ability, and lack of dynamic response, by integrating a digital twin ontology framework with an improved unsupervised learning algorithm. This method requires no labeled data, adapts to different building configurations, and achieves fully automated diagnosis from anomaly detection and root cause localization to severity assessment. Compared to traditional PCA methods, this application significantly reduces the false alarm rate and greatly improves fault localization accuracy.

[0077] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0078] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0079] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0080] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0081] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0082] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0083] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0084] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0085] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0086] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0087] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of this specification.

Claims

1. A fault diagnosis method for a commercial building HVAC system, characterized in that, The method includes: Define the hierarchical relationships between relevant equipment entities in the basic data of the HVAC system to obtain the equipment hierarchical semantic network; The operating status of each device entity in the HVAC system is simulated to obtain a dynamic virtual model; and the device-level semantic network and the dynamic virtual model are combined to obtain a digital twin ontology framework based on the HVAC system. Based on the preprocessed historical steady-state training set of the HVAC system, an unsupervised fault monitoring model is constructed. The unsupervised fault monitoring model is used to perform real-time dynamic fault detection and fault trigger judgment on the real-time operating data of the HVAC system, and to determine the current fault data. Using the digital twin ontology framework, the current fault data is processed to locate the fault source under reconstruction error hierarchically, and a fault diagnosis report is generated.

2. The fault diagnosis method for a commercial building HVAC system according to claim 1, characterized in that, The hierarchical relationships between equipment entities in the basic data of the HVAC system are defined to obtain the equipment hierarchical semantic network, which specifically includes: Collect real-time sensor data, equipment specification data, and operation and maintenance history data from the HVAC system; and input the real-time sensor data, equipment specification data, and operation and maintenance history data to obtain the basic data; The device hierarchical semantic network with physical topology is generated by using RDF triples constructed through semantic modeling standards and defining hierarchical relationships between device entities based on the basic data.

3. The fault diagnosis method for a commercial building HVAC system according to claim 1, characterized in that, The operating status of each device entity in the HVAC system is simulated to obtain a dynamic virtual model, specifically including: Based on a multilayer perceptron, real-time virtual sensors are generated to reflect the operating status of each device entity in the HVAC system; wherein, the real-time virtual sensors can mimic the operating status of physical entities in real time. The virtual sensors are used to simulate the operating status of each device entity in the HVAC system. Based on the parameter change status of the physical devices and the virtual in-situ modeling technology used in the HVAC system, a dynamic virtual model is generated. The dynamic virtual model is used to map one-to-one with each device entity in the HVAC system.

4. The fault diagnosis method for a commercial building HVAC system according to claim 1, characterized in that, Based on the preprocessed historical steady-state training set of the HVAC system, an unsupervised fault monitoring model is constructed, specifically including: Collect historical operational datasets from the HVAC system; Based on the reconstructed values ​​and reconstruction errors from principal component analysis, the distribution of daily minimum values ​​in the historical operating dataset is calculated, and a data cleaning threshold is obtained based on the system operating characteristics and skewness of the HVAC system. The historical running dataset is narrowed down by removing outliers from the data cleaning threshold to generate a steady-state training set of historical data. The unsupervised fault monitoring model is constructed based on the historical steady-state training set and the corresponding data validation set.

5. The fault diagnosis method for a commercial building HVAC system according to claim 1, characterized in that, The unsupervised fault monitoring model is used to perform real-time dynamic fault detection on the real-time operating data of the HVAC system, including: By using the sliding window in the unsupervised fault monitoring model, the average value and standard deviation of the reconstruction error of the real-time operating data under the time series are calculated to obtain the average value and standard deviation of the equipment operating data based on the equipment operating parameters of the HVAC system. Based on the preset adjustable multiplier and the moving variance of the reconstruction error, a fault dynamic threshold is calculated between the average value of the equipment operating data and the standard deviation of the equipment operating data to obtain the upper fault threshold and the lower fault threshold; wherein, the upper fault threshold and the lower fault threshold are the fault dynamic detection range of the operating data of each device in the HVAC system.

6. A fault diagnosis method for a commercial building HVAC system according to claim 5, characterized in that, The unsupervised fault monitoring model is used to determine fault triggers in the real-time operating data of the HVAC system, identifying the current fault data, including: Based on the preset ideal adjustable multiplier and the daily minimum average value of the reconstruction error in the real-time running data, a benchmark threshold is determined. If the average value of the device operation data in the current real-time operation data is greater than the benchmark threshold, then the fault trigger of the real-time operation data is determined as a level one fault warning; wherein, the level one fault warning is a minor fault warning; If the average value of the device operation data in the current real-time operation data is greater than the baseline threshold, and the device operation parameters in the current real-time operation data exceed the upper or lower fault threshold, then the fault triggering of the real-time operation data is determined as a level two fault warning; wherein, the level two fault warning is a severe fault warning; The current real-time operating data that meets the requirements of the fault dynamic detection range, the first-level fault warning, and the second-level fault warning is determined as the current fault data.

7. The fault diagnosis method for a commercial building HVAC system according to claim 1, characterized in that, Using the digital twin ontology framework, the current fault data is processed to locate the fault source under reconstruction errors at a hierarchical level, generating a fault diagnosis report, specifically including: The deviation of the reconstruction error between the normal state and the fault state is calculated on the current fault data to obtain the deviation result data; If the deviation result data is detected, the current fault data is identified and judged in relation to the central-level system and regional-level system of the HVAC system through the corresponding subordinate relationship of the system level in the digital twin ontology framework, so as to obtain the equipment-level fault result; wherein, the equipment-level fault result includes: regional-level equipment fault result and central-level equipment fault result; Based on the device-level fault results, and through the device mapping relationship in the digital twin ontology framework and physical sensor data, the fault diagnosis report is generated; wherein, the fault diagnosis report includes: the specific location of the fault, the scope of the fault impact, and the fault operation and maintenance strategy.

8. A fault diagnosis method for a commercial building HVAC system according to claim 7, characterized in that, The current fault data is used to identify and determine the central-level system and regional-level equipment of the HVAC system, resulting in equipment-level results, specifically including: Extract the key fault variables from the current fault data; If the key fault variable belongs to the abnormal state of reconstruction error of regional equipment, the key fault variable is analyzed and located according to the corresponding subordinate relationship of the system level in the digital twin ontology framework to obtain the fault result of the regional equipment. If the key fault variable belongs to the abnormal state of the reconstruction error of the central-level equipment, and the reconstruction error exceeds the fault threshold in the same period, then the key fault variable is used to identify the fault anomaly propagation of the regional-level equipment under the central-level equipment, and the fault result of the central-level equipment is obtained.

9. A fault diagnosis device for a commercial building HVAC system, characterized in that, The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor to enable the at least one processor to perform a fault diagnosis method for a commercial building HVAC system according to any one of claims 1-8.

10. A non-volatile computer storage medium, characterized in that, The storage medium is a non-volatile computer-readable storage medium that stores at least one program, each program including instructions that, when executed by a terminal, cause the terminal to perform a fault diagnosis method for a commercial building HVAC system according to any one of claims 1-8.

Images