Heating and ventilation equipment abnormity online monitoring system based on Internet of Things

By introducing technologies such as multi-parameter sensor array, deep learning analysis and knowledge graphs into the HVAC monitoring system, the shortcomings in data acquisition and predictive maintenance are solved, efficient and accurate abnormality detection and personalized maintenance are achieved, and the improvement of the improvement of the Management and maintenance efficiency of HVAC equipment.

CN118915566AInactive Publication Date: 2024-11-08JINAN THERMAL CO LTD

Patent Information

Application Number
CN202411062050.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-05
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing Internet of Things-based HVAC monitoring system has shortcomings in data acquisition accuracy and stability, loss and delay in data transmission, abnormal detection threshold management, predictive maintenance accuracy and effectiveness, resulting in frequent equipment failures. and high maintenance costs.

Method used

A HVAC equipment anomaly online monitoring system based on the Internet of Things is proposed, including data acquisition module, edge computing module, communication module, central processing module, adaptive threshold management module and predictive maintenance module, adopting multi-parameter sensor array and adaptive Sampling control, hybrid communication architecture, deep learning analysis, knowledge graph and machine learning prediction models to realize real-time data acquisition, preliminary anomaly detection, data transmission optimization, dynamic threshold management and personalized maintenance planning.

Benefits of technology

It improves the management and maintenance efficiency of HVAC equipment, achieves more accurate abnormality detection and diagnosis, reduces fault downtime and repair costs, and improves the operating reliability and user satisfaction of the equipment.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915566A_ABST
    Figure CN118915566A_ABST
Patent Text Reader

Abstract

The invention discloses a heating and ventilation equipment abnormity online monitoring system based on the Internet of Things, and belongs to the technical field of heating and ventilation control. The system comprises a data acquisition module used for acquiring multi-dimensional operation data of the heating and ventilation equipment in real time; the edge calculation module is responsible for executing real-time data processing, preliminary anomaly detection and local caching at a device end; the communication module is used for ensuring the reliability, safety and efficiency of data transmission; the central processing module is used for carrying out deep data analysis by utilizing cloud computing resources, maintaining a multi-dimensional dynamic reference model and carrying out comprehensive anomaly diagnosis in combination with a knowledge graph; the self-adaptive threshold management module is used for dynamically optimizing an abnormal detection threshold and balancing the detection accuracy and efficiency; the predictive maintenance module is used for predicting equipment performance trend and potential faults based on historical fault data and real-time data, formulating a personalized maintenance plan and optimizing resource scheduling; and the man-machine interaction module is used for providing a multi-platform and visual user interface.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of HVAC control technology, and more specifically, to an abnormal online monitoring system for HVAC equipment based on the Internet of Things. Background Art

[0002] With the rapid development of Internet of Things (IoT) technology, traditional HVAC equipment management methods are gradually being replaced by intelligent and networked solutions. HVAC equipment plays a vital role in modern buildings, and its operating status directly affects the comfort and energy consumption of the building. However, HVAC equipment has a complex operating environment and a large workload, and various faults and abnormalities often occur, which not only affects the life of the equipment, but may also cause serious economic losses and user dissatisfaction. Therefore, how to effectively monitor and maintain HVAC equipment has become an urgent problem to be solved in the industry.

[0003] Traditional HVAC equipment monitoring systems mainly rely on regular inspections and passive fault responses, which is not only inefficient, but also difficult to detect potential problems with equipment in a timely manner, resulting in frequent failures and high maintenance costs. With the continuous advancement of sensor technology, communication technology, and data processing technology, intelligent monitoring systems based on the Internet of Things have begun to be used in the field of HVAC equipment. These systems can realize online monitoring and intelligent diagnosis of equipment status by collecting equipment operation data in real time, combined with big data analysis and machine learning algorithms, thereby improving the reliability of equipment operation and maintenance efficiency.

[0004] Although the existing HVAC equipment monitoring system based on the Internet of Things has improved the monitoring and diagnosis capabilities to a certain extent, it still has some shortcomings. For example, the accuracy and stability of data collection are insufficient, data loss and delay problems during transmission are prominent, there is a lack of adaptive management of abnormal detection thresholds, and the accuracy and effect of predictive maintenance need to be improved. Therefore, a more complete system is urgently needed to solve the above problems and further improve the intelligent management level of HVAC equipment.

[0005] This application proposes an online monitoring system for HVAC equipment anomalies based on the Internet of Things. By comprehensively applying a variety of advanced technologies, it innovates in multiple aspects such as data acquisition, edge computing, communication optimization, deep data analysis, adaptive threshold management and predictive maintenance. It aims to provide an efficient, reliable and intelligent HVAC equipment monitoring solution to comprehensively improve the efficiency and effectiveness of equipment operation management. Summary of the invention

[0006] In order to overcome a series of defects in the prior art, the purpose of this application is to provide an abnormal online monitoring system for HVAC equipment based on the Internet of Things, including the following modules:

[0007] Data acquisition module, used to collect multi-dimensional operation data of HVAC equipment in real time;

[0008] The edge computing module is responsible for performing real-time data processing, preliminary anomaly detection, and local caching on the device side, reducing network load and improving response speed;

[0009] The communication module adopts a hybrid communication architecture and dynamic routing algorithm to ensure the reliability, security and efficiency of data transmission;

[0010] The central processing module uses cloud computing resources to perform in-depth data analysis, maintain a multi-dimensional dynamic benchmark model, and conduct comprehensive anomaly diagnosis in combination with knowledge graphs;

[0011] Adaptive threshold management module, used to dynamically optimize anomaly detection thresholds and balance detection accuracy and efficiency;

[0012] Predictive maintenance module, based on historical failure data and real-time data, predicts equipment performance trends and potential failures, develops personalized maintenance plans and optimizes resource scheduling;

[0013] The human-computer interaction module provides a multi-platform, visual user interface, integrates AR and natural language processing, and realizes intuitive and convenient system operation and information acquisition.

[0014] Furthermore, the data acquisition module includes the following units:

[0015] Multi-parameter sensor array integrates multiple sensors to comprehensively collect operating data of HVAC equipment;

[0016] Adaptive sampling control unit dynamically adjusts the sampling frequency according to the equipment operation status and data change rate, optimizing energy consumption while ensuring data accuracy;

[0017] Self-calibration unit, which regularly performs sensor self-test and calibration procedures to ensure the accuracy and stability of long-term data acquisition;

[0018] Environmental compensation unit monitors and compensates for the impact of environmental factors on sensor readings to improve data accuracy.

[0019] Furthermore, the central processing module includes the following units:

[0020] Data preprocessing unit, used to clean, standardize and integrate data from different devices and sensors to provide high-quality input for subsequent analysis;

[0021] Multi-dimensional benchmark performance model library, storing theoretical performance models of various HVAC equipment under different environmental conditions;

[0022] An initial benchmark performance curve generating unit extracts a corresponding model from a multi-dimensional benchmark performance model library according to device information and initial environmental conditions, and generates an initial benchmark performance curve for the device;

[0023] Dynamic benchmark model update unit, based on the initial benchmark performance curve, combined with the actual operation data of the equipment, environmental changes and historical performance, continuously optimizes and updates the benchmark model of equipment performance;

[0024] Deep learning analysis unit, which uses deep learning algorithms to analyze time series data and identify complex performance patterns and trends;

[0025] Knowledge graph engine, which integrates expert knowledge, historical cases and equipment-related information to build a comprehensive knowledge base and support intelligent diagnosis;

[0026] The anomaly diagnosis unit combines the baseline model, deep learning analysis results, and knowledge graph to accurately classify and analyze the root causes of detected anomalies;

[0027] The real-time decision support unit generates operational suggestions and warning information based on the diagnosis results, supporting operation and maintenance personnel to make quick decisions.

[0028] Furthermore, the baseline model, deep learning analysis results, and knowledge graph are combined to accurately classify and analyze the root causes of detected anomalies, including the following steps:

[0029] Compare the real-time monitoring data with the dynamic benchmark model, identify abnormal data points that deviate from the normal range and perform preliminary classification;

[0030] The detected abnormal data points are fed into a trained deep learning model to identify complex patterns in time series data and output the severity, duration, and potential failure type of the anomaly;

[0031] Based on the preliminary classification and deep learning analysis results, perform knowledge graph query and matching to quickly find historical cases similar to the current anomaly;

[0032] Comprehensively analyze the benchmark model comparison results, deep learning analysis output, and knowledge graph matching information to generate a comprehensive and accurate anomaly description and preliminary diagnosis results;

[0033] Based on the comprehensive analysis information, the root cause analysis inference engine is started to automatically check the relevant sensor data, verify the fault path of each root cause, and calculate the probability of various root causes;

[0034] Based on the inference and verification results, a diagnosis report containing multiple root causes is generated.

[0035] Further, based on the comprehensive analysis information, the root cause analysis inference engine is started to automatically check the relevant sensor data, verify the fault path of each root cause, and calculate the probability of various root causes, including the following steps:

[0036] Collect sensor data D and preliminary anomaly detection results A, where D = {d 1 ,d 2 ,…,d n}, the preliminary anomaly detection result A includes abnormal equipment status and possible failure modes;

[0037] Apply the root cause failure model F to reason about the preliminary anomaly detection result A and calculate each failure mode f i The likelihood P(A|f i );

[0038] Get each failure mode f i The prior probability P(f i );

[0039] According to Bayes' theorem, the posterior probability of each failure mode is calculated as follows: P(f i |A) = P(A|f i )·P(f i ) / P(A), where P(A) is the total probability of observing the initial abnormal detection result A, calculated by the weighted probability of all possible failure modes: P(A) = ∑ k=1 m P(A|f k )·P(f k ), where m represents the total number of failure modes; P(f k ) indicates the failure mode f k The prior probability of failure mode f is k The probability of occurrence; P(A|f k ) indicates that in the failure mode f k The probability of observing the preliminary anomaly detection result A under the condition of

[0040] Based on the calculated posterior probability, determine whether each fault mode is the root cause of A and verify the validity of the fault path.

[0041] Furthermore, the adaptive threshold management module includes the following units:

[0042] Fuzzy logic controller, which converts quantitative inputs of equipment status and environmental conditions into qualitative rules to achieve smooth adjustment of thresholds;

[0043] Genetic algorithm optimization engine, which continuously optimizes the threshold parameter combination to adapt to different operating conditions by simulating the evolution process;

[0044] Multi-objective evaluation unit, which comprehensively evaluates detection accuracy, false alarm rate and system efficiency, and provides comprehensive performance indicators for the optimization process;

[0045] The dynamic threshold calculation unit calculates and updates the abnormal detection threshold of each monitoring parameter in real time based on the output of the fuzzy logic controller and the genetic algorithm optimization engine;

[0046] Historical abnormal data analysis unit, which analyzes historical abnormal data and operation and maintenance feedback to provide long-term trend and seasonal change information for threshold adjustment;

[0047] Rapid response unit, which quickly adjusts thresholds to maintain the system's detection sensitivity in response to emergencies or rapidly changing operating conditions;

[0048] The threshold effect verification unit continuously monitors the detection effect after the threshold adjustment and feeds the results back to the genetic algorithm optimization engine to form a closed-loop optimization.

[0049] Furthermore, the anomaly detection threshold of each monitoring parameter is calculated and updated in real time by the following formula: j (t) = T j (t-1)+ΔT j (t)+Δθ j (t), where T j (t) is the threshold value of the jth monitoring parameter at time t; T j (t-1) is the threshold value of the jth monitoring parameter at time t-1; ΔT j (t) is the threshold adjustment value output by the fuzzy logic controller, which represents the adjustment value based on the current device status and environmental conditions; Δθ j (t) is the threshold adjustment amount output by the genetic algorithm optimization engine, which represents the parameter adjustment amount obtained based on the optimization algorithm.

[0050] Furthermore, the detection effect after the threshold adjustment is continuously monitored and the results are fed back to the optimization engine to form a closed-loop optimization. The specific steps are as follows:

[0051] Continuously collect and analyze key indicators of anomaly detection, including detection rate, false positive rate, and false negative rate, to evaluate the effectiveness of current threshold settings;

[0052] Regularly conduct multi-dimensional evaluations of the detection effect after threshold adjustment, including analysis of detection accuracy, response time, and resource utilization efficiency of different types of anomalies;

[0053] Perform difference analysis and cause inference on the detection effect before and after threshold adjustment, identify significant difference changes, and automatically analyze possible causes;

[0054] Generate a series of optimization suggestions based on the results of effect evaluation and gap analysis, each with a quantitative estimate of the expected improvement effect;

[0055] The evaluation results and optimization suggestions are input into the genetic algorithm optimization engine, and the optimization suggestions are prioritized according to the expected effect, implementation difficulty and current status of the system;

[0056] The genetic algorithm optimization engine gradually implements threshold adjustments and other optimization measures based on feedback and priority. After each adjustment, it immediately enters a new round of monitoring cycle to quickly evaluate the actual effect of the adjustment.

[0057] Furthermore, the following formula is used to determine whether the difference is significant: Among them, h is the standardized measure of the difference in effect before and after threshold adjustment relative to the standard error; ΔX is the difference in means before and after threshold adjustment; σΔX is the standard error of the mean difference; and e is the sample size of the threshold.

[0058] Furthermore, the predictive maintenance module includes the following units:

[0059] Data integration engine collects historical fault data and real-time equipment operation data and performs pre-processing to provide a comprehensive data foundation for predictive analysis;

[0060] A time series analysis unit that uses advanced time series algorithms to analyze equipment performance data and identify long-term trends, cyclical patterns, and unusual changes;

[0061] Machine learning prediction model: train machine learning models based on historical failure data to predict the type and time of possible failures of equipment;

[0062] The health assessment unit comprehensively considers the equipment operation status, prediction results and historical maintenance records, calculates and dynamically updates the equipment health index;

[0063] The maintenance strategy generation unit develops personalized maintenance plans and recommendations for each device based on health assessment and prediction results;

[0064] Resource optimization and scheduling unit, which takes into account factors such as maintenance needs, human resources, and spare parts inventory to optimize the overall scheduling and resource allocation of maintenance work;

[0065] The effect evaluation and feedback unit tracks the actual effects of maintenance activities and feeds the results back to the prediction model and strategy generator to continuously improve system performance.

[0066] Compared with the prior art, this application has the following beneficial effects:

[0067] This application uses technologies such as data collection, edge computing, hybrid communication, cloud computing, and knowledge graphs to monitor the operating status of HVAC equipment in real time, achieve anomaly detection and diagnosis, and optimize system performance and resource utilization efficiency through adaptive threshold management and predictive maintenance, thereby providing a multi-platform, visual, and intelligent human-computer interaction interface, significantly improving the management and maintenance efficiency of HVAC equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a structural schematic diagram of an abnormal online monitoring system for HVAC equipment based on the Internet of Things disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0069] In order to make the purpose, technical scheme and advantages of the implementation of the present invention clearer, the technical scheme in the embodiment of the present invention will be described in more detail below in conjunction with the drawings in the embodiment of the present invention. In the drawings, the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions. The described embodiments are part of the embodiments of the present invention, not all of the embodiments.

[0070] Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative work shall fall within the scope of protection of the present invention.

[0071] The embodiments and directional terms described below with reference to the accompanying drawings are exemplary and intended to be used to explain the present invention, but should not be construed as limiting the present invention.

[0072] like Figure 1 As shown, an abnormal online monitoring system for HVAC equipment based on the Internet of Things includes the following modules:

[0073] Data acquisition module, used to collect multi-dimensional operation data of HVAC equipment in real time;

[0074] The edge computing module is responsible for performing real-time data processing, preliminary anomaly detection, and local caching on the device side, reducing network load and improving response speed;

[0075] The communication module adopts a hybrid communication architecture and dynamic routing algorithm to ensure the reliability, security and efficiency of data transmission;

[0076] The central processing module uses cloud computing resources to perform in-depth data analysis, maintain a multi-dimensional dynamic benchmark model, and conduct comprehensive anomaly diagnosis in combination with knowledge graphs;

[0077] Adaptive threshold management module, used to dynamically optimize anomaly detection thresholds and balance detection accuracy and efficiency;

[0078] Predictive maintenance module, based on historical failure data and real-time data, predicts equipment performance trends and potential failures, develops personalized maintenance plans and optimizes resource scheduling;

[0079] The human-computer interaction module provides a multi-platform, visual user interface, integrates AR and natural language processing, and realizes intuitive and convenient system operation and information acquisition.

[0080] The data acquisition module plays a key role in the abnormal online monitoring system of HVAC equipment based on the Internet of Things. Its main task is to obtain the multi-dimensional operation data of HVAC equipment in real time, including temperature, humidity, pressure, wind speed and other parameters. The technical effect of this module is first reflected in the accuracy and timeliness of the data, ensuring that the true status of the equipment operation can be obtained. In real-time data acquisition, the accuracy and acquisition frequency of the sensor are crucial to the quality of the final data. In addition, the data acquisition module also needs to have a certain anti-interference ability to maintain the stability of data acquisition under environmental noise, equipment failure and other conditions. Through efficient data acquisition, the overall monitoring effect of the system and the timeliness of abnormal detection can be ensured, thus laying the foundation for subsequent processing and analysis.

[0081] The edge computing module is responsible for real-time data processing, preliminary anomaly detection, and local caching on the device side. Its technical effects are mainly reflected in reducing network load and improving response speed. Edge computing transfers data processing tasks from the central server to edge devices closer to the data source, which can significantly reduce the delay of data transmission and quickly respond to real-time data changes. Preliminary anomaly detection can be completed where the data is generated, and possible anomalies can be identified in time to avoid transmitting large amounts of data to the central server, which not only reduces the burden on the network, but also reduces the pressure on the central processing module. Edge computing also has a local caching function that can save data when the network fails or is delayed to ensure data integrity and consistency. This approach improves the robustness of the system and can continue basic monitoring and analysis when the network is in poor condition, thereby improving the reliability of the system.

[0082] The communication module plays an important role in connecting different modules and devices in the whole system. It adopts a hybrid communication architecture and dynamic routing algorithm to ensure the reliability, security and efficiency of data transmission. The hybrid communication architecture combines wired and wireless communication technologies, so that the system can maintain good communication quality in various environments. For example, wired communication can be used preferentially in a stable environment, while wireless communication is used when the equipment is widely distributed or the environment is complex. The dynamic routing algorithm is used to adjust the data transmission path in real time according to the network status to avoid data loss or delay caused by network congestion or failure. The communication module also needs to implement data encryption and authentication functions to ensure the security of data during transmission and prevent data leakage or tampering. These technical measures jointly ensure the communication stability of the system and the efficiency of data transmission.

[0083] The central processing module uses cloud computing resources to perform in-depth data analysis and abnormal diagnosis. Its technical effect is reflected in the processing ability and analysis depth of large amounts of data. Cloud computing resources have high computing power and storage capacity, and can process and analyze large amounts of data collected from various devices. By establishing a multi-dimensional dynamic benchmark model, the current operating status of the device can be compared with the normal operating status in real time to accurately identify abnormal situations. Combined with the knowledge graph for comprehensive abnormal diagnosis, valuable information can be extracted from the data to provide more accurate diagnostic results. The knowledge graph organizes the operating knowledge and abnormal patterns of the equipment in the form of a graph, which can perform complex reasoning and association analysis. The powerful computing and analysis capabilities of the central processing module enable the system to detect potential problems in a timely manner, provide comprehensive fault diagnosis and processing suggestions, and thus improve the intelligence level of the system.

[0084] The adaptive threshold management module is responsible for dynamically optimizing the anomaly detection threshold to balance detection accuracy and efficiency. Traditional anomaly detection methods usually use fixed thresholds to determine whether the device is abnormal, but this method can lead to false positives or false negatives. The adaptive threshold management module dynamically adjusts the detection threshold by analyzing historical data and real-time data to adapt to changes in the operating status of the device. The module uses machine learning algorithms and data mining technology to automatically identify the normal fluctuation range in the operation of the device and adjust it according to real-time data. This dynamic adjustment mechanism can effectively improve the accuracy of anomaly detection, reduce false positives and false negatives, and improve monitoring efficiency. The technical effect of the adaptive threshold management module is that it can flexibly adapt to changes in the status of the device to achieve more intelligent and efficient anomaly detection.

[0085] The technical effects of the predictive maintenance module are mainly reflected in fault prevention and resource optimization. Based on a large amount of historical data and real-time data, the predictive maintenance module uses predictive algorithms (such as regression analysis, time series analysis, and machine learning models) to predict the future state of the equipment and possible failures. This predictive capability enables the system to perform maintenance before a failure occurs, reducing equipment downtime and repair costs. The module can also formulate personalized maintenance plans based on the prediction results, optimize the scheduling of maintenance resources, and improve the efficiency of maintenance work. Predictive maintenance not only improves the reliability of equipment, but also effectively reduces overall operating costs and achieves more efficient equipment management.

[0086] The human-computer interaction module realizes intuitive and convenient system operation and information acquisition. The technical effect of this module is mainly reflected in the improvement of user experience. Through the visual interface, users can intuitively view the operating status, historical data and abnormal information of the equipment, so as to monitor and operate more conveniently. Augmented reality technology can display device information in the form of three-dimensional graphics in the user's field of vision, providing a more intuitive operation experience. Natural language processing technology allows users to interact with the system through voice or text, query information or issue instructions. This human-computer interaction method makes system operation more natural and convenient, lowers the operation threshold, and improves the user's work efficiency. By integrating these advanced technologies, the human-computer interaction module greatly improves the ease of use and intelligence level of the system.

[0087] Furthermore, the data acquisition module includes the following units:

[0088] Multi-parameter sensor array integrates multiple sensors to comprehensively collect operating data of HVAC equipment;

[0089] Adaptive sampling control unit dynamically adjusts the sampling frequency according to the equipment operation status and data change rate, optimizing energy consumption while ensuring data accuracy;

[0090] Self-calibration unit, which regularly performs sensor self-test and calibration procedures to ensure the accuracy and stability of long-term data acquisition;

[0091] Environmental compensation unit monitors and compensates for the impact of environmental factors on sensor readings to improve data accuracy.

[0092] The multi-parameter sensor array is the core component of the data acquisition module, which is responsible for comprehensively collecting various operating data of HVAC equipment, including multiple key parameters such as temperature, humidity, pressure, wind speed, etc. These sensors can comprehensively monitor the operating status of the equipment by integrating different types of sensors, such as temperature sensors, humidity sensors, pressure sensors, and wind speed sensors. The technical effects of the multi-parameter sensor array are reflected in several aspects. First, the integration of multiple sensors can provide comprehensive equipment status information, so that the operation of the equipment can be fully understood and potential anomalies or failures can be discovered. Secondly, through the integrated design, multiple sensors can work on the same platform, reducing the complexity of wiring and installation space, and improving the overall stability and reliability of the system. In addition, the integrated sensor array can also improve the synchronization and consistency of data acquisition, ensuring the timestamp and accuracy of the data. This comprehensive monitoring capability enables the system to perform more accurate data analysis and anomaly detection, providing a solid data foundation for subsequent processing and decision-making.

[0093] The adaptive sampling control unit is designed to optimize the efficiency and energy consumption of data acquisition. The unit dynamically adjusts the sampling frequency according to the operating status of the equipment and the rate of data change. Its technical effects are reflected in two main aspects. First, by dynamically adjusting the sampling frequency, the sampling frequency can be increased when the equipment status changes rapidly to ensure the timeliness and accuracy of the data; while when the equipment is running stably, the sampling frequency is reduced, thereby reducing energy consumption and data storage requirements. The adaptive sampling control unit monitors the operating status of the equipment in real time, analyzes the trend and amplitude of data changes, and intelligently adjusts the sampling strategy. This not only ensures the accuracy of the data, but also effectively reduces the burden of data acquisition and processing, extends the service life of the equipment, and reduces the energy consumption of the system. The design of this unit makes the data acquisition system more efficient and economical, and can adapt to the needs of different equipment and application scenarios.

[0094] The self-calibration unit is responsible for regularly performing self-check and calibration procedures on the sensor to ensure the long-term accuracy and stability of data collection. The technical effect of this unit is mainly reflected in improving data quality and system reliability. During long-term use, the sensor may cause reading deviations due to environmental changes, aging or other factors. The self-calibration unit automatically adjusts the measurement parameters of the sensor to maintain its accuracy by running self-check and calibration procedures regularly. The self-calibration process usually involves comparing the differences between the sensor readings and known standard values, and correcting these differences by adjusting the sensor's calibration coefficients. This design ensures that the sensor readings are reliable even during long-term operation, reducing the need for human intervention and improving the system's automation level and maintenance efficiency. In addition, the automatic calibration function can also reduce maintenance costs and operational complexity, allowing the system to operate more stably and permanently.

[0095] The environmental compensation unit is used to monitor and compensate for the impact of environmental factors on sensor readings, thereby improving the accuracy of data. Sensor readings may be affected by factors such as ambient temperature, humidity, and electromagnetic interference, which may cause deviations in measurement results. The environmental compensation unit monitors environmental conditions in real time and uses compensation algorithms to adjust sensor readings to eliminate these environmental effects. Its technical effects are reflected in several aspects. First, environmental compensation can significantly improve the accuracy of data, making sensor readings more consistent with the actual operating status of the device. Second, the compensation mechanism can reduce data noise caused by environmental changes and improve data reliability. The environmental compensation unit usually includes an environmental sensor and a compensation algorithm. The environmental sensor monitors environmental changes in real time, and the compensation algorithm adjusts the sensor readings based on the monitoring data. In this way, more stable and accurate data can be provided to support high-quality data analysis and decision-making. This compensation mechanism is crucial to maintaining data quality under various environmental conditions, especially for devices operating in complex or extreme environments.

[0096] Furthermore, the central processing module includes the following units:

[0097] Data preprocessing unit, used to clean, standardize and integrate data from different devices and sensors to provide high-quality input for subsequent analysis;

[0098] Multi-dimensional benchmark performance model library, storing theoretical performance models of various HVAC equipment under different environmental conditions;

[0099] An initial benchmark performance curve generating unit extracts a corresponding model from a multi-dimensional benchmark performance model library according to device information and initial environmental conditions, and generates an initial benchmark performance curve for the device;

[0100] Dynamic benchmark model update unit, based on the initial benchmark performance curve, combined with the actual operation data of the equipment, environmental changes and historical performance, continuously optimizes and updates the benchmark model of equipment performance;

[0101] Deep learning analysis unit, which uses deep learning algorithms to analyze time series data and identify complex performance patterns and trends;

[0102] Knowledge graph engine, which integrates expert knowledge, historical cases and equipment-related information to build a comprehensive knowledge base and support intelligent diagnosis;

[0103] The anomaly diagnosis unit combines the baseline model, deep learning analysis results, and knowledge graph to accurately classify and analyze the root causes of detected anomalies;

[0104] The real-time decision support unit generates operational suggestions and warning information based on the diagnosis results, supporting operation and maintenance personnel to make quick decisions.

[0105] The data preprocessing unit is an important part of the central processing module. It is responsible for cleaning, standardizing and integrating data from different devices and sensors to provide high-quality input for subsequent analysis. Its technical effects are mainly reflected in the following aspects. First, data cleaning can remove noise, errors and missing values ​​in the data, which is crucial to improving the accuracy and reliability of the data. Common data cleaning operations include removing outliers, filling missing data and correcting erroneous data. Secondly, data standardization ensures that data from different sensors and devices have a unified format and dimension, which is convenient for subsequent processing and analysis. For example, all data are uniformly converted to the same unit or scale to make the data comparable. The integration operation combines data from different sources to form a comprehensive data set. This comprehensive data set can more comprehensively reflect the actual operation of the equipment and provide a more accurate basis for subsequent analysis. Through the operation of the data preprocessing unit, the data quality can be significantly improved, the errors in the analysis can be reduced, and the analysis accuracy and reliability of the overall system can be improved.

[0106] The multidimensional benchmark performance model library is a key component in the central processing module, which is used to store theoretical performance models of various HVAC equipment under different environmental conditions. These performance models cover the theoretical performance of the equipment under different working conditions, environmental conditions and load conditions. The technical effects are mainly reflected in the following aspects. First, the multidimensional benchmark performance model library provides a systematic reference framework that can help analyze and compare the differences between actual operating data and theoretical models. By comparing actual data with theoretical performance models, the operating status of the equipment can be evaluated and possible anomalies or performance deviations can be found. Secondly, the benchmark performance model library can support performance prediction under different environmental conditions. For example, in extreme environments such as high temperature, high humidity or low temperature, the model library can provide the expected performance of the equipment. Through these multidimensional models, the performance of the equipment can be evaluated more accurately, and more targeted diagnosis and maintenance recommendations can be provided. The effective management and update of the multidimensional benchmark performance model library can enhance the system's understanding and prediction capabilities of equipment performance, thereby optimizing the maintenance and operation strategies of the equipment.

[0107] The initial benchmark performance curve generation unit extracts the corresponding model from the multi-dimensional benchmark performance model library based on the basic information of the equipment and the initial environmental conditions, and generates the initial benchmark performance curve of the equipment. Its technical effect is reflected in the generation of a performance benchmark curve that best matches the current state of the equipment. As the starting reference point for equipment performance, the initial benchmark performance curve is crucial for subsequent performance monitoring and anomaly detection. The unit first selects a matching theoretical performance model based on the model, specification, installation location and initial environmental conditions of the equipment, and generates the initial performance curve of the equipment based on these models. The initial performance curve provides a benchmark for the normal operating range and performance level of the equipment, and can compare and analyze actual data in real time during the operation of the equipment. In this way, the system can more accurately identify performance changes and potential anomalies, providing a reliable reference for long-term monitoring and maintenance of the equipment.

[0108] The dynamic benchmark model update unit is responsible for continuously optimizing and updating the benchmark model of equipment performance based on the initial benchmark performance curve, combined with the actual operation data of the equipment, environmental changes and historical performance. Its technical effect is mainly reflected in improving the adaptability and accuracy of the benchmark model. The performance of the equipment may be affected by various factors, such as environmental changes, equipment aging, load changes, etc. The dynamic benchmark model update unit continuously monitors the actual operation data of the equipment and combines these data to make real-time adjustments and optimizations to the initial performance curve. The unit uses a data-driven approach to continuously update the benchmark model to reflect the actual operation status and performance trends of the equipment. Through this dynamic adjustment mechanism, it can better adapt to the operation changes of the equipment and improve the accuracy and practicality of the performance benchmark model. In addition, the dynamic benchmark model update unit can also identify and correct long-term trends and periodic changes through the analysis of historical data, making the benchmark model more stable and reliable, thereby improving the overall diagnosis and prediction capabilities of the system.

[0109] The deep learning analysis unit uses deep learning algorithms to analyze time series data and identify complex performance patterns and trends. By building complex neural network models, deep learning technology can automatically extract features from a large amount of time series data and identify potential patterns and anomalies. Its technical effects are reflected in the following aspects. First, deep learning algorithms have powerful pattern recognition capabilities and can discover complex patterns and trends that are difficult to detect with traditional algorithms. For example, through models such as long short-term memory networks (LSTM) or convolutional neural networks (CNN), the operating data of equipment can be effectively analyzed to identify long-term trends, periodic changes, and transient anomalies. Second, deep learning analysis can process a large amount of time series data to achieve efficient feature extraction and anomaly detection. By training deep learning models, the ability to identify changes in equipment performance can be improved and the need for manual intervention can be reduced. The application of deep learning analysis units enables the system to achieve more intelligent data analysis and provide more accurate fault diagnosis and performance evaluation.

[0110] The knowledge graph engine integrates expert knowledge, historical cases, and equipment-related information to build a comprehensive knowledge base to support intelligent diagnosis. Its technical effect is mainly reflected in the organization and utilization of knowledge. The knowledge graph organizes domain knowledge, equipment operation history, failure mode and other information in the form of a graph, enabling the system to perform complex reasoning and association analysis. By building a knowledge graph, it is possible to integrate the expert's experience knowledge and historical data to provide support for abnormal diagnosis. For example, the knowledge graph can help the system identify common failure modes, the causes of related equipment failures, and apply this knowledge to real-time diagnosis. The associated information in the graph can also help the system discover hidden failure modes and provide more comprehensive diagnostic results. The technical effect of the knowledge graph engine is to improve the system's intelligence level and diagnostic capabilities, making the equipment failure analysis more accurate and comprehensive, thereby optimizing the equipment's maintenance and management strategies.

[0111] The technical effect of the abnormality diagnosis unit is reflected in improving the accuracy of abnormality detection and diagnosis. First, the abnormality diagnosis unit can identify abnormal conditions in equipment operation by comparing actual data with the benchmark model. Based on the results of deep learning analysis, the abnormalities can be further classified and evaluated. The knowledge graph provides rich information on failure modes and causes to help the system perform root cause analysis. Combined with this information, the abnormality diagnosis unit can generate detailed abnormality reports and provide accurate fault classification and root cause analysis. This not only improves the accuracy of anomaly detection, but also provides targeted solutions and maintenance suggestions, thereby improving the overall diagnostic capabilities and maintenance efficiency of the system.

[0112] The technical effect of the real-time decision support unit is reflected in improving decision-making efficiency and response speed. By analyzing the operating data and abnormal conditions of the equipment in real time, the decision support unit can generate specific operational suggestions, such as adjusting equipment parameters, performing maintenance operations, or replacing parts. In addition, the unit can also send early warning information to remind operation and maintenance personnel to pay attention to potential failures or equipment abnormalities. Real-time decision support units usually include a visual interface and an automated notification mechanism, allowing operation and maintenance personnel to quickly obtain information and respond. In this way, the efficiency of operation and maintenance work can be significantly improved, the downtime of equipment failures can be reduced, resource scheduling can be optimized, and ultimately the operating reliability of the equipment and the management level of the overall system can be improved.

[0113] Furthermore, the baseline model, deep learning analysis results, and knowledge graph are combined to accurately classify and analyze the root causes of detected anomalies, including the following steps:

[0114] Compare the real-time monitoring data with the dynamic benchmark model, identify abnormal data points that deviate from the normal range and perform preliminary classification;

[0115] The detected abnormal data points are fed into a trained deep learning model to identify complex patterns in time series data and output the severity, duration, and potential failure type of the anomaly;

[0116] Based on the preliminary classification and deep learning analysis results, perform knowledge graph query and matching to quickly find historical cases similar to the current anomaly;

[0117] Comprehensively analyze the benchmark model comparison results, deep learning analysis output, and knowledge graph matching information to generate a comprehensive and accurate anomaly description and preliminary diagnosis results;

[0118] Based on the comprehensive analysis information, the root cause analysis inference engine is started to automatically check the relevant sensor data, verify the fault path of each root cause, and calculate the probability of various root causes;

[0119] Based on the inference and verification results, a diagnosis report containing multiple root causes is generated.

[0120] Further, based on the comprehensive analysis information, the root cause analysis inference engine is started to automatically check the relevant sensor data, verify the fault path of each root cause, and calculate the probability of various root causes, including the following steps:

[0121] Collect sensor data D and preliminary anomaly detection results A, where D = {d 1 ,d 2 ,…,d n}, the preliminary anomaly detection result A includes abnormal equipment status and possible failure modes;

[0122] Apply the root cause failure model F to reason about the preliminary anomaly detection result A and calculate each failure mode f i The likelihood P(A|f i );

[0123] Get each failure mode f i The prior probability P(f i );

[0124] According to Bayes' theorem, the posterior probability of each failure mode is calculated as follows: P(f i |A) = P(A|f i )·P(f i ) / P(A), where P(A) is the total probability of observing the initial abnormal detection result A, calculated by the weighted probability of all possible failure modes: P(A) = ∑ k=1 m P(A|f k )·P(f k ), where m represents the total number of failure modes; P(f k ) indicates the failure mode f k The prior probability of failure mode f is k The probability of occurrence; P(A|f k ) indicates that in the failure mode f k The probability of observing the preliminary anomaly detection result A under the condition of

[0125] Based on the calculated posterior probability, determine whether each fault mode is the root cause of A and verify the validity of the fault path.

[0126] Furthermore, the adaptive threshold management module includes the following units:

[0127] Fuzzy logic controller, which converts quantitative inputs of equipment status and environmental conditions into qualitative rules to achieve smooth adjustment of thresholds;

[0128] Genetic algorithm optimization engine, which continuously optimizes the threshold parameter combination to adapt to different operating conditions by simulating the evolution process;

[0129] Multi-objective evaluation unit, which comprehensively evaluates detection accuracy, false alarm rate and system efficiency, and provides comprehensive performance indicators for the optimization process;

[0130] The dynamic threshold calculation unit calculates and updates the abnormal detection threshold of each monitoring parameter in real time based on the output of the fuzzy logic controller and the genetic algorithm optimization engine;

[0131] Historical abnormal data analysis unit, which analyzes historical abnormal data and operation and maintenance feedback to provide long-term trend and seasonal change information for threshold adjustment;

[0132] Rapid response unit, which quickly adjusts thresholds to maintain the system's detection sensitivity in response to emergencies or rapidly changing operating conditions;

[0133] The threshold effect verification unit continuously monitors the detection effect after the threshold adjustment and feeds the results back to the genetic algorithm optimization engine to form a closed-loop optimization.

[0134] In the adaptive threshold management module, the fuzzy logic controller is responsible for converting the quantitative inputs of the device status and environmental conditions into qualitative rules to achieve smooth adjustment of the threshold. Its technical effects are mainly reflected in the following aspects. The fuzzy logic controller can handle the complexity and uncertainty of the device status and environmental conditions, and achieve smooth control of the threshold by mapping these inputs to a set of fuzzy rules. Unlike traditional precise logic control, the fuzzy logic controller can cope with the ambiguity and uncertainty in reality. For example, the operating state of the device may be affected by multiple factors, and the changes of these factors are not always linear. By defining fuzzy rules (such as "high" temperature, "medium" humidity, etc.), the fuzzy logic controller can dynamically adjust the anomaly detection threshold according to the actual situation. This method makes the threshold adjustment more flexible and adaptable to environmental changes, avoiding the false alarm or missed alarm problem that may be caused by rigidly setting the threshold. In addition, the fuzzy logic controller can also model complex device behaviors and comprehensively consider multiple input factors, making the threshold adjustment process smoother and in line with the actual situation. In this way, more intelligent anomaly detection can be achieved, and the accuracy of overall monitoring and response can be improved.

[0135] The genetic algorithm optimization engine continuously optimizes the threshold parameter combination in the adaptive threshold management module by simulating the evolutionary process to adapt to different operating conditions. Its technical effect is reflected in improving the accuracy and adaptability of threshold settings. Genetic algorithm is an optimization method based on the principles of natural selection and genetics. It continuously generates and evaluates different threshold combinations by simulating the evolutionary process (including operations such as selection, crossover, and mutation) to find the optimal parameter settings. The genetic algorithm optimization engine first sets the current threshold initially, then generates multiple threshold combinations through an iterative process, and selects the optimal threshold parameters by evaluating the performance of these combinations (such as detection accuracy, false alarm rate, etc.). Compared with traditional optimization methods, genetic algorithms can effectively handle complex and multi-dimensional threshold optimization problems and improve the accuracy of threshold settings through multi-generation evolutionary processes. In addition, genetic algorithms have the ability to optimize globally, and can jump out of local optimal solutions and find threshold combinations that are more in line with actual conditions. Through this optimization mechanism, the threshold can be automatically adjusted under different operating conditions to ensure the accuracy and efficiency of detection.

[0136] The multi-objective evaluation unit is responsible for comprehensively evaluating the detection accuracy, false alarm rate and system efficiency in the adaptive threshold management module, and provides comprehensive performance indicators for the threshold optimization process. Its technical effect is mainly reflected in the comprehensiveness and scientificity of evaluating and optimizing the threshold setting. The multi-objective evaluation unit can provide a comprehensive performance evaluation system by comprehensively evaluating the detection performance of different threshold settings. For example, the detection accuracy index reflects the ability of the threshold setting to identify actual anomalies, the false alarm rate index measures the misjudgment rate of the threshold setting for normal data, and the system efficiency index evaluates the impact of the threshold setting on the overall system performance. By comprehensively considering these indicators, the threshold setting can be optimized to balance the detection accuracy and system efficiency. The multi-objective evaluation unit usually adopts a weighted scoring system or a multi-dimensional evaluation method to quantify various performance indicators and perform comprehensive optimization based on these indicators. Such an evaluation method can provide a more comprehensive and scientific threshold optimization solution to ensure the best performance and efficiency under different operating conditions.

[0137] The dynamic threshold calculation unit calculates and updates the anomaly detection threshold of each monitoring parameter in real time based on the output of the fuzzy logic controller and the genetic algorithm optimization engine. Its technical effect is reflected in improving the real-time and accuracy of threshold adjustment. The main task of the dynamic threshold calculation unit is to apply the rules provided by the fuzzy logic controller and the optimal threshold combination generated by the genetic algorithm optimization engine to actual data monitoring. By calculating and updating the threshold in real time, the detection threshold can be dynamically adjusted according to the current state of the device and environmental changes. For example, when the operating state of the device changes significantly, the dynamic threshold calculation unit can adjust the threshold in real time to maintain the sensitivity and accuracy of the detection. The design of this unit also includes the processing of real-time data streams to ensure that the update of the threshold can timely reflect the actual situation of the device. Through this dynamic adjustment mechanism, it is possible to adapt to changing operating conditions, optimize the performance of anomaly detection, reduce false alarms and missed alarms, and improve the effectiveness of the overall monitoring system.

[0138] The technical effect of the historical abnormal data analysis unit is mainly reflected in the long-term stability and adaptability of threshold adjustment. The historical abnormal data analysis unit identifies long-term trends and periodic changes by conducting in-depth analysis of past abnormal events and equipment operation and maintenance records. For example, by analyzing historical data, abnormal patterns of equipment in different seasons, operating cycles or workloads can be discovered, thereby providing a basis for threshold adjustment. The unit uses statistical analysis methods or machine learning techniques to perform pattern recognition and trend prediction on historical data. Based on these analysis results, the threshold can be adjusted to adapt to long-term trends and seasonal changes, and to avoid false alarms or omissions caused by environmental changes. The effective application of the historical abnormal data analysis unit can make more accurate threshold adjustments based on long-term data accumulation, improving the stability and reliability of the system.

[0139] The technical effect of the rapid response unit is mainly reflected in improving the system's response speed and sensitivity to sudden changes. During the operation of the equipment, sudden events may occur, such as equipment failure, environmental changes, etc., which can cause the equipment status to change dramatically. By monitoring the operating status and environmental changes of the equipment in real time, the rapid response unit can quickly adjust the detection threshold when an abnormal situation is found. For example, when a sharp fluctuation in the operating parameters of the equipment is detected, the rapid response unit can adjust the threshold in real time to maintain the sensitivity and accuracy of the detection. The unit includes an efficient data processing and adjustment mechanism that can complete the reset of the threshold in a short time, ensuring that the system can respond to changes in a timely manner and avoid missed reports or false alarms. Through this rapid response mechanism, stable detection performance can be maintained in various sudden situations, enhancing the robustness and response capabilities of the overall system.

[0140] The threshold effect verification unit is responsible for continuously monitoring the detection effect after the threshold adjustment, and feeding back the results to the genetic algorithm optimization engine to form a closed-loop optimization. Its technical effect is mainly reflected in the closed-loop optimization and continuous improvement of the threshold adjustment. The threshold effect verification unit can determine whether the adjustment has achieved the expected effect by monitoring and evaluating the actual detection effect after the threshold adjustment. For example, the effectiveness of the threshold adjustment can be verified by monitoring indicators such as detection accuracy, false alarm rate, and missed alarm rate. If the adjusted threshold leads to improvement or degradation of detection performance, the verification unit will feed back these results to the genetic algorithm optimization engine. The genetic algorithm optimization engine further optimizes the threshold setting based on this feedback information. Through this closed-loop optimization mechanism, the threshold setting can be continuously adjusted and improved to improve the accuracy and efficiency of anomaly detection. The design of the threshold effect verification unit ensures the continuous optimization and improvement of the system, can adapt to the changing equipment status and environmental conditions, and realize intelligent and adaptive threshold management.

[0141] Furthermore, the anomaly detection threshold of each monitoring parameter is calculated and updated in real time by the following formula: j (t) = T j (t-1)+ΔT j (t)+Δθ j (t), where T j (t) is the threshold value of the jth monitoring parameter at time t; T j (t-1) is the threshold value of the jth monitoring parameter at time t-1; ΔT j (t) is the threshold adjustment value output by the fuzzy logic controller, which represents the adjustment value based on the current device status and environmental conditions; Δθ j (t) is the threshold adjustment amount output by the genetic algorithm optimization engine, which represents the parameter adjustment amount obtained based on the optimization algorithm.

[0142] Furthermore, the detection effect after the threshold adjustment is continuously monitored and the results are fed back to the optimization engine to form a closed-loop optimization. The specific steps are as follows:

[0143] Continuously collect and analyze key indicators of anomaly detection, including detection rate, false positive rate, and false negative rate, to evaluate the effectiveness of current threshold settings;

[0144] Regularly conduct multi-dimensional evaluations of the detection effect after threshold adjustment, including analysis of detection accuracy, response time, and resource utilization efficiency of different types of anomalies;

[0145] Perform difference analysis and cause inference on the detection effect before and after threshold adjustment, identify significant difference changes, and automatically analyze possible causes;

[0146] Generate a series of optimization suggestions based on the results of effect evaluation and gap analysis, each with a quantitative estimate of the expected improvement effect;

[0147] The evaluation results and optimization suggestions are input into the genetic algorithm optimization engine, and the optimization suggestions are prioritized according to the expected effect, implementation difficulty and current status of the system;

[0148] The genetic algorithm optimization engine gradually implements threshold adjustments and other optimization measures based on feedback and priority. After each adjustment, it immediately enters a new round of monitoring cycle to quickly evaluate the actual effect of the adjustment.

[0149] Furthermore, the following formula is used to determine whether the difference is significant: Among them, h is the standardized measure of the difference in effect before and after threshold adjustment relative to the standard error; ΔX is the difference in means before and after threshold adjustment; σΔX is the standard error of the mean difference; and e is the sample size of the threshold.

[0150] Furthermore, the predictive maintenance module includes the following units:

[0151] Data integration engine collects historical fault data and real-time equipment operation data and performs pre-processing to provide a comprehensive data foundation for predictive analysis;

[0152] A time series analysis unit that uses advanced time series algorithms to analyze equipment performance data and identify long-term trends, cyclical patterns, and unusual changes;

[0153] Machine learning prediction model: train machine learning models based on historical failure data to predict the type and time of possible failures of equipment;

[0154] The health assessment unit comprehensively considers the equipment operation status, prediction results and historical maintenance records, calculates and dynamically updates the equipment health index;

[0155] The maintenance strategy generation unit develops personalized maintenance plans and recommendations for each device based on health assessment and prediction results;

[0156] Resource optimization and scheduling unit, which takes into account factors such as maintenance needs, human resources, and spare parts inventory to optimize the overall scheduling and resource allocation of maintenance work;

[0157] The effect evaluation and feedback unit tracks the actual effects of maintenance activities and feeds the results back to the prediction model and strategy generator to continuously improve system performance.

[0158] The data integration engine is responsible for collecting historical fault data and real-time equipment operation data in the predictive maintenance module, and preprocessing it to provide a comprehensive data foundation for predictive analysis. The technical effect of this process is mainly reflected in the establishment of a data-driven predictive analysis platform to ensure that subsequent analysis models and prediction systems can work on the basis of high-quality data. The data integration engine will first extract relevant information from multiple data sources, including historical fault records of equipment, real-time operation data, environmental conditions, etc. By integrating this data into a unified data platform, a comprehensive database can be formed to support subsequent predictive analysis. By establishing an efficient data integration mechanism, the data integration engine not only improves the availability of data, but also lays a solid foundation for accurate predictive analysis.

[0159] The time series analysis unit uses advanced time series algorithms to analyze equipment performance data and identify long-term trends, periodic patterns, and abnormal changes. The technical effect of this process is mainly reflected in the in-depth analysis of equipment operation data to reveal the time-dependent characteristics of equipment performance and potential abnormal patterns. The time series analysis unit uses a variety of advanced algorithms, such as autoregressive moving average (ARMA), autoregressive integrated moving average (ARIMA), and long short-term memory network (LSTM), to analyze the historical operation data of the equipment. These algorithms can capture long-term trends in the data (such as the gradual decline in equipment performance), periodic patterns (such as fluctuations in equipment performance over a specific time period), and sudden abnormal changes. By identifying these patterns, the future performance of the equipment can be predicted and potential signs of failure can be detected in a timely manner. For example, the identification of periodic patterns can help predict the maintenance cycle of the equipment, while the detection of abnormal changes can promptly warn of possible equipment failures. The in-depth analysis of the time series analysis unit not only improves the accuracy of the prediction, but also enhances the system's understanding of the equipment's operating status, thereby providing a reliable basis for subsequent predictive maintenance.

[0160] The machine learning prediction model trains the machine learning model based on historical fault data to predict the type and time of possible equipment failures. The technical effect of this process is mainly reflected in providing accurate fault predictions through data-driven models to improve the maintenance efficiency of equipment. The machine learning prediction model trains various machine learning algorithms (such as decision trees, random forests, support vector machines, and deep learning models) using historical fault data to identify potential patterns of equipment failures. The training process includes feature selection (extracting key features from historical data), model training (adjusting model parameters to improve prediction accuracy), and model validation (evaluating the prediction effect of the model). Once the model training is completed, it can analyze real-time data and predict the type and time of possible equipment failures. For example, by predicting the type of equipment failure, the maintenance team can prepare the required spare parts and tools in advance, while predicting the time of failure can help formulate maintenance plans and avoid sudden equipment downtime. The application of machine learning prediction models can significantly improve the accuracy of fault predictions and reduce the occurrence of unexpected failures, thereby improving the operational reliability and maintenance efficiency of equipment.

[0161] The technical effect of the health assessment unit is mainly reflected in providing equipment health status assessment through comprehensive analysis of various types of information, thereby supporting scientific maintenance decisions. The health assessment unit first summarizes the real-time operating status of the equipment (such as temperature, pressure, vibration, etc.), the fault prediction results of the machine learning prediction model, and the historical maintenance records of the equipment (such as past faults and maintenance history). After comprehensive analysis of this information, a dynamically updated equipment health index (HealthIndex) is calculated, which reflects the overall health status of the equipment and the risk of possible future failures. The calculation of the health index is usually combined with a weighted scoring model to comprehensively consider the influence of different factors. For example, the operating status of the equipment may occupy the main weight of the health index, while the historical maintenance records serve as auxiliary factors. By dynamically updating the health index, the health status of the equipment can be monitored in real time and data support can be provided for maintenance decisions. The effective application of the health assessment unit improves the scientificity and pertinence of equipment maintenance, reduces the risk of failures, and improves the reliability and stability of the equipment.

[0162] The maintenance strategy generation unit formulates personalized maintenance plans and recommendations for each device based on the health assessment and prediction results. The technical effect of this process is mainly reflected in providing targeted maintenance strategies to optimize the operation and maintenance management of the equipment. The maintenance strategy generation unit first evaluates the maintenance needs and priorities of each device based on the health index of the device and the output of the prediction model. Then, a personalized maintenance plan is formulated based on these evaluation results, including different types of maintenance activities such as regular inspections, preventive maintenance, predictive maintenance, and corrective maintenance. The maintenance strategy generation unit also considers factors such as the equipment's use environment, operating load, and historical fault records to ensure the comprehensiveness and effectiveness of the maintenance plan. For example, for equipment with a lower health index, the maintenance plan may include more frequent inspections and preventive maintenance to reduce the risk of failure; while for equipment with a higher health index, it may be recommended to reduce the maintenance frequency and optimize resource allocation. By formulating personalized maintenance plans, the system can improve the maintenance efficiency of the equipment, reduce maintenance costs, extend the service life of the equipment, and ultimately achieve the best operating state and maintenance effect of the equipment.

[0163] The resource optimization scheduling unit optimizes the overall scheduling and resource allocation of maintenance work by considering factors such as maintenance needs, human resources, and spare parts inventory. The technical effect of this process is mainly reflected in improving the efficiency and economy of maintenance work through scientific scheduling and resource management. The resource optimization scheduling unit formulates the optimal maintenance plan and resource allocation plan by analyzing factors such as equipment maintenance needs, available maintenance personnel, spare parts inventory, and maintenance time windows. The unit uses optimization algorithms (such as linear programming, integer programming, or heuristic algorithms) to solve resource constraints in maintenance scheduling to achieve efficient use of maintenance resources. For example, maintenance tasks are reasonably arranged according to the skills and available time of maintenance personnel; spare parts procurement and deployment are optimized according to spare parts inventory and usage frequency. The resource optimization scheduling unit also needs to monitor the progress of maintenance work in real time and adjust the resource allocation plan according to actual conditions. Through this scientific scheduling and resource optimization, the system can reduce maintenance time and cost, improve the overall efficiency of maintenance work, and ensure that equipment operates in the best state.

[0164] The effect evaluation and feedback unit tracks the actual effect of maintenance activities and feeds back the results to the prediction model and strategy generator to continuously improve system performance. The technical effect of this process is mainly reflected in the continuous optimization of system performance through the feedback mechanism. The effect evaluation and feedback unit first collects the actual effect data after the maintenance activities, including the operating status of the equipment, the failure rate after maintenance, the actual cost and time of maintenance, etc. Then, these data are analyzed to evaluate the gap between the actual effect and the expected effect of the maintenance activities. For example, by comparing the equipment failure rate after maintenance with the predicted failure rate, the accuracy of the prediction model is evaluated; by analyzing the actual maintenance cost with the budgeted cost, the economic feasibility of the maintenance strategy is evaluated. After the evaluation results are fed back to the prediction model and strategy generator, adjustments and optimizations can be made based on these feedback information. For example, the prediction model can be retrained and adjusted according to the actual effect data to improve the accuracy of the prediction; the strategy generator can optimize the maintenance plan according to the feedback results to improve the effect and efficiency of maintenance. Through this continuous effect evaluation and feedback, the system can continuously improve its performance and enhance the scientificity and effectiveness of equipment maintenance.

[0165] Finally, it should be pointed out that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, a person skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An abnormal online monitoring system for HVAC equipment based on the Internet of Things, characterized in that: Includes the following modules: Data acquisition module, used to collect multi-dimensional operation data of HVAC equipment in real time; The edge computing module is responsible for performing real-time data processing, preliminary anomaly detection, and local caching on the device side, reducing network load and improving response speed; The communication module adopts a hybrid communication architecture and dynamic routing algorithm to ensure the reliability, security and efficiency of data transmission; The central processing module uses cloud computing resources to perform in-depth data analysis, maintain a multi-dimensional dynamic benchmark model, and conduct comprehensive anomaly diagnosis in combination with knowledge graphs; Adaptive threshold management module, used to dynamically optimize anomaly detection thresholds and balance detection accuracy and efficiency; Predictive maintenance module, based on historical failure data and real-time data, predicts equipment performance trends and potential failures, develops personalized maintenance plans and optimizes resource scheduling; The human-computer interaction module provides a multi-platform, visual user interface, integrates AR and natural language processing, and realizes intuitive and convenient system operation and information acquisition.

2. According to the Internet of Things-based HVAC equipment abnormality online monitoring system according to claim 1, it is characterized in that: The data acquisition module includes the following units: Multi-parameter sensor array integrates multiple sensors to comprehensively collect operating data of HVAC equipment; Adaptive sampling control unit dynamically adjusts the sampling frequency according to the equipment operation status and data change rate, optimizing energy consumption while ensuring data accuracy; Self-calibration unit, which regularly performs sensor self-test and calibration procedures to ensure the accuracy and stability of long-term data acquisition; Environmental compensation unit monitors and compensates for the impact of environmental factors on sensor readings to improve data accuracy.

3. According to the Internet of Things-based HVAC equipment abnormality online monitoring system of claim 1, it is characterized in that: The central processing module includes the following units: Data preprocessing unit, used to clean, standardize and integrate data from different devices and sensors to provide high-quality input for subsequent analysis; Multi-dimensional benchmark performance model library, storing theoretical performance models of various HVAC equipment under different environmental conditions; An initial benchmark performance curve generating unit extracts a corresponding model from a multi-dimensional benchmark performance model library according to device information and initial environmental conditions, and generates an initial benchmark performance curve for the device; Dynamic benchmark model update unit, based on the initial benchmark performance curve, combined with the actual operation data of the equipment, environmental changes and historical performance, continuously optimizes and updates the benchmark model of equipment performance; Deep learning analysis unit, which uses deep learning algorithms to analyze time series data and identify complex performance patterns and trends; Knowledge graph engine, which integrates expert knowledge, historical cases and equipment-related information to build a comprehensive knowledge base and support intelligent diagnosis; The anomaly diagnosis unit combines the baseline model, deep learning analysis results, and knowledge graph to accurately classify and analyze the root causes of detected anomalies; The real-time decision support unit generates operational suggestions and warning information based on the diagnosis results, supporting operation and maintenance personnel to make quick decisions.

4. According to the Internet of Things-based HVAC equipment abnormality online monitoring system of claim 3, it is characterized in that: Combining the baseline model, deep learning analysis results, and knowledge graph, the detected anomalies are accurately classified and root cause analyzed, including the following steps: Compare the real-time monitoring data with the dynamic benchmark model, identify abnormal data points that deviate from the normal range and perform preliminary classification; The detected abnormal data points are fed into a trained deep learning model to identify complex patterns in time series data and output the severity, duration, and potential failure type of the anomaly; Based on the preliminary classification and deep learning analysis results, perform knowledge graph query and matching to quickly find historical cases similar to the current anomaly; Comprehensively analyze the benchmark model comparison results, deep learning analysis output, and knowledge graph matching information to generate a comprehensive and accurate anomaly description and preliminary diagnosis results; Based on the comprehensive analysis information, the root cause analysis inference engine is started to automatically check the relevant sensor data, verify the fault path of each root cause, and calculate the probability of various root causes; Based on the inference and verification results, a diagnosis report containing multiple root causes is generated.

5. According to the Internet of Things-based HVAC equipment abnormality online monitoring system of claim 4, it is characterized in that: Based on the comprehensive analysis information, the root cause analysis inference engine is started to automatically check the relevant sensor data, verify the fault path of each root cause, and calculate the probability of various root causes, including the following steps: Collect sensor data D and preliminary anomaly detection results A, where D = {d1, d2, …, d n }, the preliminary anomaly detection result A includes abnormal equipment status and possible failure modes; Apply the root cause failure model F to reason about the preliminary anomaly detection result A and calculate each failure mode f i The likelihood P(A|f i ); Get each failure mode f i The prior probability P(f i ); According to Bayes' theorem, the posterior probability of each failure mode is calculated as follows: P(f i |A) = P(A|f i )·P(f i ) / P(A), where P(A) is the total probability of observing the initial abnormal detection result A, calculated by the weighted probability of all possible failure modes: P(A) = ∑ k=1 m P(A|f k )·P(f k ), where m represents the total number of failure modes; P(f k ) indicates the failure mode f k The prior probability of failure mode f is k The probability of occurrence; P(A|f k ) indicates that in the failure mode f k The probability of observing the preliminary anomaly detection result A under the condition of Based on the calculated posterior probability, determine whether each fault mode is the root cause of A and verify the validity of the fault path.

6. According to the Internet of Things-based HVAC equipment abnormality online monitoring system of claim 1, it is characterized in that: The adaptive threshold management module includes the following units: Fuzzy logic controller, which converts quantitative inputs of equipment status and environmental conditions into qualitative rules to achieve smooth adjustment of thresholds; Genetic algorithm optimization engine, which continuously optimizes the threshold parameter combination to adapt to different operating conditions by simulating the evolution process; Multi-objective evaluation unit, which comprehensively evaluates detection accuracy, false alarm rate and system efficiency, and provides comprehensive performance indicators for the optimization process; The dynamic threshold calculation unit calculates and updates the abnormal detection threshold of each monitoring parameter in real time based on the output of the fuzzy logic controller and the genetic algorithm optimization engine; Historical abnormal data analysis unit, which analyzes historical abnormal data and operation and maintenance feedback to provide long-term trend and seasonal change information for threshold adjustment; Rapid response unit, which quickly adjusts thresholds to maintain the system's detection sensitivity in response to emergencies or rapidly changing operating conditions; The threshold effect verification unit continuously monitors the detection effect after the threshold adjustment and feeds the results back to the genetic algorithm optimization engine to form a closed-loop optimization.

7. The abnormal online monitoring system for HVAC equipment based on the Internet of Things according to claim 6 is characterized in that: The anomaly detection threshold of each monitoring parameter is calculated and updated in real time through the following formula: j (t) = T j (t-1)+ΔT j (t)+Δθ j (t), where T j (t) is the threshold value of the jth monitoring parameter at time t; T j (t-1) is the threshold value of the jth monitoring parameter at time t-1; ΔT j (t) is the threshold adjustment value output by the fuzzy logic controller, which represents the adjustment value based on the current device status and environmental conditions; Δθ j (t) is the threshold adjustment amount output by the genetic algorithm optimization engine, which represents the parameter adjustment amount obtained based on the optimization algorithm.

8. The abnormal online monitoring system for HVAC equipment based on the Internet of Things according to claim 6 is characterized in that: Continuously monitor the detection effect after the threshold adjustment and feed the results back to the optimization engine to form a closed-loop optimization. The specific steps are as follows: Continuously collect and analyze key indicators of anomaly detection, including detection rate, false positive rate, and false negative rate, to evaluate the effectiveness of current threshold settings; Regularly conduct multi-dimensional evaluations of the detection effect after threshold adjustment, including analysis of detection accuracy, response time, and resource utilization efficiency of different types of anomalies; Perform difference analysis and cause inference on the detection effect before and after threshold adjustment, identify significant difference changes, and automatically analyze possible causes; Generate a series of optimization suggestions based on the results of effect evaluation and gap analysis, each with a quantitative estimate of the expected improvement effect; The evaluation results and optimization suggestions are input into the genetic algorithm optimization engine, and the optimization suggestions are prioritized according to the expected effect, implementation difficulty and current status of the system; The genetic algorithm optimization engine gradually implements threshold adjustments and other optimization measures based on feedback and priority. After each adjustment, it immediately enters a new round of monitoring cycle to quickly evaluate the actual effect of the adjustment.

9. The abnormal online monitoring system for HVAC equipment based on the Internet of Things according to claim 8 is characterized in that: The following formula is used to determine whether the difference is significant: Among them, h is the standardized measure of the difference in effect before and after threshold adjustment relative to the standard error; ΔX is the difference in means before and after threshold adjustment; σΔX is the standard error of the mean difference; and e is the sample size of the threshold.

10. An abnormal online monitoring system for HVAC equipment based on the Internet of Things according to any one of claims 1 to 9, characterized in that: The predictive maintenance module includes the following units: Data integration engine collects historical fault data and real-time equipment operation data and performs pre-processing to provide a comprehensive data foundation for predictive analysis; A time series analysis unit that uses advanced time series algorithms to analyze equipment performance data and identify long-term trends, cyclical patterns, and unusual changes; Machine learning prediction model: train machine learning models based on historical failure data to predict the type and time of possible failures of equipment; The health assessment unit comprehensively considers the equipment operation status, prediction results and historical maintenance records, calculates and dynamically updates the equipment health index; The maintenance strategy generation unit develops personalized maintenance plans and recommendations for each device based on health assessment and prediction results; Resource optimization and scheduling unit, which takes into account factors such as maintenance needs, human resources, and spare parts inventory to optimize the overall scheduling and resource allocation of maintenance work; The effect evaluation and feedback unit tracks the actual effects of maintenance activities and feeds the results back to the prediction model and strategy generator to continuously improve system performance.

Citation Information

Patent Citations

  • Internet of Things based online monitoring system for abnormality of heating and air conditioning equipment

    CN106979590A

  • Pump equipment state monitoring dynamic threshold early warning method

    CN114135477A

  • Building energy consumption real-time monitoring control system and method

    CN117494032A

  • Wind turbine generator remote fault diagnosis method and system based on cloud platform

    CN117930815A

  • Intelligent monitoring system for urban infrastructure

    CN118348883A

Cited By

  • Communication fault sniffing method and system based on Internet of Things

    CN119135565A

  • Early warning method, device and equipment for cable joint and storage medium

    CN119149977A

  • End real-time monitoring and abnormal behavior analysis method based on deep learning large model

    CN119150241A

  • Mobile phone charging station state monitoring system for urban rail train

    CN119209928A

  • Safety detection method for boiler flue gas waste heat recovery power system

    CN119223662A