Intelligent operation and maintenance system for air conditioning unit based on multi-element perception

Abstract

This invention belongs to the field of air conditioning operation and maintenance technology, specifically involving an intelligent operation and maintenance system for air conditioning units based on multi-sensor technology. The system collects operating parameters of the air conditioning unit through a data acquisition module and transmits these parameters to a noise reduction module via a data transmission module. The noise reduction module filters, reduces noise, enhances features, and normalizes the data to form a standardized analysis dataset. A fault diagnosis module analyzes the standardized analysis dataset output by the noise reduction module to identify the fault type and severity level of the air conditioning unit, triggering an alarm if a fault is detected. A prediction module predicts future parameters based on historical data and the standardized analysis dataset output by the noise reduction module, triggering a warning when the predicted value exceeds a preset threshold. An energy consumption optimization module generates energy consumption optimization control commands based on the fault diagnosis results and trend prediction results. Finally, the system adjusts the operating parameters of each air conditioning unit according to the energy consumption optimization control commands, reducing costs, increasing efficiency, saving manpower, and improving the safety of air conditioning operation and maintenance.

Description

Technical Field

[0001] This invention belongs to the field of air conditioning operation and maintenance technology, specifically relating to an intelligent operation and maintenance system for air conditioning units based on multi-sensor perception. Background Technology

[0002] Air conditioning units are important infrastructure in complex and large public buildings such as hospitals, airport terminals, and shopping malls. These places are generally public places with complex building structures and high traffic. In order to ensure the orderly operation of production and life, it is essential to ensure that the ambient temperature inside the building meets the needs of production and life. Therefore, the monitoring and maintenance of the operating status of air conditioning units in buildings is crucial.

[0003] Traditional air conditioning unit operation status mainly relies on inspection personnel to visit the air conditioning unit locations one by one to check the operation status, which has the following shortcomings: (1) The inspectors mainly rely on their personal experience to inspect the air conditioners. Different inspectors have different judgment standards, and the results of their judgments on the same air conditioner under the same conditions vary greatly. (2) The environment where the air conditioner is located and the working years of the air conditioner will affect the working status of the air conditioner. These conditions will cause parameter drift during the operation of the air conditioner, affecting the accuracy of the inspection results. (3) Air conditioner compressors and other components are generally located outdoors, and some air conditioner units are located at high places. During inspection, staff need to climb to high places to work, which poses a safety hazard. (4) The structural condition of the compressor system, refrigeration system and lubrication structure can only be observed by disassembling the casing. For example, the mechanical structure of the compressor lacks lubricating oil or the structure is worn too much, which reduces the efficiency of inspection. (5) Personnel inspection can only inspect the air conditioner on a periodic basis. They can only obtain local static parameters during the inspection. They cannot achieve real-time continuous monitoring of the core parameters of the air conditioner, making it difficult to capture dynamic changes in parameters and predict potential faults of the air conditioner. As a result, maintenance is only carried out when the air conditioner is in an abnormal operating state, which affects the user experience.

[0004] In summary, the traditional manual inspection mode is not only labor-intensive, inefficient, and unsafe, but it also cannot accurately monitor the real-time dynamic status of the air conditioning unit. The monitoring results have large deviations and cannot predict potential operational abnormalities of the air conditioning, lacking the scientific and timely nature of air conditioning operation and maintenance. Summary of the Invention

[0005] The technical problem to be solved by this invention is to provide an intelligent operation and maintenance system for air conditioning units based on multi-sensor technology, which can efficiently, accurately and in real time realize the automatic monitoring and control of air conditioning units, thereby saving energy and manpower.

[0006] The technical solution adopted by the present invention to solve this technical problem is: an intelligent operation and maintenance system for air conditioning units based on multi-sensoring, including multiple air conditioning units, a data acquisition module for collecting the operating parameters of the air conditioning units, and a data transmission module whose input end is communicatively connected to the output end of the data acquisition module; It also includes a noise reduction module, a fault diagnosis module, a prediction module, an energy consumption optimization module, a control module, and a human-machine interaction module for displaying the system's operating status and receiving user commands; The output of the data transmission module is communicatively connected to the input of the noise reduction module. The output of the noise reduction module is communicatively connected to the inputs of the fault diagnosis module and the prediction module. The outputs of the fault diagnosis module and the prediction module are communicatively connected to the inputs of the energy consumption optimization module. The output of the energy consumption optimization module is communicatively connected to the inputs of the human-machine interaction module and the control module. The output of the control module is communicatively connected to the human-machine interaction module and the air conditioning unit in the air conditioning unit. The human-machine interaction module is used to display system information and receive external operation commands; the output of the human-machine interaction module is communicatively connected to the input of the control module, the fault diagnosis module, the prediction module and the energy consumption optimization module.

[0007] Furthermore, the data acquisition module includes a speed sensor for monitoring the compressor motor speed, a vibration sensor for monitoring the compressor housing vibration amplitude, a first temperature sensor for monitoring the compressor suction port temperature, a second temperature sensor for monitoring the compressor discharge port temperature, a first pressure sensor for monitoring the evaporator evaporation pressure, a condensation pressure sensor for monitoring the condenser pressure, a level sensor for monitoring the refrigerant level in the evaporator, a third temperature sensor for monitoring the evaporator inlet temperature, a fourth temperature sensor for monitoring the evaporator outlet temperature, a fifth temperature sensor for monitoring the condenser inlet temperature, and a fifth temperature sensor for monitoring the condenser outlet temperature. The following sensors are included: a sixth temperature sensor; an oil level sensor for monitoring the oil level in the compressor's lubricating oil tank; an oil temperature sensor for monitoring the oil temperature in the compressor's lubricating oil tank; an oil pressure sensor for monitoring the pressure at the compressor's oil supply port; a differential pressure sensor for monitoring the pressure difference across the compressor's oil filter element; a voltage sensor for monitoring the air conditioning circuit; a current sensor for monitoring the air conditioning circuit current; a frequency sensor for monitoring the inverter output frequency in the air conditioning circuit; an insulation resistance sensor for monitoring the insulation resistance between the compressor motor windings and ground; an ambient temperature sensor for monitoring the indoor temperature corresponding to a single air conditioner; a timer for monitoring the air conditioner's running time; and a counter for monitoring the number of times the air conditioner starts and stops. The output terminals of the speed sensor, vibration sensor, first temperature sensor, second temperature sensor, first pressure sensor, condensation pressure sensor, liquid level sensor, third temperature sensor, fourth temperature sensor, fifth temperature sensor, sixth temperature sensor, oil level sensor, oil temperature sensor, oil pressure sensor, differential pressure sensor, voltage sensor, current sensor, frequency sensor, insulation resistance sensor, ambient temperature sensor, timer, and counter are all communicatively connected to the noise reduction module through the data transmission module.

[0008] Furthermore, the data transmission module includes an industrial Ethernet switch, a 5G industrial router, and an edge computing node; the output of the data acquisition module is connected to the edge computing node through the industrial Ethernet switch and the 5G industrial router, respectively, and the output of the edge computing node is connected to the noise reduction module through the industrial Ethernet switch and the 5G industrial router, respectively. The edge computing node is used to receive the output data of the data acquisition module and convert the data signal into a unified format through a protocol; The 5G industrial router is normally closed, and is turned on when the Ethernet link fails.

[0009] Furthermore, the edge computing node is an edge computing gateway.

[0010] Furthermore, the noise reduction module stores an isolated forest algorithm for preliminary filtering of received data, a Kalman filter algorithm for data noise reduction, a time-domain statistical feature algorithm for data feature enhancement, and a Min-Max normalization algorithm for normalizing the enhanced data.

[0011] Furthermore, the fault diagnosis module stores a deep learning model and an ensemble learning algorithm. The deep learning model is a Transformer large model, and the ensemble learning algorithm is the XGBoost algorithm. The fault diagnosis module uses its internally stored Transformer large model and the XGBoost algorithm to analyze and diagnose system faults.

[0012] Furthermore, the prediction module stores an LSTM neural network model for predicting parameter change trends, and the energy consumption optimization module stores a SAC (SoftActor-Critic) algorithm model for energy consumption optimization calculations.

[0013] Furthermore, the human-machine interaction module includes an industrial touchscreen, computer, and mobile phone that perform data synchronization and access management through a unified cloud platform or edge server. The human-machine interaction module has alarm functions and work order closed-loop management functions.

[0014] Compared with existing technologies, the beneficial effects of this invention are: it provides an intelligent operation and maintenance system for air conditioning units based on multi-sensor processing. After the data acquisition module automatically collects the operating parameters of the air conditioning unit in real time, the noise reduction module sequentially filters, reduces noise, enhances features, and normalizes the monitored data to remove outliers caused by data drift and restores the measured electrical signals as accurately as possible. All features after normalization are mapped to the same scale to ensure that the final standardized feature vector is treated equally in subsequent steps. The normalized data is then input into the fault diagnosis module and the prediction module respectively to simultaneously diagnose and predict the operating status of the air conditioning unit. The diagnostic and prediction results are transmitted to the energy consumption optimization module. This module aims to minimize energy consumption while maintaining comfort levels. Based on the system's diagnostic and prediction results, it generates optimized control commands and automatically adjusts the air conditioning unit's operating mode in real-time, efficiently, and accurately. The human-machine interface module of the intelligent air conditioning unit operation and maintenance system displays the system's diagnostic results, prediction results, and control commands, allowing maintenance personnel to promptly understand the real-time operating status of the air conditioning and the risks in the future. This makes the air conditioning unit more energy-efficient, meets environmental requirements, improves environmental comfort, saves manpower, and enhances the safety of system detection and control. Through the integration of the Transformer large model and the XGBoost algorithm, the AI ​​adaptive system fault diagnosis dynamically adapts to changes in operating conditions and equipment aging, improving the accuracy and adaptability of fault diagnosis. In short, this invention constructs an integrated closed-loop system of "monitoring-fusion-diagnosis-prediction-control-operation and maintenance," connecting the entire chain of data acquisition, intelligent analysis, optimized control, and operation and maintenance management to achieve cost reduction and efficiency improvement.

[0015] The system's fault diagnosis results trigger an alarm. Upon receiving the alarm, maintenance personnel promptly repair the corresponding faulty air conditioner, prioritizing those with more severe faults to expedite repairs. When the system's predicted value exceeds a preset threshold, an alarm is triggered. Staff then preemptively maintain the corresponding air conditioner based on the alarm's content to prevent downtime and ensure user comfort, and to avoid a backlog of repair work orders due to multiple simultaneous malfunctions. Furthermore, by modifying the thresholds for fault diagnosis and prediction results, and switching between manual and automatic control modes, the system's diagnostic, predictive, and automatic control standards can be quickly adjusted to adapt to different working environments, making it more flexible and adaptable. Attached Figure Description

[0016] Figure 1This is a schematic diagram of the intelligent operation and maintenance system for air conditioning units of the present invention; Figure 2 This is a flowchart of the intelligent operation and maintenance method for air conditioning units according to the present invention; Figure reference numerals: 1-Data acquisition module; 101-Speed ​​sensor; 102-Vibration sensor; 103-First temperature sensor; 104-Second temperature sensor; 105-First pressure sensor; 106-Condensation pressure sensor; 107-Liquid level sensor; 108-Third temperature sensor; 109-Fourth temperature sensor; 110-Fifth temperature sensor; 111-Sixth temperature sensor; 112-Oil level sensor; 113-Oil temperature sensor; 114-Oil pressure sensor; 115-Differential pressure sensor; 116-Voltage sensor; 117-Current sensor; 118-Frequency sensor; 119-Insulation resistance sensor; 120-Ambient temperature sensor; 121-Timer; 122-Counter; 2-Data transmission module; 31-Noise reduction module; 32-Fault diagnosis module; 33-Prediction module; 34-Energy consumption optimization module; 4-Control module; 5-Human-machine interaction module; 90-Air conditioning unit. Detailed Implementation

[0017] The present invention will be further described below with reference to the accompanying drawings and embodiments. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0018] As attached Figure 1-2As shown, an intelligent operation and maintenance system for air conditioning units based on multi-sensor perception includes multiple air conditioning units 90, a data acquisition module 1 for collecting operating parameters of the air conditioning units, and a data transmission module 2 whose input end is communicatively connected to the output end of the data acquisition module 1. It also includes a noise reduction module 31, a fault diagnosis module 32, a prediction module 33, an energy consumption optimization module 34, a control module 4, and a human-machine interaction module 5 for displaying system operating status and receiving user commands. The output end of the data transmission module 2 is communicatively connected to the input end of the noise reduction module 31, and the output end of the noise reduction module 31 is communicatively connected to the input ends of the fault diagnosis module 32 and the prediction module 33, respectively. The output terminals of the fault diagnosis module 32 and the prediction module 33 are respectively communicatively connected to the input terminal of the energy consumption optimization module 34. The output terminal of the energy consumption optimization module 34 is communicatively connected to the input terminal of the human-machine interaction module 5 and the input terminal of the control module 4. The output terminal of the control module 4 is communicatively connected to the human-machine interaction module 5 and the air conditioning unit 90 within the air conditioning unit. The human-machine interaction module 5 is used to display system information and receive external operation commands. The output terminal of the human-machine interaction module 5 is communicatively connected to the input terminals of the control module 4, the fault diagnosis module 32, the prediction module 33, and the energy consumption optimization module 34. The communication connections described in this invention can be wired or wireless.

[0019] When the system of this invention is working, the data acquisition module 1 continuously monitors various data indicators of the air conditioner in real time and transmits the monitored data to the noise reduction module 31 through the data transmission module 2 for cleaning and normalization. The small conversion analysis data formed after processing by the noise reduction module 31 is transmitted to the fault diagnosis module 32 and the prediction module 33 respectively. The fault diagnosis module 32 determines whether there is an abnormal operating condition of the air conditioner in the air conditioning unit by analyzing the data and transmits the fault diagnosis result to the energy consumption optimization module 34. The prediction module 33 predicts the future parameter changes of each air conditioner by analyzing the data. When the predicted value is close to or exceeds the preset threshold, an early warning is triggered, and the prediction result is transmitted to the energy consumption optimization module 34. The energy consumption optimization module 34 generates an optimization control command based on the prediction result of the prediction module 33 and the fault diagnosis result of the fault diagnosis module 32 and transmits the control command to the control module 4. At the same time, the prediction result and control command information are also uploaded to the human-machine interaction module 5 for display. The control module 4 adjusts the operating parameters such as the fan speed and temperature of the specified air conditioner in the air conditioning unit and the operating mode of the air conditioner according to the received control command, and feeds back the control status, control parameters and other data to the human-machine interaction module 5 for display. The human-machine interaction module 5 is not only used to display information such as fault diagnosis results, prediction results, control commands, and control status, but also to receive operation commands such as modifying the thresholds of fault diagnosis and prediction results and switching between manual and automatic control modes. After receiving the operation commands, the human-machine interaction module 5 will send the received commands back to the control module 4, the fault diagnosis module 32, the prediction module 33, and the energy consumption optimization module 34, so that the control module 4, the fault diagnosis module 32, the prediction module 33, and the energy consumption optimization module 34 can operate according to the reset thresholds and commands.

[0020] The intelligent operation and maintenance system for air conditioning units described in this invention automatically collects the operating parameters of the air conditioning units in real time through the data acquisition module 1. The monitoring data is then cleaned and normalized by the noise reduction module 31. The processed data is then input into the fault diagnosis module 32 and the prediction module 33 to simultaneously diagnose and predict the operating status of the air conditioning units. The diagnosis and prediction results are transmitted to the energy consumption optimization module 34. The energy consumption optimization module 34 generates optimized control commands based on the system's diagnosis and prediction results, and automatically adjusts the air conditioning unit's operating mode in real time, efficiently, and accurately through the control module 4. The human-machine interaction module 5 of the intelligent operation and maintenance system displays the system's diagnostic results, prediction results, control commands, and other information, allowing maintenance personnel to promptly understand the real-time operating status of the air conditioning and the risks in the future. This makes the air conditioning units more energy-efficient, meets environmental requirements, improves environmental comfort, saves manpower, and enhances the safety of system detection and control. Furthermore, by modifying the thresholds of the fault diagnosis and prediction results and switching between manual and automatic control modes, the diagnostic, prediction, and automatic control standards of the entire system can be quickly adjusted to adapt to different working environments, making it more flexible and adaptable.

[0021] Data acquisition module 1 is used to continuously monitor various data indicators of the air conditioner in real time and output the monitored data. The air conditioner mainly includes components such as an evaporator installed indoors, a compressor, condenser, and heat exchanger installed outdoors. To comprehensively and accurately monitor the operating status of the air conditioner, sensors are needed to monitor the performance of each component. Specifically, data acquisition module 1 includes a speed sensor 101 for monitoring the compressor motor speed, a vibration sensor 102 for monitoring the vibration amplitude of the compressor housing, a first temperature sensor 103 for monitoring the compressor suction port temperature, a second temperature sensor 104 for monitoring the compressor discharge port temperature, a first pressure sensor 105 for monitoring the evaporator evaporation pressure, a condensation pressure sensor 106 for monitoring the condenser pressure, a level sensor 107 for monitoring the refrigerant level in the evaporator, and a third temperature sensor 108 for monitoring the evaporator inlet temperature. A fourth temperature sensor 109 for monitoring the evaporator outlet temperature; a fifth temperature sensor 110 for monitoring the condenser inlet temperature; a sixth temperature sensor 111 for monitoring the condenser outlet temperature; an oil level sensor 112 for monitoring the oil level in the compressor lubricating oil tank; an oil temperature sensor 113 for monitoring the oil temperature in the compressor lubricating oil tank; an oil pressure sensor 114 for monitoring the compressor oil supply port pressure; a differential pressure sensor 115 for monitoring the pressure difference across the compressor oil filter element; a voltage sensor for monitoring the air conditioning circuit; and a current sensor for monitoring the air conditioning circuit. The output terminals of the following sensors are connected to the input terminals of the data transmission module 2: current sensor 117, frequency sensor 118 (inverter output frequency in air conditioning circuit), insulation resistance sensor 119 (insulation resistance sensor for monitoring insulation resistance between compressor motor windings and ground), ambient temperature sensor 120 (ambient temperature sensor for monitoring indoor temperature corresponding to a single air conditioner), timer 121 (timer for monitoring air conditioner running time), and counter 122 (counter for monitoring air conditioner start-stop frequency). That is, the output terminals of the speed sensor 101, vibration sensor 102, first temperature sensor 103, second temperature sensor 104, and first... The output terminals of pressure sensor 105, condensation pressure sensor 106, liquid level sensor 107, third temperature sensor 108, fourth temperature sensor 109, fifth temperature sensor 110, sixth temperature sensor 111, oil level sensor 112, oil temperature sensor 113, oil pressure sensor 114, differential pressure sensor 115, voltage sensor 116, current sensor 117, frequency sensor 118, insulation resistance sensor 119, ambient temperature sensor 120, timer 121, and counter 122 are all communicatively connected to the noise reduction module 31 through the data transmission module 2.When the air conditioner compressor is a magnetic levitation centrifugal compressor, a gap sensor should also be installed to monitor the levitation position of the compressor motor rotor, and a levitation current sensor should be installed to monitor the output current of the compressor's magnetic bearing power amplifier, in order to further comprehensively monitor the compressor's tooling status. The sampling frequency of the above sensors is generally set to more than once every 10 seconds to ensure real-time monitoring of system performance.

[0022] The data transmission module 2 transmits the sensor's monitoring data to the noise reduction module 31. The data transmission module 2 can use wired or wireless transmission, such as NVIDIA Jetson series edge computing nodes or Lantronix SmartEDGE G600 edge computing gateways, or industrial intelligent gateways like the Rubonton edge computing gateway or Zongheng Intelligent Control data security gateway. It can also be an industrial Ethernet switch or industrial router. Preferably, the data transmission module 2 includes an industrial Ethernet switch, a 5G industrial router, and an edge computing node. The output of the data acquisition module 1 is connected to the edge computing node via the industrial Ethernet switch and the 5G industrial router, respectively. The output of the edge computing node is connected to the noise reduction module 31 via the industrial Ethernet switch and the 5G industrial router, respectively. The edge computing node receives the output data from the data acquisition module 1 and converts the data signal into a unified format using a protocol. The 5G industrial router is normally closed and is activated when the Ethernet link fails. As the primary transmission device, the industrial Ethernet switch automatically switches to a 5G wireless link when the Ethernet link fails. Simultaneously, the edge computing node receives data from the data acquisition module 1 and converts the data signal into a unified format for easy processing. The system employs AES-256 encryption to encrypt transmitted data, combined with a device authentication mechanism to prevent data leakage and tampering. It supports mainstream industrial communication protocols such as Modbus, BACnet, and OPCUA, and is compatible with the interface requirements of different brands of air conditioning units.

[0023] Edge computing nodes can be edge computing gateways, PLC controllers, or industrial computers. Edge computing gateways are preferred as edge computing nodes. They have built-in multiple industrial communication protocols, enabling plug-and-play connection of devices from different brands and using different protocols, unifying data into a standard format. They offer flexible deployment, lower costs, and convenient remote management, allowing for remote configuration of all field gateways from a remote or central platform.

[0024] The noise reduction module 31 filters, denoises, and enhances the features of the data acquired by the sensors in the data acquisition module 1, and then normalizes it to form a standardized dataset. This dataset prepares for the subsequent fault diagnosis by the fault diagnosis module 32 and the prediction by the prediction module 33, thereby improving the accuracy of the subsequent data processing results. The noise reduction module 31 can be implemented using hardware such as an edge server or an FPGA accelerator card. The noise reduction module 31 first uses algorithms such as the isolated forest algorithm or box plot method to detect outliers to filter the data, avoiding incorrect judgments (such as false alarms) caused by a single outlier, and also preventing outliers from skewing the normalization parameters and affecting the distribution of the entire dataset. Then, it uses one or more of the moving average filtering algorithm, Kalman filtering algorithm, or median filtering algorithm to perform noise reduction processing, suppressing random noise and preserving or restoring the true signal change trend. It uses time-domain statistical feature algorithms, frequency-domain feature algorithms, or empirical model decomposition algorithms to perform feature enhancement processing to construct higher-dimensional features with more physical meaning and discriminative ability from the original data. Finally, it uses Min-Max normalization algorithm, Z-Score normalization algorithm, etc. to normalize the data after feature enhancement to ensure that all features are mapped to the same scale, eliminating the influence of units and value ranges, so that the final standardized feature vector is treated equally when input into the models in the fault diagnosis module 32 and prediction module 33, accelerating model convergence and improving the stability and generalization ability of the models in the fault diagnosis module 32 and prediction module 33. Specifically, preferably, the noise reduction module 31 stores an isolated forest algorithm for preliminary filtering of received data, a Kalman filter algorithm for data noise reduction, a time-domain statistical feature algorithm for data feature enhancement, and a Min-Max normalization algorithm for normalizing the enhanced data.

[0025] The fault diagnosis module 32 is mainly used to analyze the standardized analysis dataset processed by the noise reduction module 31 to determine whether the air conditioner has a fault, and to identify the fault type and severity level. The fault diagnosis module 32 generally includes hardware devices such as a GPU server, CPU host, and storage devices. Preferably, the fault diagnosis module 32 stores a deep learning model and an ensemble learning algorithm, wherein the deep learning model is a Transformer large model, and the ensemble learning algorithm is the XGBoost algorithm. The fault diagnosis module 32 completes the analysis and diagnosis of system faults through its internally stored Transformer large model and the XGBoost algorithm. The Transformer algorithm is responsible for automatically learning complex, long-distance dependency features from massive, high-dimensional sensor time-series data, further performing deep encoding of the time-series features, capturing long-term dependencies, and dynamically adjusting the fault judgment threshold according to factors such as changes in air conditioner operating conditions and equipment aging. The XGBoost algorithm, based on these extracted high-quality features, performs efficient and accurate classification and risk probability prediction, identifies single / compound fault types, outputs the fault type and severity level, and ranks the faults by feature importance. The fault diagnosis module 32 of this type can output multiple fault types at the same time, avoiding the limitations of a single fault assumption. It can dynamically adjust the fault judgment threshold to adapt to different load rates, ambient temperatures and equipment aging, reduce false alarms. At the same time, it sorts the output features according to their importance, so that maintenance personnel can clearly know the parameters of the faulty components, which is convenient for maintenance, improves the overall maintenance efficiency of the air conditioning unit and saves manpower.

[0026] The prediction module 33 is mainly used to analyze the standardized analysis dataset processed by the noise reduction module 31 and predict the parameter change trend of each air conditioner in the future. When the predicted value approaches the safety threshold, an early warning signal is generated and sent to the human-computer interaction module 5. Generally, it is sufficient to predict the parameter change results in the next 48-72 hours. The prediction module 33 stores a prediction model, which can be an LSTM (main model), XGBoost + time series feature engineering, or Transformer model. Because the changes in key air conditioner parameters are nonlinear, slow, and coupled with multiple variables, preferably, the prediction module stores an LSTM neural network model for predicting parameter change trends. This model can effectively learn dependencies that last for several hours or even days through a gating mechanism and performs well on medium-length sequences, such as predicting the data trend for the next 3 days based on the data of the past 7 days.

[0027] The energy consumption optimization module 34 is mainly used to analyze the energy consumption efficiency of the air conditioner under different combinations of operating parameters. Based on the fault diagnosis results output by the fault diagnosis module 32 and the prediction results output by the prediction module 33, it determines the key operating parameter combination that minimizes energy consumption and meets comfort requirements, and outputs the optimized control parameter setpoints to the control module 4. Key operating parameters may include compressor speed, condenser outlet water temperature, etc. The energy consumption optimization module 34 can generate energy consumption optimization parameters through PID control algorithm, adaptive control algorithm, learning regression algorithm, reinforcement learning algorithm, or XGBoost energy consumption model + GA optimization algorithm. Preferably, the energy consumption optimization module 34 stores the SAC (SoftActor-Critic) algorithm model for energy consumption optimization calculation. This algorithm is suitable for the continuous operation space of the air conditioning unit, has high sample processing efficiency and strong robustness, and can be trained by offline pre-training combined with online fine-tuning. It adapts to equipment aging and seasonal changes and is more suitable for global optimization of multiple parameters of the air conditioning system.

[0028] Control module 4 is used to adjust the operating parameters such as fan speed and temperature of designated air conditioners in the air conditioning unit, as well as the operating mode of the air conditioner, according to the control commands output by energy optimization module 34. Control module 4 communicates with the actuators of the air conditioning unit, such as the compressor inverter, water pump inverter, electronic expansion valve, and fan, to achieve remote intelligent control and adaptive adjustment of the air conditioning unit. Specifically, control module 4 can be a PLC (Programmable Logic Controller), an industrial computer, or a microcontroller.

[0029] The human-machine interaction module 5 is used to display system operating status parameters such as air conditioning unit operating parameters, fault diagnosis conclusions, trend prediction curves, and energy consumption optimization reports, and to receive user operation commands such as alarm confirmation, work order processing, parameter setting, and remote control. The human-machine interaction module 5 supports multi-terminal collaborative work across computers, mobile devices, and industrial field terminals, achieving comprehensive visualized operation and maintenance management. Preferably, the human-machine interaction module 5 includes an industrial touchscreen, a computer, and a mobile phone that synchronize data and manage permissions through a unified cloud platform or edge server. The mobile phone and computer, along with the human-machine interaction module 5, possess alarm functions and work order closed-loop management functions. Maintenance personnel can remotely monitor system operation and confirm alarms and process work orders via mobile phones and computers.

[0030] The intelligent operation and maintenance method for air conditioning units using the above-mentioned intelligent operation and maintenance system includes the following steps: S1. The data acquisition module 1 collects the operating parameters of the air conditioning unit in real time and transmits the collected operating parameters to the noise reduction module 31 through the data transmission module 2. S2. The noise reduction module 31 performs data filtering, noise reduction, feature enhancement, and normalization on the operating parameters output in step S1 to form a standardized analysis dataset. S3. The fault diagnosis module 32 analyzes the standardized analysis dataset output from step S2 based on machine learning algorithms to identify the fault type and severity level of the air conditioning unit. If a fault exists, an alarm is triggered. S4. Prediction module 33 predicts future parameter change trends based on historical data and the standardized analysis dataset output from step S2. When the predicted value exceeds the preset threshold, an early warning is triggered. S5. The energy consumption optimization module 34 generates energy consumption optimization control instructions based on the fault diagnosis results output in step S3 and the trend prediction results output in step S4. S6. Based on the energy consumption optimization control command generated in step S5, switch the operating mode of each air conditioner in the air conditioning unit and adjust the operating parameters of each air conditioner in the air conditioning unit.

[0031] The air conditioning operation and maintenance method of this invention automatically monitors and acquires the real-time operating status of air conditioning units. The collected air conditioning operating parameters are then filtered, denoised, and enhanced before normalization to remove outliers caused by data drift during data acquisition. Noise reduction and feature enhancement methods are used to accurately reconstruct the measured electrical signals from the collected data. All features after normalization are mapped to the same scale to ensure that the resulting standardized feature vectors are treated equally in subsequent steps, improving the accuracy of subsequent system fault diagnosis and air conditioning unit operating trend prediction. This leads to the generation of more accurate energy consumption optimization control commands. Finally, the energy consumption optimization control commands are executed to regulate the air conditioning within the system to ensure the lowest energy consumption while maintaining comfort, thus saving energy. Specifically, if a fault is detected in the system fault diagnosis, an alarm is triggered. Upon receiving the alarm, maintenance personnel promptly repair the corresponding faulty air conditioner, prioritizing those with higher severity faults to expedite repairs. If the predicted value in step S4 exceeds a preset threshold, an alarm is triggered. Staff pre-maintain the corresponding air conditioner based on the alarm content to prevent air conditioning malfunctions from causing downtime and affecting user comfort, and to avoid a backlog of repair work orders due to multiple air conditioner malfunctions simultaneously.

[0032] The operating parameters of the air conditioning unit collected in step S1 are generally adjusted based on factors such as the age of the air conditioner, the working environment, and the frequency of start-stop operations. Specifically, the operating parameters of the air conditioning unit collected in step S1 include compressor motor speed, compressor housing amplitude, oil level and oil pressure in the compressor's lubricating oil tank, temperature of the compressor's suction and discharge ports, pressure in the evaporator and condenser chambers, refrigerant level in the evaporator, temperature of the evaporator's inlet and outlet, temperature of the condenser's inlet and outlet, air conditioner voltage, inverter output frequency, insulation resistance between the air conditioner compressor motor windings and ground, corresponding indoor ambient temperature, air conditioner start-up frequency, and number of start-stop operations, in order to monitor the air conditioner's operating status as comprehensively as possible.

[0033] Step S2 can first use algorithms such as the Isolation Forest algorithm or the Box Plot method to detect outliers and filter the data, avoiding incorrect judgments (such as false alarms) caused by a single outlier in the fault diagnosis module, and also preventing outliers from skewing the normalization parameters and affecting the distribution of the entire dataset. Then, one or more of the moving average filtering algorithm, Kalman filtering algorithm, or median filtering algorithm are used for noise reduction to suppress random noise and retain or restore the true signal change trend. Time-domain statistical feature algorithm, frequency-domain feature algorithm, or empirical model decomposition algorithm are used for feature enhancement to construct higher-dimensional features with more physical meaning and discriminative ability from the original data. Finally, the data after feature enhancement is normalized using the Min-Max normalization algorithm, Z-Score normalization algorithm, etc., to ensure that all features are mapped to the same scale, eliminate the influence of units and value ranges, and ensure that the final standardized feature vector is treated equally in the models in the fault diagnosis module 32 and the prediction module 33, accelerating model convergence and improving the stability and generalization ability of the models in the fault diagnosis module 32 and the prediction module 33.

[0034] Because the entire system involves a large number of samples and is complex in type, the fault diagnosis in step S3 preferably uses deep learning algorithms suitable for big data and complex time-series patterns, such as deep neural network algorithms, convolutional neural network algorithms, gated recurrent unit algorithms, and Transformer (time-series encoder) algorithms. Multiple algorithm integration models can also be used, such as extracting local features through convolutional neural networks and modeling time-series relationships through long short-term memory networks. Preferably, the machine learning algorithm in step S3 uses a deep learning model and an ensemble learning algorithm to dynamically adjust the fault judgment threshold. The deep learning model is a large Transformer model used to learn the operating rules of the entire equipment lifecycle; the ensemble learning algorithm is the XGBoost algorithm used to construct the fault diagnosis model. Based on the large Transformer model learning the operating rules of the entire equipment lifecycle, the long-term dependencies between parameters are captured through a self-attention mechanism. Combined with the XGBoost algorithm, feature importance ranking and automatic hyperparameter optimization are introduced to construct the fault diagnosis model, dynamically adjusting the fault judgment threshold to adapt to dynamic scenarios such as changing operating conditions and equipment aging; identifying single faults and compound faults, and outputting fault type, severity level, and source analysis.

[0035] The prediction model used in step S4 can be LSTM (the primary model), XGBoost+time series feature engineering model, or Transformer model. Because the changes in key air conditioning parameters are nonlinear, slow, and involve multivariate coupling, step S4 preferably uses an LSTM neural network model to predict the parameter change trend for the next 48-72 hours. When the predicted value approaches or exceeds a safety threshold, an early warning is triggered. The warning threshold is dynamically adjusted based on the equipment's operating status, and the warning lead time is optimized to 48-72 hours.

[0036] Steps S3 and S4 can be performed sequentially, or step S4 can be performed first and then step S3. Preferably, steps S3 and S4 are performed in parallel to improve the operation and maintenance efficiency of the air conditioning unit.

[0037] The energy consumption optimization module 34 in step S5 aims to minimize energy consumption and maximize operating efficiency. It dynamically optimizes operating parameters such as air conditioner compressor speed, inverter frequency, and condenser pressure setpoint through the DQN deep reinforcement learning algorithm to adapt to real-time load and environmental changes. For example, when the load rate is below 30%, it automatically adjusts the speed to below the inefficient range to reduce energy consumption.

[0038] In step S6, the control of each group of air conditioners can be automatically controlled by control module 4 or manually controlled. To improve control efficiency, automatic maintenance is preferred; that is, control module 4 controls the corresponding air conditioner according to the control command output in step S5. To achieve efficient management, it can be linked through mobile APP, WeChat mini-program, etc., so that maintenance personnel can receive faults and view key parameters in a timely manner, such as providing real-time operating status parameters of the host such as speed, temperature, and pressure, supporting real-time curves that can be switched between 1 hour / 24 hours / 7 days time dimensions, statistical distribution of fault types, frequency of occurrence, energy consumption statistics reports, and other key parameters. In addition, alarm notifications can be pushed through multiple channels such as pop-ups, SMS, APP push, and email. After an alarm is triggered, a maintenance work order is automatically generated, including fault details, handling suggestions, and spare parts list, supporting the full-process management of work order allocation, progress tracking, maintenance confirmation, and closed-loop archiving.

[0039] The embodiments described herein are preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape, and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. An intelligent operation and maintenance system for air conditioning units based on multi-sensor perception, comprising multiple air conditioning units (90), a data acquisition module (1) for collecting operating parameters of the air conditioning units, and a data transmission module (2) whose input end is communicatively connected to the output end of the data acquisition module (1), characterized in that, It also includes a noise reduction module (31), a fault diagnosis module (32), a prediction module (33), an energy consumption optimization module (34), a control module (4), and a human-machine interaction module (5) for displaying the system's operating status and receiving user instructions. The output of the data transmission module (2) is communicatively connected to the input of the noise reduction module (31). The output of the noise reduction module (31) is communicatively connected to the inputs of the fault diagnosis module (32) and the prediction module (33). The outputs of the fault diagnosis module (32) and the prediction module (33) are communicatively connected to the inputs of the energy consumption optimization module (34). The output of the energy consumption optimization module (34) is communicatively connected to the inputs of the human-machine interaction module (5) and the control module (4). The output of the control module (4) is communicatively connected to the human-machine interaction module (5) and the air conditioning unit (90) in the air conditioning unit. The human-machine interaction module (5) is used to display system information and receive external operation commands; the output end of the human-machine interaction module (5) is connected to the input end of the control module (4), the fault diagnosis module (32), the prediction module (33) and the energy consumption optimization module (34) respectively.

2. The intelligent operation and maintenance system for air conditioning units based on multi-sensoring as described in claim 1, characterized in that: The data acquisition module (1) includes a speed sensor (101) for monitoring the compressor motor speed, a vibration sensor (102) for monitoring the vibration amplitude of the compressor housing, a first temperature sensor (103) for monitoring the compressor suction port temperature, a second temperature sensor (104) for monitoring the compressor discharge port temperature, a first pressure sensor (105) for monitoring the evaporator evaporation pressure, a condensation pressure sensor (106) for monitoring the condenser pressure, a level sensor (107) for monitoring the refrigerant level in the evaporator, a third temperature sensor (108) for monitoring the evaporator inlet temperature, a fourth temperature sensor (109) for monitoring the evaporator outlet temperature, a fifth temperature sensor (110) for monitoring the condenser inlet temperature, and a sixth temperature sensor for monitoring the condenser outlet temperature. Sensors (111), oil level sensor for monitoring the oil level in the compressor lubricating oil tank (112), oil temperature sensor for monitoring the oil temperature in the compressor lubricating oil tank (113), oil pressure sensor for monitoring the oil supply port pressure of the compressor (114), differential pressure sensor for monitoring the pressure difference across the filter element of the compressor oil filter (115), voltage sensor for monitoring the air conditioning circuit (116), current sensor for monitoring the current in the air conditioning circuit (117), frequency sensor for monitoring the output frequency of the inverter on the air conditioning circuit (118), insulation resistance sensor for monitoring the insulation resistance value between the compressor motor winding and ground (119), ambient temperature sensor for monitoring the indoor temperature corresponding to a single air conditioner (120), timer for monitoring the running time of the air conditioner (121), and counter for monitoring the number of times the air conditioner starts and stops (122). The output terminals of the speed sensor (101), vibration sensor (102), first temperature sensor (103), second temperature sensor (104), first pressure sensor (105), condensation pressure sensor (106), liquid level sensor (107), third temperature sensor (108), fourth temperature sensor (109), fifth temperature sensor (110), sixth temperature sensor (111), oil level sensor (112), oil temperature sensor (113), oil pressure sensor (114), differential pressure sensor (115), voltage sensor (116), current sensor (117), frequency sensor (118), insulation resistance sensor (119), ambient temperature sensor (120), timer (121), and counter (122) are all connected to the noise reduction module (31) through the data transmission module (2).

3. The intelligent operation and maintenance system for air conditioning units based on multi-sensoring as described in claim 1, characterized in that: The data transmission module (2) includes an industrial Ethernet switch, a 5G industrial router and an edge computing node; the output end of the data acquisition module (1) is connected to the edge computing node through the industrial Ethernet switch and the 5G industrial router respectively, and the output end of the edge computing node is connected to the noise reduction module (31) through the industrial Ethernet switch and the 5G industrial router respectively. The edge computing node is used to receive the output data of the data acquisition module 1 and convert the data signal into a unified format through a protocol; The 5G industrial router is normally closed, and is turned on when the Ethernet link fails.

4. The intelligent operation and maintenance system for air conditioning units based on multi-sensoring as described in claim 3, characterized in that: The edge computing node is an edge computing gateway.

5. The intelligent operation and maintenance system for air conditioning units based on multi-sensoring as described in claim 3, characterized in that: The noise reduction module (31) stores an isolated forest algorithm for preliminary filtering of received data, a Kalman filter algorithm for data noise reduction, a time-domain statistical feature algorithm for data feature enhancement, and a Min-Max normalization algorithm for normalizing data after feature enhancement.

6. The intelligent operation and maintenance system for air conditioning units based on multi-sensoring according to any one of claims 1-5, characterized in that: The fault diagnosis module (32) stores a deep learning model and an ensemble learning algorithm. The deep learning model is a Transformer large model, and the ensemble learning algorithm is the XGBoost algorithm. The fault diagnosis module (32) completes the analysis and diagnosis of system faults through the Transformer large model stored in its internal storage and the XGBoost algorithm.

7. The intelligent operation and maintenance system for air conditioning units based on multi-sensoring as described in claim 6, characterized in that: The prediction module (33) stores an LSTM neural network model for predicting parameter change trends, and the energy consumption optimization module (34) stores an SAC algorithm model for energy consumption optimization calculation.

8. The intelligent operation and maintenance system for air conditioning units based on multi-sensoring as described in claim 6, characterized in that: The human-machine interaction module (5) includes an industrial touch screen, computer and mobile phone that perform data synchronization and permission management through a unified cloud platform or edge server. The human-machine interaction module (5) has alarm function and work order closed-loop management function.

Images