Air conditioning equipment running state monitoring method based on digital twinning

Abstract

The invention relates to the technical field of control system state monitoring, and discloses a digital twinning-based air conditioning equipment running state monitoring method, which comprises the following steps of: when a closed-loop control system controller is in a linear working area, injecting a harmonic probe signal which has no substantial influence on the physical output of the closed-loop control system controller; synchronously demodulating the error signal to construct a real-time complex gain vector representing the dynamic state of the system; and when the controller enters an output saturation region, injection of probe signals is paused, and saturation duration is obtained to serve as a judgment basis of limit performance, so that a parallel information obtaining path is established on the premise of not interfering equipment to maintain stable physical output; the inherent problem that the gradual physical attenuation of the controlled object is covered due to the successful operation of the control system is solved, and the on-line quantification of the integrity of the digital twin model and the distinguishing of the early failure mode are realized.

Description

TECHNICAL FIELD

[0001] The application relates to an air conditioning equipment operation state monitoring method based on digital twinning, and belongs to the technical field of control system state monitoring. BACKGROUND

[0002] At present, especially for application scenarios such as data centers which have strict requirements on the environment, a closed-loop feedback control system based on a digital twinning model is used to adjust a precision air conditioning equipment, which is a common technical method to ensure efficient and stable operation of the air conditioning equipment. Such a system compares the measured value of a temperature sensor with a preset target value at a high frequency, and continuously adjusts an actuator such as a compressor frequency converter or a valve opening degree by using a controller to avoid errors between the two, which has achieved success in maintaining the high stability of key physical parameters such as the temperature of the air outlet of the machine room.

[0003] However, the success of such a control system in eliminating apparent errors brings an implicit cost that has been long ignored when it faces the inevitable challenge of long-term continuous operation of the equipment. In actual engineering, the physical components of the air conditioning equipment, such as the heat exchanger fins, will accumulate dust at the micron level due to the electrostatic effect, or the refrigerant will have a small amount of leakage that is difficult to detect, which will cause the physical properties of the air conditioning equipment as a controlled object, such as heat exchange efficiency, to gradually and slowly decrease. In order to compensate for this small decrease in efficiency and maintain the same refrigeration capacity, the closed-loop controller must unknowingly slightly increase its average output value, which results in that, from all external physical monitoring data, the temperature output of the system is stable at the preset value, and the error signal is almost zero, but the digital twinning model, which is the basis of the control system, has actually deviated from the physical reality gradually and deeply.

[0004] To solve this problem, one intuitive idea is to add more physical sensors to the device, such as vibration sensors or acoustic sensors, to capture other associated signals during the physical attenuation process, but this approach not only increases the hardware cost and complexity of the system, but also introduces new diagnostic uncertainties due to the lack of direct and clear physical correlation between the new data stream and the original thermal model. This approach does not address the root cause of the problem, which is that the existing control system framework itself lacks an internal means to self-examine whether its model cognitive basis is still valid. Specifically, the existing technology mainly has the following deficiencies: 1. The monitoring information source relies on the stable physical appearance maintained by the controller. When the controlled object undergoes slow internal characteristic attenuation, the controller actively compensates for the attenuation to maintain stable output, resulting in an actual blind area for all physical output-based monitoring methods; 2. There is a lack of a non-intrusive online measurement method that can directly quantify the dynamic response capability changes of the controlled object without interrupting the normal operation of the device and interfering with its main control task; 3. The deviation between the digital twin model and the physical entity is masked and continuously accumulated by the normally operating controller, until the physical attenuation reaches a critical point, at which point the system will fail in a sudden manner, lacking early warning capabilities for potential risks. Therefore, how to develop a method that can penetrate the stable appearance of the control system itself to continuously evaluate the dynamic characteristics of the controlled object and judge the integrity of the digital twin model without adding additional high-cost hardware and affecting the normal control task of the system has become a technical problem to be solved by the present application. SUMMARY

[0005] The present application provides a digital twin-based air conditioning equipment operation state monitoring method, which mainly aims to solve the problem that the gradual attenuation of the physical characteristics of the controlled object is masked by the successful operation of the existing control system, resulting in a deviation between the digital twin model and reality that cannot be detected in a timely manner.

[0006] To achieve the above purpose, the digital twin-based air conditioning equipment operation state monitoring method provided by the present application is applied to a closed-loop control system comprising a controller and a controlled air conditioning equipment. In the application scenario, the digital twin-based air conditioning equipment operation state monitoring method establishes a dual-mode continuous interrogation procedure covering the linear operating region and the output saturation region. This dual-mode continuous interrogation procedure is a monitoring method that adaptively switches according to the working state of the controller, covering two working modes: one is when the controller is in the linear operating region, the dynamic characteristics of the controlled object are continuously interrogated by actively injecting harmonic probes and synchronously demodulating; the other is when the controller enters the output saturation region, the limit performance of the device is interrogated by obtaining the duration of the saturation state. The method comprises: Step a, injecting a harmonic probe signal in one control signal channel of the closed-loop control system when the controller is in the linear operating region, the harmonic probe signal having a frequency higher than the cutoff frequency of the physical output of the controlled air conditioning equipment and an amplitude not causing the physical output to exceed the noise fluctuation range of the physical output in normal operation; Step b, extracting in-phase component and quadrature component of the same frequency response signal in the error signal of the closed-loop control system by synchronous demodulation with the frequency of the harmonic probe signal, and constructing a real-time complex gain vector; Step c, comparing the real-time complex gain vector with a reference complex gain vector obtained when the controlled air conditioning equipment is in a preset healthy state to generate a model integrity index; Step d, determining that the controller enters the output saturation state when the controller enters the output saturation region, and immediately suspending the injection of the harmonic probe signal, and obtaining the duration of the output saturation state during the controller is in the output saturation state, and comparing the duration with a reference saturation duration to generate a limit performance index.

[0007] Preferably, the method further comprises: recording a trajectory formed by the real-time complex gain vector over time; calculating the first time derivative of the trajectory to obtain the evolution rate of the trajectory, and calculating the variance of the trajectory within a time window to obtain the stability of the trajectory; and generating a state flag representing the risk level of the air conditioning equipment operation based on the evolution rate and the stability.

[0008] Preferably, the method further comprises: continuously changing the frequency of the harmonic probe signal in a diagnostic frequency range with the cutoff frequency of the controlled air conditioning equipment as the lower limit, and synchronously calculating the real-time complex gain vector corresponding to each frequency to obtain a gain-frequency response curve; identifying a frequency with the maximum difference value as a characteristic frequency based on the difference between the gain-frequency response curve and a reference response curve; and updating the frequency of the harmonic probe signal to the characteristic frequency.

[0009] Preferably, the generation of the model integrity index specifically comprises calculating the ratio of the modulus of the real-time complex gain vector to the modulus of the reference complex gain vector.

[0010] Preferably, the step of comparing the real-time complex gain vector with the reference complex gain vector further comprises: calculating a deviation vector formed by subtracting the reference complex gain vector from the real-time complex gain vector; and distinguishing the energy-type decay mode and the delay-type decay mode according to the angle between the evolution direction of the deviation vector on the complex plane and the direction of the reference complex gain vector.

[0011] Preferably, the method further comprises: performing a notch filtering process on the error signal, the center frequency of the notch filtering process being the same as the frequency of the harmonic probe signal, so as to filter out the harmonic probe signal and its response signal component, thereby obtaining a noise accompanying signal; calculating a spectral entropy of a power spectrum of the noise accompanying signal; and comparing the spectral entropy with a reference spectral entropy, so as to determine whether a random disturbance type fault exists.

[0012] Preferably, the method further comprises: in a calibration mode, directly connecting an output path of the control signal to an input path of the error signal, so as to form an internal calibration loop; in the internal calibration loop, performing steps a to c, so as to obtain a calibration complex gain vector characterizing the self-distortion characteristics of the signal measurement link; and in a normal monitoring mode, compensating the real-time complex gain vector according to the calibration complex gain vector.

[0013] Preferably, the frequency of the harmonic probe signal is set to be in a range from five times to fifty times higher than a cutoff frequency of the physical output of the controlled air conditioning equipment.

[0014] Preferably, the synchronous demodulation is specifically point-by-point multiplication operation of the error signal with the harmonic probe signal and with a signal orthogonal to the harmonic probe signal by ninety degrees in phase, and low-pass filtering processing is performed on the results of the multiplication operation.

[0015] Preferably, the control signal is a signal output by a controller to a variable frequency compressor, and the error signal is a difference between a temperature setting value at an input end of the controller and a temperature measurement value of a temperature sensor.

[0016] Compared with the prior art, the present application has the following beneficial effects: 1. The present application establishes a parallel information acquisition path without interfering with the primary task of maintaining the stability of the physical output of the existing closed-loop control system of the air conditioning equipment; by injecting a probe signal with a frequency and amplitude that do not substantially affect the final temperature and other physical quantities into the output end of the controller, and simultaneously extracting the weak response of the probe from the error signal at the input end of the controller, a continuous inquiry into the dynamic characteristics of the controlled object is realized, and this way makes the monitoring of the accuracy of the digital twin model free from the dependence on the macroscopic error or the change of the physical appearance of the system, and solves the inherent problem that the control system covers up the gradual physical decay of the controlled object due to its own successful operation.

[0017] 2、By synchronously extracting the in-phase component and the quadrature component of the probe response in the error signal and constructing a complex gain vector, the originally one-dimensional system response amplitude evaluation is converted into two-dimensional state space trajectory analysis; when the physical characteristics of the device decay in different properties, for example, one type mainly affects the energy transmission efficiency of the system, and the other type mainly affects the system response time delay, the deviation of the complex gain vector will evolve in different directions on the complex plane, and this mechanism enables the system to add new calculation without increasing additional physical sensors, and the additional calculation mainly comes from synchronous demodulation and vector operation, which is negligible compared with the main control algorithm, that is, different early fault modes can be distinguished, and more targeted basis is provided for subsequent maintenance decision.

[0018] 3、The application further records the trajectory formed by the above-mentioned complex gain vector over time, and calculates the evolution rate and local stability of the trajectory; when two different fault modes show similar in the static position of the vector, but one is a slow and smooth gradual process, and the other is a rapid and irregular mutation process, by quantifying the dynamic characteristics of the trajectory, the system monitoring can not only identify the type of fault, but also evaluate the urgency of its development, so as to distinguish the operation risk in time scale and support the hierarchical response strategy from planned maintenance to emergency intervention.

[0019] 4、By continuously monitoring the running state of the controller main control signal, when it is identified that the controller enters the nonlinear working area of output saturation in response to extreme load, the injection of the harmonic probe signal will be suspended immediately, and then restored after the controller exits the saturation state; this not only avoids the pollution of invalid response signal to the continuity of model integrity during the actual disconnection of the control loop, ensures the purity of the basic diagnostic data, but also utilizes the duration of the saturation state as a new measurement dimension, which directly reflects the reserve capacity and control margin of the device under extreme working conditions, and converts a system state that is usually regarded as a monitoring blind area into a stress test of the limit performance of the device. BRIEF DESCRIPTION OF DRAWINGS

[0020] Fig. 1 Flow chart of the dual-mode processing and key index generation method of the application; Fig. 2 Waveform diagram of the probe excitation error response and synchronous demodulation signal of the application; Fig. 3 Module interaction diagram for fault classification and risk assessment of the application. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail below. It should be noted that the descriptive embodiments herein are intended to explain the present invention and are not intended to limit the scope of protection of the present invention.

[0022] This invention provides a digital twin-based method for monitoring the operational status of air conditioning equipment. Applied to a closed-loop control system comprising a controller and the controlled air conditioning equipment, this method establishes a dual-modal continuous challenge mode covering both the linear operating region and the output saturation region. This mode primarily includes an online dynamic characteristic calibration module, a controller status monitoring module, a saturation range performance evaluation module, and a diagnostic information comprehensive judgment module. Specifically, when the controller is in the linear operating region, the online dynamic characteristic calibration module actively injects harmonic probes and performs synchronous demodulation to continuously generate complex gain vectors characterizing the dynamic characteristics of the controlled object. The controller status monitoring module determines in real time whether the controller has entered the output saturation state and gates the operation of the aforementioned calibration module accordingly. Simultaneously, it activates the saturation range performance evaluation module, which records process parameters during the controller's saturation state. Finally, the diagnostic information comprehensive judgment module processes data from different modes to generate quantitative evaluation indicators for the integrity of the digital twin model and the equipment's ultimate performance.

[0023] In the operation monitoring of precision air conditioning equipment in data centers, a technical problem exists: the physical characteristics of the equipment slowly degrade due to factors such as dust accumulation in the heat exchanger or refrigerant micro-leakage. The closed-loop controller, by fine-tuning its output, maintains a stable outlet temperature, thus masking early signs of degradation. To address this problem, the method of this invention is configured to execute an active online challenge process within the linear operating range of the controller. This process first performs a signal injection step, i.e., step a, in a control signal channel of the closed-loop control system, for example, in the controller's main control signal... Before sending the signal to the variable frequency compressor, inject a harmonic probe signal into it. The harmonic probe signal here has two defining characteristics, one of which is frequency. The choice is, secondly, the amplitude. The setting; to determine the frequency The specific values ​​require a one-time offline system identification process. This involves applying a step input to the system during the equipment commissioning phase, recording the response curve of its physical output (such as temperature), and calculating its -3dB cutoff frequency. This cutoff frequency reflects the thermal inertia of the system. For precision air conditioning equipment, its typical value is on the order of 0.01 Hz. According to deterministic procedures, the probe frequency... It is set to a fixed multiple higher than the cutoff frequency, preferably within the range of five to fifty times the cutoff frequency. For example, if measured... If the frequency is 0.01Hz, then you can select... The frequency is set to 10Hz. This setting is based on the fact that the physical system of the controlled air conditioning unit acts as a low-pass filter, exhibiting response attenuation for probe signals with frequencies much higher than its cutoff frequency. This ensures that the injection of this signal will not cause fluctuations in the final physical output (temperature) that exceed the normal operating noise range. Regarding the amplitude... The principle behind this setting is to ensure that the power of the injected probe signal is lower than the control signal noise power caused by normal load fluctuations in the system. The specific value can be calculated by collecting control signals for a period of time during stable system operation and calculating their root mean square value. Then the amplitude By setting the value to a small percentage of the root mean square value, such as 10%, the harmonic probe signal provides a continuous excitation source for the online calibration of the system's dynamic characteristics without interfering with the device's primary control tasks.

[0024] After continuously injecting the aforementioned harmonic probe signal, the control loop information will include a weak response to the probe. Therefore, the system executes the following procedures. That is, from an error signal of the closed-loop control system Extract the response, where the error signal is... The difference between the temperature setpoint at the controller input and the temperature sensor measurement; to extract the weak response component with the same frequency as the probe signal from broadband noise, the system employs a synchronous demodulation algorithm, which specifically extracts the acquired error signal... This is achieved by performing point-by-point multiplication with both the original probe signal and a reference signal that is 90 degrees orthogonal to the original probe signal, followed by low-pass filtering of the two multiplied results. The mathematical expression is as follows: First, generate an in-phase reference signal. and quadrature reference signals Then calculate the in-phase components. and orthogonal components Where LPF represents low-pass filtering, and the cutoff frequency of this low-pass filter is set below the probe frequency. For example, 0.1Hz, its function is to filter out all non- The frequency noise and second harmonic component, and finally the DC component. and These represent the amplitudes of the in-phase and quadrature components in the response signal, respectively. The system then constructs these two components into a two-dimensional vector in the complex plane, namely the real-time complex gain vector. The magnitude of the vector This reflects the gain of the system at the probe frequency, while its phase angle... which reflects the phase delay of the system at this frequency point, and the two dimensions together represent the dynamic response characteristics of the controlled object at a specific frequency.

[0025] To use the real-time complex gain vector obtained by the above measurement for the evaluation of the health status of the digital twin model, the system needs to establish a benchmark in advance. In view of this, the method is configured to perform steps a and b above when the controlled air conditioning equipment is in a preset healthy state, for example, after the equipment is commissioned after initial installation, and store the complex gain vector obtained at this time as a benchmark complex gain vector In the subsequent continuous operation monitoring, the system performs step c, that is, compares the real-time calculated complex gain vector with the benchmark complex gain vector to generate a model integrity index A specific calculation procedure is to define the index as the ratio of the two vector modules, that is, For example, in the healthy state, is calibrated to 1.0 (normalized value), and if is measured as 0.8 at a certain time, then The decrease in the index quantitatively indicates that the dynamic response ability of the controlled object has decreased, and the physical reality has deviated from the healthy state initially represented by the digital twin model; further, to distinguish different types of decay modes, the system is also configured with analysis logic for the deviation vector, that is, to calculate a deviation vector formed by subtracting the benchmark complex gain vector from the real-time complex gain vector, and to distinguish energy decay mode and delay decay mode according to the angle between the evolution direction of the deviation vector on the complex plane and the direction of the benchmark complex gain vector. The specific judgment rule is: if the direction of the deviation vector is mainly opposite to the direction of the benchmark vector (the angle is close to 180 degrees), it is determined that the energy decay is mainly reduced, and if the direction of the deviation vector is mainly perpendicular to the direction of the benchmark vector (the angle is close to 90 degrees), it is determined that the delay decay is mainly increased.

[0026] When responding to the sudden online of the cabinet cluster in the data center and other conditions that generate instantaneous heat load, the controller will push its output to the physical upper limit (for example, 100% power), at which time the control loop enters the nonlinear output saturation zone. During this period, the output of the controller loses its association with its input (error signal), and the injected harmonic probe cannot produce a meaningful response; to address this problem and avoid the pollution of invalid data on the diagnosis result, the method of the present application is configured to perform step d, that is, to establish a dual-mode switching procedure. The system continuously monitors the value of the main control signal of the controller, and as soon as it detects that it reaches the preset physical upper or lower limit, it immediately determines that the controller has entered the output saturation state, and immediately suspends the harmonic probe signal the injection of the probe, at the same time, the system uses this saturation event to test the limit performance of the device, i.e. during the saturation state of the controller, a timer is started, and the duration of the saturation state is obtained After the main control signal is out of the saturation state, the probe injection is restored, and the system compares the measured duration with a pre-stored reference saturation duration under the same load impact in a healthy state to generate a limit performance index A physically decaying system, the control margin decreases, and when facing the same thermal load impact, the controller needs longer saturation time to suppress the temperature, so the duration will be longer, and the EPI will be reduced accordingly. This mechanism converts a common monitoring blind area into a process of quantitative evaluation of the reserve capacity of the device.

[0027] To improve the depth of diagnosis, the method of the application can further include analysis of the time series of complex gain vectors; in long-term operation, some slow progressive faults and sudden unstable faults may exhibit similar complex gain vector deviations at a certain moment, but their risk levels are different; to distinguish between them, the system is configured to record a trajectory formed by the real-time complex gain vectors over time and calculate the dynamic characteristics of the trajectory. Specifically, the system obtains the evolution rate of the trajectory by calculating the first-order time derivative of the trajectory, and the modulus quantifies the speed of the current decay, while the system calculates the statistical variance of the trajectory itself within a sliding time window to obtain the stability of the trajectory. A trajectory with high jitter has a higher variance value. Finally, based on a two-dimensional decision matrix, the system generates a state flag representing the risk level of operation according to the combination of evolution rate and stability. For example, low rate and high variance may correspond to intermittent faults such as poor relay contact, while high rate and high variance may indicate an emergency instability fault such as valve piece breakage, thereby providing a basis for hierarchical response for operation and maintenance. The generation of the operation risk state flag relies on a decision logic based on pre-set thresholds. The thresholds here come from the reference data collected under the healthy state of the device, in which the rate threshold is set to five times the maximum value of the modulus of the first-order time derivative of the complex gain vector trajectory measured within one hour under the healthy state, and the stability threshold is set to three times the standard deviation of the trajectory coordinate points within the same period; during monitoring, the system calculates the evolution rate and stability of the current trajectory within a sliding time window in real time, and performs the following judgments: if and , the output risk level is level 1 - normal wear; if but then output risk level 2 - intermittent disturbance; if but then output risk level 3 - accelerated decay; if and then output risk level 4 - instability warning, each level corresponds to a set of predefined operation response procedures, so as to convert continuous dynamic indicators into discrete risk indicators with clear operation direction; in addition, in order to cope with the problem that different physical decay modes may show different sensitivities at different frequency points, the method of the present application can also include an adaptive probe frequency optimization procedure; since the probe with fixed frequency may not be the best observation point for all fault types, the system is configured to enter a diagnostic sweep mode under preset conditions, such as when the MII first falls below a certain threshold or periodic self-checking, in which mode the system continuously changes the frequency of the harmonic probe signal in a diagnostic frequency range with the cutoff frequency of the controlled air conditioning equipment as the lower limit, for example from 5Hz to 15Hz, and simultaneously calculates the real-time complex gain vector corresponding to each frequency, thereby obtaining a gain-frequency response curve, by comparing the real-time curve with the pre-stored reference response curve, the system identifies the frequency point with the largest difference between the two, and defines it as the characteristic frequency under the current state , then the system updates the frequency of the harmonic probe to the characteristic frequency and switches back to the normal single-frequency detection mode, which enables the detection behavior to dynamically focus on the observation window with the largest amount of information.

[0028] On the other hand, in order to cope with the faults that may exist in the air conditioning system characterized by injecting random noise, such as refrigerant two-phase flow or early damage to the fan bearing, which cannot be captured by the synchronous demodulation algorithm based on deterministic signals, the method of the present application can also include a parallel random disturbance analysis branch; in order to open up this monitoring dimension without increasing hardware costs, the system performs parallel processing on the same error signal , one of which enters the aforementioned synchronous demodulation module, and the other first passes through a digital notch filter whose center frequency is set to be the same as the frequency of the current harmonic probe signal, which filters out the actively injected probe signal and its response component, thereby obtaining a noise signal , then the system performs a fast Fourier transform on the noise signal to obtain its power spectrum, and calculates the spectral entropy of the power spectrum, which is a single numerical value that quantifies the degree of disorder of the spectrum, a healthy system has a relatively flat background noise spectrum, and the spectral entropy value is high, while when a random fault source appears, it will form a specific peak or shape on the noise spectrum, resulting in a decrease in the spectral entropy value, by comparing the real-time spectral entropy with the reference spectral entropy, the system can determine whether there is a random disturbance type fault, thereby forming a complementary double-channel monitoring mode with the deterministic fault diagnosis based on the complex gain vector; finally, in order to ensure the long-term accuracy of the entire measurement system, the measurement chain itself needs to be considered, for example, the measurement error that may be introduced by the aging or temperature drift of the ADC / DAC chip and analog circuit inside the controller may be incorrectly attributed to the change of the controlled object; in order to solve this problem, the method of the application can further include a self-calibration procedure, which connects an output path of the control signal directly to an input path of the error signal through an analog switch in a preset calibration mode (for example, when the system is on standby), forming an internal calibration loop, at this time, the measured object is no longer an air conditioning device, but an electrical circuit, the system performs the same steps a to c in this loop as in the normal monitoring, and obtains a calibration complex gain vector, any deviation from the ideal value of the vector, for example, for a direct loop, the ideal value is [1, 0], is attributed to the distortion characteristics of the signal measurement chain itself, the system stores this distortion characteristics as a calibration model, and in the normal monitoring mode, the real-time measured complex gain vector is compensated according to the calibration model, thereby obtaining a purified complex gain vector that excludes the error of the measurement tool itself, and all subsequent diagnoses are based on this data, which ensures the long-term reliability of the diagnosis conclusion.

[0029] Example 1: In a high-density computing data center, a precision air conditioning device runs uninterruptedly throughout the year, and its operating environment has specific requirements for temperature and humidity, when the device is working stably, the temperature sensor reading at the outlet of the device is maintained at a set value , the mean value of the error signal is maintained near zero, at the same time, the surface of the heat exchanger fins is undergoing a gradual dust accumulation process due to electrostatic adsorption effect, which causes a slow decrease in heat exchange efficiency; applying the monitoring method to this device, the system continuously performs online dynamic characteristic calibration when the controller is in the linear working area, and records that the model integrity index of the device has decreased from the initial 1.0 to 0.85 during this period, and no observable deviation has occurred in the physical temperature output; at this time, due to the start of a large-scale data analysis task, the instantaneous heat load of the cabinet cluster increases, and the main control signal of the air conditioning device controller is pushed to the physical upper limit of 100%, and the system enters the output saturation state.

[0030] Upon determining that the controller has entered output saturation, the method immediately pauses the injection of harmonic probe signals and times the duration of the saturation state. After 350 seconds of continuous operation, the controller output exits the saturation state, and the system calculates the duration of this saturation event accordingly. By comparing the baseline saturation time of the device under a healthy state with that stored in the digital twin, when faced with an equivalent load shock. Its value is 210 seconds. By comparison, the system generates a limit performance index. Its value is 0.6; a model integrity index. The continuous decline and the ultimate performance index A level-two early warning, triggered by values ​​falling below a baseline, was sent to the operations and maintenance platform. This warning included the current values ​​of two indices, characterizing the degree of degradation of the controlled object's physical properties and its reduced ability to handle peak loads. Based on this, operations and maintenance personnel scheduled preventative maintenance, during which they discovered dust adhering to the heat exchanger fins. After cleaning the fins and restarting the system, the system collected real-time complex gain vectors. The modulus is recovered to the reference complex gain vector. Consistent level, model integrity index Returning to version 1.0, in a subsequent simulated load shock test, the extreme performance index... It also returned to the baseline level, and the equipment returned to the operating state at the initial calibration time.

[0031] Example 2: The experiment in this example was conducted on a test platform consisting of a precision air conditioning unit with a rated cooling capacity of 25kW, a closed environmental test chamber capable of outputting a continuously adjustable heat load from 0kW to 30kW, and a data acquisition and control system; wherein, the data acquisition and control system includes a precision of A temperature sensor is used to measure the outlet temperature of the air conditioning unit, a power meter is used to monitor the real-time power consumption of the compressor, and a controller that runs the method of the present invention is capable of acquiring error signals in the control loop at a frequency of 200 Hz. To simulate the gradual physical degradation of the heat exchanger of the air conditioning unit due to dust accumulation, a digitally variable baffle with porosity controlled by a computer program is installed at the air inlet of the condenser of the unit, and the decrease in heat exchange efficiency is equivalently simulated by gradually reducing its porosity.

[0032] The experiment was conducted in two groups: a control group relying solely on conventional monitoring parameters and the experimental group applying the method of this invention. Before the experiment, both groups were operated under a constant heat load of 15kW in a healthy state with no obstructions (i.e., 100% heat exchange efficiency) until the outlet air temperature stabilized at [temperature value missing]. For the sample group of this invention, the reference complex gain vector is completed at this stage. calibrated and normalized to 1.0; then, the test enters the accelerated aging simulation phase, the shielding rate of the shielding plate is set to increase by 2% per hour, linearly increasing from 0% to 20%, the whole process lasts for 10 hours, during which the heat load is maintained at 15 kW constant, and the monitoring data of each group are recorded continuously; during the 10-hour test, the recorded outlet air temperature is maintained in the range of in both groups; the compressor power consumption of both groups increases with the increase of the shielding rate, and the total change is less than 5%; the model integrity index calculated by the inventive group decreases with the increase of the shielding rate, when the shielding rate is 4%, the value is 0.95, when the shielding rate is 20%, the value is 0.78; the detailed data are recorded in Table 1.

[0033] Table 1: Comparison table of key parameters during the test.

[0034] ; The test results show that under the condition that the outlet air temperature is maintained stable due to the compensation of the controller, the value of the model integrity index obtained by the method of the present application has a negative correlation with the simulated physical attenuation degree, i.e. the shielding rate; the change of the index reflects the progressive degradation of the health degree of the controlled object which is masked by the operation of the controller.

[0035] Example 3: This example combines Figs. 1 to 3 to explain a digital twin-based air conditioning equipment operation state monitoring method, as shown in Fig. 1 The flow is applied to a closed-loop control system comprising a controller and a controlled air conditioning equipment, and the core is a controller state monitoring link which determines the working mode of the controller in real time and starts the corresponding processing path; when it is determined that the controller is in the linear working area, the flow enters the left branch, first executes the step of injecting a harmonic probe signal, injects a continuous excitation into the control signal channel, then executes the step of synchronously demodulating the error signal to extract the same frequency response component of the probe, and then constructs a real-time complex gain vector to represent the current dynamic characteristics of the controlled object; based on the vector, the system executes two analysis paths in parallel, one is to compare it with the reference vector to quantify the deviation between the model and the reality, thereby generating the model integrity index ​​second, into the complex gain vector trajectory analysis section, by recording the trajectory formed by the vector over time, and further calculating the evolution rate and stability, to quantify the speed and fluctuation of the attenuation process, so as to generate the operation risk state symbol, to distinguish the urgency of the fault development, when the controller state monitoring section determines that the controller enters the output saturation area, the flow enters the right branch, first executes the action of pausing the injection of the probe signal, to avoid introducing noise when the control loop is disconnected, at the same time, the step of obtaining the output saturation duration is executed, the time when the controller is in the limit state is recorded, then the duration is compared with the reference saturation duration, to compare the limit response ability under the healthy state, and finally generate the limit performance index , to evaluate the reserve capacity of the equipment to cope with extreme load.

[0036] As shown in Fig. 2 , the horizontal axis of the figure is time s, and the vertical axis is amplitude, wherein the solid line marked as the probe signal represents a stable, pure sinusoidal excitation, the thin dotted line marked as the error signal represents a signal containing the response of the excitation and superimposed with other noise and interference, and the thick dotted line marked as the demodulated in-phase component represents a direct current component extracted after the error signal is multiplied by the in-phase reference signal and low-pass filtered, the amplitude of the component stably reflects the strength of the in-phase component in the response signal; as shown in Fig. 3 , the process starts from the trajectory recording module, which records the time sequence of the complex gain vector and provides the trajectory data to the deviation analysis module, the deviation analysis module calculates the deviation vector based on the data, and analyzes the included angle and direction to obtain the deviation feature, then the deviation feature is transmitted to the fault classifier, the fault classifier determines the fault type according to the received feature, for example, distinguishes between energy type attenuation or delay type attenuation, and sends the fault type information to the diagnosis report module, at the same time, the deviation analysis module also calculates and provides the evolution rate and stability variance of the trajectory to the risk assessment module, the risk assessment module generates the risk level symbol according to the above, and sends the risk level information to the diagnosis report module, finally, the diagnosis report module integrates the fault type information and the risk level information received, to generate a complete diagnosis report.

[0037] Example 4: Before a newly installed precision air conditioning equipment is put into long-term continuous operation, a set of parameter calibration and system self-calibration procedures are executed, which are used to determine the characteristic frequency of a harmonic probe signal, and to quantify and compensate the signal transmission characteristics of the controller hardware itself; the system is first placed in a diagnostic sweep mode, in which the system continuously changes the frequency of the harmonic probe signal in a diagnostic frequency range of 5Hz to 15Hz with a step of 0.1Hz and at each frequency point, the corresponding real-time complex gain vector is calculated according to steps a to c, thereby obtaining a baseline gain-frequency response curve representing the healthy state of the device; subsequently, the speed of the condenser fan is reduced by 5% through controller instructions to introduce a reproducible physical disturbance, and the above-mentioned sweep frequency process is repeated under this disturbance state to obtain a second response curve; the system further calculates the absolute value of the difference between the two curves at each frequency point, and identifies the frequency point with the largest difference, which is 11.2 Hz in this procedure, and sets it as the characteristic frequency used in subsequent long-term monitoring of the device and the frequency of the harmonic probe signal is updated to this value.

[0038] After the characteristic frequency is determined, the system starts the self-calibration mode; the system disconnects the output path of the control signal from the variable frequency compressor through an internal analog switch, and directly loops back to the input path of the error signal to form an internal calibration loop; in this loop, the system injects the harmonic probe signal at the determined characteristic frequency and performs the same response extraction and characteristic calibration steps to obtain a calibration complex gain vector representing the distortion characteristics of the signal measurement link itself , and the value measured in this calibration is This measurement value quantifies the gain deviation and phase angle deviation introduced by the entire electronic measurement path; after calibration is complete, the system switches to normal monitoring mode and uses the calibration result to compensate subsequent measurements; the compensation algorithm performs dual correction of amplitude and phase on each original real-time complex gain vector measured in normal monitoring mode, specifically by multiplying its modulus by the reciprocal of the modulus of the calibration vector and subtracting the phase angle of the calibration vector from its phase angle to obtain the purified complex gain vector ; by performing this procedure, the monitoring system works at the determined characteristic frequency point, and all subsequent output diagnostic data of the system have excluded the influence of the drift of the controller hardware itself.

[0039] Example 5: After the air conditioning device is first deployed or after major component replacement is completed, a baseline data acquisition procedure is performed, which, after confirming that the device is in a healthy state, applies a stable heat load of 60% of the rated capacity of the device, and after the outlet temperature of the device stabilizes, the system runs continuously for 1 hour and calculates and stores the baseline complex gain vector and the baseline spectral entropy during this period; subsequently, the system performs a load shock test, instantaneously increasing the heat load from 20% to 90% to make the controller enter the output saturation state, and the system records the time length from entering saturation to exiting saturation of the controller and stores this time length value as the baseline saturation time length .

[0040] ​The data used to calibrate the state flag decision threshold is derived from the 1-hour health data collected under the steady thermal load condition. The system analyzes the time series trajectory of the complex gain vector in this period, and calculates the standard deviation of the trajectory position coordinates . The threshold used to determine the stability of the trajectory is set to . The system calculates the maximum value of the health trajectory evolution rate . The threshold used to determine the evolution rate of the trajectory is set to . According to this procedure, the quantitative indicators for determining instability failure and acute decay are determined.

[0041] Example 6: A procedure for constructing an offline diagnostic model is performed on a representative sample machine of a specific type of air conditioning equipment. The procedure first determines the amplitude of the harmonic probe , the steps of which are to place the sample machine in a stable operating condition, start with an initial low amplitude value, and gradually increase the amplitude of the injected probe signal , and simultaneously record the signal standard deviation of the in-phase component and the quadrature component output after demodulation at each amplitude point, as well as the standard deviation of the outlet temperature reading. Through data analysis, the critical amplitude value at which the outlet temperature standard deviation begins to exceed its inherent noise baseline is determined, and the probe amplitude for this type of equipment is set to be lower than this critical value and to obtain a value of 95% of the maximum standard deviation of the and components.

[0042] After the probe amplitude is determined, the response characteristics of different failure modes are calibrated. A micro-metering pump is used to extract refrigerant from the system at a known rate of 2 grams per hour to introduce energy-type decay failure, and the trajectory of the deviation vector is continuously recorded during this process. Then, the system is restored to a healthy state, and a programmable digital time delay is introduced between the controller and the actuator, gradually increasing from 0 ms to 500 ms to introduce delay-type decay failure, and the trajectory of the deviation vector is recorded during this process. Finally, statistical analysis is performed on the two sets of trajectory data collected to determine the angular distribution regions on the complex plane related to energy-type decay and delay-type decay, respectively, and the boundaries of these regions are set as the decision rules for distinguishing between the two types of decay for this type of equipment.

[0043] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0044] Finally, it should be noted that the above examples are merely intended to illustrate the technical solutions of the present application and not to limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the present application.

Claims

1. A digital-twin-based air conditioning equipment operation state monitoring method applied to a closed-loop control system comprising a controller and a controlled air conditioning equipment, characterized in that, The method comprises: Step a, injecting a harmonic probe signal in one control signal channel of the closed-loop control system when the controller is in the linear operating region, the frequency of the harmonic probe signal being higher than the cut-off frequency of the physical output of the controlled air conditioning equipment, and the amplitude of the harmonic probe signal not causing the physical output to exceed the noise fluctuation range of the physical output in normal operation; Step b, extracting in-phase component and quadrature component of the same frequency response signal in the error signal of the closed-loop control system by synchronous demodulation with the frequency of the harmonic probe signal, and constructing a real-time complex gain vector; Step c, comparing the real-time complex gain vector with a reference complex gain vector obtained when the controlled air conditioning equipment is in a preset healthy state to generate a model integrity index; Step d, determining that the controller enters the output saturation state when the controller enters the output saturation region, and immediately suspending the injection of the harmonic probe signal, and obtaining the duration of the output saturation state during the controller is in the output saturation state, and comparing the duration with a reference saturation duration to generate a limit performance index.

2. The air conditioning equipment operation state monitoring method based on digital twinning according to claim 1, characterized in that, The method further comprises: recording a trajectory formed by the real-time complex gain vector over time; calculating the first time derivative of the trajectory to obtain the evolution rate of the trajectory, and calculating the variance of the trajectory within a time window to obtain the stability of the trajectory; and generating a state flag representing the risk level of the air conditioning equipment based on the evolution rate and the stability. 3.The air conditioning equipment operation state monitoring method based on digital twinning of claim 1, wherein, The method further comprises: continuously changing the frequency of the harmonic probe signal in a diagnostic frequency range with the cut-off frequency of the controlled air conditioning equipment as the lower limit, and synchronously calculating the real-time complex gain vector corresponding to each frequency to obtain a gain-frequency response curve; identifying a frequency with the maximum difference value as a characteristic frequency based on the difference between the gain-frequency response curve and a reference response curve; and updating the frequency of the harmonic probe signal to the characteristic frequency.

4. The air conditioning equipment operation state monitoring method based on digital twinning according to claim 1, characterized in that, The generation of the model integrity index specifically comprises calculating the ratio of the modulus of the real-time complex gain vector to the modulus of the reference complex gain vector.

5. The digital-twin-based air conditioning equipment operation state monitoring method according to claim 1, characterized by, The step of comparing the real-time complex gain vector with the reference complex gain vector further comprises: calculating a deviation vector formed by subtracting the reference complex gain vector from the real-time complex gain vector; and distinguishing the energy-type attenuation mode and the delay-type attenuation mode according to the included angle between the evolution direction of the deviation vector on the complex plane and the direction of the reference complex gain vector.

6. The air conditioning equipment operation state monitoring method based on digital twinning according to claim 1, characterized in that, The method further comprises: performing notch filtering processing on the error signal, the center frequency of the notch filtering processing being the same as the frequency of the harmonic probe signal, to filter out the harmonic probe signal and its response signal component, thereby obtaining a companion noise signal; calculating the spectral entropy of the power spectrum of the companion noise signal; and comparing the spectral entropy with a reference spectral entropy to determine whether a random disturbance type fault exists.

7. The digital-twin-based air conditioning equipment operation state monitoring method according to claim 1, characterized by, The method further comprises: in a calibration mode, directly connecting an output path of the control signal to an input path of the error signal to form an internal calibration loop; in the internal calibration loop, performing steps a-c to obtain a calibration complex gain vector representing the distortion characteristics of the signal measurement link itself; and in a normal monitoring mode, compensating the real-time complex gain vector based on the calibration complex gain vector. 8.The air conditioning equipment operation state monitoring method based on digital twinning of claim 1, wherein, The frequency of the harmonic probe signal is set to be in a range from five times to fifty times higher than a cut-off frequency of a physical output of the controlled air conditioning device. 9.The air conditioning equipment operation state monitoring method based on digital twinning of claim 1, wherein, The synchronous demodulation is specifically point-by-point multiplication operation of the error signal with the harmonic probe signal and with a signal orthogonal to the harmonic probe signal by 90 degrees in phase, and low-pass filtering processing of the results of the multiplication operation. 10.The air conditioning equipment operation state monitoring method based on digital twinning of claim 1, wherein, The control signal is a signal output by the controller to a variable frequency compressor, and the error signal is a difference between a temperature set value at an input end of the controller and a temperature measurement value of a temperature sensor.

Images