Method, system, equipment and medium for diagnosing sensor faults in air handling units

Abstract

The present invention discloses a method, system, equipment and medium for diagnosing sensor faults in an air handling unit, comprising: obtaining real-time data of sensors in the air handling unit to be predicted; using the real-time data of sensors in the air handling unit to be predicted as input of a pre-built air handling unit sensor minor fault diagnosis model, and outputting a sensor fault diagnosis result in the air handling unit to be predicted; wherein, a particle swarm algorithm is used to optimize the kernel parameters of a kernel principal component correlation analysis algorithm to obtain an optimized kernel principal component correlation analysis algorithm; the Euclidean distance in the kernel principal component analysis algorithm is replaced by the correlation distance to optimize the Gaussian radial basis kernel function in the kernel principal component analysis algorithm to obtain the kernel principal component correlation analysis algorithm; the present invention can diagnose and detect minor sensor faults in the air handling unit, obtain the fault area and the fault sensor, and has a high fault detection efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of air conditioning systems, and in particular relates to a method, system, equipment and medium for diagnosing sensor faults in an air handling unit. Background Art

[0002] Heating, ventilation, and air conditioning (HVAC) systems, the primary equipment for ensuring indoor thermal comfort, account for approximately 50% of a building's total energy consumption during operation. In commercial and mixed-use buildings, air conditioning energy consumption can reach over 65%. Therefore, improving HVAC system energy efficiency is key to saving building energy. To address this issue, various optimization strategies have been applied to HVAC systems, such as optimization, monitoring, and automated diagnosis. Research has shown that the use of fault detection and diagnosis technologies in commercial buildings can reduce HVAC system energy consumption by 20%-30%.

[0003] Sensors, such as temperature sensors, humidity sensors, pressure sensors, and flow sensors, are widely used components in HVAC systems as signal sources for intelligent control. Therefore, ensuring the accuracy of the measurement data from these sensors is crucial for achieving intelligent control of building energy systems. However, sensor failures are inevitable during HVAC system operation. When a sensor used for control fails, the control system may issue erroneous instructions, causing the actuator to perform incorrect operations. Although such failures can be resolved by manually adjusting the control strategy, as sensor failures continue to worsen, the energy consumption cost of the HVAC system will increase. At the same time, as the structural complexity of HVAC systems in commercial buildings gradually increases, the maintenance cost of sensors also increases.

[0004] Therefore, the research on sensor fault diagnosis based on HVAC system is of great significance to building energy conservation and environmental protection; the air conditioning system is usually composed of wind system and water system; among them, the wind system (i.e. air handling unit) is the core component of the air conditioning system; the sensors in the air handling unit are in a high-temperature working environment for a long time, which can easily lead to sensor failure, and minor faults with low degree of failure are very easy to occur; relatively speaking, the occurrence of sensor failure is not overnight, it has a potential and gradual process; therefore, if the fault signal is captured in time at the early stage of the sensor failure, it will greatly reduce the impact of the fault on the system operation. It can be seen that research on the detection of minor faults is very necessary.

[0005] Air handling units are key components of air conditioning systems and have complex control systems. Failure of air handling unit sensors can lead to a decrease in indoor comfort and even damage the equipment. Currently, existing sensor fault diagnosis methods are mostly limited to sensor faults with higher fault severity. The diagnosis and detection methods for minor sensor faults with lower fault severity are still insufficient. Since the characteristics of the above-mentioned minor faults are not very obvious, it is still difficult to extract the fault characteristics in the early stage of the fault. Therefore, there is an urgent need to provide a method for diagnosing minor faults of air handling unit sensors. Summary of the Invention

[0006] In response to the technical problems existing in the prior art, the present invention provides a sensor fault diagnosis method, system, equipment and medium in an air handling unit to solve the technical problem that the existing sensor fault diagnosis methods are mostly limited to sensor faults with higher fault levels and cannot meet the diagnosis and detection of minor sensor faults with lower fault levels.

[0007] In order to achieve the above object, the technical solution adopted by the present invention is:

[0008] The present invention provides a method for diagnosing sensor faults in an air handling unit, comprising:

[0009] Obtain real-time sensor data in the air handling unit to be predicted;

[0010] Using the real-time sensor data in the air handling unit to be predicted as the input of a pre-built air handling unit sensor minor fault diagnosis model, and outputting a sensor fault diagnosis result in the air handling unit to be predicted;

[0011] The construction process of the pre-built air handling unit sensor minor fault diagnosis model is as follows:

[0012] The kernel parameters of the kernel principal component correlation analysis algorithm are optimized using a particle swarm algorithm to obtain an optimized kernel principal component correlation analysis algorithm; wherein the Euclidean distance in the kernel principal component analysis algorithm is replaced by the correlation distance to optimize the Gaussian radial basis kernel function in the kernel principal component analysis algorithm to obtain the kernel principal component correlation analysis algorithm;

[0013] The optimized kernel principal component correlation analysis algorithm is trained using historical sensor data of a normal air handling unit to obtain the pre-built air handling unit sensor minor fault diagnosis model; wherein the historical sensor data of the normal air handling unit includes normal data and preset fault data.

[0014] Furthermore, the normal data includes historical data of sensors in a normal air handling unit; wherein, the historical data of sensors in a normal air handling unit includes chilled water valve opening, fresh air temperature, fresh air humidity, supply air temperature, supply air humidity, return air temperature and return air humidity.

[0015] Furthermore, the preset fault data is collected by presetting fault forms of sensors in normal air handling units; wherein the preset fault forms include 5%-20% drift fault forms or 5%-20% deviation fault forms.

[0016] Furthermore, the real-time sensor data in the air handling unit to be predicted is used as the input of the pre-built air handling unit sensor minor fault diagnosis model, and the process of outputting the sensor fault diagnosis result in the air handling unit to be predicted is as follows:

[0017] According to the real-time data of the sensors in the air handling unit to be predicted, the initial data matrix X is constructed. N×m ;

[0018] Combined with the principle that information entropy remains unchanged before and after linear change, the initial data matrix X N×m Perform linear transformation to obtain homogeneous data matrix Z N×m ;

[0019] The optimized kernel principal component correlation analysis algorithm is used to analyze the initial data matrix X N×m and the homogeneous data matrix Z N×m Perform dimensionality reduction processing to obtain the initial data dimensionality reduction result Y N×m And the homogeneous data dimensionality reduction result Y′ N×m ;

[0020] Calculate the initial data dimensionality reduction result Y N×m And the homogeneous data dimensionality reduction result Y′ N×m The correlation distance D i ;

[0021] The relevant distance D i The preset minimum correlation distance D min For comparison, if the relevant distance D i Less than the preset minimum correlation distance D min , the principal component contribution rate is used to extract the fault characteristics, and the output is the sensor fault diagnosis result in the air handling unit to be predicted.

[0022] Furthermore, the initial data matrix X N×m for:

[0023]

[0024] Where N is the number of real-time sensor data, m is the number of sensors; x′ Nm The real-time data of the Nth sensor of the mth sensor;

[0025] The homogeneous data matrix Z N×m for:

[0026]

[0027]

[0028] Among them, Z ij ′ is the result of linear transformation of the real-time data of the i-th sensor of the j-th sensor; x i ' j is the real-time data of the i-th sensor of the j-th sensor; x i ' (j+1) is the real-time data of the i-th sensor of the j+1-th sensor; x i ' m is the real-time data of the i-th sensor of the m-th sensor; x i ′1 is the real-time data of the ith sensor of the first sensor.

[0029] Furthermore, if the relevant distance D i Greater than or equal to the preset minimum correlation distance D min , then using the particle swarm algorithm to update the kernel parameters of the kernel principal component correlation analysis algorithm to obtain an updated kernel principal component correlation analysis algorithm;

[0030] Using the updated kernel principal component correlation analysis algorithm, the initial data matrix X N×m and the homogeneous data matrix Z N×m Re-perform dimensionality reduction and calculate related distances.

[0031] Furthermore, when the principal component contribution rate is used to extract fault features, the contribution rate CPV(i) of the i-th principal component is:

[0032]

[0033] Among them, λ i is the i-th principal component of the principal component analysis method, that is, the eigenvalue of the i-th kernel function; n is the total number of principal components of the principal component analysis method;

[0034] The cumulative contribution rate CPV of the first p principal components is:

[0035]

[0036] The present invention also provides a sensor fault diagnosis system in an air handling unit, comprising:

[0037] A data acquisition module is used to obtain real-time data from sensors in the air handling unit to be predicted;

[0038] a diagnostic output module, configured to use the real-time sensor data within the air handling unit to be predicted as input to a pre-built air handling unit sensor minor fault diagnosis model, and output a diagnostic result of the sensor fault within the air handling unit to be predicted;

[0039] The construction process of the pre-built air handling unit sensor minor fault diagnosis model is as follows:

[0040] The kernel parameters of the kernel principal component correlation analysis algorithm are optimized using a particle swarm algorithm to obtain an optimized kernel principal component correlation analysis algorithm; wherein the Euclidean distance in the kernel principal component analysis algorithm is replaced by the correlation distance to optimize the Gaussian radial basis kernel function in the kernel principal component analysis algorithm to obtain the kernel principal component correlation analysis algorithm;

[0041] The optimized kernel principal component correlation analysis algorithm is trained using historical sensor data of a normal air handling unit to obtain the pre-built air handling unit sensor minor fault diagnosis model; wherein the historical sensor data of the normal air handling unit includes normal data and preset fault data.

[0042] The present invention also provides a sensor fault diagnosis device in an air handling unit, comprising:

[0043] memory for storing computer programs;

[0044] A processor is used to implement the steps of the method for diagnosing sensor faults in an air handling unit when executing the computer program.

[0045] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for diagnosing sensor faults in an air handling unit are implemented.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention provides a method and system for diagnosing sensor faults in an air handling unit. By adding preset fault data to the historical sensor data of a normal air handling unit, various fault states that may occur in the air handling unit are simulated; the kernel parameters in the KPCCA method are optimized using the PSO algorithm, and then the optimized KPCCA algorithm is used to extract features from the sensor data to obtain specific fault types, effectively solving the problem that minor sensor faults are difficult to identify during the diagnosis process, effectively improving the efficiency of diagnosing minor sensor faults in the air handling unit, and improving the diagnosis success rate; the present invention does not require manual identification of suspicious targets, and automatically performs fault detection through the overall operating status of the data; it can diagnose and detect minor sensor faults in the air handling unit, and obtain the fault area and the faulty sensor; the fault detection efficiency is high, and the minor fault type can be accurately obtained, effectively filling the gap in the diagnosis of minor sensor faults in the air handling unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of the method for diagnosing faults of sensors in an air handling unit according to the present invention;

[0049] Figure 2 Flowchart of the particle swarm algorithm in the present invention;

[0050] Figure 3 Detection results of the existing KPCA and the PSO-KPCCA in the embodiment under different degrees of drift faults;

[0051] Figure 4 1. A diagram showing classification of 5% sensor drift fault data of a test sample by conventional KPCA and PSO-KPCCA in an embodiment;

[0052] Figure 5 10% sensor drift fault data of the test sample is classified by the existing KPCA and the PSO-KPCCA in the embodiment;

[0053] Figure 6 15% sensor drift fault data of the test sample is classified by the existing KPCA and the PSO-KPCCA in the embodiment;

[0054] Figure 7 20% sensor drift fault data of the test sample is classified by the existing KPCA and the PSO-KPCCA in the embodiment;

[0055] Figure 8 Detection results of the existing KPCA and the PSO-KPCCA in the embodiment under different degrees of bias faults;

[0056] Figure 9 1. A diagram showing classification of 5% sensor bias fault data of a test sample by conventional KPCA and PSO-KPCCA in an embodiment;

[0057] Figure 10 10% sensor bias fault data of the test sample is classified by the existing KPCA and the PSO-KPCCA in the embodiment;

[0058] Figure 11 15% sensor bias fault data of the test sample is classified by the existing KPCA and the PSO-KPCCA in the embodiment;

[0059] Figure 12 Graph showing classification results of 20% sensor bias fault data of test samples by conventional KPCA and PSO-KPCCA in the embodiment. DETAILED DESCRIPTION

[0060] In order to make the technical problems, technical solutions and beneficial effects solved by the present invention more clearly understood, the present invention is further described in detail in the following specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0061] As attached Figure 1 As shown, the present invention provides a method for diagnosing sensor faults in an air handling unit, comprising the following steps:

[0062] Step 1: Obtain real-time sensor data in the air handling unit to be predicted.

[0063] Step 2: Construct a minor fault diagnosis model for the air handling unit sensor to obtain a pre-constructed minor fault diagnosis model for the air handling unit sensor. The construction process is as follows:

[0064] The kernel parameters of the kernel principal component correlation analysis algorithm are optimized using a particle swarm algorithm to obtain an optimized kernel principal component correlation analysis algorithm; wherein the Euclidean distance in the kernel principal component analysis algorithm is replaced by the correlation distance to optimize the Gaussian radial basis kernel function in the kernel principal component analysis algorithm to obtain the kernel principal component correlation analysis algorithm;

[0065] The optimized kernel principal component correlation analysis algorithm is trained using historical sensor data of a normal air handling unit to obtain the pre-built air handling unit sensor minor fault diagnosis model.

[0066] In the present invention, the sensor historical data of the normal air handling unit includes normal data and preset fault data; wherein, the normal data includes the sensor historical data in the normal air handling unit; wherein, the sensor historical data in the normal air handling unit includes the opening of the chilled water valve, fresh air temperature, fresh air humidity, supply air temperature, supply air humidity, return air temperature and return air humidity; the preset fault data is collected after the preset fault form of the sensor in the normal air handling unit; wherein, the preset fault form includes a drift fault form of 5%-20% or a deviation fault form of 5%-20%.

[0067] Step 3: Using the real-time sensor data in the air handling unit to be predicted as input to a pre-built air handling unit sensor minor fault diagnosis model, and outputting a sensor fault diagnosis result in the air handling unit to be predicted;

[0068] The specific process is as follows:

[0069] Step 31: Obtain the initial data matrix X according to the real-time sensor data in the air handling unit to be predicted. N×m ;

[0070] Among them, the initial data matrix X N×m for

[0071]

[0072] Where N is the number of real-time sensor data, m is the number of sensors; x′ Nm is the real-time data of the Nth sensor of the mth sensor.

[0073] Step 32: Based on the principle that information entropy remains unchanged before and after linear changes, the initial data matrix X N×m Perform linear transformation to obtain homogeneous data matrix Z N×m Specifically, based on the principle that information entropy remains unchanged before and after linear changes, the initial data matrix X N×m Perform linear transformation to obtain a matrix containing the original data information, that is, to obtain the homogeneous data matrix Z N×m ;

[0074] Wherein, the homogeneous data matrix is ​​Z N×m :

[0075]

[0076]

[0077] Among them, Z ij ′ is the result of linear transformation of the real-time data of the i-th sensor of the j-th sensor; xi ' j is the real-time data of the i-th sensor of the j-th sensor; x i ' (j+1) is the real-time data of the i-th sensor of the j+1-th sensor; x i ' m is the real-time data of the i-th sensor of the m-th sensor; x i ′1 is the real-time data of the ith sensor of the first sensor.

[0078] Step 33: Utilize the optimized kernel principal component correlation analysis algorithm to analyze the initial data matrix X N×m and the homogeneous data matrix Z N×m Perform dimensionality reduction processing to obtain the initial data dimensionality reduction result Y N×m And the homogeneous data dimensionality reduction result Y′ N×m .

[0079] Step 34: Calculate the initial data dimensionality reduction result Y N×m And the homogeneous data dimensionality reduction result Y′ N×m The correlation distance D i .

[0080] Step 35: The relevant distance D i The preset minimum correlation distance D min For comparison, if the relevant distance D i Less than the preset minimum correlation distance D min , then the principal component contribution rate is used to extract the fault characteristics, and the output is the sensor fault diagnosis result in the air handling unit to be predicted;

[0081] Among them, when the principal component contribution rate is used to extract fault features, the contribution rate CPV(i) of the i-th principal component is:

[0082]

[0083] Among them, λ i is the i-th principal component of the principal component analysis method, that is, the eigenvalue of the i-th kernel function; n is the total number of principal components of the principal component analysis method;

[0084] The cumulative contribution rate CPV of the first p principal components is:

[0085]

[0086] Step 36: If the relevant distance D i Greater than or equal to the preset minimum correlation distance D min, then the particle swarm algorithm is used to update the kernel parameters of the kernel principal component correlation analysis algorithm; the updated kernel principal component correlation analysis algorithm is obtained, and the process returns to steps 33-34 to re-perform the dimensionality reduction process and the operation of calculating the correlation distance; wherein, the specific process of using the particle swarm algorithm to update the kernel parameters of the kernel principal component correlation analysis algorithm is as shown in the attached figure. Figure 2 shown.

[0087] The present invention relates to the fault diagnosis of sensors in air handling units. Based on the kernel principal component analysis (KPCA) method and according to the information entropy value invariance theorem, the traditional Euclidean distance is replaced by the correlation distance to optimize the parameters of the Gaussian radial basis kernel function. The kernel principal component correlation analysis (KPCCA) method is proposed for fault feature extraction and diagnosis. The kernel parameters in the KPCCA algorithm are optimized through the particle swarm optimization (PSO) method to complete the feature extraction of minor sensor faults for diagnosis.

[0088] The present invention also provides a sensor fault diagnosis system in an air handling unit, including a data acquisition module and a diagnostic output module; the data acquisition module is used to obtain real-time data of sensors in the air handling unit to be predicted; the diagnostic output module is used to use the real-time data of sensors in the air handling unit to be predicted as input to a pre-built air handling unit sensor minor fault diagnosis model, and output the sensor fault diagnosis results in the air handling unit to be predicted.

[0089] The present invention also provides a device for diagnosing faults of sensors in an air handling unit, comprising: a memory for storing a computer program; and a processor for implementing the steps of a method for diagnosing faults of sensors in an air handling unit when executing the computer program.

[0090] When the processor executes the computer program, it implements the steps of the above-mentioned method for diagnosing sensor faults in the air handling unit, for example: obtaining real-time data of sensors in the air handling unit to be predicted; using the real-time data of sensors in the air handling unit to be predicted as input of a pre-built air handling unit sensor minor fault diagnosis model, and outputting the sensor fault diagnosis results in the air handling unit to be predicted.

[0091] Alternatively, when the processor executes the computer program, it implements the functions of each module in the above-mentioned system, for example: a data acquisition module, used to obtain real-time data of sensors in the air handling unit to be predicted; a diagnostic output module, used to use the real-time data of sensors in the air handling unit to be predicted as input of a pre-built air handling unit sensor minor fault diagnosis model, and output the sensor fault diagnosis result in the air handling unit to be predicted.

[0092] Exemplarily, the computer program can be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments that can complete preset functions, and the instruction segments are used to describe the execution process of the computer program in the sensor fault diagnosis device in the air handling unit. For example, the computer program can be divided into a data acquisition module and a diagnostic output module; the specific functions of each module are as follows: a data acquisition module, used to obtain real-time data of sensors in the air handling unit to be predicted; a diagnostic output module, used to use the real-time data of sensors in the air handling unit to be predicted as the input of a pre-built air handling unit sensor minor fault diagnosis model, and output the sensor fault diagnosis results in the air handling unit to be predicted.

[0093] The air handling unit internal sensor fault diagnosis device can be a computing device such as a desktop computer, laptop, PDA, or cloud server. The air handling unit internal sensor fault diagnosis device can include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above is an example of an air handling unit internal sensor fault diagnosis device and does not constitute a limitation on the air handling unit internal sensor fault diagnosis device. The air handling unit internal sensor fault diagnosis device can include more components than those described above, or a combination of certain components, or different components. For example, the air handling unit internal sensor fault diagnosis device can also include input and output devices, network access devices, buses, etc.

[0094] The processor may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may be any conventional processor, etc. The processor is the control center of the sensor fault diagnosis device in the air handling unit, and uses various interfaces and lines to connect various parts of the sensor fault diagnosis device in the entire air handling unit.

[0095] The memory can be used to store the computer program and / or module, and the processor implements various functions of the sensor fault diagnosis device in the air handling unit by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory.

[0096] The memory may mainly include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, the memory may include a high-speed random access memory and may also include a non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0097] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method for diagnosing sensor faults in an air handling unit are implemented.

[0098] If the module / unit integrated in the air handling unit internal sensor fault diagnosis system is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0099] Based on this understanding, the present invention implements all or part of the process of the above-mentioned method for diagnosing sensor faults in an air handling unit by instructing related hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of the above-mentioned method for diagnosing sensor faults in an air handling unit. The computer program includes computer program code, which can be in source code form, object code form, executable file, or a preset intermediate form.

[0100] The computer-readable storage medium may include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0101] It should be noted that the content contained in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practices in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practices, computer-readable storage media do not include electrical carrier signals and telecommunication signals.

[0102] Example

[0103] This embodiment provides a method for diagnosing sensor faults in an air handling unit, comprising the following steps:

[0104] Step 1. Place normal sensors and faulty sensors at sensor points of normal air handling units at different time periods according to preset fault modes; in this embodiment, the preset fault modes include drift and deviation fault modes of 5%, 10%, 15%, and 20%; collect monitoring data of the normal sensors and faulty sensors to obtain sensor historical data of the normal air handling unit; wherein, the sensor historical data of the normal air handling unit includes normal data and preset fault data; the preset fault data is obtained by collecting monitoring data of faulty sensors placed in the normal air handling unit; the sensor historical data of the normal air handling unit includes chilled water valve opening, fresh air temperature, fresh air humidity, supply air temperature, supply air humidity, return air temperature, and return air humidity.

[0105] Typically, the characteristic parameters of each HVAC system have different sensitivities to different types and degrees of faults under different operating conditions, and the correlations between different characteristic parameters are also different under different conditions. Therefore, using the kernel principal component analysis method to extract the fault characteristics of sensors in air handling units can reduce the correlation between features, reduce redundancy, increase sensitivity, and reduce dimensionality.

[0106] Assuming that the research object contains m sensors, the sensor data set of the tth measurement is expressed as:

[0107]

[0108] Among them, Y (t) is the sensor historical data set of the tth measurement; The sensor historical data of the mth sensor measured at the tth time.

[0109] When the historical data of N measurements of the above m sensors are expressed in matrix form, the sensor historical data matrix of N measurements of all sensors is:

[0110]

[0111] Where Y is the sensor historical data matrix of N measurements of all sensors; y Nmis the sensor history data of the Nth measurement of the mth sensor; therefore, the N sampling data of the sensor measurement value under normal operating conditions can be constructed as Y∈R N×m The matrix of .

[0112] Step 2: performing normalization processing on the sensor historical data of the normal air handling unit according to a normalization processing formula to eliminate the influence of different sensor dimensions on the fault feature extraction, and obtaining a normalized matrix of the historical sampling data;

[0113] The normalization formula is:

[0114]

[0115] Among them, x ij The normalized result of the sensor historical data measured by the jth sensor for the i-th time; y ij The sensor history data of the j-th sensor measured at the i-th time; is the average value of the historical sensor data of the jth sensor; is the standard deviation of the sensor history data of the jth sensor.

[0116] The standardized matrix of the historical sampling data is:

[0117]

[0118] Wherein, X is the standardized matrix of the historical sampling data.

[0119] Step 3: Based on the principle that information entropy remains unchanged before and after linear changes, a linear transformation is performed on the standardized matrix X of the historical sampled data to obtain a historical homogeneous data matrix Z. Specifically, based on the principle that information entropy remains unchanged before and after linear changes, a linear transformation is performed on the standardized matrix X of the historical sampled data to obtain a matrix containing the original historical data information, that is, the historical homogeneous data matrix Z is obtained.

[0120] Wherein, the historical homogeneous data matrix is ​​Z:

[0121]

[0122]

[0123] Among them, Z ij is the result of linear transformation of the historical data of the i-th sensor of the j-th sensor; x ij is the historical data of the i-th sensor of the j-th sensor; x i(j+1) is the historical data of the i-th sensor of the j+1-th sensor; x imis the historical data of the i-th sensor of the m-th sensor; x i1 is the historical data of the ith sensor of the 1st sensor.

[0124] Step 4: Combined with the correlation method, the particle swarm algorithm is used to optimize the kernel parameters of the Gaussian radial basis kernel function to obtain the optimized kernel parameters. The specific process is as follows:

[0125] Set the kernel parameter σ i The empirical value interval [σ min ,σ max ];

[0126] The kernel parameter σ is optimized by using particle swarm optimization. i Make updates;

[0127] The updated kernel parameters are used as the kernel parameters of the Gaussian radial basis kernel function, and the KPCA method is used to perform dimensionality reduction processing on the standardized matrix X of the historical sampling data and the historical homogeneous data matrix Z, respectively, to obtain the dimensionality reduction result Y of the historical sampling data. N×r And the dimensionality reduction result of historical homogeneous data Y′ N×r ;

[0128] Calculate the initial data dimensionality reduction result Y N×m And the homogeneous data dimensionality reduction result Y′ N×m The correlation distance D i , in order to measure the deviation of the dimensionality reduction result; wherein, the correlation distance D is calculated according to the following formula i ;

[0129] D i =1-ρ i

[0130]

[0131] When the relevant distance D i Less than the preset minimum correlation distance D min , then take the minimum correlation distance D i The corresponding kernel parameters are the optimal kernel parameters, that is, the optimized kernel parameters are obtained; otherwise, i=i+1, and the operation of step 4 is repeated.

[0132] Step 5: Use the optimized kernel parameters to construct a Gaussian radial basis kernel function; wherein the Gaussian radial basis kernel function is:

[0133]

[0134] Among them, σ is the optimized kernel parameter.

[0135] Step 6: Use the following formula to normalize the Gaussian radial basis kernel function to obtain a normalized kernel function;

[0136]

[0137]

[0138] in, is the normalized Gaussian radial basis kernel function.

[0139] Step 7: Calculate the eigenvalue and eigenvector of the normalized Gaussian radial basis kernel function using the following formula:

[0140]

[0141] Where λ is the eigenvalue of the normalized Gaussian radial basis kernel function; is the eigenvector of the normalized Gaussian radial basis kernel function.

[0142] Step 8: Use the following formula to normalize the feature connections;

[0143]

[0144] Step 9: Determine the number of pivots using the pivot contribution rate

[0145] Research has found that if too many pivots are selected, the residual subspace will contain less information; if too few pivots are selected, information redundancy will occur, making it impossible to extract effective information. Therefore, the number of pivots directly affects feature extraction. In this embodiment, fault features are extracted based on the Cumulative Percent Variance (CPV).

[0146] Among them, when the principal component contribution rate is used to extract fault features, the contribution rate CPV(i) of the i-th principal component is:

[0147]

[0148] Among them, λ i is the i-th principal component of the principal component analysis method, that is, the eigenvalue of the i-th kernel function; n is the total number of principal components of the principal component analysis method;

[0149] The cumulative contribution rate CPV of the first p principal components is:

[0150]

[0151] The contribution rate of cumulative variance reflects the ability of the selected principal component to represent the overall data. If the CPV value is too small, it indicates that the selected principal component contains less information overall, resulting in inaccurate research based on it and failure to reflect relevant content truthfully. If the CPV value is too large, the information covered by the selected principal component is too complex and contains a large amount of secondary or useless information, which leads to the neglect of the main information. Based on past experience, the CPV value is taken at 90%.

[0152] Step 10: Determine the extracted features

[0153] The eigenvectors corresponding to the first p rows are selected from the homogeneous data matrix, and the matrix is ​​constructed after normalization to obtain the reduced-dimensional samples, thereby realizing the extraction of fault features and obtaining a trained PSO-KPCCA fault diagnosis model, that is, a minor fault diagnosis model for air handling unit sensors.

[0154] Step 11: Acquire real-time sensor data in the air handling unit to be predicted.

[0155] Step 12: Using the real-time sensor data in the air handling unit to be predicted as the input of a pre-built air handling unit sensor minor fault diagnosis model, and outputting a sensor fault diagnosis result in the air handling unit to be predicted.

[0156] Specific test instructions:

[0157] Taking the fault diagnosis process of the air handling unit of a certain air conditioning system as an example, the collected sensor data is summer operating condition data.

[0158] In this embodiment, a total of 1000 sets of sensor data of normal chillers at different time periods are collected, including normal data and eight preset fault data; wherein, 300 sets of sensor data are collected under each preset fault mode; the above data are processed using the existing KPCA method verified by the European method and the air handling unit sensor fault diagnosis method (PSO-KPCCA) described in this embodiment, and the diagnostic results using the existing KPCA method and the air handling unit sensor fault diagnosis method described in this embodiment are respectively obtained, as follows:

[0159] Utilizing historical sensor data from normal air handling units, this embodiment primarily employs kernel principal component correlation analysis (KPCCA) for fault feature extraction and diagnosis. The kernel parameters in the KPCCA algorithm are optimized using a particle swarm optimization (PSO) algorithm to diagnose minor sensor faults. At sensor locations of normal air handling units at different time periods, normal sensors or faulty sensors are placed according to preset fault patterns, including 5%, 10%, 15%, and 20% drift and deviation faults. Data from normal or faulty sensors at different time periods are collected to obtain real-time sensor data from the air handling units.

[0160] In this embodiment, a total of eight preset fault modes are set; sensor data under normal conditions and at different time periods of the eight preset fault modes are used as experiments to obtain real-time collection of sensor parameters in the air-conditioning heat exchanger, and obtain real-time sensor data of the normal air handling unit.

[0161] The obtained chilled water valve opening sensor information, fresh air temperature sensor information, fresh air humidity sensor information, supply air humidity sensor information, supply air temperature sensor information, return air temperature sensor information, and return air humidity sensor information are then used to construct a sensor data matrix in chronological order using the sensor data at the same time in the sample, and normalized. The mathematical expression of the sensor data matrix is ​​as follows: according to the preset fault data form, the normalized sensor data matrix is ​​divided into two groups in chronological order to obtain a training data group and a test data group; each training data group and test data group contains normal data and various preset fault data; the training data group is used as the training sample of the PSO-KPCCA model to obtain a trained fault diagnosis model; the trained fault diagnosis model is a PSO-KPCCA fault feature extraction and diagnosis model with trained parameters.

[0162] The test data group is used as a test sample for the trained PSO-KPCCA fault feature extraction and diagnosis model, and the parameters of the trained PSO-KPCCA fault feature extraction and diagnosis model are evaluated and adjusted. After finding the optimal kernel function value, the iteration is stopped to obtain the PSO-KPCCA model with the optimal parameters, that is, the fault feature extraction and diagnosis model.

[0163] From the attached Figure 3-7It can be seen that the sensor fault diagnosis method for the air handling unit described in this embodiment has good diagnostic capabilities. The fault diagnosis accuracy rates under the four drift fault groupings are 65.67%, 67.67%, 68.67% and 70% respectively. Compared with the existing PSO-KPCCA method that uses Euclidean distance to optimize kernel parameters, the accuracy of the correlation method is improved by 10.67%, 11.00%, 9.67% and 9.67% respectively under drift faults of 5%, 10%, 15% and 20%. The overall drift fault diagnosis success rate reaches 68.1%, which effectively improves the accuracy of the diagnosis results of minor faults.

[0164] From the attached Figure 8-12 It can be seen that the method for diagnosing faults in air handling units described in this embodiment has good diagnostic capabilities, and the fault diagnosis accuracy under the four levels of deviation fault grouping is 99.33%. Compared with the existing KPCA method that uses Euclidean distance to optimize kernel parameters, the fault diagnosis accuracy is improved by about 2% at a 5% deviation fault level, and the accuracy is close at other fault levels.

[0165] The description of the relevant parts of the sensor fault diagnosis system, device and computer-readable storage medium in an air handling unit provided in this embodiment can be found in the detailed description of the corresponding parts of the sensor fault diagnosis method in an air handling unit described in this embodiment, and will not be repeated here.

[0166] The present invention avoids the current problem that the diagnosis of air handling system sensor faults is limited to fault detection using neural networks or clustering methods; the fault diagnosis process does not require manual identification of suspicious targets, nor does it require detection of suspicious targets, and can automatically detect faults through the overall operating status of sensor data; the present invention can extract features of sensor faults of lower severity and diagnose them, thereby detecting minor sensor faults and finding the fault area and the faulty sensor; the PSO-KPCCA algorithm proposed in the present invention improves the fault detection rate compared to the KPCA algorithm optimized using the traditional Euclidean method, and adopts the concept of feature extraction to greatly reduce the probability of confusion in the diagnosis of minor faults that are difficult to distinguish.

[0167] The above embodiment is only one of the implementation methods that can realize the technical solution of the present invention. The scope of protection claimed by the present invention is not limited only to this embodiment, but also includes changes, replacements and other implementation methods that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention.

Claims

1. A method for diagnosing sensor faults in an air handling unit, characterized in that: include: Obtain real-time sensor data in the air handling unit to be predicted; Using the real-time sensor data in the air handling unit to be predicted as the input of a pre-built air handling unit sensor minor fault diagnosis model, and outputting a sensor fault diagnosis result in the air handling unit to be predicted; The construction process of the pre-built air handling unit sensor minor fault diagnosis model is as follows: The kernel parameters of the kernel principal component correlation analysis algorithm are optimized using a particle swarm algorithm to obtain an optimized kernel principal component correlation analysis algorithm; wherein the Euclidean distance in the kernel principal component analysis algorithm is replaced by the correlation distance to optimize the Gaussian radial basis kernel function in the kernel principal component analysis algorithm to obtain the kernel principal component correlation analysis algorithm; The optimized kernel principal component correlation analysis algorithm is trained using historical sensor data of a normal air handling unit to obtain the pre-built air handling unit sensor minor fault diagnosis model; wherein the historical sensor data of the normal air handling unit includes normal data and preset fault data.

2. A method for diagnosing sensor faults in an air handling unit according to claim 1, characterized in that: The normal data includes historical sensor data in a normal air handling unit; wherein, the historical sensor data in the normal air handling unit includes chilled water valve opening, fresh air temperature, fresh air humidity, supply air temperature, supply air humidity, return air temperature and return air humidity.

3. A method for diagnosing sensor faults in an air handling unit according to claim 1, characterized in that: The preset fault data is collected by presetting fault forms of sensors in normal air handling units; wherein the preset fault forms include 5%-20% drift fault forms or 5%-20% deviation fault forms.

4. A method for diagnosing sensor faults in an air handling unit according to claim 1, characterized in that: The process of using the real-time sensor data in the air handling unit to be predicted as the input of the pre-built air handling unit sensor minor fault diagnosis model and outputting the sensor fault diagnosis result in the air handling unit to be predicted is as follows: According to the real-time data of the sensors in the air handling unit to be predicted, the initial data matrix X is constructed. N×m ; Combined with the principle that information entropy remains unchanged before and after linear change, the initial data matrix X N×m Perform linear transformation to obtain homogeneous data matrix Z N×m ; The optimized kernel principal component correlation analysis algorithm is used to analyze the initial data matrix X N×m and the homogeneous data matrix Z N×m Perform dimensionality reduction processing to obtain the initial data dimensionality reduction result Y N×m And the homogeneous data dimensionality reduction result Y′ N×m ; Calculate the initial data dimensionality reduction result Y N×m And the homogeneous data dimensionality reduction result Y′ N×m The correlation distance D i ; The relevant distance D i The preset minimum correlation distance D min For comparison, if the relevant distance D i Less than the preset minimum correlation distance D min , the principal component contribution rate is used to extract the fault characteristics, and the output is the sensor fault diagnosis result in the air handling unit to be predicted.

5. A method for diagnosing sensor faults in an air handling unit according to claim 4, characterized in that: The initial data matrix X N×m for: Where N is the number of real-time sensor data, m is the number of sensors; x′ Nm The real-time data of the Nth sensor of the mth sensor; The homogeneous data matrix Z N×m for: Among them, Z ij ′ is the result of linear transformation of the real-time data of the i-th sensor of the j-th sensor; x i ' j is the real-time data of the i-th sensor of the j-th sensor; x i ' (j+1) is the real-time data of the i-th sensor of the j+1-th sensor; x i ' m is the real-time data of the i-th sensor of the m-th sensor; x i ′1 is the real-time data of the ith sensor of the first sensor.

6. A method for diagnosing sensor faults in an air handling unit according to claim 4, characterized in that: If the relevant distance D i Greater than or equal to the preset minimum correlation distance D min , then using the particle swarm algorithm to update the kernel parameters of the kernel principal component correlation analysis algorithm to obtain an updated kernel principal component correlation analysis algorithm; Using the updated kernel principal component correlation analysis algorithm, the initial data matrix X N×m and the homogeneous data matrix Z N×m Re-perform dimensionality reduction and calculate related distances.

7. A method for diagnosing sensor faults in an air handling unit according to claim 4, characterized in that: When the principal component contribution rate is used to extract fault features, the contribution rate CPV(i) of the i-th principal component is: Among them, λ i is the i-th principal component of the principal component analysis method, that is, the eigenvalue of the i-th kernel function; n is the total number of principal components of the principal component analysis method; The cumulative contribution rate CPV of the first p principal components is:

8. A sensor fault diagnosis system in an air handling unit, characterized in that: include: A data acquisition module is used to obtain real-time data from sensors in the air handling unit to be predicted; a diagnostic output module, configured to use the real-time sensor data within the air handling unit to be predicted as input to a pre-built air handling unit sensor minor fault diagnosis model, and output a diagnostic result of the sensor fault within the air handling unit to be predicted; The construction process of the pre-built air handling unit sensor minor fault diagnosis model is as follows: The kernel parameters of the kernel principal component correlation analysis algorithm are optimized using a particle swarm algorithm to obtain an optimized kernel principal component correlation analysis algorithm; wherein the Euclidean distance in the kernel principal component analysis algorithm is replaced by the correlation distance to optimize the Gaussian radial basis kernel function in the kernel principal component analysis algorithm to obtain the kernel principal component correlation analysis algorithm; The optimized kernel principal component correlation analysis algorithm is trained using historical sensor data of a normal air handling unit to obtain the pre-built air handling unit sensor minor fault diagnosis model; wherein the historical sensor data of the normal air handling unit includes normal data and preset fault data.

9. A sensor fault diagnosis device in an air handling unit, characterized in that: include: memory for storing computer programs; A processor is configured to implement the steps of the method for diagnosing a fault of a sensor in an air handling unit according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for diagnosing faults of sensors in an air handling unit according to any one of claims 1 to 7 are implemented.

Images