Heating ventilation air conditioner fault detection method and system based on condition driving and hybrid detection

Through conditional drive and hybrid detection methods, the spatiotemporal characteristics of the HVAC system are extracted and combined with nonlinear and linear detection, which solves the problem of non-stationary operation and feature interleaving of the system under different operating conditions, and improves the accuracy and efficiency of fault detection. .

CN118912639AActive Publication Date: 2024-11-08SHANDONG JIANZHU UNIV
View PDF 6 Cites 0 Cited by

Patent Information

Application Number
CN202410952637.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-08
Estimated Expiration
2044-07-16

AI Technical Summary

Technical Problem

The existing HVAC fault detection methods fail to effectively consider the large-scale non-stationary characteristics of the system under different steady-state operating conditions, resulting in low detection efficiency and poor accuracy.

Method used

Using a method based on conditional drive and mixed detection, the time slice data is converted into conditional slice data by selecting indoor temperature, air supply volume, and air duct pressure as indicator variables, and the characteristics are extracted using the spatiotemporal variational autocoding model, combining with nonlinear typical Mixed detection of variables and linear typical variables to establish a fault detection model.

Benefits of technology

It improves the accuracy and efficiency of HVAC fault detection, can effectively deal with the system's non-stationary operation and linear-nonlinear characteristics mixed problems under different working conditions, and achieves accurate fault detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118912639A_ABST
    Figure CN118912639A_ABST
Patent Text Reader

Abstract

The invention discloses a heating ventilation air conditioner fault detection method and system based on condition driving and hybrid detection, and the method comprises the steps: selecting the indoor temperature, air output and air duct pressure of a heating ventilation air conditioner system in normal operation as the indication variables of condition driving, and converting time slice data into condition slice data; performing feature extraction on each condition slice by using a space-time variation self-coding model to obtain a space-time variation self-coding similarity coefficient; performing similarity analysis by using a space-time variation self-coding similarity coefficient, and clustering the condition slices to the same working condition; calculating a standard control limit corresponding to each type of working conditions by adopting a non-linear canonical variable and linear canonical variable mixed detection mode; the method comprises the following steps: acquiring online data, obtaining a working condition class to which the online measurement data belongs, obtaining an online control limit of the online measurement data, comparing the online control limit with a standard control limit of the working condition class to which the online measurement data belongs, and obtaining a classification result of whether the online measurement data has a fault or not.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of HVAC fault detection, and in particular to a HVAC fault detection method and system based on conditional drive and hybrid detection. Background Art

[0002] HVAC system refers to the system or related equipment responsible for indoor heating, ventilation and air conditioning. HVAC system can control the temperature and humidity of the air and improve the indoor comfort. It is an important part of medium and large industrial buildings or office buildings. During the long-term operation of the air conditioning system, due to the degradation of component performance and untimely maintenance, various faults such as sensor drift, damper and other actuators stuck, and pipe scaling will occur. These faults will directly affect the environmental comfort and increase energy consumption. Therefore, it is necessary to find a HVAC fault detection method with high detection efficiency and accurate detection results.

[0003] In the actual operation of the HVAC system, affected by various factors such as indoor temperature, air volume, duct pressure and human control, the HVAC system will operate under different steady-state conditions and switch between different operating states. Therefore, the HVAC system exhibits a wide range of non-stationary characteristics during use, including different steady states and transition states. For non-stationary processes, the process operation state changes frequently, resulting in frequent changes in the variance, mean, dynamic relationship and other data characteristics of the process variables along the time direction. At the same time, there is also the problem of coexistence of linear and nonlinear characteristics in the operation of the HVAC system. For example, there are linear characteristics such as indoor temperature and time, air flow rate and fan speed in the operation of the HVAC system. At the same time, there are nonlinear characteristics such as turbulent velocity and fan speed, thermal radiation heat transfer rate and indoor temperature difference in the operation of the air conditioner. In the HVAC system, linear characteristics and nonlinear characteristics are usually intertwined and affect each other. Therefore, for the monitoring, control and optimization of the system, it is necessary to comprehensively consider linear and nonlinear characteristics. Existing HVAC fault detection methods generally do not consider the above characteristics, and there are the following problems:

[0004] (1) In the actual operation of HVAC systems, changes in process operating conditions will be affected by multiple indicator variables. Due to the frequent changes in operating conditions, these variables exhibit large-scale non-stationary characteristics. The existing single-modal or multi-modal research methods are time-driven. This traditional time axis analysis method often analyzes the temporal change pattern and uses clustering methods to reveal changes in process operating conditions. However, due to the frequent state switching of HVAC systems and the different target states of switching, this brings great difficulties to modeling and understanding.

[0005] (2) When dividing the working conditions, it is necessary to extract the key features of the data for cluster analysis. However, the existing clustering methods generally do not consider the correlation information and dynamic characteristics between the various variables in the HVAC system process data, and the data feature extraction is insufficient.

[0006] (3) Since the linear and nonlinear characteristics of the actual operation process of the HVAC system are mixed, a single nonlinear model is no longer the best choice in this case. Therefore, hybrid linear-nonlinear modeling can provide a feasible alternative method for mining process data characteristics. Summary of the invention

[0007] In order to solve the deficiencies of the prior art, the present invention provides a HVAC fault detection method and system based on conditional drive and hybrid detection;

[0008] On the one hand, a HVAC fault detection method based on conditional drive and hybrid detection is provided, including:

[0009] (1) Select the three variables of indoor temperature, air supply volume, and duct pressure of a normally operating HVAC system as condition-driven indicator variables, and convert the time slice data into condition slice data;

[0010] (2) Using the spatiotemporal variational autoencoder model to extract features on each conditional slice, the spatiotemporal variational autoencoder similarity coefficient is obtained; then the spatiotemporal variational autoencoder similarity coefficient is used to perform similarity analysis and cluster the conditional slices into the same working condition;

[0011] (3) Using a mixed detection method of nonlinear typical variables and linear typical variables, the standard control limits corresponding to each type of operating condition are calculated;

[0012] (4) Obtain online measurement data of the HVAC system, determine the operating condition category to which the online measurement data belongs based on the same method as (1) to (2), determine the online control limit of the online measurement data based on the same method as (3), compare the online control limit with the standard control limit of the operating condition category to which the online measurement data belongs, and obtain a classification result of whether the online measurement data is faulty.

[0013] On the other hand, a HVAC fault detection system based on condition-driven and hybrid detection is provided, including:

[0014] A conversion module is configured to: select three variables of indoor temperature, air supply volume, and air duct pressure of a normally operating HVAC system as condition-driven indicator variables, and convert the time slice data into condition slice data;

[0015] A feature extraction module is configured to: perform feature extraction on each conditional slice using a spatiotemporal variational autoencoder model to obtain a spatiotemporal variational autoencoder similarity coefficient; then perform similarity analysis using the spatiotemporal variational autoencoder similarity coefficient to cluster the conditional slices into the same working condition;

[0016] A hybrid detection module is configured to: calculate the standard control limit corresponding to each type of working condition by hybrid detection of nonlinear typical variables and linear typical variables;

[0017] The classification output module is configured to: obtain online measurement data of the HVAC system, derive the operating condition class to which the online measurement data belongs based on the same method as the conversion module and the feature extraction module, derive the online control limit of the online measurement data based on the same method as the hybrid detection module, compare the online control limit with the standard control limit of the operating condition class to which the online measurement data belongs, and obtain a classification result of whether the online measurement data is faulty.

[0018] On the other hand, there is also provided an electronic device, comprising:

[0019] a memory for non-transitory storage of computer-readable instructions; and

[0020] a processor for executing the computer readable instructions,

[0021] When the computer-readable instructions are executed by the processor, the method described in the first aspect is executed.

[0022] On the other hand, a storage medium is provided, which non-temporarily stores computer-readable instructions, wherein when the non-temporary computer-readable instructions are executed by a computer, the instructions of the method described in the first aspect are executed.

[0023] On the other hand, a computer program product is provided, comprising a computer program, wherein the computer program is used to implement the method described in the first aspect when running on one or more processors.

[0024] The above technical solution has the following advantages or beneficial effects:

[0025] In view of the problem of large-scale non-stationary characteristics in HVAC systems, we found that although the operating conditions vary with the time axis with uncertainty, under the same operating conditions, the process characteristics are similar to a large extent and may follow a certain relationship, that is, the process characteristics of the HVAC system change with certain regularity along the condition direction. Therefore, the present invention uses a condition-driven method to reconstruct the three-dimensional data of the HVAC system. Since variables such as indoor temperature, air supply volume, and duct pressure are the most intuitive manifestations of the HVAC system adjustment process, the three variables of indoor temperature, air supply volume, and duct pressure are selected as condition-driven indicator variables, and the time slice data is converted into condition slice data to realize the reconstruction of the three-dimensional time-varying data of the HVAC system from the time axis to the condition axis. The present invention uses a method based on multivariable condition-driven to reconstruct the time-varying three-dimensional data of the HVAC system, and selects the indicator variable as the condition-driven. By reconstructing the time-varying three-dimensional data from the time axis to the condition axis, the problem of large-scale non-stationary operation of the HVAC system between batches and within batches can be effectively solved, and the accuracy of fault detection can be effectively improved.

[0026] Aiming at the correlation and dynamic characteristics of data between multiple conditional slices, in order to fully extract the process characteristics of data and accurately divide multiple working conditions, a spatiotemporal variational autoencoder (ST-VAE) is used to extract features on each conditional slice of the HVAC system. ST-VAE reduces the dimension of the data to achieve the purpose of extracting the characteristic values ​​of the conditional slices, and then uses the spatiotemporal variational autoencoder similarity coefficient to perform similarity analysis, clustering the conditional slices with high similarity to the same working condition, and realizing the accurate division of the multi-condition model. The process characteristics of the conditional slices under the same working condition are similar, and the process characteristics of the conditional slices under different working conditions are different. Fault detection models are established for different working conditions respectively, which can realize the accurate detection of HVAC faults. The present invention uses a method based on spatiotemporal variational autoencoder to cluster conditional slice data with large similarity, and encodes the conditional data into a hidden space through an encoder. In order to improve the interpretability and stability of the autoencoder, the latent vector is modeled as a Gaussian distribution, and the decoder is used to reconstruct the data, and the spatiotemporal variational autoencoder similarity coefficient is used to calculate the similarity of the latent vector, and the conditional slices with high similarity are clustered to the same working condition.

[0027] Aiming at the problem that the data of each working condition after the HVAC system is divided have mixed linear and nonlinear characteristics, the present invention adopts the hybrid detection method of kernel canonical variable dissimilarity analysis (KCVDA) and canonical variable dissimilarity analysis (CVDA) to model the divided working conditions. First, KCVDA is applied to the collected process data to extract nonlinear canonical variables (NCVs) in the main subspace of KCVDA, and then CVDA is applied in the residual subspace of KCVDA to extract linear canonical variables (CVs) from the main subspace of CVDA. Fault monitoring statistic T2 and Q are used to detect the residual matrix of the residual subspace of the linear-nonlinear mixed typical variables and CVDA respectively. Since the prior distribution information of the nonlinear principal component characteristics extracted from the normal data of each sub-operating condition is unknown, the kernel density estimation is used to determine the control limit of the monitoring statistic. The present invention combines the kernel typical variable dissimilarity analysis and the typical variable dissimilarity analysis algorithm to establish a fault detection model based on nonlinear-linear mixed monitoring in each operating condition, and extracts independent linear typical variables and nonlinear typical variables from each operating condition data set. With the help of this method, the problem of mixed linear and nonlinear characteristics can be effectively handled, and it has better accuracy than a single linear or nonlinear model. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0029] Figure 1 is a system structure diagram of Embodiment 1;

[0030] Figure 2 This is a flow chart of obtaining conditional slices in Example 1;

[0031] Figure 3 This is a structural diagram of the spatiotemporal variational autoencoder of Example 1;

[0032] Figure 4 Schematic diagram of hybrid detection statistical modeling in Example 1. DETAILED DESCRIPTION

[0033] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0034] Embodiment 1

[0035] This embodiment provides a HVAC fault detection method based on conditional drive and hybrid detection;

[0036] like Figure 1 As shown, the HVAC fault detection method based on conditional drive and hybrid detection includes:

[0037] S101: Selecting three variables, namely, indoor temperature, air supply volume, and air duct pressure of a normally operating HVAC system as condition-driven indicator variables, and converting time slice data into condition slice data;

[0038] S102: extracting features using a spatiotemporal variational autoencoder model on each conditional slice to obtain a spatiotemporal variational autoencoder similarity coefficient; then performing similarity analysis using the spatiotemporal variational autoencoder similarity coefficient to cluster the conditional slices into the same working condition;

[0039] S103: using a mixed detection method of nonlinear typical variables and linear typical variables to calculate the standard control limit corresponding to each type of working condition;

[0040] S104: Obtain online measurement data of the HVAC system, obtain the operating condition category of the online measurement data in the same way as S101 to S102, obtain the online control limit of the online measurement data in the same way as S103, compare the online control limit with the standard control limit of the operating condition category to which the online measurement data belongs, and obtain a classification result of whether the online measurement data is faulty.

[0041] Further, the step S101: selecting the three variables of indoor temperature, air volume and air duct pressure of a normally operating HVAC system as condition-driven indicator variables, and converting the time slice data into condition slice data, specifically includes:

[0042] S101-1: In the time-varying three-dimensional normal operation data set X(I×K×J) of the HVAC system, a slice operation is performed perpendicular to the time axis to obtain two-dimensional time slices X at different times k (I×J); where k represents the sampling time, k=1, 2,…, K, I represents the batch, and J represents the variable;

[0043] S101-2: In the kth two-dimensional time slice X k In the variable J of (I×J), calculate the mean of the indoor temperature variables of different batches at the same time Average air volume Mean pipeline pressure Indicator variable, where k = 1, 2, ..., K, the formula is as follows:

[0044]

[0045] Among them, T k (I) represents the indoor temperature value of the Ith batch at the kth moment, V k (I) represents the air supply volume value of the Ith batch at the kth time, P k (I) represents the duct pressure value of the Ith batch at k moments.

[0046] In order to better realize the working condition division, the multi-variable is converted into a weighted variable. The weighted variable combines multiple indicator variables in a weighted manner. First, After standardization, the weights of the three variables are calculated to achieve the synthetic representation of multiple indicator variables. The weighted variable formula is as follows:

[0047]

[0048] S101-3: Slice the weighted variable condition according to the weighted variable increment [x min ,x min+ ,…,x i ,x max ] to obtain the weighted variable x on the conditional axis. k Slice X according to the conditions that change from small to large m (I×J); where X m (I×J), m=1,2,…,K represents the mth conditional slice.

[0049] Furthermore, after S101-2 and before S101-3, the following is further included:

[0050] S101-23: Use the chimpanzee optimization algorithm to optimize the three parameters α, β, and γ in the formula:

[0051] S101-231: Chimpanzees select their initial position by random selection;

[0052] S101-232: Position update step: The position is updated by updating the position formula to select different parameters. The updated position formula is as follows:

[0053]

[0054] in, is the position of individual i at time step t. is the coefficient of the individual with the lowest objective function value, is the position of another randomly selected individual in the population. R1 and R2 are random numbers, usually following a normal distribution with a mean of 0 and a standard deviation of 1.

[0055] S101-233: Weight update step: In order to find the optimal weight, an objective function is introduced to update the weight: Among them, y k Represents the true value.

[0056] S101-233: Repeat the position updating step and the weight updating step until a predetermined number of iterations is reached or a stopping condition is met.

[0057] It should be understood that manually selecting the three weight parameters is not necessarily the optimal solution. In order to achieve the optimal performance of the algorithm, the discrete chimpanzee optimization algorithm based on the nearest neighbor traction operator is used to optimize the three parameters α, β, and γ in the formula. This method simulates the sociological behavior and learning mechanism in the chimpanzee group and combines the nearest neighbor traction operator to achieve adaptive adjustment of parameters. To minimize the weighted variable x k The difference from the actual situation is the optimization target. Through loop iteration, the weight parameters α, β, and γ are finally optimized.

[0058] It should be understood that the HVAC system refers to: a system for providing and adjusting the comfort of the internal environment of a building, which integrates the three functions of heating, ventilation and air conditioning to ensure that the temperature, humidity and air quality of the indoor environment meet the requirements of human comfort. The heating system increases the indoor temperature by heating air or water, the ventilation system improves the air quality by introducing fresh air and exhausting dirty air, and the air conditioning system reduces the indoor temperature by cooling or dehumidifying the air. An efficient HVAC system can not only provide a comfortable indoor environment, but also achieve energy conservation and emission reduction, and reduce operating costs. Its goals include temperature control, humidity control, air quality and air circulation.

[0059] It should be understood that the time-varying three-dimensional normal operation data set includes: Time-varying: the data changes over time, that is, the data set contains multiple data samples collected at different time points. The time-varying data set reflects the dynamic changes of the system or process over time, and can capture trends and patterns that change over time. Three-dimensional: The data set has three dimensions, which can be understood as each data sample has three different attributes at a certain point in time. In the present invention, these three attributes are time K, batch I, and variable J.

[0060] It should be understood that the two-dimensional time slice includes: in the time-varying three-dimensional data set, each moment K contains two attributes, batch and variable, so this moment K is a two-dimensional data, so the moment K is regarded as a two-dimensional time slice of the time-varying three-dimensional data set.

[0061] It should be understood that, considering the operation status of the HVAC system for several consecutive days, the operation status of a working day is defined as an operation batch I, and each operation batch contains two-dimensional data of time K and variable J. The normal operating condition data of different operation batches are arranged together to form a time-varying three-dimensional data set of the HVAC system. There are batch-wise time-varying dynamic characteristics between the operation data of different batches in several consecutive days; and on each day, the data in a batch also presents a large range of non-stationary operation characteristics in time.

[0062] In view of the fact that the operation data of the HVAC system for consecutive days has time-varying dynamic characteristics in batches, and the operation data of each day also has a large range of non-stationary operation in time, a multi-variable conditional driving method is used to reconstruct the three-dimensional time-varying normal operation data of the HVAC system from the time axis to the condition axis, which can effectively deal with the non-stationary operation problem of frequent switching of multiple working conditions in the actual operation of the HVAC system. The data processing flow is as follows Figure 2 shown.

[0063] It should be understood that the time-varying three-dimensional data set of the HVAC system is reconstructed from the time axis to the condition axis to obtain condition slices arranged in order from small to large according to the weighted variables.

[0064] Furthermore, if Figure 3 As shown, the S102: extracting features on each conditional slice to obtain a similarity coefficient; then performing similarity analysis using the similarity coefficient to cluster the conditional slices into the same working condition, specifically includes:

[0065] Each conditional slice is standardized, and the trained spatiotemporal variational autoencoder model is used to extract features on each standardized conditional slice to obtain the spatiotemporal variational autoencoder similarity coefficient;

[0066] The trained spatiotemporal variational autoencoder model comprises: an encoder, a hidden layer and a decoder connected in sequence; the encoder comprises: a first convolutional layer, a temporal convolutional network (TCN) and a second convolutional layer connected in sequence; the decoder comprises: a third convolutional layer, a graph convolutional neural network (GCN) and a fourth convolutional layer connected in sequence;

[0067] The similarity analysis is performed using the spatiotemporal variational autoencoder similarity coefficient to cluster the conditional slices into the same working condition.

[0068] Furthermore, the standardization process for each conditional slice includes:

[0069] Conditional slicing X m (I×J) is standardized to obtain the standard condition slice, the formula is as follows:

[0070]

[0071] in, represents the mth standard condition slice, Represents the conditional slice X m The mean value of D(X m ) represents the conditional slice X m The variance of .

[0072] Furthermore, the trained spatiotemporal variational autoencoder model includes:

[0073] Encoder, normalizes the conditional slices The conditional step size is embedded through 1×1 convolution, and multiple layers of one-dimensional expanded convolution are used to gradually learn the dependencies between batches and compress the data length. The expanded convolution 1×k operation is performed using the filter f(i) with a kernel size of K and an expansion factor of d, which is defined as:

[0074]

[0075] Where t(s) represents the conditional step size s of the compressed sequence t. The encoder compresses the original data of length T into a hidden state of length 1 through multiple one-dimensional expansion convolutions, and then outputs the hidden state h of the inherent pattern through 1×1 convolution.

[0076] Based on the idea of ​​variational inference, the hidden layer is used to fit μ and σ, as shown in the following formula:

[0077] μ=TCN μ (X) (3)

[0078] logσ 2 =TCN σ (X) (4)

[0079] Where μ is the mean vector matrix and σ is the variance matrix; TCN μ (X) is responsible for extracting the features of the input data X and generating a mean vector matrix μ, TCN σ (X) Generates a log-variance matrix logσ for calculating the distribution of latent variables 2 .

[0080] The decoder, for the hidden state z(s), uses the filter f(i) with a kernel size of 1×k and an expansion factor of d to perform a transposed expansion convolution operation, which is defined as:

[0081]

[0082] Furthermore, the training process of the trained spatiotemporal variational autoencoder model includes:

[0083] Constructing a training set, wherein the training set is time-varying three-dimensional data under normal operating conditions of the HVAC system;

[0084] The training set is used to train the spatiotemporal variational autoencoder model. When the loss function value of the spatiotemporal variational autoencoder model no longer decreases, the training is stopped to obtain the trained spatiotemporal variational autoencoder model.

[0085] Encode the original data X into the hidden layer, and then decode the hidden layer into the reconstructed data In order to minimize the reconstruction error of the spatiotemporal variational autoencoder model, the loss function is designed as follows:

[0086]

[0087] in, is the mean square error loss function, KL[q(·)||p(·)] is the Kullback-Leibler divergence between q(·) and p(·), where the Kullback-Leibler divergence is an indicator used to measure the difference between two probability distributions.

[0088] Furthermore, the feature extraction is performed on each normalized conditional slice using the trained spatiotemporal variational autoencoder model to obtain the spatiotemporal variational autoencoder similarity coefficient, including:

[0089] Through the trained spatiotemporal variational autoencoder model, the loss function L is minimized. Then the latent vector of the hidden layer is output through the trained spatiotemporal variational autoencoder model:

[0090] Z=q(X|N(μ,σ 2 )) (9)

[0091] The latent vector of the hidden layer is the spatiotemporal variational autoencoder similarity coefficient.

[0092] Furthermore, the similarity analysis is performed using the spatiotemporal variational autoencoder similarity coefficient to cluster the conditional slices into the same working condition, including:

[0093] After outputting the latent vector matrix, the mth conditional slice X is calculated using the spatiotemporal variational autoencoder similarity coefficient. m and the m+1th conditional slice X m+1 The similarity ρ between two conditional slices is as follows:

[0094]

[0095] In the formula, z m 、z m+1 They are conditional slices X m , X m+1 The latent vector of z m 、z m+1 The angle between

[0096] If the similarity coefficient of adjacent multivariate conditional slices is 0.5≤ρ≤1, it indicates that the feature similarity of the two multivariate conditional slices is high, and the two multivariate conditional slices are clustered under the same condition.

[0097] If the spatiotemporal variational autoencoder similarity coefficient of adjacent multivariate conditional slices is 0≤ρ≤0.5, it indicates that the feature similarity of the two adjacent multivariate conditional slices is low, and the mth multivariate conditional slice is taken as the last multivariate conditional slice of the previous working condition, and the m+1th multivariate conditional slice is taken as the first multivariate conditional slice of the next working condition. The spatiotemporal variational autoencoder model is used to achieve accurate division of the Z sub-conditional models.

[0098] It should be understood that the spatiotemporal variational autoencoder similarity coefficient actually calculates the mean of the cosine value of the angle between the potential vectors of the two types of data, which should be in the interval [0,1]. The similarity coefficient between conditional slices of the same type will be close to 1, which is greater than the similarity coefficient between conditional slices of different types.

[0099] It should be understood that, considering that the existing clustering models generally do not consider the correlation information and dynamic characteristics between the variables in the HVAC system data, the problem of inaccurate division of multiple operating conditions of the HVAC system is caused. The present invention adopts a multi-condition division module based on spatiotemporal variational autoencoder to conditional slice X m (I×J) is used for feature extraction, and then the spatiotemporal variational autoencoder similarity coefficient is used to cluster the conditional slices with great similarity into the same working condition, so as to obtain data of multiple working conditions with large differences, and realize the accurate division of the multi-working condition model.

[0100] It should be understood that first, autoencoding is performed within the divided conditional slices to extract eigenvalues, and then the spatiotemporal variational autoencoding similarity coefficient is used to perform similarity analysis, clustering the conditional slices with high correlation into the same working condition, thereby achieving accurate division of multi-condition data sets based on conditional slices.

[0101] Furthermore, the S103: using a mixed detection method of nonlinear typical variables and linear typical variables to calculate the standard control limits corresponding to each type of working condition includes:

[0102] For each type of working condition, a corresponding fault detection model based on nonlinear-linear hybrid monitoring is established;

[0103] S103-1: Based on the nonlinear-linear hybrid monitoring fault detection model, the process data of the current working condition data set is subjected to kernel canonical variable dissimilarity analysis: nonlinear canonical variables are extracted in the main subspace of the kernel canonical variable dissimilarity analysis; canonical variable dissimilarity analysis is applied in the residual subspace of the kernel canonical variable dissimilarity analysis; linear canonical variables and residual variables are extracted from the main subspace of the canonical variable dissimilarity analysis;

[0104] S103-2: Calculate a first statistic based on the nonlinear canonical variables and the linear canonical variables;

[0105] S103-3: Calculate a second statistic based on the residual variable;

[0106] S103-4: Based on the first statistic and the second statistic, a kernel density estimation algorithm is used to calculate the standard control limits corresponding to each type of operating condition.

[0107] Furthermore, the S103-1: based on the nonlinear-linear hybrid monitoring fault detection model, performing a kernel canonical variable dissimilarity analysis on the process data of the current operating condition data set: extracting nonlinear canonical variables in the main subspace of the kernel canonical variable dissimilarity analysis; applying canonical variable dissimilarity analysis in the residual subspace of the kernel canonical variable dissimilarity analysis; extracting linear canonical variables and residual variables from the main subspace of the canonical variable dissimilarity analysis, specifically includes:

[0108] set up is the process input at time k, is the process output at time k, and defines the past data vector containing past inputs and outputs and a future data vector containing future outputs

[0109]

[0110]

[0111] where p and f are the past and future data vectors z respectively p (k) and y f (k) Time lag.

[0112] Assume a training data set with N measurements u(k) and y(k), k = 1, 2, ..., N collected under normal operating conditions, for all k ∈ [p + 1, p + M], by z p (k) and y f (k) Construct the past and future Hankel matrices Z p and Y f ,as follows:

[0113]

[0114] Kernel Canonical Variable Dissimilarity Analysis (KCVDA) uses kernel functions to map original data into a high-dimensional feature space. This mapping captures nonlinear relationships into features, so Kernel Canonical Variable Dissimilarity Analysis (KCVDA) is used to extract nonlinear features.

[0115] Assume that the input vector The nonlinear transformation Φ1: and the future input vector The nonlinear transformation Φ2: Where Zp ,Y f is a data sequence matrix containing past and future information, H i It is called feature space.

[0116] Then, the goal is to find f1∈H1 and f2∈H2 such that f1(Z p )=<Φ1(Z p ),f1> and f2(Y f )=<Φ2(Y f ),f2> has the largest correlation. f1 and f2 are located in linear spaces S1 and S2 respectively, which are formed by the images of Φ1 and Φ2. Therefore:

[0117]

[0118] Using the kernel matrix K p ,K f Mapping the original data into a high-dimensional feature space, K p ,K f is defined as follows:

[0119]

[0120] The KCVDA uses the canonical variable dissimilarity (KCVD) between the canonical variables of the past and future projections. The KCVD at time k is s (k) is defined as follows:

[0121] d s (k) = AK f -BK p (14)

[0122] Among them, A∈(β1,β2,...,β s ) consists of the first s singular values, B∈(β s+1 ,β s+2 ,...,β M ) consists of the last s to m singular values.

[0123] Canonical variable dissimilarity analysis (CVDA) is applied in the residual subspace of kernel canonical variable dissimilarity analysis (KCVDA). Canonical variable dissimilarity analysis (CVDA) focuses on finding the most discriminative linear projection direction between data sets, thereby using canonical variable dissimilarity analysis (CVDA) to extract linear features. Canonical variable dissimilarity analysis (CVDA) uses canonical variable dissimilarity (CVD) between canonical variables in the residual subspace. CVDd at time k n (k) is defined as follows:

[0124] d n (k) = Ln y f (k)-S n J n z p (k) (15)

[0125] in, U n Contains the first n columns of U, S n The n largest singular values ​​S n =diag(λ1,λ2,...,λ n )composition.

[0126] It should be understood that process input data refers to input information received by a system or process. These data can be any information that affects the behavior or output of the system, and are usually provided when the system or process starts executing.

[0127] It should be understood that process output data refers to the results or outputs generated by a system or process after execution. These data usually reflect the response or processing results of the system or process to the input data. Process output data can be any information or signal generated by the system, such as measurement values, simulation results, predicted values, state variables, etc. These data are the reflection of the input data after processing, and are used to analyze system behavior, evaluate performance or make further decisions.

[0128] It should be understood that the future data vector of future output refers to a set of f output data generated by the HVAC system in the future that is predicted forward.

[0129] It should be understood that a Hankel matrix is ​​a square matrix in which the elements on each anti-diagonal are the same. A Hankel matrix is ​​a symmetric matrix in which each element a ij Depends only on the sum of i+j.

[0130] It should be understood that Kernel Canonical Variate Dissimilarity Analysis (KCVDA) is a statistical technique that combines kernel methods and canonical variate analysis, aiming to identify and separate different classes or groups in data.

[0131] It should be understood that the canonical variable dissimilarity is to find the difference between linear data.

[0132] Further, the S103-2: calculating a first statistic based on the nonlinear canonical variable and the linear canonical variable includes:

[0133] Calculate HotellingT at time k 2 The statistical fault detection quantity T 2 :

[0134] T 2 (k) = d(k)d(k) T (16)

[0135] Among them, T 2 The measured state vector d(k) changes, d(k) contains the nonlinear canonical variable dissimilarity d s (k) and the linear canonical variable dissimilarity d n (k).

[0136] Furthermore, the S103-3: calculating a second statistic based on the residual variable includes:

[0137] Assume y r is the data in the CVDA residual space, y r The calculation formula is as follows:

[0138]

[0139] Introduce the statistic Q:

[0140] Q(k)=y r (k)y r (k) T (18)

[0141] Here, Q(k) measures the change in the linear residual space.

[0142] It should be understood that HotellingT 2 The statistic is a statistical method that extends the univariate t-test and is used to compare means in multivariate data contexts. It can be seen as a distance measure in the multivariate variable space, measuring the degree of difference between the sample mean vector and the hypothesized mean vector.

[0143] Furthermore, if Figure 4 As shown, the S103-4: based on the first statistic and the second statistic, using the kernel density estimation algorithm, calculates the standard control limit corresponding to each type of working condition, including:

[0144] Since the prior distribution information of the mixed typical variable characteristics is unknown, the kernel density estimation is used to construct the T based on the normal working condition data. 2 and control limits for the Q statistic.

[0145] Assume that the density δ(·) is unknown are mixed canonical variable features g1, g2, …g N T 2 Statistics. T 2 The kernel density estimate of the statistic is:

[0146]

[0147] where K(·) is a non-negative function with an integral of 1 and a mean of zero, h is the bandwidth parameter, and N is the number of samples of the sub-condition;

[0148] Assume that the density η(·) is unknown. N for y r (1),y r (2),…,y r (N), the kernel density estimate of the SPE statistic is:

[0149]

[0150] It should be understood that the normal data of different sub-conditions are input into the fault detection model respectively. First, NCVs are extracted by performing KCVDA, and then CVDA is applied to extract linear features from CVs in the residual subspace of KCVDA to construct monitoring statistics T respectively. 2 and Q.

[0151] It should be understood that the SPE statistic stands for Squared Prediction Error, which is a statistic commonly used in process monitoring and fault detection to measure the difference between the observed value and the model predicted value.

[0152] Due to the problem of mixing linear and nonlinear characteristics in the actual operation process of the HVAC system, a single nonlinear model is no longer the best choice in this case. Aiming at the problem that the data of each working condition after the HVAC system is divided has mixed linear and nonlinear characteristics, the present invention adopts the KCVDA-CVDA hybrid detection method to model each divided sub-working condition.

[0153] Furthermore, the S104: comparing the online control limit with the standard control limit of the working condition class to which the online measurement data belongs, to obtain a classification result of whether the online measurement data has a fault, includes:

[0154] Statistics for online test data like This means that the test dataset X for the HVAC system test It is normal, otherwise, it indicates a fault.

[0155] For the statistics Q of online test data test ,like This means that the test dataset X for the HVAC system test It is normal, otherwise, it indicates a fault.

[0156] Embodiment 2

[0157] This embodiment provides a HVAC fault detection system based on conditional drive and hybrid detection;

[0158] HVAC fault detection system based on condition-driven and hybrid detection, including:

[0159] A conversion module is configured to: select three variables of indoor temperature, air supply volume, and air duct pressure of a normally operating HVAC system as condition-driven indicator variables, and convert the time slice data into condition slice data;

[0160] A feature extraction module is configured to: perform feature extraction on each conditional slice using a spatiotemporal variational autoencoder model to obtain a spatiotemporal variational autoencoder similarity coefficient; then perform similarity analysis using the spatiotemporal variational autoencoder similarity coefficient to cluster the conditional slices into the same working condition;

[0161] A hybrid detection module is configured to: calculate the standard control limit corresponding to each type of working condition by hybrid detection of nonlinear typical variables and linear typical variables;

[0162] The classification output module is configured to: obtain online measurement data of the HVAC system, derive the operating condition class to which the online measurement data belongs based on the same method as the conversion module and the feature extraction module, derive the online control limit of the online measurement data based on the same method as the hybrid detection module, compare the online control limit with the standard control limit of the operating condition class to which the online measurement data belongs, and obtain a classification result of whether the online measurement data is faulty.

[0163] It should be noted that the above-mentioned conversion module, feature extraction module, hybrid detection module and classification output module correspond to steps S101 to S104 in Embodiment 1, and the examples and application scenarios implemented by the above-mentioned modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above-mentioned Embodiment 1. It should be noted that the above-mentioned modules as part of the system can be executed in a computer system such as a set of computer executable instructions.

[0164] The description of each embodiment in the above embodiments has different emphases. For parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0165] The proposed system can be implemented in other ways. For example, the system embodiment described above is only illustrative, and the division of the modules is only a logical function division. In actual implementation, there may be other division methods, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0166] Embodiment 3

[0167] This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein the processor is connected to the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory so that the electronic device executes the method described in the above embodiment one.

[0168] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays 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 also be any conventional processor, etc.

[0169] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0170] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in a processor or an instruction in the form of software.

[0171] The method in the first embodiment can be directly embodied as a hardware processor, or a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware. To avoid repetition, it will not be described in detail here.

[0172] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0173] Embodiment 4

[0174] This embodiment further provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the method described in the first embodiment is completed.

[0175] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A HVAC fault detection method based on conditional drive and hybrid detection, characterized in that: include: (1) Select the three variables of indoor temperature, air supply volume, and duct pressure of a normally operating HVAC system as condition-driven indicator variables, and convert the time slice data into condition slice data; (2) Using the spatiotemporal variational autoencoder model to extract features on each conditional slice, and obtaining the spatiotemporal variational autoencoder similarity coefficient; Then, similarity analysis is performed using the spatiotemporal variational autoencoder similarity coefficient to cluster the conditional slices into the same working condition; (3) Using a mixed detection method of nonlinear typical variables and linear typical variables, the standard control limits corresponding to each type of operating condition are calculated; (4) Obtain online measurement data of the HVAC system, determine the operating condition category to which the online measurement data belongs based on the same method as (1) to (2), determine the online control limit of the online measurement data based on the same method as (3), compare the online control limit with the standard control limit of the operating condition category to which the online measurement data belongs, and obtain a classification result of whether the online measurement data is faulty.

2. The HVAC fault detection method based on conditional drive and hybrid detection as claimed in claim 1, characterized in that: The three variables of indoor temperature, air volume and duct pressure of a normally operating HVAC system are selected as condition-driven indicator variables, and the time slice data is converted into conditional slice data, including: In the time-varying three-dimensional normal operation data set X(I×K×J) of the HVAC system, a slice operation is performed perpendicular to the time axis to obtain two-dimensional time slices X at different times. k (I×J); where k represents the sampling time, k=1, 2,…, K, I represents the batch, and J represents the variable; In the k-th 2D time slice X k In the variable J of (I×J), calculate the mean of the indoor temperature variables of different batches at the same time Average air volume Mean pipeline pressure Indicator variable, where k = 1, 2, ..., K, the formula is as follows: Among them, T k (I) represents the indoor temperature value of the Ith batch at the kth moment, V k (I) represents the air supply volume value of the Ith batch at the kth time, P k (I) represents the air duct pressure value of the Ith batch at k moments; In order to better realize the working condition division, the multi-variable is converted into a weighted variable. The weighted variable combines multiple indicator variables in a weighted manner. First, After standardization, the weights of the three variables are calculated to achieve the synthetic representation of multiple indicator variables. The weighted variable formula is as follows: Slice the weighted variable conditionally by increasing the weighted variable [x min ,x min+ ,…,x i ,…x max ] to obtain the weighted variable x on the conditional axis. k Slice X according to the conditions that change from small to large m (I×J); where X m (I×J), m=1,2,…,K represents the mth conditional slice.

3. The HVAC fault detection method based on conditional drive and hybrid detection as claimed in claim 1, characterized in that: The spatiotemporal variational autoencoder model is used to extract features on each conditional slice to obtain the spatiotemporal variational autoencoder similarity coefficient; Then, the spatiotemporal variational autoencoder similarity coefficient is used to perform similarity analysis and cluster the conditional slices into the same working condition, including: Each conditional slice is standardized, and the trained spatiotemporal variational autoencoder model is used to extract features on each standardized conditional slice to obtain the spatiotemporal variational autoencoder similarity coefficient; The trained spatiotemporal variational autoencoder model comprises: an encoder, a hidden layer and a decoder connected in sequence; the encoder comprises: a first convolutional layer, a temporal convolutional network and a second convolutional layer connected in sequence; the decoder comprises: a third convolutional layer, a graph convolutional neural network and a fourth convolutional layer connected in sequence; The similarity analysis is performed using the spatiotemporal variational autoencoder similarity coefficient to cluster the conditional slices into the same working condition.

4. The HVAC fault detection method based on conditional drive and hybrid detection as claimed in claim 3, characterized in that: The trained spatiotemporal variational autoencoder model includes: Encoder, normalizes the conditional slices The conditional step size is embedded through 1×1 convolution, and multiple layers of one-dimensional expanded convolution are used to gradually learn the dependencies between batches and compress the data length. The expanded convolution 1×k operation is performed using the filter f(i) with a kernel size of K and an expansion factor of d, which is defined as: Where t(s) represents the conditional step size s of the compressed sequence t; the encoder compresses the original data of length T into a hidden state of length 1 through multiple one-dimensional unfolded convolutions, and then outputs the hidden state h of the inherent pattern through 1×1 convolution; Based on the idea of ​​variational inference, the hidden layer is used to fit μ and σ, as shown in the following formula: μ=BC μ (X) (3) logσ 2 =TCN σ (X) (4) Where μ is the mean vector matrix and σ is the variance matrix; TCN μ (X) is responsible for extracting the features of the input data X and generating a mean vector matrix μ, TCN σ (X) Generates a log-variance matrix logσ for calculating the distribution of latent variables 2 ; The decoder, for the hidden state z(s), uses the filter f(i) with a kernel size of 1×k and an expansion factor of d to perform a transposed expansion convolution operation, which is defined as: The feature extraction is performed on each normalized conditional slice using the trained spatiotemporal variational autoencoder model to obtain the spatiotemporal variational autoencoder similarity coefficient, including: The loss function L is minimized through the trained spatiotemporal variational autoencoder model; then the potential vector of the hidden layer is output through the trained spatiotemporal variational autoencoder model: Z=q(X|N(μ,σ 2 )) (9) The latent vector of the hidden layer is the spatiotemporal variational autoencoder similarity coefficient.

5. The HVAC fault detection method based on conditional drive and hybrid detection as claimed in claim 1, characterized in that: The similarity analysis is performed using the spatiotemporal variational autoencoder similarity coefficient to cluster the conditional slices into the same working condition, including: After outputting the potential vector matrix, the mth conditional slice X is calculated using the spatiotemporal variational autoencoder similarity coefficient. m and the m+1th conditional slice X m+1 The similarity ρ between two conditional slices is as follows: In the formula, z m 、z m+1 They are conditional slices X m , X m+1 The latent vector of z m 、z m+1 The angle between If the similarity coefficient of adjacent multivariate conditional slices is 0.5≤ρ≤1, it indicates that the feature similarity of the two multivariate conditional slices is high, and the two multivariate conditional slices are clustered under the same condition; If the spatiotemporal variational autoencoder similarity coefficient of adjacent multivariate conditional slices is 0≤ρ≤0.5, it indicates that the feature similarity of the two adjacent multivariate conditional slices is low. The mth multivariate conditional slice is taken as the last multivariate conditional slice of the previous working condition, and the m+1th multivariate conditional slice is taken as the first multivariate conditional slice of the next working condition.

6. The HVAC fault detection method based on conditional drive and hybrid detection as claimed in claim 1, characterized in that: The standard control limits corresponding to each type of working condition are calculated by using a mixed detection method of nonlinear typical variables and linear typical variables, including: For each type of working condition, a corresponding fault detection model based on nonlinear-linear hybrid monitoring is established; Based on the fault detection model of nonlinear-linear hybrid monitoring, the process data of the current working condition data set is subjected to kernel canonical variable dissimilarity analysis: nonlinear canonical variables are extracted in the main subspace of kernel canonical variable dissimilarity analysis; canonical variable dissimilarity analysis is applied in the residual subspace of kernel canonical variable dissimilarity analysis; linear canonical variables and residual variables are extracted from the main subspace of canonical variable dissimilarity analysis; Calculating a first statistic based on the nonlinear canonical variable and the linear canonical variable; Calculate a second statistic based on the residual variable; Based on the first statistic and the second statistic, the kernel density estimation algorithm is used to calculate the standard control limits corresponding to each type of operating condition.

7. The HVAC fault detection method based on conditional drive and hybrid detection as claimed in claim 6, characterized in that: Based on the fault detection model of nonlinear-linear hybrid monitoring, the kernel canonical variable dissimilarity analysis is performed on the process data of the current working condition data set: nonlinear canonical variables are extracted in the main subspace of the kernel canonical variable dissimilarity analysis; canonical variable dissimilarity analysis is applied in the residual subspace of the kernel canonical variable dissimilarity analysis; Extract linear canonical variables and residual variables from the principal subspace of canonical variable dissimilarity analysis, including: set up is the process input at time k, is the process output at time k, and defines the past data vector containing past inputs and outputs and a future data vector containing future outputs where p and f are the past and future data vectors z respectively p (k) and y f (k) time lag; Assume a training data set with N measurements u(k) and y(k), k = 1, 2, ..., N collected under normal operating conditions, for all k ∈ [p + 1, p + M], by z p (k) and y f (k) Construct the past and future Hankel matrices Z p and Y f ,as follows: Kernel Canonical Variable Dissimilarity Analysis (KCVDA) uses kernel functions to map raw data into a high-dimensional feature space. This mapping captures nonlinear relationships into features, so Kernel Canonical Variable Dissimilarity Analysis is used to extract nonlinear features. Assume that the input vector Nonlinear transformation of and the future input vector Nonlinear transformation of Where Z p ,Y f is a data sequence matrix containing past and future information, H i It is called feature space; Then, the goal is to find f1∈H1 and f2∈H2 such that f1(Z p )=<Φ1(Z p ),f1> and f2(Y f )=〈Φ2(Y f ),f2> has the greatest correlation; f1 and f2 are located in the linear space S1 and S2 respectively, which are formed by the images of Φ1 and Φ2, so: Using the kernel matrix K p ,K f Mapping the original data into a high-dimensional feature space, K p ,K f is defined as follows: The canonical variable dissimilarity KCVD between the canonical variables of the past projection and the future projection is used in the kernel canonical variable dissimilarity analysis KCVDA, KCVDd at time k s (k) is defined as follows: d s (k)=AK f -BK p (14) Among them, A∈(β1,β2,...,β s ) consists of the first s singular values, B∈(β s+1 ,β s+2 ,...,β M ) consists of the last s to m singular values; Canonical variable dissimilarity analysis (CVDA) is applied in the residual subspace of kernel canonical variable dissimilarity analysis (KCVDA). Canonical variable dissimilarity analysis (CVDA) focuses on finding the most discriminative linear projection direction between data sets, thereby using canonical variable dissimilarity analysis (CVDA) to extract linear features; Canonical variable dissimilarity analysis (CVDA) uses canonical variable dissimilarity (CVD) between canonical variables in the residual subspace; CVDd at time k n (k) is defined as follows: d n (k)=L n y f (k)-S n J n z p (k) (15) in, U n Contains the first n columns of U, S n The n largest singular values ​​S n =diag(λ1,λ2,...,λ n )composition.

8. The HVAC fault detection method based on conditional drive and hybrid detection as claimed in claim 6, characterized in that: Based on the nonlinear canonical variable and the linear canonical variable, a first statistic is calculated, including: Calculate HotellingT at time k 2 The statistical fault detection quantity T 2 : T 2 (k)=d(k)d(k) T (16) Among them, T 2 The measured state vector d(k) changes, d(k) contains the nonlinear canonical variable dissimilarity d s (k) and the linear canonical variable dissimilarity d n (k); The second statistic is calculated based on the residual variable, including: Assume y r is the data in the CVDA residual space, y r The calculation formula is as follows: Introduce the statistic Q: Q(k)=y r (k)y r (k) T (18) Here, Q(k) measures the change in the linear residual space.

9. The HVAC fault detection method based on conditional drive and hybrid detection as claimed in claim 6, characterized in that: Based on the first statistic and the second statistic, the kernel density estimation algorithm is used to calculate the standard control limits corresponding to each type of working condition, including: Since the prior distribution information of the mixed typical variable characteristics is unknown, the kernel density estimation is used to construct the T based on the normal working condition data. 2 and control limits for the Q statistic; Assume that the density δ(·) is unknown are mixed canonical variable features g1, g2, …g N T 2 Statistics; T 2 The kernel density estimate of the statistic is: Where K() is a non-negative function with an integral of 1 and a mean of zero, h is the bandwidth parameter, and N is the number of samples of the sub-condition; Assume that the density η() is unknown Q1,Q2,…,Q N for y r (1),y r (2),…,y r (N), the kernel density estimate of the SPE statistic is:

10. A HVAC fault detection system based on conditional drive and hybrid detection, characterized in that: include: A conversion module is configured to: select three variables of indoor temperature, air supply volume, and air duct pressure of a normally operating HVAC system as condition-driven indicator variables, and convert the time slice data into condition slice data; A feature extraction module is configured to: perform feature extraction on each conditional slice using a spatiotemporal variational autoencoder model to obtain a spatiotemporal variational autoencoder similarity coefficient; Then, similarity analysis is performed using the spatiotemporal variational autoencoder similarity coefficient to cluster the conditional slices into the same working condition; A hybrid detection module is configured to: calculate the standard control limit corresponding to each type of working condition by hybrid detection of nonlinear typical variables and linear typical variables; The classification output module is configured to: obtain online measurement data of the HVAC system, derive the operating condition class to which the online measurement data belongs based on the same method as the conversion module and the feature extraction module, derive the online control limit of the online measurement data based on the same method as the hybrid detection module, compare the online control limit with the standard control limit of the operating condition class to which the online measurement data belongs, and obtain a classification result of whether the online measurement data is faulty.

Citation Information

Patent Citations

  • Heating ventilation air conditioner fault detection method based on condition driving and neural component analysis

    CN117006605A

  • Water chilling unit fault detection method and system based on data image difference

    CN117076904A

  • Water chilling unit fault diagnosis method, system, medium and equipment

    CN118051829A

  • Method, apparatus and computer-readable medium of testing, adjusting, balancing of air conditioning for performance improvement of air conditioning equipment

    KR102509319B1

  • Manufacturing method for pork cutlet sirloin and pork cutlet sirloin manufactured by the same

    KR102743653B1