Data center sensor fault calibration iteration method based on Bayesian space reconstruction

Abstract

The invention belongs to the technical field of sensor calibration, and particularly discloses a data center sensor fault calibration iteration method based on Bayesian space reconstruction, and the method comprises the steps: carrying out the fitting of the data of a target system, obtaining a first regression function, carrying out the fitting of the data of the target system which does not comprise a to-be-calibrated sensor, and obtaining a second regression function; obtaining a second regression function; constructing a calibration model based on the first regression function and the second regression function; calculating posterior distribution based on the calibration model to obtain a compensation value of the sensor to be calibrated; calibrating the sensor to be calibrated based on the compensation value; and acquiring real-time data of the target system, and adjusting the compensation value based on the real-time data. The method can improve the calibration precision and efficiency.

Description

Technical Field

[0001] The present application belongs to the field of sensor calibration technology, and more specifically, relates to an iterative method for data center sensor fault calibration based on Bayesian space reconstruction. Background Art

[0002] Existing technologies generally use static calibration methods for sensor failures to calibrate sensors. However, due to the large number of sensors in data center cooling systems, which are constantly being updated, and the complex operating conditions, including multiple control modes, high heat loads, complex airflow distributions, and dynamically changing cooling requirements, the existing static calibration methods for sensor failures are difficult to adapt to changes in the system's operating status, resulting in low calibration accuracy and efficiency. Summary of the Invention

[0003] In response to the defects of the existing technology, the purpose of this application is to provide an iterative method for data center sensor fault calibration based on Bayesian space reconstruction, aiming to solve the problem of low calibration accuracy and efficiency caused by the difficulty of existing static calibration methods for sensor faults to adapt to changes in system operating status.

[0004] To achieve the above objectives, in a first aspect, the present application provides an iterative method for data center sensor fault calibration based on Bayesian space reconstruction, comprising: Fitting the data of the target system to obtain a first regression function, and fitting the data of the target system excluding the sensor to be calibrated to obtain a second regression function; constructing a calibration model based on the first regression function and the second regression function; Calculating a posterior distribution based on the calibration model to obtain a compensation value of the sensor to be calibrated; calibrating the sensor to be calibrated based on the compensation value; Real-time data of the target system is acquired, and the compensation value is adjusted based on the real-time data.

[0005] This application utilizes the high-precision data fitting capability of the regression method, establishes the first regression function and the second regression function as system model items, constructs a calibration model, and then calculates the posterior distribution through the Bayesian probability framework to obtain the compensation value of the sensor to be calibrated. It can effectively handle the uncertainty and fuzzy information in the operating data of the data center cooling system, model and reason about sensor faults in the form of probability distribution, isolate the influence of abnormal nodes, effectively suppress the propagation of sensor errors in the system, and improve calibration accuracy and efficiency. By continuously integrating the system's new real-time data into the modeling process and dynamically adjusting the compensation value, it can respond to changes in the data center cooling system mode and environment in real time, ensuring that the sensor can still maintain efficient operation in a dynamically changing environment, and improving calibration accuracy and efficiency.

[0006] According to the iterative method for data center sensor fault calibration based on Bayesian space reconstruction provided by the present application, fitting the data of the target system to obtain a first regression function, and fitting the data of the target system that does not include the sensor to be calibrated to obtain a second regression function, the method includes: Perform multiple linear regression fitting on the data of the target system to obtain a first regression function, and perform multiple linear regression fitting on the data of the target system excluding the sensor to be calibrated to obtain a second regression function.

[0007] This application uses multivariate linear regression to analyze the correlation between multidimensional variables, which can comprehensively consider multiple influencing factors and establish a linear relationship between input variables and output variables, thereby more accurately fitting sensor data. Its interaction with the Bayesian probability framework can effectively process the high-dimensional and nonlinear data relationships of the data center cooling system to accurately reflect the relationship between faults and calibration variables, thereby improving the accuracy and efficiency of sensor calibration.

[0008] According to the iterative method for data center sensor fault calibration based on Bayesian space reconstruction provided by the present application, the method of calculating the posterior distribution based on the calibration model to obtain the compensation value of the sensor to be calibrated includes: calculating a likelihood function based on the calibration model; Based on the likelihood function and the prior distribution of the preset compensation value, obtaining a posterior distribution sample by sampling; Approximating the posterior distribution sample to a probability density distribution curve; Based on the probability density distribution curve, a compensation value of the sensor to be calibrated is obtained.

[0009] According to the present application, a data center sensor fault calibration iterative method based on Bayesian space reconstruction is provided, and the method further includes: Reconstructing historical normal operating data of the target system through a deep neural network model autoencoder to obtain reconstructed data; Calculating a minimum distance metric between the historical normal operating data and the reconstructed data; calculating a screening threshold based on the minimum distance metric; Based on the screening threshold, screening training data from the historical normal operating data; Based on the training data, the deep neural network model autoencoder and the calibration model are trained.

[0010] The deep neural network model autoencoder of this application processes historical normal operation data, which can effectively extract the deep-level features of the data and significantly improve the efficiency and accuracy of data processing through nonlinear transformation and dimensionality reduction technology. The autoencoder can automatically learn the intrinsic features in the data through its unique network structure and training algorithm, effectively remove noise and abnormal data, and provide higher quality data support for subsequent calibration and analysis.

[0011] According to the iterative method for data center sensor fault calibration based on Bayesian space reconstruction provided by the present application, the historical normal operation data of the target system is reconstructed by a deep neural network model autoencoder to obtain the reconstructed data, including: Encoding and feature learning the historical normal operation data of the target system through a deep neural network model autoencoder to obtain low-dimensional data; The low-dimensional data is decoded to obtain reconstructed data.

[0012] According to the present application, a data center sensor fault calibration iterative method based on Bayesian space reconstruction is provided, and the method further includes: The screening threshold is adjusted based on real-time data from the target system.

[0013] This application continuously integrates new real-time data from the system into the modeling process and dynamically adjusts the screening threshold, which can respond to changes in the data center cooling system mode and environment in real time, ensuring that the sensor can still maintain efficient operation in a dynamically changing environment, and improving calibration accuracy and efficiency.

[0014] In a second aspect, the present application provides a data center sensor fault calibration iterative device based on Bayesian space reconstruction, comprising: a fitting module, configured to fit the data of the target system to obtain a first regression function, and to fit the data of the target system excluding the sensor to be calibrated to obtain a second regression function; A construction module, configured to construct a calibration model based on the first regression function and the second regression function; A calculation module, configured to calculate a posterior distribution based on the calibration model to obtain a compensation value of the sensor to be calibrated; a calibration module, configured to calibrate the sensor to be calibrated based on the compensation value; An adjustment module is used to obtain real-time data of the target system and adjust the compensation value based on the real-time data.

[0015] In a third aspect, the present application provides an electronic device comprising: at least one memory for storing programs; and at least one processor for executing the programs stored in the memory. When the programs stored in the memory are executed, the processor is used to execute the iterative method for data center sensor fault calibration based on Bayesian space reconstruction described in the first aspect or any possible implementation of the first aspect.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the iterative method for data center sensor fault calibration based on Bayesian space reconstruction described in the first aspect or any possible implementation of the first aspect.

[0017] In a fifth aspect, the present application provides a computer program product, which, when running on a processor, enables the processor to execute the iterative method for data center sensor fault calibration based on Bayesian space reconstruction described in the first aspect or any possible implementation of the first aspect.

[0018] It can be understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here.

[0019] In general, the above technical solutions conceived by this application have the following beneficial effects compared with the existing technologies: (1) This application utilizes the high-precision data fitting capability of the regression method, establishes the first regression function and the second regression function as system model items, constructs a calibration model, and then calculates the posterior distribution through the Bayesian probability framework to obtain the compensation value of the sensor to be calibrated. This can effectively handle the uncertainty and fuzzy information in the operating data of the data center cooling system, model and reason about sensor faults in the form of probability distribution, isolate the influence of abnormal nodes, effectively suppress the propagation of sensor errors in the system, and improve calibration accuracy and efficiency. By continuously integrating the system's new real-time data into the modeling process and dynamically adjusting the compensation value, it can respond to changes in the data center cooling system mode and environment in real time, ensuring that the sensor can still maintain efficient operation in a dynamically changing environment, and improving calibration accuracy and efficiency.

[0020] (2) This application analyzes the correlation between multidimensional variables through multivariate linear regression, which can comprehensively consider multiple influencing factors and establish a linear relationship between input variables and output variables, thereby fitting sensor data more accurately. Its interaction with the Bayesian probability framework can effectively process the high-dimensional and nonlinear data relationship of the data center cooling system to accurately reflect the relationship between faults and calibration variables, thereby improving the accuracy and efficiency of sensor calibration.

[0021] (3) The deep neural network model autoencoder of this application processes historical normal operation data, which can effectively extract the deep features of the data and significantly improve the efficiency and accuracy of data processing through nonlinear transformation and dimensionality reduction technology. The autoencoder can automatically learn the intrinsic features in the data through its unique network structure and training algorithm, effectively remove noise and abnormal data, and provide higher quality data support for subsequent calibration and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0023] Figure 1 1 is a flow chart of an iterative method for data center sensor fault calibration based on Bayesian space reconstruction provided by an embodiment of the present application; Figure 2 Schematic diagram of the principle of constructing a screening threshold by an autoencoder provided in an embodiment of the present application; Figure 3 This is a schematic diagram of the update iteration mechanism provided by an embodiment of the present application; Figure 4 This is a schematic diagram comparing energy consumption under different sensor failures provided in an embodiment of the present application; Figure 5 This is a schematic diagram comparing mean-square error (MSE) under a faulty temperature sensor in a computer room provided by an embodiment of the present application; Figure 6 This is a schematic diagram comparing the MSE under a supply air temperature sensor failure provided in an embodiment of the present application; Figure 7 1 is a schematic diagram of the structure of a data center sensor fault calibration iterative device based on Bayesian space reconstruction provided by an embodiment of the present application; Figure 8 It is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0025] The term "and / or" as used herein describes an association between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. The symbol " / " as used herein indicates that the related objects are in an "or" relationship, for example, A / B means either A or B.

[0026] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0027] In the description of the embodiments of the present application, unless otherwise specified, "multiple" means two or more, for example, multiple processing units means two or more processing units, etc.; multiple elements means two or more elements, etc.

[0028] First, let’s introduce the following contents: Traditional sensor calibration methods have several limitations. First, they often rely on specialized calibration equipment or known reference standards, which are expensive, complex to maintain, and difficult to obtain or accurately use in the volatile operating environments of data center cooling systems, limiting calibration flexibility. Second, traditional methods often rely on fixed thresholds for calibration in data center cooling systems. This approach struggles to accurately capture the impact of the dynamic operating environment on sensor measurements and fails to effectively model the complex relationships between the numerous variables in the system, leading to invalid calibration results.

[0029] Next, combine Figures 1-6 The iterative method for data center sensor fault calibration based on Bayesian space reconstruction provided in an embodiment of the present application is introduced.

[0030] Figure 1 FIG is a flow chart of an iterative method for data center sensor fault calibration based on Bayesian space reconstruction provided by an embodiment of the present application, such as Figure 1 As shown, the method includes the following steps: Step 100: Fit the data of the target system to obtain a first regression function, and fit the data of the target system excluding the sensor to be calibrated to obtain a second regression function; Optionally, the data of the target system may be operating data of the target system, and the target system may be any system including sensors, such as a cooling system of a data center.

[0031] Optionally, the high-precision data fitting capability of the regression method can be utilized to establish a first regression function and a second regression function as system model items, wherein the first regression function is fitted by the data of the target system including the sensor to be calibrated, and serves as the measurement value of the target system, and the second regression function is fitted by the data of the target system not including the sensor to be calibrated, and serves as a benchmark for a reliable system.

[0032] Step 110, constructing a calibration model based on the first regression function and the second regression function; The first regression function and the second regression function are used as system model items to construct a calibration model.

[0033] Step 120: Calculate the posterior distribution based on the calibration model to obtain the compensation value of the sensor to be calibrated; Obtaining a calibration model Finally, the posterior distribution and density function are calculated through the Bayesian probability framework to obtain the compensation value of the target sensor .

[0034] The causal reasoning mechanism of the Bayesian probability framework can isolate the impact of abnormal nodes and effectively suppress the propagation of sensor errors within the system. It can effectively handle the uncertainty and fuzzy information in the operating data of the data center cooling system, model and reason about sensor failures in the form of probability distribution, and further improve the accuracy and stability of the calibration results.

[0035] Step 130, calibrating the sensor to be calibrated based on the compensation value; The compensation function of the sensor to be calibrated is as follows:

[0036] Where, For the compensated sensor, is the sensor to be calibrated, is the compensation value.

[0037] Step 140: Acquire real-time data of the target system and adjust the compensation value based on the real-time data.

[0038] The calibrated sensors are put into the data center cooling system to achieve fault-tolerant real-time control of sensor failures. By continuously inputting new data center cooling system data, the compensation value is dynamically adjusted to ensure that the sensors maintain high-precision measurements under the complex and changeable operating conditions of the data center cooling system.

[0039] The iterative method for data center sensor fault calibration based on Bayesian space reconstruction provided in this application utilizes the high-precision data fitting capability of the regression method to establish a first regression function and a second regression function as system model items, constructs a calibration model, and then calculates the posterior distribution through the Bayesian probability framework to obtain the compensation value of the sensor to be calibrated. This can effectively handle the uncertainty and fuzzy information in the operating data of the data center cooling system, model and reason about sensor faults in the form of probability distribution, isolate the influence of abnormal nodes, effectively suppress the propagation of sensor errors in the system, and improve calibration accuracy and efficiency. By continuously integrating the system's new real-time data into the modeling process and dynamically adjusting the compensation value, it can respond to changes in the data center cooling system mode and environment in real time, ensuring that the sensor can still maintain efficient operation in a dynamically changing environment, and improving calibration accuracy and efficiency.

[0040] In some embodiments, step 100 specifically includes: Perform multiple linear regression fitting on the data of the target system to obtain a first regression function, and perform multiple linear regression fitting on the data of the target system excluding the sensor to be calibrated to obtain a second regression function.

[0041] Preferably, the first regression function and the second regression function are fitted by a multiple linear regression method. Multiple linear regression can comprehensively consider multiple influencing factors and establish a linear relationship between input variables and output variables, thereby fitting sensor data more accurately. Multiple linear regression can analyze the correlation between multi-dimensional variables. Its interaction with the Bayesian probability framework can effectively process the high-dimensional and nonlinear data relationship of the data center cooling system to accurately reflect the relationship between faults and calibration variables.

[0042] The improved Bayesian inference method of multiple linear regression can obtain a reliable calibration model as long as the regression model achieves a high fitness and a low regression residual.

[0043] Specifically, the multiple linear regression method is used to construct the system model terms, and the calibration model is obtained as shown below:

[0044]

[0045]

[0046] Where, represents the regression function containing the target sensor to be calibrated, i.e., the first regression function, as the measurement value of the actual data center cooling system, Represents all variable information in the system except the target sensor to be calibrated, that is, the second regression function, which serves as the benchmark for a reliable system. is the sensor to be calibrated; is a physical sensor other than the sensor to be calibrated; , is the constant term of the multiple linear regression model; , are the coefficients corresponding to the above variables.

[0047] In some embodiments, step 120 specifically includes: Step 1201, calculating a likelihood function based on the calibration model; Step 1202: obtaining a posterior distribution sample by sampling based on the likelihood function and the prior distribution of the preset compensation value; Step 1203, approximating the posterior distribution samples to a probability density distribution curve; Step 1204 : Obtain a compensation value of the sensor to be calibrated based on the probability density distribution curve.

[0048] In the sensor calibration problem, the Bayesian inference process is as follows: for the unknown fault amplitude, assume its value is , and satisfies the prior distribution; then, with the help of other sensor data information, the assumed prior distribution is updated and converted into a posterior distribution that is more in line with the actual situation. In this process, the focus is on minimizing the distance function, that is, the calibration model. The distance function usually represents the difference between the reference value and the fault value. After executing the above process, this difference is numerically equal to the mean of the posterior distribution, which is determined by the assumption Got it.

[0049] The basic mathematical description of the Bayesian probability framework is as follows:

[0050]

[0051]

[0052]

[0053]

[0054]

[0055]

[0056] in, It is the distance function, that is, the calibration model. Usually, the calibration model is constructed based on the principle of energy conservation. The posterior distribution is in the prior distribution and likelihood function The distance function The minimum probability density function, is the preset calibration compensation value; for The prior distribution of , usually the standard normal distribution, is the standard deviation of the prior distribution, which determines the form of the prior distribution. is the posterior distribution, is the normalization constant, is the benchmark value of the reliable system model, which is obtained from the partial output of the conservation equation without fault variables. is the number of system models, is the baseline value of the sensor model, is the number of target sensors; is the measured value of the system model, and the partial output of the conservation equation containing the fault variable is obtained. is the value after compensation of model-related variables, is the value after compensation of the unknown variables in the model, is the value after compensation of the target sensor, is the measurement value of the target sensor.

[0057] According to the central limit theorem, the prior distribution is defined as a normal distribution, and then the Markov chain Monte Carlo method is used to solve the normalization constant Among them, the Metropolis Hastings algorithm is a widely used Markov Chain Monte Carlo sampling method that can be used for Bayesian inference sampling to generate posterior distribution samples, thereby obtaining the statistical characteristics of the posterior distribution, such as mean and standard deviation. The basic steps of the Markov Chain Monte Carlo sampling algorithm are as follows: Step 1: Preset the initial parameters of the Markov chain from the prior distribution .For example, It can be defined as the mean of the prior distribution. If a normal distribution is used, then at the beginning of the entire iteration process, is 0.

[0058] Step 2: After step 1, you can preset a Gaussian probability density function , the center of the Gaussian probability density curve should be the front Sampling parameters in iterations In the In the iterations, candidate parameters are sampled It can be determined by a pre-set probability density function, which is the probability density function of the previous sampling parameter Add a random variable based on , as shown below:

[0059] Where, is a random value.

[0060] Step 3: Calculate candidate parameters according to the following formula The pass rate:

[0061] Where, Candidate parameters The posterior distribution function of For the previous parameter The posterior distribution function of Centered on The Gaussian probability density function, Centered on Gaussian probability density function.

[0062] Step 4: By applying the pass rate Another random number randomly generated in the range [0,1] is Compare to determine whether the candidate parameter should be accepted.

[0063]

[0064] Step 5: Repeat steps 2 to 4 until the maximum sampling value is reached. After repeated processing, the posterior distribution samples and probability density function curve samples can be obtained. Finally, the posterior distribution and density function can be used to determine the compensation value of the faulty sensor. .

[0065] In some embodiments, the method further comprises: Step 150 , reconstructing the historical normal operation data of the target system through a deep neural network model autoencoder to obtain reconstructed data; Step 160 , calculating the minimum distance metric between the historical normal operation data and the reconstructed data; Step 170, calculating a screening threshold based on the minimum distance metric; Step 180 , filtering training data from historical normal operation data based on a screening threshold; Step 190: Train the deep neural network model autoencoder and calibration model based on the training data.

[0066] This application reconstructs historical normal operation data based on the deep neural network model autoencoder, and reconstructs the measurement values ​​of the sensors to be calibrated by inputting historical data of the data center cooling system. , and calculate the actual measured variables The reconstruction error function The training data screening threshold is obtained based on the reconstruction error function , the operating data exceeding the threshold value does not participate in modeling, and the operating data within the threshold value is used for modeling.

[0067] Specifically, the system reconstructs the measured values ​​of the sensor to be calibrated and calculates the minimum distance between the reconstructed values ​​and the baseline values ​​of the sensor to be calibrated in historical data. The distribution range of normal data is calculated, and a threshold value is formed within a certain preset confidence interval. Using a method based on relevant threshold criteria, the system optimizes the evolution of historical data and eliminates poor values, ultimately obtaining a high-quality training set that provides reliable data support for sensor fault calibration and fault-tolerant control strategy models.

[0068] Processing historical normal operation data using a deep neural network autoencoder effectively extracts deep-level features from the data. Through nonlinear transformation and dimensionality reduction techniques, the efficiency and accuracy of data processing are significantly improved. Traditional data screening methods often rely on simple statistical rules and struggle to capture the complex and ever-changing operating patterns of data center cooling systems. Therefore, by calculating the minimum distance value and the threshold within the confidence interval, the autoencoder can dynamically optimize data and eliminate poor values, further optimizing the quality of the training set and improving the performance of subsequent modeling.

[0069] The autoencoder can transform the input Transform to hidden vector and reconstruct the output through decoder , the output of the decoder can approximately restore the original input.

[0070] The optimization goal of the autoencoder is to obtain the input And the decoded output The minimum distance metric between , called the reconstruction error function, can also be used as a quantitative model's prediction performance indicator.

[0071]

[0072] Where, is the input variable, is the reconstructed input variable, i It is i input features.

[0073] According to the reconstruction error function , calculate the training data screening threshold , those within the threshold range are used to participate in modeling, and those outside the threshold range are not used for modeling.

[0074]

[0075] Where, is the weight.

[0076] In some embodiments, step 150 specifically includes: Step 1501: Encode and learn features of historical normal operation data of the target system using a deep neural network model autoencoder to obtain low-dimensional data; Step 1502: decode the low-dimensional data to obtain reconstructed data.

[0077] Figure 2 This is a schematic diagram of the principle of constructing a screening threshold by the autoencoder provided in the embodiment of the present application. The mapping of the autoencoder includes two processes: encoding and decoding. The screening data is as follows: Figure 2 As shown in the figure, the process from the input layer to the hidden layer is the encoder, and the process from the hidden layer to the output layer is the decoder. The encoding process compresses the input high-dimensional data into low-dimensional data, and the decoding process restores the low-dimensional data to high-dimensional data. The function of the autoencoder is to perform dimensionality reduction operations, converting the original data into low-dimensional data through the encoding process, then analyzing the low-dimensional data, and converting the analysis results into high-dimensional data through the decoding process. In this way, high-dimensional processing effects can be achieved using low-dimensional processing methods.

[0078] In some embodiments, the method further comprises: Adjust filtering thresholds based on real-time data from target systems.

[0079] As the data center cooling system operates, new data is constantly generated, and the system status and sensor characteristics may also change. This application uses dynamic update iteration to achieve real-time fault-tolerant control of sensor failures and adaptive optimization of the system. Traditional fault calibration methods are often static and difficult to adapt to changes in system operating status. This application dynamically adjusts the compensation value by continuously integrating new system data into the modeling process. and the minimum distance metric , which can respond to changes in the data center cooling system mode and environment in real time, ensuring that the sensors can continue to operate efficiently in a dynamically changing environment.

[0080] Figure 3 This is a schematic diagram of the update iteration mechanism provided by the embodiment of the present application, such as Figure 3 As shown, this application designs an update and iteration mechanism. When faced with continuously updated operating data, the new data is processed in real time through update and iteration and integrated into the modeling process of the first two steps to dynamically adjust the compensation value. and the minimum distance metric , ensuring the real-time and accuracy of the model, real-time fault-tolerant control of sensor failures, and stable system operation. Furthermore, the dynamic update and iteration process also requires handling complex system dynamics. By combining the advantages of Bayesian spatial reconstruction error and deep neural network models, it is possible to rapidly respond to changes in system parameters and ensure the accuracy of compensation values ​​and data thresholds. This approach not only adapts to dynamic system changes but also continuously improves model performance and system reliability through continuous learning and optimization.

[0081] Figure 4 FIG. 1 is a schematic diagram comparing energy consumption under different sensor failures provided in an embodiment of the present application. Figure 4 As shown in the figure, a comparative analysis of energy consumption prediction results for different modeling data sizes s reveals that the calibrated model in this application performs significantly differently under different fault scenarios: In the case of a computer room temperature sensor failure, an s of 400 yields the best result, with energy consumption of 5863.587 MJ, a 0.2% reduction compared to the baseline model MLR-BI. However, an s of 450 results in energy consumption of 5862.076 MJ, a 0.22% reduction. In energy consumption prediction for a supply air temperature sensor failure, an s of 450 results in a 0.05% reduction compared to the MLR-BI.

[0082] These models are evaluated using the mean squared error (MSE) performance metric. When the MSE of a prediction model is smaller, its prediction performance is better. The calculation formula is as follows:

[0083] Where, and are the actual and predicted values ​​of room temperature, n is the total number of samples.

[0084] Figure 5 This is a schematic diagram comparing MSE under a faulty temperature sensor in a computer room provided by an embodiment of the present application. Figure 6 : is a schematic diagram of MSE comparison under the condition of air supply temperature sensor failure provided by the embodiment of the present application, such as Figure 5 and Figure 6 As shown in the figure, the analysis shows that the model of this application with s of 450 performs best in the supply air temperature sensor failure scenario, with the MSE value reaching the lowest 0.0036, which is significantly lower than the MLR-BI model by 36.8%, verifying the superiority of the model in predicting supply air temperature failure under this volume.

[0085] Although the MSE of the computer room temperature sensor failure is optimal when s is 400 (0.04618), it still performs well when s is 450 (0.05264).

[0086] In summary, when s is 450, it shows a clear performance advantage in SAT fault prediction and is the best choice in this scenario. At the same time, it also maintains acceptable prediction accuracy in other fault types, reflecting the advantages of the model of this application in balancing performance and data efficiency.

[0087] Compared with the MLR-BI model, the proposed model achieved the best overall performance when s was 450, with improvements in both MSE and energy consumption. In the case of a room temperature sensor failure, energy consumption was 0.22% lower than that of the MLR-BI model, with an MSE value of 0.05264 at s of 450. In the case of a supply air temperature sensor failure, energy consumption continued to decrease as s increased from 50 to 450, reaching its lowest point at 450, a 0.05% decrease compared to the MLR-BI model. The MSE value at 450 was 0.0036, a 36.8% decrease compared to the MLR-BI model (0.0057), the lowest among all s.

[0088] Figure 7 is a structural diagram of a data center sensor fault calibration iterative device based on Bayesian space reconstruction provided by an embodiment of the present application, such as Figure 7 As shown, the system includes a fitting module 710, a construction module 720, a calculation module 730, a calibration module 740 and an adjustment module 750, wherein: A fitting module 710 is configured to fit the data of the target system to obtain a first regression function, and to fit the data of the target system excluding the sensor to be calibrated to obtain a second regression function; A construction module 720 is configured to construct a calibration model based on the first regression function and the second regression function; A calculation module 730 is used to calculate the posterior distribution based on the calibration model to obtain a compensation value of the sensor to be calibrated; A calibration module 740, configured to calibrate the sensor to be calibrated based on the compensation value; The adjustment module 750 is configured to obtain real-time data of the target system and adjust the compensation value based on the real-time data.

[0089] Based on the method in the above embodiment, Figure 8 An example of a physical structure diagram of an electronic device is shown below. Figure 8As shown, an embodiment of the present application provides an electronic device, which may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the iterative method for data center sensor fault calibration based on Bayesian space reconstruction in the above embodiment.

[0090] In addition, the logic instructions in the aforementioned memory 830 can be implemented in the form of a software functional unit and, when sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the portion that contributes to the prior art, or the portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the iterative method for data center sensor fault calibration based on Bayesian space reconstruction described in various embodiments of the present application.

[0091] Based on the method in the above embodiment, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a processor, the processor executes the data center sensor fault calibration iterative method based on Bayesian space reconstruction in the above embodiment.

[0092] Based on the method in the above embodiment, an embodiment of the present application provides a computer program product. When the computer program product runs on a processor, the processor executes the data center sensor fault calibration iterative method based on Bayesian space reconstruction in the above embodiment.

[0093] It is understood that the processor in the embodiments of the present application may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field programmable gate arrays (FPGA), other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor may be a microprocessor or any conventional processor.

[0094] The method steps in the embodiments of the present application can be implemented by hardware or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, which can be stored in random access memory (RAM), flash memory, read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, hard disks, mobile hard disks, CD-ROMs, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an ASIC.

[0095] The above embodiments can be implemented in whole or in part through software, hardware, firmware, or any combination thereof. When implemented using software, they can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded and executed on a computer, the computer program instructions fully or partially produce the processes or functions described in the embodiments of this application. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted via the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer, or a data storage device such as a server or data center that integrates one or more available media. The available medium can be magnetic media (e.g., floppy disk, hard disk, tape), optical media (e.g., DVD), or semiconductor media (e.g., solid-state drive (SSD)).

[0096] It will be understood that the various numerical numbers involved in the embodiments of the present application are merely distinctions for the convenience of description and are not intended to limit the scope of the embodiments of the present application.

[0097] It is easy for those skilled in the art to understand that the above is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. An iterative method for data center sensor fault calibration based on Bayesian space reconstruction, characterized in that: include: Fitting the data of the target system to obtain a first regression function, and fitting the data of the target system excluding the sensor to be calibrated to obtain a second regression function; constructing a calibration model based on the first regression function and the second regression function; Calculating a posterior distribution based on the calibration model to obtain a compensation value of the sensor to be calibrated; calibrating the sensor to be calibrated based on the compensation value; Real-time data of the target system is acquired, and the compensation value is adjusted based on the real-time data.

2. The iterative method for data center sensor fault calibration based on Bayesian space reconstruction according to claim 1 is characterized in that: The step of fitting the data of the target system to obtain a first regression function and fitting the data of the target system excluding the sensor to be calibrated to obtain a second regression function comprises: Perform multiple linear regression fitting on the data of the target system to obtain a first regression function, and perform multiple linear regression fitting on the data of the target system excluding the sensor to be calibrated to obtain a second regression function.

3. The iterative method for data center sensor fault calibration based on Bayesian space reconstruction according to claim 1 is characterized in that: The step of calculating the posterior distribution based on the calibration model to obtain the compensation value of the sensor to be calibrated includes: calculating a likelihood function based on the calibration model; Based on the likelihood function and the prior distribution of the preset compensation value, obtaining a posterior distribution sample by sampling; Approximating the posterior distribution sample to a probability density distribution curve; Based on the probability density distribution curve, a compensation value of the sensor to be calibrated is obtained.

4. The iterative method for data center sensor fault calibration based on Bayesian space reconstruction according to claim 1 is characterized in that: The method further comprises: Reconstructing historical normal operating data of the target system through a deep neural network model autoencoder to obtain reconstructed data; Calculating a minimum distance metric between the historical normal operating data and the reconstructed data; calculating a screening threshold based on the minimum distance metric; Based on the screening threshold, screening training data from the historical normal operating data; Based on the training data, the deep neural network model autoencoder and the calibration model are trained.

5. The iterative method for data center sensor fault calibration based on Bayesian space reconstruction according to claim 3 is characterized in that: The reconstructing the historical normal operating data of the target system by using the deep neural network model autoencoder to obtain the reconstructed data includes: Encoding and feature learning the historical normal operation data of the target system through a deep neural network model autoencoder to obtain low-dimensional data; The low-dimensional data is decoded to obtain reconstructed data.

6. The iterative method for data center sensor fault calibration based on Bayesian space reconstruction according to claim 4 is characterized in that: The method further comprises: The screening threshold is adjusted based on real-time data from the target system.

7. A data center sensor fault calibration iterative device based on Bayesian space reconstruction, characterized in that: include: a fitting module, configured to fit the data of the target system to obtain a first regression function, and to fit the data of the target system excluding the sensor to be calibrated to obtain a second regression function; A construction module, configured to construct a calibration model based on the first regression function and the second regression function; A calculation module, configured to calculate a posterior distribution based on the calibration model to obtain a compensation value of the sensor to be calibrated; a calibration module, configured to calibrate the sensor to be calibrated based on the compensation value; An adjustment module is used to obtain real-time data of the target system and adjust the compensation value based on the real-time data.

8. An electronic device, characterized in that: include: at least one memory for storing a computer program; At least one processor is used to execute the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the data center sensor fault calibration iterative method based on Bayesian space reconstruction as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program runs on a processor, the processor is enabled to execute the iterative method for data center sensor fault calibration based on Bayesian space reconstruction according to any one of claims 1 to 6.

10. A computer program product, characterized in that When the computer program product runs on a processor, the processor is enabled to execute the iterative method for data center sensor fault calibration based on Bayesian space reconstruction according to any one of claims 1 to 6.

Images