An Online Monitoring and Warning Method for a Water Cooling System Based on the Internet of Things

Abstract

The present invention relates to the technical field of data processing, and particularly to an online monitoring and early warning method for a water-cooling system based on the Internet of Things. This method collects the outlet water temperature during the operation of the water-cooling system to obtain a time-series temperature data sequence; obtains the initial abnormal data in the time-series temperature data sequence, as well as the difference coefficient and autoregressive coefficient when using the ARIMA prediction model to predict the outlet water temperature of the water-cooling system at future times; constructs an initial window, and obtains the optimal moving average coefficient of the ARIMA prediction model according to the degree of interference and data fluctuation characteristics of the initial window; predicts the outlet water temperature of the water-cooling system at future times according to the autoregressive coefficient, difference coefficient and optimal moving average coefficient to obtain a predicted value, and conducts abnormal early warning according to the predicted value. The present invention accurately predicts the outlet water temperature of the water-cooling system at future times, and then effectively warns of the outlet water temperature of the water-cooling system.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to an online monitoring and warning method for a water cooling system based on the Internet of Things. Background Art

[0002] With the continuous progress of industrial technology and the rapid development of Internet of Things technology, water cooling systems are increasingly widely used in multiple fields, such as the petrochemical, power, steel, and metallurgical industries. The water cooling system is crucial for ensuring the stable operation of high-temperature equipment. However, once heat fails to be removed in time, serious safety hazards may be triggered. The traditional methods for fault detection and handling of water cooling systems mainly rely on manual inspection and manual intervention, with slow response speed, unable to prevent and handle system faults in a timely manner, and prone to equipment shutdown and production interruption. Therefore, it is necessary to monitor and predict the outlet water temperature of the water cooling system in real time to prevent equipment operation problems.

[0003] The traditional method uses the ARIMA prediction model to determine the moving average coefficient by analyzing the ACF graph and PACF graph of time series data, and predicts future data based on the moving average coefficient. However, since the outlet water temperature of the water cooling system is affected by equipment operating conditions, heat exchange efficiency, environmental temperature, etc., the outlet water temperature is fluctuating data. And during the process of collecting the outlet water temperature of the water cooling system using an NTC thermistor, there may be various external interference factors in the system operating environment, such as temperature changes, mechanical vibrations, electromagnetic interference, etc. These factors may affect the measurement of the NTC thermistor, causing the collected outlet water temperature to be interfered and exhibit abnormal performance. Therefore, the abnormal data caused by external interference may show characteristics similar to the real abnormal data in the ACF graph and PACF graph, resulting in the moving average coefficient not being the optimal moving average coefficient, thus leading to inaccurate prediction of future data and making it difficult to effectively warn of the outlet water temperature of the water cooling system.

[0004] Therefore, how to accurately predict the outlet water temperature of the water cooling system in the future and then effectively warn of the outlet water temperature of the water cooling system has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, the embodiments of the present invention provide an online monitoring and warning method for a water cooling system based on the Internet of Things to solve the problem of how to accurately predict the outlet water temperature of the water cooling system in the future and then effectively warn of the outlet water temperature of the water cooling system.

[0006] The embodiments of the present invention provide an online monitoring and warning method for a water cooling system based on the Internet of Things, and the method includes the following steps:

[0007] Collect the outlet water temperature during the operation of the water cooling system within a preset time range to obtain a time series temperature data sequence;

[0008] Perform time series analysis on the time series temperature data sequence to obtain the difference coefficient and autoregressive coefficient when using the ARIMA prediction model to predict the outlet water temperature of the water cooling system in the future. According to the autoregressive coefficient of the ARIMA prediction model, construct an initial window corresponding to the initial moving average coefficient of the ARIMA prediction model in the time series temperature data sequence;

[0009] According to the data fluctuation characteristics in the time series temperature data sequence, obtain the initial abnormal data in the time series temperature data sequence. According to the recoverability and continuity of the initial abnormal data in the initial window in the time series temperature data sequence, obtain the degree of interference of the initial window. According to the degree of interference of the initial window and the data fluctuation characteristics in the initial window, obtain the best moving average coefficient of the ARIMA prediction model;

[0010] According to the autoregressive coefficient, difference coefficient, and best moving average coefficient of the ARIMA prediction model, predict the outlet water temperature of the water cooling system in the future to obtain a predicted value, and perform abnormal warning on the outlet water temperature of the water cooling system in the future according to the predicted value.

[0011] The beneficial effects of the embodiments of the present invention compared with the prior art are:

[0012] The present invention collects the outlet water temperature during the operation of the water cooling system within a preset time range to obtain a time series temperature data sequence; performs time series analysis on the time series temperature data sequence to obtain the difference coefficient and autoregressive coefficient when using the ARIMA prediction model to predict the outlet water temperature of the water cooling system at a future time, and constructs an initial window corresponding to the initial moving average coefficient of the ARIMA prediction model in the time series temperature data sequence according to the autoregressive coefficient of the ARIMA prediction model; obtains the initial abnormal data in the time series temperature data sequence according to the data fluctuation characteristics in the time series temperature data sequence, obtains the degree of interference of the initial window according to the recoverability and continuity of the initial abnormal data in the initial window in the time series temperature data sequence, and obtains the optimal moving average coefficient of the ARIMA prediction model according to the degree of interference of the initial window and the data fluctuation characteristics in the initial window; predicts the outlet water temperature of the water cooling system at a future time according to the autoregressive coefficient, difference coefficient, and optimal moving average coefficient of the ARIMA prediction model to obtain a predicted value, and performs abnormal warning on the outlet water temperature of the water cooling system at a future time according to the predicted value. By obtaining the autoregressive coefficient of the ARIMA prediction model, the present invention constructs an initial window corresponding to the initial moving average coefficient of the ARIMA prediction model, and obtains the optimal moving average coefficient of the ARIMA prediction model according to the degree of interference of the initial window and the data fluctuation characteristics. The optimal moving average coefficient of the ARIMA prediction model can make the prediction of the outlet water temperature of the water cooling system at a future time more accurate, and thus effectively warn of the outlet water temperature of the water cooling system at a future time. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0014] Figure 1 FIG. is a flowchart of a method for online monitoring and warning of a water cooling system based on the Internet of Things provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] The following will describe in detail the embodiments of the present disclosure. The examples of the embodiments are shown in the drawings. The embodiments described below with reference to the drawings are exemplary and are intended to explain the present disclosure and should not be construed as limiting the present disclosure.

[0016] It should be noted that the terms "first", "second", etc. in the description of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present disclosure described here can be implemented in an order other than those illustrated or described here. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. On the contrary, they are only examples of devices and methods consistent with some aspects of the present disclosure.

[0017] In order to illustrate the technical solution of the present invention, the following will be described through specific embodiments.

[0018] See Figure 1 , which is a method flow chart of an online monitoring and warning method for a water-cooled system based on the Internet of Things provided in the first embodiment of the present invention. As Figure 1 shown, the method may include:

[0019] Step S101, collect the outlet water temperature when the water-cooled system is running within a preset time range to obtain a time-series temperature data sequence.

[0020] In this embodiment, for industrial equipment with a water-cooled system operating in a fixed mode, an NTC thermistor is added to the water-cooled system. When the industrial equipment is working normally, the NTC thermistor is used to collect the outlet water temperature of the water-cooled system in real time. The outlet water temperature of the water-cooled system in industrial equipment is generally -5°C - 0°C.

[0021] When collecting the outlet water temperature of the water-cooled system, the factors affecting the collection frequency include: (1) the scale of the water-cooled system: a large water-cooling system may require more frequent collection to ensure the stability and efficiency of the system; (2) the complexity of the water-cooled system: a complex water-cooling system may require more frequent collection to monitor and control multiple parameters; (3) the control accuracy of the water-cooled system: high-precision control requirements may require more frequent collection to achieve precise temperature regulation. Since this embodiment is for industrial water-cooled systems, industrial water-cooled systems require a relatively high collection frequency, so the collection frequency is set to once per minute. This is not limited here and can be set according to specific implementation scenarios.

[0022] In this embodiment, with a collection frequency of once per minute, the outlet water temperature of the water-cooled system during 24 hours of operation is collected. This is not limited here and can be set according to specific implementation scenarios. The outlet water temperature of the water-cooled system during 24 hours of operation is formed into a time-series temperature data sequence.

[0023] Step S102: Perform time series analysis on the time series temperature data sequence to obtain the difference coefficient and autoregressive coefficient when using the ARIMA prediction model to predict the outlet water temperature of the water cooling system in the future. According to the autoregressive coefficient of the ARIMA prediction model, construct an initial window corresponding to the initial moving average coefficient of the ARIMA prediction model in the time series temperature data sequence.

[0024] In order to predict the outlet water temperature of the water cooling system in the future, it is necessary to perform time series analysis on the time series temperature data sequence to obtain the difference coefficient, autoregressive coefficient, and moving average coefficient when using the ARIMA prediction model to predict the outlet water temperature of the water cooling system in the future. However, since the outlet water temperature of the water cooling system is affected by factors such as the equipment operation state, heat exchange efficiency, and ambient temperature, the outlet water temperature is fluctuating data. Moreover, during the process of collecting the outlet water temperature of the water cooling system using an NTC thermistor, various external interference factors may exist in the system operation environment, such as temperature changes, mechanical vibrations, and electromagnetic interference. These factors may affect the measurement of the NTC thermistor, causing the collected outlet water temperature to be interfered and exhibit abnormal manifestations. Therefore, the abnormal data caused by external interference may show characteristics similar to the true abnormal data in the ACF graph and PACF graph, resulting in the moving average coefficient not being the optimal moving average coefficient, thus leading to inaccurate prediction of future data.

[0025] Therefore, it is necessary to obtain the optimal moving average coefficient of the ARIMA prediction model to accurately predict the outlet water temperature of the water cooling system in the future.

[0026] To solve the above problems, first, it is necessary to obtain the difference coefficient and autoregressive coefficient when using the ARIMA prediction model to predict the outlet water temperature of the water cooling system in the future. Then, according to the autoregressive coefficient of the ARIMA prediction model, construct an initial window corresponding to the initial moving average coefficient of the ARIMA prediction model in the time series temperature data sequence. Furthermore, based on the degree of interference of the initial window and the data fluctuation characteristics in the initial window, obtain the optimal moving average coefficient of the ARIMA prediction model.

[0027] Among them, obtaining the difference coefficient and autoregressive coefficient when using the ARIMA prediction model to predict the outlet water temperature of the water cooling system in the future includes:

[0028] In the time-series temperature data sequence, the differencing coefficient of the ARIMA prediction model is obtained through the differencing method and the ADF test method; the autocorrelation coefficient and partial autocorrelation coefficient of the time-series temperature data sequence are obtained. By analyzing the ACF (autocorrelation coefficient) graph and PACF (partial autocorrelation coefficient) graph, the autoregressive coefficient of the ARIMA prediction model is obtained. Among them, the differencing method, the ADF test method, the autocorrelation coefficient, and the partial autocorrelation coefficient belong to the prior art and will not be elaborated here.

[0029] Among them, according to the autoregressive coefficient of the ARIMA prediction model, an initial window corresponding to the initial moving average coefficient of the ARIMA prediction model is constructed in the time-series temperature data sequence, including:

[0030] Taking the autoregressive coefficient of the ARIMA prediction model as the window length, an initial window including the last outlet water temperature data in the time-series temperature data sequence is constructed as the initial window corresponding to the initial moving average coefficient of the ARIMA prediction model.

[0031] In an embodiment, taking the autoregressive coefficient p of the ARIMA prediction model as an example, taking the last outlet water temperature data in the time-series temperature data sequence as the last outlet water temperature data in the initial window, an initial window with a window length of p is constructed as the initial window corresponding to the initial moving average coefficient of the ARIMA prediction model.

[0032] Thus, the initial window corresponding to the initial moving average coefficient of the ARIMA prediction model is obtained.

[0033] Step S103, according to the data fluctuation characteristics in the time-series temperature data sequence, obtain the initial abnormal data in the time-series temperature data sequence. According to the recoverability and continuity of the initial abnormal data in the initial window in the time-series temperature data sequence, obtain the degree of interference of the initial window. According to the degree of interference of the initial window and the data fluctuation characteristics in the initial window, obtain the best moving average coefficient of the ARIMA prediction model.

[0034] After obtaining the initial window corresponding to the initial moving average coefficient of the ARIMA prediction model, it is necessary to first obtain the initial abnormal data in the time-series temperature data sequence according to the data fluctuation characteristics in the time-series temperature data sequence, then obtain the degree of interference of the initial window according to the recoverability and continuity of the initial abnormal data in the initial window in the time-series temperature data sequence, and further obtain the best moving average coefficient of the ARIMA prediction model according to the degree of interference of the initial window and the data fluctuation characteristics in the initial window.

[0035] Among them, the method for obtaining the initial abnormal data in the time-series temperature data sequence according to the data fluctuation characteristics in the time-series temperature data sequence is as follows:

[0036] Obtain the upper edge value in the time-series temperature data sequence through the box plot of the time-series temperature data sequence, and use the outlet water temperature data in the time-series temperature data sequence that is greater than the upper edge value as the initial abnormal data in the time-series temperature data sequence.

[0037] In one embodiment, obtain the box plot of the time-series temperature data sequence, obtain the upper edge value of the box plot of the time-series temperature data sequence, use the upper edge value of the box plot of the time-series temperature data sequence as the abnormal threshold for judging the initial abnormal data, and use the outlet water temperature data in the time-series temperature data sequence that is greater than the upper edge value of the box plot of the time-series temperature data sequence as the initial abnormal data in the time-series temperature data sequence, so as to obtain all the initial abnormal data in the time-series temperature data sequence.

[0038] Since interference has the characteristics of being accidental, recoverable, and discontinuous, the initial abnormal data caused by external interference is relatively isolated. Moreover, the difference between the initial abnormal data and the normal outlet water temperature data is relatively obvious, and it will return to the normal outlet water temperature data after the external interference on the outlet water temperature data ends. This leads to the trend change relationship between the initial abnormal data caused by external interference and the surrounding outlet water temperature data conforming to the quadratic curve characteristics. Since the recoverability of real abnormal data is relatively low, and real abnormal data often has a certain continuity after it appears, the initial abnormal data caused by external interference is often relatively isolated and has a large distance from the rest of the initial abnormal data, while real abnormal data is relatively dense and may have a small distance from other initial abnormal data. Therefore, the continuity frequency and recoverability probability of the initial abnormal data existing in the initial window can be obtained according to the distance between the initial abnormal data and other initial abnormal data and the number of initial abnormal data in the initial window, so as to obtain the interference value of each initial abnormal data in the initial window according to the recoverability probability and continuity frequency of the initial abnormal data, and further obtain the interference degree of the initial window according to the interference value of each initial abnormal data in the initial window.

[0039] Among them, the specific method for obtaining the interference degree of the initial window according to the recoverability and continuity of the initial abnormal data in the initial window in the time-series temperature data sequence includes:

[0040] (1) Denote any initial abnormal data in the initial window as the data to be processed, and in the time-series temperature data sequence, detect whether the two adjacent outlet water temperature data to the data to be processed are initial abnormal data.

[0041] In one embodiment, taking any initial abnormal data A in the initial window as an example, the initial abnormal data A is the i-th outlet water temperature data in the time-series temperature data sequence. Denote any initial abnormal data A in the initial window as the data A to be processed. In the time-series temperature data sequence, obtain the (i - 1)-th outlet water temperature data B and the (i + 1)-th outlet water temperature data C adjacent to the data A to be processed. According to the obtaining method of the initial abnormal data in the time-series temperature data sequence, detect whether the outlet water temperature data B and the outlet water temperature data C are initial abnormal data.

[0042] (2) If the two outlet water temperature data adjacent to the data to be processed are not initial abnormal data, then with the data to be processed as the center, form a sub-window of the data to be processed by combining the consecutive non-initial abnormal data adjacent to the data to be processed and the data to be processed. According to the data fluctuation characteristics in the sub-window, obtain the interference value of the data to be processed.

[0043] In one embodiment, if the outlet water temperature data B and the outlet water temperature data C adjacent to the data A to be processed are not initial abnormal data, then with the data A to be processed as the center, in the time-series temperature data sequence, respectively obtain the non-initial abnormal data consecutive with the outlet water temperature data B and the non-initial abnormal data consecutive with the outlet water temperature data C. Combine the data A to be processed, the outlet water temperature data B, the outlet water temperature data C, the non-initial abnormal data consecutive with the outlet water temperature data B, and the non-initial abnormal data consecutive with the outlet water temperature data C to form a sub-window of the data A to be processed. According to the data fluctuation characteristics in the sub-window, obtain the interference value of the data to be processed.

[0044] Illustrate with an example: Suppose the time-series temperature data sequence is (1, 2, 3, D, 4, 5, 6, 7, 8, B, A, C, 9, 10), and the initial window is (7, 8, B, A, C, 9, 10). Among them, the numbers 1 - 10 are non-initial abnormal data, and the letters A and D are initial abnormal data. If the outlet water temperature data B and the outlet water temperature data C are not initial abnormal data, then the non-initial abnormal data consecutive with the outlet water temperature data B is 4 - 8, and the non-initial abnormal data consecutive with the outlet water temperature data C is 9 - 10. At this time, the formed sub-window is (4, 5, 6, 7, 8, B, A, C, 9, 10).

[0045] Among them, the method for obtaining the interference value of the data to be processed according to the data fluctuation characteristics in the sub-window is as follows:

[0046] Taking the data to be processed as the last water outlet temperature data, a small window is established in the sub-window, and the water outlet temperature data in the small window is fitted to obtain the fitting line of the data to be processed. The fitting lines of each water outlet temperature data in the sub-window except the first water outlet temperature data are obtained respectively, corresponding to a fitting line sequence. The correlation coefficients between each fitting line in the fitting line sequence except the first fitting line and the previous fitting line are calculated respectively, forming a correlation coefficient sequence. The correlation coefficient sequence is fitted to obtain the fitting curve of the correlation coefficient.

[0047] In one embodiment, taking the data A to be processed as an example, taking the data A to be processed as the last water outlet temperature data of the small window and the first water outlet temperature data of the sub-window as the first water outlet temperature data of the small window, a small window is established in the sub-window, and the least squares method is used to fit the water outlet temperature data in the small window to obtain the fitting line of the data A to be processed. According to the acquisition method of the fitting line of the data A to be processed, the fitting lines of each water outlet temperature data starting from the second water outlet temperature in the sub-window are obtained. The fitting lines of each water outlet temperature data starting from the second water outlet temperature in the sub-window form a fitting line sequence. The Spearman correlation coefficients between every two adjacent fitting lines in the fitting line sequence are calculated respectively. The Spearman correlation coefficient belongs to the prior art and will not be elaborated here. A correlation coefficient sequence is formed. The least squares method is used to fit the correlation coefficient sequence to obtain the fitting curve of the correlation coefficient, where the least squares method belongs to the prior art and will not be elaborated here.

[0048] b Obtain the derivative of the correlation coefficient corresponding to the data to be processed on the fitting curve, calculate the addition result of the derivative and the constant 1, denoted as the first addition result. Obtain the number of extreme points on the fitting curve, calculate the absolute value of the difference between the constant 1 and the number of extreme points on the fitting curve, denoted as the first absolute difference. Calculate the addition result of the first absolute difference and the second preset value, denoted as the second addition result. Obtain the mean of the reciprocal of the first addition result and the reciprocal of the second addition result as the recoverability probability of the data to be processed.

[0049] In one embodiment, taking the data A to be processed as an example, the data A to be processed is the i-th outlet water temperature data in the time-series temperature data sequence. First, the derivative method is used to obtain the extreme points on the fitting curve and the number of extreme points on the fitting curve. The derivative method belongs to the prior art and will not be elaborated here. Then, in the sequence of fitting lines, starting from the second fitting line, the Spearman correlation coefficient between each fitting line and the previous fitting line is denoted as the Spearman correlation coefficient corresponding to each fitting line, and the derivative of the Spearman correlation coefficient corresponding to the data A to be processed on the fitting curve is obtained. According to the number of extreme points on the fitting curve and the derivative of the Spearman correlation coefficient corresponding to the data A to be processed on the fitting curve, the recoverability probability of the data A to be processed is calculated:

[0050]

[0051] where X is the recoverability probability of the data A to be processed; is the derivative of the Spearman correlation coefficient corresponding to the data A to be processed on the fitting curve; b is the number of extreme points on the fitting curve; c is the second preset value; i is the serial number of the outlet water temperature data in the time-series temperature data sequence; | | is the absolute value symbol.

[0052] It should be noted that the second preset value c is a constant 2, which is used to prevent the formula calculation from going wrong when the denominator takes a value of 0. There is no limitation here, and it can be set according to the specific implementation scenario; The closer it is to 0, the greater the probability that the Spearman correlation coefficient corresponding to the data A to be processed is an extreme point, and the greater the recoverability probability of the data A to be processed; the closer b is to 1, the more the fitting curve conforms to the characteristics of a quadratic curve, and the greater the recoverability probability of the data A to be processed; the greater the recoverability probability of the data A to be processed, the greater the possibility that the data A to be processed is an interfered data, and the greater the interfered value of the data A to be processed.

[0053] On the fitting curve of c, the correlation coefficient corresponding to the data to be processed is obtained, and the mean value between the correlation coefficients corresponding to the previous outlet water temperature data and the next outlet water temperature data of the data to be processed is denoted as the first mean value. The absolute value of the difference between the correlation coefficient corresponding to the data to be processed and the first mean value is denoted as the second absolute difference. The difference between the constant 1 and the reciprocal of the second absolute difference is used as the influence degree of the data to be processed.

[0054] In one embodiment, taking the data A to be processed as an example, the data A to be processed is the i-th outlet water temperature data in the time-series temperature data sequence. The Spearman correlation coefficient corresponding to the data A to be processed on the fitting curve, and the Spearman correlation coefficients corresponding to the previous outlet water temperature data and the next outlet water temperature data of the data A to be processed are obtained, and the influence degree of the data A to be processed is calculated:

[0055]

[0056] Wherein, Y is the influence degree of the data A to be processed; is the Spearman correlation coefficient corresponding to the data A to be processed on the fitting curve; is the Spearman correlation coefficient corresponding to the next water outlet temperature data of the data A to be processed on the fitting curve; is the Spearman correlation coefficient corresponding to the previous water outlet temperature data of the data A to be processed on the fitting curve; i is the serial number of the water outlet temperature data in the time series temperature data sequence; | | is the absolute value symbol.

[0057] It should be noted that represents the influence degree of the data A to be processed on the trends of the previous water outlet temperature data and the next water outlet temperature data, the larger it is, the greater the influence degree of the data A to be processed; the greater the influence degree of the data A to be processed, it indicates that the data A to be processed is more affected by interference, and the interference value of the data A to be processed is larger.

[0058] d Obtain the number of initial abnormal data in the initial window, respectively obtain the number of water outlet temperature data between each initial abnormal data and the data to be processed in the initial window, form a number sequence, obtain the minimum number in the number sequence, calculate the ratio of the number of initial abnormal data in the initial window to the minimum number, denoted as the first ratio, and perform normalization processing on the first ratio to obtain the continuity frequency of the data to be processed.

[0059] In one embodiment, taking the data A to be processed as an example, the data A to be processed is the i-th water outlet temperature data in the time series temperature data sequence. In the initial window, obtain the number of all initial abnormal data, and obtain the number of water outlet temperature data between each initial abnormal data and the data A to be processed, form a number sequence, obtain the minimum number in the number sequence, and calculate the continuity frequency of the data A to be processed:

[0060]

[0061] Wherein, Z is the continuity frequency of the data A to be processed; M is the number of initial abnormal data in the initial window; is the number sequence; is the minimum number in the number sequence; norm() is the normalization function; i is the serial number of the water outlet temperature data in the time series temperature data sequence.

[0062] It should be noted that The larger it is, it indicates that the interval quantity between the data A to be processed and the remaining initial abnormal data in the initial window is larger, and the continuity frequency of the data A to be processed is smaller; the smaller the continuity frequency of the data A to be processed, it indicates that the possibility of the data A to be processed being interfered is greater, and the interference value of the data A to be processed is larger.

[0063] e Calculate the average values of the recoverability probability, influence degree, and continuity frequency of the data to be processed as the interference value of the data to be processed.

[0064] In one embodiment, taking the data A to be processed as an example, calculate the interference value of the data A to be processed:

[0065]

[0066] Wherein, is the interference value of the data A to be processed; X is the recoverability probability of the data A to be processed; Y is the influence degree of the data A to be processed; Z is the continuity frequency of the data A to be processed.

[0067] It should be noted that the larger the recoverability probability of the data A to be processed, the greater the possibility that the data A to be processed is interfered data, and the larger the interference value of the data A to be processed; the greater the influence degree of the data A to be processed, it indicates that the interference influence on the data A to be processed is greater, and the interference value of the data A to be processed is larger; the smaller the continuity frequency of the data A to be processed, it indicates that the possibility of the data A to be processed being interfered is greater, and the interference value of the data A to be processed is larger.

[0068] (3) If there is an initial abnormal data among the two outlet water temperature data adjacent to the data to be processed, record the initial abnormal data adjacent to the data to be processed as the adjacent abnormal data, and record the non-data to be processed adjacent to the adjacent abnormal data as the data to be detected. Detect whether the data to be detected is an initial abnormal data. When the data to be detected is not an initial abnormal data, with the data to be processed as the center, form the sub-window of the data to be processed by the continuous non-initial abnormal data adjacent to the data to be processed, the continuous non-initial abnormal data adjacent to the adjacent abnormal data, and the data to be processed. According to the data fluctuation characteristics in the sub-window, obtain the interference value of the data to be processed.

[0069] In one embodiment, taking the data A to be processed as an example, if there is an initial abnormal data among the outlet water temperature data B and the outlet water temperature data C adjacent to the data A to be processed, then the initial abnormal data in the outlet water temperature data B and the outlet water temperature data C is recorded as the adjacent abnormal data. Among the two outlet water temperature data adjacent to the adjacent abnormal data, the non-data to be processed is recorded as the data to be detected. The data to be detected is detected. When the data to be detected is not the initial abnormal data, centered on the data A to be processed, in the time-series temperature data sequence, the non-initial abnormal data consecutive with the data to be detected and the non-initial abnormal data consecutive with the data A to be processed are respectively obtained. The non-initial abnormal data consecutive with the data to be detected, the non-initial abnormal data consecutive with the data A to be processed, as well as the data A to be processed and the data to be detected form the sub-window of the data A to be processed. According to the method for obtaining the interference value of the data A to be processed as described above, the interference value of the data A to be processed is obtained.

[0070] Illustrative example: Suppose the time-series temperature data sequence is (1, 2, 3, D, 4, 5, 6, 7, 8, B, A, C, 9, 10), and the initial window is (7, 8, B, A, C, 9, 10). Among them, the numbers 1 - 10 are non-initial abnormal data, and the letters A and D are initial abnormal data. If the outlet water temperature data B is the initial abnormal data and the outlet water temperature data C is not the initial abnormal data, then the outlet water temperature data B is recorded as the adjacent abnormal data B, and the outlet water temperature data 8 is recorded as the data to be detected. When the data to be detected 8 is not the initial abnormal data, centered on the data A to be processed, in the time-series temperature data sequence, the non-initial abnormal data 4 - 7 consecutive with the data to be detected 8 and the non-initial abnormal data C - 10 consecutive with the data A to be processed. At this time, the formed sub-window is (4, 5, 6, 7, 8, A, C, 9, 10).

[0071] (4) When the data to be detected is the initial abnormal data, then the interference value of the data to be processed is set to the first preset value.

[0072] In one embodiment, taking the data A to be processed as an example, if there is an initial abnormal data among the outlet water temperature data B and the outlet water temperature data C adjacent to the data A to be processed, then the initial abnormal data in the outlet water temperature data B and the outlet water temperature data C is recorded as the adjacent abnormal data. Among the two outlet water temperature data adjacent to the adjacent abnormal data, the non-data to be processed is recorded as the data to be detected. The data to be detected is detected. When the data to be detected is the initial abnormal data, the data A to be processed is directly recorded as the abnormal data, and the interference value of the data A to be processed is set to 0.1. There is no limitation here and it can be set according to the specific implementation scenario.

[0073] (5) If two adjacent outlet water temperature data to the data to be processed are both initial abnormal data, set the interference value of the data to be processed to a first preset value.

[0074] In one embodiment, taking the data A to be processed as an example, if the adjacent outlet water temperature data B and the outlet water temperature data C to the data A to be processed are both initial abnormal data, directly mark the data A to be processed as abnormal data, and set the interference value of the data A to be processed to 0.1. There is no limitation here, and it can be set according to the specific implementation scenario.

[0075] (6) Obtain the interference value of each data to be processed in the initial window, and obtain the interference degree of the initial window according to the interference value of each data to be processed in the initial window.

[0076] Specifically, obtain the summation result of the interference values of all the data to be processed in the initial window, denoted as the third summation result, calculate the ratio of the number of data to be processed in the initial window to the number of outlet water temperature data in the initial window, denoted as the second ratio, and calculate the product of the second ratio and the third summation result as the interference degree of the initial window.

[0077] In one embodiment, according to the method for obtaining the interference value of the above-mentioned data A to be processed, obtain the interference values of all the data to be processed in the initial window, and calculate the interference degree of the initial window:

[0078]

[0079] Wherein, is the interference degree of the initial window; M is the number of data to be processed in the initial window; n is the number of outlet water temperature data in the initial window; is the interference value of the j-th data to be processed in the initial window; j is the serial number of the data to be processed in the initial window.

[0080] It should be noted that the more the number of data to be processed in the initial window, the greater the interference value of each data to be processed in the initial window, and the greater the interference degree of the initial window.

[0081] After obtaining the interference degree of the initial window, the best moving average coefficient of the ARIMA prediction model can be obtained according to the interference degree of the initial window and the data fluctuation characteristics in the initial window.

[0082] The method for obtaining the best moving average coefficient of the ARIMA prediction model is as follows:

[0083] (1)Calculate the difference between the constant 1 and the interference value of each data to be processed respectively as the outlier of each data to be processed, set an outlier threshold, and mark the data to be processed in the initial window whose outlier is greater than or equal to the outlier threshold as abnormal data.

[0084] In one embodiment, taking the data to be processed A as an example, calculate the outlier of the data to be processed A:

[0085]

[0086] wherein, is the outlier of the data to be processed A; is the interference value of the data to be processed A.

[0087] It should be noted that the greater the interference value of the data to be processed A, the more the data to be processed A is affected by interference, that is, the abnormal situation of the data to be processed A is more likely to be caused by external interference rather than the real abnormal water outlet temperature. Therefore, the outlier of the data to be processed A is smaller.

[0088] According to the method for obtaining the outlier of the data to be processed A, obtain the outlier of each data to be processed in the initial window, set the outlier threshold to 0.6 (the setting is not limited and can be set according to the specific implementation scenario), and mark the data to be processed in the initial window whose outlier is greater than or equal to 0.6 as abnormal data, so as to obtain all the abnormal data in the initial window.

[0089] (2)Combine the non-initial abnormal data and abnormal data in the initial window to form a reference data sequence, and obtain the fluctuation value of the reference data sequence according to the data fluctuation characteristics in the reference data sequence.

[0090] After obtaining all the abnormal data in the initial window, obtain all the non-initial abnormal data in the initial window, combine all the non-initial abnormal data and all the abnormal data in the initial window to form a reference data sequence, and obtain the fluctuation value of the reference data sequence according to the data fluctuation characteristics in the reference data sequence.

[0091] Among them, the method for obtaining the fluctuation value of the reference data sequence according to the data fluctuation characteristics in the reference data sequence is:

[0092] Calculate the absolute value of the difference between each two adjacent reference data in the reference data sequence respectively, perform normalization processing on the absolute value of the difference between each two adjacent reference data in the reference data sequence to obtain a normalized absolute difference value sequence, and obtain the mean value of the normalized absolute difference value sequence, denoted as the third mean value;

[0093] Form the abnormal data in the initial window into an abnormal data sequence, obtain the variance of the abnormal data sequence, calculate the addition result of the variance and the constant 1 to obtain a fourth addition result, and calculate the difference between the constant 1 and the reciprocal of the fourth addition result, denoted as the first difference;

[0094] Perform a weighted summation process on the third mean and the first difference to obtain the fluctuation value of the reference data sequence.

[0095] In an embodiment, form all the abnormal data in the initial window into an abnormal data sequence, obtain the variance of the abnormal data sequence, and calculate the fluctuation value of the reference data sequence:

[0096]

[0097] where, is the fluctuation value of the reference data sequence; is the k-th reference data in the reference data sequence; is the (k - 1)-th reference data in the reference data sequence; k is the serial number of the reference data in the reference data sequence; m is the number of reference data in the reference data sequence; is the variance of the abnormal data sequence; norm() is a normalization function; | | is an absolute value symbol; is 's weight; is 's weight. In this embodiment, is set to 0.4, is set to 0.6. There is no limitation here and it can be set according to specific implementation scenarios.

[0098] It should be noted that the greater the difference between every two adjacent reference data in the reference data sequence, the greater the fluctuation of the reference data in the reference data sequence, and the greater the fluctuation value of the reference data sequence; the greater the variance of the abnormal data sequence, the greater the fluctuation of the abnormal data in the reference data sequence, and the greater the fluctuation value of the reference data sequence.

[0099] (3) Adjust the initial window according to the fluctuation value of the reference data sequence and the degree of interference of the initial window, correspondingly obtain an adjusted initial window, and use the length of the adjusted initial window as the best moving average coefficient of the ARIMA prediction model.

[0100] Since the window can smooth the noise data, expanding the initial window can reduce the degree of interference of the data in the initial window. However, there are large fluctuations in the abnormal data in the initial window, that is, the short-term changes are obvious. Expanding the initial window is likely to cause the loss of the short-term fluctuation details of the abnormal data. Therefore, the initial window can be adjusted according to the fluctuation value of the reference data sequence and the degree of interference of the initial window. When the degree of interference of the initial window is large, the window is expanded; when the fluctuation value of the reference data sequence in the initial window is large, the window is reduced, and the adjusted initial window is obtained.

[0101] Among them, the specific method of adjusting the initial window according to the fluctuation value of the reference data sequence and the degree of interference of the initial window to obtain the adjusted initial window is as follows:

[0102] a Calculate the difference between the constant 1 and the fluctuation value of the reference data sequence, denoted as the second difference, and obtain the mean value of the degree of interference of the initial window and the second difference as the adjustment coefficient of the initial window.

[0103] In an embodiment, calculate the adjustment coefficient of the initial window:

[0104]

[0105] Among them, is the adjustment coefficient of the initial window; is the degree of interference of the initial window; is the fluctuation value of the reference data sequence.

[0106] It should be noted that the greater the degree of interference of the initial window, the greater the adjustment coefficient of the initial window, indicating that the initial window needs to be expanded; the greater the fluctuation value of the reference data sequence, the smaller the adjustment coefficient of the initial window, indicating that the initial window needs to be reduced.

[0107] b Obtain the adjustment coefficient threshold interval. If the adjustment coefficient of the initial window is within the adjustment coefficient threshold interval, there is no need to adjust the initial window.

[0108] In an embodiment, the obtained adjustment coefficient threshold interval is [0.5, 0.6], which is not limited here and can be set according to the specific implementation scenario. If the adjustment coefficient of the initial window is within [0.5, 0.6], there is no need to adjust the initial window.

[0109] If the adjustment coefficient of the initial window is less than the minimum value of the adjustment coefficient threshold interval, calculate the difference between the minimum value of the adjustment coefficient threshold interval and the adjustment coefficient of the initial window, denoted as the third difference. Calculate the product of the number of outlet water temperature data in the initial window and the third difference, denoted as the first product. Calculate the difference between the number of outlet water temperature data in the initial window and the first product, and use it as the length of the first window after the initial window is reduced. Reduce the length of the initial window to the length of the first window to obtain the adjusted initial window.

[0110] In one embodiment, if the adjustment coefficient of the initial window is less than 0.5, it indicates that the fluctuation value of the reference data sequence in the initial window is too large. At this time, it is necessary to reduce the initial window and calculate the length of the first window after the initial window is reduced:

[0111]

[0112] Where Is is the length of the first window after the initial window is reduced; n is the number of outlet water temperature data in the initial window; is the adjustment coefficient of the initial window.

[0113] It should be noted that the smaller the adjustment coefficient of the initial window, the greater the degree to which the initial window needs to be reduced, and the smaller the length of the first window after the initial window is reduced.

[0114] After obtaining the length of the first window after the initial window is reduced, reduce the length of the initial window to the length of the first window to obtain the adjusted initial window.

[0115] If the adjustment coefficient of the initial window is greater than the maximum value of the adjustment coefficient threshold interval, calculate the difference between the adjustment coefficient of the initial window and the maximum value of the adjustment coefficient threshold interval, denoted as the fourth difference. Calculate the product of the number of outlet water temperature data in the initial window and the fourth difference, denoted as the second product. Calculate the sum of the number of outlet water temperature data in the initial window and the first product, and use it as the length of the second window after the initial window is enlarged. Enlarge the length of the initial window to the length of the second window to obtain the adjusted initial window.

[0116] In one embodiment, if the adjustment coefficient of the initial window is greater than 0.6, it indicates that the degree of interference of the initial window is too large. At this time, it is necessary to enlarge the initial window and calculate the length of the second window after the initial window is enlarged:

[0117]

[0118] Where Ik is the length of the second window after the initial window is enlarged; n is the number of outlet water temperature data in the initial window; is the adjustment coefficient of the initial window.

[0119] It should be noted that the larger the adjustment coefficient of the initial window, the greater the degree to which the initial window needs to be expanded, and the greater the length of the second window after the initial window is expanded.

[0120] After obtaining the length of the second window after the initial window is expanded, expand the length of the initial window to the length of the second window to obtain the adjusted initial window.

[0121] After obtaining the adjusted initial window, use the length of the adjusted initial window as the best moving average coefficient of the ARIMA prediction model.

[0122] Thus, the best moving average coefficient of the ARIMA prediction model is obtained.

[0123] Step S104, according to the autoregressive coefficient, difference coefficient, and best moving average coefficient of the ARIMA prediction model, predict the outlet water temperature of the water cooling system at future times to obtain a predicted value, and perform an anomaly warning on the outlet water temperature of the water cooling system at future times according to the predicted value.

[0124] After obtaining the best moving average coefficient of the ARIMA prediction model, predict the outlet water temperature of the water cooling system at future times according to the obtained difference coefficient, autoregressive coefficient, and best moving average coefficient of the ARIMA prediction model to obtain a predicted value, set an outlet water temperature threshold, and perform an anomaly warning on the outlet water temperature of the water cooling system at future times according to the predicted value of the outlet water temperature of the water cooling system at future times and the outlet water temperature threshold. Among them, the ARIMA prediction model belongs to the prior art and will not be elaborated here.

[0125] Among them, the specific method for performing an anomaly warning on the outlet water temperature of the water cooling system at future times according to the predicted value of the outlet water temperature of the water cooling system at future times and the outlet water temperature threshold is as follows:

[0126] Set the outlet water temperature threshold. If the predicted value is greater than or equal to the outlet water temperature threshold, perform a warning process on the outlet water temperature of the water cooling system at future times.

[0127] In an embodiment, set the outlet water temperature threshold to 1°C. If the predicted value of the outlet water temperature of the water cooling system at future times is greater than or equal to 1°C, perform an anomaly warning on the outlet water temperature of the water cooling system at future times. If the predicted value of the outlet water temperature of the water cooling system at future times is less than 1°C, it means that the outlet water temperature of the water cooling system at future times is normal and no anomaly warning is performed on it.

[0128] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. The online monitoring and early warning method of water cooling system based on Internet of Things is characterized by: The method comprises: Collect the outlet water temperature of the water cooling system during operation within a preset time range to obtain a time series temperature data sequence; Performing time series analysis on the time series temperature data sequence, obtaining the differential coefficient and the autoregressive coefficient when predicting the outlet water temperature of the water cooling system at a future time using the ARIMA prediction model, and constructing an initial window corresponding to the initial sliding average coefficient of the ARIMA prediction model in the time series temperature data sequence according to the autoregressive coefficient of the ARIMA prediction model; According to the data fluctuation characteristics in the time series temperature data sequence, initial abnormal data in the time series temperature data sequence is obtained, according to the recoverability and continuity of the initial abnormal data in the initial window in the time series temperature data sequence, the interference degree of the initial window is obtained, and according to the interference degree of the initial window and the data fluctuation characteristics in the initial window, the optimal sliding average coefficient of the ARIMA forecasting model is obtained; According to the autoregressive coefficient, the differential coefficient, and the optimal sliding average coefficient of the ARIMA prediction model, the outlet water temperature of the water cooling system at a future time is predicted to obtain a predicted value, and according to the predicted value, an abnormal warning is issued for the outlet water temperature of the water cooling system at a future time; The obtaining the interference degree of the initial window according to the recoverability and continuity of the initial abnormal data in the initial window in the time series temperature data sequence includes: Record any initial abnormal data in the initial window as data to be processed, and detect whether two water outlet temperature data adjacent to the data to be processed are initial abnormal data in the time series temperature data sequence; If the two water outlet temperature data adjacent to the data to be processed are not initial abnormal data, then the data to be processed is taken as the center, and the continuous non-initial abnormal data adjacent to the data to be processed and the data to be processed are used to form a subwindow of the data to be processed, and the disturbed value of the data to be processed is obtained according to the data fluctuation characteristics in the subwindow; If there is an initial abnormal data in the two water outlet temperature data adjacent to the data to be processed, the initial abnormal data adjacent to the data to be processed is recorded as adjacent abnormal data, and the non-to-be-processed data adjacent to the adjacent abnormal data is recorded as the data to be detected, and it is detected whether the data to be detected is the initial abnormal data. When the data to be detected is not the initial abnormal data, with the data to be processed as the center, the continuous non-initial abnormal data adjacent to the data to be processed, the continuous non-initial abnormal data adjacent to the adjacent abnormal data, and the data to be processed are used to form a subwindow of the data to be processed, and according to the data fluctuation characteristics in the subwindow, the disturbed value of the data to be processed is obtained; When the data to be detected is initial abnormal data, setting the interference value of the data to be processed to a first preset value; If two water outlet temperature data adjacent to the data to be processed are both initial abnormal data, setting the interference value of the data to be processed to a first preset value; The interference value of each to-be-processed data in the initial window is obtained, and the interference degree of the initial window is obtained according to the interference value of each to-be-processed data in the initial window.

2. The online monitoring and early warning method for a water cooling system based on the Internet of Things according to claim 1 is characterized in that: The step of constructing an initial window corresponding to the initial sliding average coefficient of the ARIMA prediction model in the time series temperature data sequence according to the autoregressive coefficient of the ARIMA prediction model includes: Taking the autoregressive coefficient of the ARIMA prediction model as the window length, an initial window including the last outlet water temperature data in the time series temperature data sequence is constructed as the initial window corresponding to the initial sliding average coefficient of the ARIMA prediction model.

3. The online monitoring and early warning method for a water cooling system based on the Internet of Things according to claim 1 is characterized in that: The step of acquiring initial abnormal data in the time series temperature data sequence according to the data fluctuation characteristics in the time series temperature data sequence comprises: The upper edge value in the time series temperature data sequence is obtained through the box plot of the time series temperature data sequence, and the outlet water temperature data in the time series temperature data sequence that is greater than the upper edge value is used as the initial abnormal data in the time series temperature data sequence.

4. The online monitoring and early warning method for a water cooling system based on the Internet of Things according to claim 1 is characterized in that: The obtaining, according to the data fluctuation characteristics in the sub-window, the disturbed value of the data to be processed comprises: Taking the data to be processed as the last water outlet temperature data, a small window is established in the subwindow, the water outlet temperature data in the small window is fitted to obtain a fitting straight line for the data to be processed, and the fitting straight line of each water outlet temperature data except the first water outlet temperature data in the subwindow is obtained respectively, and a corresponding fitting straight line sequence is obtained, and the correlation coefficient between each fitting straight line except the first fitting straight line in the fitting straight line sequence and the previous fitting straight line is calculated respectively to form a correlation coefficient sequence, and the correlation coefficient sequence is fitted to obtain a fitting curve of the correlation coefficient; Obtaining the derivative of the correlation coefficient corresponding to the data to be processed on the fitting curve, calculating the addition result of the derivative and constant 1, recording it as a first addition result, obtaining the number of extreme points on the fitting curve, calculating the absolute value of the difference between constant 1 and the number of extreme points on the fitting curve, recording it as a first absolute value of the difference, calculating the addition result of the first absolute value of the difference and a second preset value, recording it as a second addition result, obtaining the average of the reciprocal of the first addition result and the reciprocal of the second addition result as the recoverability probability of the data to be processed; On the fitting curve, obtain the correlation coefficient corresponding to the data to be processed, obtain the mean between the correlation coefficients corresponding to the previous water outlet temperature data and the next water outlet temperature data of the data to be processed, record it as a first mean, calculate the absolute value of the difference between the correlation coefficient corresponding to the data to be processed and the first mean, record it as a second absolute value of the difference, calculate the difference between a constant 1 and the reciprocal of the second absolute value of the difference, as the influence degree of the data to be processed; Obtaining the number of initial abnormal data in the initial window, respectively obtaining the number of outlet water temperature data between each initial abnormal data and the data to be processed in the initial window to form a number sequence, obtaining the minimum number in the number sequence, calculating the ratio of the number of initial abnormal data in the initial window to the minimum number, recording it as a first ratio, normalizing the first ratio, and obtaining the continuity frequency of the data to be processed; The average values ​​of the recoverability probability, the impact degree and the continuity frequency of the data to be processed are calculated as the interference value of the data to be processed.

5. The online monitoring and early warning method for a water cooling system based on the Internet of Things according to claim 1 is characterized in that: The obtaining the interference degree of the initial window according to the interference degree of each to-be-processed data in the initial window comprises: Obtain the sum of the interference values ​​of all the data to be processed in the initial window, recorded as the third addition result, calculate the ratio of the number of data to be processed in the initial window to the number of water outlet temperature data in the initial window, recorded as the second ratio, and calculate the product of the second ratio and the third addition result as the interference degree of the initial window.

6. The online monitoring and early warning method for a water cooling system based on the Internet of Things according to claim 1 is characterized in that: The step of obtaining the optimal sliding average coefficient of the ARIMA forecasting model according to the interference degree of the initial window and the data fluctuation characteristics in the initial window includes: Calculate the difference between the constant 1 and the disturbed value of each to-be-processed data respectively as the abnormal value of each to-be-processed data, set the abnormal value threshold, and record the to-be-processed data whose abnormal value in the initial window is greater than or equal to the abnormal value threshold as abnormal data; The non-initial abnormal data and the abnormal data in the initial window form a reference data sequence, and according to the data fluctuation characteristics in the reference data sequence, obtain the fluctuation value of the reference data sequence; According to the fluctuation value of the reference data sequence and the interference degree of the initial window, the initial window is adjusted to obtain an adjusted initial window, and the length of the adjusted initial window is used as the optimal sliding average coefficient of the ARIMA prediction model.

7. The online monitoring and early warning method for a water cooling system based on the Internet of Things according to claim 6 is characterized in that: The obtaining the fluctuation value of the reference data sequence according to the data fluctuation feature in the reference data sequence includes: Respectively calculating the absolute value of the difference between every two adjacent reference data in the reference data sequence, normalizing the absolute value of the difference between every two adjacent reference data in the reference data sequence to obtain a normalized difference absolute value sequence, and obtaining a mean of the normalized difference absolute value sequence, which is recorded as a third mean; The abnormal data in the initial window are combined into an abnormal data sequence, the variance of the abnormal data sequence is obtained, the sum of the variance and a constant 1 is calculated to obtain a fourth sum result, and the difference between the constant 1 and the reciprocal of the fourth sum result is calculated and recorded as a first difference; A weighted summation process is performed on the third mean and the first difference to obtain a fluctuation value of the reference data sequence.

8. The online monitoring and early warning method for a water cooling system based on the Internet of Things according to claim 6 is characterized in that: The adjusting the initial window according to the fluctuation value of the reference data sequence and the interference degree of the initial window to obtain the adjusted initial window includes: Calculate the difference between the constant 1 and the fluctuation value of the reference data sequence, record it as a second difference, and obtain the average of the interference degree of the initial window and the second difference as the adjustment coefficient of the initial window; Acquire an adjustment coefficient threshold interval, and if the adjustment coefficient of the initial window is within the adjustment coefficient threshold interval, there is no need to adjust the initial window; If the adjustment coefficient of the initial window is less than the minimum value of the adjustment coefficient threshold interval, the difference between the minimum value of the adjustment coefficient threshold interval and the adjustment coefficient of the initial window is calculated, recorded as the third difference, the product of the number of outlet water temperature data in the initial window and the third difference is calculated, recorded as the first product, the difference between the number of outlet water temperature data in the initial window and the first product is calculated as the first window length after the initial window is reduced, the length of the initial window is reduced to the first window length, and the adjusted initial window is obtained; If the adjustment coefficient of the initial window is greater than the maximum value of the adjustment coefficient threshold interval, the difference between the adjustment coefficient of the initial window and the maximum value of the adjustment coefficient threshold interval is calculated, recorded as the fourth difference, the product of the number of outlet water temperature data in the initial window and the fourth difference is calculated, recorded as the second product, the sum of the number of outlet water temperature data in the initial window and the first product is calculated as the second window length after the initial window is expanded, the length of the initial window is expanded to the second window length, and the adjusted initial window is obtained.

9. The online monitoring and early warning method for a water cooling system based on the Internet of Things according to claim 8 is characterized in that: The step of providing an abnormal warning for the outlet water temperature of the water cooling system at a future time according to the predicted value includes: A water outlet temperature threshold is set, and if the predicted value is greater than or equal to the water outlet temperature threshold, an early warning process is performed on the water outlet temperature of the water cooling system at a future time.

Images