A valve optimization control method and system based on machine learning

Abstract

This invention belongs to the field of valve control technology. It provides a valve optimization control method and system based on machine learning, comprising: performing anomaly analysis on the valve's opening and closing response time when fluid is present to determine whether the valve's opening and closing response time is abnormal; if abnormal, analyzing the possibility of the instantaneous flow velocity affecting the valve's opening and closing response time by combining the instantaneous flow velocity corresponding to the valve's opening and closing response time; if the possibility of an effect is high, determining whether the effect of the instantaneous flow velocity on the valve's opening and closing response time is regular; constructing a prediction model and judging the prediction accuracy of the prediction model; determining the opening and closing response time adjustment value based on the prediction accuracy judgment result, and adjusting the valve's opening and closing response time. This valve uses the opening and closing response time adjustment value to quantitatively guide flow velocity control, and further drives the model's adaptive evolution through dynamic feedback of monitoring data.

Description

Technical Field

[0001] This invention belongs to the field of valve control technology, specifically a valve optimization control method and system based on machine learning. Background Technology

[0002] In industrial process control, valves are key actuators in fluid systems, and the stability of their opening and closing response time directly affects the safety and efficiency of the system.

[0003] However, existing technologies have significant shortcomings in valve response time control: insufficient multi-dimensional data correlation analysis capabilities, making it difficult to locate the fluid dynamic causes of response time anomalies by combining instantaneous flow velocity; lack of dynamic adaptation mechanism in prediction models, resulting in a significant decrease in accuracy when facing equipment aging or changes in operating conditions; and reliance on human experience in adjustment strategies, failing to achieve refined control based on the relationship between valve opening and closing response time and instantaneous flow velocity.

[0004] Therefore, the present invention provides a valve optimization control method and system based on machine learning. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] In a first aspect, the present invention provides a valve optimization control method based on machine learning, comprising:

[0007] Based on the valve opening and closing response time in historical time periods, an opening and closing response time analysis library is constructed. By performing anomaly analysis on the opening and closing response time when there is fluid, it is determined whether the valve opening and closing response time is abnormal.

[0008] If the valve's opening and closing response time is abnormal, analyze the possibility that the instantaneous flow velocity has an impact on the valve's opening and closing response time by combining the instantaneous flow velocity corresponding to the valve's opening and closing response time. If the instantaneous flow velocity has a high probability of having an impact on the opening and closing response time, determine whether the impact of the instantaneous flow velocity on the valve's opening and closing response time is regular.

[0009] Based on the analysis results of whether the influence of instantaneous flow velocity on valve opening and closing response time is regular, a prediction model is constructed, the range of instantaneous flow velocity for regulation is output, and the prediction accuracy of the prediction model is judged.

[0010] Based on the prediction accuracy of the prediction model, the adjustment value for the opening and closing response time is determined, and the opening and closing response time of the valve is adjusted accordingly.

[0011] Preferably, the specific process for determining whether the valve's opening and closing response time is abnormal is as follows:

[0012] The historical period is divided into several historical time points according to equal time intervals. The opening and closing response times of valves at all historical time points are exported from the opening and closing response time analysis library. The opening and closing response times of all historical time points within the historical period are analyzed to determine the abnormal response ratio. If it is greater than or equal to the abnormal response ratio threshold, it indicates that the valve's opening and closing response time is abnormal.

[0013] Preferably, the process for determining the abnormal response ratio is as follows:

[0014] The valve opening and closing response time at a historical time point is subtracted from the standard value of the opening and closing response time, and the absolute value of the difference is taken to obtain the response time deviation value. If it is greater than or equal to the standard value of the response time deviation, the corresponding historical time point is recorded as the response abnormal time point.

[0015] Calculate the percentage of abnormal response times among all historical time points, and denote it as the abnormal response ratio.

[0016] Preferably, the specific process for analyzing the possibility of the instantaneous flow velocity affecting the valve opening and closing response time is as follows:

[0017] Extract the instantaneous flow velocity corresponding to the opening and closing response time at the abnormal response time point, and use the Spearman coefficient formula to calculate the monotonic correlation value between the valve opening and closing response time and the instantaneous flow velocity by taking the absolute value. If it is greater than or equal to the monotonic correlation value threshold, it indicates that the instantaneous flow velocity is likely to have an impact on the opening and closing response time.

[0018] Preferably, the specific process for determining whether the influence of instantaneous flow velocity on the valve opening and closing response time is regular is as follows:

[0019] The least squares method was used to fit the opening and closing response time and the corresponding instantaneous flow velocity at the abnormal response time point. The slope and intercept of the fitted line were calculated to construct a fitting model of the opening and closing response time and instantaneous flow velocity.

[0020] The goodness-of-fit formula is used to calculate the regularity characterization value of the fitted straight line between the opening and closing response time and the instantaneous flow velocity. If it is greater than or equal to the regularity characterization threshold, it indicates that the influence of the instantaneous flow velocity on the opening and closing response time is linear.

[0021] Preferably, the process for determining the instantaneous flow velocity range is as follows:

[0022] The standard value of the valve opening and closing response time is input into the prediction model to obtain the instantaneous flow velocity when the valve opening and closing response time reaches the standard value. This instantaneous flow velocity is recorded as the arrival instantaneous flow velocity. The difference between this instantaneous flow velocity and the current instantaneous flow velocity is processed to obtain the regulating instantaneous flow velocity. The interval between the regulating instantaneous flow velocities is recorded as the regulating instantaneous flow velocity interval.

[0023] Preferably, the process of judging the prediction accuracy of the prediction model is as follows:

[0024] Within the range of instantaneous flow rate adjustment, a monitoring range is preset. The opening and closing response time of the valve and the instantaneous flow rate are monitored in real time within the monitoring range. By comparing and analyzing the different instantaneous flow rates monitored in real time within the monitoring range with the predicted instantaneous flow rate, the asynchronous ratio value is determined. If it is greater than or equal to the asynchronous ratio threshold, it indicates that the prediction model is accurate; otherwise, it indicates that the prediction model is inaccurate.

[0025] Preferably, the process for determining the asynchronous proportional value is as follows:

[0026] Based on any instantaneous flow velocity within a preset monitoring interval; if the instantaneous flow velocity within the preset monitoring interval is subtracted from the predicted instantaneous flow velocity, and the absolute value is taken, the relative deviation of the instantaneous flow velocity is obtained;

[0027] The relative deviation of instantaneous flow velocity is compared with the standard value of the relative deviation of instantaneous flow velocity. If it is greater than or equal to the standard value of the relative deviation of instantaneous flow velocity, the corresponding predicted instantaneous flow velocity is recorded as asynchronous instantaneous flow velocity. The proportion of asynchronous instantaneous flow velocities in the monitoring interval is calculated to obtain the asynchronous ratio value.

[0028] Preferably, the process for determining the opening / closing response time adjustment value is as follows:

[0029] If the prediction model is inaccurate, the instantaneous flow velocity and corresponding opening and closing response time monitored in the monitoring interval are added to the training dataset of the prediction model to correct the prediction model, and the adjustment instantaneous flow velocity interval is re-output and used as the analysis interval.

[0030] If the prediction model is accurate, the output instantaneous flow velocity range of the already trained prediction model will be used as the analysis range.

[0031] Extract the predicted instantaneous flow velocity and the corresponding opening and closing response time within the analysis interval, and perform fitting analysis again;

[0032] If it conforms to a linear law, the slope of the fitted line will be used as an adjustment coefficient.

[0033] If the linearity is not met, the analysis interval is divided into several analysis segments with equal time intervals. The difference between the maximum and minimum opening and closing response times within each analysis segment is calculated to obtain the segment time difference. The difference between the instantaneous velocity corresponding to the maximum and minimum opening and closing response times within each analysis segment is calculated to obtain the instantaneous velocity difference. The ratio of the instantaneous velocity difference to the segment time difference is calculated to obtain the rate of change. The rates of change of all analysis segments are integrated into a rate of change sequence, and the maximum rate of change in the rate of change sequence is extracted as an adjustment coefficient.

[0034] The absolute value of the difference between the current instantaneous flow velocity and the predicted instantaneous flow velocity is taken to obtain the instantaneous flow velocity change value. The instantaneous flow velocity change value is then multiplied by the adjustment coefficient to obtain the opening and closing response time adjustment value.

[0035] Secondly, the present invention also provides a valve optimization control system based on machine learning, the system comprising:

[0036] Anomaly detection module: Based on the valve opening and closing response time in historical time periods, an opening and closing response time analysis library is built. By performing anomaly analysis on the opening and closing response time when there is fluid, it is determined whether the valve opening and closing response time is abnormal.

[0037] Regularity Analysis Module: If the valve opening and closing response time is abnormal, combine the instantaneous flow velocity corresponding to the valve opening and closing response time to analyze the possibility that the instantaneous flow velocity has an impact on the valve opening and closing response time. If the instantaneous flow velocity has a high probability of having an impact on the opening and closing response time, determine whether the impact of the instantaneous flow velocity on the valve opening and closing response time is regular.

[0038] Model optimization module: Based on the analysis results of whether the influence of instantaneous flow velocity on valve opening and closing response time is regular, a prediction model is constructed, the range of instantaneous flow velocity for regulation is output, and the prediction accuracy of the prediction model is judged.

[0039] Control and Analysis Module: Based on the prediction accuracy of the prediction model, determine the adjustment value of the opening and closing response time, and adjust the opening and closing response time of the valve.

[0040] The beneficial effects of this invention are as follows:

[0041] 1. This invention ensures data reliability by constructing a historical data-driven opening and closing response time analysis library. It uses timestamps to map instantaneous flow velocity and response time to achieve multi-dimensional data correlation analysis. Based on historical experience, it sets standard values ​​and quantifies the degree of abnormality through abnormal response ratio, which can effectively identify the overall abnormal state of valve response time. It introduces Spearman coefficient analysis to analyze monotonic correlation, which can accurately determine the possibility of flow velocity affecting response time. By fitting a linear model using the least squares method and combining it with goodness-of-fit test, it can further clarify the regularity of flow velocity influence. The slope of the fitted line provides a quantitative basis for valve opening and closing time adjustment.

[0042] 2. This invention intelligently selects linear regression or machine learning models based on the influence of flow velocity, and significantly improves the accuracy of response time prediction by combining training set-validation set partitioning and mean square error optimization. By setting a monitoring interval with an interval length smaller than the adjustment interval, it achieves quantitative evaluation of the convergence of predicted values ​​and real-time flow velocity. The asynchronous proportional threshold comparison mechanism effectively identifies model deviations and provides clear triggering conditions for subsequent model iterations. Furthermore, it innovatively proposes a dynamic adjustment coefficient calculation strategy, which supports the direct application of slope in linear scenarios and achieves refined control under complex working conditions by extracting the segment change rate in nonlinear scenarios. It not only guides flow velocity control through quantitative adjustment values ​​of opening and closing response time, but also drives the adaptive evolution of the model through dynamic feedback of monitoring data, significantly improving the intelligence, response speed, and long-term stability of the valve control system. Attached Figure Description

[0043] The invention will now be further described with reference to the accompanying drawings.

[0044] Figure 1 This is a flowchart illustrating the steps of a valve optimization control method based on machine learning according to an embodiment of the present invention.

[0045] Figure 2 This is a system block diagram of a valve optimization control system based on machine learning according to an embodiment of the present invention. Detailed Implementation

[0046] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0047] Example 1

[0048] See also Figure 1 As shown in the figure, a valve optimization control method based on machine learning according to an embodiment of the present invention includes the following steps:

[0049] Step 1: Based on the valve opening and closing response time in historical time periods, construct an opening and closing response time analysis library. By performing anomaly analysis on the opening and closing response time when there is fluid, determine whether the valve opening and closing response time is abnormal.

[0050] It should be noted that the historical period includes, but is not limited to, three months, six months, and one year. The opening and closing response time refers to the complete action time of the valve from fully open to fully closed or from fully closed to fully open.

[0051] Valve opening and closing response times over historical periods are obtained from data acquisition and monitoring control systems and distributed control systems. Instantaneous flow velocity data during valve opening and closing responses are collected using flow velocity sensors. The collected data is preprocessed, including using the IQR method to remove outliers, interpolating to complete missing data, and using Min-Max normalization. Valve opening and closing response times are mapped to instantaneous flow velocities using timestamps. The opening and closing response times and instantaneous flow velocity data are stored in a MySQL database to build an opening and closing response time analysis library.

[0052] Understandably, data is collected at a fixed sampling frequency for a fixed fluid type and valve type within a historical period.

[0053] The historical period is divided into several historical time points according to equal time intervals. The opening and closing response times of valves at all historical time points are exported from the opening and closing response time analysis library. The opening and closing response times of valves at historical time points are subtracted from the standard values ​​of opening and closing response times, and the absolute value of the difference is taken to obtain the response time deviation value.

[0054] Compare the response time deviation value with the standard response time deviation value:

[0055] If the response time deviation is greater than or equal to the standard response time deviation, the corresponding historical time point will be recorded as the response anomaly time point.

[0056] If the response time deviation is less than the standard response time deviation, the corresponding historical time point is recorded as the normal response time point.

[0057] It is understandable that the abnormal response time point indicates that the valve's opening and closing response time deviates from the standard value, reflecting an abnormal situation in the valve's opening and closing response time. The normal response time point indicates that the valve's opening and closing response time does not deviate from the standard value, reflecting a normal situation in the valve's opening and closing response time. The standard value of the opening and closing response time and the standard value of the response time deviation are both set by those skilled in the art based on historical experience.

[0058] The number of response anomalies is counted among all historical time points, and the percentage of response anomalies among all historical time points is calculated and denoted as the anomaly response ratio.

[0059] In some embodiments, the anomaly response ratio is compared with an anomaly response ratio threshold:

[0060] If the abnormal response ratio is greater than or equal to the abnormal response ratio threshold, it indicates that the valve's opening and closing response time is abnormal.

[0061] If the abnormal response ratio is less than the abnormal response ratio threshold, it indicates that the valve's opening and closing response time is normal.

[0062] Step 2: If the valve opening and closing response time is abnormal, analyze the possibility that the instantaneous flow velocity has an impact on the valve opening and closing response time by combining the instantaneous flow velocity corresponding to the valve opening and closing response time. If the instantaneous flow velocity has a high probability of having an impact on the opening and closing response time, determine whether the impact of the instantaneous flow velocity on the valve opening and closing response time is regular.

[0063] Extract the instantaneous flow velocity corresponding to the opening and closing response time at the abnormal response time point, using the Spearman coefficient formula: The absolute value is used to calculate the monotonic correlation between the valve opening / closing response time and the instantaneous flow velocity, where, This represents the monotonic correlation between the valve opening / closing response time and the instantaneous flow rate, where n represents the sample size. It represents the difference between the rank of the instantaneous flow velocity and the rank of the opening and closing response time in the i-th sample;

[0064] If the monotonic correlation value between the valve opening / closing response time and the instantaneous flow velocity is greater than or equal to the monotonic correlation value threshold, it indicates that the instantaneous flow velocity is likely to have an impact on the opening / closing response time.

[0065] If the monotonic correlation value between the valve opening / closing response time and the instantaneous flow velocity is less than the monotonic correlation value threshold, it indicates that the instantaneous flow velocity is unlikely to have an impact on the opening / closing response time.

[0066] Given that instantaneous flow velocity is likely to have a significant impact on opening and closing response time, the least squares method is used to fit the opening and closing response time at abnormal response time points with the corresponding instantaneous flow velocity. The fitted line equation is y=kx+b, where k represents the slope of the fitted line, b represents the intercept of the fitted line, x represents the instantaneous flow velocity, and y represents the opening and closing response time.

[0067] By formula: Calculate the slope k of the fitted line, where, This represents the i-th instantaneous flow velocity sample value. This represents the average value of the instantaneous flow velocity samples. This represents the sample value of the opening and closing response time for the i-th time. This represents the average value of the sample opening and closing response times; expressed by the formula: Calculate the intercept of the fitted line;

[0068] It should be noted that because the timestamp maps the valve opening and closing response time to the instantaneous flow rate, therefore and The meaning of 'i' in the text is the same;

[0069] Based on the slope and intercept of the fitted line obtained from the calculation, a fitting model of opening and closing response time and instantaneous flow velocity is constructed.

[0070] Using the goodness-of-fit formula The regularity characterization value of the fitted straight line between the opening and closing response time and the instantaneous flow velocity was calculated. In the formula, SSE represents the residual sum of squares, and SST represents the total sum of squares;

[0071] It is understandable that the residual sum of squares represents the error between the predicted value and the actual value of the fitted model, while the total sum of squares represents the total degree of variation of the data itself.

[0072] Compare the regularity representation value with the regularity representation threshold:

[0073] If the regularity value is greater than or equal to the regularity threshold, it indicates that the effect of instantaneous flow velocity on opening and closing response time follows a linear law.

[0074] If the regularity value is less than the regularity threshold, it indicates that the effect of instantaneous flow velocity on the opening and closing response time is nonlinear.

[0075] It should be noted that the necessity of using the least squares method to fit the opening and closing response time and instantaneous flow velocity at the abnormal response time point is that the slope of the fitted line can clarify the unit influence of the valve's instantaneous flow velocity on the opening and closing response time, providing reliable support for subsequent adjustment of the valve's opening and closing time.

[0076] The technical solution of this embodiment is as follows: This invention ensures data reliability by constructing a historical data-driven opening and closing response time analysis library, and uses timestamps to map instantaneous flow velocity and response time to achieve multi-dimensional data correlation analysis; based on historical experience, standard values ​​are set and the degree of abnormality is quantified by the abnormal response ratio, which can effectively identify the overall abnormal state of valve response time; Spearman coefficient is introduced to analyze monotonic correlation, which can accurately determine the possibility of flow velocity affecting response time; by fitting a linear model using the least squares method and combining it with goodness-of-fit test, the regularity of flow velocity influence can be further clarified, and the slope of the fitted line provides a quantitative basis for valve opening and closing time adjustment.

[0077] Example 2

[0078] See also Figure 1 As shown in the figure, a valve optimization control method based on machine learning according to an embodiment of the present invention includes the following steps:

[0079] Step 3: Based on the analysis results of whether the influence of instantaneous flow velocity on valve opening and closing response time is regular, construct a prediction model, output the range of instantaneous flow velocity for regulation, and analyze it to judge the prediction accuracy of the prediction model;

[0080] The opening and closing response times and corresponding instantaneous flow velocities at all abnormal response time points are integrated into a dataset, which is then divided into a training set (70%) and a validation set (30%) according to time order. If the influence of instantaneous flow velocity on the opening and closing response time is linear, a prediction model is constructed based on a linear regression model. If the influence of instantaneous flow velocity on the opening and closing response time is nonlinear, a prediction model is constructed based on a machine learning model. The selected prediction model is trained using the training dataset to obtain the trained prediction model. The trained prediction model is then validated using the validation set. Mean squared error is used as the loss function, and Adam is used as the optimizer.

[0081] Input the standard value of valve opening and closing response time into the prediction model to obtain the instantaneous flow velocity when the valve opening and closing response time reaches the standard value, which is recorded as the arrival instantaneous flow velocity. The difference between the arrival instantaneous flow velocity and the current instantaneous flow velocity is processed to obtain the regulating instantaneous flow velocity. The interval between the regulating instantaneous flow velocities is recorded as the regulating instantaneous flow velocity interval.

[0082] Within the range of instantaneous flow rate adjustment, a monitoring range is preset. Within the monitoring range, the valve opening and closing response time and instantaneous flow rate are monitored in real time. By comparing and analyzing the different instantaneous flow rates monitored in real time within the monitoring range with the predicted instantaneous flow rate, the degree of convergence between the instantaneous flow rate monitored in real time within the range of instantaneous flow rate adjustment and the predicted instantaneous flow rate is determined.

[0083] Among them, the interval length corresponding to the preset monitoring interval is less than the interval length corresponding to the instantaneous flow velocity adjustment interval;

[0084] It should be noted that the purpose of setting a monitoring interval within the instantaneous flow velocity range is to analyze the accuracy of the prediction by monitoring in real time whether the instantaneous flow velocity within the preset monitoring interval is close to the predicted instantaneous flow velocity.

[0085] Based on any instantaneous flow velocity within a preset monitoring interval; if the instantaneous flow velocity within the preset monitoring interval is subtracted from the predicted instantaneous flow velocity, and the absolute value is taken, the relative deviation of the instantaneous flow velocity is obtained;

[0086] The relative deviation of instantaneous flow velocity is compared with the standard value of the relative deviation of instantaneous flow velocity. If the relative deviation of instantaneous flow velocity is greater than or equal to the standard value of the relative deviation of instantaneous flow velocity, the corresponding predicted instantaneous flow velocity is recorded as asynchronous instantaneous flow velocity; if the relative deviation of instantaneous flow velocity is less than the standard value of the relative deviation of instantaneous flow velocity, the corresponding predicted instantaneous flow velocity is recorded as synchronous instantaneous flow velocity.

[0087] It should be noted that the standard value for the relative deviation of instantaneous flow velocity was set by those skilled in the art based on historical experience;

[0088] The number of asynchronous instantaneous flow velocities within the monitoring interval is counted, and the proportion of asynchronous instantaneous flow velocities within the monitoring interval is calculated to obtain the asynchronous ratio value;

[0089] In some embodiments, the asynchronous ratio value is compared with the asynchronous ratio threshold;

[0090] If the asynchronous ratio value is greater than or equal to the asynchronous ratio threshold, it means that the instantaneous flow rate change trend of the valve in the monitoring range is similar to the predicted instantaneous flow rate change trend, indicating that the prediction model is accurate.

[0091] If the asynchronous ratio value is less than the asynchronous ratio threshold, it means that the instantaneous flow rate change trend of the valve in the monitoring range is not similar to the predicted instantaneous flow rate change trend, indicating that the prediction model is inaccurate.

[0092] Step 4: Based on the prediction accuracy of the prediction model, determine the adjustment value for the opening and closing response time, and adjust the opening and closing response time of the valve.

[0093] If the prediction model is inaccurate, the instantaneous flow velocity and corresponding opening and closing response time monitored in the monitoring interval are added to the training dataset of the prediction model to correct the prediction model, and the adjustment instantaneous flow velocity interval is re-output and used as the analysis interval.

[0094] If the prediction model is accurate, the output instantaneous flow velocity range of the already trained prediction model will be used as the analysis range.

[0095] Extract the predicted instantaneous flow velocity and the corresponding opening and closing response time within the analysis interval, and perform the current fitting analysis again;

[0096] If it conforms to a linear law, the slope of the fitted line obtained from the fitting analysis will be used as the adjustment coefficient.

[0097] If the linearity is not met, the analysis interval is divided into several analysis segments with equal time intervals. The difference between the maximum and minimum opening and closing response times within each analysis segment is calculated to obtain the segment time difference. The difference between the instantaneous velocity corresponding to the maximum and minimum opening and closing response times within each analysis segment is calculated to obtain the instantaneous velocity difference. The ratio of the instantaneous velocity difference to the segment time difference is calculated to obtain the rate of change. The rates of change of all analysis segments are integrated into a rate of change sequence, and the maximum rate of change in the rate of change sequence is extracted as an adjustment coefficient.

[0098] Based on the adjustment coefficient, the absolute value of the difference between the current instantaneous flow velocity and the predicted instantaneous flow velocity is taken to obtain the instantaneous flow velocity change value. The instantaneous flow velocity change value is then multiplied by the adjustment coefficient to obtain the opening and closing response time adjustment value.

[0099] The valve opening and closing response time is adjusted according to the opening and closing response time adjustment value. An accurate response time can avoid the fluid water hammer effect when the valve is opened and closed.

[0100] The technical solution of this embodiment is as follows: Based on a model with a suitable flow velocity influence law, this invention significantly improves the accuracy of response time prediction by combining training set-validation set partitioning and mean square error optimization; by setting a monitoring interval with an interval length smaller than the adjustment interval, it achieves quantitative evaluation of the convergence of predicted values ​​and real-time flow velocities, and the asynchronous proportional threshold comparison mechanism effectively identifies model deviations, providing clear triggering conditions for subsequent model iterations; further, it innovatively proposes a dynamic adjustment coefficient calculation strategy, which supports the direct application of slope in linear scenarios and achieves refined control under complex working conditions by extracting the segment change rate in nonlinear scenarios; it not only guides flow velocity control through quantitative adjustment values ​​of opening and closing response times, but also drives the model's adaptive evolution through dynamic feedback of monitoring data, significantly improving the intelligence, response speed, and long-term stability of the valve control system.

[0101] Example 3

[0102] See also Figure 2 As shown in the embodiment of the present invention, a valve optimization control system based on machine learning includes the following modules:

[0103] Anomaly detection module: Based on the valve opening and closing response time in historical time periods, an opening and closing response time analysis library is built. By performing anomaly analysis on the opening and closing response time when there is fluid, it is determined whether the valve opening and closing response time is abnormal.

[0104] Regularity Analysis Module: If the valve opening and closing response time is abnormal, combine the instantaneous flow velocity corresponding to the valve opening and closing response time to analyze the possibility that the instantaneous flow velocity has an impact on the valve opening and closing response time. If the instantaneous flow velocity has a high probability of having an impact on the opening and closing response time, determine whether the impact of the instantaneous flow velocity on the valve opening and closing response time is regular.

[0105] Model optimization module: Based on the analysis results of whether the influence of instantaneous flow velocity on valve opening and closing response time is regular, a prediction model is constructed, the range of instantaneous flow velocity for regulation is output, and the prediction accuracy of the prediction model is judged.

[0106] Control and Analysis Module: Based on the prediction accuracy of the prediction model, determine the adjustment value of the opening and closing response time, and adjust the opening and closing response time of the valve.

[0107] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A valve optimization control method based on machine learning, characterized in that: include: Based on the valve opening and closing response time in historical time periods, an opening and closing response time analysis library is constructed. By performing anomaly analysis on the opening and closing response time when there is fluid, it is determined whether the valve opening and closing response time is abnormal. If the valve's opening and closing response time is abnormal, analyze the possibility that the instantaneous flow velocity has an impact on the valve's opening and closing response time by combining the instantaneous flow velocity corresponding to the valve's opening and closing response time. If the instantaneous flow velocity has a high probability of having an impact on the opening and closing response time, determine whether the impact of the instantaneous flow velocity on the valve's opening and closing response time is regular. Based on the analysis results of whether the influence of instantaneous flow velocity on valve opening and closing response time is regular, a prediction model is constructed, the range of instantaneous flow velocity for regulation is output, and the prediction accuracy of the prediction model is judged. Based on the prediction accuracy of the prediction model, determine the adjustment value for the opening and closing response time, and adjust the opening and closing response time of the valve. The process of judging the prediction accuracy of the prediction model is as follows: Within the range of instantaneous flow rate adjustment, a monitoring range is preset. The opening and closing response time of the valve and the instantaneous flow rate are monitored in real time within the monitoring range. By comparing and analyzing the different instantaneous flow rates monitored in real time within the monitoring range with the predicted instantaneous flow rate, the asynchronous ratio value is determined. If it is greater than or equal to the asynchronous ratio threshold, it indicates that the prediction model is accurate; otherwise, it indicates that the prediction model is inaccurate. The process for determining the opening and closing response time adjustment value is as follows: If the prediction model is inaccurate, the instantaneous flow velocity and corresponding opening and closing response time monitored in the monitoring interval are added to the training dataset of the prediction model to correct the prediction model, and the adjustment instantaneous flow velocity interval is re-output and used as the analysis interval. If the prediction model is accurate, the output instantaneous flow velocity range of the already trained prediction model will be used as the analysis range. Extract the predicted instantaneous flow velocity and the corresponding opening and closing response time within the analysis interval, and perform fitting analysis again; If it conforms to a linear law, the slope of the fitted line will be used as an adjustment coefficient. If the linearity is not met, the analysis interval is divided into several analysis segments with equal time intervals. The difference between the maximum and minimum opening and closing response times within each analysis segment is calculated to obtain the segment time difference. The difference between the instantaneous velocity corresponding to the maximum and minimum opening and closing response times within each analysis segment is calculated to obtain the instantaneous velocity difference. The ratio of the instantaneous velocity difference to the segment time difference is calculated to obtain the rate of change. The rates of change of all analysis segments are integrated into a rate of change sequence, and the maximum rate of change in the rate of change sequence is extracted as an adjustment coefficient. The absolute value of the difference between the current instantaneous flow velocity and the predicted instantaneous flow velocity is taken to obtain the instantaneous flow velocity change value. The instantaneous flow velocity change value is then multiplied by the adjustment coefficient to obtain the opening and closing response time adjustment value.

2. The valve optimization control method based on machine learning according to claim 1, characterized in that: The specific process for determining whether the valve's opening and closing response time is abnormal is as follows: The historical period is divided into several historical time points according to equal time intervals. The opening and closing response times of valves at all historical time points are exported from the opening and closing response time analysis library. The opening and closing response times of all historical time points within the historical period are analyzed to determine the abnormal response ratio. If it is greater than or equal to the abnormal response ratio threshold, it indicates that the valve's opening and closing response time is abnormal.

3. The valve optimization control method based on machine learning according to claim 2, characterized in that: The process for determining the abnormal response ratio is as follows: The valve opening and closing response time at a historical time point is subtracted from the standard value of the opening and closing response time, and the absolute value of the difference is taken to obtain the response time deviation value. If it is greater than or equal to the standard value of the response time deviation, the corresponding historical time point is recorded as the response abnormal time point. Calculate the percentage of abnormal response times among all historical time points, and denote it as the abnormal response ratio.

4. The valve optimization control method based on machine learning according to claim 2, characterized in that: The specific process for analyzing the potential impact of instantaneous flow velocity on valve opening and closing response time is as follows: Extract the instantaneous flow velocity corresponding to the opening and closing response time at the abnormal response time point, and use the Spearman coefficient formula to calculate the monotonic correlation value between the valve opening and closing response time and the instantaneous flow velocity by taking the absolute value. If it is greater than or equal to the monotonic correlation value threshold, it indicates that the instantaneous flow velocity is likely to have an impact on the opening and closing response time.

5. The valve optimization control method based on machine learning according to claim 2, characterized in that: The specific process for determining whether the effect of instantaneous flow velocity on valve opening and closing response time is regular is as follows: The least squares method was used to fit the opening and closing response time and the corresponding instantaneous flow velocity at the abnormal response time point. The slope and intercept of the fitted line were calculated to construct a fitting model of the opening and closing response time and instantaneous flow velocity. The goodness-of-fit formula is used to calculate the regularity characterization value of the fitted straight line between the opening and closing response time and the instantaneous flow velocity. If it is greater than or equal to the regularity characterization threshold, it indicates that the influence of the instantaneous flow velocity on the opening and closing response time is linear.

6. The valve optimization control method based on machine learning according to claim 3, characterized in that: The process for determining the instantaneous flow velocity range is as follows: The standard value of the valve opening and closing response time is input into the prediction model to obtain the instantaneous flow velocity when the valve opening and closing response time reaches the standard value. This instantaneous flow velocity is recorded as the arrival instantaneous flow velocity. The difference between this instantaneous flow velocity and the current instantaneous flow velocity is processed to obtain the regulating instantaneous flow velocity. The interval between the regulating instantaneous flow velocities is recorded as the regulating instantaneous flow velocity interval.

7. The valve optimization control method based on machine learning according to claim 1, characterized in that: The process for determining the asynchronous proportional value is as follows: Based on any instantaneous flow velocity within a preset monitoring interval; if the instantaneous flow velocity within the preset monitoring interval is subtracted from the predicted instantaneous flow velocity, and the absolute value is taken, the relative deviation of the instantaneous flow velocity is obtained; The relative deviation of instantaneous flow velocity is compared with the standard value of the relative deviation of instantaneous flow velocity. If it is greater than or equal to the standard value of the relative deviation of instantaneous flow velocity, the corresponding predicted instantaneous flow velocity is recorded as asynchronous instantaneous flow velocity. The proportion of asynchronous instantaneous flow velocities in the monitoring interval is calculated to obtain the asynchronous ratio value.

8. A valve optimization control system based on machine learning, characterized in that, The system is used to perform the method according to any one of claims 1-7, and the system comprises: Anomaly detection module: Based on the valve opening and closing response time in historical time periods, an opening and closing response time analysis library is built. By performing anomaly analysis on the opening and closing response time when there is fluid, it is determined whether the valve opening and closing response time is abnormal. Regularity Analysis Module: If the valve opening and closing response time is abnormal, combine the instantaneous flow velocity corresponding to the valve opening and closing response time to analyze the possibility that the instantaneous flow velocity has an impact on the valve opening and closing response time. If the instantaneous flow velocity has a high probability of having an impact on the opening and closing response time, determine whether the impact of the instantaneous flow velocity on the valve opening and closing response time is regular. Model optimization module: Based on the analysis results of whether the influence of instantaneous flow velocity on valve opening and closing response time is regular, a prediction model is constructed, the range of instantaneous flow velocity for regulation is output, and the prediction accuracy of the prediction model is judged. Control and Analysis Module: Based on the prediction accuracy of the prediction model, determine the adjustment value of the opening and closing response time, and adjust the opening and closing response time of the valve.

Images