Pump station abnormal data detection method and system

The abnormal data in the intelligent pump station is screened through the one-variable linear regression and linear transformation method to ensure the data quality, and the problem that abnormal data in the intelligent pump station affects the accuracy of unit characteristics is solved, and the accuracy of intelligent optimization and prediction and maintenance is improved.

CN119004310APending Publication Date: 2024-11-22BEIJING UNIV OF TECH +1
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410983638.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

Due to the abnormal data in the historical data in the intelligent pump station, the accuracy of the unit flow frequency characteristics is affected, resulting in poor intelligent optimization and prediction and maintenance results.

Method used

By obtaining the frequency and flow data of the pump station unit, using one-element linear regression to fit the linear relationship between the frequency and flow, performing pair-by-two linear transformations, determining the slope and intercept threshold range of the normal point, and screening out abnormal data. Ensure data quality.

Benefits of technology

By effectively screening abnormal data, accurate unit flow and frequency characteristics are constructed, and the optimization and prediction and maintenance accuracy of intelligent pump stations are improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004310A_ABST
    Figure CN119004310A_ABST
Patent Text Reader

Abstract

The invention relates to a pump station abnormal data detection method and system. The method comprises the steps of obtaining historical working data of a pump station unit; the historical data comprises pump station unit frequency data and pump station working flow data; fitting a linear relation between the frequency and the flow based on historical data according to unary linear regression, and obtaining a fitting slope and a fitting intercept value of a fitting straight line; performing pairwise linear transformation on the historical data to obtain detection slopes and detection intercept values of multiple groups of straight lines; determining the slope and intercept value threshold range of the normal point based on the fitting slope and the fitting intercept value in combination with the multiple groups of detection slopes and detection intercept values; and determining whether the detection slopes and the detection intercept values of the multiple groups of straight lines are within the slope and intercept threshold value range of the normal point or not so as to screen abnormal data, the frequency and the flow of each unit of the intelligent pump station are detected, and then abnormal data points are found. And therefore, better frequency characteristics of the intelligent pump station unit can be obtained for operation and maintenance of the unit.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of automatic control technology, and in particular to a method and system for detecting abnormal data in a pump station. Background Art

[0002] With the development of deep learning and neural network technology, pump stations continue to explore more complex intelligent optimization and predictive maintenance methods, and more and more intelligent pump stations have begun to be put into operation. Complex intelligent optimization and predictive maintenance are mainly realized based on deep learning and neural network models. However, deep learning and neural network models cannot be separated from training with high-quality data.

[0003] Smart pumping stations are usually composed of several pumping units and a forebay. Deep learning and neural network models need to be trained based on historical data such as the pumping frequency, forebay water level, and pumping station working flow of the pumping station to build the flow frequency characteristics of the pumping station unit, thereby achieving intelligent optimization and predictive maintenance. At present, during the operation of pumping station units, the flow frequency characteristics of pumping station units are often built by fitting historical data, but there are often abnormal data in the historical data due to various external reasons, which will seriously affect the accuracy of the constructed unit characteristics. Summary of the invention

[0004] The present application provides a pump station abnormal data detection method and system, which performs abnormal detection on the water pump unit frequency data generated by the intelligent pump station and the working flow data corresponding to the water pump unit frequency, so as to ensure that accurate unit flow and frequency characteristics are constructed through high-quality historical data.

[0005] In a first aspect, the present application provides a method for detecting abnormal data in a pump station, comprising the following steps: obtaining historical working data of a pump station unit; wherein the historical data includes frequency data of the pump station unit and working flow data of the pump station; fitting the linear relationship between frequency and flow according to univariate linear regression based on the historical data to obtain the fitting slope and fitting intercept value of the fitting straight line; performing pairwise linear transformation on the historical data to obtain detection slopes and detection intercept values ​​of multiple groups of straight lines; determining a threshold range of slopes and intercept values ​​of normal points based on the fitting slopes and fitting intercept values ​​and in combination with multiple groups of detection slopes and detection intercept values; determining whether the detection slopes and detection intercept values ​​of multiple groups of straight lines fall within the threshold range of the slopes and intercepts of normal points to screen abnormal data.

[0006] Optionally, a linear relationship between frequency and flow is fitted based on the historical data according to a univariate linear regression to obtain a fitting slope and a fitting intercept value of the fitting line, including: calculating the average operating frequency of each working water pump at each moment, which is described by formula (1):

[0007]

[0008] Where: represents the frequency mean of each working water pump at time t, represents the frequency of the i-th working water pump at time t; the mean frequency is set as the independent variable, and the working flow rate (Q t ) is set as the dependent variable, and the fitting line is set as a linear function of one variable, which is described by formulas (2), (3), and (4):

[0009]

[0010] y i =Q t i=1,2,3,..,k,i=t(3),

[0011] y=a·x+b (4)

[0012] Where, X i is the ith independent variable, Y i is the ith dependent variable, a represents the summary slope of the solved fitting line, and b represents the summary intercept value of the solved fitting line. Based on univariate linear regression, the fitting slope and fitting intercept values ​​are accurately obtained according to the actual flow value and the fitting flow value on the corresponding fitting line.

[0013] Optionally, based on univariate linear regression, according to the actual flow value and the fitted flow value on the corresponding fitting straight line, the fitting slope and the fitting intercept value are accurately obtained, including: based on univariate linear regression, calculating the error sum of squares between the actual flow value and the fitted flow value on the corresponding fitting straight line and the partial derivative of the error sum of squares, which are described by formula (5):

[0014]

[0015] In the formula, J(a, b) represents the difference between the actual flow value and the fitted flow value on the corresponding fitting straight line. It means the partial derivative of the error sum of squares function with respect to slope a. The error square sum function is used to obtain the partial derivative of the vertical intercept b. Based on the error square sum, the error sum is minimized to obtain the final slope a and intercept value b, which are described by formula (6):

[0016]

[0017] Optionally, performing a pairwise linear transformation on the historical data to obtain multiple sets of straight line detection slopes and detection intercept values ​​includes: selecting a frequency-flow data point at time t in the historical data, and calculating the detection slope and detection intercept value of the straight line formed by the data point and the frequency-flow data points at other times, which is described by formula (7):

[0018]

[0019] In the formula, It represents the detection slope value of the straight line formed by the frequency-flow data point at time t and the frequency-flow data point at time j. represents the intercept value of the straight line formed by the frequency-flow data point at time t and the frequency-flow data point at time j, x t represents the frequency value at time t, y t Represents the flow value at time t; repeat the above steps until the frequency-flow data points at all times have been linearly transformed to obtain the detection slope and detection intercept values ​​of multiple sets of straight lines.

[0020] Optionally, the method of determining the slope and intercept value threshold range of normal points based on the fitting slope and the fitting intercept value and combining multiple groups of the detection slope and the detection intercept value includes: projecting all the detection slopes and the detection vertical intercepts of the straight line formed by the obtained frequency-flow data points at time t and the frequency-flow data points at the remaining times into the corresponding slope-intercept value (kb) parameter space; based on the obtained fitting slope a and the fitting intercept value b as a reference, and combining the data situation in which the detection slope and the detection intercept fall in the kb parameter space, setting the number threshold N of normal points in the kb parameter space, and determining the normal threshold range of the fitting slope k and the fitting intercept value b.

[0021] Optionally, the obtained fitting slope a and the fitting intercept value b are used as a benchmark, and the number threshold N of normal points in the kb parameter space is set in combination with the data situation where the detection slope and the detection intercept fall in the kb parameter space, and the normal threshold range of the fitting slope a and the fitting intercept value b is determined, including: setting the number threshold N of normal points in the kb parameter space in combination with the distribution of the detection slope and the detection intercept data points projected in the kb parameter space, and presetting the normal threshold range of the fitting slope a to a±x, and the normal threshold range of the fitting intercept value b to b±y according to the range where the number threshold N is located; wherein the setting of the size of x and y affects the setting of the number threshold N for normal points.

[0022] Optionally, the step of determining whether the detection slope and detection intercept values ​​of multiple groups of straight lines fall within the slope and vertical intercept threshold range of normal points to screen abnormal data includes: based on the detection slope-detection intercept point projected at time t in the slope-intercept value (kb) parameter space, if the number of detection slope-detection intercept value data points within the range of the normal threshold is greater than or equal to the set quantity threshold, then the frequency-flow data point at time t is considered to be a normal point, otherwise it is an abnormal point; repeating the above steps until the frequency-flow data points at all times have been detected to detect all abnormal data points.

[0023] In the second aspect, the present application provides a pump station abnormal data detection system, which includes: an acquisition module for acquiring historical working data of the pump station unit, wherein the historical data includes pump station unit frequency data and pump station working flow data; a construction module for fitting the linear relationship between frequency and flow based on the historical data according to univariate linear regression to obtain the fitting slope and fitting intercept value of the fitting straight line; a processing module for performing pairwise linear transformation on the historical data to obtain detection slopes and detection intercept values ​​of multiple groups of straight lines; a threshold determination module for determining the slope and intercept value threshold range of normal points based on the fitting slope and the fitting intercept value and in combination with multiple groups of detection slopes and the detection intercept values; a detection module for determining whether the detection slopes and the detection intercept values ​​of multiple groups of straight lines fall within the slope and intercept threshold range of normal points to screen abnormal data.

[0024] This application has at least the following advantages:

[0025] According to the technical content provided in the embodiments of the present application, the above steps are mainly to obtain the frequency data of the pump station unit and the working flow data of the pump station, so as to fit the linear relationship between frequency and flow according to univariate linear regression, obtain a linear function, determine the slope and intercept values, and then linearly transform all the historical data obtained according to the obtained linear function, so as to obtain multiple groups of detection slopes and detection intercept values, project the obtained detection slopes and detection intercept values ​​and the slopes and intercept values ​​of the determined normal points into the parameter space, determine the normal threshold as needed, and finally compare all the detection slopes and detection intercept values ​​with the normal threshold to screen out abnormal data, obtain high-quality historical data and then construct accurate unit flow and frequency characteristics, so as to obtain better intelligent pump station unit frequency characteristics for the operation and maintenance of the unit. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic diagram showing a flow chart of a method for detecting abnormal data of a pumping station in one embodiment;

[0027] Figure 2A schematic diagram showing the structure of the pump station frequency and the pump station working flow rate in one embodiment;

[0028] Figure 3 A schematic diagram of a process for fitting a linear relationship between frequency and flow rate according to a univariate linear regression in one embodiment;

[0029] Figure 4 A schematic diagram of a process for obtaining a fitting slope and a fitting intercept value according to an actual flow value and a fitting flow value on a corresponding fitting straight line in one embodiment;

[0030] Figure 5 A schematic diagram of a process of performing pairwise linear transformation on historical data to obtain multiple groups of detection slopes and detection intercept values ​​in one embodiment;

[0031] Figure 6 A schematic diagram of a structure showing, in one embodiment, determining a slope and an intercept value threshold range based on a fitting slope and a fitting intercept value combined with a detection slope and a detection intercept value;

[0032] Figure 7 A schematic diagram of a structure and a flow chart showing a method of determining a slope and an intercept value threshold range based on a fitting slope and a fitting intercept value in combination with a detection slope and a detection intercept value in one embodiment;

[0033] Figure 8 A schematic diagram showing a process flow of screening abnormal data in one embodiment;

[0034] Fig. 9 A structural block diagram showing a pump station abnormal data detection system in one embodiment. DETAILED DESCRIPTION

[0035] The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0036] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when "including" and / or "comprising" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0037] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings. However, it will be appreciated by those skilled in the art that in the embodiments of the present application, many technical details are proposed in order to enable the reader to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical scheme claimed in the present application can also be implemented. The markings of the following embodiments are for convenience of description and should not constitute any limitation to the specific implementation of the present application. The various embodiments can be combined with each other and quoted from each other without contradiction.

[0038] Figure 1 A flow chart of a method for detecting abnormal data in a pumping station provided in an embodiment of the present application is shown in FIG. Figure 1 As shown, the method may include the following steps:

[0039] S201, obtaining historical working data of pump station units;

[0040] Among them, historical data include pump station unit frequency data and pump station working flow data;

[0041] S202, fitting the linear relationship between frequency and flow rate according to univariate linear regression based on historical data, and obtaining the fitting slope and fitting intercept value of the fitting line;

[0042] S203, performing pairwise linear transformation on the historical data to obtain detection slopes and detection intercept values ​​of multiple groups of straight lines;

[0043] S204, determining a slope and intercept value threshold range of a normal point based on the fitting slope and the fitting intercept value and in combination with multiple groups of detection slopes and detection intercept values;

[0044] S205: Determine whether the detection slopes and detection intercept values ​​of the plurality of groups of straight lines fall within the slope and intercept threshold ranges of normal points to screen abnormal data.

[0045] By acquiring the frequency data of the pump station unit and the working flow data of the pump station, the linear relationship between frequency and flow is fitted to obtain a linear function, and the slope and intercept values ​​are determined. All the historical data obtained are then linearly transformed according to the obtained linear function to obtain multiple groups of detection slopes and detection intercept values. The normal threshold is determined by combining the slope and intercept values. Finally, all the detection slopes and detection intercept values ​​are compared with the normal threshold to screen out abnormal data, thereby obtaining better frequency characteristics of the intelligent pump station unit for operation and maintenance of the unit.

[0046] The following is a detailed description of each step:

[0047] Please refer to Figure 1 , Figure 2As shown, step S201, obtaining the working history data of the pump station unit;

[0048] Among them, historical data include pump station unit frequency data and pump station working flow data;

[0049] In this embodiment, it should be noted that the operating frequency data of each pump station unit and the forebay water level data are collected based on the sensors of each pump unit in the pump station and the forebay water level sensor, so as to obtain the pump station unit frequency data {f1, f2, ..., fn} and the pump station working flow data {Q1, Q2, ..., Qn}, so as to provide a data basis for subsequent detection.

[0050] Please continue to refer to Figure 1-Figure 3 As shown, step S202, based on the historical data, a linear relationship between frequency and flow is fitted according to a univariate linear regression to obtain a fitting slope and a fitting intercept value of the fitting line;

[0051] In this embodiment, it should be noted that the univariate linear regression can be specifically used to fit the linear function curve of the mean frequency of each unit at all times to the flow rate of the pump station forebay by the least squares method, and finally the fitting slope and fitting intercept value are obtained.

[0052] Among them, step S202 fits the linear relationship between frequency and flow rate according to univariate linear regression based on historical data to obtain the fitting slope and fitting intercept value of the fitting line, which specifically includes:

[0053] Step S2021, calculate the average operating frequency of each working water pump at each moment, which is described by formula (1):

[0054]

[0055] Where: represents the frequency mean of each working water pump at time t, It represents the frequency of the i-th working water pump at time t, and k represents that there are k pump station units in total.

[0056] Step S2022, set the frequency mean as the independent variable, the working flow (Q t ) is set as the dependent variable, and the fitting line is set as a linear function of one variable, which is described by formulas (2), (3), and (4):

[0057]

[0058] y i =Q t i=1,2,3,...,k,i=t(3),

[0059] y=a·x+b (4)

[0060] Where, X i is the ith independent variable, Y i is the ith dependent variable, a represents the summary slope of the fitted line, and b represents the summary intercept value of the fitted line.

[0061] Step S2023, based on univariate linear regression, obtain the fitting slope and the fitting intercept value according to the actual flow value and the fitting flow value on the corresponding fitting straight line.

[0062] In this embodiment, it should be noted that the average value of the pump station unit operating frequency is obtained based on the obtained multiple groups of pump station unit operating frequencies and a functional relationship is constructed with the corresponding pump station working flow to obtain a one-dimensional linear relationship, and then the summary slope and summary intercept values ​​can be obtained. According to the obtained functional relationship, the fitting flow value is calculated and combined with the actual flow value to finally obtain the fitting slope and fitting intercept value.

[0063] Reference Figure 3 , Figure 4 As shown, step S2023, based on univariate linear regression, accurately obtains the fitting slope and the fitting intercept value according to the actual flow value and the fitting flow value on the corresponding fitting straight line, specifically including:

[0064] S20231. Based on univariate linear regression, the sum of square errors between the actual flow value and the corresponding fitted flow value on the fitted straight line and the partial derivative of the sum of square errors are calculated, which are described by formula (5):

[0065]

[0066] In the formula, J(a, b) represents the difference between the actual flow value and the fitted flow value on the corresponding fitting straight line. It represents the partial derivative of the error sum of squares function with respect to the summary slope a, It represents the partial derivative of the error sum of squares function with respect to the summary intercept value b;

[0067] S20232. Minimize the error sum based on the error square sum, and obtain the final fitting slope a and fitting intercept value b, which are described by formula (6):

[0068]

[0069] According to the least squares method, the sum of squares of the errors between the actual flow value and the fitted flow value is minimized, so as to be as close to the actual value as possible, reduce the error, and finally obtain an accurate function, thereby determining the final fitting slope and fitting intercept values.

[0070] Reference Figure 1 , Figure 5As shown, in step S203, the historical data are linearly transformed in pairs to ensure that the historical data are linearly transformed, and multiple groups of detection slopes and detection intercept values ​​are obtained, specifically:

[0071] Step S2031: Select the frequency-flow data point at time t in the historical data, and find the detection slope and detection intercept value of the straight line formed by the data point and the frequency-flow data points at other times, which can be described by formula (8):

[0072]

[0073] In the formula, It represents the detection slope value of the straight line formed by the frequency-flow data point at time t and the frequency-flow data point at time j. represents the intercept value of the straight line formed by the frequency-flow data point at time t and the frequency-flow data point at time j, x t represents the frequency value at time t, y t represents the flow value at time t;

[0074] Step S2032, repeat the above steps until all frequency-flow data points at all times have been linearly transformed to obtain multiple groups of detection slopes and detection intercept values.

[0075] According to the frequency-flow data point at a selected moment, the slope and intercept values ​​after linear transformation with the remaining moment data points are calculated, and all frequency-flow data points are linearly transformed to obtain multiple groups of detection slope and detection intercept values.

[0076] Reference Figure 1 , Figure 6 as well as Figure 7 As shown, step S204, based on the fitting slope and the fitting intercept value and in combination with multiple groups of detection slopes and detection intercept values, the slope and intercept value threshold range of the normal point is determined, specifically:

[0077] Step S2041, projecting all the detection slopes and detection vertical intercepts of the straight line formed by the obtained frequency-flow data point at time t and the frequency-flow data points at the remaining times into the corresponding slope-intercept value (kb) parameter space;

[0078] Step S2042: Based on the obtained fitting slope a and fitting intercept value b as a benchmark, and in combination with the data situation where the detection slope and the detection intercept fall in the kb parameter space, set the number threshold N of normal points in the kb parameter space, and determine the normal threshold range of the fitting slope a and the fitting intercept value b.

[0079] like Figure 6As shown, various straight lines formed by the frequency-flow data points at time t and the frequency-flow data points at the remaining times are obtained in the frequency and flow space, and the corresponding detection slope and detection intercept values ​​are obtained according to each straight line obtained, and are projected into the corresponding kb parameter space. Then, based on the obtained slope a and intercept value b as a benchmark, combined with the detection slope and detection vertical intercept data projected into the kb parameter space, such as relatively dense areas, a quantity threshold N is set, and then the normal threshold range of the slope a and the normal threshold range of the intercept value b are preset according to the quantity threshold N.

[0080] Based on the obtained fitting slope k and fitting intercept value b as a benchmark, the number threshold N of normal points in the kb parameter space is set in combination with the data situation where the detection slope and the detection intercept fall in the kb parameter space, and the normal threshold range of the fitting slope k and the fitting intercept value b is determined. Specifically, the number threshold N of normal points in the kb parameter space is set in combination with the distribution of the detection slope and the detection intercept data points projected in the kb parameter space, and the normal threshold range of the slope k is preset to k±x according to the range where the number threshold N is located, and the normal threshold range of the intercept value b is preset to b±y; wherein, the setting of x and y size affects the setting of the number threshold N of normal points. The larger the x and y settings are, the more data points are included in the normal range.

[0081] Specifically, in one example, in the kb parameter space, it is found that multiple points of the detection slope-detection intercept obtained at time t are concentratedly distributed. Seven concentrated points can be taken as needed, and the quantity threshold N can be set to 7. Combined with the quantity threshold, the normal threshold range of the fitting slope a is preset as a±x, x can be set to 5, and the normal threshold range of the intercept value b is b±y, y can be set to 10. At this time, there are 7 detection slope-detection intercept points within the threshold range of the fitting slope-fitting intercept.

[0082] Reference Figure 1 , Figure 8 As shown, step S205 determines whether the detection slopes and intercept values ​​of multiple groups of straight lines fall within the slope and intercept threshold range of normal points to screen abnormal data, specifically:

[0083] S2051. According to the detection slope-detection intercept point projected at time t in the slope-intercept value (kb) parameter space, if the number of detection slope-detection intercept value data points within the normal threshold value is greater than or equal to the set number threshold value, the frequency-flow data point at time t is considered to be a normal point, otherwise it is an abnormal point;

[0084] S2052, repeat the above steps until the frequency-flow data points at all times have been detected to detect all abnormal data points.

[0085] In this embodiment, it should be noted that in the kb parameter space, it is determined whether the number of points corresponding to the detection slope and detection intercept of the data point at time t that fall within the range (k±x, b±10) set by the preset normal threshold is greater than or equal to the number threshold 7. If so, the points corresponding to the detection slope and detection intercept at time t are considered normal points, otherwise they are abnormal points. Then, it is determined whether the number of points corresponding to the multiple groups of detection slopes and detection intercepts obtained from the data point at time t+1 and the remaining data points that fall within the range (k±x, b±10) set by the preset normal threshold is greater than or equal to the number threshold 7, and the above steps are repeated until all frequency-flow data points at all times have been detected to detect all abnormal data points.

[0086] In other embodiments, the slope and intercept threshold ranges for determining normal points can be set based on the summary slope and intercept of the straight line fitted by the univariate linear regression, and the slope and intercept of the straight line formed by the selected data point at time t and the data points at the remaining times are all projected into the corresponding kb parameter space, and a dispersion threshold for the data mapping in the parameter space is set and the dispersion of the data after being mapped into the parameter space is calculated, which is represented by the following formula:

[0087]

[0088]

[0089] σ k-b : The degree of data dispersion after the frequency-flow data point at a certain moment is mapped to the kb parameter space by performing a pairwise linear transformation with the frequency-flow data points at all other moments;

[0090] The area of ​​the kb parameter space region determined by the set normal point slope and vertical intercept threshold range.

[0091] According to the dispersion threshold of the set parameter space, it is determined whether the selected frequency-flow data point at a certain moment is a normal point. Generally, the higher the quality of the data point, the greater the spatial dispersion value calculated after being projected into the kb parameter space. Because the slope and vertical intercept values ​​of the straight line formed by the normal frequency-flow data point at a certain moment and the frequency-flow data points at other moments are often within the set slope and vertical intercept threshold range, the dispersion value of the parameter space is large, and vice versa. The setting of the dispersion threshold is often based on the setter's requirements for data quality. If the data quality requirements are high, then the parameter space dispersion threshold will be high. If the data quality requirements are not stringent, then the parameter space dispersion threshold will be relatively low.

[0092] The implementation principle of this embodiment: The above steps are mainly to obtain the frequency data of the pump station unit and the working flow data of the pump station, and then fit the linear relationship between frequency and flow according to univariate linear regression to obtain a linear function, determine the slope and intercept values, and then linearly transform all the historical data obtained according to the obtained linear function to obtain multiple groups of detection slopes and detection intercept values, project the obtained detection slopes and detection intercept values ​​and the slopes and intercept values ​​of the determined normal points into the parameter space, determine the normal threshold as needed, and finally compare all the detection slopes and detection intercept values ​​with the normal threshold to screen out abnormal data, so as to ensure that accurate unit flow and frequency characteristics are constructed through high-quality historical data, thereby improving the accuracy of intelligent optimization and predictive maintenance of pump stations.

[0093] Reference Fig. 9 As shown, the present application also provides a pump station abnormal data detection system, which may include: an acquisition module 301, a construction module 302, a processing module 303, a threshold determination module 304, and a detection module 305. The main functions of each component module are as follows:

[0094] The acquisition module 301 is used to acquire the historical working data of the pump station unit, wherein the historical data includes the frequency data of the pump station unit and the working flow data of the pump station;

[0095] A construction module 302 is used to fit the linear relationship between frequency and flow rate according to a univariate linear regression based on historical data, and obtain a fitting slope and a fitting intercept value of the fitting line;

[0096] The processing module 303 is used to perform pairwise linear transformation on the historical data to obtain detection slopes and detection intercept values ​​of multiple groups of straight lines;

[0097] A threshold determination module 304 is used to determine a threshold range of slope and intercept values ​​of a normal point based on the fitting slope and the fitting intercept value and in combination with multiple sets of detection slopes and detection intercept values;

[0098] The detection module 305 is used to determine whether the detection slopes and detection intercept values ​​of multiple groups of straight lines fall within the slope and intercept threshold ranges of normal points to screen abnormal data.

[0099] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A method for detecting abnormal data in a pump station, characterized in that: The following steps are involved: Obtain historical working data of pump station units; Wherein, the historical data includes the frequency data of the pump station units and the working flow data of the pump station; based on the historical data, the linear relationship between the frequency and the flow is fitted according to the univariate linear regression to obtain the fitting slope and the fitting intercept value of the fitting line; Performing pairwise linear transformation on the historical data to obtain detection slopes and detection intercept values ​​of multiple groups of straight lines; Determining a slope and an intercept value threshold range of a normal point based on the fitting slope and the fitting intercept value and in combination with a plurality of groups of the detection slopes and the detection intercept values; Determine whether the detection slopes and the detection intercept values ​​of multiple groups of straight lines fall within the slope and intercept threshold ranges of normal points to screen abnormal data.

2. The method for detecting abnormal data of a pumping station according to claim 1, characterized in that: The method of fitting the linear relationship between frequency and flow rate based on the historical data according to univariate linear regression to obtain the fitting slope and the fitting intercept value of the fitting line includes: Calculate the average operating frequency of each working water pump at each moment and describe it through formula (1): Where: represents the frequency mean of each working water pump at time t, represents the frequency of the i-th working water pump at time t; The frequency mean is set as the independent variable, and the workload (Q t ) is set as the dependent variable, and the fitting line is set as a linear function of one variable, which is described by formulas (2), (3), and (4): y i =Q t i=1,2,3,...,k,i=t (3), y=a·x+b (4), Where, X i is the ith independent variable, Y i is the i-th dependent variable, a represents the summary slope of the fitted line, and b represents the summary intercept value of the fitted line; Based on univariate linear regression, the fitting slope and fitting intercept values ​​are accurately obtained according to the actual flow value and the fitting flow value on the corresponding fitting straight line.

3. The method for detecting abnormal data of a pump station according to claim 2, characterized in that: The method based on univariate linear regression accurately obtains the fitting slope and the fitting intercept value according to the actual flow value and the fitting flow value on the corresponding fitting straight line, including: calculating the error sum of squares between the actual flow value and the fitting flow value on the corresponding fitting straight line and the partial derivative of the error sum of squares based on univariate linear regression, which is described by formula (5): In the formula, J(a, b) represents the difference between the actual flow value and the fitted flow value on the corresponding fitting straight line. It means the partial derivative of the error sum of squares function with respect to slope a. It represents the partial derivative of the error sum of squares function with respect to the vertical intercept b; Based on the error square sum, the error sum is minimized to obtain the final slope a and intercept value b, which are described by formula (6):

4. The method for detecting abnormal data of a pumping station according to claim 2 or 3, characterized in that: The pairwise linear transformation of the historical data is performed to obtain a plurality of groups of straight line detection slopes and detection intercept values, including: Select the frequency-flow data point at time t in the historical data, and find the detection slope and detection intercept value of the straight line formed by the data point and the frequency-flow data points at other times, which can be described by formula (7): In the formula, It represents the detection slope value of the straight line formed by the frequency-flow data point at time t and the frequency-flow data point at time j. represents the intercept value of the straight line formed by the frequency-flow data point at time t and the frequency-flow data point at time j, x t represents the frequency value at time t, y t represents the flow value at time t; The above steps are repeated until the frequency-flow data points at all times have been linearly transformed to obtain multiple groups of straight line detection slopes and detection intercept values.

5. The method for detecting abnormal data of a pumping station according to claim 4, characterized in that: The determining of the slope and intercept value threshold range of the normal point based on the fitting slope and the fitting intercept value and in combination with multiple groups of the detection slopes and the detection intercept values ​​includes: Projecting the detected slope and the detected vertical intercept of the straight line formed by the obtained frequency-flow data point at time t and the frequency-flow data points at the remaining times into the corresponding slope-intercept value (kb) parameter space; Based on the obtained fitting slope a and the fitting intercept value b as a benchmark, the number threshold N of normal points in the kb parameter space is set in combination with the data situation that the detection slope and the detection intercept fall in the kb parameter space, and the normal threshold range of the fitting slope k and the fitting intercept value b is determined.

6. The method for detecting abnormal data of a pumping station according to claim 5, characterized in that: The method of setting the normal point quantity threshold N in the kb parameter space based on the obtained fitting slope a and the fitting intercept value b as a benchmark and combining the data situation that the detection slope and the detection intercept fall in the kb parameter space, and determining the normal threshold range of the fitting slope k and the fitting intercept value b includes: Combined with the distribution of the detection slope and the detection intercept data points projected in the kb parameter space, a number threshold N of normal points in the kb parameter space is set, and according to the range of the number threshold N, the normal threshold range of the fitting slope a is preset to be a±x, and the normal threshold range of the fitting intercept value b is preset to be b±y; Among them, the setting of x and y size affects the setting of the number threshold N of normal points included.

7. The method for detecting abnormal data of a pumping station according to claim 5 or 6, characterized in that: The step of determining whether the detection slopes and detection intercept values ​​of the plurality of groups of straight lines fall within the slope and vertical intercept threshold range of the normal point to screen abnormal data includes: According to the detection slope-detection intercept point projected at time t in the slope-intercept value (kb) parameter space, if the number of the detection slope-detection intercept value data points within the range of the normal threshold is greater than or equal to the set number threshold, then the frequency-flow data point at time t is considered to be a normal point, otherwise it is an abnormal point; The above steps are repeated until the frequency-flow data points at all times have been detected to detect all abnormal data points.

8. A pump station abnormal data detection system, characterized in that: include: An acquisition module, used for acquiring historical working data of the pump station unit, wherein the historical data includes frequency data of the pump station unit and working flow data of the pump station; A construction module is used to fit the linear relationship between frequency and flow rate according to univariate linear regression based on the historical data, and obtain the fitting slope and fitting intercept value of the fitting line; A processing module, used for performing pairwise linear transformation on the historical data to obtain detection slopes and detection intercept values ​​of multiple groups of straight lines; A threshold determination module, used to determine a threshold range of slope and intercept values ​​of a normal point based on the fitting slope and the fitting intercept value and in combination with a plurality of groups of the detection slopes and the detection intercept values; The detection module is used to determine whether the detection slopes and the detection intercept values ​​of multiple groups of straight lines fall within the slope and intercept threshold ranges of normal points to screen abnormal data.

Citation Information

Cited By

  • Hydraulic support anomaly detection method, device and equipment based on displacement data and medium

    CN120761006A