Automatic meter reading method and system for water meter data

Abstract

The invention relates to the technical field of automatic meter reading, solves the technical problems that in the prior art, meter reading efficiency is low, real-time performance is poor, and monitoring of abnormal water consumption data and visual display of water consumption characteristics are lacked, and particularly relates to an automatic meter reading method for water meter data. Preliminary abnormal value elimination is carried out on the original data, and non-abnormal data is obtained; the method comprises the steps of S1, carrying out data processing on the water consumption data, S2, carrying out deep preprocessing on the non-abnormal data to obtain preprocessed data, and S3, dividing the preprocessed data into a plurality of water consumption data Wa (t) according to time periods, and calculating characteristic variables of the water consumption data Wa (t). According to the method, dimension reduction processing is carried out on the water consumption data, the problem of dimension disasters can be reduced, the performance of the method is improved, and meanwhile the accuracy of the method is improved. And the water consumption condition obtained by the user can be a three-dimensional visual image, so that the user can obtain the water consumption information more intuitively and quickly.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic meter reading, and particularly to a method and system for automatically reading water meter data. Background Art

[0002] As a key tool for water resource management, the accuracy and real-time nature of water meter data are crucial for energy distribution, billing, and resource optimization. Traditional methods for collecting water meter data mainly rely on manual on-site meter reading at regular intervals. This process involves on-site staff recording the readings of each water meter and then manually inputting them into the management system. Although this method is technically simple to implement, manual meter reading consumes a large amount of human resources, and as the number of water meters increases, the required human and time costs increase linearly, and the real-time nature is poor. Existing meter reading methods obtain water volume through digital devices, but in the process of obtaining water consumption data, errors are easily caused by sensor measurements, and abnormal water consumption behaviors such as water theft and leakage cannot be monitored. When users obtain water consumption data, they cannot intuitively obtain the characteristics of water consumption, resulting in a very poor user experience. Summary of the Invention

[0003] Aiming at the deficiencies of the prior art, the present invention provides a method and system for automatically reading water meter data, which solves the technical problems of low meter reading efficiency, poor real-time nature, lack of monitoring of abnormal water consumption data, and lack of intuitive display of water consumption characteristics in the prior art, and achieves the purpose of improving meter reading efficiency and real-time nature, regularly monitoring water consumption data, and improving the intuitiveness of water consumption data.

[0004] To solve the above technical problems, the present invention provides the following technical solution: A method for automatically reading water meter data, the method comprising the following steps:

[0005] S1. Collect the original data of the water meter running at any stage, and perform preliminary outlier rejection on the original data to obtain anomaly-free data;

[0006] S2. Perform in-depth preprocessing on the anomaly-free data to obtain preprocessed data;

[0007] S3. Divide the preprocessed data into multiple water volume data Wa (t) according to time periods, and calculate the characteristic variables of the water volume data Wa (t) ;

[0008] S4. Classify users according to the characteristic variables to obtain a user classification set, and obtain water volume prediction data Yc w through the time convolutional network method;

[0009] S5. Calculate the dimensionality reduction value Jw w according to the water volume prediction data Yc z and, based on the dimensionality reduction value Jwz Generate a visual three-dimensional graph;

[0010] S6. Send the visual three-dimensional graph to the user data center.

[0011] Preferably, in step S1, the specific implementation steps are as follows:

[0012] S11. Arrange the original data in ascending order to generate a processing sequence C a , and calculate the first extreme value A1 and the second extreme value A2 according to the processing sequence C a . The calculation formula is:

[0013] A1 = (C max + C min ) × 25%

[0014] A2 = (C max + C min ) × 75%

[0015] Among them, C max and C min respectively represent the maximum value and the minimum value in the processing sequence C a ;

[0016] S12. Calculate the measured value B c according to the first extreme value A1 and the second extreme value A2. The calculation formula is:

[0017] B c = A2 - A1

[0018] Among them, B c represents the c-th measured value;

[0019] S13. Calculate the maximum value J c and the minimum value J a representing the normal data range in the processing sequence C jd according to the measured value B jx . The calculation formula is:

[0020] J jd = A2 + 1.5 × B c

[0021] J jx = A1 - 1.5 × B c

[0022] Among them, J jd represents the d-th maximum value, and J jx represents the x-th minimum value;

[0023] S14. According to the maximum value J jd and the minimum value J jxReject the outliers in the processing sequence C a ;

[0024] If C a ≤J jx or C a ≥J jd , then the processing sequence C a is an outlier and is rejected, obtaining the vacant position K cy ;

[0025] If J jx <C a <J jd , then the processing sequence C a is a normal value and is retained;

[0026] S15. Restore the data order of the processing sequence C cy containing the vacant position K a , and obtain the data without outliers.

[0027] Preferably, in step S2, the specific implementation steps are as follows:

[0028] S21. Determine the adjacent vacant positions K cy and K cy+1 according to the vacant position K cy-1 , and calculate the filling value Tc cy for filling the vacant position K h in the data without outliers. The calculation formula is:

[0029]

[0030] where Tc h represents the h-th filling value;

[0031] S22. Fill the filling value Tc h into the vacant position K cy in the data without outliers, obtaining the data without missing values Ws f ;

[0032] S23. Calculate the detection value fw z of the data without missing values. The calculation formula is:

[0033]

[0034] where erf -1 represents the inverse function of the error function, and p represents the cumulative probability value;

[0035] S24. Obtain the data quantity N of the data without missing values by the counting method, and calculate the quantile value Fw z according to the detection value fw d . The calculation formula is:

[0036]

[0037] Among them, Fw d represents the d-th quantile value, and fw z represents the z-th test value;

[0038] S25. Sort the data Ws without missing data f in ascending order to obtain the sorted data set P;

[0039] S26. Obtain the test coefficient H by looking up a table according to the quantile value Fw d and the number of data N, and calculate the test value Jh b , and the calculation formula is:

[0040]

[0041]

[0042] Among them, Jh b represents the b-th test value, represents the average value of the data Ws without missing data f . Preferably, in step S26, the specific implementation steps are as follows:

[0043] S261. Obtain the critical value p by looking up a table according to the number of data N and the test value Jh b ;

[0044] S262. Judge whether the data Ws without missing data f is normally distributed according to the critical value p;

[0045] If p < 0.05, it means that it does not follow the normal distribution, and go to step S263;

[0046] If p ≥ 0.05, it means that it follows the normal distribution, and go to step S265;

[0047] S263. Calculate the non-positive test value Jy z under non-normal distribution, and the calculation formula is:

[0048]

[0049] Among them, Ws f-1 and Ws1 represent the (f - 1)-th and the 1st data without missing data respectively;

[0050] S264. Detect outliers for the data Ws without missing data z according to the non-positive test value Jy f ;

[0051] If Jyz If it is < 2, it is a normal value, and preprocessed data is generated;

[0052] If Jy z ≥ 2, it is an outlier, and the first replacement value Th is calculated e and the outlier is replaced; S265. Calculate the normal inspection value Zj under the normal distribution c , and the calculation formula is:

[0053]

[0054] where s represents the sample standard deviation of the non-missing data Ws f ;

[0055] S266. Detect outliers for the non-missing data Ws according to the normal inspection value Zj c ; f ;

[0056] If Zj c < 2, it is a normal value, and preprocessed data is generated;

[0057] If Zj c ≥ 2, it is an outlier, and the second replacement value Te is calculated h and the outlier is replaced.

[0058] Preferably, the calculation formula of the first replacement value Th e and the second replacement value Te h is:

[0059]

[0060] where Ws f+1 and Ws f-1 respectively represent the (f + 1)-th and (f - 1)-th non-missing data on both sides of the non-missing data Ws f .

[0061] Preferably, in step S3, the specific implementation steps are as follows:

[0062] S31. Define the water volume data in the t-th time period as Wa (t) , and calculate the periodic model Zq of the water volume data Wa (t) , and the calculation formula is: m ;

[0063] Zq m = D·sin(ρt + δ)+ G

[0064] where D represents the model coefficient, ρ represents the periodic coefficient, δ represents the initial phase, and G represents the model offset value;

[0065] S32. Calculate the weekly water consumption Zz y , and the calculation formula is:

[0066]

[0067] where n represents the number of water volume data Wa (t) within a week, and Wa (t) represents the water volume data at the t-th time period;

[0068] S33. Calculate the fluctuation period T of the water consumption according to the periodic coefficient ρ b , and the calculation formula is:

[0069]

[0070] where T b represents the b-th fluctuation period;

[0071] S34. Calculate the growth rate θ of the water volume data z , and the calculation formula is:

[0072]

[0073] where Zq m+1 represents the periodic model value of the (m + 1)-th water volume data.

[0074] Preferably, in step S4, the specific implementation steps are as follows:

[0075] S41. The characteristic variables include the weekly water consumption Zz y , the fluctuation period T b and the growth rate θ z . Integrate the weekly water consumption Zz y , the fluctuation period T b and the growth rate θ z in the same time period into the water use vector Yw s , and the expression is:

[0076] Yw s = [Zz y , T b , θ z

[0077] where Yw s represents the s-th water use vector;

[0078] S42. Calculate the standard value B according to the water use vector Yw s , and the calculation formula is: z , and the formula is:

[0079]

[0080] where​ Denote the average of the water consumption vector Yw s as, and m represents the number of the water consumption vectors Yw s ;

[0081] S43. Take a standard value B z as the initial center Zx c , and calculate the numerical interval Rj z according to the standard value B c and the initial center Zx s . The calculation formula is:

[0082] Rj s = ||B z - Zx c || 2

[0083] where Rj s represents the s-th numerical interval;

[0084] S43. Change the initial center Zx c and repeat step S43 to obtain multiple numerical intervals Rj s , calculate the optimal interval value Rzj k , and take the initial center Zx k corresponding to the optimal interval value Rzj c as the classification center fl x . The expression of the optimal interval value Rzj k is:

[0085]

[0086] where L represents the number of the numerical intervals Rj s ;

[0087] S44. Calculate the group average value WC x according to the classification center fl s . The calculation formula is:

[0088]

[0089] where B represents the number of groups of the standard value B z and the classification center fl x ;

[0090] S45. Calculate the optimal classification value Zy s according to the group average value WC v . The calculation formula is:

[0091]

[0092] where Zy vRepresents the v-th optimal classification value, Indicates that the (s - 1)-th population average value is divided into Q classification quantities;

[0093] S46. Repeat step S45 to obtain multiple optimal classification values Zy v , and use the classification quantity Q corresponding to the minimum value in the optimal classification value Zy v as the optimal classification scheme, and obtain the user classification set;

[0094] S47. Construct a water consumption prediction model through the time convolution network method, and input the data in the user classification set into the water consumption prediction model to obtain the water volume prediction data Yc w .

[0095] Preferably, in step S5, the specific implementation steps are as follows:

[0096] S51. Convert the water volume prediction data Yc w into an m×n analysis matrix Fx through the data conversion method j , and the expression is:

[0097]

[0098] where R 11 represents the data value in the first row and the first column of the analysis matrix Fx j , n represents the number of types of characteristic variables of the water volume prediction data Yc w , and m represents the quantity of the water volume prediction data Yc in each type of characteristic variable w ;

[0099] S52. Calculate the variance matrix Xf of the analysis matrix Fx j , and the calculation formula is: j where Xf

[0100]

[0101] represents the j-th variance matrix; j

[0102] S53. Perform eigenvalue decomposition on the variance matrix Xf through the eigenvalue decomposition method j to obtain the eigenmatrix Tz j and the diagonal value λ corresponding to the eigenmatrix Tz j i ;

[0103]

[0104] S54. Calculate the dimensionality reduction value Jw according to the eigenmatrix Tz j , and the calculation formula is: z ​​​

[0105] Among them, F j represents the average value of the analysis matrix Fx j ;

[0106] S55. Calculate the weight value Qz i according to the contribution value λ j and the eigenmatrix Tz d , and the calculation formula is:

[0107]

[0108] Among them, Qz d represents the d-th weight value;

[0109] S56. Sort the dimensionality reduction value Jw d from large to small according to the weight value Qz z , and map the sorting result to the three-dimensional space to obtain a visualized three-dimensional graph.

[0110] This technical solution also provides a system for the above automatic meter reading method, and the system includes:

[0111] A preliminary processing module, which collects the original data of the water meter running at any stage, and performs preliminary outlier removal on the original data to obtain anomaly-free data;

[0112] A preprocessing module, which performs in-depth preprocessing on the anomaly-free data to obtain preprocessed data;

[0113] A feature module, which divides the preprocessed data into multiple water volume data Wa (t) according to time periods, and calculates the characteristic variables of the water volume data Wa (t) ;

[0114] A classification and prediction module, which classifies users according to the characteristic variables to obtain a user classification set, and obtains water volume prediction data Yc w through the time convolution network method;

[0115] A dimensionality reduction module, which calculates the dimensionality reduction value Jw w according to the water volume prediction data Yc z , and generates a visualized three-dimensional graph based on the dimensionality reduction value Jw z ;

[0116] A visualization module, which sends the visualized three-dimensional graph to the user data center.

[0117] By means of the above technical solution, the present invention provides an automatic meter reading method for water meter data, which has at least the following beneficial effects:

[0118] 1. Through preliminary processing and preprocessing, the present invention completes the preparation of user water consumption data. To prevent large errors in user water consumption data, through outlier detection in preprocessing, outliers in water consumption can be initially discovered, and then the abnormal water consumption data of users can be marked and processed.

[0119] 2. By obtaining water consumption data at regular intervals, the present invention can complete the regular monitoring of water consumption data, and obtain multiple water consumption characteristics for calculating water consumption data, which can more accurately analyze the water consumption habits of users, facilitate the comprehensive prediction of users' water consumption situations, and give the future water consumption situations of users. This enables users to not only quickly and efficiently obtain water consumption information but also know the future water consumption, enhancing the user experience.

[0120] 3. By performing dimensionality reduction processing on water consumption data, the present invention can reduce the data dimension, reduce the problem of dimensionality disaster, improve the method performance, and at the same time enable the visualized data after dimensionality reduction to be more vivid and comprehensive. As a result, the water consumption situation obtained by users is a three-dimensional visualized image, enabling users to obtain water consumption information more intuitively and quickly. BRIEF DESCRIPTION OF THE DRAWINGS

[0121] The drawings described herein are used to provide a further understanding of the present application and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings:

[0122] Figure 1 is a flowchart of a method for automatically reading water meter data according to the present invention;

[0123] Figure 2 is a structural block diagram of a system for automatically reading water meter data according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0124] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments. This enables a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects and implement accordingly.

[0125] Due to the technical problems of low meter reading efficiency and poor real-time performance in the prior art, and the lack of monitoring of abnormal water consumption data and intuitive display of water consumption characteristics, please refer to Figure 1 , this embodiment provides a method for automatically reading water meter data, which can improve meter reading efficiency and real-time performance, regularly monitor water consumption data, and improve the intuitiveness of water consumption data. The method includes the following steps:

[0126] S1. Collect the original data of the water meter during any operation stage, and perform preliminary outlier removal on the original data to obtain anomaly-free data. After obtaining the original data of the water meter, due to the influence of equipment sensors or the environment, outliers may appear in the original data, and the outliers need to be processed. The specific implementation steps are as follows:

[0127] S11. Arrange the original data in ascending order to generate a processing sequence C a , and based on the processing sequence C a calculate the first extreme value A1 and the second extreme value A2. The calculation formula is:

[0128] A1 = (C max + C min ) × 25%

[0129] A2 = (C max + C min ) × 75%

[0130] where C max and C min respectively represent the maximum value and the minimum value in the processing sequence C a ;

[0131] S12. Calculate the measured value B c based on the first extreme value A1 and the second extreme value A2. The calculation formula is:

[0132] B c = A2 - A1

[0133] where B c represents the c-th measured value;

[0134] S13. Calculate the maximum value J c and the minimum value J a representing the normal data range in the processing sequence C jd based on the measured value B jx . The calculation formula is:

[0135] J jd = A2 + 1.5 × B c

[0136] J jx = A1 - 1.5 × B c

[0137] where J jd represents the d-th maximum value, and J jx represents the x-th minimum value;

[0138] S14. Based on the maximum value J jd and the minimum value J jx for the processing sequence Ca The outliers are removed;

[0139] If C a ≤J jx or C a ≥J jd , then process sequence C a is an outlier and is removed, resulting in a vacant position K cy ;

[0140] If J jx <C a <J jd , then process sequence C a Normal value and keep it;

[0141] S15, the free position K cy The processing sequence C a The data sequence is restored and no abnormal data is obtained. Through preliminary processing and preprocessing, the preparation of user water use data is completed. In order to prevent large errors in user water use data, the abnormal value detection in preprocessing can initially find the abnormal value of water use, and then mark the abnormal water use data of users.

[0142] S2. Perform deep preprocessing on the data without abnormalities to obtain preprocessed data. After removing the abnormal values, there will still be data that does not conform to the normal distribution due to the influence of the acquired circuit or environment, which affects the calculation of subsequent steps. In order to facilitate the accurate calculation of subsequent steps, the specific implementation steps are as follows:

[0143] S21, according to the free position K cy Determine the adjacent vacancy K cy+1 and K cy-1 , calculate the K used to fill the empty space in the data without abnormality cy Fill value Tc h , the calculation formula is:

[0144]

[0145] Among them, Tc h Indicates the hth filling value;

[0146] S22, fill value Tc h Fill in the remaining spaces K without abnormal data cy On the above, we get no missing data Ws f ;

[0147] S23. Calculate the detection value fw of the non-missing data z , the calculation formula is:

[0148]

[0149] Among them, erf -1 represents the inverse function of the error function, and p represents the cumulative probability value;

[0150] S24. Obtain the number of data N without missing data through the counting method, and calculate the quantile value Fw according to the detected value fw z The calculation formula is: d

[0151]

[0152] Among them, Fw d represents the d-th quantile value, and fw z represents the z-th detected value;

[0153] S25. Sort the data Ws without missing data f in ascending order to obtain the sorted data set P;

[0154] S26. Obtain the test coefficient H through the look-up table method according to the quantile value Fw d and the number of data N, and calculate the test value Jh according to the test coefficient H b The calculation formula is:

[0155]

[0156] Among them, Jh b represents the b-th test value, represents the average value of the data Ws without missing data f . The specific implementation steps are as follows: The look-up table method is a commonly used method to obtain the test coefficient H, which can be used to test whether the data conforms to the normal distribution or the skewed distribution in the subsequent steps, and will not be elaborated here.

[0157] S261. Obtain the critical value p through the look-up table method according to the number of data N and the test value Jh b ;

[0158] S262. Judge whether the data Ws without missing data f is normally distributed according to the critical value p;

[0159] If p < 0.05, it means that it does not follow the normal distribution, and go to step S263;

[0160] If p ≥ 0.05, it means that it follows the normal distribution, and go to step S265;

[0161] S263. Calculate the non-positive test value Jy under the non-normal distribution z , and the calculation formula is:

[0162] ​

[0163] Among them, Ws f-1 and Ws1 respectively represent the (f - 1)-th and the 1-st data without missing values; a non-normal distribution can be understood as a skewed distribution.

[0164] S264. Detect outliers for the data without missing values Ws z based on the non-normal test value Jy f ;

[0165] If Jy z < 2, it is a normal value, and preprocessed data is generated;

[0166] If Jy z ≥ 2, it is marked as an outlier, the first replacement value Th e is calculated and the outlier is replaced. The calculation formulas for the first replacement value Th e and the second replacement value Te h are as follows:

[0167]

[0168] Among them, Ws f+1 and Ws f-1 respectively represent the (f + 1)-th and the (f - 1)-th data without missing values located on both sides of the data without missing values Ws f .

[0169] S265. Calculate the normal test value Zj c under the normal distribution, and the calculation formula is:

[0170]

[0171] Among them, s represents the sample standard deviation of the data without missing values Ws f ;

[0172] S266. Detect outliers for the data without missing values Ws c based on the normal test value Zj f ;

[0173] If Zj c < 2, it is a normal value, and preprocessed data is generated;

[0174] If Zj c ≥ 2, it is marked as an outlier, the second replacement value Te h is calculated and the outlier is replaced. Through preliminary processing and preprocessing, the preparation work for the user's water consumption data is completed. In order to prevent large errors in the user's water consumption data, through outlier detection in preprocessing, outliers in water consumption can be initially discovered, and then the abnormal water consumption data of users can be marked.

[0175] S3. Divide the preprocessed data into multiple water volume data Wa according to time periods (t) , and calculate the characteristic variables of the water volume data Wa (t) . In order to monitor the water volume regularly and prevent water leakage when users use water, the specific implementation steps are as follows:

[0176] S31. Define the water volume data of the t-th time period as Wa (t) , and calculate the periodic model Zq (t) of the water volume data Wa m . The calculation formula is:

[0177] Zq m = D·sin(ρt + δ) + G

[0178] where D represents the model coefficient, ρ represents the periodic coefficient, δ represents the initial phase, and G represents the model offset value;

[0179] S32. Calculate the weekly water consumption Zz y . The calculation formula is:

[0180]

[0181] where n represents the number of water volume data Wa (t) within a week, and Wa (t) represents the water volume data of the t-th time period;

[0182] S33. Calculate the fluctuation period T b of the water consumption according to the periodic coefficient ρ. The calculation formula is:

[0183]

[0184] where T b represents the b-th fluctuation period;

[0185] S34. Calculate the growth rate θ z of the water volume data. The calculation formula is:

[0186]

[0187] where Zq m+1 represents the periodic model value of the (m + 1)-th water volume data. By obtaining the water usage data regularly, it is possible to complete the regular monitoring of the water usage data, and obtain multiple water usage characteristics of the water usage data, which can analyze the user's water usage habits more accurately, facilitate the comprehensive prediction of the user's water usage situation, and give the user's future water usage situation, enabling the user to not only quickly and efficiently obtain the water usage situation, but also know the future water consumption, improving the user's usage experience.

[0188] S4. Classify users according to characteristic variables to obtain a user classification set, and obtain water volume prediction data Yc through the time convolution network method w ; To solve the problem that users can obtain a more comprehensive understanding of their water usage situation, enable users to have a general understanding of future water usage, and improve the user experience, the specific implementation steps are as follows:

[0189] S41. The characteristic variables include the weekly water consumption Zz y , the fluctuation period T b and the growth rate θ z . Integrate the weekly water consumption Zz y , the fluctuation period T b and the growth rate θ z in the same time period into a water usage vector Yw s . The expression is:

[0190] Yw s = [Zz y , T b , θ z

[0191] Among them, Yw s represents the s-th water usage vector; the weekly water consumption Zz y , the fluctuation period T b and the growth rate θ z are only three types of analysis variables for water consumption. In practice, more types of variables will be used to analyze water consumption. In this invention, these three variables are used as examples for illustration.

[0192] S42. Calculate the standard value B s according to the water usage vector Yw z . The calculation formula is:

[0193]

[0194] Among them, represents the average of the water usage vector Yw s , and m represents the number of the water usage vector Yw s ;

[0195] S43. Take a standard value B z as the initial center Zx c , and calculate the numerical spacing Rj z according to the standard value B c and the initial center Zx s . The calculation formula is:

[0196] Rj s = ||B z - Zx c || 2 ​

[0197] Among them, Rj s represents the s-th numerical interval;

[0198] S43. Change the initial center Zx c and repeat step S43 to obtain multiple numerical intervals Rj s , calculate the optimal interval value Rzj k , and use the initial center Zx k corresponding to the optimal interval value Rzj c as the classification center fl x , and the expression of the optimal interval value Rzj k is:

[0199]

[0200] Among them, L represents the number of numerical intervals Rj s ;

[0201] S44. Calculate the group average value WC x according to the classification center fl s , and the calculation formula is:

[0202]

[0203] Among them, B represents the standard value B z and the number of groups of the classification center fl x ;

[0204] S45. Calculate the optimal classification value Zy s according to the group average value WC v , and the calculation formula is:

[0205]

[0206] Among them, Zy v represents the v-th optimal classification value, represents the number of Q classification quantities divided in the (s - 1)-th group average value;

[0207] S46. Repeat step S45 to obtain multiple optimal classification values Zy v , and use the classification quantity Q corresponding to the minimum value in the optimal classification value Zy v as the optimal classification scheme, and obtain the user classification set;

[0208] S47. Construct a water use prediction model through the time convolution network method, input the data in the user classification set into the water use prediction model, and obtain the water volume prediction data Yc w, The temporal convolutional network method is a model for predicting water consumption established by combining a convolutional neural network in the time dimension. It is a commonly used method for establishing models and will not be elaborated here. By obtaining water usage data at regular intervals, it is possible to complete the regular monitoring of water usage data and obtain multiple water usage characteristics for calculating water usage data, enabling a more accurate analysis of users' water usage habits, facilitating a comprehensive prediction of users' water usage situations, and providing the future water usage situations of users. This allows users to not only quickly and efficiently obtain water usage information but also know the future water consumption, enhancing the user experience.

[0209] S5. According to the water volume prediction data Yc w Calculate the dimensionality reduction value Jw z , and based on the dimensionality reduction value Jw z Generate a visualized three-dimensional graph; the water volume prediction data Yc obtained through the water usage prediction model w Contains data in multiple dimensions, which will increase the operation steps of the algorithm during the analysis process. To solve this problem, it is necessary to perform dimensionality reduction on the water volume prediction data Yc w The specific implementation steps are as follows:

[0210] S51. Convert the water volume prediction data Yc w Into an m×n analysis matrix Fx through the data conversion method j , and the expression is:

[0211]

[0212] Among them, R 11 Represents the data value in the first row and first column of the analysis matrix Fx j , n represents the number of types of characteristic variables of the water volume prediction data Yc w , and m represents the number of water volume prediction data Yc in each type of characteristic variable w ;

[0213] S52. Calculate the variance matrix Xf of the analysis matrix Fx j , and the calculation formula is: j

[0214]

[0215] Among them, Xf j Represents the jth variance matrix;

[0216] S53. Perform eigenvalue decomposition on the variance matrix Xf through the eigenvalue decomposition method j To obtain the eigenmatrix Tz j And the diagonal value λ corresponding to the eigenmatrix Tz j ; i ;

[0217] S54. Calculate the dimensionality reduction value Jw according to the feature matrix Tz j The calculation formula is as follows: z

[0218] Jw z =(Fx j -F j )×Tz j

[0219] where F j represents the average value of the analysis matrix Fx j ;

[0220] S55. Calculate the weight value Qz according to the contribution value λ i and the feature matrix Tz j , and the calculation formula is as follows: d

[0221]

[0222] where Qz d represents the d-th weight value;

[0223] S56. Sort the dimensionality reduction value Jw from large to small according to the weight value Qz d and map the sorted result into a three-dimensional space to obtain a visualized three-dimensional graph. By performing dimensionality reduction processing on the water usage data, the data dimension can be reduced, the curse of dimensionality problem can be alleviated. While improving the method performance, it can also make the dimensionality-reduced data be visualized more vividly and comprehensively. Furthermore, the water usage situation obtained by the user is a three-dimensional visualized image, enabling the user to obtain water usage information more intuitively and quickly. z

[0224] S6. Send the visualized three-dimensional graph to the user data center, and the user data center will display the visualized three-dimensional graph to the user through a display screen. During the data transmission and storage process, in order to ensure the privacy of user data, the present invention also encrypts the data through a hash algorithm. Since the hash algorithm is a commonly used encryption method, it will not be elaborated here.

[0225] Please refer to Figure 2 , which shows the structural block diagram of the automatic meter reading system for water meters provided in this embodiment. The automatic meter reading system includes a preliminary processing module, a preprocessing module, a feature module, a classification and prediction module, a dimensionality reduction module, and a visualization module.

[0226] The preliminary processing module is used to collect the original data of the water meter running at any stage and perform preliminary outlier rejection on the original data to obtain anomaly-free data; the preprocessing module is used to perform in-depth preprocessing on the anomaly-free data to obtain preprocessed data; the feature module is used to divide the preprocessed data into multiple water volume data Wa according to time periods​​​(t) , calculate the water volume data Wa (t) Characteristic variables; a classification and prediction module for classifying users according to the characteristic variables to obtain a user classification set, and obtaining water volume prediction data Yc by means of a temporal convolutional network method w ; a dimensionality reduction module for calculating a dimensionality reduction value Jw according to the water volume prediction data Yc w and generating a visualized three-dimensional graph based on the dimensionality reduction value Jw z , and a visualization module for sending the visualized three-dimensional graph to the user data center z Those of ordinary skill in the art can understand that all or part of the steps in implementing the above-described embodiment method can be completed by instructing relevant hardware through a program. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes

[0227] The above embodiments have been described in detail for the present invention. Specific examples are used herein to illustrate the principle and implementation of the present invention. The description of the above embodiments is only for helping to understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention

[0228] ​

Claims

1. An automatic meter reading method for water meters, characterized in that, The method includes the following steps: S1. Collect the original data of the water meter running at any stage, and perform preliminary outlier rejection on the original data to obtain anomaly-free data; S2. Perform in-depth preprocessing on the anomaly-free data to obtain preprocessed data; S3. Divide the preprocessed data into multiple water volume data Wa according to time periods (t) , and calculate the characteristic variables of the water volume data Wa (t) . S4. Classify users according to characteristic variables to obtain a user classification set, and obtain water volume prediction data Yc through the temporal convolutional network method w ; S5. According to the water volume prediction data Yc w Calculate the dimensionality reduction value Jw z , and based on the dimensionality reduction value Jw z Generate a visual three-dimensional graph; S6. Send the visualized 3D map to the user data center.

2. The automatic meter reading method according to claim 1, wherein In step S1, the specific implementation steps are as follows: S11. Arrange the original data in ascending order to generate a processing sequence C a , and calculate the first extreme value A1 and the second extreme value A2 according to the processing sequence C a . The calculation formulas are as follows: A1 = (C max + C min ) × 25% A2 = (C max + C min ) × 75% Among them, C max and C min respectively represent the maximum value and the minimum value in the processing sequence C a ; S12. Calculate the measured value B based on the first extreme value A1 and the second extreme value A2 c , and the calculation formula is: B c = A2 - A1 Among them, B c represents the c-th measured value; S13. According to the measured value B c Calculate the maximum value J a representing the normal data range in the processing sequence C jd and the minimum value J jx , and the calculation formula is: J jd = A2 + 1.5 × B c J jx = A1 - 1.5 × B c Among them, J jd represents the d-th maximum value, and J jx represents the x-th minimum value; S14. According to the maximum value J jd and the minimum value J jx to eliminate the outliers of the processing sequence C a ; If C a ≤ J jx or C a ≥ J jd , then the processing sequence C a is an outlier and is removed, obtaining the vacant position K cy ; If J jx <C a <J jd , then the processing sequence C a is a normal value and is retained; S15. Restore the data order of the processing sequence C cy containing the empty bit K a to obtain anomaly-free data.

3. The automatic meter reading method according to claim 1, characterized in that, In step S2, the specific implementation steps are as follows: S21. According to the vacant bit K cy Determine the adjacent vacant bits K cy+1 and K cy-1 , calculate the filling value Tc cy for filling the vacant bit K h in the anomaly-free data. The calculation formula is: where Tc h represents the h-th padding value; S22. Fill the filling value Tc h into the vacant bit K cy of the data without anomalies to obtain the data without missing values Ws f ; S23. Calculate the detection value fw without missing data z , and the calculation formula is as follows: where erf -1 represents the inverse function of the error function, and p represents the cumulative probability value; S24. Obtain the number of data N without missing data by the counting method, and calculate the quantile value Fw according to the detection value fw z The calculation formula is as follows: d ​ Among them, Fw d represents the d-th quantile value, and fw z represents the z-th detected value; S25. Sort the data set Ws without missing data f in ascending order to obtain the sorted data set P; S26. Obtain the test coefficient H by looking up a table according to the quantile value Fw d and the number of data N, and calculate the test value Jh according to the test coefficient H b , and the calculation formula is: Among them, Jh b represents the b-th test value, represents the average value of the non-missing data Ws f of.

4. The automatic meter reading method according to claim 3, wherein In step S26, the specific implementation steps are as follows: S261. Obtain the critical value p by looking up a table according to the number of data N and the test value Jh b Obtain the critical value p by looking up a table S262. Determine whether the data Ws without missing values is a normal distribution according to the critical value p f or not; If p < 0.05, it indicates non-compliance with the normal distribution, and proceed to step S263; If p ≥ 0.05, it indicates compliance with the normal distribution, and proceed to step S265; S263. Calculate the non-positive detection value Jy under non-normal distribution z , and the calculation formula is as follows: Among them, Ws f-1 and Ws1 respectively represent the (f - 1)-th and the 1-st data without missing values; S264. Detect outliers for the non-missing data Ws according to the non-positive inspection value Jy z for the non-missing data Ws f by performing outlier detection; If Jy z < 2, it is a normal value and preprocessed data is generated; If Jy z ≥ 2, it is marked as an outlier, and the first replacement value Th e is calculated and the outlier is replaced; S265. Calculate the normal inspection value Zj under the normal distribution c , and the calculation formula is as follows: where s represents the sample standard deviation of the data Ws without missing data f ; S266. Detect outliers for the data set without missing values Ws based on the normal inspection value Zj c for the data set without missing values Ws f to perform outlier detection; If Zj c < 2, it is a normal value and preprocessed data is generated; If Zj c ≥ 2, it is marked as an outlier, and the second replacement value Te h is calculated and the outlier is replaced.

5. The automatic meter reading method according to claim 4, wherein The first replacement value Th e and the second replacement value Te h are calculated by the following formula: Among them, Ws f+1 and Ws f-1 respectively represent the (f + 1)-th and (f - 1)-th data without missing data located on both sides of the data without missing data Ws f on both sides.

6. The automatic meter reading method according to claim 1, wherein In step S3, the specific implementation steps are as follows: S31. Define the water volume data for the t-th time period as Wa (t) , and calculate the periodic model Zq (t) of the water volume data Wa m . The calculation formula is: Zq m = D·sin(ρt + δ) + G Where D represents the model coefficient, ρ represents the periodic coefficient, δ represents the initial phase, and G represents the model offset value; S32. Calculate the weekly water consumption Zz y , and the calculation formula is: Among them, n represents the number of water volume data Wa (t) within one week, and Wa (t) represents the water volume data at the t-th time period; S33. Calculate the fluctuation period T of water consumption according to the periodic coefficient ρ b , and the calculation formula is: Among them, T b represents the b-th fluctuation period; S34. Calculate the growth rate θ of the water volume data z , and the calculation formula is as follows: Among them, Zq m+1 represents the periodic model value of the (m + 1)-th water volume data.

7. The automatic meter reading method according to claim 1, wherein In step S4, the specific implementation steps are as follows: S41. The characteristic variables include the weekly water consumption Zz y , the fluctuation period T b and the growth rate θ z . The weekly water consumption Zz y , the fluctuation period T b and the growth rate θ z in the same time period are integrated into a water consumption vector Yw s . The expression is as follows: Yw s =(Zz y , T b , θ z ​ Among them, Yw s represents the s-th water consumption vector; S42. Calculate the standard value B based on the water consumption vector Yw s The calculation formula is as follows: z ​ Among them, represents the average of the water consumption vector Yw s , m represents the number of the water consumption vectors Yw s . S43. Take a standard value B z as the initial center Zx c , and calculate the numerical distance Rj z according to the standard value B c and the initial center Zx s . The calculation formula is: Rj s = ||B z -Zx c || 2 where Rj s represents the s-th numerical interval; S43. Change the initial center Zx c And repeat step S43 to obtain multiple numerical spacings Rj s , calculate the optimal spacing value Rzj k , and the optimal spacing value Rzj k The corresponding initial center Zx c Is used as the classification center fl x , the optimal spacing value Rzj k The expression of is: where L represents the number of numerical spacings Rj s ; S44. According to the classification center fl x Calculate the group average value WC s , and the calculation formula is: Among them, B represents the standard value B z and the classification center fl x the number of groups; S45. Calculate the optimal classification value Zy based on the group average value WC s v The calculation formula is as follows:​ Among them, Zy v represents the v-th optimal classification value, indicating the number of Q classifications divided in the (s - 1)-th population average value; S46. Repeat step S45 to obtain multiple optimal classification values Zy v , and use the classification quantity Q corresponding to the minimum value among the optimal classification values Zy v as the optimal classification scheme, and obtain the user classification set; S47. Construct a water consumption prediction model through the temporal convolutional network method, and input the data of user classification concentration into the water consumption prediction model to obtain the water volume prediction data Yc w .

8. The automatic meter reading method according to claim 1, characterized in that, In step S5, the specific implementation steps are as follows: S51. Convert the water volume prediction data Yc w into an m×n analysis matrix Fx through the data conversion method j . The expression is as follows: Among them, R 11 represents the data value of the first row and the first column in the analysis matrix Fx j , n represents the number of types of characteristic variables of the water volume prediction data Yc w , and m represents the number of water volume prediction data Yc w in each type of characteristic variable; S52. Calculate and analyze the matrix Fx j Variance matrix Xf of j , and the calculation formula is: Among them, Xf j represents the j-th variance matrix; S53. Perform eigen - decomposition on the variance matrix Xf j to obtain the eigen - matrix Tz j and the diagonal values λ j corresponding to the eigen - matrix Tz i ; S54. According to the feature matrix Tz j Calculate the dimensionality reduction value Jw z , and the calculation formula is: Jw z = (Fx j - F j ) × Tz j Among them, F j represents the average value of the analysis matrix Fx j ; S55. According to the contribution value λ i and the feature matrix Tz j calculate the weight value Qz d , and the calculation formula is: Among them, Qz d represents the d-th weight value; S56. According to the weight value Qz d Arrange the dimensionality reduction value Jw z in descending order, and map the sorting result into a three-dimensional space to obtain a visualized three-dimensional graph.

9. A system applied to the automatic meter reading method described in any one of the above claims 1-8, characterized in that, The system includes: A preliminary processing module for collecting the original data of the water meter running at any stage and performing preliminary outlier rejection on the original data to obtain anomaly-free data; A preprocessing module for performing in-depth preprocessing on the anomaly-free data to obtain preprocessed data; A feature module for dividing the preprocessed data into multiple water volume data Wa according to time periods (t) and calculating the feature variables of the water volume data Wa (t) ; A classification prediction module, which is used to classify users according to feature variables to obtain a user classification set, and obtain water volume prediction data Yc through the temporal convolutional network method w ; Dimensionality reduction module, used to calculate the dimensionality reduction value Jw based on the water volume prediction data Yc w and generate a visualized three-dimensional graph based on the dimensionality reduction value Jw z ; z ​ A visualization module for sending the visualized 3D map to the user data center.

Images