CN116221627A - Intelligent identifying and early warning method, equipment and storage medium for leakage event of water supply network system - Google Patents
Intelligent identifying and early warning method, equipment and storage medium for leakage event of water supply network system Download PDFInfo
- Publication number
- CN116221627A CN116221627A CN202310037279.XA CN202310037279A CN116221627A CN 116221627 A CN116221627 A CN 116221627A CN 202310037279 A CN202310037279 A CN 202310037279A CN 116221627 A CN116221627 A CN 116221627A
- Authority
- CN
- China
- Prior art keywords
- value
- time
- predicted
- data
- monitoring data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F17—STORING OR DISTRIBUTING GASES OR LIQUIDS
- F17D—PIPE-LINE SYSTEMS; PIPE-LINES
- F17D5/00—Protection or supervision of installations
- F17D5/02—Preventing, monitoring, or locating loss
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Mechanical Engineering (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses an intelligent identifying and early warning method, intelligent identifying and early warning equipment and a storage medium for leakage events of a water supply network system, and relates to the technical field of water supply network management and control. Wherein the method comprises the following steps: acquiring flow monitoring data and preprocessing the flow monitoring data; inputting the feature vector into a prediction model which is trained, and outputting a predicted value at the time t+1; feedback correction is carried out on the predicted value output by the predicted model, and whether the input of the predicted model at the time t+2 is adjusted or not is confirmed according to the deviation of the predicted value and the measured value; judging whether the number of times that the residual error of the predicted value and the measured value exceeds the specified multiple standard deviation exceeds a preset value or not by adopting the Laida criterion: if yes, identifying that a leakage event occurs and giving an alarm; otherwise, no alarm operation is performed. Compared with the prior art, the method and the device have the advantages that a large amount of historical flow detection data are not required to be input into the model, errors caused by the input of abnormal data are avoided through feedback correction, long-term stable use of the prediction model is realized, and early warning precision of leakage events is improved.
Description
Technical Field
The invention relates to the technical field of water supply pipe management and control, in particular to an intelligent identifying and early warning method, equipment and storage medium for leakage events of a water supply pipe network system.
Background
The water supply network is a main component of urban life line engineering and plays a role in maintaining normal running and economic functions of cities. When the water supply network is leaked, the data collected by the flow sensor near the leakage point can be abnormal, so that the abnormal detection can be carried out by adopting a prediction-classification method in data driving, thereby realizing intelligent identification and early warning of leakage events.
In the conventional prediction-classification method, the model training process in the prediction stage is often to input more samples into the model so that the model can keep higher precision and usability in the future, but the data acquisition and monitoring control system is built later in the domestic water supply network system, so that huge historical monitoring data of 3-5 years may not be provided, and even though water companies in partially developed cities may be able to provide the data, the following problems still face: (1) The urban development of China is rapid, population transition is large, the topology structure of a water supply pipe network and the water consumption mode of users can be greatly changed in the past 3-5 years, the original industrial area can become a residential area and the like, the correlation between historical data and future data is low, and the training value is not achieved; (2) The model training difficulty can be greatly increased by inputting huge data sets for 3-5 years at a time, and the requirement on the computational power of a computer is too high.
The model prediction process in the prediction stage of the traditional prediction-classification method often adopts a traditional iterative prediction method, and errors of the traditional iterative prediction method are accumulated continuously along with a time axis in the actual use process, so that larger errors are caused. In the traditional prediction-classification method, single-threshold classification is adopted in the classification stage, the data collected by the flow sensor in the water supply network under the influence of water fluctuation of a user can show larger fluctuation, and the valve scheduling and the quality problem of the flow sensor in the water supply network can cause abnormal flow monitoring data, so that accurate leakage signals are difficult to identify in the single-threshold classification.
Disclosure of Invention
The invention provides an intelligent identifying and early warning method, equipment and a storage medium for leakage events of a water supply network system, which are used for overcoming the defects of the prior art that a large amount of historical data are needed and the accuracy is poor.
In order to solve the technical problems, the technical scheme of the invention is as follows:
in a first aspect, a method for intelligently identifying and early warning a leakage event of a water supply network system includes:
acquiring flow monitoring data and preprocessing the flow monitoring data; the flow monitoring data comprises a historical flow actual measurement value and a current t-moment flow actual measurement value;
generating a feature vector based on the preprocessed flow monitoring data;
inputting the feature vector into a prediction model which is trained, and outputting a predicted value at the time t+1; the prediction model comprises an input layer, a hidden layer and an output layer, wherein the hidden layer comprises a GRU layer, a Dropout layer and an Attention layer;
feedback correction is carried out on the predicted value output by the predicted model, and whether the input of the predicted model at the time t+2 is adjusted or not is confirmed according to the deviation of the predicted value and the measured value;
judging whether the number of times that the residual error of the predicted value and the measured value of a single day exceeds the standard deviation of a designated multiple exceeds a preset value or not by adopting the Laida criterion: if yes, identifying that a leakage event occurs and giving an alarm; otherwise, no alarm operation is performed.
In a second aspect, a computer device comprises a memory storing a computer program and a processor implementing the method of the first aspect when executing the computer program.
In a third aspect, a computer storage medium has instructions stored therein, which when executed on a computer, cause the computer to perform the method according to the first aspect.
Compared with the prior art, the technical scheme of the invention has the beneficial effects that: according to the method, a large amount of historical flow detection data is not needed to be used as an input training model, errors caused by the input of abnormal data are avoided through feedback correction, and long-term stable use of the prediction model is realized; in addition, the invention reduces false alarms caused by single abnormal values, abnormal sensors, valve scheduling and the like, and improves the early warning precision of leakage events.
Drawings
FIG. 1 is a schematic flow chart of a method for intelligently identifying and pre-warning leakage events of a water supply network system in embodiment 1;
FIG. 2 is a schematic diagram of a feedback calibration flow chart in embodiment 1;
FIG. 3 is a schematic diagram of the topology of the leak location and flow sensor locations in example 2;
FIG. 4 is a diagram of historical flow monitoring data in example 2;
FIG. 5 is a schematic diagram of the filling and masking flow in embodiment 2;
FIG. 6 is a schematic diagram of the structure of a prediction model in example 2;
FIG. 7 is a graph showing the comparison of predicted and measured values of 2021-11-20 flow data in example 2;
FIG. 8 is a graph showing the comparison of predicted and measured values of 2021-11-21 traffic data in example 2;
FIG. 9 is a graph showing the comparison of predicted and measured values of 2021-11-22 flow data in example 2;
FIG. 10 is a graph showing the comparison of predicted and measured values of 2021-11-23 flow data in example 2;
FIG. 11 is a graph of predicted and measured flow data for 2021-11-24 in example 2.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the present patent;
for the purpose of better illustrating the embodiments, certain elements of the drawings may be omitted, enlarged or reduced and do not represent the actual product dimensions;
it will be appreciated by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
Example 1
The embodiment provides a method for intelligently identifying and early warning leakage events of a water supply network system, referring to fig. 1, which comprises the following steps:
acquiring flow monitoring data and preprocessing the flow monitoring data; the flow monitoring data comprises a historical flow actual measurement value and a current t-moment flow actual measurement value;
generating a feature vector based on the preprocessed flow monitoring data;
inputting the feature vector into a prediction model which is trained, and outputting a predicted value at the time t+1; the prediction model comprises an input layer, a hidden layer and an output layer, wherein the hidden layer comprises a GRU layer, a Dropout layer and an Attention layer;
feedback correction is carried out on the predicted value output by the predicted model, and whether the input of the predicted model at the time t+2 is adjusted or not is confirmed according to the deviation of the predicted value and the measured value;
judging whether the number of times that the residual error of the predicted value and the measured value of a single day exceeds the standard deviation of a designated multiple exceeds a preset value or not by adopting the Laida criterion: if yes, identifying that a leakage event occurs and giving an alarm; otherwise, no alarm operation is performed.
In the existing iterative prediction scheme, the flow prediction data of the current moment is mainly transmitted to the flow prediction data of the next moment to realize rolling prediction. In the embodiment, an improved multi-step prediction method is provided, predicted flow data is not transmitted any more, actual measurement flow data at the current moment is transmitted to perform next prediction, whether the input of a t+2 moment prediction model needs to be adjusted is only confirmed in a feedback correction link according to the deviation between a predicted value and an actual measurement value, and the accuracy of prediction is improved.
In a preferred embodiment, the standard deviation is expressed as follows:
in sigma t The standard deviation at time t; x is x tj An actual measurement value at the j-th day t;the average value of the flow data measured at the time t for all days of model training is represented as follows:
where n represents the total number of days of flow data used for model training.
In a preferred embodiment, the preprocessing comprises:
for the repeated values in the flow monitoring data, one of the repeated values is reserved, and the rest of the repeated values are deleted;
and filling missing values in the flow monitoring data by adopting a linear interpolation method, wherein the interpolation process is shown in the following formula:
wherein x is t The value which needs to be filled at the time t is represented as a missing value; x is x t+1 The measured value at the time t+1 is the next time of the missing value; x is x t-1 The measured value at the time t-1 is the previous time of the missing value; Δt represents the time difference between the time t+1 and the time t-1;
for inconsistent acquisition frequency of flow monitoring data, resampling a time sequence and converting the time sequence into an expected frequency by a linear interpolation method;
normalizing the flow monitoring data, wherein the normalization process is as follows:
wherein X is min Is the minimum value in the input sequence data; x is X max Is the maximum value in the input sequence data; x is X i Is the original value in the input sequence data; x is X i ' is the normalized value of the raw data in the input sequence data.
Resampling refers to the conversion of a time sequence from one time step to another.
In a specific implementation process, for the problem of inconsistent acquisition frequency, the time sequence is firstly up-sampled, namely encrypted into a high-precision time sequence with consistent frequency, then the flow value corresponding to the newly added time sequence is complemented by adopting a linear filling method, and finally the time sequence is processed into the expected frequency through down-sampling.
In a preferred embodiment, the generating the feature vector based on the preprocessed flow monitoring data includes:
extracting the last 30 days of flow monitoring data from the preprocessed flow monitoring data, generating 23 x k samples, which take into account weekly periodicity and daily trends, comprising two sets of feature vectors, as follows:
wherein x represents actual measurement monitoring data, and k represents the data quantity acquired in one day;
and introducing a filling mechanism to supplement a value of '1' as a placeholder into the feature vector with a shorter length, keeping the length of the feature vector in the sample consistent, and introducing a masking mechanism to block the placeholder.
In the preferred embodiment, features and tags are input into a predictive model in pairs to achieve supervised learning, wherein the tags represent predicted values based on corresponding times of the single day, the predicted values being generated based on the flow monitoring data acquired k times before the corresponding times and the flow monitoring data acquired at the same time as the previous 7 days.
Furthermore, in the preferred embodiment, since the lengths of the feature vectors of the two sets of data in the samples are not identical, and the GRU requires that the input samples have the same length, when a plurality of samples are input and the lengths between the samples are different, a filling mechanism and a masking mechanism are added, so that the normal operation of the model can be ensured. Since there is no negative value in the historical traffic monitoring data, treating-1 as a placeholder to treat two eigenvectors of different lengths as equal length vectors, and then masking-1 by a masking mechanism to ensure that the introduced placeholder does not affect the model accuracy.
In a preferred embodiment, in the hidden layer of the prediction model, the GRU layer includes a first GRU layer and a second GRU layer, and the dropout layer includes a first dropout layer and a second dropout layer; the first GRU layer, the first dropout layer, the second GRU layer, the second dropout layer and the Attention layer are sequentially connected.
In the preferred embodiment, a Dropout layer is added after each GRU layer, so that partial neurons are deactivated randomly, the overfitting phenomenon of the prediction model can be ensured, and the generalization force of the model is enhanced.
In a preferred embodiment, referring to FIG. 2, x t The flow actual measurement value at the time t is shown; y is tt And (3) representing a predicted value of flow at the time t, wherein the predicted value output by the predicted model is subjected to feedback correction, and whether the input of the predicted model at the time t+2 is adjusted is confirmed according to the deviation between the predicted value and the measured value, specifically:
if the deviation between the predicted value of the t+1 time output by the prediction model and the obtained measured value of the t+1 time exceeds 1 time standard deviation, in the t+2 time, the input of the prediction model comprises the measured value of the historical flow and the predicted value of the t+1 time, and does not comprise the measured value of the t+1 time.
In a preferred embodiment, the step of determining whether the number of times that the residual error between the predicted value and the measured value on a single day exceeds the standard deviation of the specified multiple exceeds the preset value by using the rad criterion includes:
(1) Single day predicted and measured valuesResidual error exceeding sigma t The time mark of (1) is 1, and when 8 times are continuously marked to be more than or equal to 1, an alarm is given;
(2) Residual error of single day predicted value and measured value exceeds 2 sigma t The time mark of (2) is 2, and when 4 times are continuously marked to be more than or equal to 2, an alarm is given;
(3) Residual error between single day predicted value and measured value exceeds 4σ t The time mark of (2) is 4, and when 2 times are continuously marked to be more than or equal to 4, an alarm is given;
(4) Residual error between single day predicted value and measured value exceeds 8σ t And (3) the time mark is 8, and when 1 time mark is greater than or equal to 8 continuously, an alarm is given.
In an alternative embodiment, the predictive model is trained using a feedback correction training method, comprising:
enumerating a plurality of combinations by adopting an enumeration method, and constructing initial prediction models of different neuron numbers and related parameters;
and writing flow monitoring data under the normal working condition for 30 days into a characteristic matrix of 30 Xk, wherein the flow monitoring data are as follows:
wherein x represents actual measurement monitoring data, and k represents the number of data acquired by the sensor in one day;
the feature matrix is input into a prediction model for initial training, a mean square error MSE is calculated according to a predicted value and an actual measured value of the 31 st day output by the prediction model, a prediction model parameter when the mean square error MSE is minimum is reserved as an optimal parameter, and the mean square error MSE expression is as follows:
wherein x is i As a result of the actual measurement of the value,output for predictive modelA predicted value, n, is a number of data;
classifying the predicted value of the 31 st day by adopting a Laida criterion, and when the classification result is displayed as normal, directly retraining the model by using the measured values of the 2 nd day to the 31 st day and predicting the data of the 32 nd day; when the classification result shows abnormality, the model is retrained with the actual measurement values from the 2 nd day to the 30 th day and the predicted value from the 31 st day, and the data from the 32 nd day is predicted again.
Because the flow monitoring data in the water supply network is influenced by the day-night change of water used by users, the daily monitoring data can show severe fluctuation of wave crests and wave troughs. In the preferred embodiment, the abnormality detection is performed in a mode of changing along with the date, so that the adaptability of the detection result to fluctuation is improved.
In a specific implementation process, when the classification result shows that the classification result is abnormal, the water supply network system is shown to have a pipe explosion phenomenon.
It can be understood that the model is continuously trained and predicted for the following days according to the method, and can be used all the time when the model is good in performance, and the model structure is properly adjusted and then used when the model accuracy error is large.
Example 2
To verify the feasibility of the method of example 1, the present example performed simulation experiments on the method of example 1 with leakage event data on a DN500 pipe section of a water supply pipe network system.
FIG. 3 illustrates the leak location and flow sensor location topology on the pipe segment; FIG. 4 shows the flow monitoring data for the pipe sections 2021/10/20-2021/11/24, knowing that the service time actually recorded by the water service crew is 2021/11/22 13:00, and the flow sensor records flow data every 5 minutes.
The 2021/10/20-2021/11/23 flow monitoring data are obtained and preprocessed, and the preprocessed partial flow monitoring data are shown in table 1:
table 1 flow monitoring data after pretreatment
Feature extraction is performed on the preprocessed flow monitoring data, and the feature extraction comprises the following steps:
(1) Flow monitoring data for 30 days of normal conditions from 2021-10-20:00 to 2021-11-18:55 were written into a 30 x 288 feature matrix as follows:
(2) Considering the weekly periodicity and daily trend, 6624 samples were generated as follows:
referring to fig. 5, a padding mechanism is introduced to supplement a value "-1" as a placeholder to a feature vector with a shorter length, so that the length of the feature vector in a sample is kept consistent, and a masking mechanism is introduced to block the placeholder.
Because the samples contain two feature vectors, the hidden layer of the prediction model adopts a structure of connecting 2 GRU layers, 2 dropout layers and 1 Attention layer, see fig. 6. The number of neurons and other related parameters are searched through an enumeration method, and a plurality of initial prediction models are constructed.
Inputting a feature matrix corresponding to 30-day flow monitoring data of 2021-10-20:00 to 2021-11-18:55 into a plurality of initial prediction models for initial training, predicting the flow of 2021-11-19, and evaluating the accuracy of the plurality of initial models by adopting a mean square error, wherein the evaluation results are shown in table 2:
table 2 comparison of mean square error results for different predictive models
It can be seen that the optimal parameter combination is the prediction model parameter corresponding to experiment number 1, and the prediction model corresponding to the group of parameters is used for the subsequent prediction experiment.
The flow data of 2021-11-20 is predicted using the above-described optimal parameter combination correspondence model, and fig. 7 shows a comparison of the predicted and measured results for 2021-11-20 flow data. It can be seen that the day is normal, no leakage is detected, and the predicted result is fitted with the measured result.
And the flow data of 2021-11-21 is continuously predicted by adopting the prediction model, and because 2021-11-20 is in a normal working condition, the measured data of 2021-11-20 is directly replaced by the data of 2021-10-21 and is input into the prediction model. Fig. 8 shows a comparison of the predicted and measured results of 2021-11-20 traffic data, and it can be seen that the residual error between the measured value and the predicted value output by the prediction model exceeds 2 times of standard deviation after 03:05, and the prediction model gives out a leakage alarm after 4 moments, i.e. 03:20, based on the radon criterion, so that compared with the water supply staff, 33 hours and 40 minutes in advance, a great amount of water resource waste is avoided.
And (3) continuously predicting the flow data of 2021-11-22 by adopting the prediction model, and because 2021-11-21 is a leakage working condition, adjusting the input of the prediction model by feedback correction, and replacing the monitoring data of 2021-10-22 by the predicted value of 2021-11-21 to be input into the prediction model. FIG. 9 illustrates a comparison of the predicted and measured results of 2021-11-22 flow data, with the predictive model identifying a leak event and alerting, i.e., identifying a sustained leak condition.
And (3) continuously predicting the flow data of 2021-11-23 by adopting the prediction model, and inputting the monitoring data of 2021-10-23 into the prediction model by replacing the predicted value of 2021-11-22 by the input of the prediction model through feedback correction because 2021-11-22 is a leakage working condition. FIG. 10 illustrates a comparison of the predicted and measured results of 2021-11-23 flow data, with the predictive model identifying sustained leak conditions.
And (3) continuously predicting the flow data of 2021-11-24 by adopting the prediction model, and because 2021-11-22 is a leakage working condition, adjusting the input of the prediction model by feedback correction, and replacing the monitoring data of 2021-10-24 by the predicted value of 2021-11-23 to be input into the prediction model. FIG. 11 shows a comparison of the predicted and measured results of 2021-11-24 traffic data, with the predicted values output by the predictive model after 01:30 showing that the network is restored to normal, with the residual error between the predicted and measured values being within a reasonable deviation range.
Example 3
The present embodiment proposes a computer device comprising a memory storing a computer program and a processor implementing the method of embodiment 1 when executing the computer program.
Example 4
This embodiment proposes a computer storage medium, wherein instructions are stored in the computer storage medium, and when executed on a computer, cause the computer to perform the method described in embodiment 1.
The same or similar reference numerals correspond to the same or similar components;
the terms describing the positional relationship in the drawings are merely illustrative, and are not to be construed as limiting the present patent;
it is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Other variations or modifications of the above teachings will be apparent to those of ordinary skill in the art. It is not necessary here nor is it exhaustive of all embodiments. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (10)
1. The intelligent identifying and early warning method for the leakage event of the water supply network system is characterized by comprising the following steps of:
acquiring flow monitoring data and preprocessing the flow monitoring data; the flow monitoring data comprises a historical flow actual measurement value and a current t-moment flow actual measurement value;
generating a feature vector based on the preprocessed flow monitoring data;
inputting the feature vector into a prediction model which is trained, and outputting a predicted value at the time t+1; the prediction model comprises an input layer, a hidden layer and an output layer, wherein the hidden layer comprises a GRU layer, a Dropout layer and an Attention layer;
feedback correction is carried out on the predicted value output by the predicted model, and whether the input of the predicted model at the time t+2 is adjusted or not is confirmed according to the deviation of the predicted value and the measured value;
judging whether the number of times that the residual error of the predicted value and the measured value of a single day exceeds the standard deviation of a designated multiple exceeds a preset value or not by adopting the Laida criterion: if yes, identifying that a leakage event occurs and giving an alarm; otherwise, no alarm operation is performed.
2. The intelligent identifying and early warning method for leakage events of a water supply network system according to claim 1, wherein the expression of the standard deviation is as follows:
in sigma t The standard deviation at time t; x is x tj An actual measurement value at the j-th day t;the average value of the flow data measured at the time t for all days of model training is represented as follows:
where n represents the total number of days of flow data used for model training.
3. The intelligent identifying and early warning method for leakage events of a water supply network system according to claim 1, wherein the preprocessing comprises the following steps:
for the repeated values in the flow monitoring data, one of the repeated values is reserved, and the rest of the repeated values are deleted;
and filling missing values in the flow monitoring data by adopting a linear interpolation method, wherein the interpolation process is shown in the following formula:
wherein x is t The value which needs to be filled at the time t is represented as a missing value; x is x t+1 The measured value at the time t+1 is the next time of the missing value; x is x t-1 The measured value at the time t-1 is the previous time of the missing value; Δt represents the time difference between the time t+1 and the time t-1;
for inconsistent acquisition frequency of flow monitoring data, resampling a time sequence and converting the time sequence into an expected frequency by a linear interpolation method;
normalizing the flow monitoring data, wherein the normalization process is as follows:
wherein X is min Is the minimum value in the input sequence data; x is X max Is the maximum value in the input sequence data; x is X i Is the original value in the input sequence data; x is X i ' is the normalized value of the raw data in the input sequence data.
4. The intelligent identifying and early warning method for leakage events of a water supply network system according to claim 1, wherein the generating feature vectors based on the preprocessed flow monitoring data comprises:
extracting the last 30 days of flow monitoring data from the preprocessed flow monitoring data, generating 23 x k samples, which take into account weekly periodicity and daily trends, comprising two sets of feature vectors, as follows:
wherein x represents actual measurement monitoring data, and k represents the data quantity acquired in one day;
and introducing a filling mechanism to supplement a value of '1' as a placeholder into the feature vector with a shorter length, keeping the length of the feature vector in the sample consistent, and introducing a masking mechanism to block the placeholder.
5. The intelligent identifying and early warning method for leakage events of a water supply network system according to claim 1, wherein in the hidden layer of the prediction model, the GRU layer comprises a first GRU layer and a second GRU layer, and the dropoff layer comprises a first dropoff layer and a second dropoff layer; the first GRU layer, the first dropout layer, the second GRU layer, the second dropout layer and the Attention layer are sequentially connected.
6. The intelligent identifying and early warning method for leakage events of a water supply network system according to claim 1, wherein the feedback correction is performed on the predicted value output by the predicted model, and whether to adjust the input of the predicted model at the time t+2 is determined according to the deviation between the predicted value and the actually measured value is specifically as follows:
if the deviation between the predicted value of the t+1 time output by the prediction model and the obtained measured value of the t+1 time exceeds 1 time standard deviation, in the t+2 time, the input of the prediction model comprises the measured value of the historical flow and the predicted value of the t+1 time, and does not comprise the measured value of the t+1 time.
7. The intelligent identifying and early warning method for leakage events of a water supply network system according to claim 1, wherein the step of adopting a rada criterion to determine whether the number of times that the residual error of the predicted value and the measured value of a single day exceeds the standard deviation of a specified multiple exceeds a preset value comprises the steps of:
(1) Residual error between single day predicted value and measured value exceeds sigma t The time mark of (1) is 1, and when 8 times are continuously marked to be more than or equal to 1, an alarm is given;
(2) Residual error of single day predicted value and measured value exceeds 2 sigma t The time mark of (2) is 2, and when 4 times are continuously marked to be more than or equal to 2, an alarm is given;
(3) Residual error between single day predicted value and measured value exceeds 4σ t The time mark of (2) is 4, and when 2 times are continuously marked to be more than or equal to 4, an alarm is given;
(4) Residual error between single day predicted value and measured value exceeds 8σ t And (3) the time mark is 8, and when 1 time mark is greater than or equal to 8 continuously, an alarm is given.
8. The intelligent identifying and early warning method for leakage events of a water supply network system according to claim 7, wherein the predictive model is trained by a feedback correction training method, and the method comprises the following steps:
enumerating a plurality of combinations by adopting an enumeration method, and constructing initial prediction models of different neuron numbers and related parameters;
and writing flow monitoring data under the normal working condition for 30 days into a characteristic matrix of 30 Xk, wherein the flow monitoring data are as follows:
wherein x represents actual measurement monitoring data, and k represents the number of data acquired by the sensor in one day;
the feature matrix is input into a prediction model for initial training, a mean square error MSE is calculated according to a predicted value and an actual measured value of the 31 st day output by the prediction model, a prediction model parameter when the mean square error MSE is minimum is reserved as an optimal parameter, and the mean square error MSE expression is as follows:
wherein x is i As a result of the actual measurement of the value,n is a number of data for the predicted value output by the prediction model;
classifying the predicted value of the 31 st day by adopting a Laida criterion, and when the classification result is displayed as normal, directly retraining the model by using the measured values of the 2 nd day to the 31 st day and predicting the data of the 32 nd day; when the classification result shows abnormality, the model is retrained with the actual measurement values from the 2 nd day to the 30 th day and the predicted value from the 31 st day, and the data from the 32 nd day is predicted again.
9. A computer device comprising a memory storing a computer program and a processor implementing the method of any of claims 1-8 when the computer program is executed by the processor.
10. A computer storage medium having instructions stored therein, which when executed on a computer, cause the computer to perform the method of any of claims 1-8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310037279.XA CN116221627A (en) | 2023-01-10 | 2023-01-10 | Intelligent identifying and early warning method, equipment and storage medium for leakage event of water supply network system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310037279.XA CN116221627A (en) | 2023-01-10 | 2023-01-10 | Intelligent identifying and early warning method, equipment and storage medium for leakage event of water supply network system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116221627A true CN116221627A (en) | 2023-06-06 |
Family
ID=86572412
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310037279.XA Pending CN116221627A (en) | 2023-01-10 | 2023-01-10 | Intelligent identifying and early warning method, equipment and storage medium for leakage event of water supply network system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116221627A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116757876A (en) * | 2023-08-21 | 2023-09-15 | 埃睿迪信息技术(北京)有限公司 | Method, device and equipment for determining water consumption of water supply partition |
CN117455709A (en) * | 2023-12-07 | 2024-01-26 | 深圳拓安信物联股份有限公司 | Leakage monitoring method and device for water supply network, electronic equipment and storage medium |
CN118195291A (en) * | 2024-05-16 | 2024-06-14 | 深圳拓安信物联股份有限公司 | Table replacement identification method, apparatus, storage medium and device |
-
2023
- 2023-01-10 CN CN202310037279.XA patent/CN116221627A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116757876A (en) * | 2023-08-21 | 2023-09-15 | 埃睿迪信息技术(北京)有限公司 | Method, device and equipment for determining water consumption of water supply partition |
CN116757876B (en) * | 2023-08-21 | 2023-11-14 | 埃睿迪信息技术(北京)有限公司 | Method, device and equipment for determining water consumption of water supply partition |
CN117455709A (en) * | 2023-12-07 | 2024-01-26 | 深圳拓安信物联股份有限公司 | Leakage monitoring method and device for water supply network, electronic equipment and storage medium |
CN118195291A (en) * | 2024-05-16 | 2024-06-14 | 深圳拓安信物联股份有限公司 | Table replacement identification method, apparatus, storage medium and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN116221627A (en) | Intelligent identifying and early warning method, equipment and storage medium for leakage event of water supply network system | |
CN111460728B (en) | Method and device for predicting residual life of industrial equipment, storage medium and equipment | |
CN110807550B (en) | Distribution transformer overload recognition and early warning method based on neural network and terminal equipment | |
CN111079989B (en) | DWT-PCA-LSTM-based water supply amount prediction device for water supply company | |
CN114611778B (en) | Reservoir water level early warning method and system based on warehousing flow | |
CN113343581B (en) | Transformer fault diagnosis method based on graph Markov neural network | |
CN114330934A (en) | Model parameter self-adaptive GRU new energy short-term power generation power prediction method | |
CN114997313B (en) | Abnormality detection method for ocean on-line monitoring data | |
CN114399235B (en) | Method and system for judging disaster risk level based on rain condition data | |
Liu et al. | Deep learning approach for sensor data prediction and sensor fault diagnosis in wind turbine blade | |
CN117725491A (en) | SCITET-based power system fault state detection and classification method | |
CN112036496A (en) | Nuclear power device fault diagnosis method and system | |
CN116522560A (en) | Active power distribution network topology identification method and system based on probabilistic neural network | |
CN116502123A (en) | Non-supervision cross-domain prediction method for residual service life of aero-engine | |
Lyu et al. | Interval prediction of remaining useful life based on convolutional auto-encode and lower upper bound estimation | |
CN109034453A (en) | A kind of Short-Term Load Forecasting Method based on multiple labeling neural network | |
GB2620602A (en) | Training of a machine learning model for predictive maintenance tasks | |
CN115293249A (en) | Power system typical scene probability prediction method based on dynamic time sequence prediction | |
CN113756786A (en) | Method for predicting time sequence indicator diagram of rod-pumped well based on deep learning | |
Mounce et al. | Development of artificial intelligence systems for analysis of water supply system data | |
CN117633456B (en) | Marine wind power weather event identification method and device based on self-adaptive focus loss | |
CN113551157B (en) | Oil-gas data acquisition system and method based on sensor network | |
CN118296388A (en) | Construction method and application method of drainage pipe network water quality monitoring model | |
CN118189074A (en) | Leakage detection method, leakage detection device, computer readable storage medium and terminal equipment | |
CN113780689A (en) | Energy router service life prediction method and device based on artificial intelligence |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |