JP2023150219A - Gas leak estimation system and method for estimating gas leak - Google Patents

Abstract

To provide a gas leak estimation system and a method for estimating a gas leak that can monitor a leak of gas from a gas pipe.SOLUTION: A gas leak estimation system has a first point and a second point along a gas pipe, and estimates a leakage of gas in the second point on the basis of the pressure of gas in the first point. The gas leak estimation system includes: gas pressure estimation means for determining the estimated value of information on the pressure of gas in the second point from the pressure of gas in the first point measured by a measuring unit by using a learned estimation model; and determination means for determining whether there is a gas leak in the second point on the basis of the estimated value of information on the pressure of gas in the second point and an actual measured value of pressure in a previous normal time in the second point.SELECTED DRAWING: Figure 3

Description

The present disclosure relates to a gas leak estimation system and a gas leak estimation method.

As a method for detecting gas pipe leakage, for example, Patent Document 1 discloses a method of emitting laser light from a vehicle onto the road surface and detecting reflected light to detect gas components leaking from gas pipes buried in the road. The technology is disclosed. Moreover, Patent Document 2 discloses a technique for detecting gas leakage by comparing the flow rates of gas at two points provided in a pipe.

Japanese Patent Application Publication No. 2016-80629 Japanese Patent Application Publication No. 9-61283

The technique disclosed in Patent Document 1 detects gas leakage by driving a vehicle, and therefore is not suitable for monitoring leakage from gas piping on a daily basis. In addition, the technology of Patent Document 2 is mainly aimed at detecting gas leaks at end homes or in the vicinity, and may not be suitable for detecting leaks in gas piping upstream from high-pressure gas pipes. It will be done. An object of the present disclosure is to provide a gas leak estimation system and a gas leak estimation method that can monitor leaks in gas pipes.

A gas leak estimation system according to an embodiment of the present disclosure includes a first point and a second point located along a gas pipe, and estimates a gas leak at the second point based on the gas pressure at the first point. The gas leak estimation system includes a measuring device located at a first point of the gas pipe and measuring the gas pressure in the gas pipe, and a measuring device that measures the gas pressure at the second point from the gas pressure at the first point. An estimation model for estimating information regarding pressure, which uses a learned estimation model by performing machine learning using actual measured values of gas pressure at normal times at the first point and the second point as training data. , a gas pressure estimating means for calculating an estimated value of information regarding the gas pressure at the second point from the gas pressure at the first point measured by the measuring device; and an estimated value of information regarding the gas pressure at the second point. , and determining means for determining whether or not a gas leak has occurred at the second point based on the actual measured value of the pressure at the second point during normal conditions in the past.

According to an embodiment of the present disclosure, a gas leak estimation system and a gas leak estimation method capable of monitoring gas pipe leaks are provided.

FIG. 1 is a schematic diagram illustrating an example of a gas pipe to which the gas leak estimation system and gas leak estimation method of the present embodiment can be applied. FIG. 2 is a schematic diagram showing the configuration of the gas leak estimation system 101 of this embodiment. FIG. 3 is a block diagram showing the functions of the information processing device 30. FIG. 4 is a flowchart showing a method for estimating gas leakage according to this embodiment. FIG. 5 is a diagram showing an example of the results of the example. FIG. 6 is a diagram showing the relationship between the threshold value, the time required to detect a gas leak, and the number of false detections in the example. FIG. 7 is a diagram showing the relationship between the threshold value, the time required to detect a gas leak, and the number of false detections in the example.

In recent years, AI (Artificial Intelligence) technology has developed, and attempts are being made to utilize AI in various fields. The inventor of the present application has considered estimating gas pipe leakage by using machine learning, and has come up with a gas leak estimation system and a gas leak estimation method. Hereinafter, one embodiment of the present invention will be described based on the drawings. Note that the present invention is not limited to the following embodiments, and design changes can be made as appropriate within the scope that satisfies the configuration of the present invention. In addition, in the following description, the same parts or parts having similar functions will be denoted by the same reference numerals in different drawings, and repeated description thereof will be omitted. Moreover, each structure described in the embodiment and the modified example may be combined or changed as appropriate. Further, in order to make the explanation easier to understand, in the drawings referred to below, the configuration is shown in a simplified or schematic manner, or a part of the configuration is omitted.

FIG. 1 is a schematic diagram illustrating an example of piping to which the gas leak estimation system and gas leak estimation method of the present disclosure can be applied. The gas pipe M1 is, for example, a high-pressure gas pipe, and gas flows in the direction of the arrow. For example, the inner diameter of the gas pipe M1 is, for example, 140.2 mm to 724.0 mm in diameter, and city gas is transferred at a pressure of about 1.00 MMpa to 3.92 MPa.

Points A, B, C, and D are located in the middle of the gas pipe M1, and a measuring device for measuring the gas pressure is placed at each point. Point A is on the upstream side, and point D is on the downstream side. In other words, points A to D are located on the upstream side in alphabetical order. Further, a branching gas pipe M2 is connected to the point D.

The intervals between points A to D may all be equal, or two or more intervals may be different from each other. Points A to D may be governor stations for regulating gas pressure, or valve stations where valves are provided for maintenance. Since land is required to locate these facilities, the intervals between points A to D on the actual piping are different from each other.

As described in detail below, the gas leak estimation system and gas leak estimation method of the present disclosure include a first point and a second point located along a gas pipe, based on the gas pressure at the first point, Estimate the gas leak at the second location. Specifically, an estimation model that estimates information about the gas pressure at the second point from the gas pressure at the first point is generated by machine learning using normal data, and using the learned estimation model, Information regarding the gas pressure at the second location is estimated from the gas pressure at the first location. This estimation model is specific to the location where the data used during learning was acquired. In other words, the estimation model that estimates the pressure from point A to point B can estimate gas leakage at these points with the highest accuracy.

The gas leak at the second point refers to gas leaking from the gas pipe within a predetermined distance upstream and downstream from the second point. Gas leakage in this range reduces the measured pressure at the second point. For example, if the gas pressure cannot be measured at the second point for some reason, it is possible to estimate whether gas leakage has occurred from the actual measured value of the gas pressure at the first point. If the gas pressure can be measured at the second point, the measured value and estimated value can be used to determine whether there is an abnormality in the measuring device or to confirm the measurement result.

FIG. 2 is a schematic diagram showing the configuration of the gas leak estimation system 101 of this embodiment. The gas leak estimation system of this embodiment includes measuring instruments 10A, 10B, 10C, and 10D that measure the pressure of gas in a gas pipe, and an information processing device 30. Measuring instruments 10A, 10B, 10C, and 10D are placed at points A, B, C, and D, respectively. As described above, the gas leak estimation system 101 estimates the gas leak at the second point based on the gas pressure at the first point, so it is sufficient to include at least a measuring device at a point corresponding to the first point. However, as will be described later, in order to learn the estimation model, actual measured values of pressure during normal times at the first point and the second point are required. Therefore, in order to obtain the actual measured value of the pressure at the second point, it is preferable to include a measuring device at a point corresponding to the second point.

The first point and the second point may be any two points selected from a plurality of points arranged along the gas pipe M1, and either of them may be located upstream. For example, the first point and the second point may be point A and point C. Optimum weighting is given to points A and C in the estimation model by machine learning. However, in order to obtain an estimation model that allows estimation with higher accuracy, it is preferable that the first point and the second point are adjacent to each other. Moreover, it is preferable that the first point is located upstream of the second point.

Flowmeters 11A to 11D may be further arranged at points A to D.

Data on the pressure at that point measured by the measuring instruments 10A, 10B, 10C, and 10D is transmitted to the information processing device 30 via the network 20. The network 20 may be an Internet line, a telephone line, a mobile phone line, another wireless communication line, or a dedicated line.

FIG. 3 is a block diagram showing the functions of the information processing device 30. The information processing device 30 includes a gas pressure estimation means 31, a determination means 32, and a storage means 33. In FIG. 3, the information processing device 30 is, for example, a microcomputer, a server, etc., and includes an information processing unit such as a CPU, a storage device such as a RAM, ROM, HDD, or SSD, and communication for receiving data from the network 20. Equipped with equipment. With these configurations, the gas pressure estimating means 31 and the determining means 32 are realized as software. Furthermore, the storage device 33 is realized as hardware.

Although the information processing device 30 is shown as one device in FIG. 3, some of the gas pressure estimation means 31, the determination means 32, and the storage means 33 are physically separate devices, and the network 20 It may be configured such that information can be sent and received by. For example, the storage means 33 may be a storage medium on the cloud, and the gas pressure estimating means 31 described later may also be stored on the cloud.

The gas pressure estimating means 31 estimates information regarding the gas pressure at the second point from the gas pressure at the first point using the learned estimation model. That is, upon receiving the measured value of the gas pressure at the first point, an estimated value of information regarding the gas pressure at the second point is output. The learned estimation model estimates information regarding the gas pressure at the second point from the gas pressure at the first point. The estimation model is subjected to machine learning in advance using actual measured gas pressure values at the first point and the second point under normal conditions as training data.

A branching gas pipe M2 is connected to the gas pipe M1 (FIG. 1), and a general house that consumes gas is connected downstream. Furthermore, such branching gas pipes and gas-consuming houses are not uniformly located. Therefore, the learned estimation model is an estimation model specific to the first point and the second point where machine learning was performed. Specifically, when point A and point B are selected as the first point and second point, the learned estimation model that is obtained receives the measured value of gas pressure at point A, and calculates the gas pressure at point B. It is optimized as a model that outputs estimated values of information regarding pressure.

Preferably, the estimation model estimates information regarding the gas pressure at the second location from the gas pressure at the first location and the time and date when the pressure was acquired. Further, it is preferable that the estimation model is trained using actual measured pressure values at the first point and the second point, and the times and dates when these were obtained as training data. In this case, the gas pressure estimating means 31 uses the learned estimation model to obtain an estimated value of information regarding the pressure at the second point from the gas pressure, time and date at the first point measured by the measuring instruments 10A to 10D. .

The pressure of the gas in the gas pipe changes depending on the amount of gas used, and the change in the amount of gas used is thought to have a cycle of approximately one day. Furthermore, changes in the amount of gas used over a year are also considered to have periodicity. For example, it is thought that the amount of gas used increases throughout the day from evening to night, and that the amount of gas used increases throughout the year during the winter season. Therefore, it is considered that the estimation model can more accurately estimate the pressure at the second point by further using the time and date when the gas pressure at the first point was acquired.

The estimation model may estimate information regarding the pressure of the gas at the second location from the pressure and flow rate of the gas at the first location. In this case, it is preferable that the estimation model is trained using the actual measured values of the pressures at the first point and the second point and the flow rate at the first point as training data.

Further, the gas pressure estimating means 31 uses the learned estimation model to obtain an estimated value of information regarding the pressure at the second point from the gas pressure and gas flow rate at the first point measured by the measuring instruments 10A to 10D. Since the pressure of the gas in the gas pipe is considered to change depending on the flow rate of the gas, it is considered that the estimation model can more accurately estimate the pressure at the second point by also using the flow rate at the first point.

The estimation model may estimate information regarding the gas pressure at the second point from the gas pressure at the first point and the temperature at the time the pressure was obtained. In this case, it is preferable that the estimation model is trained using actual measured values of the pressure at the first point and the second point and the temperature at the time the pressure was acquired as training data. Further, the gas pressure estimating means 31 uses the learned estimation model to obtain an estimated value of information regarding the pressure at the second point from the gas pressure and temperature at the first point measured by the measuring instruments 10A to 10D.

Since the amount of gas used is also affected by the temperature, it is thought that the estimation model can more accurately estimate the pressure at the second point by using the temperature as well.

The information regarding the gas pressure at the second point determined by the estimation model may be various parameters. For example, the information regarding the gas pressure at the second point may be the gas pressure at the second point, or may be the ratio of the pressure at the second point to the pressure at the first point.

The determining means 32 determines whether or not a gas leak has occurred at the second point based on an estimated value of information regarding the gas pressure at the second point and an actual measured value of pressure at the second point during normal times in the past. Determine whether

For example, the determining means 32 sets the estimated value of the gas pressure at the second point Ps as the estimated value of the information regarding the gas pressure at the second point, sets the actual measured value of the pressure at the second point during normal times as Pr, Calculate the error rate E obtained by the following formula.
E=100×(Ps-Pr)/Pr
Further, the determining means 32 determines that a gas leak has occurred when E exceeds a predetermined threshold value Rt.

The estimated value of the information regarding the gas pressure at the second point is a value calculated by the gas pressure estimating means 31, and the actual measured value of the pressure at the second point during normal times is stored in the storage means 33, for example. .

As explained below, the threshold value Rt may be, for example, 0.015 or more. Thereby, it is possible to determine that there is a possibility of gas leakage in as short a time as possible while suppressing false detection.

The gas leak estimation system 101 may determine the possibility of gas leak only once at any timing. However, from the viewpoint of monitoring the gas pipe M1, it is preferable to continuously repeat the monitoring at predetermined time intervals. Thereby, the presence or absence of gas leakage can be estimated by continuously monitoring the gas pipe M1.

Next, a method for estimating gas leakage according to this embodiment will be explained. FIG. 4 is a flowchart showing a method for estimating gas leakage according to this embodiment. The gas leak estimation method of the present embodiment includes a step (A) of measuring the gas pressure in the gas pipe, a step (B) of obtaining an estimated value of information regarding the gas pressure at a second point, and a step (B) of measuring the gas pressure in the gas pipe. and a step (C) of determining whether or not.

(1) Preparing a trained estimation model Before executing the gas leak estimation method of this embodiment, first prepare a trained estimation model. As mentioned above, the actual measured values of gas pressure at the first and second points during normal conditions are used as training data in the estimation model that estimates information about the gas pressure at the second point from the gas pressure at the first point. A trained estimation model is obtained by machine learning. Machine learning using training data can be performed with any information processing device. For example, the estimation model can be trained using various commercially available machine learning libraries. Furthermore, a library suitable for the gas leak estimation system and gas leak estimation method of the present disclosure can be selected from various machine learning libraries. The actual measured values of the gas pressure at normal times at the first and second points are determined by selecting any two points from points A, B, C, and D as the first and second points, and measuring at the selected points. It can be obtained using the instruments 10A to 10D.

As mentioned above, the estimation model calculates, in addition to the gas pressure at the first point, the time and date when the gas pressure at the first point was obtained, the gas flow rate at the first point, and the gas pressure at the first point. At least one selected from the temperature at the time of acquisition may be further used as an input parameter.

The obtained trained estimation model is stored in the information processing device that performs step (B) so that it can be used in the following step (B), or the trained estimation model obtained by a network etc. is stored. It is preferable that the information processing device and the information processing device that performs step (B) be able to send and receive information.

(2) Step (A) of measuring the pressure of gas in the gas pipe (S1)
At any of the points A to D selected as the first point, the pressure of the gas is measured using a measuring device. From the viewpoint of monitoring the gas pipe M1, it is preferable to repeat measurements at predetermined time intervals.

(3) Step (B) of obtaining an estimated value of information regarding the gas pressure at the second point (S2)
The gas pressure at the first point obtained in step (A) is input to the obtained trained estimation model, and an estimated value of information regarding the gas pressure at the second point is obtained.

(4) Step (C) of determining whether a gas leak has occurred (S3)
Based on the estimated value of the information regarding the pressure at the second point and the actual measured value of the pressure at the second point during normal times in the past, it is determined whether a gas leak has occurred at the second point.

The estimated value obtained in step (B) is used as the estimated value of the information regarding the pressure at the second point. As the actual measured value of the past normal pressure at the second point, for example, a value used as training data during machine learning of the estimation model can be used.

Furthermore, as described above, for example, the estimated value of the gas pressure at the second point Ps is used as the estimated value of the information regarding the gas pressure at the second point, and the actual measured value of the pressure at the second point during normal times is Pr. , the error rate E is calculated using the following formula.
E=100×(Ps-Pr)/Pr
Furthermore, if E exceeds a predetermined threshold value Rt, it is determined that a gas leak has occurred. The threshold value Rt may be, for example, 0.015 or more.

As described above, according to the gas leak estimation system and gas leak estimation method of the present disclosure, the gas pressure at the second point is estimated from the gas pressure at the first point, and furthermore, it is possible to estimate the gas pressure at the second point from the gas pressure at the first point. Estimate whether it has occurred. Therefore, even if it is not possible to measure the gas pressure at the second point due to an accident, breakdown, power outage, etc., the gas pressure at the second point can be estimated from the gas pressure at the first point. , it is possible to estimate the presence or absence of a gas leak. Furthermore, if the gas pressure can be measured at the second point, the measured value and the estimated value can be used to determine whether there is an abnormality in the measuring device or to confirm the measurement result.

By using such an estimation model based on machine learning, it is possible to determine the presence or absence of a gas leak without installing new measuring equipment. Furthermore, since gas pipes are equipped with multiple measuring instruments, gas leaks in gas pipes can be continuously monitored at each location by using the gas leak estimation system or gas leak estimation method of the present disclosure. Is possible.

The gas leak estimation system and gas leak estimation method of the present disclosure are not limited to the above embodiments, and various modifications can be made. For example, the configuration of the information processing device is not limited to the above embodiment. Furthermore, part or all of the configuration of the shape processing device may be, for example, a cloud server. Further, the estimation model may perform machine learning using teacher data, for example, periodically or irregularly at an appropriate timing. Furthermore, the estimation model may estimate the pressure at the second point using other parameters in addition to the input parameters described in this embodiment.

Examples of a gas leak estimation system and a gas leak estimation method will be described below. The gas leak estimation system and gas leak estimation method of this embodiment were verified using the measurement results of normal gas pressure at two adjacent points arranged in a gas pipe.

(1) The actual measured values of the pressure of the gas prepared as the first point and the second point were obtained every minute from 1:00 a.m. until midnight the next day. The number of data points was 1381 points. Therefore, the array of measured values of gas pressure at the first and second points also represents the time at which the pressure was acquired. For example, the actual measured value of the third gas pressure is the gas pressure obtained at 1:02 AM.

(2) The following function was set as the estimation model.
Y=F(X)
Objective variable (Y) = Pressure at the second point (estimated) / Pressure at the first point (actual measurement)
Explanatory variable (X) = Pressure at the first point Using the actual measured values of gas pressure at normal times at the first and second points described above as training data, obtain a trained estimation model using the following Python library. Ta.
・pandas
・Numpy
・scipy
・matplotlib
・japanize_matplotlib
・tqdm
・datetime
・scikit-learn
・lightgbm

(3) Since it is difficult to obtain pressure data when a gas leak actually occurs from a gas pipe, the actual measured gas pressure at the first point under normal conditions is abnormal for one hour from 12:00 p.m. Assuming that this occurred, the actual measured values were replaced with abnormal data, and this was set as verification data 1. Similarly, it was assumed that an abnormality occurred for one hour from 9:00 pm on the actual measured value of the gas pressure at the first point during normal conditions, and the actual measured value was replaced with abnormal data, which was set as verification data 2.

The abnormality data was calculated by converting the actual value of the gas ejection amount at the time of a past medium-pressure pipe breakage into a gas leak in a high-pressure state, and assumed that a gas leak of 10,000 Nm ³ /h would occur. Assuming that the amount of gas demanded increased by 10,000 Nm ³ /h at the time when the above abnormality occurred compared to normal times, we analyzed the abnormality using analysis software that can simulate the pressure state of gas pipes using the amount of gas demanded as input data. The gas pressure at the time was determined by calculation. For the pressure data during normal conditions, data from a weekday in summer was used.

(4) The error rate E at each time was calculated according to the following formula.
E=100×(Ps-Pr)/Pr
Ps is a predicted value at the second point obtained by inputting verification data 1 and verification data 2 into the learned model, and Pr is an actual value of the gas pressure at the second point during normal conditions. In addition, as normal data, the predicted value at the second point obtained by inputting the measured value of the gas pressure at the first point into the trained model is Ps, and the gas pressure at the second point during normal conditions is Ps. The error rate E was determined in the same manner, using the actual measured value of the pressure as Pr.

The results are shown in FIGS. 5 to 7. In FIG. 5, N indicates the error rate E of normal data, and E1 and E2 indicate the error rate E of the verification data 1 and 2, respectively. As can be seen from FIG. 5, the error rate E of E1 shows a large value between 12:00 pm and 1:00 pm. Similarly, the error rate of E2 shows a large value between 9:00 pm and 10:00 pm. From these results, it is considered that the learned estimation model can appropriately estimate the pressure abnormality at the second point based on the pressure at the first point.

FIG. 6 shows the relationship between the time required until the error rate E exceeds the threshold Rt and it is determined that a gas leak has occurred and the number of false detections when the threshold Rt is changed in E1. The number of false detections indicates the number of times in which the error rate E exceeds the threshold value Rt in a range other than the set abnormal data. Similarly, FIG. 7 shows the relationship between the time required until the error rate E exceeds the threshold Rt and it is determined that a gas leak has occurred and the number of false positives when the threshold Rt is changed in E2. show. The number of false detections indicates the number of times in which the error rate E exceeds the threshold value Rt in a range other than the set abnormal data.

As shown in FIGS. 6 and 7, if the threshold value Rt is made smaller, the time required to determine that a gas leak has occurred will be shortened, but the number of false detections will also increase. These results show that, for example, if Rt is set to 0.015 or more, the number of false detections can be sufficiently reduced and the occurrence of gas leakage can be determined.

When a gas leak is determined using the gas leak estimation system and the gas leak estimation method of this embodiment, and based on the determination result, for example, when deploying personnel to respond to the gas leak, the time required for the personnel to be dispatched is , the threshold value Rt can be determined according to the allowable number of false positives.

As described above, according to the present example, it was found that the pressure at the second point of the gas pipe can be estimated from the pressure at the first point of the gas pipe using the estimation model learned by machine learning.

The gas leak estimation system and gas leak estimation method of the present disclosure can also be explained as follows.

The gas leak estimation system according to the first configuration includes a first point and a second point located along a gas pipe, and a gas leak estimation system that estimates a gas leak at the second point based on the gas pressure at the first point. The system includes a measuring device located at a first point in a gas pipe and measuring the pressure of gas in the gas pipe, and an estimation model that estimates information about the gas pressure at a second point from the gas pressure at the first point. By machine learning using the actual measured values of gas pressure at normal times at the first and second points as training data, the learned estimation model is used to calculate the value at the first point measured by the measuring device. a gas pressure estimating means for calculating an estimated value of information regarding the gas pressure at the second point from the gas pressure; an estimated value of the information regarding the gas pressure at the second point; and a past normal pressure at the second point. A determining means is provided for determining whether or not a gas leak has occurred at the second location based on the actual measurement value.

In the gas leak estimation system according to the second configuration, in the first configuration, a plurality of points including a first point and a second point are located in the gas pipe, and the first point and the second point are adjacent to each other. The gas leak estimation system according to claim 1.

The gas leak estimation system according to claim 2, wherein the first point is located upstream of the second point in the second structure.

In the gas leakage estimation system according to the fourth configuration, in any one of the first to third configurations, the estimation model is based on the gas pressure at the first point and the time and date at which the pressure was obtained. The gas pressure estimating means estimates the information about the gas pressure at the first point and the second point, and learns the actual measured pressure values at the first point and the second point, and the time and date when these were obtained as training data, and the gas pressure estimation means uses the learned estimation. Using the model, an estimate of information regarding the pressure at the second location is determined from the gas pressure at the first location, time, and date measured by the measuring device.

In the gas leak estimation system according to the fifth configuration, in any one of the first to fourth configurations, the estimation model estimates information regarding the gas pressure at the second point from the gas pressure and flow rate at the first point. , the actual measured values of the pressure at the first point and the second point, and the flow rate at the first point are used as training data, and the gas pressure estimating means uses the learned estimation model to estimate the first point measured by the measuring device. An estimated value of information regarding the pressure at a second point is obtained from the gas pressure and gas flow rate at one point.

In the gas leakage estimation system according to the sixth configuration, in any one of the first to fifth configurations, the estimation model is based on the gas pressure at the first point and the temperature at the time when the pressure is acquired, and the gas leakage estimation system at the second point. The gas pressure estimating means uses the learned estimation model to estimate information about the pressure of Using this method, an estimated value of information regarding the pressure at the second point is obtained from the gas pressure and temperature at the first point measured by the measuring device.

In the gas leak estimation system according to the seventh configuration, in any one of the first to sixth configurations, the determination means sets the estimated value of pressure to Ps, and determines the actual measured value of the pressure at the second point during normal times. As Pr, calculate the error rate E obtained by the following formula,
E=100×(Ps-Pr)/Pr
If E exceeds a predetermined threshold value Rt, it is determined that a gas leak has occurred.

In the gas leak estimation system according to the eighth configuration, in the seventh configuration, Rt is 0.015 or more.

A gas leak estimation method according to a ninth configuration includes a first point and a second point located along a gas pipe, and estimates a gas leak at the second point based on the gas pressure at the first point. A method for estimating gas leakage, which includes a step (A) of measuring the gas pressure at a first point of a gas pipe, and estimating information about the gas pressure at the second point from the gas pressure at the first point. The estimation model is a trained estimation model that is machine-trained using actual measured values of gas pressure at normal times at the first and second points as training data, and is used to calculate the estimated value at the first point measured by the measuring instrument. Step (B) of calculating an estimated value of information regarding the gas pressure at the second point from the gas pressure; (C) determining whether a gas leak has occurred at two points.

In the gas leak estimation method according to the tenth configuration, in the ninth configuration, a plurality of points including a first point and a second point are located in the gas pipe, and the first point and the second point are adjacent to each other. ing.

In the gas leak estimation method according to the eleventh configuration, in the tenth configuration, the first point is located upstream of the second point.

The method for estimating gas leakage according to the twelfth configuration is that in any one of the ninth to eleventh configurations, the estimation model is based on the gas pressure at the first point and the time and date at which the pressure was obtained. Information regarding the gas pressure at the point is estimated, and the actual measured values of the pressure at the first point and the second point, as well as the time and date when these were obtained, are learned as training data. An estimated value of information regarding the pressure at the second point is obtained from the gas pressure at the first point measured by, time and date.

In the gas leak estimation method according to the thirteenth configuration, in any one of the ninth to twelfth configurations, the estimation model estimates information regarding the gas pressure at the second point from the gas pressure and flow rate at the first point. However, the actual measured values of the pressure at the first point and the second point and the flow rate at the first point are used as teacher data for learning, and step (B) is based on the gas pressure and the flow rate at the first point measured by the measuring device. An estimated value of information regarding the pressure at the second point is obtained from the gas flow rate.

In the gas leakage estimation method according to the fourteenth configuration, in any one of the ninth to thirteenth configurations, the estimation model is based on the gas pressure at the first point and the temperature at the time when the pressure is obtained. Information regarding the gas pressure is estimated and learned using the actual measured pressure values at the first and second points and the temperature at the time the pressure was obtained as training data. An estimated value of information regarding the pressure at the second point is obtained from the gas pressure and temperature at the first point.

In the gas leakage estimation method according to the fifteenth configuration, in any one of the ninth to fourteenth configurations, in step (C), the estimated value of pressure is Ps, and the pressure at the second point in the past normal time is Using the actual measurement value as Pr, calculate the error rate E obtained by the following formula,
E=100×(Ps-Pr)/Pr
If E exceeds a predetermined threshold value Rt, it is determined that a gas leak has occurred.

In the gas leak estimation method according to the 16th configuration, in the 15th configuration, Rt is 0.015 or more.

The gas leak estimation system and gas leak estimation method according to the embodiments of the present disclosure are suitably used in gas leak estimation systems and gas leak estimation methods for various gas pipes.

10A to 10D... Measuring device, 11A to 11D... Flowmeter, 20... Network, 30... Information processing device, 31... Gas pressure estimation means, 32... Judgment means, 33... Storage means, 101... Estimation system

Claims (16)

A gas leak estimation system in which a first point and a second point are located along a gas pipe, and estimates a gas leak at the second point based on the gas pressure at the first point,
a measuring device located at a first point of the gas pipe and measuring the pressure of the gas in the gas pipe;
The estimation model estimates information regarding the gas pressure at the second point from the gas pressure at the first point, and the model uses actual measured values of the gas pressure at the first point and the second point as training data during normal times. Gas pressure estimation that calculates an estimated value of information regarding the gas pressure at the second point from the gas pressure at the first point measured by the measuring device using a learned estimation model through machine learning. means and
Determining whether a gas leak has occurred at the second point based on an estimated value of information regarding the gas pressure at the second point and an actual measured value of pressure at the second point during normal times in the past. Judgment means;
Gas leak estimation system. The gas leak estimation system according to claim 1, wherein a plurality of points including the first point and the second point are located in the gas pipe, and the first point and the second point are adjacent to each other. The gas leak estimation system according to claim 2, wherein the first point is located upstream of the second point. The estimation model estimates information regarding the gas pressure at the second point from the gas pressure at the first point and the time and date when the pressure was acquired, and calculates the pressure at the first point and the second point. The actual measured values and the time and date when these were acquired are used as the training data, and
The gas pressure estimating means uses the learned estimation model to obtain an estimated value of information regarding the pressure at the second point from the gas pressure at the first point measured by the measuring device, time, and date. The gas leak estimation system according to any one of Items 1 to 3. The estimation model estimates information regarding the gas pressure at the second point from the gas pressure and flow rate at the first point, and calculates the actual measured values of the pressures at the first point and the second point, and the first point. is learned using the flow rate as the teacher data,
The gas pressure estimating means uses the learned estimation model to obtain an estimated value of information regarding the pressure at the second point from the gas pressure and gas flow rate at the first point measured by the measuring device. The gas leak estimation system according to any one of 1 to 4. The estimation model estimates information regarding the gas pressure at the second point from the gas pressure at the first point and the temperature at the time when the pressure was acquired, and calculates the actual measurement of the pressure at the first point and the second point. The value and the temperature at the time the pressure was acquired are used as training data.
2. The gas pressure estimating means uses the learned estimation model to obtain an estimated value of information regarding the pressure at the second point from the gas pressure and temperature at the first point measured by the measuring device. The gas leak estimation system according to any one of 5 to 5. The determination means calculates an error rate E obtained by the following formula, where the estimated value of the pressure is Ps, and the actual measured value of the pressure during normal times at the second point is Pr,
E=100×(Ps-Pr)/Pr
The gas leak estimation system according to any one of claims 1 to 6, wherein it is determined that a gas leak has occurred when E exceeds a predetermined threshold value Rt. The gas leak estimation system according to claim 7, wherein the Rt is 0.015 or more. A gas leak estimation method in which a first point and a second point are located along a gas pipe, and a gas leak at the second point is estimated based on the gas pressure at the first point,
(A) measuring the pressure of the gas at a first point of the gas pipe;
The estimation model estimates information regarding the gas pressure at the second point from the gas pressure at the first point, and the model uses actual measured values of the gas pressure at the first point and the second point as training data during normal times. (B) calculating an estimated value of information regarding the gas pressure at the second point from the gas pressure at the first point measured by the measuring device using a trained estimation model subjected to machine learning; ,
(C) determining whether a gas leak has occurred at the second point based on the estimated value of the information regarding the pressure and the actual measured value of the pressure at the second point during normal times;
Gas leak estimation method with A gas leak estimation method according to claim 9, wherein a plurality of points including the first point and the second point are located in the gas pipe, and the first point and the second point are adjacent to each other. . The gas leak estimation method according to claim 10, wherein the first point is located upstream of the second point. The estimation model estimates information regarding the gas pressure at the second point from the gas pressure at the first point and the time and date when the pressure was acquired, and calculates the pressure at the first point and the second point. The actual measured values and the time and date when these were acquired are used as the training data, and
Any one of claims 9 to 11, wherein the step (B) calculates an estimated value of information regarding the pressure at the second point from the gas pressure at the first point measured by the measuring device, time, and date. Gas leak estimation method described in . The estimation model estimates information regarding the gas pressure at the second point from the gas pressure and flow rate at the first point, and calculates the actual measured values of the pressures at the first point and the second point, and the first point. is learned using the flow rate as the teacher data,
According to any one of claims 9 to 12, in the step (B), an estimated value of information regarding the pressure at the second point is obtained from the gas pressure and gas flow rate at the first point measured by the measuring device. Gas leak estimation method described. The estimation model estimates information regarding the gas pressure at the second point from the gas pressure at the first point and the temperature at the time when the pressure was acquired, and calculates the actual measurement of the pressure at the first point and the second point. The value and the temperature at the time the pressure was acquired are used as training data.
According to any one of claims 9 to 13, in the step (B), an estimated value of information regarding the pressure at the second point is obtained from the gas pressure and temperature at the first point measured by the measuring device. How to estimate gas leakage. In the step (C), the estimated value of the pressure is set as Ps, and the actual measured value of the pressure during normal times at the second point is set as Pr, and an error rate E obtained by the following formula is calculated,
E=100×(Ps-Pr)/Pr
The gas leak estimation method according to any one of claims 9 to 14, wherein it is determined that a gas leak has occurred when E exceeds a predetermined threshold value Rt. The gas leak estimation method according to claim 15, wherein the Rt is 0.015 or more.

Images