阅读说明：本技术 管道气体测量系统、方法和存储介质 (Pipeline gas measurement system, method and storage medium ) 是由 谢泽琼 何军飞 高富强 漆雅庆 肖振乾 陈秀峰 黄嘉瑜 郑桂兴 于 2020-04-23 设计创作，主要内容包括：本申请公开了一种管道气体测量系统、方法和存储介质,系统包括：传感器组件,其包括多个气体质量流量传感器和安装组件,多个所述气体质量流量传感器安装在所述安装组件上,所述安装组件被安装在管道内时使得多个所述气体质量流量传感器在所述管道的内径方向上分布；温度传感器,用于检测管道内气体的温度数据；数据处理模块,用于将多个所述气体质量流量传感器的检测数据以及所述温度传感器的检测数据输入到预测模型中进行预测,得到气体质量流量的检测结果。本申请通过测量管道内多点的气体质量流量,并通过预测模型进行预测,相对于现有技术更加准确。本申请可以广泛应用于气体测量技术。(The application discloses a pipeline gas measurement system, a method and a storage medium, wherein the system comprises: a sensor assembly comprising a plurality of gas mass flow sensors and a mounting assembly on which the plurality of gas mass flow sensors are mounted, the mounting assembly being mounted within a pipe such that the plurality of gas mass flow sensors are distributed in an inner diameter direction of the pipe; the temperature sensor is used for detecting the temperature data of the gas in the pipeline; and the data processing module is used for inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a prediction model for prediction to obtain a detection result of the gas mass flow. The gas mass flow of this application through measuring multiple spot in the pipeline to predict through predictive model, it is more accurate for prior art. The gas measuring device can be widely applied to gas measuring technology.)

1. A pipeline gas measurement system, comprising:

a sensor assembly comprising a plurality of gas mass flow sensors and a mounting assembly on which the plurality of gas mass flow sensors are mounted, the mounting assembly being mounted within a pipe such that the plurality of gas mass flow sensors are distributed in an inner diameter direction of the pipe;

the temperature sensor is used for detecting the temperature data of the gas in the pipeline;

and the data processing module is used for inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a prediction model for prediction to obtain a detection result of the gas mass flow.

2. The system of claim 1, wherein the plurality of gas mass flow sensors are symmetrically distributed about a center of a cross-section of the pipe when the mounting assembly is installed in the pipe.

3. The system for measuring the pipeline gas according to claim 1, wherein a plurality of prediction models corresponding to pipe diameters and user-configured pipe diameters are stored in the data processing module;

the step of inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a prediction model for prediction specifically comprises the following steps:

selecting a prediction model according to the pipe diameter configured by the user;

and inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a selected prediction model for prediction.

4. A method for measuring pipeline gas, comprising the steps of:

acquiring detection data acquired by a plurality of gas mass flow sensors;

acquiring temperature data of gas in a pipeline;

inputting the detection data and the temperature data acquired by the plurality of gas mass flow sensors into a prediction model for prediction to obtain a detection result of the gas mass flow;

wherein a plurality of the gas mass flow sensors are mounted on a mounting assembly which is mounted within the pipe such that the plurality of gas mass flow sensors are distributed in an inner diameter direction of the pipe.

5. The pipeline gas measurement method of claim 4, wherein the prediction model is obtained by:

acquiring a training sample and a label corresponding to the training sample;

and training a prediction model according to the training samples and the labels corresponding to the training samples until a preset condition is met.

6. The method of claim 5, wherein the obtaining of the training samples and the labels corresponding to the training samples comprises:

delivering gas with preset or random mass flow into the pipeline;

acquiring detection data of a temperature sensor in the pipeline and detection data of a plurality of gas mass flow sensors as training samples;

and taking the mass flow as a label of the training sample.

7. The method of claim 4, wherein the predictive model is trained based on a support vector machine and a particle swarm optimization algorithm.

8. The pipeline gas measurement method of claim 4, wherein the prediction model is obtained by:

calculating the average voltage value of two symmetrical gas mass flow sensors, wherein the average voltage value is represented as U_i＝(u_i+u_(2n-(i-1)))/2，i＝1,2,…,n；

The temperature data T and the voltage average value U are compared_iAs independent variable, the supply air flow Q_mAs a dependent variable, establishing a fitting model;

constructing a linear regression function in a high-dimensional space: y (x) ═ w Φ (x) + b; wherein y (x) represents the supply air flow rate Q_mj，x_jRepresenting an input vector [ U_1j,U_2j,…,U_nj,T_j]W represents the slope of the regression function, b represents the intercept of the regression function, and Φ () is a preset function;

introducing a slack variable ξ_j>0 and ξ_j ^*>0, the constraint conditions for obtaining the regression function are: y is_j-wΦ(x_j)-b≤+ξ_j，wΦ(x_j)+b-y_j≤+ξ_j ^*(j ═ 1,2, …, m), giving an optimal regression function ofC is a penalty factor, wherein C is more than 0 and is used for controlling the penalty degree of the sample with the error exceeding;

introducing a Lagrange function, and converting the optimal regression function into a dual form:the constraint condition of the optimal regression function isa and a are both support vectors with solution, K (x)_j,x_k) For support vector machine kernel, l is the number of training samples, K (x)_j,x_k)＝EXP(-|x-y|²/σ²) Wherein σ is²Is a nuclear parameter;

converting a regression function to

Selecting the mean error e of the training sample as a fitness function of the particle swarm algorithm, wherein the fitness function calculation formula of the particle swarm algorithm isWherein, y_j(j ═ 1,2, …, l) is the label of the jth sample, y_j ^*(j ═ 1,2, …, l) is the predicted value of the j-th sample;

training samples are normalized and linearly mapped to [0,1]]Within the range, the calculation formula of the linear mapping is x_i′＝(x_i-x_min)/(x_max-x_min)；

Establishing a particle X consisting of a penalty factor C and a variance σ_i＝(C_i,σ_i) Initializing the particles;

randomly generating initial positions and speeds of particles, and setting maximum iteration times;

substituting the initial particles into the established support vector machine model, taking the label as an output value, inputting a training sample into the support vector machine model for training, calculating a prediction error according to the fitness function, if the prediction error is smaller than a set value, obtaining the optimal particles, and if not, entering the next step;

the particles X_i＝(C_i,σ_i) Its speed and position are updated as follows:

v_ij(t+1)＝w(t)v_ij(t)+c₁r₁(p_ij(t)-x_ij(t))+c₂r₂(g_j(t)-x_ij(t))，x_ij(t+1)＝x_ij(t)+v_ij(t+1)；

wherein w is a weight; c. C₁And c₂Is an acceleration factor; r is₁And r₂Is distributed in [0,1]]The random number of (2); p is a radical of_ijIs the particle individual extremum; g_jIs the extreme value of the whole population; x is the number of_ij(t) and v_ij(t) position and velocity at time t, respectively;

substituting the updated particles into a support vector machine model to recalculate the prediction error, judging whether the new prediction error is smaller than a preset value or not, and simultaneously judging whether the maximum iteration number is reached or not; if the maximum iteration times are reached or the new prediction error is smaller than a preset value, obtaining the optimal particles; if the maximum iteration times are not reached or the new prediction error is larger than or equal to a preset value, returning to the particle X_i＝(C_i,σ_i) Updating the speed and position thereof according to the following formula;

and substituting the optimal particles into the support vector machine model to obtain a prediction model.

9. A pipeline gas measurement system comprising a program, a memory and a processor;

the program is stored in a processor;

the program, when loaded by a processor, performs the method of any of claims 4-8.

10. A storage medium storing a program, characterized in that the program, when executed by a processor, implements the method according to any one of claims 4-8.

Technical Field

The present application relates to gas measurement technologies, and in particular, to a pipeline gas measurement system, method, and storage medium.

Background

With the rapid development of the industry, the demand of various gas used by the industry is increased day by day, and the diameter of a pipeline required by gas supply is gradually increased. However, the existing plug-in airflow measurement method usually derives the airflow of the whole pipeline by measuring the flow of a point in the middle of the pipeline. In fact, the conventional measurement method has low measurement accuracy for large pipe diameter.

Disclosure of Invention

To solve at least one of the above technical problems, the present application aims to: a pipeline gas measurement system, method, and storage medium are provided.

In a first aspect, an embodiment of the present application provides:

a pipeline gas measurement system comprising:

a sensor assembly comprising a plurality of gas mass flow sensors mounted on a mounting assembly that when mounted within a pipe causes the plurality of gas mass flow sensors to be distributed in an inner diameter direction of the pipe;

the temperature sensor is used for detecting the temperature data of the gas in the pipeline;

and the data processing module is used for inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a prediction model for prediction to obtain a detection result of the gas mass flow.

Further, when the mounting assembly is mounted in a pipeline, the plurality of gas mass flow sensors are symmetrically distributed with the center of the cross section of the pipeline as the center.

Further, a plurality of prediction models corresponding to pipe diameters and pipe diameters configured by a user are stored in the data processing module;

the step of inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a prediction model for prediction specifically comprises the following steps:

selecting a prediction model according to the pipe diameter configured by the user;

and inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a selected prediction model for prediction.

In a second aspect, embodiments of the present application provide:

a method of pipeline gas measurement comprising the steps of:

acquiring detection data acquired by a plurality of gas mass flow sensors;

acquiring temperature data of gas in a pipeline;

inputting the detection data and the temperature data acquired by the plurality of gas mass flow sensors into a prediction model for prediction to obtain a detection result of the gas mass flow;

wherein a plurality of the gas mass flow sensors are mounted on a mounting assembly which is mounted within the pipe such that the plurality of gas mass flow sensors are distributed in an inner diameter direction of the pipe.

Further, the prediction model is obtained by the following steps:

acquiring a training sample and a label corresponding to the training sample;

and training a prediction model according to the training samples and the labels corresponding to the training samples until a preset condition is met.

Further, the obtaining of the training samples and the labels corresponding to the training samples includes:

delivering gas with preset or random mass flow into the pipeline;

acquiring detection data of a temperature sensor in the pipeline and detection data of a plurality of gas mass flow sensors as training samples;

and taking the mass flow as a label of the training sample.

Further, the prediction model is obtained by the following steps:

calculating the average voltage value of two symmetrical gas mass flow sensors, wherein the average voltage value is represented as U_i＝(u_i+u_(2n-(i-1)))/2，i＝1,2,…,n；

The temperature data T and the voltage average value U are compared_iAs independent variable, the supply air flow Q_mAs a dependent variable, establishing a fitting model;

constructing a linear regression function in a high-dimensional space: y (x) ═ w Φ (x) + b; wherein y (x) represents the supply air flow rate Q_mj，x_jRepresenting an input vector [ U_1j,U_2j,…,U_nj,T_j]W represents the slope of the regression function, b represents the intercept of the regression function;

introducing a slack variable ξ_j>0 and ξ_j ^*>0, the constraint conditions for obtaining the regression function are: y is_j-wΦ(x_j)-b≤+ξ_j，wΦ(x_j)+b-y_j≤+ξ_j ^*(j ═ 1,2, …, m), giving an optimal regression function ofC is a penalty factor, wherein C is more than 0 and is used for controlling the penalty degree of the sample with the error exceeding;

introducing a Lagrange function, and converting the optimal regression function into a dual form:the constraint condition of the optimal regression function isa and a are both support vectors with solution, K (x)_j,x_k) For support vector machine kernel, l is the number of training samples, K (x)_j,x_k)＝EXP(-|x-y|²/σ²) Wherein σ is²Is a nuclear parameter;

converting a regression function to

Selecting the mean error e of the training sample as a fitness function of the particle swarm algorithm, wherein the fitness function calculation formula of the particle swarm algorithm isWherein, y_j(j ═ 1,2, …, l) is the label of the jth sample, y_j ^*(j ═ 1,2, …, l) is the predicted value of the j-th sample;

training samples are normalized and linearly mapped to [0,1]]Within the range, the calculation formula of the linear mapping is x_i’＝(x_i-x_min)/(x_max-x_min)；

Establishing a particle X consisting of a penalty factor C and a variance σ_i＝(C_i,σ_i) Initializing the particles;

randomly generating initial positions and speeds of particles, and setting maximum iteration times;

substituting the initial particles into the established support vector machine model, taking the label as an output value, inputting a training sample into the support vector machine model for training, calculating a prediction error according to the fitness function, if the prediction error is smaller than a set value, obtaining the optimal particles, and if not, entering the next step;

the particles X_i＝(C_i,σ_i) Its speed and position are updated as follows:

v_ij(t+1)＝w(t)v_ij(t)+c₁r₁(p_ij(t)-x_ij(t))+c₂r₂(g_j(t)-x_ij(t))，x_ij(t+1)＝x_ij(t)+v_ij(t+1)；

wherein w is a weight; c. C₁And c₂Is an acceleration factor; r is₁And r₂Is distributed in [0,1]]The random number of (2); p is a radical of_ijIs the particle individual extremum; g_jIs the extreme value of the whole population; x is the number of_ij(t) and v_ij(t) position and velocity at time t, respectively;

substituting the updated particles into a support vector machine model to recalculate the prediction error, judging whether the new prediction error is smaller than a preset value or not, and simultaneously judging whether the maximum iteration number is reached or not; if the maximum iteration times are reached or the new prediction error is smaller than a preset value, obtaining the optimal particles; if the maximum iteration times are not reached or the new prediction error is larger than or equal to a preset value, returning to the particle X_i＝(C_i,σ_i) Updating the speed and position thereof according to the following formula;

and substituting the optimal particles into the support vector machine model to obtain a prediction model.

In a third aspect, embodiments of the present application provide:

a pipeline gas measurement system comprising a program, a memory, and a processor;

the program is stored in a processor;

the program, when loaded by a processor, performs the pipeline gas measurement method.

In a fourth aspect, embodiments of the present application provide:

a storage medium storing a program which, when executed by a processor, implements the pipeline gas measurement method.

The beneficial effects of the embodiment of the application are that: this application has set up sensor assembly, and this sensor assembly includes a plurality of gas mass flow sensor and installation component, a plurality of gas mass flow sensor are installed installation component is last, make when installation component is installed in the pipeline a plurality ofly gas mass flow sensor is in distribute in the internal diameter direction of pipeline, this application through gather the detection data that distributes in the internal diameter direction of pipeline carries out the flow prediction, and for prior art, this application considers the factor that gas mass flow exists uneven distribution in the pipeline at the prediction in-process, and the prediction result is more accurate for prior art.

Drawings

FIG. 1 is a block diagram of a pipeline gas measurement system provided in accordance with an embodiment of the present application;

FIG. 2 is a front view of a sensor assembly mounting structure provided in accordance with an embodiment of the present application;

FIG. 3 is a side view of a sensor assembly mounting structure provided in accordance with an embodiment of the present application;

fig. 4 is a flowchart of a method for measuring pipeline gas according to an embodiment of the present application.

Detailed Description

The present application is further described with reference to the following figures and specific examples. The described embodiments should not be considered as limiting the present application, and all other embodiments obtained by a person skilled in the art without making any inventive step are within the scope of protection of the present application.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the present application only and is not intended to be limiting of the application.

First, the terms appearing in the present application are explained:

a support vector machine: the (SVM) is a generalized linear classifier (generalized linear classifier) that binary classifies data according to a supervised learning (superiattice) method, and a decision boundary of the SVM is a maximum-margin hyperplane (maximum-margin hyperplane) that solves for a learning sample.

Mass flow rate: refers to the mass of fluid per unit time that the fluid passes through the effective cross-section of a closed pipe or open channel.

Radial basis function: radial Basis Function, abbreviated as RBF, is a radially symmetric scalar Function.

Referring to fig. 1, the present embodiment discloses a pipeline gas measurement system, including:

a sensor assembly comprising a plurality of gas mass flow sensors mounted on a mounting assembly that when mounted within a pipe causes the plurality of gas mass flow sensors to be distributed in an inner diameter direction of the pipe;

the temperature sensor is used for detecting the temperature data of the gas in the pipeline;

and the data processing module is used for inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a prediction model for prediction to obtain a detection result of the gas mass flow.

Referring to fig. 2 and 3, the mounting assembly 210 is installed in the duct 220, and the gas mass flow sensors 230 are symmetrically distributed around the midpoint of the inner portion of the duct of the mounting assembly 210, which can be understood as that the gas mass flow sensors 230 are symmetrically distributed in the length direction of the mounting assembly 210 around the center of the cross section of the duct 220. The mounting assembly 210 is a bar, and the mounting position thereof overlaps the inner diameter of the pipe 220. It should be understood that the duct 220 is a duct having a circular cross-section, and the length direction of the mounting assembly 210 is the same as the diameter direction of a cross-section of the duct 220. On the other hand, the pipe 220 is installed parallel to the ground, and the installation assembly 210 is installed perpendicular to the pipe, that is, the installation assembly 210 is in a perpendicular relationship to the ground. In addition, the installation position of the temperature sensor 240 is also shown in fig. 3. Of course, the mounting assembly 210 is not necessarily mounted perpendicular to the ground, but the mounting angle of the mounting assembly 210 to the ground is consistent with the mounting angle of the mounting assembly when the model training sample is collected, and the theoretically predicted result is the best.

The inventors have found that the velocity of the gas in the conduit is greatest at the center and smallest at the edges of the conduit, and that the distribution of the gas velocity in the conduit is difficult to describe as a function of the center point, wherein the mass flow rate is positively correlated to the gas velocity. Therefore, by arranging a plurality of sensors for measurement and inputting measurement data into a model for prediction, a mass flow rate detection result with higher accuracy can be obtained.

The output of the gas mass flow sensor can be a voltage signal, and the voltage signal output by the gas mass flow sensor is positively correlated with the mass flow of the gas.

The output of the temperature sensor can also be a voltage signal, and the magnitude of the voltage signal is positively correlated with the temperature. Of course, the sensor may directly output a digital signal indicating the corresponding detection value.

The voltage signal processing can be realized by arranging an analog-to-digital conversion circuit.

In the present embodiment, a pre-trained prediction model is stored in the data processing module, and the input of the prediction model is the detection data of the plurality of sensors, and the output thereof is the detection data.

In some embodiments, the mounting assembly is mounted within a pipe, and the plurality of gas mass flow sensors are symmetrically distributed about a cross-sectional center of the pipe. The number of gas mass flow sensors is determined according to the diameter of the gas supply pipeline, and the arrangement position is determined according to a linear logarithm method. Namely, the gas supply pipeline is divided into n circular rings with equal areas during measurement, the intersection point of the circular rings and the diameter is the installation position of the gas mass flow sensor, and the number of corresponding measuring points is 2 n.

In some embodiments, the data processing module stores a plurality of prediction models corresponding to pipe diameters and user-configured pipe diameters;

the step of inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a prediction model for prediction specifically comprises the following steps:

selecting a prediction model according to the pipe diameter configured by the user;

and inputting the detection data of the plurality of gas mass flow sensors and the detection data of the temperature sensors into a selected prediction model for prediction.

In some embodiments, the prediction model may be trained for a certain pipe diameter, and thus for different pipe diameters, there are different prediction models. Different pipe diameters, the number of the sensors and the installation spacing have certain differences. The model is confirmed based on pipe diameter parameter setting in this embodiment, can let constructor through configuration pipe diameter quick match to the model that corresponds for software can be general in the equipment of different pipe diameters.

Referring to fig. 4, the embodiment discloses a method for measuring pipeline gas, which includes the following steps:

step 410, acquiring detection data acquired by a plurality of gas mass flow sensors;

step 420, acquiring temperature data of gas in the pipeline;

step 430, inputting the detection data and the temperature data acquired by the plurality of gas mass flow sensors into a prediction model for prediction to obtain a detection result of the gas mass flow;

wherein a plurality of the gas mass flow sensors are mounted on a mounting assembly which is mounted within the pipe such that the plurality of gas mass flow sensors are distributed in an inner diameter direction of the pipe. The gas mass flow sensor may be mounted in a manner as described with reference to fig. 2 and 3 and corresponding system embodiments.

Wherein the prediction model is obtained by the following steps:

(1) establishing a fitting model of a support vector machine

Calculating the average voltage value of two symmetrical gas mass flow sensors, wherein the average voltage value is represented as U_i＝(u_i+u_(2n-(i-1))) 2, i ═ 1,2, …, n; the value measured by the temperature sensor is the gas supply temperature T, and the unit is; in the same pipe diameter and the same air supply flow Q_mAnd U corresponding to the supply air temperature T_i(U₁,U₂,…,U_n) Is a group of invariantsNumerical values.

The collection frequency and time of the flow sensor and the temperature sensor are consistent, and the temperature T and the temperature U of the supplied air are equal_iAs independent variable, the supply air flow Q_mAs a dependent variable, a fitting model was established.

Constructing a linear regression function in a high-dimensional space: y (x) ═ w Φ (x) + b, where y (x) represents the supply air flow rate Q_mj，x_j(i.e., x in the formula) represents the input vector [ U [_1j,U_2j,…,U_nj,T_j]W represents the slope of the regression function, b represents the intercept of the regression function, and Φ () is a preset function, where Φ (x) is x in some embodiments.

Introducing a slack variable ξ_j>0 and ξ_j ^*>0, the constraints of the available regression function are: y is_j-wΦ(x_j)-b≤+ξ_j，wΦ(x_j)+b-y_j≤+ξ_j ^*(j ═ 1,2, …, m), giving an optimal regression function ofAnd C is used for controlling the punishment degree of the sample with the error exceeding.

Introducing a Lagrange function, and converting the optimal regression function into a dual form:the constraint condition of the optimal regression function isa and a are both support vectors with solution, K (x)_j,x_k) For support vector machine kernel, l is the number of training samples.

Choosing RBF as the kernel function of the support vector machine, then K (x)_j,x_k)＝EXP(-|x-y|²/σ²) Wherein σ is²For the nuclear parameters, the regression equation of the supply air flow can be converted intoThe optimal solution to solve the equation can thus be translated into a problem of finding the optimal penalty factor C and the functional variance σ.

(2) Particle swarm optimization support vector machine

Selecting the mean error e of the training sample as a fitness function of the particle swarm algorithm, wherein the fitness function calculation formula of the particle swarm algorithm isWherein l is the number of samples, y_j(j ═ 1,2, …, l) is the label of the jth sample, y_j ^*(j ═ 1,2, …, l) is the predicted value for the j-th sample.

Training samples are normalized and linearly mapped to [0,1]]Within the range, the calculation formula of the linear mapping is x_i’＝(x_i-x_min)/(x_max-x_min)。

Establishing a particle X consisting of a penalty factor C and a variance σ_i＝(C_i,σ_i) Initializing the particles; the initial position and velocity of the particle are randomly generated in the search space, and the maximum number of iterations is set.

Substituting the initial particles into the established support vector machine model, taking the label as an output value, inputting a training sample into the support vector machine model for training, calculating a prediction error according to the fitness function, obtaining the optimal particles if the prediction error is smaller than a set value, and otherwise, entering the next step.

The particles X_i＝(C_i,σ_i) Its speed and position are updated as follows:

v_ij(t+1)＝w(t)v_ij(t)+c₁r₁(p_ij(t)-x_ij(t))+c₂r₂(g_j(t)-x_ij(t))，x_ij(t+1)＝x_ij(t)+v_ij(t+1)；

wherein w is a weight; c. C₁And c₂Is an acceleration factor; r is₁And r₂Is distributed in [0,1]]The random number of (2); p is a radical of_ijIs the particle individual extremum;g_jis the extreme value of the whole population; x is the number of_ij(t) and v_ij(t) is the position and velocity, respectively, at time t.

Substituting the updated particles into a support vector machine model to recalculate the prediction error, judging whether the new prediction error is smaller than a preset value or not, and simultaneously judging whether the maximum iteration number is reached or not; if the maximum iteration times are reached or the new prediction error is smaller than a preset value, obtaining the optimal particles; if the maximum iteration times are not reached or the new prediction error is larger than or equal to a preset value, returning to the particle X_i＝(C_i,σ_i) And updating the speed and the position according to the following formula.

And substituting the optimal particles into the support vector machine model to obtain a prediction model.

By adopting the particle swarm optimization algorithm, the optimal particles can be quickly obtained, so that the obtaining time of the prediction model is reduced. The embodiment is based on the particle swarm optimization support vector machine, and can obtain a more accurate prediction model relative to other neural networks under the condition of the same training resources.

In some embodiments, the data input to the predictive model is normalized, i.e., the input values are mapped into the [0,1] range. Therefore, in these embodiments, the training and practical application processes both include a step of normalizing the data.

In some embodiments, training samples may also be collected based on the above method, and neural networks such as CNN and RNN may be trained to derive the predictive model.

When the neural network is adopted for training, parameter adjustment is carried out on the neural network based on the error between the prediction result of the training sample and the corresponding label until the neural network meets the preset condition. Specifically, until the neural network converges in the verification set or the number of training times reaches the maximum number.

In this embodiment, the obtaining of the training samples and the labels corresponding to the training samples includes:

delivering gas with preset or random mass flow into the pipeline;

acquiring detection data of a temperature sensor in the pipeline and detection data of a plurality of gas mass flow sensors as training samples;

and taking the mass flow as a label of the training sample.

The present embodiment uses a known mass flow rate of gas delivered to the pipeline, then uses the sensor to obtain the detection data, uses the detection data of the sensor as a training sample, and uses the known mass flow rate as a training label. The present embodiment can generate gas with different mass flow and temperature by configuring a standard gas generator, and obtain multiple training samples by the above-mentioned method.

In addition, when data are collected, the data collected by the two symmetrical sensors need to be subjected to validity check, namely, the difference between the voltage signal values measured by the two symmetrical sensors should satisfy the following condition: | u_i-u_jAnd | ≦ A, i, j ═ 1,2. Wherein, the measurement precision of the sensor is shown, and A is a correction coefficient. The value of the gas supply pipeline A which is ideal and has no interference is 1. If the data measured by the sensor does not pass the condition, the 3 data processing modules reject the set of measured data.

The embodiment discloses a pipeline gas measurement system, which comprises a program, a memory and a processor;

the program is stored in a processor;

the program, when loaded by a processor, performs the pipeline gas measurement method.

The present embodiment discloses a storage medium storing a program which, when executed by a processor, implements the pipeline gas measurement method.

The step numbers in the above method embodiments are set for convenience of illustration only, the order between the steps is not limited at all, and the execution order of each step in the embodiments can be adapted according to the understanding of those skilled in the art.

While the present application has been described with reference to the preferred embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.