CN118627381A - Construction method of natural gas pipeline network energy transmission difference calculation system integrated with machine learning - Google Patents

Abstract

The present invention proposes a method for constructing a natural gas pipeline network energy transmission difference calculation system that integrates machine learning. The energy transmission difference calculation system includes a source database unit, a prediction unit, and an optimization unit; the source database unit includes a sensor data source module and other data modules; the prediction unit includes a sufficient sample prediction module and a small sample prediction module; the optimization unit includes a measured data module and a Bayesian parameter update module. It mainly includes the following steps: determining the factors affecting the energy transmission difference to construct a basic database; clustering processing to establish a machine learning regression prediction model for the energy transmission difference of the natural gas pipeline network; real-time input of real-time sensor data of the required prediction section for energy transmission difference prediction; constructing an optimization database and updating the weights of the input parameters in the machine learning model. The present invention avoids the serious dependence of traditional methods on manual experience, and the calculation process is simple and time-consuming; it solves the small sample problem in the field of pipeline transportation and improves the generalization performance of the model.

Description

Construction method of natural gas pipeline network energy transmission difference calculation system integrated with machine learning

Technical Field

The present invention relates to the field of natural gas transportation, and in particular to a method for constructing a natural gas pipeline network energy transmission difference calculation system integrating machine learning.

Background Art

How to ensure the comprehensive coverage of natural gas resources in my country's urban areas is an important factor restricting the development of the natural gas industry. Among them, pipeline transportation has become one of the main long-distance transportation methods for natural gas and other energy sources, thanks to its significant advantages in safety, transportation costs, construction cycle and transportation volume. However, due to many uncertain factors such as metering devices, personnel measurement errors, inherent losses in aging pipelines, gas quality, gas flow status, temperature, etc., long-distance pipeline transportation of natural gas will inevitably have transmission differences. This has a significant impact on the accurate calculation of the natural gas demand side, the formulation of pipeline network gas supply plans, and the economic benefits of natural gas companies.

At present, the internationally accepted method for measuring natural gas is energy measurement units. Within the selected measurement period, the energy transmission difference E of natural gas can be calculated using formula (1). The commonly used methods for calculating the energy transmission difference of natural gas pipeline networks mainly include: traditional SCADA system, manual calculation and scheduling experience mode. However, due to the diversity and uncertainty of the above-mentioned influencing factors, the above-mentioned energy transmission difference calculation methods often have problems such as low calculation efficiency and low cost performance.

E＝(H _C ) _G (t ₁ )Q (1)

Q＝(V ₁ +Q ₁ )-(Q ₂ +Q ₃ +Q ₄ +V ₂ ) (2)

Wherein: E is the energy transmission difference, ( _HC ) _G ( _t1 ) is the actual high molar calorific value of the gas fuel, kJ/kmol, Q is the equilibrium gas transmission volume difference in the gas pipeline at a certain time, ^m3 ; _Q1 is the input gas volume value in the same time, ^m3 ; _Q2 is the output gas volume value in the same time, ^m3 ; _Q3 is the production and living gas consumption of the gas transmission unit in the same time, ^m3 ; _Q4 is the gas discharge volume in the same time, ^m3 ; _V1 is the gas storage volume in the pipeline calculation section at the beginning of the calculation time, ^m3 ; _V2 is the gas storage volume in the pipeline calculation section at the end of the calculation time, ^m3 .

For example, the invention patent with application number CN202010958923.3 discloses a visual natural gas pipeline network transmission difference optimization solution system that comprehensively considers the actual pipeline storage and transmission difference problems, significantly reducing the calculation error of energy transmission difference. However, due to the solution of multivariable differential equations, the solution process is cumbersome and time-consuming. At the same time, the calculation applicability of energy transmission difference in different sections or different service environments is poor.

Data-driven models with high-dimensional nonlinear modeling capabilities and flexible data processing capabilities have been widely used in various practical engineering regression prediction and classification problems in recent years. In the field of natural gas transportation, the rapid iteration of advanced sensor equipment provides a large amount of accurate data, and the emergence of supercomputers solves the problem of insufficient computing power. Based on this, artificial intelligence calculation methods provide a feasible solution for the rapid and accurate prediction of energy transmission differences in natural gas pipeline transportation.

Therefore, based on the monitoring data of existing sensor equipment, it is urgent to develop a natural gas pipeline network energy transmission difference calculation method with wide applicability, short time consumption and high accuracy.

Summary of the invention

The present invention proposes a method for constructing a natural gas pipeline network energy transmission difference calculation system integrating machine learning, which can realize intelligent regression prediction of energy transmission difference in any section and under any working conditions during natural gas pipeline transportation.

The technical solution of the present invention is implemented as follows: a method for constructing a natural gas pipeline network energy transmission difference calculation system integrating machine learning, wherein the energy transmission difference calculation system includes three units: a source database unit, a prediction unit, and an optimization unit;

The source database unit includes a sensor data source module and other data modules;

The prediction unit includes a sufficient sample prediction module and a small sample prediction module;

The optimization unit includes a measured data module and a Bayesian parameter updating module;

The construction process of the natural gas pipeline network energy transmission difference calculation system integrating machine learning mainly includes the following steps:

Step 1: Determine the factors affecting energy transmission difference. The source database unit collects the historical data of energy transmission difference of natural gas pipeline networks in various engineering cases and builds a basic database.

Step 2: The prediction unit first performs clustering processing on the basic data, dividing it into small sample and sufficient sample sub-datasets, and then divides the two subsets into training sets and test sets respectively, and uses the training set to establish a machine learning regression prediction model for energy transmission difference of the natural gas pipeline network;

Step 3: Input the real-time sensor data and other working data of the pipeline section to be predicted, and use the model trained in step 2 to predict the energy transmission difference;

Step 4: The measured data module in the optimization unit collects the actual data during each energy transmission difference detection, and uses the error with the predicted result as the judgment basis, retains the real sample data with an error below 10% and builds an optimization database; based on the comparison between the samples in the optimization database and the predicted results, the weights of each particle in the particle filter are defined, and the weights of the input parameters in the machine learning model are updated to further reduce the error rate.

Preferably, the influencing factors in step 1 include internal gas pressure, pressure drop, gas flow, gas temperature, actual high molar calorific value, initial flow, terminal flow, pipe section length, and external gas pressure and temperature of the natural gas pipeline network section, and the above variables in the historical case are constructed as a feature set X={x11,x12,x13,…,xij}, where xij represents the jth feature of the i-th sample number;

The energy input difference in the historical data collected in the step 1 is used as the label set Y={Y1, Y2, Y3, ..., Yi} of the basic data set.

Preferably, in step 2, the basic data is divided into small sample data subsets and sufficient sample data subsets using K-means clustering method;

The machine learning model in step two includes a sufficient sample machine learning prediction model and a small sample machine learning prediction model, wherein the sufficient sample machine learning model adopts a neural network method, and the small sample machine learning model adopts a transfer learning method.

Preferably, the sufficient samples take the feature set in step 1 as input, and are assigned initial weight factors, and a corresponding neural network model is established using the training set;

A transfer learning method is used to train the small sample training set. The above transfer learning uses a feature transfer method, which is based on feature mapping and aligns data distribution in different domains.

Preferably, in the transfer learning, the physical field of the outgoing data is the source domain M, the physical field of the incoming data is the target domain N, and the feature set and label set in the source domain and the target domain are defined as Xm, Ym, Xn, and Yn, respectively; the transfer learning model optimizes the hyperparameters derived from data fusion and further combines clustering on the basis of classical feature migration.

Preferably, the key point of the feature migration is to construct a mapping φ so that the source domain and the target domain are close to the same conditional distribution. In the present invention, the maximum mean difference, i.e., Maximum Mean Discrepancy, abbreviated as MMD, is selected to evaluate the distribution distance between the source domain and the target domain, and its calculation expression is:

Where: n ₁ and n ₂ are the number of source domain and target domain data respectively, x _mi ∈X _m , x _nj ∈X _n ;

After feature mapping, the clustering method is introduced to construct the optimization index, and the random search method is further combined to optimize the regularization parameter.

As a preferred method, K-means clustering method is used when optimizing feature mapping hyperparameters, which is specifically used for the feature matrix after feature migration mapping; at this time, a cluster center will be generated for the source domain and the target domain respectively, and the distance between the two cluster centers will represent the distribution difference between the source feature and the target domain feature;

The distance between the two cluster centers can be defined by the following formula:

Distance(K)＝||c _m -c _n || ² (4)

Among them, _cm and _cn are the unique clustering centers of the source domain and target domain features after feature mapping.

Preferably, based on the introduction of the above-mentioned K clustering, the optimization problem of the hyperparameters is transformed into a problem of minimizing Distance(K), and a random search method is used to optimize the regularization parameters within a given range.

Preferably, when predicting the energy transmission difference of the unknown pipe section in step three, it is necessary to first determine whether the problem to be predicted belongs to a small sample problem or a sufficient sample problem based on the similarity analysis between the input features to be predicted and the features in the basic data set; after determining the classification, the feature set of the sample to be predicted is input, and the current energy transmission difference prediction value can be obtained using the corresponding machine learning model.

As a preferred method, some pipe sections are randomly inspected regularly, and the actual energy transmission difference is calculated based on the measured signal and other data; that is, new samples are obtained; at the same time, within the randomly inspected pipe section, samples with a relative error of less than 10% are retained, and samples with an error of more than 10% are eliminated;

By sampling new samples, particle filter parameters can be defined to further optimize the weight factors of feature engineering in machine learning models.

Compared with the prior art, the advantages of the present invention are:

(1) The present invention avoids the heavy reliance of traditional methods on manual experience; at the same time, compared with various analytical calculations, the calculation process of the present invention is simple and time-saving.

(2) The introduction of transfer learning enables the present invention to effectively solve the small sample problem in the field of pipeline transportation and improve the generalization performance of the model, making the method more widely applicable.

(3) The establishment of an optimization system can further reversely optimize the machine learning regression prediction model established based on existing data, and can update the model parameters in real time, thus overcoming the problem of low prediction accuracy of the machine learning model under complex working conditions.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of the structure of a natural gas pipeline energy transmission difference calculation system integrating machine learning;

Figure 2 is an application flow chart of the natural gas pipeline energy transmission difference calculation system integrating machine learning.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

The present invention provides a natural gas pipeline network energy transmission difference calculation system integrating machine learning.

The energy transfer difference calculation includes three units: source database unit, prediction unit and optimization unit. The database unit includes sensor data source module and other data modules; the prediction unit includes sufficient sample prediction module and small sample prediction module; the optimization unit includes measured data module and Bayesian parameter update module.

Furthermore, the present invention provides a construction process of a natural gas pipeline network energy transmission difference calculation system integrating machine learning, which mainly includes the following steps:

Step 1: Determine the factors affecting energy transmission difference. The sensor data source module and other data modules of the source database unit collect the historical data of energy transmission difference of the natural gas pipeline network in various engineering cases to build a basic database.

Step 2: Cluster the basic data into small sample and sufficient sample sub-datasets. Then divide the two subsets into training sets and test sets, and use the training set to establish a machine learning regression prediction model for energy transmission difference in natural gas pipeline networks.

Step 3: Input the real-time sensor data and other working data of the pipeline section to be predicted, and use the model trained in step 2 to predict the energy transmission difference.

Step 4: Collect the actual data of each energy transmission difference detection, use the error with the predicted result as the basis for judgment, retain the real sample data with an error below 10% and build an optimized database. Based on the comparison between the samples in the optimized database and the predicted results, define the weight of each particle in the particle filter, and update the weight of the input parameters in the machine learning model to further reduce the error rate.

Preferably, the influencing factors in step one include internal gas pressure, pressure drop, gas flow, gas temperature, actual high molar calorific value, initial flow, terminal flow, pipe section length, and external gas pressure, temperature, etc. of the natural gas pipeline network section, and the above variables in the historical cases are constructed as a feature set X = {x11, x12, x13, ..., xij}, where xij represents the jth feature of the i-th sample number.

The energy input difference in the historical data collected in step 1 is used as the label set Y = {Y1, Y2, Y3, ..., Yi} of the basic data set.

Preferably, in step 2, the basic data is divided into small sample data and sufficient sample data subsets using K-means clustering method.

The machine learning model in step 2 includes a sufficient sample machine learning prediction model (corresponding to a sufficient sample prediction module) and a small sample machine learning prediction model (corresponding to a small sample prediction module). The sufficient sample machine learning model adopts a neural network method, and the small sample machine learning model adopts a transfer learning method.

For the sufficient sample problem, the feature set described in step 1 is used as input, and the initial weight factor is assigned, and the corresponding neural network model is established using the training set.

For small sample data problems, the prediction accuracy of traditional machine learning methods is often low. However, transfer learning can achieve mutual migration and utilization of data in different domains, which can effectively solve the small sample problem. In addition, appropriate migration criteria enable transfer learning to reduce the differences in data distribution in different domains, which can solve the problem of multi-distribution of data. Therefore, the present invention uses transfer learning methods to train small sample training sets.

The above transfer learning uses the feature transfer method, which is based on feature mapping and can align data distribution in different domains. It has good applicability to small sample problems.

In the transfer learning, the physical field of the outgoing data is the source domain M, and the physical field of the incoming data is the target domain N. The feature set and label set in the source domain and the target domain are defined as Xm, Ym, Xn, and Yn, respectively.

Preferably, the above transfer learning model optimizes the hyperparameters derived from data fusion and further combines clustering based on the classic feature migration.

The key point of feature migration is to construct a mapping φ so that the source domain and the target domain are close to the same conditional distribution. In this invention, the maximum mean discrepancy (MMD) is used to evaluate the distribution distance between the source domain and the target domain, and its calculation expression is:

Where: n ₁ and n ₂ are the number of source domain and target domain data respectively,

Based on the above formula, the construction of the feature transfer map can be determined by minimizing the maximum mean difference. Similar to other machine learning methods, by introducing kernel functions and regularization terms, the problem of minimizing the maximum mean difference can be transformed into an optimization problem.

Preferably, in order to achieve a better feature migration effect, the regularization term must be optimized. Considering the differences in data distribution, the present invention also introduces a clustering method after feature mapping to construct an optimization index, and further combines the random search method to optimize the regularization parameter.

When optimizing feature mapping hyperparameters, K-means clustering is used, which is specifically used for the feature matrix after feature migration mapping. At this time, a cluster center will be generated for the source domain and the target domain respectively, and the distance between the two cluster centers will represent the distribution difference between the source and target domain features.

The distance between the two cluster centers in the previous step can be defined by the following formula:

Distance(K)＝||c _m -c _n || ² (4)

Among them, _cm and _cn are the unique clustering centers of the source domain and target domain features after feature mapping.

Based on the introduction of K clustering, the optimization problem of hyperparameters is transformed into the problem of minimizing Distance(K). The random search method can be used to optimize the regularization parameters within a given range.

Preferably, when predicting the energy transmission difference of the unknown pipe section in step three, it is necessary to first determine whether the problem to be predicted is a small sample problem or a sufficient sample problem based on the similarity analysis between the input features to be predicted and the features in the basic data set.

After determining the classification, input the feature set of the sample to be predicted, and use the corresponding machine learning model to obtain the current energy output difference prediction value.

Preferably, due to the importance of energy transmission difference in natural gas pipelines, it is inevitable to periodically sample some pipe sections and calculate the actual energy transmission difference based on the measured signal and other data. That is, to obtain new samples. At the same time, for the sampled pipe section, there will inevitably be a certain error between the predicted value and the true value. Samples with a relative error of less than 10% are retained, and samples with an error of more than 10% are eliminated.

Furthermore, the particle filter parameters can be defined by sampling new samples, thereby further optimizing the weight factors of feature engineering in the machine learning model.

Preferably, by repeating the above steps 1 to 4, the energy transmission difference of any section of the natural gas pipeline can be obtained.

Compared with the existing energy transfer difference calculation method, the beneficial effects of the present invention are:

(1) The present invention avoids the heavy reliance of traditional methods on manual experience. At the same time, compared with various analytical calculations, the present invention has a simple calculation process and takes less time.

(2) The introduction of transfer learning enables the present invention to effectively solve the small sample problem in the field of pipeline transportation and improve the generalization performance of the model, making the method more widely applicable.

(3) The establishment of an optimization system can further reversely optimize the machine learning regression prediction model established based on existing data, and can update the model parameters in real time, thus overcoming the problem of low prediction accuracy of the machine learning model under complex working conditions.

The concept and application process of the present invention are further described in detail below in conjunction with the embodiments of the drawings, so that those skilled in the art can implement them according to the description.

As shown in Figure 1, the natural gas pipeline energy transmission difference calculation system integrating machine learning includes a source database unit, a prediction unit, and an optimization unit.

FIG2 shows an implementation process of a natural gas pipeline energy transmission difference calculation system integrating machine learning. Combined with FIG1 and FIG2, the specific implementation process of the system includes the following steps:

S1: Determine the factors that affect the energy transmission difference based on manual experience and existing research. Other data modules of the source database unit collect historical structured and unstructured data sets containing energy transmission difference from various public engineering cases, published literature, natural gas company databases, etc. Take each influencing factor as a feature set and the energy transmission difference result as a label value to construct a natural gas pipeline network energy transmission difference source database.

The source database data is preprocessed to remove missing data and abnormal data. At the same time, due to the large number of influencing factors involved, the distribution ranges of different types of data vary greatly, and even have magnitude differences. Therefore, the source data set after screening is standardized and normalized.

Using K-means clustering method, the source data set is divided into small sample data set and sufficient sample data set.

For sufficient sample data sets and small sample data sets, the normalized features are constructed into feature matrices to facilitate the subsequent training of machine learning models.

S2: Based on the clustering division of the data set in S1, machine learning model training is performed on two types of data respectively.

For sufficient sample data sets, the training set and test set are divided into 3:7 ratios. Each influencing factor is used as input, the energy input difference is used as output, and the root mean square error is used as the loss function. The neural network weights and bias terms are initialized. The random search method is used to optimize the hyperparameters within a certain range.

The above-mentioned neural network hyperparameters include activation function, regularization coefficient, learning rate, maximum number of iterations, etc. The optimal hyperparameters are selected by giving the hyperparameter optimization interval according to the sample data volume, input volume, etc., and comprehensively considering the accuracy and stability.

For small sample problems, transfer learning is used to construct a regression prediction model. The small sample source data set is divided into the source domain and the target domain, and feature mapping is used to map the source domain data and the target domain data into two closer feature matrices.

In the above mapping process, K-means clustering is introduced to realize problem transformation, and the regularization hyperparameters of feature mapping are automatically optimized by combining the random search method with the goal of minimizing the cluster center distance.

For the mapped new source and target domain data sets, neural networks are also used as predictors, and the weights of each particle in the particle filter are defined to find the best. The root mean square error and determination coefficient are used as evaluation indicators to obtain the optimal transfer learning model. The measured sensor data and other non-sensor data of the pipe section are used as input to predict the energy output difference of the current pipe section.

It is worth noting that the use of transfer learning can improve the prediction accuracy of small sample conditions. At the same time, the use of transfer learning makes it possible to accurately predict the energy transmission difference of natural gas pipeline sections in different pipeline sections and different service environments.

S3: The sensor data source module collects regular sampling data sets from each pipeline, and uses the relative error with the predicted result as an evaluation table, retaining sampling samples with a relative error of less than 10%. Combined with the Bayesian parameter update module for optimization, the weights of each particle in the particle filter are defined, with the goal of minimizing the relative error, and the network node weights are changed in reverse transmission to update the machine learning model to achieve automatic update of the regression model.

By repeating the above steps, the energy transmission difference regression prediction of any pipe section can be realized.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. The construction method of the natural gas pipe network energy transmission difference calculation system integrating machine learning is characterized by comprising the following steps of: the energy input difference calculation system comprises three units: the system comprises a source database unit, a prediction unit and an optimization unit;

the source database unit comprises a sensor data source module and other data modules;

the prediction unit comprises a sufficient sample prediction module and a small sample prediction module;

the optimizing unit comprises an actual measurement data module and a Bayesian parameter updating module;

the construction process of the natural gas pipe network energy transmission difference calculation system integrating machine learning mainly comprises the following steps:

Step one: determining energy transmission difference influence factors, collecting energy transmission difference historical data of a natural gas pipe network in various engineering cases by a source database unit, and constructing a basic database;

step two: the prediction unit firstly performs clustering treatment on basic data division, divides the basic data into a small sample sub-data set and a sufficient sample sub-data set, then respectively divides the two sub-sets into a training set and a test set, and establishes a natural gas pipe network energy transmission difference machine learning regression prediction model by using the training set;

Step three: inputting real-time sensor data and other working data of a pipe section to be predicted in real time, and performing energy transmission difference prediction by using the model trained in the second step;

Step four: the actual data module in the optimizing unit collects actual data during each energy transmission difference detection, takes the error between the actual data and a predicted result as a judgment basis, and reserves real sample data with the error below 10% through the Bayesian parameter updating module and builds an optimizing database; and defining the weight of each particle in the particle filtering based on the comparison of the sample in the optimized database and the prediction result, and updating the weight of the input parameter in the machine learning model so that the error rate is further reduced.

2. The method for constructing the natural gas pipe network energy input difference computing system integrated with machine learning according to claim 1, wherein the method comprises the following steps of: the influencing factors in the first step include internal air pressure, pressure drop, air flow, air temperature, actual high-order molar heating value, initial flow, terminal flow, pipe section length, external air pressure and temperature of the pipe section of the natural gas pipe network, and the variables in the history case are constructed as a feature set X= { X ₁₁,x₁₂,x₁₃,…,x_ij},x_ij to represent the j feature of the i sample number;

The energy input difference in the history data collected in the step one is used as a tag set y= { Y1, Y2, Y3, …, yi } of the basic data set.

3. The method for constructing the natural gas pipe network energy input difference computing system integrated with machine learning according to claim 1, wherein the method comprises the following steps of: dividing the basic data into small samples and sufficient sample data subsets by adopting a K-means clustering method;

The machine learning model in the second step comprises a sufficient sample machine learning prediction model and a small sample machine learning prediction model, wherein the sufficient sample machine learning model adopts a neural network method, and the small sample machine learning model adopts a transfer learning method.

4. The method for constructing the natural gas pipe network energy input difference computing system integrated with machine learning according to claim 3, wherein the method comprises the following steps of: the sufficient sample takes the feature set in the first step as input, initialization weight factors are given, and a training set is utilized to establish a corresponding neural network model;

And training the small sample training set by adopting a transfer learning method, wherein the transfer learning adopts a characteristic transfer method, and the method is based on characteristic mapping and aligns data distribution in different domains.

5. The method for constructing the natural gas pipe network energy input difference computing system integrated with machine learning according to claim 4, wherein the method comprises the following steps of: in the transfer learning, a physical field of the migrated data is taken as a source domain M, a physical field of the migrated data is taken as a target domain N, and a feature set and a tag set in the source domain and the target domain are respectively defined as Xm, ym, xn, yn; the migration learning model is based on classical feature migration, and is used for optimizing the super-parameters of data fusion and further combined clustering.

6. The method for constructing the natural gas pipe network energy input difference computing system integrated with machine learning according to claim 5, wherein the method comprises the following steps of: the key point of the feature migration is to construct a mapping phi so that a source domain and a target domain are close to the same-condition distribution, and the Maximum mean difference, namely Maximum MEAN DISCREPANCY, abbreviated as MMD, is selected to evaluate the distribution distance of the source domain and the target domain, wherein the calculation expression is as follows:

wherein: n ₁、n₂ is the source domain and destination domain data quantity respectively,

After the feature mapping, a clustering method is introduced for constructing an optimization index, and the regularization parameter is optimized by further combining a random search method.

7. The method for constructing the natural gas pipe network energy input difference computing system integrated with machine learning according to claim 6, wherein the method comprises the following steps: when the feature mapping super-parameters are optimized, a K-means clustering method is selected, and the K-means clustering method is specifically used for feature matrixes after feature migration mapping; at this time, the source domain and the target domain will generate a cluster center respectively, and the distance between the two cluster centers will represent the distribution difference of the features from the features of the target domain;

The distance between the two cluster centers can be defined by the following formula:

Distance(K)＝||c_m-c_n||² (4)

Wherein c _m and c _n are respectively unique cluster centers of the source domain and the target domain after feature mapping.

8. The method for constructing the natural gas pipe network energy input difference computing system integrated with machine learning according to claim 7, wherein the method comprises the following steps of: based on the introduction of K clustering, the problem of optimizing the super parameters is converted into the problem of minimizing Distance (K), and regularized parameter optimizing is performed in a given range by using a random search method.

9. The method for constructing the natural gas pipe network energy input difference computing system integrated with machine learning according to claim 1, wherein the method comprises the following steps of: when the energy transmission difference of the unknown pipe section is predicted in the third step, judging whether the problem to be predicted belongs to a small sample or a sufficient sample according to similarity analysis of the input characteristic to be predicted and the characteristic in the basic data set; after the classification is determined, the feature set of the sample to be predicted is input, and the current energy input difference prediction value can be obtained by using a corresponding machine learning model.

10. The method for constructing the natural gas pipe network energy input difference computing system integrated with machine learning according to claim 1, wherein the method comprises the following steps of: the actual measurement data module periodically performs selective inspection on part of the pipe section, and calculates actual energy transmission difference based on the actual measurement signals and other data; obtaining a new sample; meanwhile, in the sampling tube section, samples with the relative error less than 10% are reserved, and samples with the error more than 10% are removed;

Particle filter parameters can be defined by using the sampling new samples, so that the weighting factors of the feature engineering in the machine learning model are further optimized in a direction through a Bayesian parameter updating module.