CN112819035B

CN112819035B - Method and device for judging gas channeling by utilizing PVT (physical vapor transport) experiment and machine learning

Info

Publication number: CN112819035B
Application number: CN202110036346.7A
Authority: CN
Inventors: 朱维耀; 孔德彬; 黄堃; 夏静; 岳明; 李保柱; 宋智勇
Original assignee: University of Science and Technology Beijing USTB
Current assignee: University of Science and Technology Beijing USTB
Priority date: 2021-01-12
Filing date: 2021-01-12
Publication date: 2021-11-09
Anticipated expiration: 2041-01-12
Also published as: CN112819035A

Abstract

The invention provides a method and a device for judging gas channeling by utilizing PVT (physical vapor transport) experiments and machine learning, wherein the method comprises the following steps: obtaining characteristic parameters of the formation fluid sample under different injection volumes and different pressures and corresponding gas channeling judgment results through a PVT experiment to obtain PVT experiment data; acquiring characteristic parameters monitored from the well mouth of each well and corresponding gas channeling judgment results to obtain monitoring historical data; combining PVT experimental data and monitoring historical data to form an initial training sample, and training the initial training sample by a machine learning method to obtain a prediction model; and inputting new characteristic parameters monitored by the wellhead into the prediction model to obtain a corresponding gas channeling judgment result. According to the invention, accurate gas channeling judgment result can be obtained by utilizing wellhead monitoring data, and the prediction cost is reduced.

Description

Method and device for judging gas channeling by utilizing PVT (physical vapor transport) experiment and machine learning

Technical Field

The invention relates to the technical field of condensate gas reservoir gas injection development, in particular to a method and a device for judging gas channeling by utilizing PVT (physical vapor transport) experiments and machine learning.

Background

The summary of condensate gas reservoir gas injection development is referred to as a condensate gas reservoir gas injection development gas channeling quantitative evaluation method and application, an oil drilling and production process, 11 months in 2017 (volume 39), 6 th stage, and P667-P672. Condensate gas reservoirs are a class of special, complex gas reservoirs that are developed differently from oil reservoirs as well as dry gas reservoirs. Under the original condition of a stratum, condensate gas reservoir fluid exists in a single-phase gas mode, the pressure of the gas reservoir is continuously reduced along with continuous extraction of the fluid in the development process, when the formation pressure is lower than the dew point pressure, heavy hydrocarbon substances dissolved in a gas phase generate a reverse condensation phenomenon, liquid hydrocarbon begins to be separated out, and the liquid hydrocarbon is attached to the surface of a pore space of a reservoir in a liquid film mode. As the pressure continues to drop, the oil saturation in the reservoir pores increases until a maximum oil saturation is reached. However, this saturation tends to be below the critical flow saturation, thereby causing liquid hydrocarbons (condensate) of higher economic value to be lost in the formation. The recovery ratio of the condensate with higher condensate content is only 20 percent in the exhaustion type exploitation of the condensate. In order to improve the recovery ratio of the condensate gas reservoir, especially for the condensate gas reservoir with higher condensate oil content, a pressure maintaining development mode is mostly adopted, and most of injected media are dry gas. The circulating gas injection can keep the formation pressure, prevent the further precipitation of the reverse condensate, improve the seepage capability and improve the recovery ratio of the condensate. The mechanism is mainly the reverse evaporation and gas drive.

The gas channeling is a common phenomenon in the middle and later stages of condensate gas reservoir circulating gas injection development, and in the gas injection development process, due to the heterogeneity of stratum porous media and the difference of physical properties on a reservoir stratum plane and in the longitudinal direction, along with the development and the increase of injection pore volume multiple (PV), the front edge part of injected gas forms an advantageous flow channel along a high-permeability layer in the stratum, so that the gas channeling breaks through to the bottom of a gas production well. After the injected gas breaks through, the injected gas is continuously accumulated along with the injected volume, so that the displacement efficiency of the injected gas on the original formation fluid is greatly reduced.

At present, the most accurate mode for judging gas channeling is underground sampling, a sampler loads underground fluid into a sampling bin under the condition of underground temperature and pressure, and then fluid component analysis is carried out on the ground, so that accurate component distribution can be obtained, and thus, the gas channeling is judged. The sampling technology has high requirements on sampling equipment, the technology is mainly mastered in foreign oil and gas service companies, and the cost of single sampling is high. The onsite can roughly estimate the gas channeling (also called the overshoot) through the pressure change and the gas-oil ratio change, but the data changes greatly, are low in correlation and accuracy due to the fact that the onsite working system changes frequently.

In addition, tens of wells are few in a development block, and hundreds of wells are many in the development block, and the wells can generate a large amount of data in the daily development process. The data of a single well is usually only used for judging the gas channeling of the well, and other wells have no guidance function, so that the waste of data resources is caused. At present, the data for judging gas channeling has two sources, one is the fluid analysis data of underground sampling, and the other is the well head fluid analysis data. Since the condensate gas varies with the pressure, the content of the internal components changes, and a part of the condensate gas changes between a gas phase and a liquid phase. Therefore, the two types of data are different, the underground sampling result is the most accurate, but the acquisition difficulty is high, most of the data depend on foreign technologies, and well-head data are acquired well, so that the cost is lower. Wellhead data is usually judged by adopting a gas-oil ratio chart method, but the judgment reliability is poor because well switching operation is frequent and the gas-oil ratio changes violently in the actual exploitation process.

Disclosure of Invention

In order to solve at least one of the above technical problems, some aspects of the present invention provide a method and an apparatus for determining gas channeling by PVT experiments and machine learning, so that an accurate gas channeling determination result can be obtained by using wellhead monitoring data, and prediction cost is reduced.

In one aspect, a method for determining gas channeling using PVT experiments and machine learning is provided, the method comprising:

obtaining characteristic parameters of the formation fluid sample under different injection volumes and different pressures and corresponding gas channeling judgment results through a PVT experiment to obtain PVT experiment data;

acquiring the characteristic parameters and corresponding gas channeling judgment results monitored from the well mouths of the wells to obtain monitoring historical data;

combining the PVT experimental data and the monitoring historical data to form an initial training sample, and training the initial training sample by a machine learning method to obtain a prediction model;

inputting the new characteristic parameters monitored by the wellhead into the prediction model to obtain a corresponding gas channeling judgment result.

In at least one embodiment of the invention, the characteristic parameters include one or more of gas-oil ratio, C1 content, and C7 content.

In at least one embodiment of the present invention, further comprising:

inputting the new characteristic parameters monitored by the wellhead into the prediction model, and taking the new characteristic parameters monitored by the wellhead and the corresponding gas channeling judgment results as new training samples after obtaining the corresponding gas channeling judgment results;

and training the prediction model by using the new training sample.

In at least one embodiment of the present invention, the experimental apparatus used in the PVT experiment includes a pressure-resistant cylinder, a PVT cylinder, a first pressure pump, a second pressure pump, a third pressure pump, a sampler, and a dry gas bottle;

the PVT cylinder is arranged in the pressure-resistant cylinder body, and hydraulic oil is filled between the PVT cylinder and the pressure-resistant cylinder body; a piston is arranged in the PVT cylinder, and the interior of the cylinder is divided into a first cavity and a second cavity; the pressure-resistant cylinder body is provided with a confining pressure filling opening, a hydraulic oil inlet and a hydraulic oil outlet, and a sample inlet and a sample outlet; the sample inlet and outlet is communicated with the first cavity; the hydraulic oil inlet and outlet are communicated with the second cavity;

the first pressure pump is connected with the hydraulic oil inlet and outlet through a valve and a pipeline and is used for injecting or extracting hydraulic oil into or from the second cavity; the second pressure pump is connected with the confining pressure injection port through a valve and a pipeline and is used for filling hydraulic oil into the pressure-resistant cylinder body; the third pressure pump is connected with the first end of the sampler and the first end of the dry gas bottle through a valve and a pipeline; and the second end of the sampler and the second end of the dry gas cylinder are connected with the sample inlet and the sample outlet through valves and pipelines and are used for injecting or extracting the formation fluid sample or the dry gas into the first cavity.

In at least one embodiment of the present invention, the obtaining of the characteristic parameters of the formation fluid sample at different injection volumes and different pressures and the corresponding gas channeling determination results through the PVT experiment includes:

s1, filling hydraulic oil into the pressure-resistant cylinder by the second pressure pump to increase confining pressure; pressurizing the second cavity to a sampler pressure by the first pressure pump; when the formation fluid sample is injected into the first cavity through the third pressure pump and the sampler, the first pressure pump is withdrawn at constant pressure, so that the piston moves downwards; closing all valves after the injection of the formation fluid sample is completed;

s2, the first pressure pump is withdrawn to enable the piston to move downwards, the first cavity is depressurized, and after the pressure is reduced, the condensate gas expands to generate redundant gas; the first pressure pump is fed with constant pressure to move the piston upwards, and redundant gas is discharged and is restored to the original volume; measuring components of discharged gas, and measuring gas-oil ratio;

s3, the first pressure pump is withdrawn to enable the piston to move downwards, the pressure of the first cavity is reduced to be lower than the dew point pressure, and after the pressure is reduced, the condensate gas expands to generate redundant gas and partial condensate liquid; the first pressure pump is fed with constant pressure to move the piston upwards, and redundant gas is discharged and is restored to the original volume; measuring components of discharged gas, and measuring gas-oil ratio;

s4, repeating the step S3 until the pressure in the first cavity is reduced to the waste pressure.

In at least one embodiment of the present invention, further comprising:

s5, discharging all the residual formation fluid samples, and cleaning equipment;

s6, executing step S1;

s7, injecting dry gas with a preset volume into the first cavity through the third pressure pump and the dry gas bottle, and then closing all valves;

s8, performing steps S2-S4;

s9, repeating the steps S5-S8, wherein the ambient pressure and the injection volume of the dry gas are different preset values until the required number of PVT experimental data with different injection volumes and different pressures are obtained.

In at least one embodiment of the present invention, the PVT experimental data includes dry gas injection multiple, pressure, gas-oil ratio, C1 content, C7 content, and gas channeling determination result.

In at least one embodiment of the present invention, combining the PVT experimental data and the monitoring history data to form an initial training sample, and training the initial training sample by a machine learning method to obtain a prediction model includes:

training data by using a K nearest neighbor method, dividing original data into training data and verification data according to the proportion of 7:3, and calculating the accuracy; (TP + TN)/(TP + FN + FP + TN); wherein the content of the first and second substances,

FN represents an example that is determined as a negative sample, but is actually a positive sample;

FP represents an example that is determined to be a positive sample, but is in fact a negative sample;

TN represents an example of being determined as a negative sample, and in fact, a negative sample;

TP represents a sample that is determined to be positive, and is in fact an example of a witness sample.

In at least one embodiment of the present invention, inputting new characteristic parameters of wellhead monitoring into the prediction model, and obtaining the corresponding gas channeling determination result includes:

sampling from a wellhead, measuring gas-oil ratio, C1 content, C7 content and pressure data, inputting the measured data into the prediction model, and outputting a gas channeling judgment result by the prediction model.

In another aspect, an apparatus is provided, the apparatus includes a processor and a memory, the memory stores computer program instructions adapted to be executed by the processor, and the computer program instructions are executed by the processor to perform the steps of the method according to any of the above embodiments.

Compared with the prior art, the method for judging the gas channeling by using the PVT experiment and the machine learning determines the characteristic parameters of the formation fluid sample under different injection volumes and different pressures and the corresponding gas channeling judgment result through the PVT experiment, establishes and forms an initial training sample (namely an initial sample library) by combining historical data monitored by each wellhead, and trains the initial training sample through the machine learning method to obtain the prediction model. In actual operation, an accurate gas channeling judgment result can be obtained only by inputting characteristic parameters monitored by a wellhead into the prediction model. Expensive underground sampling test procedures are not required. And the prediction model can be trained by using a newly obtained gas channeling judgment result continuously, so that the prediction accuracy is improved, the prediction time is saved and the prediction cost is reduced while the training sample is enlarged.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the principles of the invention.

FIG. 1 is a schematic flow diagram of a method for determining gas channeling using PVT experiments and machine learning, according to some embodiments;

FIG. 2 is a schematic diagram of a PVT experimental setup according to some embodiments;

FIG. 3 is a schematic diagram of variations in different states during a PVT experiment according to some embodiments;

FIG. 4 is a schematic of gas-to-oil ratio as a function of injection multiple and pressure according to some embodiments;

FIG. 5 is a graphical representation of C1 content as a function of injection multiple and pressure according to some embodiments;

fig. 6 is a schematic diagram of a device configuration for determining gas channeling using PVT experiments and machine learning, according to some embodiments.

Fig. 7 is a schematic view of the blow-by gas in the present invention.

Description of the reference numerals

1-a first pressure pump; 2-a second pressure pump; 3-a third pressure pump; 4,5, 6-pressure gauge; 7,8,9,10,11,12,13,14, 30-valves; 15-pressure cylinder; 16-a piston; 17-a lower cover; 18-upper cover; 19, 20-sample import and export; 21,22,23, 24-fixing bolts; 25-a sampler; 26-confining pressure injection port; 27-annulus; 28, 29-hydraulic oil inlet and outlet; 31-dry gas cylinder; 32-PVT cartridge.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are for purposes of illustration only and are not to be construed as limitations of the invention. It should be noted that, for convenience of description, only the portions related to the present invention are shown in the drawings.

In addition, the embodiments of the present invention and the features of the embodiments may be combined with each other without conflict. The present invention will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.

It should be noted that, the step numbers in the text are only for convenience of explanation of the specific embodiments, and do not serve to limit the execution sequence of the steps.

The method provided by some embodiments of the present invention can be executed by a relevant processor, and the following description will take the processor as an example of an execution subject. The execution subject can be adjusted according to the specific case, such as a server, an electronic device, a computer, and the like.

At present, the research on gas channeling is mainly carried out on a conventional gas reservoir, and the research on gas channeling of a condensate gas reservoir is relatively less. The most effective method for judging whether the injected gas is blown by gas is to carry out underground sampling to determine the composition of the fluid by sampling while drilling, however, the method mainly depends on foreign technologies, and has high cost and poor domestic replacement.

In order to overcome the defects of the conventional gas channeling judgment method, embodiments of the present invention provide a method and an apparatus for judging gas channeling by using PVT experiments and machine learning, so as to judge a gas channeling result more accurately and at low cost. The method can be applied to the technical field of condensate gas reservoir gas injection development.

Referring to fig. 1, a schematic flow chart of a method for determining gas channeling by using PVT experiments and machine learning is shown; the method for judging gas channeling by utilizing PVT (physical vapor transport) experiments and machine learning comprises the following steps of:

step 1, obtaining characteristic parameters of a formation fluid sample under different injection volumes and different pressures and corresponding gas channeling judgment results through a PVT experiment to obtain PVT experiment data. The change of formation fluid under high temperature and high pressure is called high pressure physical property, after the system is determined, the phase state characteristics of the fluid are mainly controlled by the formation pressure, volume and temperature, so the related experiment is called PVT (pressure-volume-temperature) experiment. Formation fluid samples are obtained from downhole sampling. The injection volume refers to the volume of dry gas injected to simulate production conditions. The pressure is to simulate the bottom pressure. The parameters of the production well such as formation pressure, yield, gas-oil ratio, condensate content, crude oil density, dew point pressure, well fluid composition and the like all obey a certain change rule before and after circular gas injection. Therefore, whether the gas channeling exists or not can be judged through the characteristic parameters. The gas channeling judgment result contained in the PVT experimental data is accurate, because whether the well is in a gas channeling state or not can be accurately judged through underground sampling, and then the ground sample corresponding to the underground sample is subjected to PVT analysis. Only one time of underground sampling is needed in the step, and only wellhead data parameters need to be acquired when gas channeling judgment is subsequently carried out. Referring to fig. 7, the gas channeling is a phenomenon that the injected dry gas is light, and the gas channeling occurs from the top of the original condensate gas to form a gas channeling channel and reaches the production well in advance, and after the gas channeling occurs, the development efficiency is poor because a large number of unswept areas exist. In terms of component composition, original condensate gas is uniformly distributed in a reservoir, after dry gas is injected, the concentration of the dry gas at the top part is increased, so that the dry gas is gradually reduced along the depth direction, in addition, the dry gas can extract a part of light components in the condensate gas to cause the condensate gas to be heavier, and therefore, the middle hydrocarbon components and heavy components in the condensate gas tend to increase along with the depth.

Optionally, the characteristic parameters may include one or more of gas-oil ratio, C1 content, and C7 content.

And 2, acquiring the characteristic parameters monitored from the well mouths of the wells and the corresponding gas channeling judgment results to obtain monitoring historical data. Historical monitoring data of each well and historical data of gas channeling judgment results can be better utilized, and the method is not limited to be applicable to single wells.

And 3, combining the PVT experimental data and the monitoring historical data to form an initial training sample, combining the data from two different sources together to form a database serving as a training sample, namely an initial training sample library. And training the initial training sample by a machine learning method to obtain a prediction model. Machine learning refers to the process of using some algorithms to direct a computer to use known data to derive an appropriate model and to use this model to make decisions about new scenarios, the most critical of which is the data throughout. Therefore, the initial training sample is formed by combining the PVT experimental data and the monitoring historical data, the accuracy brought by the PVT experiment is utilized, the huge historical data in the production process of each well is effectively utilized, and the utilization rate of the data and the benefits brought by the data are improved. Machine learning currently has a plurality of different learning algorithms, and the invention can use one of the algorithms to find the internal rule from the unordered data and form a prediction model by processing the data.

And 4, inputting the new characteristic parameters monitored by the wellhead into the prediction model to obtain a corresponding gas channeling judgment result. In the step, the judgment result can be obtained only by inputting the parameters monitored by the wellhead, and an expensive underground sampling test process is not needed.

In some embodiments, the method further comprises:

and 5, inputting the new characteristic parameters monitored by the wellhead into the prediction model, and taking the new characteristic parameters monitored by the wellhead and the corresponding gas channeling judgment results as new training samples after obtaining the corresponding gas channeling judgment results. After the prediction model is established, only after the wellhead monitors new characteristic parameters, the new characteristic parameters are input into the prediction model, and the prediction model obtains a corresponding gas channeling judgment result according to rules established by training. The process can be carried out in real time, namely, the characteristic parameters are directly transmitted to the prediction model after being monitored at the wellhead, and the prediction model gives a gas channeling judgment result in real time. In many cases, when the gas channeling is discovered through underground sampling, the gas channeling is late, the gas channeling is serious, and the application of measures is useless, so that the early discovery of the gas channeling is very important. The method can obtain the judgment result of the gas channeling in real time and improve the reliability.

And 6, training the prediction model by using the new training sample. And forming a new training sample by using the newly obtained characteristic parameters at the wellhead, such as pressure, gas-oil ratio, C1 content and the like, and the corresponding gas channeling judgment result, and continuing to train the prediction model, and so on. Namely, on the basis of the initial training sample, the training sample is continuously expanded, the prediction precision is further improved, the prediction time is saved, and the prediction cost is reduced.

In some embodiments, see figure 2 for a schematic diagram of the structure of the PVT experimental apparatus. The experimental device used for the PVT experiment comprises a pressure-resistant cylinder 15, a PVT cylinder 32, a first pressure pump 1, a second pressure pump 2, a third pressure pump 3, a sampler 25 and a dry gas bottle 31. The pressure cylinder 15 may be a cylindrical cylinder, and an upper cover 18 and a lower cover 17 are respectively installed at both ends of the cylinder to enclose the inside of the cylinder, and the upper cover 18 and the lower cover 17 are respectively fixed to the end surfaces of the cylinder by fixing

bolts

21,22,23, 24. The PVT cylinder 32 is made of transparent sapphire glass and is resistant to pressure of 100 MPa. A formation fluid sample is contained within the sampler 25. The dry gas cylinder 31 stores therein dry gas (methane as a main component). Pressure gauges 4,5 and 6 are respectively arranged on pipelines connected with the first pressure pump 1, the second pressure pump 2 and the third pressure pump 3.

The PVT cylinder 32 is provided in the pressure-resistant cylinder 15, and hydraulic oil is filled between the PVT cylinder 32 and the pressure-resistant cylinder 15, so that an annular space 27 is provided between the PVT cylinder 32 and the pressure-resistant cylinder 15, and rigid contact between the PVT cylinder 32 and the pressure-resistant cylinder 15 is prevented. The PVT cylinder 32 is internally provided with a piston 16, and the piston 16 divides the interior of the PVT cylinder 32 into a first cavity and a second cavity. The pressure cylinder 15 is provided with a confining pressure injection port 26, hydraulic oil inlet and outlet ports 28 and 29 and sample inlet and outlet ports 19 and 20. The confining pressure injection port 26 is arranged on the side wall of the cylinder body and penetrates through the side wall; the hydraulic oil inlet and outlet 28 and 29 are arranged on the lower cover 17 and penetrate through the lower cover 17; sample ports 19,20 are provided in the upper cover 18 and extend through the upper cover 18. The sample inlet and outlet 19,20 communicates with the first chamber; the hydraulic oil inlet and outlet 28 and 29 are communicated with the second cavity.

The first pressure pump 1 is connected with the hydraulic oil inlet and outlet 28 and 29 through valves and pipelines and is used for injecting or extracting hydraulic oil into or from the second cavity; the second pressure pump 2 is connected with the confining pressure injection port 26 through a valve and a pipeline and is used for filling hydraulic oil into the pressure-resistant cylinder body 15 so as to simulate reservoir pressure. The third pressure pump 3 is connected with the first end of the sampler 25 and the first end of the dry gas bottle 31 through valves and pipelines; the second end of the sampler 25 and the second end of the dry gas cylinder 31 are connected to the sample inlet and outlet 19,20 through valves and pipes, and are respectively used for injecting or extracting the formation fluid sample or dry gas into or from the first cavity.

In some embodiments, referring to the schematic diagram of the variation of different states during the PVT experiment shown in fig. 3, there are 7 states from left to right, which are state 1 to state 7 sequentially. Obtaining characteristic parameters of the formation fluid sample under different injection volumes and different pressures and corresponding gas channeling judgment results through a PVT experiment, wherein the obtained PVT experiment data comprises the following steps:

s1, filling hydraulic oil into the pressure-resistant cylinder 15 by the second pressure pump 2 to increase the confining pressure; pressurizing the second cavity to sampler 25 pressure by the first pressure pump 1; while the formation fluid sample is injected into the first cavity through the third pressure pump 3 and the sampler 25, the first pressure pump 1 is released at constant pressure, so that the piston 16 moves downwards; all valves are closed after injection of the formation fluid sample is complete. Specifically, during the experiment, the second pressure pump 2 and the valve 13 are firstly opened to add confining pressure, then the valves 8 and 9 are opened to pressurize to the pressure of the sampler 25, the valves 10,11 and 12 are opened to inject the formation fluid sample, the constant pressure pump is withdrawn to move the piston 16 downwards, all the valves are closed after the transfer is completed, and the following constant volume failure experiment is carried out. This step may be referred to as the confining pressure and sample transfer step.

S2, the first pressure pump 1 is withdrawn to enable the piston 16 to move downwards, the pressure of the first cavity is reduced, and after the pressure is reduced, the condensate gas expands to generate redundant gas; the first pressure pump 1 is constant in pressure and pumps to move the piston 16 upwards, so that redundant gas is discharged and the original volume is restored; the exhaust gas was measured for composition and gas-oil ratio. Specifically, the valves 7,8 and 9 are opened, the first pressure pump 1 is withdrawn to reduce the pressure, the piston 16 moves downwards, the condensate gas expands after the pressure is reduced, the redundant gas is generated (state 2), the valves 10 and 11 are opened, the constant pressure pump is fed, the redundant gas is discharged and is restored to the original volume (state 3), the discharged gas is used for measuring components, and the gas-oil ratio is measured.

S3, the first pressure pump 1 is withdrawn to enable the piston 16 to move downwards, the pressure of the first cavity is reduced to be lower than the dew point pressure, and after the pressure is reduced, the condensate gas expands to generate redundant gas and partial condensate liquid; the first pressure pump 1 is constant in pressure and pumps to move the piston 16 upwards, so that redundant gas is discharged and the original volume is restored; the exhaust gas was measured for composition and gas-oil ratio. Specifically, the valves 7,8, and 9 are opened, the first pressure pump 1 is withdrawn to reduce the pressure, the piston 16 moves downward, and after the pressure is reduced, the condensate gas expands to generate excess gas (state 4), and partial condensate liquid is generated while the dew point pressure is lower than the dew point pressure (the dew point pressure is the pressure when the first drop of liquid occurs during the pressure reduction). The valves 10 and 11 are opened, the pump is fed at constant pressure, the excess gas is discharged and returned to the original volume (state 5), the gas is discharged for component determination, and the gas-oil ratio is determined.

And S4, repeating the step S3, wherein the pressure of the first cavity is gradually reduced and the condensate is gradually increased in the repeated process (state 6) until the pressure in the first cavity is reduced to the waste pressure, and ending the process of the constant volume exhaustion experiment. The waste pressure is also called exhaustion pressure, and as the pressure decreases, the pressure of the gas reservoir gradually decreases until the final pressure is exhausted, so that the generation of stable gas flow cannot be ensured, or the cost for maintaining operation is higher than the development benefit, and the pressure at this time is the exhaustion pressure.

In the process of the constant volume exhaustion experiment, dry gas is not injected into the first cavity, namely the condensate gas constant volume exhaustion experiment in the original state, and the corresponding injection multiple is 0.

In some embodiments, to obtain experimental data at different implantation multiples, the method further comprises:

and S5, discharging all the residual formation fluid samples and cleaning the equipment. And preparing for the next constant volume failure experiment.

S6, step S1 is performed, that is, the process of performing the step is the same as that of step S1, and the confining pressure and the sample transfer are performed.

S7, injecting a preset volume of dry gas into the first cavity through the third pressure pump 3 and the dry gas bottle 31, and then closing all valves. The step is to increase the volume of the injected dry gas by a set increment value each time according to the set injection volume increment in order to perform the constant volume exhaustion experiment with different injection volumes. For example, the increase in the volume of dry gas injected in this step is 0.1 times the volume of condensate, i.e. the ratio of the volume of dry gas to the volume of condensate is 0.1. Specifically, the valve 12 is closed, the valve 30 is opened, the dry gas 31 (main component methane) with the volume 0.1 time that of the condensate gas is injected, and then all the valves are closed to perform the constant volume exhaustion experiment.

S8, performing steps S2-S4; namely, the step of performing the constant volume exhaustion experiment again under the condition of injecting dry gas with the volume 0.1 time of the condensate gas.

S9, repeating the steps S5-S8, wherein the ambient pressure and the injection volume of the dry gas are different preset values until the required number of PVT experimental data with different injection volumes and different pressures are obtained. That is, the experiment is repeated for a plurality of times, the number of times of repetition is determined according to the required data amount, and in each repeated experiment, the applied confining pressure and the injected dry gas volume are changed to reflect the experimental data in different states.

Optionally, the PVT experimental data includes dry gas injection multiple, pressure, gas-oil ratio, C1 content, C7 content, and gas channeling determination result. The experimental data obtained are shown below by way of example.

And (4) making a data table, wherein the data table mainly comprises pressure, dry gas injection times, gas-oil ratio, C1 content, C7 content and whether gas channeling exists. The gas-oil ratio, the C1 content and the C7 content are interpolated at intervals of 0.1MPa to obtain enough data. See table 1 for the raw data table.

TABLE 1 original data sheet

Because the magnitude difference between different data is large, the result is influenced in the calculation process, and all values are normalized to be between 0 and 1 for the convenience of computer storage. See table 2 for normalized data.

TABLE 2 normalized data sheet

Multiple of injection	Pressure of	Gas-oil ratio	C1	C7	Whether or not there is gas channeling
						0.0	1.000	0.050	0.000	0.080	0
0.1	0.842	0.146	0.705	0.800	1
						0.2	0.794	0.040	0.095	0.903	0
0.2	0.636	0.056	0.119	0.820	0
						0.4	0.599	0.435	0.577	0.360	1
0.1	0.490	0.084	0.097	0.312	0
						0.1	0.413	0.193	0.114	0.000	1
0.1	0.328	0.109	0.034	0.243	0
						0.1	0.184	0.122	0.101	0.206	0
0.3	0.136	0.125	0.316	0.240	1
						0.2	0.069	0.121	0.060	0.122	0
0.2	0.042	0.137	0.004	0.280	0
						0.1	0.000	1.000	1.000	0.094	1

See the schematic diagram of the gas-oil ratio as a function of injection multiple and pressure shown in fig. 4 and the schematic diagram of the C1 content as a function of injection multiple and pressure shown in fig. 5. In the figure, all the points on the line are obtained in the case of uniform mixing, which means that the injected gas does not undergo a gas channeling phenomenon, so the points on the line are all points where no gas channeling occurs, and there is a possibility that gas channeling occurs at points off the line. By establishing key parameter databases under different injection volumes and different pressures, the occurrence of gas channeling can be judged.

In some embodiments, forming an initial training sample by combining the PVT experimental data and the monitoring historical data, and training the initial training sample by a machine learning method to obtain a prediction model includes:

training data by using a K nearest neighbor method, dividing original data into training data and verification data according to the proportion of 7:3, and calculating the accuracy; (TP + TN)/(TP + FN + FP + TN);

wherein the content of the first and second substances,

TP represents a sample that is determined to be positive, and is in fact an example of a positive sample.

See table 3 for an accuracy description.

TABLE 3 accuracy description Table

Optionally, python may be used for the training, and the algorithm may also use other machine learning algorithms such as a support vector machine.

Furthermore, the obtained prediction model can be optimized, the accuracy is improved by increasing the number of training samples, eliminating abnormal values and the like until the accuracy reaches more than 98%, and the model optimization is completed.

In some embodiments, inputting new characteristic parameters of wellhead monitoring into the prediction model, and obtaining a corresponding gas channeling judgment result comprises:

sampling from a wellhead, measuring gas-oil ratio, C1 content, C7 content and pressure data, inputting the measured data into the prediction model, and outputting a gas channeling judgment result by the prediction model. In particular, during the application, to measure the corresponding data after the well has been steadily produced, the data may be distorted if the well is suddenly opened after the well is shut in.

When the prediction model is applied, the pressure corresponding to the test sample, the component content of the ground sample, the gas-oil ratio and the dry gas injection amount are input to judge whether the sample has gas channeling or not. The surface sample means that the difference between the fluid obtained from a surface wellhead and the formation fluid is mainly different in temperature and pressure, but the surface sample is easy to obtain, and the formation fluid needs a special underground sampling tool, is complex to operate and is not easy to obtain. A specific set of predicted outcome data is shown in table 4.

TABLE 4 prediction results data sheet

Multiple of injection	Pressure of	Gas-oil ratio	C1	C7	Whether or not there is gas channeling
						0.2	53.2	1550.245	0.755	0.059	Without gas channeling
0.1	52.5	1690.654	0.776	0.058	Gas channeling
						0.2	52.0	1602.312	0.753	0.056	Without gas channeling

Some embodiments of the present invention further provide a device for determining gas channeling through PVT experiments and machine learning, which is shown in fig. 6, and includes a communication interface 1000, a memory 2000 and a processor 3000. The communication interface 1000 is used for communicating with an external device to perform data interactive transmission. The memory 2000 has stored therein a computer program that is executable on the processor 3000. The number of the memory 2000 and the processor 3000 may be one or more.

If the communication interface 1000, the memory 2000 and the processor 3000 are implemented independently, the communication interface 1000, the memory 2000 and the processor 3000 may be connected to each other through a bus to complete communication therebetween. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown, but this does not represent only one bus or one type of bus.

Optionally, in a specific implementation, if the communication interface 1000, the memory 2000, and the processor 3000 are integrated on a chip, the communication interface 1000, the memory 2000, and the processor 3000 may complete communication with each other through an internal interface.

The processor is configured to support the acquiring device to perform one or more steps of the method for determining gas channeling by using PVT experiments and machine learning according to any of the embodiments. The processor may be a Central Processing Unit (CPU), or may be other general-purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, or the like. Wherein a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.

The memory has stored therein computer program instructions adapted to be executed by the processor, the computer program instructions, when executed by the processor, performing one or more steps of the method for determining gas breakthrough using PVT experiments and machine learning as described in any of the above embodiments.

The Memory may be a Read-Only Memory (ROM) or other type of static storage device that can store static information and instructions, a Random Access Memory (RAM) or other type of dynamic storage device that can store information and instructions, an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc Read-Only Memory (CD-ROM) or other optical Disc storage, optical Disc storage (including Compact Disc, laser Disc, optical Disc, digital versatile Disc, blu-ray Disc, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited to these. The memory may be self-contained and coupled to the processor via a communication bus. The memory may also be integral to the processor.

In the description herein, reference to the description of the terms "one embodiment/mode," "some embodiments/modes," "example," "specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment/mode or example is included in at least one embodiment/mode or example of the application. In this specification, the schematic representations of the terms used above are not necessarily intended to be the same embodiment/mode or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments/modes or examples. Furthermore, the various embodiments/aspects or examples and features of the various embodiments/aspects or examples described in this specification can be combined and combined by one skilled in the art without conflicting therewith.

Furthermore, the terms "first", "second" and "first" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "a plurality" means at least two, e.g., two, three, etc., unless specifically limited otherwise. Meanwhile, in the description of the present invention, unless otherwise explicitly specified or limited, the terms "connected" and "connected" should be interpreted broadly, for example, as being fixedly connected, detachably connected, or integrally connected; the connection can be mechanical connection or electrical connection; may be directly connected or indirectly connected through an intermediate. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.

It will be understood by those skilled in the art that the foregoing embodiments are merely for clarity of description and are not intended to limit the scope of the invention. Other variations or modifications will occur to those skilled in the art based on the foregoing disclosure and are within the scope of the invention.

Claims

1. A method for determining gas channeling using PVT experiments and machine learning, the method comprising:

2. The method for determining gas channeling using PVT experiments and machine learning of claim 1, wherein said characteristic parameters include one or more of gas-oil ratio, C1 content and C7 content.

3. The method for determining gas channeling using PVT experiments and machine learning of claim 1, further comprising:

and training the prediction model by using the new training sample.

4. The method for judging gas channeling by using PVT experiments and machine learning as claimed in claim 1, wherein experimental devices adopted by the PVT experiments comprise a pressure-resistant cylinder, a PVT cylinder, a first pressure pump, a second pressure pump, a third pressure pump, a sampler and a dry gas bottle;

5. The method for judging gas channeling by utilizing PVT (physical vapor transport) experiments and machine learning according to claim 4, wherein the characteristic parameters of the formation fluid samples under different injection volumes and different pressures and corresponding gas channeling judgment results are obtained through the PVT experiments, and the obtaining of the PVT experimental data comprises the following steps:

6. The method for determining gas channeling using PVT experiments and machine learning of claim 5, further comprising:

s6, executing step S1;

s8, performing steps S2-S4;

7. The method of determining gas channeling using PVT experiments and machine learning of claim 6, wherein the PVT experimental data includes dry gas injection multiple, pressure, gas-oil ratio, C1 content, C7 content and gas channeling determination result.

8. The method for judging gas channeling by using PVT (physical vapor transport) experiments and machine learning as claimed in claim 1, wherein the step of combining the PVT experimental data and the monitoring historical data to form an initial training sample, and the step of training the initial training sample by using a machine learning method to obtain a prediction model comprises the steps of:

9. The method for judging gas channeling using PVT experiments and machine learning according to claim 1, wherein inputting new characteristic parameters monitored by a wellhead into the prediction model and obtaining a corresponding gas channeling judgment result comprises:

10. An apparatus for determining gas channeling using PVT experiments and machine learning, the apparatus comprising a processor and a memory, the memory having stored therein computer program instructions adapted to be executed by the processor, the computer program instructions when executed by the processor performing the steps of the method of any one of claims 1-9.