CN117569803A

CN117569803A - Prediction method and system for automatic drug delivery drainage of natural gas well

Info

Publication number: CN117569803A
Application number: CN202410079263.XA
Authority: CN
Inventors: 阳登科; 胡强; 邓永红
Original assignee: Sichuan Suju Zhilian Technology Co ltd
Current assignee: Sichuan Suju Zhilian Technology Co ltd
Priority date: 2024-01-19
Filing date: 2024-01-19
Publication date: 2024-02-20
Anticipated expiration: 2044-01-19
Also published as: CN117569803B

Abstract

The invention provides a prediction method and a prediction system for automatic drug delivery and drainage of a natural gas well, which are used for predicting the liquid accumulation of the natural gas well in one or more subsequent production time sequences in subsequent production activities after a past production activity after obtaining production monitoring data of a plurality of past production time sequences of the natural gas well in the past production activity, determining the loss of the liquid accumulation of the natural gas well in the production monitoring data, then adjusting the liquid accumulation of the natural gas well according to the loss to obtain the liquid accumulation of the target natural gas well in each subsequent production time sequence, and then determining the maximum liquid accumulation in the liquid accumulation of the target natural gas well. The loss of the natural gas well liquid accumulation amount can be determined and obtained on the basis of predicting the natural gas well liquid accumulation amount of the subsequent production time sequence, so that the predicted target natural gas well liquid accumulation amount not only brings complex long-time macroscopic characteristics into analysis, but also captures local changes such as short-term regular changes and the like, and the accuracy of maximum liquid accumulation amount prediction is improved.

Description

Prediction method and system for automatic drug delivery drainage of natural gas well

Technical Field

The application relates to the technical fields of data processing and machine learning, and in particular relates to a prediction method and a prediction system for automatic drug delivery and drainage of a natural gas well.

Background

During the production of natural gas wells, liquid products may form in the wellbore due to the drop in formation pressure and the production of natural gas. The presence of the liquid can affect the production of natural gas and wellhead pressure and even lead to a shut-in of the gas well. In order to maintain normal production of a gas well, it is necessary to periodically perform a water drainage operation on the gas well. It is clear that in the drainage of natural gas wells, liquid accumulation is an important indicator which directly affects the cleaning effect, the cost and the safe operation of the equipment. At present, a common drainage method is to reduce the surface tension of accumulated liquid in a shaft by adding a medicament so as to facilitate drainage. However, in the process of draining the injected medicament, the accumulated liquid amount changes due to the influence of factors such as the production condition and geological conditions of the gas well, so that the maximum accumulated liquid amount of the gas well needs to be predicted in time so as to take corresponding measures. At present, traditional liquid accumulation prediction methods are mainly based on empirical formulas and statistical models, and the methods require a large amount of field data and complex calculation, and have limited accuracy and stability under the complex and changeable actual conditions. Thus, there is a need for a more accurate and reliable predictive method for guiding the administration and drainage steps of a medicament. In recent years, artificial intelligence technology has made significant progress in the field of data analysis and prediction. Among them, techniques such as machine learning and deep learning can improve accuracy and generalization ability of prediction by learning and pattern recognition of a large amount of data. In the drainage process of a natural gas well, the maximum hydrops prediction by utilizing an artificial intelligence technology has important application value, and how to reasonably apply related technologies to perform hydrops prediction is a precondition for guaranteeing the cleaning of the hydrops and the control of the cost.

Disclosure of Invention

In view of this, the embodiments of the present application at least provide a method and a system for predicting automatic drug delivery and drainage of a natural gas well.

The technical scheme of the embodiment of the application is realized as follows:

in one aspect, an embodiment of the present application provides a prediction method for automatic drug delivery and drainage of a natural gas well, applied to a prediction system, the method including:

acquiring a past production monitoring dataset of a target natural gas well, the past production monitoring dataset comprising production monitoring data of a plurality of past production time sequences in a past production campaign;

predicting, from the production monitoring data, gas well fluid production in a subsequent production campaign, the subsequent production campaign including one or more subsequent production time sequences following the past production campaign;

determining a loss of the gas well fluid production volume from the production monitoring data, the loss being indicative of long-term regular changes and short-term regular changes that are ignored when predicting the gas well fluid production volume;

adjusting the natural gas well liquid accumulation volume according to the loss to obtain a target natural gas well liquid accumulation volume of each subsequent production time sequence;

and determining the maximum accumulated liquid amount in the target natural gas well accumulated liquid amount.

In some embodiments, said predicting natural gas well liquid production in subsequent production activities from said production monitoring data comprises:

determining the data integrity of the production monitoring data, and cleaning the production monitoring data according to the data integrity determination result to obtain target production monitoring data of each past production time sequence in the past production activities;

and extracting a maximum value prediction branch network based on a maximum fluid accumulation amount prediction neural network from the target production monitoring data to obtain a production description vector in the past production activity, and predicting to obtain one or more subsequent production time series of natural gas well fluid accumulation amounts according to the production description vector.

In some embodiments, the performing data cleansing on the production monitoring data according to the data integrity determination result to obtain target production monitoring data of each past production time sequence in the past production campaign includes:

adjusting the production monitoring data according to the data integrity determination result to obtain adjusted production monitoring data of each past production time sequence;

normalizing the adjusted production monitoring data to a preset numerical range to obtain transitional production monitoring data of each past production time sequence;

And carrying out trend decomposition on the transitional production monitoring data to obtain target production monitoring data of each past production time sequence.

In some embodiments, the production monitoring data includes a liquid accumulation volume log and a mining geological log, the liquid accumulation volume log includes a total liquid accumulation volume and a dispersed liquid accumulation volume, the production monitoring data is adjusted according to a data integrity determination result, and adjusted production monitoring data of each past production time sequence is obtained, including:

if the data integrity determination result shows that the total amount of the effusion is intermittent and vacant, supplementing the vacant total amount of the effusion to obtain production monitoring data after adjustment;

if the data integrity determination result shows that the distribution condition of the dispersed accumulated liquid in the set time interval meets the set requirement and uninterrupted gaps appear, supplementing the dispersed accumulated liquid in the gaps in the production monitoring data based on linear fitting to obtain adjusted production monitoring data;

and if the data integrity determination result shows that the mining geological log has uninterrupted gaps, determining a current past production time sequence of the mining geological log with uninterrupted gaps in the past production time sequence, and simultaneously clearing production monitoring data of the current past production time sequence to obtain adjusted production monitoring data.

In some embodiments, normalizing the adjusted production monitoring data to a preset value range to obtain transitional production monitoring data for each past production time series comprises:

acquiring average numbers and average dispersion of the adjusted production monitoring data corresponding to the plurality of past production time sequences;

respectively obtaining deviation results between the adjusted production monitoring data of each previous production time sequence and the average number of the adjusted production monitoring data;

acquiring the proportion between the deviation result and the average dispersity as transitional production monitoring data corresponding to each past production time sequence;

or normalizing the adjusted production monitoring data to a preset numerical range to obtain transitional production monitoring data of each previous production time sequence, wherein the method comprises the following steps:

extracting the maximum elements and the minimum elements of the data elements from the adjusted production monitoring data corresponding to the plurality of past production time sequences;

obtaining a deviation result between the maximum element and the minimum element, obtaining an extreme value deviation result, obtaining a deviation result between a data element of the regulated production monitoring data and the minimum element, and obtaining a target deviation result;

Acquiring the ratio of the target deviation result to the extreme value deviation result as transitional production monitoring data of each past production time sequence;

the trend decomposition is performed on the transitional production monitoring data to obtain target production monitoring data of each past production time sequence, including:

determining one or more transitional production monitoring data in the transitional time period from the transitional production monitoring data to obtain a transitional production monitoring data set;

acquiring the average number of data elements of the transitional production monitoring data in the transitional production monitoring data set to acquire the average number corresponding to the transitional time period;

and obtaining deviation results of the average number corresponding to the transitional production monitoring data and the transitional time period as target production monitoring data of each past production time sequence.

In some embodiments, the maximum value prediction branch network based on the maximum fluid accumulation prediction neural network extracts a production description vector in the past production activity from the target production monitoring data, predicts one or more subsequent production time series fluid accumulation of the natural gas well according to the production description vector, and comprises:

Determining and obtaining target production monitoring data of a target past production time sequence from the target production monitoring data;

performing description vector mining on the target production monitoring data of the target past production time sequence based on a maximum value prediction branch network of a maximum liquid accumulation amount prediction neural network to obtain a mining description vector of the target past production time sequence, wherein the mining description vector comprises a liquid accumulation amount description vector and a gate description vector;

adjusting the liquid accumulation amount description vector through the gate description vector to obtain a target liquid accumulation amount description vector corresponding to the past production activity;

determining and obtaining one or more natural gas well liquid production volumes of subsequent production time sequences according to the target liquid production volume description vector;

the determining in the production monitoring data a loss of fluid production from the gas well comprises:

determining and obtaining short-time rule data in the target production monitoring data;

adjusting a branch network to perform adjacent difference calculation on the target production monitoring data based on the short-time law of the maximum liquid accumulation amount prediction neural network to obtain adjacent difference data;

and determining the loss of the natural gas well accumulated liquid volume of each subsequent production time sequence in the adjacent difference data according to the short-time rule data.

In some embodiments, before the maximum value prediction branch network based on the maximum fluid accumulation amount prediction neural network extracts the production description vector in the previous production activity from the target production monitoring data and predicts one or more subsequent production time series of gas well fluid accumulation amounts according to the production description vector, the method further comprises:

acquiring a past production monitoring training data set of the target natural gas well, wherein the past production monitoring training data set comprises a plurality of production monitoring training data of past training production time sequences;

debugging a preset maximum value prediction branch network based on the production monitoring training data, and determining training data loss corresponding to the production monitoring training data through the debugged maximum value prediction branch network;

debugging the preset short-time law adjustment branch network according to the training data loss to obtain a short-time law adjustment branch network after debugging is completed;

the maximum liquid accumulation amount prediction neural network is generated through the maximum value prediction branch network after debugging and the short-time regulation branch network after debugging;

the debugging of the preset maximum value prediction branch network based on the production monitoring training data comprises the following steps:

Determining the data integrity of the production monitoring training data, and cleaning the data of the production monitoring training data according to the result of determining the data integrity of the training data to obtain target production monitoring training data of each past training production time sequence;

dividing the target production monitoring training data into a support set, a development set and a test set according to a preset percentage;

debugging the preset maximum value prediction branch network according to the support set to obtain a maximum value prediction branch network after debugging is completed;

the maximum value prediction branch network completed through debugging determines the training data loss corresponding to the production monitoring training data, and the method comprises the following steps:

and predicting a branch network based on the maximum value after debugging, and determining the training data loss corresponding to the production monitoring training data according to the support set and the development set.

In some embodiments, the determining, by the branch network based on the maximum value prediction after debugging, the training data loss corresponding to the production monitoring training data according to the support set and the development set includes:

predicting the natural gas well liquid accumulation amounts respectively corresponding to the support set and the development set based on the maximum value prediction branch network after debugging is completed so as to obtain predicted natural gas well liquid accumulation amounts;

Acquiring a natural gas well effusion quantity tag in the production monitoring training data, and comparing the predicted natural gas well effusion quantity with a corresponding natural gas well effusion quantity tag to obtain a debugging loss of the support set and a development loss of the development set;

taking the debugging loss and the development loss as training data loss corresponding to the production monitoring training data;

the maximum value prediction branch network completed through the debugging and the short-time regulation branch network completed through the debugging generate a maximum liquid accumulation amount prediction neural network, which comprises the following steps:

determining the maximum value prediction branch network after debugging and the short-time regulation branch network after debugging as a transitional maximum liquid accumulation amount prediction neural network;

based on the transitional maximum accumulated liquid amount prediction neural network, predicting the accumulated liquid amount of the natural gas well corresponding to the test set, and obtaining transitional predicted accumulated liquid amount of the natural gas well;

acquiring a target natural gas well effusion quantity label corresponding to the test set, and respectively acquiring the second power of a deviation result between the transitional predicted natural gas well effusion quantity of each production monitoring training data in the test set and the target natural gas well effusion quantity label to obtain a effusion quantity square error;

Obtaining the average of the squared differences of the liquid volumes so as to obtain a test value of the transient maximum liquid volume prediction neural network;

and if the test value is smaller than the set test value, determining the transitional maximum liquid accumulation prediction neural network as a maximum liquid accumulation prediction neural network.

In some embodiments, said adjusting said natural gas well fluid production based on said losses to obtain a target natural gas well fluid production for each subsequent production time series comprises:

determining and obtaining a target loss corresponding to the accumulated liquid volume of each natural gas well in the losses;

summing the target losses with the corresponding gas well fluid volumes to obtain a target gas well fluid volume for each subsequent production time sequence;

the acquiring of past production monitoring data sets for a target natural gas well comprises:

acquiring a liquid accumulation amount log and a past exploitation geological log of the target natural gas well in the past production activity;

determining and obtaining one or more past production time series accumulated liquid logs in the accumulated liquid logs, and determining and obtaining the past production time series mined geological logs in the past mined geological logs;

Determining sequence data corresponding to each past production time sequence, and taking the accumulated liquid volume log, the mining geological log and the sequence data as a past production monitoring data set of the target natural gas well;

after determining the maximum liquid accumulation amount in the target natural gas well liquid accumulation amount, the method further comprises the following steps:

and determining a subsequent production time sequence in which the maximum liquid accumulation is obtained in the subsequent production activity, and obtaining a target subsequent production time sequence corresponding to the maximum liquid accumulation.

In another aspect, the present application provides a prediction system comprising a memory and a processor, the memory storing a computer program executable on the processor, the processor implementing the steps of the method described above when the program is executed.

The beneficial effects of this application include at least:

after production monitoring data of a plurality of past production time sequences of a target natural gas well in past production activities are obtained, the production monitoring data are used for predicting the natural gas well liquid accumulation of one or more of the following production time sequences in the following production activities after the past production activities, loss of the natural gas well liquid accumulation is determined in the production monitoring data, then the natural gas well liquid accumulation is adjusted according to the loss, the target natural gas well liquid accumulation of each of the following production time sequences is obtained, and then the maximum liquid accumulation is determined in the target natural gas well liquid accumulation. The method can determine the loss of the natural gas well liquid accumulation based on the prediction of the natural gas well liquid accumulation of the subsequent production time sequence, and meanwhile, the loss can be considered to be obtained based on the long-time regular change and the short-time regular change of the captured production monitoring data, so that the predicted target natural gas well liquid accumulation not only brings complex long-time macroscopic characteristics into analysis, but also captures local changes such as short-time regular change, and further, the interference between the predicted natural gas well liquid accumulation and different neural networks of the loss can be avoided based on loss regulation, and the accuracy of the maximum liquid accumulation prediction can be increased.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the aspects of the present application.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and, together with the description, serve to explain the technical aspects of the application.

Fig. 1 is a schematic implementation flow chart of a prediction method for automatic drug delivery and drainage of a natural gas well according to an embodiment of the present application.

Fig. 2 is a schematic diagram of a composition structure of a prediction apparatus according to an embodiment of the present application.

Fig. 3 is a schematic hardware entity diagram of a prediction system according to an embodiment of the present application.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application are further elaborated below in conjunction with the accompanying drawings and examples, which should not be construed as limiting the present application, and all other embodiments obtained by those skilled in the art without making inventive efforts 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 to be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with one another without conflict. The term "first/second/third" is merely to distinguish similar objects and does not represent a specific ordering of objects, it being understood that the "first/second/third" may be interchanged with a specific order or sequence, as permitted, to enable embodiments of the present application described herein to be practiced otherwise than as illustrated or described herein. 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 the present application only and is not intended to be limiting of the present application.

The embodiment of the application provides a prediction method for automatic drug delivery and drainage of a natural gas well, which can be executed by a processor of a prediction system. The prediction system may refer to a computer device with data processing capability, such as a server, a notebook computer, a tablet computer, and a desktop computer.

Fig. 1 is a schematic implementation flow chart of a prediction method for automatic drug delivery and drainage of a natural gas well according to an embodiment of the present application, as shown in fig. 1, the method includes the following steps 110 to 150:

step S110, acquiring a past production monitoring dataset of the target natural gas well.

The past production monitoring data set comprises a plurality of past production time series production monitoring data in the past production activities. The production activities of natural gas wells that occur within a time interval prior to the current time. The past production activities include a plurality of past production time sequences, that is, a production time sequence before the current time, one production time sequence corresponds to one statistical period, and the time length of the period is not limited, and it can be understood that the shorter the time length is, the more frequent the statistics is.

The production monitoring data includes data associated with production of the fluid in the target natural gas well, and the production detection data is recorded as log data including fluid production logs, production geology logs, sequence data, and the like. The fluid accumulation log includes the fluid accumulation in the gas well collected, and specifically may include the total fluid accumulation and the dispersed fluid accumulation. The dispersed fluid production is the fluid production at different locations in a natural gas well due to non-uniform geology or irregular wellbores. The production geological log includes data records relating to the production volume of the target gas well, such as data including gas production, formation permeability, bottom hole pressure, production time, well depth, etc. The sequence data is time period data corresponding to the production time sequence, such as date (e.g., week, month, date, etc.). The method of acquiring the past production monitoring data set of the target natural gas well is, for example, to acquire a fluid accumulation log and a past production geological log of the target natural gas well in a past production activity, to determine one or more fluid accumulation logs of past production time series from the fluid accumulation log, to determine a production geological log of the past production time series from the past production geological log, to determine sequence data corresponding to each of the past production time series, and to determine the fluid accumulation log, the production geological log and the sequence data as the past production monitoring data set of the target natural gas well. The accumulated liquid log comprises data records of accumulated liquid of the target natural gas well in past production activities, the past exploitation geological log is records of exploitation conditions and geological conditions of the target natural gas well in the past production activities, one or more accumulated liquid logs of past production time sequences are determined in the accumulated liquid log, for example, the past production activities are divided into one or more past production time sequences, and then accumulated liquid of each past production time sequence is determined in the accumulated liquid log.

The method of determining the mining geological log of the past production time series from the past mining geological logs is, for example, adding the past production time series to the mining geological log of the past production time series in one period of one hour, and determining the mining geological log of each hour in the preset past production activity from the mining geological log to obtain the mining geological log of each past production time series.

The method of acquiring the sequence data corresponding to each previous production time series may be, for example, to acquire the target time of each previous production time series, and determine the sequence data corresponding to the production time series based on the target time. In one example, the set sequence data includes week, day, month, and for week, the value interval is (1, 7), and if the target time of the past production time series is wednesday, the value is 3, the week data of each of the past production time series is obtained. Accordingly, the numerical interval is (1, 31) for the day, and (1, 12) for the month. And determining the liquid accumulation amount log and the mining geological log, and determining the liquid accumulation amount log, the mining geological log and the sequence data as a past production monitoring data set of the target natural gas well after determining the liquid accumulation amount log and the mining geological log.

The past production monitoring data set may include at least one type of liquid accumulation log, at least one type of mining geological log, at least one type of sequence data, and assuming that the liquid accumulation log includes total liquid accumulation and dispersed liquid accumulation, the mining geological log includes natural gas yield, formation permeability, bottom hole pressure, mining time and well depth, the sequence data includes week data, day data and month data, then the production monitoring data set is time sequence data with 10 characteristic elements, and is labeled as SN, SN (n) may be described as a characteristic vector with a dimension of 10, and each element represents total liquid accumulation, dispersed liquid accumulation, natural gas yield, formation permeability, bottom hole pressure, mining time, well depth, week data and day data month data corresponding to time n. I.e., SN (n) = (total fluid production, dispersed fluid production, natural gas production, formation permeability, bottom hole pressure, production time, well depth, weekly data, daily data, monthly data).

Step S120, predicting the natural gas well liquid production volume in the subsequent production activities by the production monitoring data.

The subsequent production campaign is one or more subsequent production time series in the future after the past production campaign. For example, assume that a past production campaign includes an s-b to s-th past production time series, 1 < b < s, and a subsequent production campaign includes one or more subsequent production time series from an s+1 to an s+b-th subsequent production time series. The natural gas well fluid production may be considered as the total fluid production of the subsequent production time series.

The process of predicting the amount of gas well fluid production in subsequent production activities from the production monitoring data may specifically include: and carrying out data integrity determination on the production monitoring data, carrying out data cleaning on the production monitoring data according to a data integrity determination result to obtain target production monitoring data of each past production time sequence in the past production activities, extracting a mining description vector in the past production activities from the target production monitoring data based on a maximum value prediction branch network of the maximum liquid accumulation prediction neural network, and predicting one or more gas well liquid accumulation volumes in the subsequent production time sequences according to the mining description vector. Specifically, the above-mentioned process includes:

step S121, determining the data integrity of the production monitoring data, and performing data cleaning on the production monitoring data according to the data integrity determination result to obtain the target production monitoring data of each previous production time sequence in the previous production activities.

The process of determining the data integrity of the production monitoring data includes, for example: and determining the integrity of the liquid accumulation volume log and the mining geological log in the production monitoring data to obtain a data integrity determination result of the production monitoring data, or determining whether a data vacancy occurs in the liquid accumulation volume log and the mining geological log of the production monitoring data to obtain a data integrity determination result of the production monitoring data. After the data integrity determination is performed on the production monitoring data, performing data cleaning on the production monitoring data according to the data integrity determination result to obtain target production monitoring data of each past production time sequence in the past production activities. The process of data cleaning the production monitoring data includes, for example: and adjusting the production monitoring data according to the data integrity determination result to obtain adjusted production monitoring data of each past production time sequence, normalizing the adjusted production monitoring data to a preset numerical range to obtain transitional production monitoring data of each past production time sequence, and carrying out trend decomposition on the transitional production monitoring data to obtain target production monitoring data of each past production time sequence. The production monitoring data may include a fluid accumulation volume log and a production geological log, and the fluid accumulation volume log may include a total fluid accumulation volume and a dispersed fluid accumulation volume. The process of adjusting the production monitoring data in dependence on the data integrity determination result comprises, for example: and if the data integrity determination result shows that the total amount of the accumulated liquid has discontinuous gaps, supplementing the total amount of the accumulated liquid with the gaps to obtain adjusted production monitoring data, if the data integrity determination result shows that the distribution condition of the scattered accumulated liquid in a set time interval meets the set requirement, and meanwhile, when the continuous gaps occur, supplementing the scattered accumulated liquid with the gaps in the production monitoring data based on linear fitting to obtain the adjusted production monitoring data, and if the data integrity determination result shows that the continuous gaps occur in the exploitation geological log, determining a current past production time sequence in which the continuous gaps occur in the exploitation geological log in the past production time sequence, and simultaneously clearing the production monitoring data of the current past production time sequence to obtain the adjusted production monitoring data. Wherein, the intermittent vacancy is that the continuous past production time series of which the continuous past production time series does not appear has a vacancy in the production monitoring data, namely the range of the continuous past production time series of which the data vacancy appears is smaller than the set range. For the total liquid, the time series of the continuous production of the total liquid is a discontinuous time series of the continuous production, the discontinuous vacancy is that the vacancy is scattered, the discontinuous vacancy is opposite to the uninterrupted vacancy, and the continuous vacancy is that the vacancies appear in the production monitoring data of a plurality of continuous time series of the continuous production, namely the range of the continuous time series of the continuous production of the vacancy is larger than the set range. For the discrete liquid accumulation amount or the mining geological log, the previous production time sequence of the vacant discrete liquid accumulation amount or the mining geological log is a continuous previous production time sequence, for example, if the production monitoring data from the 2 nd previous production time sequence to the 4 th previous production time sequence are vacant in the mining geological log, namely, the mining geological log data are vacant continuously. Alternatively, if the total volume of production monitoring empty fluid in the 3 rd past production time series, the total volume of production monitoring empty fluid in the 8 th past production time series, and the total volume of production monitoring empty fluid in the 12 th past production time series, the total volume of fluid is intermittent.

If the data integrity determination result shows that the total effusion amount has intermittent gaps, the process of filling the total effusion amount with gaps comprises the following steps: based on an autoregressive moving average model (Autoregressive Integrated Moving Average Model, ARIMA), predicting and obtaining the total amount of the effusion in the production monitoring data according to the total amount of the effusion without the vacancy in the production monitoring data, and supplementing the total amount of the effusion without the vacancy so as to obtain the adjusted production monitoring data. Of course, in other embodiments, other models may be used to perform the gap prediction filling, for example, a propset model, a VAR model, an exponential smoothing model, and other time series prediction models, which are not limited in this application.

The method comprises the steps that a set time interval is the first half time interval of the dispersion liquid accumulation amount, or a random time interval, if a data integrity determination result shows that the data distribution condition of the dispersion liquid accumulation amount in the set time interval meets a set requirement, and an uninterrupted vacancy appears at the same time, the long-term regular characteristic and the short-term regular characteristic which are necessary for the deficiency of the dispersion liquid accumulation amount at the moment are represented, wherein the long-term regular characteristic refers to periodic fluctuation which is shown in a longer time range and is also called periodicity; the short-time law characteristic refers to periodic fluctuations that are exhibited by data over a fixed period of time, also known as seasonal. When the long-time regular characteristic and the short-time regular characteristic are lacked, the scattered accumulated liquid quantity filling the vacancy can be supplemented based on linear fitting (Linear Interpolation) so as to obtain production monitoring data after adjustment. The distribution condition meets the set requirement, specifically, the dispersed liquid accumulation amount does not have uninterrupted distribution, in other words, in the production monitoring data, the dispersed liquid accumulation amount is dispersed. The uninterrupted void may be an uninterrupted void in a global past production campaign or an uninterrupted void in a local past production campaign.

If the data integrity determination result indicates that the mining geological log has an uninterrupted vacancy, in other words, the production monitoring data of the uninterrupted plurality of past production time sequences has no mining geological log data, the current past production time sequence of the mining geological log with the uninterrupted vacancy can be determined in the past production time sequence, and meanwhile, the production monitoring data of the current past production time sequence are cleared, so that the adjusted production monitoring data are obtained. The reason for the direct purging is that the mined geological data in the past production time series that was purged cannot be refilled.

If the data integrity determination indicates one of the above conditions, cleaning, such as refilling or cleaning, is performed based on the corresponding cleaning mode, and if the data integrity determination indicates a plurality of the above conditions, cleaning is performed based on the corresponding plurality of modes to obtain adjusted production monitoring data.

After the production monitoring data is adjusted, normalizing the adjusted production monitoring data to a preset numerical range, obtaining transitional production monitoring data of each previous production time sequence, and normalizing the adjusted production monitoring data to the preset numerical range, for example, includes: the method comprises the steps of obtaining average numbers and deviation results of adjusted production monitoring data corresponding to a plurality of past production time sequences, respectively obtaining deviation results between the adjusted production monitoring data and the average numbers of each past production time sequence, obtaining a ratio between the deviation results and average dispersion degrees to serve as transitional production monitoring data corresponding to each past production time sequence, or extracting a maximum element (namely data with the largest value) and a minimum element (namely data with the smallest value) of data elements (namely data with the largest value) in the adjusted production monitoring data corresponding to the plurality of past production time sequences to obtain deviation results between the maximum element and the minimum element, namely a subtracted value of the maximum element and the minimum element, obtaining an extreme deviation result, obtaining a deviation result of the data elements of the adjusted production monitoring data and the minimum element, obtaining a target deviation result, obtaining a ratio of the target deviation result and the extreme deviation result, and serving as transitional production monitoring data of each past production time sequence. In this embodiment, the "transition" indicates that the corresponding data object is in an intermediate state, and is not the data of the final result, and the data object is further operated on to obtain the target object data. Average dispersion is a statistic describing the degree of dispersion of production monitoring data, which can be expressed using standard deviation (Standard Deviation) or Variance (Variance).

Specifically, the process of normalizing the adjusted production monitoring data based on the average number and the average dispersion can be implemented by the following formula:

equation one: s is S _n1 =(S _n -S _m )/S _k

Wherein S is _n1 Transitional production monitoring data for the nth previous production time sequence, S _n Production monitoring data after adjustment for nth previous production time series, S _m For an average of the adjusted production monitoring data of a plurality of past production time sequences, S _k Average dispersion of the production monitoring data for a plurality of past production time series.

The process of normalizing the adjusted production monitoring data based on the extremum of the adjusted production monitoring data may be implemented by the following formula two:

formula II: s is S _n1 =(S _n -S _a )/(S _b -S _a )

Wherein S is _n1 Transitional production monitoring data for the nth previous production time sequence, S _n Production monitoring data after adjustment for nth previous production time series, S _b For the values of the largest elements in the adjusted production monitoring data of a plurality of past production time sequences, S _a The values of the very small elements in the production monitoring data are adjusted for a plurality of past production time sequences.

Based on normalization, some characteristic data can be prevented from being ignored due to the fact that the numerical value is too small, network acceleration debugging is facilitated, and prediction accuracy is improved.

And normalizing the regulated production monitoring data to a preset numerical range, performing trend decomposition (Detrending) on the normalized transitional production monitoring data, and removing trend components in the time series data to obtain target production monitoring data of each past production time series. The process of trend decomposition of the transitional production monitoring data comprises the following steps: and determining one or more transitional production monitoring data in a transitional time period from the transitional production monitoring data, obtaining a transitional production monitoring data set, obtaining the average number of data elements of the transitional production monitoring data in the transitional production monitoring data set, obtaining the average number corresponding to the transitional time period, and obtaining the deviation result of the transitional production monitoring data and the average number as target production monitoring data of each past production time sequence. For example, assuming that the transition time period is 12h, a moving average (SMA) is used to obtain a moving average result (i.e., an average number of data elements) with the 12h period, and then the moving average result is subtracted from the transition production monitoring data to obtain target production monitoring data of each previous production time sequence, where the target production monitoring data reciprocally changes around the moving average result.

Step S122, extracting a production description vector in the past production activity from the target production monitoring data based on the maximum value prediction branch network of the maximum fluid accumulation prediction neural network, and predicting to obtain one or more gas well fluid accumulation of the subsequent production time sequence according to the production description vector.

In an embodiment of the application, the maximum fluid production prediction neural network comprises a maximum fluid production prediction branch network and a short-time law adjustment branch network, wherein the maximum fluid production prediction branch network is configured to predict the fluid production of a natural gas well of each subsequent production time sequence. The short term regulation branch network is configured to capture long term and short term changes that are ignored in predicting the gas well fluid production to obtain a predicted loss of gas well fluid production.

The extraction description vector is a feature vector for vector representation of the integrated liquid related feature, and in the embodiment of the present application, the process of extracting the extraction description vector in the past production activity from the target production monitoring data based on the maximum value prediction branch network of the maximum integrated liquid amount prediction neural network is as follows:

determining target production monitoring data of a target past production time sequence from the target production monitoring data, carrying out description vector mining (namely, completing feature extraction) on the target production monitoring data of the target past production time sequence based on a maximum value prediction branch network of a maximum accumulated liquid prediction neural network to obtain mining description vectors of the target past production time sequence, wherein the mining description vectors comprise accumulated liquid description vectors and gate description vectors, regulating the accumulated liquid description vectors through the gate description vectors so as to obtain target accumulated liquid description vectors corresponding to past production activities, and determining and obtaining one or more natural gas well accumulated liquid of the subsequent production time sequence according to the target accumulated liquid description vectors. Wherein the gate description vector is a memory feature. For example, in a long-short-term memory neural network, a memory cell such as a memory unit stores past information, and updates memory based on the state of the memory cell at the current input and the last moment is introduced to store long-term information of sequence data, and the storage and forgetting of information is controlled through an input gate, a forgetting gate, and an output gate.

In the above process, the target past production time sequence is, for example, the past production time sequence that is the longest from the current time, or another set past production time sequence. As noted above, the structure of the max predicted branch network includes memory cells, then the max predicted branch network is a recurrent neural network or a long-short term memory network. If the maximum predictive branch network is a long-short-time memory network, feature extraction can be completed by carrying out description vector mining on target production monitoring data of a target past production time sequence based on the long-short-time memory network to obtain a liquid accumulation volume description vector and a gate description vector of the target past production time sequence, then the target production monitoring data of a next past production time sequence (which can be called a transition past production time sequence) of the target past production time sequence is determined in the target production monitoring data, then the liquid accumulation volume description vector and the gate description vector of the transition past production time sequence are extracted from the target production monitoring data of the transition past production time sequence according to the gate description vector, the updated liquid accumulation volume description vector of the transition past production time sequence is obtained, the transition past production time sequence is determined as the target past production time sequence, the step of returning to the target production monitoring data of the next past production time sequence (the transition past production time sequence) of the target past production time sequence is determined in the target production monitoring data is carried out, the liquid accumulation volume description vector is returned when each past production time sequence is the target past production time sequence, the liquid accumulation volume description vector is obtained corresponding to the target past production time sequence, and a liquid accumulation volume description vector is determined according to the target past production time sequence or a plurality of liquid accumulation volume description.

In a feasible design, before predicting the accumulated liquid of the natural gas well in one or more subsequent production time sequences corresponding to the past production time sequences according to the target production monitoring data, the prediction method for automatically adding medicament and draining the natural gas well provided by the embodiment of the application further comprises a neural network debugging step, and specifically comprises the following steps: the method comprises the steps of obtaining a past production monitoring training data set of a target natural gas well, wherein the past production monitoring training data set comprises a plurality of production monitoring training data of past training production time sequences, debugging a preset maximum value prediction branch network based on the production monitoring training data, determining training data loss corresponding to the production monitoring training data through the debugged maximum value prediction branch network, debugging a preset short-time regulation branch network according to the training data loss, obtaining a debugged short-time regulation branch network, and generating a maximum liquid accumulation amount prediction neural network through the debugged maximum value prediction branch network and the debugged short-time regulation branch network.

Specifically, the method comprises the following steps:

And step S1, acquiring a past production monitoring training data set of the target natural gas well.

The past production monitoring training data set comprises a plurality of past production monitoring training data of a past training production time sequence. The process of acquiring the past production monitoring training data set of the target natural gas well is the same as the process of acquiring the past production monitoring data set of the target natural gas well.

And S2, debugging a preset maximum value prediction branch network based on the production monitoring training data, and determining the training data loss corresponding to the production monitoring training data through the debugged maximum value prediction branch network.

The process of debugging the preset maximum value prediction branch network based on the production monitoring training data includes, for example: and carrying out data integrity determination on the production monitoring training data, carrying out data cleaning on the production monitoring training data according to the result of the data integrity determination to obtain target production monitoring training data of each past training production time sequence, dividing the target production monitoring training data into a Support Set (Support Set), a development Set (Validation Set) and a Test Set (Test Set) by a preset percentage, and debugging the maximum value prediction branch network according to the Support Set to obtain a debugged maximum value prediction branch network. The process of data integrity determination for production monitoring training data is the same as the process of data integrity determination for production monitoring data. And the process of carrying out data cleaning on the production monitoring training data according to the training data integrity determination result is the same as the process of carrying out data cleaning on the production monitoring data.

In addition, it can be understood that after the production monitoring training data is subjected to data cleaning, the cleaned target production monitoring training data can be divided into a support set, a development set and a test set according to a preset percentage, and the preset percentage is not limited, for example, the support set is 50%, the development set is 25% and the test set is 25%.

After the target production monitoring training data are divided into a support set, a development set and a test set, the preset maximum value prediction branch network is debugged according to the support set, so that the maximum value prediction branch network after debugging is completed is obtained. The process of debugging the preset maximum value prediction branch network according to the support set includes, for example: according to the support set, predicting the natural gas well liquid accumulation amount of the subsequent production time sequence corresponding to the support set based on the preset maximum value prediction branch network, obtaining the current predicted natural gas well liquid accumulation amount, extracting the natural gas well liquid accumulation amount label of the subsequent production time sequence in the support set, comparing the current predicted natural gas well liquid accumulation amount with the natural gas well liquid accumulation amount label, obtaining a prediction error corresponding to the support set, debugging the preset maximum value prediction branch network according to the prediction error (for example, adjusting network parameters in the error reducing direction based on a gradient optimization algorithm), and obtaining the maximum value prediction branch network after debugging.

When the preset maximum value prediction branch network is a long-short-time memory neural network, in the debugging process, the association between the execution data and the output data is learned. In the process of predicting the total liquid accumulation amount, each execution data is data in a past production time sequence, namely SN (n-a, n), a is the number of the past production time sequences, each output data is a numerical value in the same past production time sequence, namely SP (n+1, n+b), and b is the number of predicted subsequent production time sequences and can be 1. One time n corresponds to one execution data/output data group, and the sequence with the length of D is divided into D-a-b execution data/output data groups.

After the preset maximum value prediction branch network is debugged according to the support set, determining the training data loss corresponding to the production monitoring training data according to the support set and the development set based on the debugged maximum value prediction branch network. The training data loss is the deviation result of the natural gas well effusion quantity corresponding to the predicted support set and the development set and the actual natural gas well effusion quantity. The process of determining the loss of training data corresponding to the production monitoring training data according to the support set and the development set comprises the following steps: and predicting the natural gas well effusion amount corresponding to the support set and the development set respectively based on the maximum value of the debugging completion, acquiring a natural gas well effusion amount label in the production monitoring training data, comparing the predicted natural gas well effusion amount with the corresponding natural gas well effusion amount label to acquire the debugging loss of the support set and the development loss of the development set, and taking the debugging loss and the development loss as the training data loss corresponding to the production monitoring training data.

And step S3, debugging the preset short-time law adjustment branch network according to the training data loss to obtain a debugged short-time law adjustment branch network.

For example, the loss of the predicted natural gas well effusion is predicted based on the preset short-time law regulating branch network, the predicted loss is obtained, the predicted loss is compared with the training data loss to obtain the loss predicted loss (namely, the loss generated by the predicted loss), and the preset short-time law regulating branch network is debugged according to the loss predicted loss to obtain the debugged short-time law regulating branch network. The architecture of the short-time law regulating branch network is, for example, a Prophet model.

And S4, generating a maximum liquid accumulation amount prediction neural network through the maximum value prediction branch network after debugging and the short-time regulation branch network after debugging.

For example, the maximum value prediction branch network after debugging and the short-time regulation branch network after debugging are determined to be the transition maximum liquid accumulation prediction neural network, the transition maximum liquid accumulation prediction neural network is tested based on a test set to obtain a test value, and if the test value is smaller than a set test value, the transition maximum liquid accumulation prediction neural network is determined to be the maximum liquid accumulation prediction neural network. The process of testing the transient maximum liquid accumulation amount prediction neural network based on the test set comprises the following steps: based on the transitional maximum fluid accumulation prediction neural network, the fluid accumulation of the natural gas well corresponding to the test set is obtained, the transitional predicted fluid accumulation of the natural gas well is obtained, the target fluid accumulation label of the natural gas well corresponding to the test set is obtained, the second power of the deviation result between the transitional predicted fluid accumulation of the natural gas well and the target fluid accumulation label of each production monitoring training data in the test set is respectively obtained, the fluid accumulation square difference is obtained, and the average of the fluid accumulation square differences is obtained, so that the test value of the transitional maximum fluid accumulation prediction neural network is obtained. The process of predicting the gas well liquid accumulation amount corresponding to the test set based on the transient maximum liquid accumulation amount prediction neural network includes: predicting the branch network based on the maximum value of the debugging to obtain the natural gas well liquid accumulation amount of each production monitoring training data in the test set, obtaining the transitional natural gas well liquid accumulation amount, adjusting the branch network based on the short-time rule of the debugging to obtain the loss of the transitional natural gas well liquid accumulation amount, obtaining the predicted loss, and adjusting the transitional natural gas well liquid accumulation amount according to the predicted loss to obtain the transitional predicted natural gas well liquid accumulation amount of each production monitoring training data in the test set.

The process of obtaining the average of the squared differences of the liquid volumes includes, for example: the average of the squared differences of the liquid accumulation amounts is obtained, and square root is calculated on the average to obtain a test value of the transient maximum liquid accumulation amount prediction neural network, and the test value can be specifically realized according to the following formula III:

and (3) a formula III:

wherein, T is the test value, the smaller the T, the smaller the difference between the predicted result and the real value of the network, rx is the transitional predicted natural gas well effusion of the xth production monitoring training data in the test set, qx is the target natural gas well effusion label of the xth production monitoring training data in the test set, and m is the quantity of the production monitoring training data in the test set.

After the transient natural gas well effusion prediction neural network is tested, if the test value is smaller than the set test value, the error of the predicted transient predicted natural gas well effusion and the actual target natural gas well effusion label is in an expected interval, and the transient maximum effusion prediction neural network is determined to be the maximum effusion prediction neural network.

Step S130, determining the loss of the accumulated liquid of the natural gas well in the production monitoring data.

The losses herein may also be referred to as residuals, which are used to indicate ignored long term and short term changes in predicting natural gas well fluid production. The process of determining the loss of fluid production from a gas well in production monitoring data includes, for example: and determining short-time rule data in the target production monitoring data, adjusting a branch network to perform adjacent difference calculation on the target production monitoring data based on the short-time rule of the maximum fluid accumulation prediction neural network to obtain adjacent difference data (namely, performing differential calculation to obtain a differential result), and determining the loss of the fluid accumulation of the natural gas well in each subsequent production time sequence in the adjacent difference data according to the short-time rule data. The short-time rule data is sequence data corresponding to short-time variation, for example, week data in the sequence data, and the short-time rule data may be 7. If the short-time law regulating branch network is a Prophet model, the loss of the accumulated liquid of the natural gas well can be predicted based on the Prophet model.

And step S140, adjusting the natural gas well liquid accumulation amount according to the loss, and obtaining the target natural gas well liquid accumulation amount of each subsequent production time sequence.

For example, determining a target loss corresponding to each natural gas well fluid production volume from the losses, and summing the target loss with the corresponding natural gas well fluid production volume to obtain a target natural gas well fluid production volume for each subsequent production time series. Wherein the process of summing the target loss with the corresponding natural gas well liquid product amount includes, for example: summing the target loss and the corresponding gas well fluid volumes to obtain a target gas well fluid volume for each subsequent production time series, or determining a weighting value, weighting the target loss and the corresponding gas well fluid volumes according to the weighting value, and summing the weighted target loss and gas well fluid volumes to obtain a target gas well fluid volume for each subsequent production time series.

Assuming that the maximum liquid accumulation amount prediction neural network comprises a long-short-time memory neural network and a Prophet model, the prediction accuracy of the long-short-time memory neural network is enhanced based on the Prophet model. In the implementation process, the long-short-term memory neural network is obtained by debugging on the support set, and the prediction result of the long-short-term memory neural network on the support set and the deviation result between the real numerical values of the support set are regarded as debugging loss; and taking the deviation result between the prediction result of the long-and-short-term memory neural network on the development set and the real numerical value of the development set as the development loss. Determining the debugging loss as a support set, determining the development loss as a development set, loading the support set into a Prophet model to obtain a prediction loss corresponding to the support set, debugging the Prophet model until convergence through the prediction loss and the labeling loss of the support set, performing trafficability verification on the debugged Prophet model through the development set, and determining the Prophet model as the debugged Prophet model after the trafficability verification is completed. Predicting the predicted natural gas well effusion of the test set based on the long-short-time memory neural network after debugging, and predicting the predicted natural gas well effusion loss based on the long-short-time memory neural network after debugging. And summing the predicted loss and the corresponding predicted natural gas well accumulated liquid amount to finish the adjustment of the predicted natural gas well accumulated liquid amount predicted by the long-short-time memory neural network so as to obtain more accurate predicted natural gas well accumulated liquid amount and increase the prediction accuracy of the long-short-time memory neural network.

And step S150, determining and obtaining the maximum accumulated liquid amount in the target natural gas well accumulated liquid amount.

The maximum fluid production may be the maximum of the fluid production of the natural gas well in one or more subsequent production time sequences, and the process of determining the maximum fluid production at the target natural gas well fluid production may include, for example: and determining the maximum element in the target natural gas well liquid accumulation volume, and determining the maximum element as the maximum liquid accumulation volume, or dividing the subsequent production time sequence into a plurality of maximum value intervals, and determining the maximum target natural gas well liquid accumulation volume in the target natural gas well liquid accumulation volume corresponding to each maximum value interval to obtain the maximum liquid accumulation volume of each maximum interval. The minimum dividing unit of the maximum section is not limited, and in one example, the minimum dividing unit of the subsequent production time series is 1h, and the maximum section is 2h, 4h, 8h, or the like, for example. If the maximum interval is 24h, for example, the 12 subsequent production time sequences are divided into a maximum interval, namely, the maximum liquid accumulation amount in 12h is determined.

Optionally, after determining the maximum fluid accumulation in the target fluid accumulation of the natural gas well, the embodiment of the application may determine a subsequent production time sequence in which the maximum fluid accumulation is obtained in a subsequent production activity, so as to obtain a target subsequent production time sequence corresponding to the maximum fluid accumulation. After the maximum accumulated liquid is predicted, proper medicine amount can be reserved according to the predicted maximum accumulated liquid, and on the premise of ensuring the high efficiency of accumulated liquid cleaning, the medicine input amount is reduced, the resources are saved, and the cost is reduced.

In summary, after obtaining production monitoring data of a plurality of past production time sequences of a target natural gas well in past production activities, predicting the natural gas well liquid accumulation of one or more subsequent production time sequences in subsequent production activities after the past production activities through the production monitoring data, determining the loss of the obtained natural gas well liquid accumulation in the production monitoring data, adjusting the natural gas well liquid accumulation according to the loss to obtain the target natural gas well liquid accumulation of each subsequent production time sequence, and determining the maximum liquid accumulation in the target natural gas well liquid accumulation; the process can be used for predicting the liquid accumulation amount of the natural gas well to obtain the subsequent production time sequence, and determining the loss of the liquid accumulation amount of the natural gas well, meanwhile, the loss can be considered to be obtained based on capturing long-time regular changes and short-time regular changes of production monitoring data, so that the predicted target liquid accumulation amount of the natural gas well can not only bring complex long-time macroscopic characteristics into analysis, but also capture local changes such as short-time regular changes, and further, interference between the predicted liquid accumulation amount of the natural gas well and different neural networks of the loss can be avoided based on loss regulation, and the accuracy of the maximum liquid accumulation amount prediction can be improved.

The method provided in the embodiment of the present application is described in detail below based on the above description.

In the embodiment of the application, an execution main body of the method is a prediction system, the execution main body is computer equipment, a production time sequence is set to be 1h, a maximum value prediction branch network of a maximum liquid accumulation amount prediction neural network is a long-short-time memory neural network, and a short-time law regulation branch network of the maximum liquid accumulation amount prediction neural network is a Prophet model.

Firstly, the debugging process of the maximum liquid accumulation amount prediction neural network comprises the following steps:

1. the computer device obtains a past production monitoring training dataset of the target natural gas well.

2. The computer equipment is used for debugging the preset long-time memory neural network based on the production monitoring training data, and determining the training data loss corresponding to the production monitoring training data through the long-time memory neural network after the debugging is completed.

For example, the computer device performs data integrity determination on the production monitoring training data, and performs data cleaning on the production monitoring training data according to the result of the determination of the integrity of the training data, so as to obtain target production monitoring training data of each past training production time sequence. The target production monitoring training data after the data cleaning is completed are divided into 50% supporting sets, 25% developing sets and 25% testing sets. The computer equipment predicts the natural gas well liquid accumulation amount of the subsequent production time sequence corresponding to the support set based on the preset long-short time memory neural network according to the support set, obtains the current predicted natural gas well liquid accumulation amount, extracts the natural gas well liquid accumulation amount label of the subsequent production time sequence in the support set, compares the current predicted natural gas well liquid accumulation amount with the natural gas well liquid accumulation amount label to obtain the target prediction loss corresponding to the support set, and debugs the preset long-short time memory neural network according to the loss until reaching convergence conditions (such as the number of times of debugging reaches the preset number of times, the loss reaches the minimum, and the like) so as to obtain the long-short time memory neural network after the debugging is completed.

The method comprises the steps that based on the natural gas well effusion amount respectively corresponding to a neural network prediction support set and a development set after debugging, the computer equipment obtains a natural gas well effusion amount label in production monitoring training data, compares the predicted natural gas well effusion amount with the corresponding natural gas well effusion label to obtain debugging loss of the support set and development loss of the development set, and determines the debugging loss and the development loss as training data loss corresponding to the production monitoring training data.

3. The computer equipment debugs the preset propset model according to the training data loss to obtain a debugged propset model.

The computer device may predict a loss of predicted natural gas well fluid production based on a preset propset model, obtain a predicted loss, and compare the predicted loss with training data loss to obtain a loss predicted loss, debug the preset propset model according to the loss predicted loss to obtain a debugged propset model.

4. And the computer equipment generates the maximum liquid accumulation amount prediction neural network through the debugged long-short-term memory neural network and the debugged Prophet model.

The method comprises the steps that a computer device takes a debugged long-short-term memory neural network and a debugged Prophet model as a transitional maximum accumulated liquid prediction neural network, natural gas well accumulated liquid of each production monitoring training data in a test set is obtained based on the debugged long-short-term memory neural network, transitional natural gas well accumulated liquid is obtained, loss of the transitional natural gas well accumulated liquid is obtained based on the debugged Prophet model, prediction loss is obtained, and the transitional natural gas well accumulated liquid is adjusted according to the prediction loss so as to obtain transitional predicted natural gas well accumulated liquid of each production monitoring training data in the test set. The method comprises the steps that a computer device obtains a target natural gas well effusion quantity label corresponding to a support set, and obtains the second power of a deviation result between transitional prediction natural gas well effusion quantity of each production monitoring training data in a test set and the target natural gas well effusion quantity label respectively, so that the effusion quantity square error is obtained. The mean of the squared differences of the liquid volumes is obtained, and square root is taken on the mean to obtain the test value of the transient maximum liquid volume prediction neural network, and the formula III can be referred to. And if the test value is smaller than the set test value, determining the transitional maximum liquid accumulation prediction neural network as the maximum liquid accumulation prediction neural network.

The computer device then predicts a maximum fluid production volume for one or more subsequent production time series of the target natural gas well based on the debugged maximum fluid production prediction neural network to obtain a maximum fluid production volume. The maximum liquid accumulation prediction neural network may include a long and short term memory neural network and a Prophet model.

The prediction method for automatic drug delivery and drainage of the natural gas well specifically comprises the following steps:

in step S210, the computer device obtains past production monitoring data sets for the target natural gas well.

The method comprises the steps of acquiring a liquid accumulation log and a past exploitation geological log of a target natural gas well in past production activities, determining and obtaining one or more liquid accumulation logs (total liquid accumulation and scattered liquid accumulation) of past production time sequences in the liquid accumulation log, determining and obtaining exploitation geological logs (natural gas yield, stratum permeability, bottom hole pressure, exploitation time and well depth) of the past production time sequences in the past exploitation geological log, determining sequence data (week, day and month) corresponding to each past production time sequence, and taking the liquid accumulation logs, the exploitation geological logs and the sequence data as a past production monitoring data set of the target natural gas well.

In step S220, the computer device performs data integrity determination on the production monitoring data, and obtains a data integrity determination result.

For example, the computer device determines the data integrity of the fluid accumulation volume log and the production geological log in the production monitoring data to obtain a data integrity determination result of the production monitoring data, or determines whether a void occurs in the fluid accumulation volume log and the production geological log in the production monitoring data to obtain a data integrity determination result of the production monitoring data.

Step S230, according to the data integrity determination result, the production monitoring data is subjected to data cleaning to obtain target production monitoring data of each previous production time sequence in the previous production activities.

And if the data integrity determination result shows that the total amount of the effusion is in a discontinuous vacancy, the computer equipment supplements the total amount of the effusion in the vacancy to obtain adjusted production monitoring data, if the data integrity determination result shows that the distribution condition of the scattered effusion (scattered effusion) in a set time interval meets the set requirement, and meanwhile, when the continuous vacancy is in the same time, the computer equipment supplements the missing scattered effusion in the production monitoring data based on linear fitting to obtain the adjusted production monitoring data, and if the data integrity determination result shows that the continuous vacancy is in the exploitation geological log, the current past production time sequence of the continuous vacancy is determined in the past production time sequence, and meanwhile, the production monitoring data of the current past production time sequence is cleared to obtain the adjusted production monitoring data. The computer equipment obtains the average number and deviation results of the adjusted production monitoring data corresponding to the plurality of past production time sequences, respectively obtains the deviation results between the adjusted production monitoring data and the average number of each past production time sequence, obtains the proportion between the deviation results and the average dispersity, and is used as the transitional production monitoring data corresponding to each past production time sequence, or extracts the maximum element and the minimum element of the data element from the adjusted production monitoring data corresponding to the plurality of past production time sequences, obtains the deviation results between the maximum element and the minimum element, obtains the extreme value deviation results, and obtains the deviation results of the data element and the minimum element of the adjusted production monitoring data, obtains the target deviation results, obtains the proportion between the target deviation results and the extreme value deviation results, and is used as the transitional production monitoring data of each past production time sequence.

In step S240, the computer device extracts the production description vector in the past production activity from the target production monitoring data based on the long-short-term memory neural network of the maximum fluid accumulation prediction neural network, and predicts the fluid accumulation of the natural gas well according to the production description vector, and one or more subsequent production time sequences.

The computer equipment can determine the target production monitoring data of the target past production time sequence from the target production monitoring data, perform description vector mining on the target production monitoring data of the target past production time sequence based on the long-short-time memory neural network to obtain a mining description vector of the target past production time sequence, wherein the mining description vector comprises a liquid accumulation volume description vector and a gate description vector, then determine the target production monitoring data of the next past production time sequence (transition past production time sequence) of the target past production time sequence from the target production monitoring data, then extract the liquid accumulation volume description vector and the gate description vector of the target production monitoring data of the transition past production time sequence to obtain an updated liquid accumulation volume description vector of the transition past production time sequence, and return to execute the determination of the target production monitoring data of the next past production time sequence (transition past production time sequence) of the target past production time sequence from the target production monitoring data until each of the target past production time sequence is the target past production time sequence, obtain a corresponding liquid accumulation volume description vector of the target past production time sequence, and the gas well.

In step S250, the computer device determines a loss of natural gas well fluid production from the production monitoring data.

The computer equipment can determine and obtain short-time rule data in the target production monitoring data, and perform adjacent difference calculation on the target production monitoring data based on a Prophet model of the maximum liquid accumulation prediction neural network to obtain adjacent difference data, and determine and obtain the loss of the liquid accumulation of the natural gas well in each subsequent production time sequence in the adjacent difference data according to the short-time rule data.

In step S260, the computer device adjusts the gas well fluid production based on the losses to obtain a target gas well fluid production for each subsequent production time series.

The computer device may determine a target loss corresponding to each natural gas well fluid production volume from the losses, sum the target loss with the corresponding natural gas well fluid production volume to obtain a target natural gas well fluid production volume for each subsequent production time series, or weight the target loss with the corresponding natural gas well fluid production volume based on a weight set in advance, sum the weighted target loss and the natural gas well fluid production volume to obtain a target natural gas well fluid production volume for each subsequent production time series.

In step S270, the computer device determines a maximum fluid production from the target gas well fluid production.

The computer device may determine a maximum element in the target gas well fluid production volume and determine the maximum element as a maximum fluid production volume, or divide the subsequent production time series into a plurality of maximum intervals, and determine a maximum target gas well fluid production volume in the target gas well fluid production volume corresponding to each maximum interval to obtain a maximum fluid production volume for each maximum interval.

Optionally, the computer device determines a subsequent production time sequence in which the maximum fluid accumulation is obtained in the subsequent production activity, so as to obtain a target subsequent production time sequence corresponding to the maximum fluid accumulation.

In summary, after obtaining production monitoring data of a plurality of past production time sequences of a target natural gas well in a past production activity, the computer equipment predicts the natural gas well liquid accumulation of one or more of the following production time sequences in the following production activity after the past production activity through the production monitoring data, determines the loss of the obtained natural gas well liquid accumulation in the production monitoring data, adjusts the natural gas well liquid accumulation according to the loss, obtains the target natural gas well liquid accumulation of each of the following production time sequences, and then determines the maximum liquid accumulation in the target natural gas well liquid accumulation; the process can be used for predicting the liquid accumulation amount of the natural gas well to obtain the subsequent production time sequence, and determining the loss of the liquid accumulation amount of the natural gas well, meanwhile, the loss can be considered to be obtained based on capturing long-time regular changes and short-time regular changes of production monitoring data, so that the predicted target liquid accumulation amount of the natural gas well can not only bring complex long-time macroscopic characteristics into analysis, but also capture local changes such as short-time regular changes, and further, interference between the predicted liquid accumulation amount of the natural gas well and different neural networks of the loss can be avoided based on loss regulation, and the accuracy of the maximum liquid accumulation amount prediction can be improved.

Based on the foregoing embodiments, the embodiments of the present application provide a prediction apparatus, where each unit included in the apparatus, and each module included in each unit may be implemented by a processor in a computer device; of course, the method can also be realized by a specific logic circuit; in practice, the processor may be a central processing unit (Central Processing Unit, CPU), microprocessor (Microprocessor Unit, MPU), digital signal processor (Digital Signal Processor, DSP) or field programmable gate array (Field Programmable Gate Array, FPGA), etc.

Fig. 2 is a schematic structural diagram of a prediction apparatus according to an embodiment of the present application, and as shown in fig. 2, a prediction apparatus 200 includes:

a data acquisition module 210 for acquiring a past production monitoring dataset of a target natural gas well, the past production monitoring dataset comprising a plurality of past production time-series production monitoring data in a past production campaign;

a hydrops prediction module 220 for predicting, from the production monitoring data, a natural gas well hydrops in a subsequent production campaign including one or more subsequent production time sequences following the past production campaign;

A loss determination module 230 for determining a loss of the gas well fluid production volume from the production monitoring data, the loss being indicative of long term regular changes and short term regular changes that are ignored when predicting the gas well fluid production volume;

a hydrops adjustment module 240, configured to adjust the natural gas well hydrops according to the loss, to obtain a target natural gas well hydrops for each subsequent production time sequence;

the effusion determination module 250 is used for determining the maximum effusion amount in the target natural gas well effusion amount.

The description of the apparatus embodiments above is similar to that of the method embodiments above, with similar advantageous effects as the method embodiments. In some embodiments, functions or modules included in the apparatus provided in the embodiments of the present application may be used to perform the methods described in the embodiments of the methods, and for technical details that are not disclosed in the embodiments of the apparatus of the present application, please refer to the description of the embodiments of the methods of the present application for understanding.

It should be noted that, in the embodiment of the present application, if the above prediction method for automatically delivering the chemical and draining the water in the natural gas well is implemented in the form of a software functional module, and sold or used as an independent product, the prediction method may also be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application may be essentially or portions contributing to the related art, and the software product may be stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read Only Memory (ROM), a magnetic disk, an optical disk, or other various media capable of storing program codes. Thus, embodiments of the present application are not limited to any specific hardware, software, or firmware, or to any combination of hardware, software, and firmware.

The embodiment of the application provides a prediction system, which comprises a memory and a processor, wherein the memory stores a computer program capable of running on the processor, and the processor executes the program to realize part or all of the steps of the method. Embodiments of the present application provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs some or all of the steps of the above-described method. The computer readable storage medium may be transitory or non-transitory. Embodiments of the present application provide a computer program comprising computer readable code which, when run in a computer device, performs some or all of the steps for implementing the above method. Embodiments of the present application provide a computer program product comprising a non-transitory computer-readable storage medium storing a computer program which, when read and executed by a computer, performs some or all of the steps of the above-described method. The computer program product may be realized in particular by means of hardware, software or a combination thereof. In some embodiments, the computer program product is embodied as a computer storage medium, in other embodiments the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), or the like.

It should be noted here that: the above description of various embodiments is intended to emphasize the differences between the various embodiments, the same or similar features being referred to each other. The above description of apparatus, storage medium, computer program and computer program product embodiments is similar to that of method embodiments described above, with similar advantageous effects as the method embodiments. For technical details not disclosed in the embodiments of the apparatus, storage medium, computer program and computer program product of the present application, please refer to the description of the method embodiments of the present application.

Fig. 3 is a schematic hardware entity diagram of a prediction system according to an embodiment of the present application, as shown in fig. 3, the hardware entity of the prediction system 1000 includes: a processor 1001 and a memory 1002, wherein the memory 1002 stores a computer program executable on the processor 1001, the processor 1001 implementing the steps in the method of any of the embodiments described above when the program is executed.

The memory 1002 stores a computer program executable on the processor, and the memory 1002 is configured to store instructions and applications executable by the processor 1001, and may also cache data (e.g., image data, audio data, voice communication data, and video communication data) to be processed or already processed by each module in the processor 1001 and the prediction system 1000, which may be implemented by a FLASH memory (FLASH) or a random access memory (Random Access Memory, RAM). The processor 1001 executes a program to implement the method for predicting automatic chemical delivery and drainage in a natural gas well according to any one of the above. The processor 1001 generally controls the overall steps of the prediction system 1000. Embodiments of the present application provide a computer storage medium storing one or more programs executable by one or more processors to implement the steps of the method for predicting gas well automatic drug administration and drainage of any of the embodiments above.

It should be noted here that: the description of the storage medium and apparatus embodiments above is similar to that of the method embodiments described above, with similar benefits as the method embodiments. For technical details not disclosed in the embodiments of the storage medium and the apparatus of the present application, please refer to the description of the method embodiments of the present application for understanding. The processor may be at least one of a target application integrated circuit (Application Specific Integrated Circuit, ASIC), a digital signal processor (Digital Signal Processor, DSP), a digital signal processing device (Digital Signal Processing Device, DSPD), a programmable logic device (Programmable Logic Device, PLD), a field programmable gate array (Field Programmable Gate Array, FPGA), a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, and a microprocessor. It will be appreciated that the electronic device implementing the above-mentioned processor function may be other, and embodiments of the present application are not specifically limited. The computer storage medium/Memory may be a Read Only Memory (ROM), a programmable Read Only Memory (Programmable Read-Only Memory, PROM), an erasable programmable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), an electrically erasable programmable Read Only Memory (Electrically Erasable Programmable Read-Only Memory, EEPROM), a magnetic random access Memory (Ferromagnetic Random Access Memory, FRAM), a Flash Memory (Flash Memory), a magnetic surface Memory, an optical disk, or a Read Only optical disk (Compact Disc Read-Only Memory, CD-ROM); but may also be various terminals such as mobile phones, computers, tablet devices, personal digital assistants, etc., that include one or any combination of the above-mentioned memories.

It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. It should be understood that, in various embodiments of the present application, the sequence number of each step/process described above does not mean that the execution sequence of each step/process should be determined by the function and the internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application. The foregoing embodiment numbers of the present application are merely for describing, and do not represent advantages or disadvantages of the embodiments. It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other ways. The above described device embodiments are only illustrative, e.g. the division of the units is only one logical function division, and there may be other divisions in practice, such as: multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. In addition, the various components shown or discussed may be coupled or directly coupled or communicatively coupled to each other via some interface, whether indirectly coupled or communicatively coupled to devices or units, whether electrically, mechanically, or otherwise.

The units described above as separate components may or may not be physically separate, and components shown as units may or may not be physical units; can be located in one place or distributed to a plurality of network units; some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment. In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may be separately used as one unit, or two or more units may be integrated in one unit; the integrated units may be implemented in hardware or in hardware plus software functional units. Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, and the foregoing program may be stored in a computer readable storage medium, where the program, when executed, performs steps including the above method embodiments; and the aforementioned storage medium includes: a mobile storage device, a Read Only Memory (ROM), a magnetic disk or an optical disk, or the like, which can store program codes.

Alternatively, the integrated units described above may be stored in a computer readable storage medium if implemented in the form of software functional modules and sold or used as a stand-alone product. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the related art in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a removable storage device, a ROM, a magnetic disk, or an optical disk. The foregoing is merely an embodiment of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think about changes or substitutions within the technical scope of the present application, and the changes and substitutions are intended to be covered in the protection scope of the present application.

Claims

1. A method for predicting drainage of an automatic dosing agent of a natural gas well, applied to a prediction system, the method comprising:

2. The method for predicting automated chemical delivery drainage for a gas well of claim 1, wherein predicting gas well liquid production in a subsequent production campaign from the production monitoring data comprises:

3. The method for predicting automated chemical delivery drainage for a gas well of claim 2, wherein the performing data cleansing on the production monitoring data based on the data integrity determination results to obtain target production monitoring data for each past production time sequence in the past production campaign comprises:

4. A method of predicting automated chemical delivery drainage for a gas well according to claim 3, wherein the production monitoring data comprises a liquid accumulation log and a production geological log, the liquid accumulation log comprising a total amount of liquid accumulation and a dispersed liquid accumulation, the production monitoring data being adjusted according to a data integrity determination result to obtain adjusted production monitoring data for each past production time sequence, comprising:

5. The method for predicting automated chemical delivery drainage for a natural gas well according to claim 3 or 4, wherein normalizing the adjusted production monitoring data to a preset value range to obtain transitional production monitoring data for each past production time sequence comprises:

6. The method for predicting automatic chemical delivery and drainage of a gas well according to claim 2, wherein the maximum value prediction branch network based on the maximum liquid accumulation prediction neural network extracts a production description vector in the past production activity from the target production monitoring data, predicts one or more gas well liquid accumulation volumes of subsequent production time sequences according to the production description vector, and comprises:

7. The method for predicting automated chemical delivery drainage for a gas well of claim 6, wherein the maximum value prediction branch network based on the maximum liquid accumulation prediction neural network extracts a production description vector in the past production activity from the target production monitoring data, and predicts one or more subsequent production time series of liquid accumulation for the gas well according to the production description vector, and further comprising:

8. The method for predicting automated chemical delivery drainage for a gas well of claim 7, wherein the predicting a branch network based on a maximum value of a completion of the debugging, determining a training data loss corresponding to the production monitoring training data according to the support set and the development set, comprises:

9. The method for predicting automated chemical delivery drainage for a natural gas well of claim 1, wherein said adjusting said natural gas well volumetric capacity based on said losses to obtain a target natural gas well volumetric capacity for each subsequent production time series comprises:

10. A prediction system comprising a memory and a processor, the memory storing a computer program executable on the processor, wherein the processor performs the steps of the method of any one of claims 1 to 9 when the program is executed.