CN109784539A - The determining method and apparatus of operation are adjusted for the training of index prediction model and well group - Google Patents
The determining method and apparatus of operation are adjusted for the training of index prediction model and well group Download PDFInfo
- Publication number
- CN109784539A CN109784539A CN201811544048.3A CN201811544048A CN109784539A CN 109784539 A CN109784539 A CN 109784539A CN 201811544048 A CN201811544048 A CN 201811544048A CN 109784539 A CN109784539 A CN 109784539A
- Authority
- CN
- China
- Prior art keywords
- well group
- long
- term
- oil
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses the method and apparatus using machine learning algorithm training oil/gas well group long-term objective prediction model, and determine the method and apparatus that well group adjusts operation using prediction model.Model training method includes: the history data set obtained about multiple well groups；Feature extraction is carried out to history data set to obtain adjusting the training sample feature set of operation and well group status data based on well group and based on the corresponding label data of well group long-term objective；And training sample feature set and corresponding label data are based on, it is trained using preset machine learning algorithm, obtains housebroken oil/gas well group long-term objective prediction model.The above-mentioned model obtained can predict the long-term objective of each well group, so as to which the practical operation of well group is adjusted according to the long-term objective of prediction.Further, the long-term and short run target that prediction model can merge well group carries out intensified learning as label data, obtains the operating parameter for being able to ascend both short-term and long-term objectives.
Description
Technical field
The present invention relates to oil-gas mining field, it is true to adjust operation for the more particularly, to training of index prediction model and well group
Fixed method and apparatus.
Background technique
In the oil-gas mining field of oil and natural gas, it is only necessary to using natural energy with regard to minable flowing well relatively
It is few.With the reduction of natural energy, need to improve the oil recovery of oil reservoir to stratum oil-bearing stratum flooding or particular chemicals
Rate.It, then need to be by steam assisted gravity drainage techniques (SAGA) for viscous crude and oil-sand.The technology into stratum by continuously infusing
Enter high-temperature steam heating oil reservoir, so that Crude viscosity significantly reduces and is able to flow under the influence of gravity into the production of steam injection downhole
In well, and it is lifted to ground, is achieved in effective exploitation to viscous crude, super-viscous oil and oil-sand.
Natural gas is also equally imbedded among the geological structure of underground sealed with crude oil, some and oil reservoir are in same layer
Position, some individualisms.It can be mined together with crude oil for the natural gas with oil reservoir in same layer position.For only
There is gas reservoir existing for natural gas, recovery method was not only quite similar with crude oil production method, but also the place for having its special.
The either exploitation of oil field, gas field or oil gas field, it usually needs while multiple groups operating parameter is controlled.Often
The operating parameter of rule needs to rely on experienced staff and manually adjusts.Due between different well groups mining conditions difference compared with
Greatly, therefore operating parameter is manually adjusted highly dependent upon personal experience, and accuracy is not high.
For this reason, it may be necessary to a kind of operating parameter regulation scheme that is more accurate and can adapt to various well group conditions.
Summary of the invention
In order to solve the problems, such as above at least one, the present invention using machine learning algorithm be based on more well group historical datas training
Practice well group long-term objective prediction model, above-mentioned model can predict the long-term objective of well group, so as to according to prediction
Long-term objective the operation of practical well group is adjusted.Further, prediction model can merge the long-term and short-term of well group
Index carries out intensified learning as label data, to obtain the operation for being able to ascend both short-term and long-term objectives of well group
Parameter.
According to an aspect of the present invention, it proposes a kind of method of trained oil/gas well group long-term objective prediction model, packets
It includes: obtaining the history data set about multiple well groups, every historical data which concentrates includes: that well group adjusts behaviour
Make, execute the well group status data when well group adjusts operation and execute the well group long-term objective after well group operation；To described
History data set carries out feature extraction processing to obtain the training sample feature for adjusting operation and well group status data based on well group
Collection and the corresponding label data based on well group long-term objective；And it is based on the training sample feature set and corresponding label number
According to being trained using preset machine learning algorithm, obtain housebroken oil/gas well group long-term objective prediction model.As a result,
Pass through machine learning algorithm efficiently use the historical data from more well groups and accurately found out by repetitive exercise operation with it is long-term
The inner link of index, to realize that the long-term objective for all kinds of well groups is predicted.Here, preset machine learning algorithm is for example
It can be deep neural network (DNN) algorithm.
Preferably, this method further include: the phase at least adjusting operation based on the well group for model modification well group and obtaining
Well group achievement data is answered, the housebroken oil/gas well group long-term objective prediction model is updated.Thus, it is possible to through just
The model of instruction is updated further to promote the accuracy of prediction model.
Preferably, above-mentioned update operation may include: and execute well group to model modification well group to adjust operation；Acquisition and well group
Adjust the well group status data and well group short run target for operating the corresponding model modification well group；With the model modification well group
Well group status data and the well group adjust operation constitute forecast sample, and using the oil/gas well group long-term objective predict mould
The well group long-term objective of the type prediction model modification well group corresponding with well group adjusting operation；And it is based on the well group
It adjusts operation and corresponding well group status data obtains sample characteristics, the well group of the well group short run target and prediction that merge acquisition is long
Phase index obtains the training for updating the housebroken oil/gas well group long-term objective prediction model as corresponding label data
Data.Further improve the degree of association of long-term objective and short run target that prediction model is predicted as a result,.The short run target can
To include at least one following: well group abnormal conditions；Oil yield efficiency；Steam injection cost；Short-term recovery ratio；And short-term gas oil ratio.
The well group status data then may include well group state feature and well group history time sequence status feature based on the moment.
Preferably, executing well group to adjust operation to the model modification well group includes: the current of acquisition model modification well group
Well group status data, and obtain the well group adjusting operational set for model modification well group；Based on the housebroken oil gas
Well group long-term objective prediction model, each well group adjusting operation for predicting to adjust with the well group in operational set are corresponding long-term
Index；And it is adjusted from the well group and chooses the optimal well group adjusting operation of corresponding long-term objective in operational set in institute
It states and is executed on model modification well group.Training/update efficiency of prediction model is promoted as a result,.
Preferably, obtaining about the history data set of multiple well groups includes: under the different moments for obtaining corresponding each well group
Well group adjust operation and well group status data and obtain corresponding well group long-term objective.As a result, by for from multiple
The timing training sample of the history data set of well group constructs, and excavates the relevance respectively operated in historical data with long-term objective, by
The subsequent prediction to long-term objective of this implementation model.
Preferably, well group adjusts the adjusting that operates and can be for SAGD well group, may include at least one of following: well group
Steam injection rate adjusts operation；Well group steam injection mass dryness fraction adjusts operation；Well group steam injecting temperature adjusts operation；And well group production-injection ratio is adjusted
Operation.
In addition, well group long-term objective may include: that the long-term recovery ratio of well group and/or well group add up gas oil ratio.
According to another aspect of the present invention, a kind of determining method of oil gas well group adjusting operation is proposed, comprising: for finger
Determine well group, obtain the well group status data of the well group, by the well group status data of the well group respectively be directed to the well group well group
It adjusts each well group in operational set and adjusts operation combination, obtain multiple forecast samples；The multiple forecast sample is carried out
Feature extraction processing, obtains multiple forecast sample features；It is long-term that the multiple forecast sample feature is inputted into oil/gas well group respectively
Index prediction model is predicted, corresponding multiple well group long-term objectives are obtained；And behaviour is adjusted from the well group for the well group
Make to choose the optimal well group adjusting operation of corresponding long-term objective in gathering.Thus, it is possible to which counter from prediction model release most
Good well group adjusts operation.Then, selected well group can be executed on the specified well group and adjusts operation, be achieved in pair
Effective control of practical well group.Preferably, which can be according to above-mentioned training method institute
It obtains.
Preferably, which can also include: and obtain the specified well group to adjust under operation in performed well group
Short run target；Operation, which is adjusted, with the well group status data of the specified well group and performed well group is constituted forecast sample number
According to, and it is corresponding with performed well group adjusting operation described specified using oil/gas well group long-term objective prediction model prediction
The well group long-term objective of well group；Operation is adjusted based on performed well group and corresponding well group status data obtains sample characteristics,
The well group long-term objective of the well group short run target and prediction that obtain is merged as corresponding label data, obtains updating the oil gas
The training data of well group long-term objective prediction model；Summarize the update oil/gas well group long-term objective increased newly in preset time
The training data that prediction model is used is as model modification training dataset；And at least it is based on the model modification training data
Collection, updates the oil/gas well group long-term objective prediction model.Thus, it is possible to still carry out model more in model service stage
Newly.
Similarly, short run target may include at least one following: well group abnormal conditions；Oil yield efficiency；Steam injection cost；
Short-term recovery ratio；And short-term gas oil ratio.Well group long-term objective is then can include: the long-term recovery ratio of well group and/or well group add up oil and gas
Than.
According to a further aspect of the invention, a kind of dress of trained oil/gas well group long-term objective prediction model is additionally provided
It sets, comprising: data capture unit, for obtaining the history data set about multiple well groups, every of historical data concentration is gone through
History data include: after well group adjusts operation, executes the well group status data when well group adjusts operation and executes well group operation
Well group long-term objective；Feature extraction unit, for carrying out feature extraction processing to the history data set to obtain based on well
Group adjusts the training sample feature set of operation and well group status data and the corresponding label data based on well group long-term objective；With
And model training unit, for being based on the training sample feature set and corresponding label data, using preset machine learning
Algorithm is trained, and obtains housebroken oil/gas well group long-term objective prediction model.
Preferably, which can also include: model modification unit, at least based on the well for being directed to model modification well group
The corresponding well group achievement data that group adjusts operation and obtains carries out the housebroken oil/gas well group long-term objective prediction model
It updates.
Preferably, the model modification unit includes: that well group adjusts operation execution unit, for the model modification well
Group executes well group and adjusts operation；First data acquisition subelement, for obtaining the mould corresponding with well group adjusting operation
The well group status data and well group short run target of type update well group；First prediction subelement, for the model modification well group
Well group status data and the well group adjust operation constitute forecast sample, and using the oil/gas well group long-term objective predict mould
The well group long-term objective of the type prediction model modification well group corresponding with well group adjusting operation；And data fusion list
Member obtains sample characteristics for adjusting operation and corresponding well group status data based on the well group, and the well group for merging acquisition is short
Phase index and the well group long-term objective of prediction obtain updating the housebroken oil/gas well group long-term as corresponding label data
The training data of index prediction model.
Preferably, it includes: the second data acquisition subelement that the well group, which adjusts operation execution unit, for obtaining model more
The current well group status data of new well group, and obtain the well group adjusting operational set for model modification well group；Second prediction
Subelement, for being based on the housebroken oil/gas well group long-term objective prediction model, prediction adjusts operation set with the well group
Each well group in conjunction, which is adjusted, operates corresponding long-term objective；Optimal well group adjusts operation selection unit, adjusts from the well group
The optimal well group adjusting operation of corresponding long-term objective is chosen in operational set to execute on the model modification well group.
Preferably, the short run target includes at least one following: well group abnormal conditions；Oil yield efficiency；Steam injection cost；
Short-term recovery ratio；And short-term gas oil ratio.
Preferably, the well group under the different moments that the data capture unit is used to obtain corresponding each well group adjusts operation
With well group status data and the corresponding well group long-term objective of acquisition.
Preferably, the well group status data includes the well group state feature based on the moment and well group history time sequence status spy
Sign.
Preferably, it includes at least one of following that the well group, which adjusts operation: well group steam injection rate adjusts operation；Well group steam injection
Mass dryness fraction adjusts operation；Well group steam injecting temperature adjusts operation；Well group production-injection ratio adjusts operation.
Preferably, the well group long-term objective includes: that the long-term recovery ratio of well group and/or well group add up gas oil ratio.
Preferably, the preset machine learning algorithm is deep neural network (DNN) algorithm.
According to another aspect of the invention, a kind of oil gas well group adjusting operation determining device is additionally provided, comprising: pre- test sample
This acquiring unit, for obtaining the well group status data of the well group for specified well group, by the well group status data of the well group point
The each well group adjusting operation adjusted with the well group for the well group in operational set is combined, and multiple forecast samples are obtained；It is special
Extraction unit is levied, for carrying out feature extraction processing to the multiple forecast sample, obtains multiple forecast sample features；Prediction is single
Member predicts for the multiple forecast sample feature to be inputted oil/gas well group long-term objective prediction model respectively, obtains pair
The multiple well group long-term objectives answered；And optimal adjustment operates selection unit, for adjusting operation from the well group for the well group
An optimal well group of corresponding long-term objective is chosen in set adjusts operation.
Preferably, which may further include: well group adjusts operation execution unit, in the specified well group
It executes selected well group and adjusts operation.
Preferably, the oil/gas well group long-term objective prediction model is obtained according to as above any one method.
Preferably, which can also include: third data acquisition subelement, held for obtaining the specified well group
Capable well group adjusts the short run target under operation；Third predicts subelement, for the well group status data of the specified well group
Operation is adjusted with performed well group and constitutes forecast sample data, and is predicted using the oil/gas well group long-term objective prediction model
The well group long-term objective of the specified well group corresponding with performed well group adjusting operation；Data fusion unit, for being based on
Performed well group adjusts operation and corresponding well group status data obtains sample characteristics, merge acquisition well group short run target and
The well group long-term objective of prediction obtains updating the oil/gas well group long-term objective prediction model as corresponding label data
Training data；Data summarization unit, for summarizing the update oil/gas well group long-term objective prediction mould increased newly in preset time
The training data that type is used is as model modification training dataset；Model modification unit, at least being instructed based on the model modification
Practice data set, updates the oil/gas well group long-term objective prediction model.
Preferably, the short run target includes at least one following: well group abnormal conditions；Oil yield efficiency；Steam injection cost；
Short-term recovery ratio；And short-term gas oil ratio, the well group long-term objective include: that the long-term recovery ratio of well group and/or well group add up oil and gas
Than.
According to another aspect of the invention, a kind of determining system of oil gas well group adjusting operation is additionally provided, including data obtain
Take unit, feature extraction unit, model training unit, predicting unit and optimal adjustment operation selection unit, the oil/gas well
Group, which adjusts to operate, determines systematic training oil/gas well group long-term objective prediction model, is predicted using the model, is instructed in model
Practice the stage: the data capture unit is used to obtain history data set about multiple well groups, and every of historical data concentration
Historical data includes: that well group adjusts operation, executes the well group status data when well group adjusts operation and executes well group operation
Well group long-term objective afterwards；The feature extraction unit is used to carry out the history data set feature extraction processing to obtain base
The training sample feature set of operation and well group status data is adjusted in well group and based on the corresponding label number of well group long-term objective
According to；And the model training unit is used for preset based on the training sample feature set and corresponding label data, use
Machine learning algorithm is trained, and housebroken oil/gas well group long-term objective prediction model is obtained, in the model prediction stage: described
Data capture unit is used to the well group status data of the well group is obtained, by the well group status data of the well group for specified well group
The each well group adjusting operation adjusted respectively with the well group for the well group in operational set is combined, and multiple forecast samples are obtained；
The feature extraction unit is used to carry out feature extraction processing to the multiple forecast sample, obtains multiple forecast sample features；
The predicting unit is used to inputting the multiple forecast sample feature into oil/gas well group long-term objective prediction model respectively and carry out in advance
It surveys, obtains corresponding multiple well group long-term objectives；And the optimal adjustment operates selection unit from the well group for being directed to the well group
It adjusts and chooses the optimal well group adjusting operation of corresponding long-term objective in operational set.
Preferably, the oil gas well group adjusts operation and determines the practical adjustments operation progress mould that system is also carried out based on prediction
Type updates, in the model modification stage: the data capture unit is for obtaining the specified well group in performed well group
Adjust the short run target under operation；The feature extraction unit is used for the well group status data of the specified well group and performed
Well group adjust operation constitute forecast sample data, and using the oil/gas well group long-term objective prediction model prediction with it is performed
Well group adjust the well group long-term objective for operating the corresponding specified well group；Operation is adjusted based on performed well group and is corresponded to
Well group status data obtain sample characteristics, merge acquisition well group short run target and prediction well group long-term objective as correspond to
Label data, obtain the training data for updating the oil/gas well group long-term objective prediction model；Summarize new in preset time
The training data that the update oil/gas well group long-term objective prediction model of increasing is used is as model modification training dataset；The mould
Type training unit is at least based on the model modification training dataset, updates the oil/gas well group long-term objective prediction model.
According to another aspect of the invention, a kind of calculating equipment is additionally provided, comprising: processor；And memory, thereon
It is stored with executable code, when the executable code is executed by the processor, the processor is executed in as above and appoints
Method described in what one.
According to another aspect of the invention, a kind of non-transitory machinable medium is additionally provided, is stored thereon with
Executable code executes the processor as above any when the executable code is executed by the processor of electronic equipment
Method described in.
Train well group long-term objective pre- by the present invention in that being based on more well group history data sets with machine learning algorithm as a result,
Model is surveyed, above-mentioned model can predict the long-term objective of well group, so as to the long-term objective according to prediction to reality
The operation of well group is adjusted.Further, prediction model can merge the long-term and short run target of well group as label data
Intensified learning is carried out, to obtain the operating parameter for being able to ascend both short-term and long-term objectives of well group.
Detailed description of the invention
Disclosure illustrative embodiments are described in more detail in conjunction with the accompanying drawings, the disclosure above-mentioned and its
Its purpose, feature and advantage will be apparent, wherein in disclosure illustrative embodiments, identical reference label
Typically represent same parts.
Fig. 1 shows the method for trained oil/gas well group long-term objective prediction model according to an embodiment of the invention.
Fig. 2 shows examples prediction model training and updated.
Fig. 3 shows oil gas well group according to an embodiment of the invention and adjusts the determining method of operation.
Fig. 4 shows showing for the device of trained oil/gas well group long-term objective prediction model according to an embodiment of the invention
Meaning property block diagram.
Fig. 5 shows the schematic block diagram that oil gas well group according to an embodiment of the invention adjusts operation determining device.
Fig. 6 shows the example that oil gas well group operation according to the present invention adjusts the system of determination.
Fig. 7 shows the structural schematic diagram according to an embodiment of the invention for calculating equipment.
Specific embodiment
The preferred embodiment of the disclosure is more fully described below with reference to accompanying drawings.Although showing the disclosure in attached drawing
Preferred embodiment, however, it is to be appreciated that may be realized in various forms the disclosure without the embodiment party that should be illustrated here
Formula is limited.On the contrary, these embodiments are provided so that this disclosure will be more thorough and complete, and can be by the disclosure
Range is completely communicated to those skilled in the art.
In oilfield exploitation, it usually needs while multiple groups operating parameter is controlled.Conventional operating parameter needs to rely on
Experienced staff manually adjusts.Since mining conditions differ larger, the hand of operating parameter between different well groups
Dynamic adjust relies on personal experience, and accuracy is not high.
In order to promote oil gas harvest efficiency, the present invention is based on more well group history data sets training well using machine learning algorithm
Group long-term objective prediction model, above-mentioned model can predict the long-term objective of well group, so as to according to the length of prediction
The operation of practical well group is adjusted in phase index.Further, prediction model can merge the long-term and short run target of well group
Intensified learning is carried out as learning objective, to obtain the operation tune for being able to ascend both short-term and long-term objectives of well group
Section.
Fig. 1 shows the method for trained oil/gas well group long-term objective prediction model according to an embodiment of the invention.?
This, " oil/gas well group long-term objective prediction model " refers to the model of the long-term objective for predicting oil/gas well group." oil gas well group "
It may refer to petroleum well group, natural gas well group, or the well group including both oil and natural gas." well group " is typically referred to infuse
Centered on well, the basic development block in the oil field that the oil well being associated with surrounding is constituted.In this application, " well group " can be used for
One or more wells with incidence relation are referred to, for example, single oil well, single gas well, single oil gas well, or including water filling
One group of well of well or steam well and one or more oil wells associated therewith.It is long that " long-term objective " reference can reflect well group
The index parameter of the oil/gas harvesting situation of phase, such as the accumulative gas oil ratio of the long-term recovery ratio of well group and/or SAGD well group.In addition,
It will be obvious that in this application, " oil/gas well group long-term objective prediction model " may refer to be exclusively used in oil exploitation well group
Prediction model, be exclusively used in the prediction model of natural gas extraction well group, air-fuel mixture model etc..
In step s 110, the history data set about multiple well groups is obtained.It, can in order to promote the universality of prediction model
Well group as much as possible, or the data from representational well group are come to obtain in model training stage, is predicted as a result,
Model acquistion can predict all kinds of well group situations from data, and even different well group types are (for example, individual well or SAGD well
Group or other water filling well groups) ability.
Specifically, every historical data which concentrates may include: that well group adjusts operation, executes the well group tune
The well group long-term objective after well group status data and execution well group operation when section operation.Here, well group adjusting operation includes
For each generic operation of well group, aforesaid operations can have an impact short-term and long-term objective (for example, recovery ratio).In SAGD well group
In the case where, it for example may include well group production-injection ratio, steam injection rate, steam injection mass dryness fraction and/or and steam injecting temperature that well group, which adjusts operation,
Adjusting operation.It should be understood that well group adjusts operation also in the case where other kinds of well group for example fills the water well group
It may include the operation of the adjusting such as waterflood injection rate and temperature.Well group status data then may include based on the moment (for example, adjusting
Save operation the correspondence moment) well group state feature and well group history time sequence status feature.Above-mentioned state feature may include oil
Mild air pressure etc..Since compared to instantaneous state, the integrality in a duration can more reflect the real conditions of well group,
It therefore should also include history time sequence status feature except the status data based on the moment, such as the adjusting operates before carrying out
A period of time (for example, 30 minutes) well head state, for example, average value.
In one embodiment of the present of invention crowd, described obtain about the history data set of multiple well groups includes: to obtain to correspond to
Well group under the different moments of each well group adjusts operation and well group status data and obtains corresponding well group long-term objective.
Then, in step S120, feature extraction processing is carried out to obtain training sample feature set and mark to history data set
Sign data.In one embodiment, above-mentioned data include that the training sample spy of operation and well group status data is adjusted based on well group
Collection and the corresponding label data based on well group long-term objective.Due to need from historical data acquistion to Secular Variation Tendency
Predictive ability, therefore the training sample of each well group is preferably constructed with timing.Then, step S120 may include with corresponding each
Well group under the different moments of a well group adjusts operation, well group status data and well group long term object to construct training sample and right
Answer label data.For example, a collection of timing training sample can be constructed for well group A, a collection of timing training sample is constructed for well group B
This, and so on.Facilitate following model in the subsequent training stage as a result, by well group the variation tendency between learning data, from
And realize the prediction to long-term objective.
In step S130, the training sample feature set and corresponding label data are based on, using preset machine learning
Algorithm is trained, and obtains housebroken oil/gas well group long-term objective prediction model.Here, any appropriate machine can be used
Learning algorithm is trained.In one embodiment, DNN (deep learning network) can be used to be trained, it is more to be based on
The feature of layer network is transmitted, and the relevance between data is more accurately held.
The present invention realizes the training to prediction model via step S110~S130 and its preferred embodiment as a result,.?
In one embodiment, the prediction model that can also be completed to above-mentioned training is updated.At this point, above-mentioned training can be regarded as needle
To the first instruction of prediction model.In the present invention, it is preferred to be updated using intensified learning technology to the prediction model through just instructing.
By constantly learning, the using effect of prediction model will not reduce over time, can obtain instead more optimize it is pre-
Survey result.
For this purpose, training method of the invention can also include at least adjusting operation based on the well group for model modification well group
And the corresponding well group achievement data obtained, the housebroken oil/gas well group long-term objective prediction model is updated." model
Update well group " it may refer to generally can be practical work for the well group for generating the training data for updating prediction model herein
Well group in industry.Above-mentioned true well group can carry out operation adjusting under the guidance of the prediction model through just instructing, the correspondence got
Data can be used for the update of prediction model in turn.
In a preferred embodiment, it can be combined as learning objective with short-term and long-term objective, utilizes extensive chemical
Technology innovation prediction model is practised, so that adjusting each time can comprehensively consider short in prediction model when being subsequently used for adjusting
Phase and long-term objective, are more in line with production requirement.As previously mentioned, long-term objective be reflect well group relatively long period (for example,
Half a year) interior harvesting situation index parameter, such as the long-term recovery ratio of well group and/or well group add up gas oil ratio.Short run target is then
The index situation of the interior well group of relatively short time (for example, half an hour) after operation is adjusted, including but not limited to: well group exception feelings
Condition；Oil yield efficiency；Steam injection cost；Short-term recovery ratio；And short-term gas oil ratio.
Specifically, model modification operation may include: and execute well group to the model modification well group to adjust operation；Obtain with
The well group adjusts the well group status data and well group short run target for operating the corresponding model modification well group；With the model
The well group status data (executing state when well group adjusts operation) and the well group for updating well group adjust operation and constitute pre- test sample
This, and use the oil/gas well group long-term objective prediction model prediction model modification corresponding with well group adjusting operation
The well group long-term objective of well group；And operation and corresponding well group status data are adjusted based on the well group and obtain sample characteristics,
The well group long-term objective of the well group short run target and prediction that obtain is merged as corresponding label data, obtains updating described through instructing
The training data of experienced oil/gas well group long-term objective prediction model.
Long-term and short run target can be merged based on any suitable method.In one embodiment, it can incite somebody to action
The long-term objective of model prediction is added after centainly losing (for example, multiplied by a weight less than 1) with short run target.Index
Fused result can be used as the target (i.e. as label) of model training, and will adjust operation and well group state accordingly
Data value updates model parameter, completes a closed loop learning process as mode input.It so constantly repeats, so that prediction mould
Type is by intensified learning, and voluntarily acquistion is more close to the model parameter that can optimize both shot and long term indexs.
In actual operation, it is preferable to use the operation tune most preferably influenced can be generated on the index of following a period of time
Section.Thereby, it is possible to lift scheme update efficiency while, Optimized model update well group itself recovery ratio and/or cost with
Time efficiency.The selection that above-mentioned Optimum Operation is adjusted can be realized by prediction model.In one embodiment, to model modification
It may include: the current well group status data for obtaining model modification well group that well group, which executes well group and adjusts operation, and obtains and be directed to
The well group of model modification well group adjusts operational set；Based on housebroken oil/gas well group long-term objective prediction model, prediction and institute
It states each well group that well group is adjusted in operational set and adjusts the corresponding long-term objective of operation；And it is adjusted in operational set from well group
The optimal well group adjusting operation of corresponding long-term objective is chosen to execute on the model modification well group.
In one embodiment of the invention, the well group adjusting operational set packet obtained for model modification well group
Include: the possible well group of whole obtained for current model modification well group adjusts operation, obtains well group and adjusts operational set.?
The possible well group in part that can rule of thumb obtain for current model modification well group adjusts operation, obtains well group and adjusts behaviour
Work is gathered.
Fig. 2 shows examples prediction model training and updated.
Shown on the left of Fig. 2, related data and the adjusting for obtaining each well group history manual adjustment first are corresponding related
Instantaneous or short term state (for example, recovery ratio and accumulative gas oil ratio, temperature and air pressure etc.) is then based on above-mentioned state (or one
Part) calculate the long-term objective of each period.For example, long-term recovery ratio and accumulative gas oil ratio can pass through the short term state that adds up
It obtains.
Each well group can be regarded as a training sample at every point of time.Training sample is obtained here, reasonably selecting
This temporal frequency, the feature of extraction operation dimension and index dimension, the long-term objective of various time points is as model training mesh
Mark, i.e. prediction well group have done the long-term recovery ratio after some operation and accumulative oil-gas ratio at some time point.
Index effect that prediction model (for example, intensified learning model) generates after being adjusted every time according to history learns
Whether such operation is the optimum choice under situation at that time, the experience of good absorbing and avoids the occurrence of the mistake that manual adjustment commits excesses
Accidentally.This is the process of an initialization study.The model is completed just to instruct and have certain regulating power as a result,.
Shown on the right side of figure, the model of Jing Chuxun can then be applied to each update well group, predict each well group
Available optimal recovery ratio and accumulative oil-gas ratio in the case where which type of adjusts operation, and adjusting is thus made respectively.Then,
The short run target after adjusting, and the long-term objective that can be obtained via the model prediction operation are calculated, the two is passed through certain
Mode feeds back to model after merging, for example, by model prediction result after some lose (multiplied by a weight less than 1) with
Short run target is added.The data of target of the fused result of index as model training, current state (adjust operation and work as
Before/history well group state) and it is used as mode input, model parameter is updated, a closed loop learning process is completed, is so constantly repeated.
As above Fig. 1 and Fig. 2 is combined to describe model training of the invention and preferred update method.The above-mentioned model of training
Purpose is to instruct true oil-gas mining activity using model.For this purpose, the invention also includes a kind of determining oil gas well groups to adjust
Method.Fig. 3 shows oil gas well group according to an embodiment of the invention and adjusts the determining method of operation.The above method is to be directed to
What specified practical well group was made.
In step S310, the well group status data of well group is obtained, by the well group status data of the well group respectively and for this
The well group of well group adjusts each well group in operational set and adjusts operation combination, obtains multiple forecast samples.It is right in step S320
Multiple forecast samples carry out feature extraction processing, obtain multiple forecast sample features.It is in step S330, multiple forecast samples are special
Sign respectively predicted by input oil/gas well group long-term objective prediction model, obtains corresponding multiple well group long-term objectives.In step
S340 is adjusted from the well group for the well group and is chosen the optimal well group adjusting operation of corresponding long-term objective in operational set.
In a preferred embodiment, this method can also include specifying at this and executing selected well group on well group and adjust operation.
The oil/gas well group long-term objective prediction model obtained based on the above-mentioned training of the present invention and/or update method can be used for reality
Border well group adjusts the determination of operation.Further, above-mentioned true adjusting operation and its caused corresponding states and achievement data
It can be used as a part that prediction model itself updates (intensified learning).It is closed for example, can equally be seen as right side shown in Fig. 2
A part of ring study.
For this purpose, it can also include: to obtain the specified well group to adjust in performed well group that the well group, which adjusts the method for determination,
Short run target under operation；Operation, which is adjusted, with the well group status data of the specified well group and performed well group is constituted pre- test sample
Notebook data, and it is corresponding with performed well group adjusting operation described using oil/gas well group long-term objective prediction model prediction
The well group long-term objective of specified well group；Operation is adjusted based on performed well group and corresponding well group status data obtains sample spy
Sign, merges the well group short run target of acquisition and the well group long-term objective of prediction as corresponding label data, obtains described in update
The training data of oil/gas well group long-term objective prediction model；It is long-term to summarize the update oil/gas well group increased newly in preset time
The training data that index prediction model is used is as model modification training dataset；At least it is based on the model modification training data
Collection, updates the oil/gas well group long-term objective prediction model.
Similarly, above-mentioned short run target may include at least one following: well group abnormal conditions；Oil yield efficiency；Steam injection
Cost；Short-term recovery ratio；And short-term gas oil ratio.Well group long-term objective is then can include: the long-term recovery ratio of well group and/or well group are accumulative
Gas oil ratio.
So far, it has been combined attached drawing 1-3 and model training of the invention, prediction and the method for update is described in detail.This
Inventing method as described above can be realized by its corresponding device.
It is (as follows that Fig. 4 shows trained oil/gas well group long-term objective prediction model device according to an embodiment of the invention
Abbreviation training device 400) schematic block diagram.Fig. 5 shows oil gas well group according to an embodiment of the invention and adjusts operation
The schematic block diagram of determining device (following abbreviation determining device 500).Wherein, the functional module of device can be by realizing the present invention
The combination of the hardware of principle, software or hardware and software is realized.It will be appreciated by persons skilled in the art that Fig. 4 or Fig. 5 institute
The functional module of description can combine or be divided into submodule, to realize the principle of foregoing invention.Therefore, this paper
Description can be supported to any possible combination or division of functions described herein module or further restriction.
Training device 400 shown in Fig. 4 can be used to realize training method shown in FIG. 1, determining device shown in fig. 5
500 can be shown in Fig. 3 for realizing.The functional module that only training device 400 and determining device 500 can have below
And the operation that each functional module can execute is described briefly, for the detail section being directed to may refer to above in association with
The description of Fig. 1 or 3, which is not described herein again.
As shown in figure 4, training device 400 may include data capture unit 410, feature extraction unit 420 and model instruction
Practice unit 430.
The available history data set about multiple well groups of data capture unit 410, every of historical data concentration
Historical data includes: that well group adjusts operation, executes the well group status data when well group adjusts operation and executes well group operation
Well group long-term objective afterwards.Feature extraction unit 420 can carry out feature extraction processing to the history data set to obtain base
The training sample feature set of operation and well group status data is adjusted in well group and based on the corresponding label number of well group long-term objective
According to.Model training unit 430 can be based on the training sample feature set and corresponding label data, using preset engineering
It practises algorithm to be trained, obtains housebroken oil/gas well group long-term objective prediction model.Preset machine learning algorithm for example may be used
To be deep neural network (DNN) algorithm.
In a preferable example, which can also include model modification unit (not shown).Model is more
The corresponding well group achievement data that new unit at least can be adjusted operation and be obtained based on the well group for model modification well group, to institute
Housebroken oil/gas well group long-term objective prediction model is stated to be updated.
Specifically, model modification unit may include: well group adjust operation execution unit, the first data acquisition subelement,
First prediction subelement and data integrated unit (not shown).
Well group adjusts operation execution unit can execute well group adjusting operation to the model modification well group.First data obtain
Take subelement available to adjust the well group status data and well group of the corresponding model modification well group of operation with the well group
Short run target.First prediction subelement can adjust operation with the well group status data of the model modification well group and the well group
Forecast sample is constituted, and uses oil/gas well group long-term objective prediction model prediction institute corresponding with well group adjusting operation
State the well group long-term objective of model modification well group.Data fusion unit can adjust operation and corresponding well group based on the well group
Status data obtains sample characteristics, merges the well group short run target of acquisition and the well group long-term objective of prediction as corresponding label
Data obtain the training data for updating the housebroken oil/gas well group long-term objective prediction model.
In one embodiment, it may include the second data acquisition subelement, that the well group, which adjusts operation execution unit,
Two prediction subelements and optimal well group adjust operation selection unit (not shown).
The current well group status data of the available model modification well group of second data acquisition subelement, and obtain and be directed to
The well group of model modification well group adjusts operational set.Second prediction subelement can be long-term based on the housebroken oil/gas well group
Index prediction model predicts long-term objective corresponding with each well group adjusting operation in well group adjusting operational set.Most
Excellent well group adjusts operation selection unit, adjusts in operational set from the well group and chooses an optimal well of corresponding long-term objective
Group adjusts operation and executes on the model modification well group.
In a preferable example, the data capture unit 410 was used to obtain under the different moments of corresponding each well group
Well group adjust operation and well group status data and obtain corresponding well group long-term objective.
The short run target may include at least one following: well group abnormal conditions；Oil yield efficiency；Steam injection cost；It is short
Phase recovery ratio；And short-term gas oil ratio.The well group status data may include that well group state feature based on the moment and well group are gone through
History time sequence status feature.
It may include at least one of following that the well group, which adjusts operation: well group steam injection rate adjusts operation；Well group steam injection is dry
Degree adjusts operation；Well group steam injecting temperature adjusts operation；Well group production-injection ratio adjusts operation.Preferably, the well group long-term objective packet
Include: the long-term recovery ratio of well group and/or well group add up gas oil ratio.
It should be understood that the first and second subelements involved in above-mentioned update and operation execution are (for example, the first data
Obtain subelement, the second data acquisition subelement) it can be separated unit, it is also possible to and the corresponding unit in model training
Identical unit, for example, the data capture unit for model training can also be used for the data in model modification and well group practical operation
It obtains.Correspondingly, model prediction can be carried out using identical predicting unit in practical operation by updating.
As shown in figure 5, determining device 500 of the invention may include forecast sample acquiring unit 510, feature extraction unit
520, predicting unit 530 and optimal adjustment operate selection unit 540.
Forecast sample acquiring unit 510 can obtain the well group status data of the well group, by the well group for specified well group
Well group status data respectively with for the well group well group adjust operational set in each well group adjust operation combined, obtain
Multiple forecast samples.Feature extraction unit 520 can carry out feature extraction processing to the multiple forecast sample, obtain multiple pre-
Test sample eigen.The multiple forecast sample feature can be inputted the prediction of oil/gas well group long-term objective by predicting unit 530 respectively
Model is predicted, corresponding multiple well group long-term objectives are obtained.Optimal adjustment operates selection unit 540 can be from for the well
The well group of group, which is adjusted, chooses the optimal well group adjusting operation of corresponding long-term objective in operational set.
In one embodiment, which may further include: well group adjusts operation execution unit (in figure
It is not shown).Well group adjusts operation execution unit can execute selected well group adjusting operation on the specified well group.
The oil/gas well group long-term objective prediction model is obtained according to as above any one method.
In a preferred embodiment, the device can also include third data acquisition subelement, third prediction subelement,
Data fusion unit, data summarization unit and model modification unit (not shown).
The available specified well group of third data acquisition subelement adjusts short-term under operation in performed well group
Index.Third prediction subelement can adjust operation with the well group status data of the specified well group and performed well group and be constituted
Forecast sample data, and it is corresponding with performed well group adjusting operation using oil/gas well group long-term objective prediction model prediction
The specified well group well group long-term objective.Data fusion unit can based on performed well group adjust operation and it is corresponding
Well group status data obtains sample characteristics, merges the well group short run target of acquisition and the well group long-term objective of prediction as corresponding
Label data obtains the training data for updating the oil/gas well group long-term objective prediction model.Data summarization unit can converge
The training data that the update oil/gas well group long-term objective prediction model increased newly in total preset time is used is instructed as model modification
Practice data set.Model modification unit can at least be based on the model modification training dataset, and it is long-term to update the oil/gas well group
Index prediction model.
The short run target may include at least one following: well group abnormal conditions；Oil yield efficiency；Steam injection cost；It is short
Phase recovery ratio；And short-term gas oil ratio, the well group long-term objective is then can include: the long-term recovery ratio of well group and/or the accumulative oil of well group
Vapour ratio.
In one embodiment, determining device shown in fig. 5 can be combined with training device shown in Fig. 4, to constitute
A kind of determining system of oil gas well group operation adjusting.Fig. 6 shows oil gas well group operation according to the present invention and adjusts the system of determination
One example.As shown in fig. 6, system 600 may include data capture unit 610, feature extraction unit 620, model training list
Member 630, predicting unit 640 and optimal adjustment operate selection unit 650.The system 600 utilizes the history from each well group
Data train long-term objective prediction model, and the prediction carried out by using the prediction model, obtain for the excellent of practical well group
Change operation to adjust, to promote production efficiency and the cost of implementation optimization of well group.Further, which can be based on above-mentioned
Well group state caused by optimizing regulation and the variation of shot and long term index, further update prediction model, are closed to realize
Ring learning process.
Specifically, in model training stage: data capture unit 610 can be used for obtaining the history number about multiple well groups
According to collection, every historical data which concentrates includes: that well group adjusts operation, executes the well group when well group adjusts operation
Well group long-term objective after status data and execution well group operation.Feature extraction unit 620 can be used for the history number
According to collection carry out feature extraction processing with obtain based on well group adjust operation and well group status data training sample feature set and
Corresponding label data based on well group long-term objective.Model training unit 630 then can based on the training sample feature set and
Corresponding label data are trained using preset machine learning algorithm, obtain housebroken oil/gas well group long-term objective prediction
Model.System obtains housebroken prediction model as a result,.
Then, in the model prediction stage: data capture unit 610 can be used for obtaining the well group for specified well group
The well group status data of the well group is adjusted each of operational set with the well group for the well group respectively by well group status data
Well group adjusts operation and combines, and obtains multiple forecast samples；Feature extraction unit 620 can be used for the multiple forecast sample into
Row feature extraction processing, obtains multiple forecast sample features；Predicting unit 640 can be used for the multiple forecast sample feature
Input oil/gas well group long-term objective prediction model is predicted respectively, obtains corresponding multiple well group long-term objectives；And it is optimal
It adjusts operation selection unit 650 and adjusts optimal one of the corresponding long-term objective of selection in operational set from the well group for the well group
Well group adjusts operation.
Further, oil gas well group, which adjusts to operate, determines that system 600 is also based on the practical adjustments operation that prediction carries out
Carry out model modification.Then, in the model modification stage: data capture unit 610 can be used for obtaining the specified well group in institute
The well group of execution adjusts the short run target under operation；Feature extraction unit 620 can be used for the well group shape of the specified well group
State data and performed well group adjust operation and constitute forecast sample data, and predict mould using the oil/gas well group long-term objective
The well group long-term objective of the type prediction specified well group corresponding with performed well group adjusting operation；Based on performed well group
It adjusts operation and corresponding well group status data obtains sample characteristics, the well group of the well group short run target and prediction that merge acquisition is long
Phase index obtains the training data for updating the oil/gas well group long-term objective prediction model as corresponding label data；It converges
The training data that the update oil/gas well group long-term objective prediction model increased newly in total preset time is used is instructed as model modification
Practice data set；Model training unit 630 is at least based on the model modification training dataset, updates the oil/gas well group and refers to for a long time
Mark prediction model.
In one embodiment, it is disposed on line in the prediction model that such as training obtains and is based on model prediction service
API carries out the centralized management that sample data obtains in real time on line and prediction and operation are adjusted or other are related to prediction data remittance
Total situation.
Fig. 7 shows the structural schematic diagram according to an embodiment of the invention for calculating equipment.
Referring to Fig. 7, calculating equipment 700 includes memory 710 and processor 720.
Processor 720 can be the processor of a multicore, also may include multiple processors.In some embodiments,
Processor 720 may include a general primary processor and one or more special coprocessors, such as graphics process
Device (GPU), digital signal processor (DSP) etc..In some embodiments, the circuit reality of customization can be used in processor 720
It is existing, such as application-specific IC (ASIC, Application Specific Integrated Circuit) or scene
Programmable gate array (FPGA, Field Programmable Gate Arrays).
Memory 710 may include various types of storage units, such as Installed System Memory, read-only memory (ROM), and forever
Long storage device.Wherein, ROM can store the static data of other modules needs of processor 720 or computer or refer to
It enables.Permanent storage can be read-write storage device.Permanent storage can be after computer circuit breaking not
The non-volatile memory device of the instruction and data of storage can be lost.In some embodiments, permanent storage device uses
Mass storage device (such as magnetically or optically disk, flash memory) is used as permanent storage.In other embodiment, permanently deposit
Storage device can be removable storage equipment (such as floppy disk, CD-ROM drive).Installed System Memory can be read-write storage equipment or
The read-write storage equipment of volatibility, such as dynamic random access memory.Installed System Memory can store some or all processors
The instruction and data needed at runtime.In addition, memory 710 may include the combination of any computer readable storage medium,
Including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read only memory), disk and/or
CD can also use.In some embodiments, memory 710 may include that removable storage that is readable and/or writing is set
It is standby, for example, laser disc (CD), read-only digital versatile disc (such as DVD-ROM, DVD-dual layer-ROM), read-only Blu-ray Disc,
Super disc density, flash card (such as SD card, min SD card, Micro-SD card etc.), magnetic floppy disc etc..It is computer-readable to deposit
It stores up medium and does not include carrier wave and the momentary electron signal by wirelessly or non-wirelessly transmitting.
It is stored with executable code on memory 710, when executable code is handled by processor 720, can make to handle
Device 720 executes the method addressed above.
It is above described in detail by reference to attached drawing according to the present invention for the training of index prediction model and well group tune
The determining method and apparatus of section operation.
In addition, being also implemented as a kind of computer program or computer program product, the meter according to the method for the present invention
Calculation machine program or computer program product include the calculating for executing the above steps limited in the above method of the invention
Machine program code instruction.
Alternatively, the present invention can also be embodied as a kind of (or the computer-readable storage of non-transitory machinable medium
Medium or machine readable storage medium), it is stored thereon with executable code (or computer program or computer instruction code),
When the executable code (or computer program or computer instruction code) by electronic equipment (or calculate equipment, server
Deng) processor execute when, so that the processor is executed each step according to the above method of the present invention.
Those skilled in the art will also understand is that, various illustrative logical blocks, mould in conjunction with described in disclosure herein
Block, circuit and algorithm steps may be implemented as the combination of electronic hardware, computer software or both.
The flow chart and block diagram in the drawings show the possibility of the system and method for multiple embodiments according to the present invention realities
Existing architecture, function and operation.In this regard, each box in flowchart or block diagram can represent module, a journey
A part of sequence section or code, a part of the module, section or code include one or more for realizing defined
The executable instruction of logic function.It should also be noted that in some implementations as replacements, the function of being marked in box can also
To be occurred with being different from the sequence marked in attached drawing.For example, two continuous boxes can actually be basically executed in parallel,
They can also be executed in the opposite order sometimes, and this depends on the function involved.It is also noted that block diagram and/or stream
The combination of each box in journey figure and the box in block diagram and or flow chart, can the functions or operations as defined in executing
Dedicated hardware based system realize, or can realize using a combination of dedicated hardware and computer instructions.
Various embodiments of the present invention are described above, above description is exemplary, and non-exclusive, and
It is not limited to disclosed each embodiment.Without departing from the scope and spirit of illustrated each embodiment, for this skill
Many modifications and changes are obvious for the those of ordinary skill in art field.The selection of term used herein, purport
In the principle, practical application or improvement to the technology in market for optimally explaining each embodiment, or make the art
Other those of ordinary skill can understand each embodiment disclosed herein.
Claims (10)
1. a kind of method of trained oil/gas well group long-term objective prediction model, comprising:
The history data set about multiple well groups is obtained, every historical data which concentrates includes: that well group adjusts behaviour
Make, execute the well group status data when well group adjusts operation and execute the well group long-term objective after well group operation；
Feature extraction processing is carried out to the history data set to obtain the instruction for adjusting operation and well group status data based on well group
Practice sample characteristics collection and the corresponding label data based on well group long-term objective；And
Based on the training sample feature set and corresponding label data, it is trained, is obtained using preset machine learning algorithm
To housebroken oil/gas well group long-term objective prediction model.
2. the method as described in claim 1, further includes:
The corresponding well group achievement data at least adjusting operation based on the well group for model modification well group and obtaining, to described through instructing
Experienced oil/gas well group long-term objective prediction model is updated.
3. method according to claim 2, wherein it is described at least based on for model modification well group well group adjust operation and
The corresponding well group achievement data obtained, is updated the housebroken oil/gas well group long-term objective prediction model and includes:
Well group is executed to the model modification well group and adjusts operation；
Obtain the well group status data and well group short run target of the model modification well group corresponding with well group adjusting operation；
Operation is adjusted with the well group status data of the model modification well group and the well group and is constituted forecast sample, and described in use
The well group of the oil/gas well group long-term objective prediction model prediction model modification well group corresponding with well group adjusting operation is long
Phase index；And
Operation is adjusted based on the well group and corresponding well group status data obtains sample characteristics, and the well group for merging acquisition refers in short term
Mark and the well group long-term objective of prediction obtain updating the housebroken oil/gas well group long-term objective as corresponding label data
The training data of prediction model.
4. method as claimed in claim 3, wherein executing well group adjusting operation to the model modification well group includes:
The current well group status data of model modification well group is obtained, and obtains the well group adjusting operation for model modification well group
Set；
Based on the housebroken oil/gas well group long-term objective prediction model, predict to adjust with the well group every in operational set
A well group, which is adjusted, operates corresponding long-term objective；
It is adjusted from the well group and chooses the optimal well group adjusting operation of corresponding long-term objective in operational set in the mould
Type updates to be executed on well group.
5. a kind of oil gas well group, which is adjusted, operates the method for determination, comprising:
For specified well group, the well group status data of the well group is obtained, by the well group status data of the well group respectively and for this
The well group of well group adjusts each well group in operational set and adjusts operation combination, obtains multiple forecast samples；
Feature extraction processing is carried out to the multiple forecast sample, obtains multiple forecast sample features；
The multiple forecast sample feature is inputted oil/gas well group long-term objective prediction model to predict respectively, is obtained corresponding
Multiple well group long-term objectives；And
It is adjusted from the well group for the well group and chooses the optimal well group adjusting operation of corresponding long-term objective in operational set.
6. a kind of device of trained oil/gas well group long-term objective prediction model, comprising:
Data capture unit, for obtaining the history data set about multiple well groups, every history number of historical data concentration
According to well group status data when including: well group adjusting operation, executing well group adjusting operation and execute the well after well group operation
Group long-term objective；
Feature extraction unit, for the history data set carry out feature extraction processing with obtain based on well group adjust operation and
The training sample feature set of well group status data and corresponding label data based on well group long-term objective；And
Model training unit, for being based on the training sample feature set and corresponding label data, using preset engineering
It practises algorithm to be trained, obtains housebroken oil/gas well group long-term objective prediction model.
7. a kind of oil gas well group adjusts operation determining device, comprising:
Forecast sample acquiring unit, for the well group status data of the well group being obtained, by the well group of the well group for specified well group
Each well group adjusting operation that status data is adjusted respectively with the well group for the well group in operational set is combined, and is obtained multiple pre-
Test sample sheet；
Feature extraction unit obtains multiple forecast sample features for carrying out feature extraction processing to the multiple forecast sample；
Predicting unit carries out in advance for the multiple forecast sample feature to be inputted oil/gas well group long-term objective prediction model respectively
It surveys, obtains corresponding multiple well group long-term objectives；And
Optimal adjustment operates selection unit, and corresponding long-term objective is chosen in operational set for adjusting from the well group for the well group
An optimal well group adjusts operation.
8. a kind of oil gas well group, which is adjusted, operates the system of determination, including data capture unit, feature extraction unit, model training list
Member, predicting unit and optimal adjustment operate selection unit, and the oil gas well group, which adjusts to operate, determines systematic training oil gas well group
Long-term objective prediction model predicted using the model,
In model training stage:
The data capture unit is used to obtain the history data set about multiple well groups, every history which concentrates
Data include: after well group adjusts operation, executes the well group status data when well group adjusts operation and executes well group operation
Well group long-term objective；
The feature extraction unit is used to carry out the history data set feature extraction processing to obtain adjusting behaviour based on well group
Corresponding label data of the work with the training sample feature set of well group status data and based on well group long-term objective；And
The model training unit is used to be based on the training sample feature set and corresponding label data, using preset machine
Learning algorithm is trained, and obtains housebroken oil/gas well group long-term objective prediction model,
In the model prediction stage:
The data capture unit is used to the well group status data of the well group is obtained, by the well group of the well group for specified well group
Each well group adjusting operation that status data is adjusted respectively with the well group for the well group in operational set is combined, and is obtained multiple pre-
Test sample sheet；
The feature extraction unit is used to carry out feature extraction processing to the multiple forecast sample, and it is special to obtain multiple forecast samples
Sign；
The predicting unit be used for by the multiple forecast sample feature input respectively oil/gas well group long-term objective prediction model into
Row prediction, obtains corresponding multiple well group long-term objectives；And
The optimal adjustment operates selection unit and adjusts the corresponding long-term objective of selection in operational set from the well group for the well group
An optimal well group adjusts operation.
9. a kind of calculating equipment, comprising:
Processor；And
Memory is stored thereon with executable code, when the executable code is executed by the processor, makes the processing
Device executes the method as described in any one of claim 1-5.
10. a kind of non-transitory machinable medium, is stored thereon with executable code, when the executable code is electric
When the processor of sub- equipment executes, the processor is made to execute the method as described in any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811544048.3A CN109784539B (en) | 2018-12-17 | 2018-12-17 | Method and apparatus for index prediction model training and well group adjustment operation determination |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811544048.3A CN109784539B (en) | 2018-12-17 | 2018-12-17 | Method and apparatus for index prediction model training and well group adjustment operation determination |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109784539A true CN109784539A (en) | 2019-05-21 |
| CN109784539B CN109784539B (en) | 2021-06-15 |
Family
ID=66498019
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811544048.3A Active CN109784539B (en) | 2018-12-17 | 2018-12-17 | Method and apparatus for index prediction model training and well group adjustment operation determination |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109784539B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110439513A (en) * | 2019-07-30 | 2019-11-12 | 北京雅丹石油技术开发有限公司 | A kind of optimization method of plunger lift liquid discharging gas producing production system |
| CN111256757A (en) * | 2020-02-25 | 2020-06-09 | 深圳哈维生物医疗科技有限公司 | Medical equipment monitoring system and method based on cloud computing |
| CN112855127A (en) * | 2019-11-28 | 2021-05-28 | 北京国双科技有限公司 | Gas well accumulated liquid identification method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101415905A (en) * | 2006-04-07 | 2009-04-22 | 国际壳牌研究有限公司 | Method for optimising the production of a cluster of wells |
| CN104951842A (en) * | 2014-03-27 | 2015-09-30 | 中国石油化工股份有限公司 | Novel method for predicting oil field output |
-
2018
- 2018-12-17 CN CN201811544048.3A patent/CN109784539B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101415905A (en) * | 2006-04-07 | 2009-04-22 | 国际壳牌研究有限公司 | Method for optimising the production of a cluster of wells |
| CN104951842A (en) * | 2014-03-27 | 2015-09-30 | 中国石油化工股份有限公司 | Novel method for predicting oil field output |
Non-Patent Citations (1)
| Title |
|---|
| 倪红梅: "基于智能计算的蒸汽驱开发效果预测", 《中国博士学位论文全文数据库工程科技Ⅰ辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110439513A (en) * | 2019-07-30 | 2019-11-12 | 北京雅丹石油技术开发有限公司 | A kind of optimization method of plunger lift liquid discharging gas producing production system |
| CN110439513B (en) * | 2019-07-30 | 2021-08-31 | 北京雅丹石油技术开发有限公司 | Optimization method of plunger gas lift liquid drainage gas production system |
| CN112855127A (en) * | 2019-11-28 | 2021-05-28 | 北京国双科技有限公司 | Gas well accumulated liquid identification method and device |
| CN111256757A (en) * | 2020-02-25 | 2020-06-09 | 深圳哈维生物医疗科技有限公司 | Medical equipment monitoring system and method based on cloud computing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109784539B (en) | 2021-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zandvliet et al. | Adjoint-based well-placement optimization under production constraints | |
| US8214186B2 (en) | Oilfield emulator | |
| US8352227B2 (en) | System and method for performing oilfield simulation operations | |
| Yeten | Optimum deployment of nonconventional wells | |
| US10458207B1 (en) | Reduced-physics, data-driven secondary recovery optimization | |
| Asadollahi et al. | Waterflooding optimization using gradient based methods | |
| CN109784539A (en) | The determining method and apparatus of operation are adjusted for the training of index prediction model and well group | |
| Hanea et al. | Reservoir management under geological uncertainty using fast model update | |
| Salazar et al. | Combining decline-curve analysis and capacitance-resistance models to understand and predict the behavior of a mature naturally fractured carbonate reservoir under gas injection | |
| Asadollahi | Waterflooding optimization for improved reservoir management | |
| Edmunds et al. | Advanced solvent-additive processes by genetic optimization | |
| NO342368B1 (en) | System and method for performing oil field simulation operations | |
| Ahmed Elfeel et al. | Employing smart flow control valves for fast closed-loop reservoir management | |
| Kosmala et al. | Coupling of a surface network with reservoir simulation | |
| Sankaran et al. | Data Analytics in Reservoir Engineering | |
| Al-Ajmi et al. | Innovative pattern balancing and waterflood optimization of a super giant carbonate Mauddud reservoir, Raudhatain field, North Kuwait | |
| Al-Khazraji et al. | Development of heterogeneous immature Brownfield with Waterdrive using dynamic opportunity index: a case study from Iraqi oilfields | |
| Spyrou et al. | An approach to alternative waterflood designs and operations using streamline simulation: application to an oil field in the North German Basin | |
| Haghighat Sefat et al. | Field Management by Proactive Optimisation of Intelligent Wells-A Practical Approach | |
| Regtien | Extending the smart fields concept to enhanced oil recovery | |
| Haghighat Sefat | Proactive optimisation of intelligent wells under uncertainty | |
| Saputelli et al. | Waterflood recovery optimization using intelligent wells and decision analysis | |
| Abbas et al. | An Efficient Method to Control Oilfield Wells Through Advanced Field Management Ensemble Based Optimization Capabilities | |
| Sharma et al. | The Use of Voronoi Mapping for Production Growth in a Heavy Oil Field | |
| Rezaei et al. | Utilizing a Global Sensitivity Analysis and Data Science to Identify Dominant Parameters Affecting the Production of Wells and Development of a Reduced Order Model for the Eagle Ford Shale |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |