CN109784539B

CN109784539B - Method and apparatus for index prediction model training and well group adjustment operation determination

Info

Publication number: CN109784539B
Application number: CN201811544048.3A
Authority: CN
Inventors: 秦川; 周振华; 刘勇
Original assignee: 4Paradigm Beijing Technology Co Ltd
Current assignee: 4Paradigm Beijing Technology Co Ltd
Priority date: 2018-12-17
Filing date: 2018-12-17
Publication date: 2021-06-15
Anticipated expiration: 2038-12-17
Also published as: CN109784539A

Abstract

The invention discloses a method and a device for training a long-term index prediction model of an oil-gas well group by adopting a machine learning algorithm, and a method and a device for determining well group regulation operation by using the prediction model. The model training method comprises the following steps: obtaining a historical data set for a plurality of well groups; performing feature extraction on the historical data set to obtain a training sample feature set based on well group adjusting operation and well group state data and corresponding label data based on well group long-term indexes; and training by adopting a preset machine learning algorithm based on the training sample characteristic set and the corresponding label data to obtain a trained oil-gas well group long-term index prediction model. The obtained model can predict the long-term index of each well group, so that the actual operation of the well group can be adjusted according to the predicted long-term index. Further, the prediction model can fuse long-term and short-term indexes of the well group as tag data to perform reinforcement learning, and obtain operation parameters capable of improving the short-term and long-term indexes.

Description

Method and apparatus for index prediction model training and well group adjustment operation determination

Technical Field

The invention relates to the field of oil and gas exploitation, in particular to a method and a device for index prediction model training and well group adjustment operation determination.

Background

In the field of oil and gas exploitation of oil and gas, relatively few artesian wells can be exploited by using only natural energy. As natural energy decreases, it is necessary to inject water or specific chemicals into the formation or reservoir to increase the oil recovery from the reservoir. For heavy oil and oil sands, steam assisted gravity drainage (SAGA) is used. According to the technology, high-temperature steam is continuously injected into a stratum to heat an oil reservoir, so that the viscosity of crude oil is obviously reduced, the crude oil can flow into a production well below a steam injection well under the action of gravity and is lifted to the ground, and therefore, the heavy oil, the super heavy oil and the oil sand can be effectively exploited.

Natural gas is also buried in underground closed geological structures like crude oil, and some are stored at the same level as crude oil and some exist independently. Natural gas for reservoirs co-located with crude oil is produced along with the crude oil. For the gas reservoir only with natural gas, the exploitation method is very similar to the exploitation method of crude oil, and the method has special places.

Whether it be the exploitation of an oil, gas or oil-gas field, it is often necessary to control multiple sets of operating parameters simultaneously. Conventional operating parameters require manual adjustment by an experienced worker. Because the production conditions vary greatly between different well groups, manual adjustment of operating parameters is highly dependent on personal experience and is not highly accurate.

To this end, there is a need for an operating parameter adjustment scheme that is more accurate and can accommodate various well group conditions.

Disclosure of Invention

To address at least one of the problems described above, the present invention trains a well group long term index prediction model, which is capable of predicting long term indexes of a well group, based on a multi-well group historical data set using a machine learning algorithm, so that operation of an actual well group can be adjusted according to the predicted long term indexes. Further, the prediction model may fuse the long-term and short-term indicators of the well group as tag data for reinforcement learning, thereby obtaining operating parameters that can promote both the short-term and long-term indicators of the well group.

According to one aspect of the invention, a method for training a long-term index prediction model of an oil and gas well group is provided, which comprises the following steps: obtaining a historical data set for a plurality of well groups, each historical data in the historical data set comprising: well group adjusting operation, well group state data when the well group adjusting operation is executed and well group long-term indexes after the well group adjusting operation is executed; performing feature extraction processing on the historical data set to obtain a training sample feature set based on well group adjusting operation and well group state data and corresponding label data based on well group long-term indexes; and training by adopting a preset machine learning algorithm based on the training sample characteristic set and the corresponding label data to obtain a trained oil-gas well group long-term index prediction model. Therefore, historical data from the multiple well groups are effectively utilized through a machine learning algorithm, and internal relation between operation and long-term indexes is accurately found through iterative training, so that long-term index prediction for various well groups is realized. Here, the preset machine learning algorithm may be, for example, a Deep Neural Network (DNN) algorithm.

Preferably, the method further comprises: updating the trained oil and gas well group long-term index prediction model at least based on the well group adjustment operation aiming at the model updating well group and the obtained corresponding well group index data. Therefore, the initially trained model can be updated to further improve the accuracy of the prediction model.

Preferably, the update operation may include: performing well group adjustment operation on the model updating well group; acquiring well group state data and well group short-term indexes of the model updating well group corresponding to the well group adjusting operation; forming a prediction sample by using the well group state data of the well group updated by the model and the well group adjusting operation, and predicting the well group long-term index of the model updated well group corresponding to the well group adjusting operation by using the oil and gas well group long-term index prediction model; and obtaining sample characteristics based on the well group adjusting operation and corresponding well group state data, and fusing the obtained short-term well group indexes and the predicted long-term well group indexes as corresponding label data to obtain training data for updating the trained oil and gas well group long-term index prediction model. This further improves the degree of correlation between the long-term index and the short-term index predicted by the prediction model. The short-term indicator may include at least one of: abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and short-term oil-to-steam ratio. The well group status data may then include time-based well group status characteristics and well group historical timing status characteristics.

Preferably, performing a well group adjustment operation on the model update well group comprises: obtaining current well group state data of the model updating well group and obtaining a well group adjusting operation set aiming at the model updating well group; predicting a long-term index corresponding to each well group adjustment operation in the well group adjustment operation set based on the trained oil and gas well group long-term index prediction model; and selecting one well group adjusting operation with the optimal corresponding long-term index from the well group adjusting operation set to be executed on the model updating well group. Therefore, the training/updating efficiency of the prediction model is improved.

Preferably, obtaining historical data sets for a plurality of well groups comprises: well group adjusting operation and well group state data corresponding to each well group at different moments are obtained, and corresponding well group long-term indexes are obtained. Thus, through the time sequence training sample structure of the historical data set from a plurality of well groups, the relevance of each operation in the historical data and the long-term index is mined, and the model can predict the long-term index subsequently.

Preferably, the well group adjustment operation may be an adjustment to a SAGD well group, which may include at least one of: regulating the steam injection speed of the well group; adjusting the steam injection dryness of the well group; regulating the steam injection temperature of the well group; and well group production-injection ratio adjustment operation.

Additionally, well group long-term indicators may include: long term recovery from a well group and/or cumulative oil to steam ratio for a well group.

According to another aspect of the invention, a method of determining an adjustment operation for a hydrocarbon well group is provided, comprising: acquiring well group state data of a specified well group, and combining the well group state data of the well group with each well group adjusting operation in a well group adjusting operation set of the well group to obtain a plurality of prediction samples; performing feature extraction processing on the plurality of prediction samples to obtain a plurality of prediction sample features; inputting the characteristics of the plurality of prediction samples into an oil and gas well group long-term index prediction model respectively for prediction to obtain a plurality of corresponding well group long-term indexes; and selecting one well group adjusting operation which corresponds to the optimal long-term index from the well group adjusting operation set aiming at the well group. Thus, the optimal well group adjustment operation can be deduced back from the predictive model. Subsequently, selected well group adjustment operations may be performed on the designated well group, thereby enabling effective control of the actual well group. Preferably, the oil and gas well group long-term index prediction model can be obtained according to the training method.

Preferably, the determination method may further include: obtaining a short-term index of the designated well group under the executed well group adjusting operation; forming prediction sample data by using the well group state data of the specified well group and the executed well group adjusting operation, and predicting the well group long-term index of the specified well group corresponding to the executed well group adjusting operation by using the oil and gas well group long-term index prediction model; obtaining sample characteristics based on the executed well group adjusting operation and corresponding well group state data, and fusing the obtained short-term well group indexes and the predicted long-term well group indexes as corresponding label data to obtain training data for updating the oil and gas well group long-term index prediction model; summarizing newly-added training data for updating the long-term index prediction model of the oil and gas well group in preset time to serve as a model updating training data set; and updating the long-term index prediction model of the oil and gas well group at least based on the model updating training data set. Thus, the model can still be updated in the model using stage.

Similarly, the short-term indicators may include at least one of: abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and short-term oil-to-steam ratio. Well group long-term metrics may then include: long term recovery from a well group and/or cumulative oil to steam ratio for a well group.

According to another aspect of the present invention, there is also provided an apparatus for training a long-term index prediction model for a hydrocarbon well group, comprising: a data acquisition unit to acquire a historical data set for a plurality of well groups, each historical data in the historical data set comprising: well group adjusting operation, well group state data when the well group adjusting operation is executed and well group long-term indexes after the well group adjusting operation is executed; the characteristic extraction unit is used for carrying out characteristic extraction processing on the historical data set to obtain a training sample characteristic set based on well group adjusting operation and well group state data and corresponding label data based on well group long-term indexes; and the model training unit is used for training by adopting a preset machine learning algorithm based on the training sample characteristic set and the corresponding label data to obtain a trained oil-gas well group long-term index prediction model.

Preferably, the apparatus may further comprise: and the model updating unit is used for updating the trained oil and gas well group long-term index prediction model at least based on the well group adjusting operation aiming at the model updating well group and the obtained corresponding well group index data.

Preferably, the model updating unit includes: the well group adjusting operation executing unit is used for executing well group adjusting operation on the model updating well group; the first data acquisition subunit is used for acquiring well group state data and well group short-term indexes of the model updating well group corresponding to the well group adjusting operation; the first prediction subunit is used for forming a prediction sample by using the well group state data of the model updating well group and the well group adjusting operation, and predicting the well group long-term index of the model updating well group corresponding to the well group adjusting operation by using the oil and gas well group long-term index prediction model; and the data fusion unit is used for obtaining sample characteristics based on the well group adjusting operation and the corresponding well group state data, fusing the obtained short-term indexes of the well group and the predicted long-term indexes of the well group as corresponding label data, and obtaining training data for updating the trained oil and gas well group long-term index prediction model.

Preferably, the well group adjustment operation performing unit includes: the second data acquisition subunit is used for acquiring current well group state data of the model updating well group and acquiring a well group adjusting operation set aiming at the model updating well group; a second prediction subunit for predicting a long-term index corresponding to each well group adjustment operation in the set of well group adjustment operations based on the trained hydrocarbon well group long-term index prediction model; and the optimal well group adjusting operation selecting unit is used for selecting one corresponding well group adjusting operation with the optimal long-term index from the well group adjusting operation set to execute on the model updating well group.

Preferably, the short-term indicator comprises at least one of: abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and short-term oil-to-steam ratio.

Preferably, the data acquisition unit is used for acquiring well group adjustment operation and well group state data at different moments corresponding to each well group and acquiring a corresponding well group long-term index.

Preferably, the well group status data comprises time-based well group status characteristics and well group historical time-series status characteristics.

Preferably, the well group adjustment operation comprises at least one of: regulating the steam injection speed of the well group; adjusting the steam injection dryness of the well group; regulating the steam injection temperature of the well group; and (5) adjusting the production-injection ratio of the well group.

Preferably, the well group long-term indicators include: long term recovery from a well group and/or cumulative oil to steam ratio for a well group.

Preferably, the preset machine learning algorithm is a Deep Neural Network (DNN) algorithm.

According to yet another aspect of the present invention, there is also provided a hydrocarbon well group adjustment operation determining apparatus comprising: the prediction sample acquisition unit is used for acquiring well group state data of a specified well group and combining the well group state data of the well group with each well group adjusting operation in a well group adjusting operation set of the well group to obtain a plurality of prediction samples; the characteristic extraction unit is used for carrying out characteristic extraction processing on the plurality of prediction samples to obtain a plurality of prediction sample characteristics; the prediction unit is used for inputting the characteristics of the prediction samples into the oil and gas well group long-term index prediction model respectively for prediction to obtain a plurality of corresponding well group long-term indexes; and the optimal adjustment operation selection unit is used for selecting one well group adjustment operation which corresponds to the optimal long-term index from the well group adjustment operation set aiming at the well group.

Preferably, the apparatus may further comprise: and the well group adjusting operation executing unit is used for executing the selected well group adjusting operation on the specified well group.

Preferably, the long-term index prediction model of the oil and gas well group is obtained according to any one of the methods.

Preferably, the apparatus may further comprise: a third data acquisition subunit, configured to acquire a short-term index of the specified well group under the performed well group adjustment operation; the third prediction subunit is used for forming prediction sample data by using the well group state data of the specified well group and the executed well group adjusting operation, and predicting the well group long-term index of the specified well group corresponding to the executed well group adjusting operation by using the oil and gas well group long-term index prediction model; the data fusion unit is used for obtaining sample characteristics based on the executed well group adjusting operation and corresponding well group state data, fusing the obtained well group short-term indexes and the predicted well group long-term indexes as corresponding label data, and obtaining training data for updating the oil and gas well group long-term index prediction model; the data summarizing unit is used for summarizing training data for updating the long-term index prediction model of the oil and gas well group, which is newly added in preset time, and taking the training data as a model updating training data set; and the model updating unit is used for updating the oil and gas well group long-term index prediction model at least based on the model updating training data set.

Preferably, the short-term indicator comprises at least one of: abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and short-term oil-to-gas ratio, wherein the long-term indexes of the well group comprise: long term recovery from a well group and/or cumulative oil to steam ratio for a well group.

According to another aspect of the invention, there is also provided an oil and gas well group adjustment operation determination system, including a data acquisition unit, a feature extraction unit, a model training unit, a prediction unit, and an optimal adjustment operation selection unit, the oil and gas well group adjustment operation determination system trains an oil and gas well group long-term index prediction model, performs prediction using the model, and in a model training phase: the data acquisition unit is configured to acquire a historical data set for a plurality of well groups, each piece of historical data in the historical data set including: well group adjusting operation, well group state data when the well group adjusting operation is executed and well group long-term indexes after the well group adjusting operation is executed; the characteristic extraction unit is used for carrying out characteristic extraction processing on the historical data set to obtain a training sample characteristic set based on well group adjusting operation and well group state data and corresponding label data based on well group long-term indexes; and the model training unit is used for training by adopting a preset machine learning algorithm based on the training sample characteristic set and the corresponding label data to obtain a trained oil-gas well group long-term index prediction model, and in a model prediction stage: the data acquisition unit is used for acquiring well group state data of a specified well group and combining the well group state data of the well group with each well group adjusting operation in a well group adjusting operation set of the well group to obtain a plurality of prediction samples; the feature extraction unit is used for performing feature extraction processing on the plurality of prediction samples to obtain a plurality of prediction sample features; the prediction unit is used for inputting the characteristics of the prediction samples into the oil and gas well group long-term index prediction model respectively for prediction to obtain a plurality of corresponding well group long-term indexes; and the optimal adjustment operation selection unit selects one well group adjustment operation with the optimal long-term index from the well group adjustment operation set aiming at the well group.

Preferably, the hydrocarbon well group adjustment operation determination system further performs a model update based on predicted actual adjustment operations performed, in the model update phase: the data acquisition unit is used for acquiring a short-term index of the specified well group under the executed well group regulation operation; the characteristic extraction unit is used for forming prediction sample data by using the well group state data of the specified well group and the executed well group adjusting operation, and predicting the well group long-term index of the specified well group corresponding to the executed well group adjusting operation by using the oil and gas well group long-term index prediction model; obtaining sample characteristics based on the executed well group adjusting operation and corresponding well group state data, and fusing the obtained short-term well group indexes and the predicted long-term well group indexes as corresponding label data to obtain training data for updating the oil and gas well group long-term index prediction model; summarizing newly-added training data for updating the long-term index prediction model of the oil and gas well group in preset time to serve as a model updating training data set; and the model training unit updates the oil and gas well group long-term index prediction model at least based on the model updating training data set.

According to yet another aspect of the present invention, there is also provided a computing device comprising: a processor; and a memory having executable code stored thereon, which when executed by the processor, causes the processor to perform the method as any one of above.

According to yet another aspect of the invention, there is also provided a non-transitory machine-readable storage medium having stored thereon executable code, which when executed by a processor of an electronic device, causes the processor to perform the method of any of the above.

Thus, the present invention trains well group long term index prediction models that can predict long term indices of well groups by using machine learning algorithms based on a multi-well group historical data set, thereby enabling adjustments to be made to the operation of an actual well group based on the predicted long term indices. Further, the prediction model may fuse the long-term and short-term indicators of the well group as tag data for reinforcement learning, thereby obtaining operating parameters that can promote both the short-term and long-term indicators of the well group.

Drawings

The above and other objects, features and advantages of the present disclosure will become more apparent by describing in greater detail exemplary embodiments thereof with reference to the attached drawings, in which like reference numerals generally represent like parts throughout.

FIG. 1 illustrates a method of training a long-term index prediction model for a hydrocarbon well group according to one embodiment of the present invention.

FIG. 2 illustrates an example of predictive model training and updating.

FIG. 3 illustrates a method of determining a hydrocarbon well group adjustment operation according to one embodiment of the present invention.

FIG. 4 shows a schematic block diagram of an apparatus for training a long-term index prediction model for a hydrocarbon well group according to one embodiment of the present invention.

FIG. 5 shows a schematic block diagram of a hydrocarbon well group adjustment operation determination device in accordance with one embodiment of the present invention.

FIG. 6 illustrates one example of a hydrocarbon well group operation adjustment determination system in accordance with the present invention.

FIG. 7 shows a schematic structural diagram of a computing device according to one embodiment of the invention.

Detailed Description

Preferred embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While the preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

In oilfield exploitation, it is often necessary to control multiple sets of operating parameters simultaneously. Conventional operating parameters require manual adjustment by an experienced worker. Because the production conditions vary greatly between different well groups, manual adjustment of operating parameters is dependent on personal experience and is not highly accurate.

In order to improve the oil and gas recovery efficiency, the long-term index prediction model of the well group is trained by using a machine learning algorithm based on the multi-well group historical data set, and the model can predict the long-term index of the well group, so that the operation of the actual well group can be adjusted according to the predicted long-term index. Further, the predictive model may fuse the long-term and short-term indicators of the well group as learning objectives for reinforcement learning, thereby obtaining operational adjustments that can promote both the short-term and long-term indicators of the well group.

FIG. 1 illustrates a method of training a long-term index prediction model for a hydrocarbon well group according to one embodiment of the present invention. Herein, the "oil and gas well group long-term index prediction model" refers to a model for predicting a long-term index of an oil and gas well group. A "hydrocarbon well group" may refer to an oil well group, a gas well group, or a well group that includes both oil and gas. "well group" generally refers to the basic development unit of an oil field consisting of oil wells centered around water injection wells and connected to the surroundings. In this application, "well group" may be used to refer to one or more wells having an associated relationship, such as, for example, a single oil well, a single gas well, a single oil or gas well, or a group of wells including a water injection well or a steam well and one or more oil wells associated therewith. "Long-term indicator" refers to an indicator parameter that may reflect the long-term oil/gas recovery conditions of a well group, such as the long-term recovery of a well group and/or the cumulative oil-to-gas ratio of a SAGD well group. Additionally, it should be appreciated that in the present application, a "long-term index prediction model for a hydrocarbon well group" may refer to a prediction model dedicated to a hydrocarbon production well group, a prediction model dedicated to a gas production well group, a hydrocarbon blending model, and the like.

In step S110, historical data sets are acquired for a plurality of well groups. To improve the universality of the predictive model, data from as many well groups as possible, or from representative well groups, may be obtained during the model training phase, whereby the predictive model learns the ability to predict various well group conditions, even different well group types (e.g., single well, or SAGD well groups, or other waterflooding well groups) from the data.

Specifically, each piece of historical data in the historical data set may include: well group adjusting operation, well group state data when the well group adjusting operation is executed and well group long-term indexes after the well group adjusting operation is executed. Here, the well group adjustment operations include various types of operations for a well group that may have an effect on short-term and long-term indicators (e.g., recovery factors). In the case of a SAGD well group, the well group adjustment operations may include, for example, adjustment operations for well group production ratio, steam injection rate, steam injection dryness, and/or steam injection temperature. It should be understood that in the case of other types of well groups, such as water injection well groups, the well group conditioning operations may also include conditioning operations such as water injection rate and temperature. The well group status data may then include well group status characteristics based on time of day (e.g., corresponding time of the adjustment operation) and well group historical timing status characteristics. The above condition characteristics may include oil temperature and air pressure, etc. Since the overall state over a duration of time is more reflective of the true condition of the well group than the instantaneous state, historical time series state characteristics, such as the well-head state, e.g., average, for a period of time (e.g., 30 minutes) before the conditioning operation was performed, should be included in addition to the time-based state data.

In one embodiment of the present invention, the obtaining historical data sets for a plurality of well groups comprises: well group adjusting operation and well group state data corresponding to each well group at different moments are obtained, and corresponding well group long-term indexes are obtained.

Subsequently, in step S120, a feature extraction process is performed on the historical data set to obtain a training sample feature set and label data. In one embodiment, the data includes a training sample feature set based on well group adjustment operations and well group status data and corresponding label data based on long-term indicators for the well group. Since predictive power for long term trends needs to be learned from historical data, it is preferable to construct training samples for each well group in a time series. Step S120 may then include constructing training samples and corresponding label data with well group adjustment operations, well group status data, and well group long term goals at different times for each well group. For example, a batch of timing training samples may be constructed for well group a, a batch of timing training samples may be constructed for well group B, and so on. Therefore, the subsequent model can conveniently learn the variation trend among data well group by well group in the subsequent training stage, and the long-term index can be predicted.

In step S130, based on the training sample feature set and the corresponding label data, a preset machine learning algorithm is used for training to obtain a trained oil-gas well group long-term index prediction model. Any suitable machine learning algorithm may be employed for training herein. In one embodiment, DNN (deep learning network) may be used for training in order to more accurately grasp the association between data based on feature transfer of a multi-layer network.

Thus, the present invention implements training of the predictive model through steps S110-S130 and its preferred embodiment. In one embodiment, the trained predictive model may be updated. In this case, the training may be regarded as initial training for the prediction model. In the present invention, reinforcement learning techniques are preferably employed to update the initially trained predictive models. Through continuous learning, the use effect of the prediction model is not reduced along with the time, and a more optimized prediction result is obtained.

To this end, the training method of the present invention may further comprise updating the trained long-term index prediction model for the hydrocarbon well group based at least on well group adjustment operations for the model-updated well group and the corresponding well group index data obtained. A "model-updated well group" may refer herein to a well group used to generate training data for updating a predictive model, which may generally be a well group in actual operation. The real well group can be operated and adjusted under the guidance of the primarily trained prediction model, and the obtained corresponding data can be used for updating the prediction model in return.

In a preferred embodiment, the combination of the short-term and long-term indexes is taken as a learning target, and the prediction model is updated by using a reinforcement learning technology, so that the short-term and long-term indexes can be comprehensively considered in each adjustment of the prediction model when the prediction model is subsequently used for adjustment, and the production requirement is better met. As previously mentioned, the long-term indicator is an indicator parameter that reflects the recovery status of the well group over a relatively long period of time (e.g., half a year), such as the well group long-term recovery factor and/or the cumulative oil-to-gas ratio of the well group. Short-term indicators are then indicative conditions of the well group within a relatively short time (e.g., half an hour) after the operational adjustment, including but not limited to: abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and short-term oil-to-steam ratio.

Specifically, the model update operation may include: performing a well group adjustment operation on the model update well group; acquiring well group state data and well group short-term indexes of the model updating well group corresponding to the well group adjusting operation; forming a prediction sample by using the well group state data (the state when the well group adjusting operation is executed) of the model updating well group and the well group adjusting operation, and predicting the well group long-term index of the model updating well group corresponding to the well group adjusting operation by using the oil and gas well group long-term index prediction model; and obtaining sample characteristics based on the well group adjusting operation and corresponding well group state data, and fusing the obtained short-term well group indexes and the predicted long-term well group indexes as corresponding label data to obtain training data for updating the trained oil and gas well group long-term index prediction model.

The long-term and short-term indicators may be fused based on any suitable method. In one embodiment, the model predicted long-term metrics may be discounted (e.g., multiplied by a weight less than 1) and added to the short-term metrics. The result after the index fusion can be used as a target (namely as a label) of model training, corresponding adjusting operation and well group state data values are used as model input, model parameters are updated, and a closed-loop learning process is completed. The above steps are repeated continuously, so that the prediction model can learn model parameters closer to the model parameters capable of optimizing long and short-term indexes by self through reinforcement learning.

In actual operation, it is preferable to use an operation adjustment that can have the best effect on the index for a future period of time. Thus, the recovery factor and/or the cost and time efficiency of the model update well group itself can be optimized while improving the model update efficiency. The selection of the optimal operational adjustment may be achieved by a predictive model. In one embodiment, performing a well group adjustment operation on the model update well group may include: obtaining current well group state data of the model updating well group and obtaining a well group adjusting operation set aiming at the model updating well group; predicting a long-term index corresponding to each well group adjustment operation in the well group adjustment operation set based on the trained oil and gas well group long-term index prediction model; and selecting one well group adjusting operation with the optimal corresponding long-term index from the well group adjusting operation set to be executed on the model updating well group.

In one embodiment of the invention, said obtaining a set of well group adjustment operations for the model update well group comprises: and acquiring all possible well group adjusting operations of the current model updating well group to obtain a well group adjusting operation set. Part of the possible well group adjustment operations for the current model update well group may also be obtained empirically, resulting in a well group adjustment operation set.

FIG. 2 illustrates an example of predictive model training and updating.

As shown on the left side of fig. 2, data relating to historical manual adjustments for each well group is first obtained and the corresponding associated instantaneous or short-term conditions (e.g., recovery and cumulative oil-to-gas ratio, temperature and pressure, etc.) are adjusted, and then long-term indicators for each time period are calculated based on the conditions (or a portion thereof). For example, long term recovery and cumulative oil to gas ratios may be obtained by accumulating short term conditions.

Each well group can be considered a training sample at each time point. Herein, the time frequency of obtaining the training samples is reasonably selected, the operation dimensionality and the characteristics of the index dimensionality are extracted, and the long-term indexes of all time points are used as model training targets, namely, the long-term recovery ratio and the accumulated oil-gas ratio of a well group after certain operation is carried out at a certain time point are predicted.

The predictive model (e.g., a reinforcement learning model) can learn whether such operations are optimal choices in the current situation based on the index effects that are historically produced after each adjustment, absorb good experience and avoid errors made by manual adjustments. This is a process of initializing learning. Therefore, the model completes the initial training and has certain adjusting capacity.

Referring to the right hand side of the figure, the model, once trained, can then be applied to each of the updated well groups to predict under what conditioning operations each well group will receive the optimal recovery and cumulative gas oil ratio and make adjustments accordingly. Then, the adjusted short-term index is calculated, the long-term index obtained by the operation is predicted through the model, the two long-term indexes are fused in a certain mode and fed back to the model, and for example, the model prediction result is subjected to certain breakage (multiplied by a weight smaller than 1) and added with the short-term index. The result after the index fusion is taken as the target of model training, the data (adjusting operation and current/historical well group state) of the current state is taken as the model input, the model parameters are updated, the closed-loop learning process is completed, and the process is repeated continuously.

The model training and preferred updating method of the present invention is described above in connection with fig. 1 and 2. The purpose of training the above models is to guide real hydrocarbon production activities using the models. To this end, the invention also includes a method of determining adjustments to a hydrocarbon well group. FIG. 3 illustrates a method of determining a hydrocarbon well group adjustment operation according to one embodiment of the present invention. The above method is made for a specified set of actual wells.

In step S310, well group status data for a well group is obtained, and the well group status data for the well group is combined with each well group adjustment operation in the set of well group adjustment operations for the well group, respectively, to obtain a plurality of prediction samples. In step S320, feature extraction processing is performed on the plurality of prediction samples to obtain a plurality of prediction sample features. In step S330, the multiple prediction sample characteristics are respectively input into the oil and gas well group long-term index prediction model for prediction, so as to obtain corresponding multiple well group long-term indexes. In step S340, a well group adjustment operation that is optimal for the long-term index is selected from the set of well group adjustment operations for the well group. In a preferred embodiment, the method may further comprise performing a selected well group adjustment operation on the designated well group.

The oil and gas well group long-term index prediction model obtained based on the training and/or updating method can be used for determining the adjustment operation of the actual well group. Further, the above-described real adjustment operations and their resulting corresponding states and metric data may also be part of the update (reinforcement learning) of the prediction model itself. It can also be considered as part of the right-hand closed-loop learning shown in fig. 2, for example.

To this end, the well group adjustment determination method may further include: obtaining a short-term index of the designated well group under the executed well group adjusting operation; forming prediction sample data by using the well group state data of the specified well group and the executed well group adjusting operation, and predicting the well group long-term index of the specified well group corresponding to the executed well group adjusting operation by using the oil and gas well group long-term index prediction model; obtaining sample characteristics based on the executed well group adjusting operation and corresponding well group state data, and fusing the obtained short-term well group indexes and the predicted long-term well group indexes as corresponding label data to obtain training data for updating the oil and gas well group long-term index prediction model; summarizing newly-added training data for updating the long-term index prediction model of the oil and gas well group in preset time to serve as a model updating training data set; and updating the long-term index prediction model of the oil and gas well group at least based on the model updating training data set.

Similarly, the short-term index may include at least one of: abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and short-term oil-to-steam ratio. Well group long-term metrics may then include: long term recovery from a well group and/or cumulative oil to steam ratio for a well group.

The model training, prediction and updating methods of the present invention have been described in detail with reference to FIGS. 1-3. The methods of the present invention as described above can be implemented by their corresponding apparatuses.

FIG. 4 illustrates a schematic block diagram of a device for training a long-term index prediction model of a hydrocarbon well group (hereinafter referred to as a training device 400) according to an embodiment of the present invention. FIG. 5 shows a schematic block diagram of a hydrocarbon well group adjustment operation determination device (hereinafter referred to simply as determination device 500) according to one embodiment of the present invention. Wherein the functional blocks of the device can be implemented by hardware, software, or a combination of hardware and software implementing the principles of the present invention. It will be appreciated by those skilled in the art that the functional blocks described in fig. 4 or 5 may be combined or divided into sub-blocks to implement the principles of the invention described above. Thus, the description herein may support any possible combination, or division, or further definition of the functional modules described herein.

The training apparatus 400 shown in fig. 4 can be used to implement the training method shown in fig. 1, and the determining apparatus 500 shown in fig. 5 can be used to implement the method shown in fig. 3. Only the functional modules that the training apparatus 400 and the determining apparatus 500 can have and the operations that each functional module can perform are briefly described below, and for the details related thereto, reference may be made to the description above in conjunction with fig. 1 or 3, and details are not described here again.

As shown in fig. 4, the training apparatus 400 may include a data acquisition unit 410, a feature extraction unit 420, and a model training unit 430.

The data acquisition unit 410 may acquire a historical data set regarding a plurality of well groups, each piece of historical data in the historical data set including: well group adjusting operation, well group state data when the well group adjusting operation is executed and well group long-term indexes after the well group adjusting operation is executed. Feature extraction unit 420 may perform feature extraction processing on the historical data set to obtain a training sample feature set based on well group adjustment operations and well group status data and corresponding label data based on well group long-term indicators. The model training unit 430 may perform training using a preset machine learning algorithm based on the training sample feature set and the corresponding label data to obtain a trained oil and gas well group long-term index prediction model. The preset machine learning algorithm may be, for example, a Deep Neural Network (DNN) algorithm.

In a preferred example, the training apparatus may further include a model updating unit (not shown in the figure). The model update unit may update the trained hydrocarbon well group long term index prediction model based at least on well group adjustment operations for the model update well group and the obtained corresponding well group index data.

Specifically, the model updating unit may include: the well group adjustment operation execution unit, the first data acquisition subunit, the first prediction subunit and the data fusion unit (not shown in the figure).

The well group adjustment operation performing unit may perform a well group adjustment operation on the model update well group. The first data acquisition subunit may acquire well group status data and well group short-term indicators of the model-updated well group corresponding to the well group adjustment operation. The first prediction subunit may constitute a prediction sample with the model-updated well group status data and the well group adjustment operation for the well group, and predict a well group long-term index for the model-updated well group corresponding to the well group adjustment operation using the hydrocarbon well group long-term index prediction model. The data fusion unit may obtain sample characteristics based on the well group adjustment operation and corresponding well group status data, fuse the obtained well group short-term index and the predicted well group long-term index as corresponding tag data, and obtain training data for updating the trained oil and gas well group long-term index prediction model.

In one embodiment, the well group adjustment operation performing unit may include a second data obtaining subunit, a second predicting subunit, and an optimal well group adjustment operation selecting unit (not shown in the drawings).

The second data acquisition subunit may acquire current well group status data for the model update well group and acquire a set of well group adjustment operations for the model update well group. The second prediction subunit may predict a long-term indicator corresponding to each well group adjustment operation in the set of well group adjustment operations based on the trained hydrocarbon well group long-term indicator prediction model. And the optimal well group adjusting operation selecting unit is used for selecting one corresponding well group adjusting operation with the optimal long-term index from the well group adjusting operation set to execute on the model updating well group.

In a preferred example, the data acquisition unit 410 is configured to acquire well group adjustment operation and well group status data at different times corresponding to each well group and acquire a corresponding well group long-term index.

The short-term indicator may include at least one of: abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and short-term oil-to-steam ratio. The well group status data may include time-of-day based well group status characteristics and well group historical timing status characteristics.

The well group adjustment operation may include at least one of: regulating the steam injection speed of the well group; adjusting the steam injection dryness of the well group; regulating the steam injection temperature of the well group; and (5) adjusting the production-injection ratio of the well group. Preferably, the well group long-term indicators include: long term recovery from a well group and/or cumulative oil to steam ratio for a well group.

It should be understood that the first and second subunits (e.g., the first data acquisition subunit, the second data acquisition subunit) involved in the above-described updating and operation execution may be separate units or may be the same units as the corresponding units in the model training, e.g., the data acquisition unit used for the model training may also be used for the model updating and data acquisition in the well group practice. Accordingly, the same prediction unit can be used for model prediction in the updating and the real operation.

As shown in fig. 5, the determining apparatus 500 of the present invention may include a prediction sample acquiring unit 510, a feature extracting unit 520, a predicting unit 530, and an optimal adjustment operation selecting unit 540.

The prediction sample obtaining unit 510 may obtain well group status data of a given well group, and combine the well group status data of the well group with each well group adjustment operation in the well group adjustment operation set for the well group, respectively, to obtain a plurality of prediction samples. The feature extraction unit 520 may perform feature extraction processing on the plurality of prediction samples to obtain a plurality of prediction sample features. The prediction unit 530 may input the multiple predicted sample characteristics into the oil and gas well group long-term index prediction model respectively for prediction, so as to obtain multiple corresponding well group long-term indexes. The optimal adjustment operation selecting unit 540 may select one well group adjustment operation that is optimal for the long-term index from the well group adjustment operation set for the well group.

In one embodiment, the determining means 500 may further comprise: a well group adjustment operation performing unit (not shown in the drawings). The well group adjustment operation performing unit may perform the selected well group adjustment operation on the designated well group.

The oil and gas well group long-term index prediction model is obtained according to any one method.

In a preferred embodiment, the apparatus may further include a third data obtaining subunit, a third predicting subunit, a data fusing unit, a data summarizing unit, and a model updating unit (not shown in the figure).

The third data acquisition subunit may acquire a short-term indicator of the specified well group under the performed well group adjustment operation. The third prediction subunit may constitute prediction sample data with the well group status data of the specified well group and the performed well group adjustment operation, and predict a well group long-term index of the specified well group corresponding to the performed well group adjustment operation using the oil and gas well group long-term index prediction model. The data fusion unit can obtain sample characteristics based on the executed well group adjusting operation and corresponding well group state data, fuse the obtained well group short-term indexes and the predicted well group long-term indexes as corresponding label data, and obtain training data for updating the oil and gas well group long-term index prediction model. The data summarizing unit can summarize newly added training data for updating the long-term index prediction model of the oil and gas well group in preset time to serve as a model updating training data set. A model update unit may update the long-term index prediction model for the hydrocarbon well group based at least on the model update training dataset.

The short-term indicator may include at least one of: abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and short-term oil-to-gas ratio, and the well group long-term indexes comprise: long term recovery from a well group and/or cumulative oil to steam ratio for a well group.

In one embodiment, the determination device shown in FIG. 5 may be combined with the training device shown in FIG. 4 to form a hydrocarbon well group operation adjustment determination system. FIG. 6 illustrates one example of a hydrocarbon well group operation adjustment determination system in accordance with the present invention. As shown in fig. 6, the system 600 may include a data acquisition unit 610, a feature extraction unit 620, a model training unit 630, a prediction unit 640, and an optimal adjustment operation selection unit 650. The system 600 trains a long-term index prediction model using historical data from each well group and obtains optimal operation adjustments for the actual well group through predictions made using the prediction model to improve production efficiency of the well group and achieve cost optimization. Further, the system 600 may further update the predictive model based on the well group status and long and short term indicator changes caused by the optimization adjustments described above, thereby implementing a closed loop learning process.

Specifically, in the model training phase: data acquisition unit 610 may be configured to acquire a historical data set for a plurality of well groups, each historical data in the historical data set including: well group adjusting operation, well group state data when the well group adjusting operation is executed and well group long-term indexes after the well group adjusting operation is executed. Feature extraction unit 620 may be configured to perform feature extraction processing on the historical data set to obtain a training sample feature set based on well group adjustment operations and well group status data and corresponding label data based on well group long-term indicators. The model training unit 630 may perform training using a preset machine learning algorithm based on the training sample feature set and the corresponding label data to obtain a trained oil and gas well group long-term index prediction model. Thus, the system obtains a trained predictive model.

Subsequently, in the model prediction phase: the data obtaining unit 610 may be configured to obtain, for a specified well group, well group status data of the well group, and combine the well group status data of the well group with each well group adjustment operation in a well group adjustment operation set for the well group, respectively, to obtain a plurality of prediction samples; the feature extraction unit 620 may be configured to perform feature extraction processing on the multiple prediction samples to obtain multiple prediction sample features; the prediction unit 640 may be configured to input the multiple prediction sample characteristics into the oil and gas well group long-term index prediction model respectively for prediction, so as to obtain multiple corresponding well group long-term indexes; and the optimal adjustment operation selection unit 650 selects one well group adjustment operation that is optimal for the long-term index from the well group adjustment operation set for the well group.

Further, the hydrocarbon well group adjustment operation determination system 600 may also perform model updates based on actual adjustment operations that are predicted to be performed. Thus, in the model update phase: the data acquisition unit 610 may be configured to acquire a short-term indicator of the specified well group under the performed well group adjustment operation; the feature extraction unit 620 may be configured to construct prediction sample data with the well group status data of the specified well group and the performed well group adjustment operation, and predict a well group long-term index of the specified well group corresponding to the performed well group adjustment operation using the hydrocarbon well group long-term index prediction model; obtaining sample characteristics based on the executed well group adjusting operation and corresponding well group state data, and fusing the obtained short-term well group indexes and the predicted long-term well group indexes as corresponding label data to obtain training data for updating the oil and gas well group long-term index prediction model; summarizing newly-added training data for updating the long-term index prediction model of the oil and gas well group in preset time to serve as a model updating training data set; model training unit 630 updates the long-term index prediction model for the hydrocarbon well group based at least on the model update training dataset.

In one embodiment, centralized management of online real-time sample data acquisition and prediction and operational throttling or other scenarios involving predictive data aggregation is performed, such as on-line deployment of trained predictive models and on-line based on model predictive service APIs.

FIG. 7 illustrates a schematic structural diagram of a computing device according to an embodiment of the invention.

Referring to fig. 7, computing device 700 includes memory 710 and processor 720.

Processor 720 may be a multi-core processor or may include multiple processors. In some embodiments, processor 720 may include a general-purpose host processor and one or more special purpose coprocessors such as a Graphics Processor (GPU), Digital Signal Processor (DSP), or the like. In some embodiments, processor 720 may be implemented using custom circuits, such as an Application Specific Integrated Circuit (ASIC) or a Field Programmable Gate Array (FPGA).

The memory 710 may include various types of storage units, such as system memory, Read Only Memory (ROM), and permanent storage. Wherein the ROM may store static data or instructions that are required by processor 720 or other modules of the computer. The persistent storage device may be a read-write storage device. The persistent storage may be a non-volatile storage device that does not lose stored instructions and data even after the computer is powered off. In some embodiments, the persistent storage device employs a mass storage device (e.g., magnetic or optical disk, flash memory) as the persistent storage device. In other embodiments, the permanent storage may be a removable storage device (e.g., floppy disk, optical drive). The system memory may be a read-write memory device or a volatile read-write memory device, such as a dynamic random access memory. The system memory may store instructions and data that some or all of the processors require at runtime. In addition, the memory 710 may include any combination of computer-readable storage media, including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), magnetic and/or optical disks, may also be employed. In some embodiments, memory 710 may include a removable storage device that is readable and/or writable, such as a Compact Disc (CD), a digital versatile disc read only (e.g., DVD-ROM, dual layer DVD-ROM), a Blu-ray disc read only, an ultra-dense disc, a flash memory card (e.g., SD card, min SD card, Micro-SD card, etc.), a magnetic floppy disk, or the like. Computer-readable storage media do not contain carrier waves or transitory electronic signals transmitted by wireless or wired means.

The memory 710 has stored thereon executable code that, when processed by the processor 720, causes the processor 720 to perform the methods described above.

The method and apparatus for index predictive model training and well group adjustment operation determination according to the present invention has been described in detail above with reference to the accompanying drawings.

Furthermore, the method according to the invention may also be implemented as a computer program or computer program product comprising computer program code instructions for carrying out the above-mentioned steps defined in the above-mentioned method of the invention.

Alternatively, the invention may also be embodied as a non-transitory machine-readable storage medium (or computer-readable storage medium, or machine-readable storage medium) having stored thereon executable code (or a computer program, or computer instruction code) which, when executed by a processor of an electronic device (or computing device, server, etc.), causes the processor to perform the steps of the above-described method according to the invention.

Those of skill would further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the disclosure herein may be implemented as electronic hardware, computer software, or combinations of both.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems and methods according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

Having described embodiments of the present invention, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen in order to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method of training an index prediction model, comprising:

obtaining a historical data set for a plurality of well groups, each historical data in the historical data set comprising: well group adjusting operation, well group state data when the well group adjusting operation is executed and well group long-term indexes after the well group adjusting operation is executed;

performing feature extraction processing on the historical data set to obtain a training sample feature set based on well group adjusting operation and well group state data and corresponding label data based on well group long-term indexes;

training by adopting a preset machine learning algorithm based on the training sample characteristic set and the corresponding label data to obtain a trained oil-gas well group long-term index prediction model;

forming a prediction sample by using well group state data of the model updating well group and well group adjusting operation, and predicting the well group long-term index of the model updating well group by using the oil and gas well group long-term index prediction model; and

and acquiring sample characteristics based on the well group adjusting operation of the model updating well group and corresponding well group state data, and fusing the well group short-term index and the predicted well group long-term index after the well group adjusting operation is performed on the model updating well group as corresponding label data to acquire training data for updating the trained oil and gas well group long-term index prediction model.

2. The method of claim 1, further comprising:

performing a well group adjustment operation on the model update well group;

and acquiring well group state data and well group short-term indexes of the model updating well group corresponding to the well group adjusting operation.

3. The method of claim 2, wherein performing a well group adjustment operation on the model update well group comprises:

obtaining current well group state data of the model updating well group and obtaining a well group adjusting operation set aiming at the model updating well group;

predicting a long-term index corresponding to each well group adjustment operation in the well group adjustment operation set based on the trained oil and gas well group long-term index prediction model;

and selecting one well group adjusting operation with the optimal corresponding long-term index from the well group adjusting operation set to be executed on the model updating well group.

4. The method of claim 1, wherein the short-term indicators comprise at least one of:

abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and short-term oil-to-steam ratio.

5. The method of claim 1, wherein obtaining historical data sets for a plurality of well groups comprises:

well group adjusting operation and well group state data corresponding to each well group at different moments are obtained, and corresponding well group long-term indexes are obtained.

6. The method of claim 5, wherein the well group status data comprises time-of-day based well group status characteristics and well group historical timing status characteristics.

7. The method of claim 1, wherein the well group adjustment operation comprises at least one of:

regulating the steam injection speed of the well group;

adjusting the steam injection dryness of the well group;

regulating the steam injection temperature of the well group;

and (5) adjusting the production-injection ratio of the well group.

8. The method of claim 1, wherein the well group long term indicators comprise:

long term recovery from a well group and/or cumulative oil to steam ratio for a well group.

9. The method of claim 1, wherein the preset machine learning algorithm is a Deep Neural Network (DNN) algorithm.

10. A method of determining an adjustment operation for a hydrocarbon well group, comprising:

acquiring well group state data of a specified well group, and combining the well group state data of the well group with each well group adjusting operation in a well group adjusting operation set of the well group to obtain a plurality of prediction samples;

performing feature extraction processing on the plurality of prediction samples to obtain a plurality of prediction sample features;

inputting the characteristics of the plurality of prediction samples into an oil and gas well group long-term index prediction model respectively for prediction to obtain a plurality of corresponding well group long-term indexes;

selecting a well group adjusting operation with the optimal long-term index from the well group adjusting operation set aiming at the well group;

forming prediction sample data by using the well group state data of the specified well group and the selected well group adjusting operation, and predicting the well group long-term index of the specified well group corresponding to the selected well group adjusting operation by using the oil and gas well group long-term index prediction model; and

and obtaining sample characteristics based on the selected well group adjusting operation and the corresponding well group state data, and fusing the well group short-term index and the predicted well group long-term index after the selected well group adjusting operation is performed on the specified well group as corresponding label data to obtain training data for updating the oil and gas well group long-term index prediction model.

11. The method of claim 10, further comprising:

performing the selected well group adjustment operation on the designated well group.

12. The method of claim 11 wherein the hydrocarbon well group long term index prediction model is derived from any one of claims 1-9.

13. The method of claim 11, further comprising:

obtaining a short-term index of the designated well group under the executed well group adjusting operation;

summarizing newly-added training data for updating the long-term index prediction model of the oil and gas well group in preset time to serve as a model updating training data set;

and updating the long-term index prediction model of the oil and gas well group at least based on the model updating training data set.

14. The method of claim 10, wherein,

the short-term indicators include at least one of: abnormal conditions of the well group; oil outlet efficiency; steam injection cost; short-term recovery; and the short-term oil-to-gas ratio,

the well group long-term indicators include: long term recovery from a well group and/or cumulative oil to steam ratio for a well group.

15. An apparatus for training an index prediction model, comprising:

a data acquisition unit to acquire a historical data set for a plurality of well groups, each historical data in the historical data set comprising: well group adjusting operation, well group state data when the well group adjusting operation is executed and well group long-term indexes after the well group adjusting operation is executed;

the characteristic extraction unit is used for carrying out characteristic extraction processing on the historical data set to obtain a training sample characteristic set based on well group adjusting operation and well group state data and corresponding label data based on well group long-term indexes;

the model training unit is used for training by adopting a preset machine learning algorithm based on the training sample characteristic set and the corresponding label data to obtain a trained oil-gas well group long-term index prediction model;

the first prediction subunit is used for forming a prediction sample by using the well group state data and the well group adjusting operation of the model updating well group and predicting the well group long-term index of the model updating well group by using the oil and gas well group long-term index prediction model; and

and the data fusion unit is used for obtaining sample characteristics based on the well group adjusting operation of the model updating well group and corresponding well group state data, fusing the well group short-term index and the predicted well group long-term index after the well group adjusting operation is performed on the model updating well group as corresponding label data, and obtaining training data for updating the trained oil and gas well group long-term index prediction model.

16. The apparatus of claim 15, further comprising:

the well group adjusting operation executing unit is used for executing well group adjusting operation on the model updating well group;

and the first data acquisition subunit is used for acquiring well group state data and well group short-term indexes of the model updating well group corresponding to the well group adjusting operation.

17. The apparatus as defined in claim 16, wherein the well group adjustment operation performing unit comprises:

the second data acquisition subunit is used for acquiring current well group state data of the model updating well group and acquiring a well group adjusting operation set aiming at the model updating well group;

a second prediction subunit for predicting a long-term index corresponding to each well group adjustment operation in the set of well group adjustment operations based on the trained hydrocarbon well group long-term index prediction model;

and the optimal well group adjusting operation selecting unit is used for selecting one corresponding well group adjusting operation with the optimal long-term index from the well group adjusting operation set to execute on the model updating well group.

18. The apparatus of claim 15, wherein the short-term indicators comprise at least one of:

19. The apparatus of claim 15, wherein the data acquisition unit is configured to acquire well group adjustment operation and well group status data at different times for each well group and to acquire a corresponding well group long term indicator.

20. The apparatus of claim 19, wherein the well group status data comprises time-of-day based well group status characteristics and well group historical timing status characteristics.

21. The apparatus of claim 15, wherein the well group adjustment operation comprises at least one of:

regulating the steam injection speed of the well group;

adjusting the steam injection dryness of the well group;

regulating the steam injection temperature of the well group;

and (5) adjusting the production-injection ratio of the well group.

22. The apparatus of claim 15, wherein the well group long term indicators comprise:

23. The apparatus of claim 15, wherein the preset machine learning algorithm is a Deep Neural Network (DNN) algorithm.

24. An oil and gas well group adjustment operation determination device comprising:

the prediction sample acquisition unit is used for acquiring well group state data of a specified well group and combining the well group state data of the well group with each well group adjusting operation in a well group adjusting operation set of the well group to obtain a plurality of prediction samples;

the characteristic extraction unit is used for carrying out characteristic extraction processing on the plurality of prediction samples to obtain a plurality of prediction sample characteristics;

the prediction unit is used for inputting the characteristics of the prediction samples into the oil and gas well group long-term index prediction model respectively for prediction to obtain a plurality of corresponding well group long-term indexes;

the optimal adjustment operation selection unit is used for selecting a well group adjustment operation with the optimal long-term index from the well group adjustment operation set aiming at the well group;

the third prediction subunit is used for forming prediction sample data by using the well group state data of the specified well group and the selected well group adjusting operation, and predicting the well group long-term index of the specified well group corresponding to the selected well group adjusting operation by using the oil and gas well group long-term index prediction model; and

and the data fusion unit is used for obtaining sample characteristics based on the selected well group adjusting operation and the corresponding well group state data, fusing the well group short-term index and the predicted well group long-term index after the selected well group adjusting operation is performed on the specified well group as corresponding label data, and obtaining training data for updating the oil and gas well group long-term index prediction model.

25. The apparatus of claim 24, further comprising:

and the well group adjusting operation executing unit is used for executing the selected well group adjusting operation on the specified well group.

26. The apparatus of claim 24 wherein the hydrocarbon well group long term index prediction model is obtained according to the method of any one of claims 1-9.

27. The apparatus of claim 25, further comprising:

a third data acquisition subunit, configured to acquire a short-term index of the specified well group under the performed well group adjustment operation;

the data summarizing unit is used for summarizing training data for updating the long-term index prediction model of the oil and gas well group, which is newly added in preset time, and taking the training data as a model updating training data set;

and the model updating unit is used for updating the oil and gas well group long-term index prediction model at least based on the model updating training data set.

28. The apparatus of claim 24, wherein,

29. An oil and gas well group adjusting operation determining system comprises a data obtaining unit, a characteristic extracting unit, a model training unit, a predicting unit and an optimal adjusting operation selecting unit, wherein the oil and gas well group adjusting operation determining system trains an oil and gas well group long-term index predicting model and uses the model to predict,

in the model training phase:

the data acquisition unit is configured to acquire a historical data set for a plurality of well groups, each piece of historical data in the historical data set including: well group adjusting operation, well group state data when the well group adjusting operation is executed and well group long-term indexes after the well group adjusting operation is executed;

the characteristic extraction unit is used for carrying out characteristic extraction processing on the historical data set to obtain a training sample characteristic set based on well group adjusting operation and well group state data and corresponding label data based on well group long-term indexes; and is

The model training unit is used for training by adopting a preset machine learning algorithm based on the training sample characteristic set and the corresponding label data to obtain a trained oil-gas well group long-term index prediction model,

in the model prediction phase:

the data acquisition unit is used for acquiring well group state data of a specified well group and combining the well group state data of the well group with each well group adjusting operation in a well group adjusting operation set of the well group to obtain a plurality of prediction samples;

the feature extraction unit is used for performing feature extraction processing on the plurality of prediction samples to obtain a plurality of prediction sample features;

the prediction unit is used for inputting the characteristics of the prediction samples into the oil and gas well group long-term index prediction model respectively for prediction to obtain a plurality of corresponding well group long-term indexes; and is

The optimal adjustment operation selection unit selects a well group adjustment operation with an optimal long-term index from the well group adjustment operation set aiming at the well group,

the hydrocarbon well group conditioning operation determination system also performs model updates based on predicted actual conditioning operations to be performed,

in the model update phase:

the characteristic extraction unit is used for forming prediction sample data by using the well group state data of the specified well group and the selected well group adjusting operation, and predicting the well group long-term index of the specified well group corresponding to the selected well group adjusting operation by using the oil and gas well group long-term index prediction model;

30. The system of claim 29, wherein,

in the model update phase:

the data acquisition unit is used for acquiring a short-term index of the specified well group under the executed well group regulation operation;

and the model training unit updates the oil and gas well group long-term index prediction model at least based on the model updating training data set.

31. A computing device, comprising:

a processor; and

a memory having executable code stored thereon, which when executed by the processor, causes the processor to perform the method of any of claims 1-14.

32. A non-transitory machine-readable storage medium having stored thereon executable code, which when executed by a processor of an electronic device, causes the processor to perform the method of any of claims 1-14.