Method and device for optimizing oil field production working parameters in complex oil reservoir state

Abstract

The invention discloses an oil field production working parameter optimization method and device in a complex oil reservoir state, and the method comprises the steps: inputting a target state parameter into a target strategy model, and outputting a plurality of target working parameter combinations; the multiple sets of target working parameter combinations are input into a target production dynamic prediction model, target production dynamic data in the whole oil field production process are output, and the target production dynamic prediction model and the target strategy model are obtained through interactive training; determining a corresponding target production index according to the target production dynamic data, and determining a corresponding target net present value according to the target production index; and according to the target net present value and the target production index, screening a first target working parameter combination reaching the standard target production index from the multiple groups of target working parameter combinations. When the oil field production working parameters are optimized, the standard production indexes can be achieved and the net present value can be maximized at the same time in the complex and changeable oil reservoir state, and the flexibility, efficiency and accuracy of optimization are high.

Description

Technical Field

[0001] The invention relates to the technical field of tight oil reservoir development, and in particular to a method and device for optimizing oil field production working parameters under complex oil reservoir conditions. Background Art

[0002] Tight oil reservoirs are an important oil resource with broad prospects for exploration and development. The development method of tight oil reservoirs is similar to that of shale gas, and horizontal well fracturing technology is often used. However, since tight oil reservoirs are different from conventional oil reservoirs, it is difficult to establish inter-well displacement relationships. After tight oil reservoirs are fractured by horizontal well fracturing technology, the development effect is still poor. The development effect of tight oil reservoirs in the oil field can be improved by optimizing the production working parameters of the oil field.

[0003] However, due to the complexity of the actual production environment of oil fields and the complex and changeable reservoir conditions, the optimization algorithms used in existing technologies cannot simultaneously achieve standard production indicators and maximize the net present value under complex and changeable reservoir conditions when optimizing oil field production working parameters. The flexibility, efficiency and accuracy of the optimization are low.

[0004] To address the above problems, no effective solution has been proposed yet. Summary of the invention

[0005] The embodiments of this specification provide a method and device for optimizing oilfield production working parameters under complex reservoir conditions to solve the problem that the prior art cannot simultaneously achieve standard production indicators and maximize net present value under complex and changeable reservoir conditions when optimizing oilfield production working parameters, and the optimization flexibility, efficiency and accuracy are low.

[0006] In a first aspect, the embodiments of this specification provide a method for optimizing oilfield production parameters under complex reservoir conditions, including:

[0007] Inputting target state parameters reflecting the target reservoir state into the target strategy model, and outputting multiple sets of target working parameter combinations;

[0008] Inputting multiple groups of target working parameter combinations into a target production dynamic prediction model, and outputting target production dynamic data in the entire oilfield production process, wherein the target production dynamic prediction model and the target strategy model are obtained through interactive training;

[0009] Determine the target production index corresponding to each group of target working parameter combinations according to the target production dynamic data, and determine the target net present value corresponding to each group of target working parameter combinations according to the target production index;

[0010] According to the target net present value and the target production index, a first target operating parameter combination that achieves a standard target production index is screened from the multiple groups of target operating parameter combinations, the target net present value corresponding to the first target operating parameter combination is greater than a preset net present value threshold, and the standard target production index is greater than a preset production index threshold.

[0011] In some embodiments, the target state parameters include a target injection-production ratio and a target recovery factor; each group of target operating parameter combinations is a combination of at least two target parameters including a target injection volume, a target injection rate, a target injection concentration, and a target shut-in time; the target production dynamic data includes target daily oil production data corresponding to each group of target operating parameter combinations; and the target production index includes a target cumulative oil production.

[0012] In some embodiments, the target production dynamic prediction model and the target strategy model are obtained through interactive training, including:

[0013] Inputting the randomly generated multiple sets of working parameter combinations into the initial production dynamic prediction model, outputting the first production dynamic data of the whole process of oilfield production, inputting the state parameters reflecting the expected reservoir state into the Actor network in the initial strategy model, outputting the multiple sets of first working parameter combinations at the current moment, inputting the multiple sets of first working parameter combinations into the initial production dynamic prediction model, outputting the second production dynamic data of the whole process of oilfield production;

[0014] Determine a first cumulative oil production and a first production round according to the first production dynamic data, determine a first net present value according to the first cumulative oil production, and determine a second net present value according to the second production dynamic data;

[0015] When the difference between the first net present value and the second net present value is not less than a preset difference threshold, the first cumulative oil production and the first production round are used as the first state at the current moment, and the first net present value is used as the first reward at the current moment to input into the Critic network in the initial strategy model, the Critic network is used to evaluate the multiple groups of first working parameter combinations output by the Actor network at the current moment according to the first state and the first reward, obtain the evaluation result at the current moment and feed it back to the Actor network, and the Actor network is used to adjust its own strategy according to the evaluation result and output multiple groups of new working parameter combinations based on the adjusted strategy;

[0016] Inputting multiple sets of new working parameter combinations into the initial production performance prediction model, outputting third production performance data under the entire oil field production process, and determining a third net present value based on the third production performance data;

[0017] Determine whether the difference between the first net present value and the third net present value is less than a preset difference threshold. If so, use the initial production dynamic prediction model as the trained target production dynamic prediction model and the initial strategy model as the trained target strategy model.

[0018] In some embodiments, the initial production dynamic prediction model is constructed by a long short-term memory network and a self-attention module. Accordingly, the randomly generated multiple groups of working parameter combinations are input into the initial production dynamic prediction model to output the first production dynamic data under the whole process of oilfield production, including:

[0019] Inputting the randomly generated multiple groups of working parameter combinations into the long short-term memory network, and outputting the production dynamic data corresponding to each group of working parameter combinations;

[0020] Inputting the production dynamic data into a self-attention module, determining the importance of the production dynamic data based on the self-attention module, and determining the weight value of the production dynamic data corresponding to each group of working parameter combinations according to the normalized value of the importance;

[0021] The weight value is weighted and summed with the corresponding production dynamic data to obtain the first production dynamic data in the whole process of oil field production and output the first production dynamic data.

[0022] In some embodiments, the method further comprises:

[0023] When the difference between the first net present value and the second net present value is not less than a preset difference threshold, obtaining a second state at the next moment, and inputting the first state, the second state, and the first reward into an initial strategy model;

[0024] Accordingly, obtaining the evaluation result at the current moment includes:

[0025] Determine a first evaluation result at a current moment according to the first state and a first action performed by the Actor network in the first state, and determine a second evaluation result at a next moment according to the second state and a second action performed by the Actor network in the second state, wherein the first action represents a plurality of first working parameter combinations, and the second action represents a plurality of second working parameter combinations;

[0026] According to the first evaluation result, the second evaluation result, and the first reward, multiple groups of first working parameter combinations output by the Actor network at the current moment are evaluated to obtain the evaluation result at the current moment.

[0027] In some embodiments, determining the target production index corresponding to each group of target working parameter combinations according to the target production dynamic data includes:

[0028] Determine the target cumulative oil production corresponding to each set of target operating parameters according to the target daily oil production data corresponding to each set of target operating parameter combinations;

[0029] Determining the target net present value corresponding to each group of target working parameter combinations according to the target production index includes:

[0030] The target net present value corresponding to each set of target operating parameter combinations is determined based on the target cumulative oil production, the pre-acquired discount rate, the crude oil price, the throughput work cost corresponding to each set of target operating parameter combinations, and the total number of production years.

[0031] In some embodiments, the step of selecting a first target operating parameter combination that achieves a standard target production index from the plurality of sets of target operating parameter combinations according to the target net present value and the target production index comprises:

[0032] According to the target net present value and target production index corresponding to each group of target working parameter combinations, each group of target working parameter combinations is sorted in descending order;

[0033] The target operating parameter combination corresponding to the target net present value being greater than the preset net present value threshold and the target production index being greater than the preset production index threshold is selected from the sorting results as the first target operating parameter combination to achieve the standard target production index.

[0034] In a second aspect, the embodiments of this specification also provide an oilfield production working parameter optimization device under complex reservoir conditions, including:

[0035] A target strategy model output module is used to input target state parameters reflecting the target reservoir state into the target strategy model and output multiple sets of target working parameter combinations;

[0036] A target production dynamic prediction model output module is used to input multiple sets of target working parameter combinations into the target production dynamic prediction model, and output the target production dynamic data under the whole process of oilfield production. The target production dynamic prediction model and the target strategy model are obtained through interactive training;

[0037] A target net present value determination module, used to determine the target production index corresponding to each group of target working parameter combinations according to the target production dynamic data, and determine the target net present value corresponding to each group of target working parameter combinations according to the target production index;

[0038] A screening module is used to screen a first target operating parameter combination that achieves a standard target production indicator from the multiple groups of target operating parameter combinations based on the target net present value and the target production indicator, wherein the target net present value corresponding to the first target operating parameter combination is greater than a preset net present value threshold, and the standard target production indicator is greater than a preset production indicator threshold.

[0039] On the third aspect, an embodiment of the present specification also provides an electronic device, including a memory and a processor, wherein the processor and the memory are communicatively connected to each other, the memory stores computer instructions, and the processor implements the steps of the above-mentioned method for optimizing oilfield production working parameters under complex reservoir conditions by executing the computer instructions.

[0040] In a fourth aspect, the embodiments of this specification also provide a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the steps of the above-mentioned method for optimizing oilfield production working parameters under complex reservoir conditions.

[0041] The embodiment of this specification provides a method and device for optimizing oilfield production working parameters under complex reservoir conditions. First, the target state parameters reflecting the target reservoir state are input into the target strategy model, and multiple groups of target working parameter combinations are output. Then, the multiple groups of target working parameter combinations are input into the target production dynamic prediction model, and the target production dynamic data under the whole process of oilfield production are output. The target production dynamic prediction model and the target strategy model are obtained through interactive training. Then, according to the target production dynamic data, the target production index corresponding to each group of target working parameter combinations is determined, and according to the target production index, the target net present value corresponding to each group of target working parameter combinations is determined. Finally, according to the target net present value and the target production index, the first target working parameter combination that reaches the standard target production index is selected from the multiple groups of target working parameter combinations, and the target net present value corresponding to the first target working parameter combination is greater than the preset net present value threshold, and the standard target production index is greater than the preset production index threshold. In the embodiment of this specification, by inputting the target state parameters reflecting the target reservoir state into the target strategy model, different reservoir states can be adapted, and finally the optimization of oilfield production working parameters under complex and changeable reservoir states can be achieved. The multiple groups of working parameters output by the target strategy model are input into the target production dynamic prediction model to output the target production dynamic data of the whole oilfield production process, and then the target production index is determined based on the target production dynamic data, and the target net present value is determined based on the target production index. Therefore, the first target working parameter combination that achieves the standard target production index can be screened out based on the target net present value and the target production index, and the maximization of the target net present value of the first target working parameter combination can be ensured at the same time, thereby solving the problem that the prior art cannot achieve the standard production index and maximize the net present value under complex and changeable reservoir conditions when optimizing oilfield production working parameters, and the optimization flexibility, efficiency and accuracy are low. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0043] Figure 1 It is a flow chart of a method for optimizing oilfield production parameters under complex reservoir conditions provided by an embodiment of this specification;

[0044] Figure 2 It is a schematic diagram of the initial production dynamic prediction model provided in the embodiments of this specification;

[0045] Figure 3 is a schematic diagram of the model interactive training provided in the embodiments of this specification;

[0046] Figure 4 It is a schematic diagram of the structure of an oilfield production working parameter optimization device under complex oil reservoir conditions provided by an embodiment of this specification;

[0047] Figure 5 It is a schematic diagram of the structural composition of an electronic device provided in an embodiment of this specification. DETAILED DESCRIPTION

[0048] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0049] As mentioned above, the optimization algorithms used in the existing technology have great limitations when optimizing oil field production working parameters. They are usually optimized under fixed goals and conditions, and it is difficult to cope with complex and changeable reservoir characteristics and multi-objective optimization problems. That is, it is impossible to simultaneously achieve standard production indicators and maximize the net present value under complex and changeable reservoir conditions. The flexibility, efficiency and accuracy of the optimization are low, which leads to low overall benefits of oil field development and lack of reliable reference basis for oil field production management.

[0050] In order to solve the above problems, the embodiment of this specification provides an oilfield production working parameter optimization method under complex reservoir conditions. First, the target state parameters reflecting the target reservoir state are input into the target strategy model, and multiple groups of target working parameter combinations are output. Then, multiple groups of target working parameter combinations are input into the target production dynamic prediction model, and the target production dynamic data under the entire oilfield production process are output. The target production dynamic prediction model and the target strategy model are obtained through interactive training. Then, according to the target production dynamic data, the target production index corresponding to each group of target working parameter combinations is determined, and according to the target production index, the target net present value corresponding to each group of target working parameter combinations is determined. Finally, according to the target net present value and the target production index, the first target working parameter combination that reaches the standard target production index is selected from the multiple groups of target working parameter combinations, the target net present value corresponding to the first target working parameter combination is greater than the preset net present value threshold, and the standard target production index is greater than the preset production index threshold.

[0051] It should be noted that the terms "first", "second", etc. in the specification and claims of this application are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so as to describe the embodiments of the present application described herein.

[0052] It is understood that the above method provided in the embodiments of this specification can be applied to electronic devices, and the electronic devices can refer to electronic devices with data calculation, processing and storage capabilities. The electronic device can be a terminal such as a PC (Personal Computer), a tablet computer, a smart phone, a wearable device, an intelligent robot, etc.; it can also be a server. Among them, the server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0053] A method for optimizing oilfield production parameters under complex reservoir conditions provided by an embodiment of this specification will be introduced below in conjunction with the accompanying drawings.

[0054] Figure 1It is a flow chart of a method for optimizing oilfield production working parameters under complex reservoir conditions provided by an embodiment of this specification. Although this specification provides method operation steps or device structures as shown in the following embodiments or drawings, the method or device may include more or fewer operation steps or module units after partial merger based on routine or no creative labor. In the steps or structures that do not logically have a necessary causal relationship, the execution order of these steps or the module structure of the device is not limited to the execution order or module structure shown in the embodiments or drawings of this specification. When the method or module structure described is applied to an actual device, server or terminal product, it can be executed sequentially or in parallel according to the method or module structure shown in the embodiments or drawings (for example, a parallel processor or multi-threaded processing environment, or even a distributed processing, server cluster implementation environment). For specific implementation, refer to Figure 1 As shown, the method may include the following contents.

[0055] See also Figure 1 As shown, the embodiment of this specification provides a method for optimizing oilfield production parameters under complex reservoir conditions. In specific implementation, the method may include the following contents.

[0056] S101: inputting target state parameters reflecting the target reservoir state into the target strategy model, and outputting multiple sets of target working parameter combinations;

[0057] S102: inputting a plurality of target working parameter combinations into a target production dynamic prediction model, and outputting target production dynamic data in the whole oilfield production process, wherein the target production dynamic prediction model and the target strategy model are obtained through interactive training;

[0058] S103: determining the target production index corresponding to each group of target working parameter combinations according to the target production dynamic data, and determining the target net present value corresponding to each group of target working parameter combinations according to the target production index;

[0059] S104: Based on the target net present value and the target production index, a first target operating parameter combination that achieves a standard target production index is selected from the multiple groups of target operating parameter combinations, the target net present value corresponding to the first target operating parameter combination is greater than a preset net present value threshold, and the standard target production index is greater than a preset production index threshold.

[0060] Based on the above embodiments, by inputting the target state parameters reflecting the target reservoir state into the target strategy model, it is possible to adapt to different reservoir states, and ultimately achieve the optimization of oilfield production working parameters under complex and changeable reservoir states. The multiple groups of working parameters output by the target strategy model are then input into the target production dynamic prediction model, and the target production dynamic data under the entire oilfield production process is output. Then, based on the target production dynamic data, the target production index is determined, and based on the target production index, the target net present value is determined, so that the first target working parameter combination that meets or reaches the standard target production index can be screened out based on the target net present value and the target production index, and the target net present value of the first target working parameter combination can be maximized, thereby solving the problem that the prior art cannot simultaneously achieve the standard or specified production index and maximize the net present value under complex and changeable reservoir states when optimizing oilfield production working parameters, and the optimization flexibility, efficiency and accuracy are low.

[0061] In some embodiments, the above-mentioned target state parameters may include a target injection-production ratio and a target recovery factor; the above-mentioned groups of target working parameter combinations may be a combination of at least two target parameters among target injection volume, target injection rate, target injection concentration, and target shut-in time; the above-mentioned target production dynamic data may include target daily oil production data corresponding to each group of target working parameter combinations; the target production indicators include target cumulative oil production.

[0062] Specifically, the above-mentioned target reservoir state can be a complex and changeable tight reservoir state under a dynamic production environment. The target state parameters reflecting the target reservoir state can specifically include: target injection-production ratio and target recovery factor. Among them, the injection-production ratio reflects the water injection effect and equilibrium state of the reservoir. The higher the injection-production ratio, the better the water injection effect and the higher the reservoir recovery factor. The recovery factor refers to the ratio of the amount of oil that can be extracted from the reservoir to the amount of geological reserves. The higher the recovery factor, the higher the exploitation efficiency of the reservoir and the better the reservoir state.

[0063] The above-mentioned target strategy model can be a trained initial strategy model, and the initial strategy model can be constructed based on the DRL framework (deep reinforcement learning model framework), and the DRL framework specifically includes an Actor network and a Critic network. Among them, the Actor Network can be a neural network responsible for selecting and outputting actions, and the goal is to maximize the long-term return of the agent. The Critic Network can be a neural network responsible for evaluating the value of the current strategy, giving a value estimate for each action to guide the learning of the Actor network. When the model is actually applied, the above-mentioned target state parameters can be input into the target strategy model, and multiple sets of target working parameter combinations under the target reservoir state can be output. Each set of target working parameter combinations in the multiple sets of target working parameter combinations can be a combination of at least two target parameters among the target injection volume, target injection rate, target injection concentration, and target well soaking time, and can be randomly combined, and this specification does not make specific restrictions on this.

[0064] The above-mentioned target production dynamic prediction model can be a trained initial production dynamic prediction model, and the initial production dynamic prediction model can be constructed based on the LSTM framework, and the LSTM framework specifically includes LSTM and self-attention mechanism. Wherein, LSTM is a long short-term memory network, which is a special recursive neural network (RNN), which is mostly used for time series prediction and regression prediction. In the present invention, since the target working parameter combination (or working parameter combination) is random, the corresponding production duration of the target production dynamic prediction model (or production dynamic prediction model) output data is also different. Therefore, conventional time series prediction cannot achieve production dynamic prediction, and it is necessary to use LSTM regression prediction to perform production dynamic prediction. When the model is actually applied, the above-mentioned multiple groups of target working parameter combinations can be input into the target production dynamic prediction model, and the target production dynamic data under the whole process of oil field production can be output, that is, the target daily oil production data corresponding to each group of target working parameter combinations can be output, such as: the target daily oil production change or target daily oil production corresponding to each group of target working parameter combinations. Wherein, the target production dynamic data under the whole process of oil field production output is the target daily oil production data within different production durations.

[0065] The above-mentioned target net present value can be the target NPV. Since the learning strategy of the deep reinforcement learning model is a random strategy, the action decided (or made) by the strategy will not be the only solution, that is, there will be multiple sets of target working parameter combinations (or multiple sets of working parameters) to be decided. Correspondingly, multiple sets of target working parameter combinations can be input into the target production dynamic prediction model, and then the target NPV of each set of target working parameter combinations after the decision is calculated according to the output results of the target production dynamic prediction model (specifically, the target production index corresponding to each set of target working parameter combinations is first calculated according to the output results, and then the target NPV corresponding to each set of target working parameter combinations is calculated according to the target production index). Then, the target NPV is used as one of the considerations for screening, and finally the first target working parameter combination whose target net present value is greater than the preset net present value threshold is screened out. Among them, the preset net present value threshold can be set according to actual needs, and this specification does not make specific restrictions on this.

[0066] The above-mentioned target production index may include target cumulative oil production. When screening the first target operating parameter combination, in addition to considering NPV, the target production index corresponding to each group of target operating parameter combinations may also be considered, and the first target operating parameter combination that achieves a target production index greater than a preset production index threshold (standard target production index) is screened out, and the target net present value corresponding to the first target operating parameter combination is greater than the preset net present value threshold. Among them, the preset production index threshold can be set according to actual needs, and this specification does not specifically limit this.

[0067] In some embodiments, the target production dynamic prediction model and the target strategy model are obtained through interactive training, and when implemented specifically, may include:

[0068] S1: Input multiple sets of randomly generated working parameter combinations into the initial production dynamic prediction model, output the first production dynamic data of the whole process of oilfield production, input the state parameters reflecting the expected reservoir state into the Actor network in the initial strategy model, output multiple sets of first working parameter combinations at the current moment, input multiple sets of first working parameter combinations into the initial production dynamic prediction model, and output the second production dynamic data of the whole process of oilfield production;

[0069] S2: determining a first cumulative oil production and a first production round according to the first production dynamic data, determining a first net present value according to the first cumulative oil production, and determining a second net present value according to the second production dynamic data;

[0070] S3: When the difference between the first net present value and the second net present value is not less than a preset difference threshold, the first cumulative oil production and the first production round are used as the first state at the current moment, and the first net present value is used as the first reward at the current moment and input into the Critic network in the initial strategy model, the Critic network is used to evaluate the multiple groups of first working parameter combinations output by the Actor network at the current moment according to the first state and the first reward, obtain the evaluation result at the current moment and feed it back to the Actor network, the Actor network is used to adjust its own strategy according to the evaluation result and output multiple groups of new working parameter combinations based on the adjusted strategy;

[0071] S4: inputting multiple sets of new working parameter combinations into the initial production performance prediction model, outputting third production performance data under the entire oilfield production process, and determining a third net present value according to the third production performance data;

[0072] S5: Determine whether the difference between the first net present value and the third net present value is less than a preset difference threshold. If so, use the initial production dynamic prediction model as the trained target production dynamic prediction model and the initial strategy model as the trained target strategy model.

[0073] Specifically, the initial production dynamics prediction model (LSTM and self-attention mechanism) can be used to process multiple groups of working parameter combinations and output the first production dynamic data of the entire oilfield production process. The initial strategy model can interact with the initial production dynamics prediction model, and continuously improve the strategy of the initial strategy model itself through the interaction process, until the actions made by the initial strategy model (i.e., the working parameter combination generated by the Actor network) under different reservoir conditions can be the parameter combination that can maximize the net present value under the reservoir state. At this time, the target production dynamics prediction model and the target strategy model can be trained. However, since the strategy model may have certain errors, the actions made by the strategy model need to be further screened to ensure that the screened working parameters can achieve production indicators and maximize the net present value.

[0074] Specifically, the process of interactive training is as follows:

[0075] First, the reservoir numerical simulation software can be used to generate data sets, such as randomly generating multiple sets of working parameter combinations, and then inputting the multiple sets of working parameter combinations into the initial production dynamic prediction model (LSTM and self-attention mechanism), and outputting the first production dynamic data under the entire oilfield production process, that is, outputting the daily oil production data corresponding to each set of working parameter combinations. The state parameters reflecting the user's expected reservoir state can be input into the Actor network in the initial strategy model, and multiple sets of first working parameter combinations at the current moment (such as: moment t) (that is, the first action made by the Actor network at the current moment) are output, and then the first action made by the Actor network is input into the initial ecological dynamic prediction model, and the second production dynamic data under the entire oilfield production process is output.

[0076] Then, the first cumulative oil production and the first production round of the entire oil field production process can be determined based on the first production dynamic data. For example, the first cumulative oil production can be obtained by simply adding up the daily oil production, and the first production round can be obtained by calculating the time length. The first net present value (such as NPV) of the entire oil field production process can be determined based on the first cumulative oil production. 1 Similarly, the second cumulative oil production in the whole process of oil field production can be determined according to the second production dynamic data, and the second net present value (such as NPV) in the whole process of oil field production can be determined according to the second cumulative oil production. 2 ). NPV can be determined 1 and NPV 2 Is the difference result less than the preset difference threshold? If it is less than the preset difference threshold, it means NPV 1 and NPV 2 If the difference is not less than the preset difference threshold, that is, greater than or equal to the preset difference threshold, it means that NPV 1 and NPV 2 There are many differences between them. At this time, it can be considered that the first working parameter combination output by the Actor network at the current moment is still a certain distance away from the parameter combination that can maximize the net present value under the expected reservoir state. The first cumulative oil production and the first production round can be used as the first state at the current moment, and the first net present value can be used as the first reward at the current moment to input into the Critic network in the initial strategy model. Based on the first state and the first reward, the Critic network can evaluate the multiple sets of first working parameter combinations output by the Actor network at the current moment, so that the Actor network can adjust its own strategy according to the evaluation results, and then output multiple sets of new working parameter combinations.

[0077] After that, the above steps are repeated, that is, the multiple sets of new working parameter combinations output by the Actor network are input into the initial production dynamic prediction model, and the third production dynamic data under the whole process of oil field production is output. Then, the third cumulative oil production is determined according to the third production dynamic data, and the third net present value (such as NPV) under the whole process of oil field production is determined according to the third cumulative oil production. 3 ). Then determine the NPV 1 and NPV 3 Is the difference result less than the preset difference threshold? If it is less than the preset difference threshold, it means NPV 1 and NPV 3 If the difference is not less than the preset threshold, the above steps are repeated to calculate the NPV. X , until NPV 1 and NPV X If the difference result is less than the preset difference threshold, the interactive training ends.

[0078] The preset difference threshold can be set according to actual needs, and this specification does not make any specific limitation on this.

[0079] Among them, the first net present value (such as: NPV 1 ), the second net present value (such as: NPV 2 ), third net present value (such as: NPV 3 ) can be determined according to the following formula (1):

[0080]

[0081] Where NPV is the net present value (specifically, it can represent the first net present value, the second net present value, the third net present value, etc.); T is the total number of years of production; P t is the crude oil price in year t; Q t is the cumulative oil production in year t (when calculating the first net present value, it may be the first cumulative oil production; when calculating the second net present value, it may be the second cumulative oil production; when calculating the third net present value, it may be the third cumulative oil production, etc.); C t is the throughput work cost in year t (when calculating the first net present value, it can be specifically the total throughput work cost corresponding to multiple groups of working parameter combinations; when calculating the second net present value, it can be specifically the total throughput work cost corresponding to multiple groups of first working parameter combinations; when calculating the third net present value, it can be specifically the total throughput work cost corresponding to multiple groups of new working parameter combinations); r is the discount rate.

[0082] Among them, C t It can be calculated according to the following formula:

[0083]

[0084] Wherein, f is a cost function corresponding to each set of working parameters or each set of first working parameter combinations or each set of new working parameter combinations. For example, assuming that a certain set of working parameter combinations is a combination of injection volume, injection rate, injection concentration and shut-in time, then f may include f(I i )、f(ν i )、f(C i )、f(T i ), are the cost functions related to injection volume, injection rate, injection concentration and shut-in time; n is the total number of production rounds.

[0085] It should be noted that, assuming that the first net present value is calculated, the throughput work cost corresponding to each set of working parameter combinations can be obtained by the above formula (2), and then the total throughput work cost corresponding to multiple sets of working parameter combinations can be obtained by adding them up, which is used as C t , substituted into the above formula (1), the first net present value of the entire oil field production process is calculated.

[0086] In some embodiments, the initial production dynamic prediction model in S1 is constructed by a long short-term memory network and a self-attention module. Accordingly, the above S1 inputs the randomly generated multiple groups of working parameter combinations into the initial production dynamic prediction model to output the first production dynamic data of the entire oilfield production process. In specific implementation, it may include:

[0087] S11: inputting the randomly generated multiple groups of working parameter combinations into the long short-term memory network, and outputting the production dynamic data corresponding to each group of working parameter combinations;

[0088] S12: inputting the production dynamic data into a self-attention module, determining the importance of the production dynamic data based on the self-attention module, and determining the weight value of the production dynamic data corresponding to each group of working parameter combinations according to the normalized value of the importance;

[0089] S13: performing weighted summation on the weight value and the corresponding production dynamic data to obtain the first production dynamic data in the whole process of oil field production and output the first production dynamic data.

[0090] Specifically, the above-mentioned initial production dynamics prediction model can be composed of LSTM and self-attention mechanism, and the self-attention mechanism is the above-mentioned self-attention module or self-attention model. First, the randomly generated multiple groups of working parameter combinations can be input into the long short-term memory network LSTM, and the production dynamic data corresponding to each group of working parameter combinations can be output.

[0091] Then, the production dynamic data output by LSTM can be input into the self-attention module, and the importance of the production dynamic data corresponding to each group of working parameter combinations can be determined based on the self-attention module, and then the normalized value of the importance of the production dynamic data corresponding to each group of working parameter combinations can be calculated, and the weight value of the production dynamic data corresponding to each group of working parameter combinations can be determined according to the normalized value of the importance. For example, the similarity between the production dynamic data corresponding to each group of working parameter combinations and its own parameters can be calculated based on the self-attention module (such as: the similarity is determined by the product of the two, and the own parameters can be obtained by the self-attention module in advance through model training based on the historical working parameter combinations, which can represent important production dynamic data). The higher the similarity, the more important the production dynamic data output by the input working parameter combination is, and accordingly, a higher weight value can be assigned.

[0092] Finally, the weight value and the corresponding production dynamic data may be weightedly summed to obtain the first production dynamic data in the entire oil field production process and output the first production dynamic data.

[0093] Specifically, the above LSTM can be four layers. Through the deep structure of the four-layer LSTM, the model can extract richer sequence features and learn higher-level abstract representations layer by layer. Figure 2 As shown, the specific process of inputting the randomly generated multiple groups of working parameter combinations into the initial production performance prediction model and outputting the first production performance data under the whole oilfield production process can be as follows:

[0094] First, multiple sets of working parameter combinations are received through the sequence input layer (SequenceInput).

[0095] Next, the extracted features are passed through four stacked LSTM layers (LSTM 1 to 4) for time series modeling. The LSTM network can capture dependencies within high-dimensional data through its unique gating structure (input gate, forget gate, and output gate), and is very effective in processing complex nonlinear and long-term dependency problems in data. The specific gating mechanism contained in the LSTM unit and its working method are as follows:

[0096] Forget Gate:

[0097] f t =σ(W f ·[ht-1 ,x t ]+b f )(3)

[0098] Input Gate:

[0099] i t =σ(W i ·[h t-1 ,x t ]+b i )(4)

[0100] Candidate cell states:

[0101] C t =tanh(W C ·[h t-1 ,x t ]+b C )(5)

[0102] Update cell status:

[0103] C t =f t ·C t-1 +i t ·C t (6)

[0104] Output Gate:

[0105] o t =σ(Wo·[h t-1 ,x t ]+b o )(7)

[0106] Hidden state:

[0107] h t =o t tanh(C t )(8)

[0108] Among them, f t is the forget gate (used to control how much previous memory is forgotten); σ is the Sigmoid activation function; x t is the input at time t; h t-1 is the output at time t-1; b is the bias term; i t is the input gate (used to control how much of the current input information is written into the memory unit); C t is the candidate cell state (used to provide new candidate memory); C t is the neuron state after updating by the forget threshold and input threshold, that is, the neuron state at time t; C t-1 is the neuron state at time t-1; o tis the output gate (used to control the output of each time step).

[0109] Subsequently, the output after the LSTM layer is transformed nonlinearly by activation function 1 and activation function 2 (such as ReLU1 and ReLU2) to increase the expressive power of the model and enable it to better fit complex nonlinear relationships. In addition, the dropout layer is introduced for regularization to prevent overfitting and improve the generalization ability of the model. The dropout layer randomly discards some neurons during training, forcing the model to learn more robust feature representations. Among them, it first enters the dropout layer through activation function 1, and then enters the fully connected layer 1 and activation function 2.

[0110] Finally, the features after nonlinear activation and regularization are further mapped to the final output space through a fully connected layer 2 (FullyConnected 2). The fully connected layer 2 performs weighted summation on the multi-dimensional features and finally outputs the predicted value, that is, the daily output of the oil field, through the regression output layer. It is worth noting that in order to further improve the prediction accuracy of the model and enhance the performance of the model on the output sequence, we introduced the self-attention mechanism (SelfAttention) in the output stage. The self-attention mechanism can help the model capture important features in the output sequence. By weighting the output of each time step, the model can pay more attention to key time points and potential abnormal patterns when generating predictions.

[0111] Through such a multi-level model structure, the daily oil production changes in the entire production process can be accurately predicted.

[0112] Among them, LSTM outputs the hidden state h of the sequence t , which can be passed to the next LSTM layer or used as the final output of the network. The goal of training is to improve the accuracy of LSTM prediction by minimizing the loss function RMSE, where the minimization of the loss function RMSE is:

[0113]

[0114] Where N is the number of samples; y t is the true value (target value) at time t; is the predicted value at time t.

[0115] Among them, the Adam optimizer can be used when training the initial production dynamic prediction model. The Adam optimizer is an adaptive optimization algorithm that can update the weights and biases of the network by calculating the momentum of the gradient and the adaptive learning rate.

[0116] In some embodiments, the above S3 may further include:

[0117] S31: when the difference between the first net present value and the second net present value is not less than a preset difference threshold, obtaining a second state at a next moment and a second reward at a next moment, and inputting the first state, the second state, the first reward, and the second reward into an initial strategy model;

[0118] Accordingly, obtaining the evaluation result at the current moment may include, in specific implementation, the following steps:

[0119] S32: determining a first evaluation result at a current moment according to the first state and a first action performed by the Actor network in the first state, and determining a second evaluation result at a next moment according to the second state and a second action performed by the Actor network in the second state, wherein the first action represents a plurality of first working parameter combinations, and the second action represents a plurality of second working parameter combinations;

[0120] S33: According to the first evaluation result, the second evaluation result, the first reward, and the second reward, multiple groups of first working parameter combinations output by the Actor network at the current moment are evaluated to obtain the evaluation result at the current moment.

[0121] Specifically, in NPV 1 and NPV 2 When the difference is not less than the preset difference threshold, the second state (such as s t+1 ). For example, a numerical simulator can be used to generate multiple sets of working parameter combinations at the next moment, which are input into the initial production dynamic prediction model, and then the production dynamic data of the entire oil field production process at the next moment are output. According to the production dynamic data, the fourth cumulative oil production and the fourth production round are determined as the second state at the next moment.

[0122] Then, the first state at the current moment, the second state at the next moment, and the first reward at the current moment are input into the initial strategy model, and the Critic network can obtain the evaluation result at the current moment according to the following formula:

[0123]

[0124] Specifically, according to the first state (such as: s t ) and the first action made by the Actor network in the first state (a t , the first action represents a plurality of first working parameter combinations), determining the first evaluation result (Q(s t ,a t ) 1 ). According to the second state (such as: s t+1 ) and the second action made by the Actor network in the second state (a t+1, the second action represents multiple sets of second working parameter combinations), and determines the second evaluation result (Q(s t+1 ,a t+1 )). Then Q(s t ,a t ) 1 , Q(s t+1 ,a t+1 ), the first reward r at the current moment t Substitute the above formula (10) to evaluate the multiple sets of first working parameter combinations output by the Actor network at the current moment, and obtain the evaluation result Q(s) at the current moment. t ,a t ), the evaluation result is for Q(s t ,a t ) 1 The updated evaluation results. In formula (10), γ is the discount factor, which indicates the impact of future rewards; is the learning rate, which controls the step size of Q value update. In this training process, the value function of the Critic network will also change with the progress of the interaction, usually based on the time difference (TD) error, expressed as:

[0125]

[0126] In some embodiments, the Actor network in S3 is used to adjust its own strategy according to the evaluation results and output multiple sets of new working parameter combinations based on the adjusted strategy, which may include:

[0127] Adjust your strategy according to the following formula:

[0128]

[0129] Among them, θ t+1 is the Actor network parameter at the next moment (i.e. the adjusted strategy); θ t is the Actor network parameter at the current moment (i.e. the strategy before adjustment); is the learning rate; is the gradient of θ; πθ(αt|st) is the probability distribution based on the policy network; At is the advantage function, which measures the t Take action a t The superiority.

[0130] Based on the adjusted strategy, the Actor network can make the next action that is more likely to gain long-term rewards.

[0131] In some embodiments, the above S103, according to the target production dynamic data, determines the target production index corresponding to each group of target working parameter combinations, which may include:

[0132] Determine the target cumulative oil production corresponding to each set of target operating parameters according to the target daily oil production data corresponding to each set of target operating parameter combinations;

[0133] The above-mentioned determination of the target net present value corresponding to each group of target working parameter combinations according to the target production index may include:

[0134] The target net present value corresponding to each set of target operating parameter combinations is determined based on the target cumulative oil production, the pre-acquired discount rate, the crude oil price, the throughput work cost corresponding to each set of target operating parameter combinations, and the total number of production years.

[0135] Specifically, the target cumulative oil production, the pre-acquired discount rate, the crude oil price, the throughput work cost corresponding to each set of target working parameter combinations, and the total number of production years can be substituted into the above formula (1), and the target net present value corresponding to each set of target working parameter combinations can be calculated using the above formula (1). This manual will not elaborate on this.

[0136] In some embodiments, the above S104, based on the target net present value and the target production index, selects a first target operating parameter combination that achieves the standard target production index from the multiple groups of target operating parameter combinations, which may include:

[0137] According to the target net present value and target production index corresponding to each group of target working parameter combinations, each group of target working parameter combinations is sorted in descending order;

[0138] The target operating parameter combination corresponding to the target net present value being greater than the preset net present value threshold and the target production index being greater than the preset production index threshold is selected from the sorting results as the first target operating parameter combination to achieve the standard target production index.

[0139] Specifically, after determining the target net present value and target production index corresponding to each group of target working parameter combinations, the target net present value and target production index corresponding to each group of target working parameter combinations can be compared, and each group of target working parameter combinations can be sorted in descending order according to the comparison results, and then the target working parameter combinations corresponding to the target net present value greater than the preset net present value threshold and the target production index greater than the preset production index threshold are selected from the sorting results, that is, the first target working parameter combination that meets or reaches the standard target production index (that is, the target parameter combination with a large target net present value and a large target production index can be selected as the first target parameter combination). At this time, it can be achieved that the standard or specified production index is achieved under complex reservoir conditions, and the maximum net present value (NPV) is achieved, and the optimal working parameter combination (that is, the first target parameter combination) is obtained.

[0140] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. For details, please refer to the description of the above-mentioned related processing related embodiments, and no further description is given here.

[0141] The above is an explanation of the present invention, however, it is worth noting that the specific embodiment is only for the purpose of better illustrating the present application and describing a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0142] In a specific implementation scenario, see Figure 3 As shown, before the specific implementation, the interactive training process of the model is generally as follows (for details, please refer to the above steps S1-S5):

[0143] A data set can be generated by a numerical simulator, and a prediction model built based on the LSTM framework (i.e., the above-mentioned initial production dynamic prediction model) can be trained based on the generated data set. Production dynamic data can be output based on the model. The production dynamic data includes daily oil production data corresponding to each group of working parameter combinations, or daily oil production data within different production durations, and then the cumulative oil production and the number of production rounds are calculated, and then the net present value NPV is calculated. The cumulative oil production and the number of production rounds are used as the state State, and the net present value NPV is sent to the Critic network as the reward Reward. The Critic network can accept feedback from the environment. The Critic network improves the Actor network strategy based on the state State and the reward Reward, so that the action Action output by the Actor network is the action that can achieve the maximum net present value NPV under the reservoir state. That is, the action Action output by the Actor network is input into the prediction model built based on the LSTM framework, and then the net present value of the output data is calculated. If the net present value calculated by the output data is not much different from the net present value calculated by the output data of the above-mentioned prediction model, it can be considered that the Action output this time is the best.

[0144] In the specific implementation, since the Action under different reservoir conditions is directly output based on the Actor in the trained strategy model (mature strategy model), and then the Action is input into the trained production dynamic prediction model, the Action output under a certain reservoir state will not be the only solution, the best working parameter combination can be screened out. For example, NPV and cumulative oil production can be used as standards or screening factors to screen various working parameter combinations under the reservoir state, and finally n groups of working parameter combinations that meet the production indicators are selected.

[0145] Through the above method, more flexible and effective decision-making can be achieved, and the net present value and profit level can be maximized under complex reservoir conditions, while ensuring that the development effect reaches a certain production index, thereby improving the overall benefits of oilfield development. It solves the problem that traditional optimization technology cannot take into account both production indicators and maximum profits when optimizing oilfield production parameters, and can only optimize under fixed target conditions, making it difficult to effectively cope with complex production environments with multiple targets and multiple constraints.

[0146] Although this specification provides examples such as the following embodiments or the attached Figure 4The method operation steps or device structure shown in the figure, but based on routine or no creative labor, the method or device may include more or fewer operation steps or module units after partial combination. In the steps or structures that do not have a necessary causal relationship logically, the execution order of these steps or the module structure of the device is not limited to the execution order or module structure shown in the embodiments of this specification or the drawings. When the method or module structure is applied to an actual device, server or terminal product, it can be executed sequentially or in parallel according to the method or module structure shown in the embodiments or the drawings (for example, a parallel processor or multi-threaded processing environment, or even a distributed processing, server cluster implementation environment). Based on the above-mentioned method for optimizing oilfield production operating parameters under complex reservoir conditions, the embodiments of this specification also propose an embodiment of a device for optimizing oilfield production operating parameters under complex reservoir conditions. As Figure 4 As shown, the device may specifically include the following modules:

[0147] The target strategy model output module 401 can be used to input the target state parameters reflecting the target reservoir state into the target strategy model and output multiple sets of target working parameter combinations;

[0148] The target production dynamic prediction model output module 402 can be used to input multiple sets of target working parameter combinations into the target production dynamic prediction model, and output the target production dynamic data of the whole oilfield production process. The target production dynamic prediction model and the target strategy model are obtained through interactive training;

[0149] The target net present value determination module 403 may be used to determine the target production index corresponding to each group of target working parameter combinations according to the target production dynamic data, and determine the target net present value corresponding to each group of target working parameter combinations according to the target production index;

[0150] The screening module 404 can be used to screen a first target operating parameter combination that achieves a standard target production indicator from the multiple groups of target operating parameter combinations based on the target net present value and the target production indicator, wherein the target net present value corresponding to the first target operating parameter combination is greater than a preset net present value threshold, and the standard target production indicator is greater than a preset production indicator threshold.

[0151] In some embodiments, the target state parameters in the above-mentioned target strategy model output module 401 may include a target injection-production ratio and a target recovery factor; the target production dynamic data in the above-mentioned target production dynamic prediction model output module 402 may include target daily oil production data corresponding to each group of target working parameter combinations; each group of target working parameter combinations in the above-mentioned target net present value determination module 403 is a combination of at least two target parameters including target injection volume, target injection rate, target injection concentration, and target shut-in time; the target production index in the above-mentioned target net present value determination module 403 may include target cumulative oil production.

[0152] In some embodiments, the above-mentioned target production dynamic prediction model output module 402 can be specifically used to input multiple sets of randomly generated working parameter combinations into the initial production dynamic prediction model, output the first production dynamic data under the entire oil field production process, input the state parameters reflecting the expected reservoir state into the Actor network in the initial strategy model, output multiple sets of first working parameter combinations at the current moment, input multiple sets of first working parameter combinations into the initial production dynamic prediction model, and output the second production dynamic data under the entire oil field production process; determine the first cumulative oil production and the first production cycle according to the first production dynamic data, determine the first net present value according to the first cumulative oil production, and determine the second net present value according to the second production dynamic data; when the difference between the first net present value and the second net present value is not less than the preset difference threshold, the first cumulative oil production and the first production cycle are used as the first state at the current moment, and the first net present value is used as The first reward at the current moment is input into the Critic network in the initial strategy model, and the Critic network is used to evaluate multiple groups of first working parameter combinations output by the Actor network at the current moment according to the first state and the first reward, obtain the evaluation result at the current moment and feed it back to the Actor network, and the Actor network is used to adjust its own strategy according to the evaluation result and output multiple groups of new working parameter combinations based on the adjusted strategy; the multiple groups of new working parameter combinations are input into the initial production dynamic prediction model, and the third production dynamic data under the whole process of oilfield production is output, and the third net present value is determined according to the third production dynamic data; it is determined whether the difference between the first net present value and the third net present value is less than a preset difference threshold, and if so, the initial production dynamic prediction model is used as the trained target production dynamic prediction model, and the initial strategy model is used as the trained target strategy model.

[0153] In some embodiments, the above-mentioned target production dynamics prediction model output module 402 can also be specifically used to input multiple sets of randomly generated working parameter combinations into a long short-term memory network, and output production dynamics data corresponding to each set of working parameter combinations; input the production dynamics data into the self-attention module, determine the importance of the production dynamics data based on the self-attention module, and determine the weight value of the production dynamics data corresponding to each set of working parameter combinations according to the normalized value of the importance; perform weighted summation of the weight value and the corresponding production dynamics data to obtain the first production dynamics data under the entire oil field production process and output the first production dynamics data.

[0154] In some embodiments, the above-mentioned target production dynamic prediction model output module 402 can also be specifically used to obtain the second state at the next moment when the difference between the first net present value and the second net present value is not less than a preset difference threshold, and input the first state, the second state, and the first reward into the initial strategy model; determine the first evaluation result at the current moment according to the first state and the first action performed by the Actor network under the first state, and determine the second evaluation result at the next moment according to the second state and the second action performed by the Actor network under the second state, the first action represents multiple groups of first working parameter combinations, and the second action represents multiple groups of second working parameter combinations; evaluate the multiple groups of first working parameter combinations output by the Actor network at the current moment according to the first evaluation result, the second evaluation result, and the first reward to obtain the evaluation result at the current moment.

[0155] In some embodiments, the above-mentioned target net present value determination module 403 can be specifically used to determine the target cumulative oil production corresponding to each group of target operating parameter combinations based on the target daily oil production data corresponding to each group of target operating parameter combinations; determine the target net present value corresponding to each group of target operating parameter combinations based on the target cumulative oil production, the pre-acquired discount rate, the crude oil price, the throughput work cost corresponding to each group of target operating parameter combinations, and the total number of production years.

[0156] In some embodiments, the above-mentioned screening module 404 can be specifically used to sort each group of target working parameter combinations in descending order according to the target net present value and the target production index corresponding to each group of target working parameter combinations; and select the target working parameter combination corresponding to the target net present value greater than the preset net present value threshold and the target production index greater than the preset production index threshold from the sorting results as the first target working parameter combination that achieves the standard target production index.

[0157] It can be seen from the above that an oilfield production working parameter optimization device under complex reservoir conditions provided in the embodiments of this specification can simultaneously achieve standard production indicators and maximize the net present value under complex and changeable reservoir conditions, and the optimization flexibility, efficiency and accuracy are high.

[0158] The embodiment of the present specification also provides an electronic device based on the above-mentioned oilfield production working parameter optimization method under complex reservoir conditions, comprising a processor and a memory for storing processor executable programs / instructions, wherein the processor can perform the following steps according to the programs / instructions when implemented: inputting target state parameters reflecting the target reservoir state into a target strategy model, and outputting multiple sets of target working parameter combinations; inputting multiple sets of target working parameter combinations into a target production dynamic prediction model, and outputting target production dynamic data under the entire oilfield production process, wherein the target production dynamic prediction model and the target strategy model are obtained through interactive training; determining target production indicators corresponding to each set of target working parameter combinations according to the target production dynamic data, and determining target net present values ​​corresponding to each set of target working parameter combinations according to the target production indicators; and selecting a first target working parameter combination that reaches a standard target production indicator from the multiple sets of target working parameter combinations according to the target net present value and the target production indicator, wherein the target net present value corresponding to the first target working parameter combination is greater than a preset net present value threshold, and the standard target production indicator is greater than a preset production indicator threshold.

[0159] In order to complete the above instructions more accurately, refer to Figure 5 As shown, the embodiment of this specification also provides another specific electronic device, wherein the electronic device includes a network communication port 501, a processor 502 and a memory 503, and the above structures are connected through internal cables so that each structure can perform specific data interaction.

[0160] The network communication port 501 can be used to input the target state parameters reflecting the target reservoir state into the target strategy model, and output multiple sets of target working parameter combinations;

[0161] The processor 502 may be specifically used to input multiple groups of target working parameter combinations into a target production dynamic prediction model, and output target production dynamic data under the entire oilfield production process, wherein the target production dynamic prediction model and the target strategy model are obtained through interactive training; according to the target production dynamic data, determine the target production index corresponding to each group of target working parameter combinations, and according to the target production index, determine the target net present value corresponding to each group of target working parameter combinations; according to the target net present value and the target production index, select a first target working parameter combination that reaches a standard target production index from the multiple groups of target working parameter combinations, wherein the target net present value corresponding to the first target working parameter combination is greater than a preset net present value threshold, and the standard target production index is greater than a preset production index threshold;

[0162] The memory 503 may be specifically used to store corresponding instruction programs.

[0163] In this embodiment, the network communication port 501 can be a virtual port that is bound to different communication protocols so that different data can be sent or received. For example, the network communication port can be a port responsible for web data communication, a port responsible for FTP data communication, or a port responsible for email data communication. In addition, the network communication port can also be a physical communication interface or communication chip. For example, it can be a wireless mobile network communication chip, such as GSM, CDMA, etc.; it can also be a Wifi chip; it can also be a Bluetooth chip.

[0164] In this embodiment, the processor 502 may be implemented in any appropriate manner. For example, the processor may take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) executable by the (micro)processor, a logic gate, a switch, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, etc. This specification does not limit this.

[0165] In this embodiment, the memory 503 may include multiple levels. In a digital system, anything that can store binary data can be a memory; in an integrated circuit, a circuit with a storage function but no physical form is also called a memory, such as RAM, FIFO, etc.; in a system, a storage device with a physical form is also called a memory, such as a memory stick, TF card, etc.

[0166] The embodiment of the present specification also provides a computer storage medium based on the above-mentioned oilfield production working parameter optimization method under complex reservoir conditions, the computer storage medium stores a computer program / instruction, and when the computer program / instruction is executed, the following is achieved: inputting a target state parameter reflecting the target reservoir state into a target strategy model, and outputting a plurality of target working parameter combinations; inputting a plurality of target working parameter combinations into a target production dynamic prediction model, and outputting target production dynamic data under the entire oilfield production process, wherein the target production dynamic prediction model and the target strategy model are obtained through interactive training; determining a target production index corresponding to each group of target working parameter combinations based on the target production dynamic data, and determining a target net present value corresponding to each group of target working parameter combinations based on the target production index; and selecting a first target working parameter combination that meets a standard target production index from the plurality of target working parameter combinations based on the target net present value and the target production index, wherein the target net present value corresponding to the first target working parameter combination is greater than a preset net present value threshold, and the standard target production index is greater than a preset production index threshold.

[0167] In this embodiment, the storage medium includes, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a cache, a hard disk (HDD), or a memory card. The memory may be used to store computer program instructions. The network communication unit may be an interface for network connection communication set in accordance with the standard specified by the communication protocol.

[0168] In this embodiment, the functions and effects specifically implemented by the program instructions stored in the computer storage medium can be explained in comparison with other implementations and will not be described in detail here.

[0169] Although the present specification provides method operation steps as described in the embodiments or flow charts, more or less operation steps may be included based on conventional or non-creative means. The order of steps listed in the embodiments is only one way of executing the order of many steps, and does not represent a unique execution order. When the device or client product in practice is executed, it can be executed in sequence or in parallel according to the method shown in the embodiments or the drawings (for example, a parallel processor or a multi-threaded processing environment, or even a distributed data processing environment). The term "include", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, product or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such a process, method, product or device. In the absence of more restrictions, it is not excluded that there are other identical or equivalent elements in the process, method, product or device including the elements. The first, second, etc. words are used to represent the name, and do not represent any particular order.

[0170] Those skilled in the art also know that, in addition to implementing the controller in a purely computer-readable program code, the controller can be made to implement the same function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, such a controller can be considered as a hardware component, and the devices for implementing various functions included therein can also be considered as structures within the hardware component. Or even, the devices for implementing various functions can be considered as both software modules for implementing the method and structures within the hardware component.

[0171] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media including storage devices.

[0172] Through the description of the above embodiments, it can be known that those skilled in the art can clearly understand that the present specification can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present specification can essentially be embodied in the form of a software product, which can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a mobile terminal, a server, or a network device, etc.) to execute the methods described in each embodiment of the present specification or some parts of the embodiments.

[0173] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. This specification can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0174] Although the present specification is described through embodiments, persons skilled in the art will appreciate that there are many variations of the present specification without departing from the spirit of the present specification, and it is intended that the appended claims include these variations without departing from the spirit of the present specification.

Claims

1. A method for optimizing oilfield production parameters under complex reservoir conditions, characterized in that: include: Inputting target state parameters reflecting the target reservoir state into the target strategy model, and outputting multiple sets of target working parameter combinations; Inputting multiple groups of target working parameter combinations into a target production dynamic prediction model, and outputting target production dynamic data in the entire oilfield production process, wherein the target production dynamic prediction model and the target strategy model are obtained through interactive training; Determine the target production index corresponding to each group of target working parameter combinations according to the target production dynamic data, and determine the target net present value corresponding to each group of target working parameter combinations according to the target production index; According to the target net present value and the target production index, a first target operating parameter combination that achieves a standard target production index is screened from the multiple groups of target operating parameter combinations, the target net present value corresponding to the first target operating parameter combination is greater than a preset net present value threshold, and the standard target production index is greater than a preset production index threshold.

2. The method according to claim 1, characterized in that The target state parameters include a target injection-production ratio and a target recovery factor; each group of target working parameter combinations is a combination of at least two target parameters of a target injection volume, a target injection rate, a target injection concentration, and a target shut-in time; the target production dynamic data includes target daily oil production data corresponding to each group of target working parameter combinations; the target production index includes target cumulative oil production.

3. The method according to claim 1, characterized in that: The target production dynamic prediction model and the target strategy model are obtained through interactive training, including: Inputting the randomly generated multiple sets of working parameter combinations into the initial production dynamic prediction model, outputting the first production dynamic data of the whole process of oilfield production, inputting the state parameters reflecting the expected reservoir state into the Actor network in the initial strategy model, outputting the multiple sets of first working parameter combinations at the current moment, inputting the multiple sets of first working parameter combinations into the initial production dynamic prediction model, outputting the second production dynamic data of the whole process of oilfield production; Determine a first cumulative oil production and a first production round according to the first production dynamic data, determine a first net present value according to the first cumulative oil production, and determine a second net present value according to the second production dynamic data; When the difference between the first net present value and the second net present value is not less than a preset difference threshold, the first cumulative oil production and the first production round are used as the first state at the current moment, and the first net present value is used as the first reward at the current moment to input into the Critic network in the initial strategy model, the Critic network is used to evaluate the multiple groups of first working parameter combinations output by the Actor network at the current moment according to the first state and the first reward, obtain the evaluation result at the current moment and feed it back to the Actor network, and the Actor network is used to adjust its own strategy according to the evaluation result and output multiple groups of new working parameter combinations based on the adjusted strategy; Inputting multiple sets of new working parameter combinations into the initial production performance prediction model, outputting third production performance data under the entire oil field production process, and determining a third net present value based on the third production performance data; Determine whether the difference between the first net present value and the third net present value is less than a preset difference threshold. If so, use the initial production dynamic prediction model as the trained target production dynamic prediction model and the initial strategy model as the trained target strategy model.

4. The method according to claim 3, characterized in that The initial production dynamic prediction model is constructed by a long short-term memory network and a self-attention module. Accordingly, the randomly generated multiple groups of working parameter combinations are input into the initial production dynamic prediction model to output the first production dynamic data under the whole process of oil field production, including: Inputting the randomly generated multiple groups of working parameter combinations into the long short-term memory network, and outputting the production dynamic data corresponding to each group of working parameter combinations; Inputting the production dynamic data into a self-attention module, determining the importance of the production dynamic data based on the self-attention module, and determining the weight value of the production dynamic data corresponding to each group of working parameter combinations according to the normalized value of the importance; The weight value is weighted and summed with the corresponding production dynamic data to obtain the first production dynamic data in the whole process of oil field production and output the first production dynamic data.

5. The method according to claim 3, characterized in that: The method further comprises: When the difference between the first net present value and the second net present value is not less than a preset difference threshold, obtaining a second state at the next moment, and inputting the first state, the second state, and the first reward into an initial strategy model; Accordingly, obtaining the evaluation result at the current moment includes: Determine a first evaluation result at a current moment according to the first state and a first action performed by the Actor network in the first state, and determine a second evaluation result at a next moment according to the second state and a second action performed by the Actor network in the second state, wherein the first action represents a plurality of first working parameter combinations, and the second action represents a plurality of second working parameter combinations; According to the first evaluation result, the second evaluation result, and the first reward, multiple groups of first working parameter combinations output by the Actor network at the current moment are evaluated to obtain the evaluation result at the current moment.

6. The method according to claim 1, characterized in that Determining the target production index corresponding to each group of target working parameter combinations according to the target production dynamic data includes: Determine the target cumulative oil production corresponding to each set of target operating parameters according to the target daily oil production data corresponding to each set of target operating parameter combinations; Determining the target net present value corresponding to each group of target working parameter combinations according to the target production index includes: The target net present value corresponding to each set of target operating parameter combinations is determined based on the target cumulative oil production, the pre-acquired discount rate, the crude oil price, the throughput work cost corresponding to each set of target operating parameter combinations, and the total number of production years.

7. The method according to claim 1, characterized in that The step of selecting a first target operating parameter combination that achieves a standard target production index from the plurality of groups of target operating parameter combinations according to the target net present value and the target production index comprises: According to the target net present value and target production index corresponding to each group of target working parameter combinations, each group of target working parameter combinations is sorted in descending order; The target operating parameter combination corresponding to the target net present value being greater than the preset net present value threshold and the target production index being greater than the preset production index threshold is selected from the sorting results as the first target operating parameter combination to achieve the standard target production index.

8. An oilfield production working parameter optimization device under complex oil reservoir conditions, characterized in that: include: A target strategy model output module is used to input target state parameters reflecting the target reservoir state into the target strategy model and output multiple sets of target working parameter combinations; A target production dynamic prediction model output module is used to input multiple sets of target working parameter combinations into the target production dynamic prediction model, and output the target production dynamic data under the whole process of oilfield production. The target production dynamic prediction model and the target strategy model are obtained through interactive training; A target net present value determination module, used to determine the target production index corresponding to each group of target working parameter combinations according to the target production dynamic data, and determine the target net present value corresponding to each group of target working parameter combinations according to the target production index; A screening module is used to screen a first target operating parameter combination that achieves a standard target production indicator from the multiple groups of target operating parameter combinations based on the target net present value and the target production indicator, wherein the target net present value corresponding to the first target operating parameter combination is greater than a preset net present value threshold, and the standard target production indicator is greater than a preset production indicator threshold.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the processor and the memory are communicatively connected to each other, the memory stores computer instructions, and the processor implements the steps of the method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images