Patent Application
20230229907
To optimize the Injection Production ratio (IPR) of a reservoir, geoscientists and engineers must optimize the number of wells drilled, as well as the drilling and completion procedures for each well. In mature fields, the increased production from the crest necessitates drilling of up-dip injectors to support crest production. These up-dip injectors are used to increase the core pressure as well as sustaining longer production periods for wells with high water cut. Increasing number of wells requires advanced/smart completions and inflow and injection control devices (ICDs) to be deployed selectively to overcome heterogeneity, such as fractures and stratifications, and to distribute the production along the horizontal section, delaying early water breakthrough and rapid increase of water production. Smart completion includes surface to downhole sensors and related wireless accessories.
This summary is provided to introduce a selection of concepts that are further described below in the detailed description. This summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to be used as an aid in limiting the scope of the claimed subject matter.
In one aspect, embodiments disclosed herein relate to a method for optimizing water injection in a reservoir that includes obtaining, by a computer processor, a first dataset from a first pipeline system in a first reservoir. The method includes training, by the computer processor and the first dataset, a first model. The method includes determining, by the computer processor and the first model, reliability of the first dataset. The method includes upon determining that the first dataset is reliable, generating, by the computer processor, the first dataset, and a second model, a first categorized dataset, and training, by the computer processor and the first categorized dataset, a third model. The method further includes optimizing, by the computer processor and using the third model, water injection control parameters of a second reservoir in accordance to a final water injection scheme.
In one aspect, embodiments relate to a system for optimizing water injection in a reservoir. The system includes an optimization manager comprising a computer processor. The optimization manager obtains a first dataset from a first pipeline system in a first reservoir. The optimization manager trains a first model utilizing the first dataset. The optimization manager determines, by the first model, reliability of the first dataset. Upon determining that the first dataset is reliable, the optimization manager generates, by a second model and the first dataset, a first categorized dataset, and trains, by the first categorized dataset, a third model. The optimization manager optimizes, using the third model, water injection control parameters of a second reservoir in accordance to a final water injection scheme.
In one aspect, embodiments relate to a non-transitory computer readable medium storing instructions. The instructions obtains a first dataset from a first pipeline system in a first reservoir. The instructions train, by the first dataset, a first model. The instructions determine, by the first model, reliability of the first dataset. Upon determining that the first dataset is reliable, the instruction generate, by a second model and the first dataset, a first categorized dataset by a second model, and train, by the first categorized dataset, a third model. The instructions further optimizes, using the third model, water injection control parameters of a second reservoir in accordance to a final water injection scheme
Other aspects and advantages of the claimed subject matter will be apparent from the following description and the appended claims.
Specific embodiments of the disclosed technology will now be described in detail with reference to the accompanying figures. Like elements in the various figures are denoted by like reference numerals for consistency.
In the following detailed description of embodiments of the disclosure, numerous specific details are set forth in order to provide a more thorough understanding of the disclosure. However, it will be apparent to one of ordinary skill in the art that the disclosure may be practiced without these specific details. In other instances, well-known features have not been described in detail to avoid unnecessarily complicating the description.
Throughout the application, ordinal numbers (e.g., first, second, third, etc.) may be used as an adjective for an element (i.e., any noun in the application). The use of ordinal numbers is not to imply or create any particular ordering of the elements nor to limit any element to being only a single element unless expressly disclosed, such as using the terms “before”, “after”, “single”, and other such terminology. Rather, the use of ordinal numbers is to distinguish between the elements. By way of an example, a first element is distinct from a second element, and the first element may encompass more than one element and succeed (or precede) the second element in an ordering of elements.
This disclosure provides an automatic procedure that determines optimized field water injection utilizing an artificial intelligence framework incorporating inline sensors and wireless communication. Embodiments of the disclosure also provide a workflow of constructing a plurality of models that are used in the artificial intelligence framework. The disclosure provides water injection optimization that recovers reservoir production while maintaining sustainability. Embodiments disclosed herein relate to the field water injection optimization using an artificial intelligence framework of inline sensors and wireless control.
In one or more embodiments, the framework integrates data from inline chemical sensors for the data-driven determination of phase flow measurements and optimizes the injection via communication of wireless transmitters. The inline sensors measure the chemical composition of phase fluids in the production tube, which is then utilized in order to adapt the phase measurement velocity recordings, and determine the uncertainty of the production quantities. These data are then integrated into a data-driven injection optimization framework for the optimization of water injection. In one or more embodiments, wireless data transmission based on LoRa and 5G technologies is utilized for data communication between the wells and the water injector wells in order to deal with the substantial amount of data that is retrieved from the various connected devices, and maintain operability in harsh operating conditions to avoid a single point of failure.
In general, embodiments of the disclosure include a system and a method that automatically process a plurality of data obtained from a reservoir. In particular, the obtained data are categorized and analyzed for reliability. Further, the disclosed method provide recommendations on replacing components in the system based on reliabilities of the obtained data. The obtained data also improves phase-flow determination accuracy. Moreover, the disclosed system and method utilize a plurality of models and the obtained data to predict future production of the reservoir, and to further determine an optimized water injection scheme (i.e., an optimized recovery scheme) to achieve best production while considering sustainability. Moreover, embodiments of the disclosure also relate to constructing the plurality of models. By utilizing wireless communication, disclosed system and method may determine the optimized water injection scheme based on real-time data.
In one or more embodiments, the wells (111-114) may communication with each other and with other system components such as a optimization manager (200) utilizing 5G and LoRaWAN protocols as shown in
Turning to
Specifically, the historical production data (251) may refer to production data that were previously measured from one or more reservoirs. In some embodiments, the historical production data (251) may include oil/water/gas rates, well pressure, choke size, and shut-in times. The inline chemical sensor data (252) may refer to data measured by inline chemical sensors deployed in a pipe system in the one or more reservoirs. More specifically, the inline chemical sensors obtain various data that reflect chemical composition of fluid flowing in the pipe system. In some embodiments, the inline chemical sensor data (252) may include, for example, pH level of the fluid, Hydrogen Sulfide (H2S) level, Carbon dioxide (CO2) level, Nitrogen oxides (NOX) level, flow rate, and gas content/temperature/pressure. The fluid phase data (253) may refer to data obtained by a plurality of multiphase flow meters (MPFMs) deployed in the pipe system. In particular, the inline chemical sensor (252) may be used to improve readings of the MPFMs. Further, in some embodiments, the water quality data (254) may include salinity content. Finally, the injection pattern data (256) may refer to established patterns for the water injection wells deployment. In some embodiments, the above-referenced various data may include existing data that have been measured from various wells in one or more tapped reservoirs, as well as real-time measured data obtained from a reservoir to be recovered.
Keeping with
LoRa is a radio modulation technique that is based on spread-spectrum modulation techniques. LoRa mainly includes LoRa physical layer and LoRaWAN protocol. Specifically, LoRaWAN defines communication protocol and system architecture for a network while LoRa physical layer enables a long-range communication link. More specifically, LoRa utilizes a very robust wireless modulation to create a long-range communication link. In particular, for LoRa, a single gateway or base station can cover hundreds of square kilometers.
5G wireless technology is the fifth generation mobile network that delivers data with multi-Gbps peak data speeds, ultra-low latency, better reliability. 5G wireless technology provides massive network capacity and unified user experience. In addition, 5G wireless technology provides new types of network, such as Internet of Things (IoT), that are designed to virtually connect everyone and everything together including machines, objects, and devices.
Turning back to
In some embodiments, the optimization manager (200) may include a graphical user interface (GUI) (205) that receives instructions and/or inputs from users. More specifically, the users may enter various types of instructions and/or inputs via the GUI (205) to start certain actions, such as calculating, evaluating, selecting, and/or updating data or parameters. In addition, the users may obtain various types of results from the certain actions via the GUI (205) as outputs.
In some embodiments, the optimization manager (200) may include a data controller (e.g., data controller (210)). The data controller (210) may be software and/or hardware implemented on any suitable computing device, and may include functionalities for obtaining various data from the data source (250) and processing the obtained data (255). For example, the data controller (210) may obtain the historical production data (251), the inline chemical sensor data (252), the fluid phase data (253), the water quality data (254), and the injection pattern data (256) from the data resource (250). The data controller may include data processors (215) and data storage (216). Specifically, the data processors (215) process the data obtained from the data source (250) as well as the data stored in the data storage (216). The data storage (216) may store the various data obtained from the data source (250), and other data and parameters for the other sections and functionalities of the optimization manager (200).
Keeping with
In one or more embodiments, the deep learning algorithms used to train the classification model (220) may be, for example, convolutional neural networks, deep neural networks, recurrent neural networks, support vector machines, decision trees, inductive learning models, deductive learning models, supervised learning models, unsupervised learning models, reinforcement learning models, etc. In some embodiments, two or more different types of machine-learning models are integrated into a single machine-learning architecture, e.g., an ML model may include decision trees and neural networks. In some embodiments, the optimization manager (200) may generate augmented or synthetic data to produce a large amount of interpreted data for training a particular model. More specifically, supervised ML models include classification, regression models, etc. Unsupervised ML models include, for example, clustering models. DL algorithms are a part of ML algorithms based on artificial neural networks with representation learning. For example, the DL algorithm may run data through multiple layers of neural network algorithms, each of which passes a simplified representation of the data to the next layer. More specifically, with respect to neural networks, for example, a neural network may include one or more hidden layers, where a hidden layer includes one or more neurons. A neuron may be a modelling node or object that is loosely patterned on a neuron of the human brain. In particular, a neuron may combine data inputs with a set of coefficients, i.e., a set of network weights for adjusting the data inputs. These network weights may amplify or reduce the value of a particular data input, thereby assigning an amount of significance to various data inputs for a task being modeled. Through machine learning (ML), a neural network may determine which data inputs should receive greater priority in determining one or more specified outputs of the neural network. Likewise, these weighted data inputs may be summed such that this sum is communicated through a neuron's activation function to other hidden layers within the neural network. As such, the activation function may determine whether and to what extent an output of a neuron progresses to other neurons where the output may be weighted again for use as an input to the next hidden layer.
In some embodiments, the trained classification model (220) is further used to determine whether a plurality of new data should be classified as reliable or unreliable. The classification results are output as data reliability (226). The plurality of new data may be real-time data collected from the new reservoir. Data that are determined as reliable will be further used to construct and train other models, such as a production prediction model (260). Data that are determined as unreliable will not be further adopted and would result in generating the recommendations to replace corresponding components and/or devices (227), such as the inline chemical sensors and the MPFMs. More details of the classification model (220) and the deep-learning algorithm(s) (225) are explained below in
In some embodiments, the optimization manager (200) may include a category model (230) that categorizes the obtained data into a plurality of categorized data (235) based on their impact on reservoir production. In some embodiments, the obtained data (225) are clustered with weight values in 10 categories ranging from 1 to 10, wherein weight value of 1 has the least impact on production forecasts, and weight value of 10 has the most impact on production forecasts. For example, compared to injection water quantity, water salinity has a lower impact on production prediction. As such, the water salinity may have a weight value of 2 and the injection water quantity may have a weight value of 10. More details of the category model (230) are explained below in
Further, in some embodiments, the optimization manager (200) may include the production prediction model (260). In some embodiments, the production prediction model (260) utilizes one or more deep-learning algorithms to generate production prediction (267) of a reservoir, such as the new reservoir. More specifically, the production prediction (267) is a numerical value that can be represented by a rate or a cumulative production.
In some embodiments, the production prediction model (260) is one or more ML models trained by Long-Short Tern Memory Network (LSTM) (265). LSTM is an artificial recurrent neural network architecture used in deep learning, which is capable of learning order dependence in sequence prediction problems. With regard to LSTM networks, an LSTM cell may include various output lines that carry vectors of information, e.g., from the output of one LSTM cell to the input of another LSTM cell. Thus, an LSTM cell may include multiple hidden layers as well as various pointwise operation units that perform computations such as vector addition. Furthermore, the size of the LSTM network may depend on the specific application. For simple geological layers, a limited number of hidden layers may be needed. For complex geological structures, a large number of hidden layers may be used to deal with the varying settings and complexity of the reservoir. To train the production prediction model (260), a plurality of water injection patterns and their corresponding production predictions for a reservoir are taken as inputs, and the production prediction model (260) is trained to correlate the plurality of water injection patterns with the production predictions using LSTM. More specifically, the water injection pattern refers to a flooding pattern for a particular arrangement of production and injection wells. For example, water injection pattern may include injection patterns within the same well or multiple wells. More specifically, in one or more embodiments, the water injection pattern may include alternate injections between the laterals either manually or autonomously via inflow control devices (ICDs) and inflow control valves (ICVs). In another example, it may include alternate injection of different types of water, such as low salinity water, medium salinity water, and sea water. For example the water injection pattern may be a direct line drive pattern, a staggered line drive pattern, a regular five-spot pattern, a regular or inverted seven spot pattern, a peripheral flood pattern, etc.
Further, the optimization manager (200) may include an optimization model (240) that integrates the trained production prediction model (260) to determine an optimized water injection scheme (245) of the reservoir. In some embodiments, the optimized water injection scheme (245) results in minimum overall carbon footprint produced during water injection process that recovers the reservoir. In particular, the optimized water injection scheme indicates water injection rate, time duration, and choke size for each individual injection well. These indicated parameters may be referred as water injection control parameters. In some embodiments, the optimized water injection scheme (245) may be further modified by users based on expert information. The expert information refers to information provided by the users to modify the optimized recovery scheme in order to adapt certain well parameters. Details of the production prediction model (260) and the optimization model (240) are explained below in
Keeping with
In some embodiments, the data source (250) and the optimization manager (200) may be implemented on different computing systems connected by a network. In some embodiments, the data source (250), the optimization manager (200) and/or other elements, including but not limited to network elements, user equipment, user devices, servers, and/or network storage devices may be implemented on computing systems similar to the computing system (600) shown and described in
While
Turning to
In Step 401, a plurality of data are obtained from a data source. For example, the plurality of data may include to the historical production data (251), the inline chemical sensor data (252), the fluid phase data (253), the water quality data (254), and the injection pattern data (256) from one or more wells, which are further obtained by the data controller (220) from the data resource (250) as illustrated in
Within the framework, in Step 402, a classification model is trained that is further used to determine reliability of the obtained data. In some embodiments, the classification model is one or more ML models trained by one or more DL algorithms and the plurality of data obtained at Step 401. The trained classification model determines reliability of the set of new data from the new reservoir. For example, the classification model and reliability determination may refer to the classification model (220) and the data reliability (226) in
In Step 403, reliability of the new data are determined by the trained classification model from Step 401. Upon determining that the data are not reliable, the workflow goes to Step 404, wherein recommendations on component replacement are generated. For example, the recommendations on component replacement may be the component replacement (227) generated by the data processor (215) and conveyed to users via the GUI (205) in
In Step 405, upon receipt of the recommendations, users may replace corresponding component(s), such as inline chemical sensors and MFMPs, in order to obtain more accurate and reliable data from the one or more wells. Upon determining that the obtained data are reliable, the procedure goes to Step 406. Accordingly, data with high reliability are used in ML model training processes to maintain accuracy.
In Step 406, the reliable obtained data are categorized according to their impact on production forecasts by a category model. In some embodiments, the category model categorizes the obtained data with different weight values. For example, the category model and the categorized data may refer to the category model (230) and categorized data (235) in
In Step 407, a production prediction model is trained to further predict production of the new reservoir. In some embodiments, the production prediction model is one or more ML models trained by LSTM and the categorized data from Step 406. The trained production prediction model predicts production of the new reservoir. For example, the production prediction model and the predicted production may refer to the production prediction model (260) and production prediction (267) in
In Step 408, an optimized water injection scheme is determined using an optimization model. Specifically, the optimization model incorporates the trained production prediction model from Step 307, and determines the optimized water injection scheme of the new reservoir based on the production prediction from Step 407. In some embodiments, the optimized water injection would result in the minimum overall carbon footprint produced during the water injection process that recovers the new reservoir. For example, the optimization model and the optimized water injection scheme may refer to the optimization model (240) and the optimized water injection scheme (245) in
In Step 409, the optimized water injection scheme from Step 408 is updated based on expert information. In some embodiments, the optimized water injection scheme is adjusted based on the expert information to adapt to certain requirements that are captured by the data obtained in Step 401. The updated water injection scheme is determined as final water injection scheme. For example, the expert data and the final water injection scheme may refer to the expert data (268) and the final water injection scheme (269) in
In Step 410, water injection with optimized parameters is performed in accordance with the final water injection scheme from Step 409. Specifically, in order to recover production of the new reservoir, the water injection is applied. The framework optimizes water injection control parameters that are further applied to the water injection in accordance to the final water injection scheme. Similar to the description of
Particularly, the framework described in
Those skilled in the art will appreciate that the process of
Turning to
In
The reliable data (502) are further categorized into reliable data with different weights (503) based on their impact on production predictions. Similar to the description in
While
Turning to
In one or more embodiments, for example, the input device 620 may be coupled to a receiver and a transmitter used for exchanging communication with one or more peripherals connected to the network system 630. The receiver may receive information relating to data obtained from reservoir(s) as described in
Further, one or more elements of the computing system 600 may be located at a remote location and be connected to the other elements over the network system 630. The network system 630 may be a cloud-based interface performing processing at a remote location from the well site and connected to the other elements over a network. In this case, the computing system 600 may be connected through a remote connection established using a 5G connection, such as protocols established in Release 15 and subsequent releases of the 3GPP/New Radio (NR) standards.
The computing system in
While
Although only a few example embodiments have been described in detail above, those skilled in the art will readily appreciate that many modifications are possible in the example embodiments without materially departing from this invention. Accordingly, all such modifications are intended to be included within the scope of this disclosure as defined in the following claims. In the claims, means-plus-function clauses are intended to cover the structures described herein as performing the recited function and not only structural equivalents, but also equivalent structures. Thus, although a nail and a screw may not be structural equivalents in that a nail employs a cylindrical surface to secure wooden parts together, whereas a screw employs a helical surface, in the environment of fastening wooden parts, a nail and a screw may be equivalent structures. It is the express intention of the applicant not to invoke 35 U.S.C. § 112, paragraph 6 for any limitations of any of the claims herein, except for those in which the claim expressly uses the words ‘means for’ together with an associated function.
