 PREDICTION OF OIL TANK BEHAVIOR USING A PROXY FLOW MODEL TECHNICAL AREA
专利摘要:
Using production data and a production flow record based on production data, a deep neural network (RNP) is formed to model a simulation of proxy flow from a reservoir. The simulation of proxy reservoir flow is performed using an overall Kalman filter (FKen), based on the RNP formed. FKen assimilates new data by updating a current set to match historicals by minimizing the difference between a predicted production output from the proxy flow simulation and measured production data from 'a deposit. Using the current updated assembly, a second proxy reservoir flow simulation is performed based on the RNP formed. Assimilation and realization are repeated while new data is available for assimilation. The predicted behavior of the reservoir is determined based on the simulation of proxy reservoir flow. An indication of the predicted behavior is provided to facilitate the production of fluids from the reservoir. Figure to be published with the abstract: Fig. 1 公开号:FR3085053A1 申请号:FR1903855 申请日:2019-04-11 公开日:2020-02-21 发明作者:Srinath Madasu；Yevgeniy Zagayevskiy；Terry Wong；Dominic Camilleri；Charles Hai Wang；Courtney Leeann Beck；Hanzi Mao；Hui Dong；Harsh Biren Vora 申请人:Landmark Graphics Corp； IPC主号:
专利说明:
Description Title of the invention: PREDICTION OF OIL TANK BEHAVIOR USING A PROXY FLOW MODEL Technical Field This description generally relates to forecasting the production of a given petroleum reservoir, including forecasting the production of a given petroleum reservoir using, for example, machine learning techniques. BACKGROUND [0002] Reservoir simulation is an area of reservoir engineering which uses computer models to predict the flow of fluids, such as oil, water and gas, within a reservoir. Reservoir simulators are used by oil producers, for example, to determine the best way to develop new fields, as well as to generate production forecasts for the developed fields on which investment decisions are based. Brief description of the drawings [0003] [fig-1] illustrates an example of a production well suitable for the production and exploration of hydrocarbons from an underground reservoir according to certain implementations. [Fig.2] conceptually illustrates an example of a flowchart of a process for predicting the future behavior of a reservoir using an overall Kalman filter in conjunction with a deep neural network in accordance with certain implementations. [Fig-3] illustrates a graph of an example of comparing the forecasted cumulative oil production with actual production data using a multilayer neural network in accordance with certain implementations. [Fig-4] illustrates an example of a drilling module for the implementation of the processes described here in accordance with certain implementations. [Fig.5] illustrates a cable line system suitable for implementing the processes described here in accordance with certain implementations. [Fig. 6] illustrates a block diagram of a set of general components of an example computer device in accordance with certain implementations. [0009] [fig-7] illustrates a block diagram of an example of an environment for the implementation of aspects in accordance with certain implementations. In one or more implementations, the components shown in each figure may not all be required, and one or more implementations may include additional components not shown in a figure. Variants of the arrangement and of the type of the components can be made, while remaining within the scope of the description of the invention. Additional components, different components or a lower number of components can be used while remaining within the scope of the description of the invention. Description of the embodiments The detailed description presented below is intended to describe various implementations and is not intended to represent the only implementations in which the technology which is the subject of the invention can be put into practice . As will be understood by the specialist in the field, the described implementations can be modified in different ways, while remaining within the scope of the present description. Consequently, the drawings and the description should be considered as illustrative in nature, and not limiting. Data assimilation techniques can be used in petroleum engineering, these techniques also being called "historical correspondence", which may involve combining observations with "prior knowledge" (for example, mathematical representations of mechanistic relationships, numerical models, model output) to obtain an estimate of the state of a system and the uncertainty of this estimate. In one example, data assimilation techniques can be used to determine an estimate of the uncertainty of a forecast of the production of wells in oil reservoirs and to generate computer models for optimizing decision parameters. likely to improve oil production. An overall Kalman filter (FKen) may refer to an efficient data assimilation technique, such as matching the production history of an oil reservoir. In one example, the FKen is able to update the properties of static and dynamic reservoirs, such as petrophysical properties and oil extraction rates, by assimilating data of different types from different sources. FKen techniques can be applied to a number of embodiments of an oil reservoir to update the model during a data assimilation step and to characterize the uncertainty associated with the model. As used herein, an embodiment refers to a set of values for properties at each location in an oil reservoir model, for example corresponding to a volume of interest. The FKen forecasting step requires the completion of a flow simulation cycle for each model realization, from the current time step to the future time step, using, for example, a simulator of complete physical flow. In one example, the computation time required for simulating the flow is generally very high due to the complexity of the oil reservoir model to be simulated and the relatively large number of realizations (several hundreds, for example) which can be processed. in the context of the simulation, in order to guarantee an appropriate relationship between the parameters of the model for an optimal FKen update operation. In practice, it may therefore be impossible to perform as many respective iterations of a flow simulation (for example, for each number of realizations) with a full physical flow simulator. Although the implementations of the technology which is the subject of the invention described here relate to the use of FKen, it is understood that any stochastic statistical technique can be used in place of FKen while remaining within the framework of the technology of the invention. The implementations of the technology that are the subject of the invention make it possible to replace the execution of a complete physical flow simulator by a proxy model based on machine learning techniques associated with a deep neural network (RNP). ). A deep neural network can be called a network because it can be represented by connecting different functions. For example, an RNP model can be represented in the form of a graph representing the way in which the functions are connected together from an input layer, through one or more hidden layers and finally to an output layer, each layer being able to have one or more nodes. In one example, the RNP of the technology which is the subject of the invention generates one or more dynamic properties of an oil tank model calibrated on the static input parameters of the flow simulator, for example a field of permeability, with low design requirements. This relationship between static and dynamic parameters allows for historical correspondence of production data by adjusting the static parameters of the reservoir model, such as petrophysical properties. [0016] Although an RNP is discussed for the purpose of explanation, it is obvious that other machine learning techniques can also be used. In addition, it is understood that other types of neural networks can be used by the technology which is the subject of the invention. For example, a convolutional neural network, a regulatory feedback network, a radial basis function network, a recurrent neural network, a modular neural network, an instantly formed neural network, an impulse neural network, a regulatory feedback network , a dynamic neural network, a neuro-fuzzy network, a pattern generation composition network, a memory network and / or any other suitable type of neural network can be used. The implementations of the technology which are the subject of the invention provide engineers specialized in oil deposits, and reservoir managers, with a tool enabling them to quickly and accurately predict the future yields of the reservoirs as well as the associated uncertainties. and therefore optimize the production of hydrocarbons as soon as possible. The reservoir yield forecast could be based on all relevant measured data, including, but not limited to, historical production well data (rate of oil and gas production, production and water injection, pressure at the bottom of the well), core samples, well logs, large-scale seismic, electromagnetic and gravity soundings, performed repeatedly in the same region over time. Data assimilation, model update, and prediction of future reservoir performance are performed automatically by the FKen techniques described in this document. In an example scenario consisting of performing a historical correspondence on the basis of a flow simulation of a reservoir model using an existing tool (for example, a tool not using no techniques related to FKen and / or machine learning), the execution of such a flow simulation can take longer. The historical mapping and the reservoir model provided by FKen and / or the machine learning techniques described here facilitate the improvement of fluid production from a reservoir production well, facilitate the determination of the advisability of carrying out a drilling operation in relation to the reservoir and / or other operations related to the reservoir (for example, injection of fluids). The technology object of the invention improves the parameters of an original reservoir model in order to provide an improved representation of the reservoir model, which can be run for future periods in order to predict the future production of the reservoir, either with the conditions of current operation, either using different tank operating parameters. In addition, the use of FKen and machine learning techniques to perform history matching potentially reduces the amount of time and / or amount of computing resources required to perform history matching. In one implementation, the proxy model can be based in part on the data of the production system, including various measurements collected at the bottom of the well from a well drilled in a reservoir, for example in the form of '' a production well for an oil and gas reservoir. In addition, multiple production wells can be drilled to allow access to fluids from the underground tank. Measured well data can be collected regularly from each production well to track changes in conditions in the reservoir, as will be described in more detail below in connection with an example of a production well as illustrated in [fig.l]. Oil reservoirs are generally geologically complex and large. In order to facilitate decision making that optimizes oil recovery, reservoir models are generated using numerous details based on different data. In one example, two types of data commonly used in reservoir modeling are geological data and production data. Geological data, such as seismic and cable logs, describe the properties of the soil (for example, porosity) of the reservoir. For comparison purposes, production data (for example, water saturation and pressure information) relate to the flow dynamics of the fluid in the reservoir. Both types of data can therefore be taken into account so that the resulting models are more precise. Based on these models, managers and other staff can make business decisions aimed at maximizing economic profits and / or minimizing operational risks for a given tank. [Fig-1] is a diagram of an example of production well 100 with a borehole 102 which was drilled in a reservoir formation. Hole 102 can be drilled to any depth and in any direction of formation. For example, drill hole 102 can be drilled to a depth of 10,000 feet or more, and can be oriented horizontally any distance through the formation, as desired for a particular implementation. The production well 100 also comprises a casing head 104 and a casing 106, both fixed in place by cement 103. A shutter block 108 is coupled to the casing head 104 and to a production well head 110 which, together, are sealed in the wellhead and allow fluids to be extracted from the well in a safe and controlled manner. The measured well data corresponding to the aforementioned geological and / or production data can be periodically sampled and collected in the production well 100 and combined with measurements from other wells in a reservoir, making it possible to monitor and assess the general condition of the tank. These measurements can be taken using a number of different downhole and surface instruments, including, but not limited to, a temperature and pressure sensor 118 and a flow meter 120. additional devices can also be coupled in line to a production column 112 comprising, for example, a well bottom throttle valve 116 (for example, for varying a level of restriction of fluid flow), an electric pump submersible (ESP) 122 (for example to suck the fluid flowing from the perforations 125 outside the ESP 122 and the production column 112), an ESP 124 motor (for example, to drive the ESP 122 ) and a gasket 114 (for example, to isolate the production area located under the gasket from the rest of the well 100). Additional area measuring devices can be used to measure, for example, the pressure at the top of the production column and the power consumption of the ESP 124 engine. Although various examples of components of the production well 100 are described above, it is obvious that the operations related to the measurement of the well data can be applied to other components of the production well 100 than those described and / or represented on [fig.l]. For example, measured well data can be provided from components such as a crown block and a water table, an access ramp and a hoisting cable, a drilling line, a rigging platform, a mobile muffle, mast, depth sounder shelter, water tank, electric cable tray, generator sets, fuel tanks, electric control room, bulk mud component storage, reserve pits, a sludge gas separator, a vibrating screen, a throttle collector, a tube ramp, racks for storing drill pipes, an accumulator and / or other types of components of production well 100. In bids in the work described here, well data can be provided by any of the components described here in connection with the production well 100. As shown in [fig. 1], the device along the production column 112 is coupled to a cable 128, which can be fixed to an external part of the production column 112. The cable 128 can be used mainly to power the devices to which it is coupled . Cable 128 can also be used to provide signal paths (e.g. electrical or optical paths), through which control signals can be directed from the surface to downhole devices, as well as telemetry signals downhole devices at the surface. The respective control and telemetry signals can be sent and received by a control unit 132 located on the surface of the production well. The control unit 132 can be coupled to the cable 128 via the shutter block 108. In one embodiment, the field staff can use the control unit 132 to locally control and monitor the downhole devices, for example via a user interface installed on a terminal or an integrated control panel. to the control unit 132. In addition or as a variant, the downhole devices can be controlled and monitored by a remote processing system. The processing system can be used to provide various supervisory control and data acquisition (SCADA) capabilities to the production wells associated with each reservoir in a deposit. For example, a teleoperator can use a processing system to send appropriate instructions to control operations at the well location to the control unit 132. Communication between the control unit 132 and the processing system can '' Perform via one or more communication networks, for example, in the form of a wireless network (for example, a cellular network), a wired network (for example, a wired connection to the Internet) or a combination of wireless and wired networks. In one or more implementations, such a processing system can include a computer device (for example a server) and a data storage device (for example a database). Such a computing device can be implemented using any type of computing device comprising at least one processor, a memory and a network interface capable of sending and receiving data to and from the control unit 132 via a communication network, such as a processor 438 described in [fig.4] and / or the computing device 600 described in [Fig. 6], In one implementation, the control unit 132 can periodically send drilling location production data via a communication network to the processing system for processing and storage. This drilling location production data may include, for example, production system metrics from various downhole devices, as described above. In some implementations, such production data can be sent using a remote terminal unit (RTU) of the control unit 132. In one implementation, the data storage device 144 can be used to store production data received from the control unit 132. In one example, the data storage device 144 can be used to store historical production data, including a record of actual and simulated system measurements of production obtained or calculated over a period, for example multiple temporal stages of simulation, as will be described in more detail below. Although the production well 100 is described in the context of a single reservoir, it should be noted that the implementations described here are not limited to them and that the implementations described can apply to the production of fluid from multiple tanks in a multi-tank production system. Although it is a computer simulation, it is possible to integrate the production data into a reservoir model. In an existing approach, several computer simulations are performed to identify reservoir models generating fluid dynamics corresponding to the production data of histories, also called matching of histories. Due to the complexity of the calculations and the large duration (for example, several hours or days) of such computer simulations, in some cases only a small number of computer simulations are performed and the results of the matching results d Historical data may be associated with some uncertainty, which may also introduce another uncertainty into future reservoir production forecasts that are made. Consequently, the embodiments of the technology which is the subject of the invention, with the aim of improving the decisions regarding planning and development of reservoirs, could make it possible to improve the accuracy of these mapping results. historical using FKen techniques in conjunction with an RNP machine learning model. In one or more implementations, the current process for managing petroleum reservoirs comprises two main improvements, associated with the matching of histories based on an FKen of the technology object of the invention: (i) the mapping of histories of an oil reservoir is carried out automatically with an FKen by grouping together all the relevant data available, and (ii) the application of RNP in an FKen environment allows the mapping of faster, since the RNP model has already been trained on a separate training dataset. In this way, the implementations of the technology which are the subject of the invention provide techniques for matching histories based on an FKen which use a proxy flow simulator based on an RNP model, thus allowing a faster assimilation of data in the reservoir model and, therefore, better prediction of the reservoir for future periods. Using the improved forecast, staff can make faster decisions about reservoir management. A proxy flow simulator, as used herein, relates to a machine learning representation of a full physical flow simulator for a given reservoir model. In one implementation, such a proxy flow simulator behaves and functions in a manner analogous to a full physical flow simulator, but uses an RNP model capable of allowing a reduction in processing times and / or other resources. computer to run a reservoir model flow simulation. [Fig.2] conceptually illustrates an example of a process flow diagram 200 for predicting the future behavior of a reservoir using an overall Kalman filter in conjunction with a deep neural network in accordance with certain implementations. Although this figure, as well as other process illustrations in this description, may represent functional steps in a particular sequence, the processes are not necessarily limited to the particular order or steps illustrated. The different steps shown in this figure or other figures can be modified, reorganized, executed in parallel or adapted in various ways. In addition, it should be understood that certain stages or certain orders of stages can be added to the process or removed from it, while remaining within the scope of the various implementations. The process 200 can be implemented by one or more computer devices or systems in certain implementations, such as a computer device 600 described in [Fig. 6] and / or the client device 702 or the server 706 described in [Fig. 7]. In block 202, initial input data including production data and geological data relating to a reservoir are received. In one or more implementations, production data may include, for example, real and / or simulated production system measurements, sometimes including recording the production flow corresponding to at least one flow (for example, oil, gas and / or water) when producing hydrocarbons in the reservoir and / or an amount of hydrocarbon produced from the reservoir. Actual measurements of the production system may include, for example, surface and downhole well measurements from various production wells in the multi-reservoir system. These measurements may include, but are not limited to, pressure, temperature, and fluid flow measurements taken at the bottom of the well near the well perforations, along the production rod, at the well head and in the collection network before the point where fluids mix with fluids from other tanks. The simulated measurements may include, for example and without limitation, estimates of the pressure, temperature and flow of the fluid. These estimates can be determined based, for example, on simulation results from one or more previous time steps corresponding to previous time periods. In one or more implementations, the geological data can be determined from the oil well logging which collects information relating to the properties of the land formations traversed by a wellbore for drilling and production operations oil. For example, in oil well cable logging, a probe is lowered into the borehole after drilling all or part of the well and is used to determine certain properties of the formations traversed by the borehole. In one example, geological data corresponding to various properties of earth formations are measured and correlated to the position of the probe in the borehole, when the probe is pulled up from the well. These properties can, in one example, be stored in one or more well logs. These well logs therefore provide geological data corresponding to petrophysical properties, facies and other related geological attributes along the trajectory of the wells. For example, a well log may include geological data corresponding to at least one petrophysical property (for example, porosity, lithology, water saturation, permeability, density, oil / water ratio, geochemical information, palaeodata, etc.) along the trajectory of a given well used for drilling oil wells. In one or more implementations, a deep neural network (RNP) model, corresponding to a proxy flow simulation model, can be formed in the following manner, as described in the discussion which will follow blocks 203 and 204. In block 203, a number M of complete physical flow simulations is executed with various sets of input parameters (for example, porosity, permeability and / or recording of the production flow) and outputs corresponding (for example, oil production rate, gas production rate, water production rate and / or downhole pressure at each well depending on the input parameters and duration) are recorded. In one example, this number of complete physical flow simulations can be large enough to sample the uncertainty space as precisely as possible. One hundred flow simulation cycles provide, in one example, a reasonable minimum number. These input and output values are not part of the FKen set, which can be generated later. Instead, this set of a wide range of input parameter values and corresponding response values can be used as a "library" to form an RNP-based proxy flow simulation model. In one or more implementations, relationships between the geological data received and the production data can be determined during the formation of the RNP. At block 204, the RNP determines a non-linear relationship between the input parameters and the response of the model by adapting a mathematical model to a learning set of available flow simulation cycles, which is a subset of all cycles (for example, about 60% to 80%). The mathematical model can be represented by a set of weights used to weight the nonlinear transformation of the input parameters as a weighted sum. This weighted sum represents an estimate of the output parameters of the flow simulation cycles which should correspond as closely as possible to the recorded target output parameters. The remaining set of flow simulation cycles can be used for testing and cross-validation of the RNP model formed. Once the RNP model is said to correctly describe the relationship between the static and dynamic parameters of the particular dynamic system over time, this RNP model is used to match histories with the FKen, since the production configurations do not change in the field (for example, the number of wells remains the same for future time steps, the production operation remains the same for these wells in future time steps, etc.). The proxy flow simulation model obtained has a form represented in equation (1) below. Here, the expected output parameters MJ 2 can be related to a time step t n + 1 and are represented as a function / output parameters updated M u 2 to a time step tn, input parameters set to day M u ; and a time component t n . [Math.l] [In block 205, based at least in part on the initial input data received, an initial set of input parameters is generated for the simulation of proxy flow. In one example, the initial set includes a number A of realizations of the input parameters (for example, petrophysical properties) and can be generated using one or more geostatistical techniques such as sequential Gaussian simulation (SGS ) for continuous properties (for example, porosity, permeability) and the simulation of sequential indicators (SIS) for categorical properties (for example, lithological facies). The initial set of input parameters may include respective parameters corresponding to petrophysical properties such as porosity and permeability. These initial parameters are treated as inputs to the flow simulator. In an example, the values of the initial parameters are assumed not to vary over time, but are updated at each step of data assimilation. Other examples of initial parameters can be a relative permeability curve, PVT tables, geomechanical properties, etc. In one example, geostatistical techniques allow interpolation of geological data so that, for geological data corresponding to one or more locations, geostatistical techniques can provide an interpolated value of geological data at a different location. In block 206, the number N of realizations of input parameters is used for the proxy flow simulation. In one example, the initial set is used as an entry in the RNP corresponding to the proxy flow simulation model. In block 208, an FKen forecast operation is performed in which the proxy flow simulation model is executed A times from the time step t n to t n + i. For example, using an overall Kalman filter, the simulation of the reservoir's proxy flow is performed on the basis of the RNP formed. Each realization from the set of static input parameters, presented in a three-dimensional space, is used as an input parameter in the RNP proxy flow simulation model formed described by equation (1) to predict a realization of the values of the output parameters for the next time step t n +] from the current time step t n . This forecasting procedure is repeated for each member of the set (achievement) of the set. In one example, the result of the FKen forecast operation is an unmodified three-dimensional input model (for example, a petrophysical reservoir model) and the corresponding production rates predicted from time step t n to t n +1 . In block 210, it is determined whether new production and / or geological data have been received. In one example, the new production and / or geological data can be received from well logging data (or from any other data source) generated since the reception of the previous production and / or geological data at block 202. If new production and / or geological data has not been received, the process 200 can go to block 214 described below. In block 212, if new production and / or geological data have been received, an update operation of the FKen is carried out, in which the petrophysical model is updated, using the Kalman filter overall, based on the determination of a covariance matrix of the initial input data, input and output parameters, and / or new production and / or geological data. For example, the results of the proxy flow simulation, such as the oil production rate, and the petrophysical model are updated in the data sampled at the current time step t n +] based on the matrix of covariance of the modeled system. In one example, the new production and / or geological data is assimilated using the set Kalman filter, updating a current set to obtain a matching of histories while minimizing the differences between a planned production output from the proxy flow simulation and production data measured from a deposit, in which the deposit may include one or more reservoirs located in a given geographic area and / or region. Equation (2) below shows how the update is performed. In one example, M I2 is the complete set in a matrix form made up of both static parameters M ; , for example, a petrophysical model, and dynamic variables M 2 , for example, the rate of oil production, as described in equation (3) below; the upper indices / u and a represent a predicted model of equation (1), an updated model and a real (true) model; K (t n + 1 ) is the Kalman gain, which is calculated from the covariance of the modeled system C 12 (t n + 1 ) and from the measurement error covariance of the data ε (ί n +] ) in accordance in equation (4) below; the covariance between the model variable M J2 (t n +] ), · and the model variable M J2 (t n +] ) 7 is calculated as in equation (5) below, r being an index of achievement and N being the size of the set or the total number of realizations, <...> Is the average operator indicated in equation (6) below; D 12 (t n +] ) represents the data sampled at the time step t n +] for the static variables M } and dynamic variables M 2 ; and H (t n + 1 ) is the observation matrix of 0 and 1 connecting the real model M 12 a (t n + 1 ) sought and the data values D ] 2 (t n + 1 ) in the notation matrix of the current time step t n + 1 as indicated in equation (7) below. [Math.2] [0043] [Math.3] M12 e Ml U M2 [Math.4] „ + 1) = C12 (t„ + +1 ) (H (t n + 1 ) * C12 (t fl + i) , f HT (t n + χ ) + ε (ί η + J) - 1 [0045] [Math.5] C12 (tn + l) z, j> r - l (M12 (tn + l) z, r - <M12 (tn + l) i>) * (M12 (tn + l) j, r - <M12 (tn + l) j>) / (JV - 1) [0046] [Math.6] <M12 (tn + l) z> = Σ λ ε = 1 M12 (tn + l) ï, r / N [0047] [ Math.7] D12 (tn + 1) = H (tn + 1) * M12 a (fn + 1) The process 200 can then repeat the operation in block 206 by carrying out, using the set Kalman filter , the reservoir proxy flow simulation based on the RNP formed and the updated petrophysical model, then repeat the operation at block 208 by executing the proxy flow simulation model at block 208 to predict the production parameters for the next time step t n + 2 using equation (1). These operations can be repeated until all the data is assimilated in the system (for example, while new data is available for assimilation). In one example, using the current updated set (mentioned above), a second simulation of proxy reservoir flow is performed based on the RNP formed. In block 214, based at least in part on the proxy flow simulation performed from the reservoir, the future behavior of the reservoir hey at least at a future rate of oil production from the reservoir is predicted. In one example, a predicted future behavior of the reservoir is determined based at least on the simulation of proxy flow performed from the reservoir. [Fig.3] illustrates a plot 300 of an example of comparison of the cumulative oil production forecast with the actual production data using a multilayer neural network for an overall size where N = 100. [Fig.3] shows the comparison between the expected cumulative oil production rate (for example, line 302) and real (for example, line 304) at a certain time step t for the hundred achievements using a multilayer neural network. In the example of [Fig. 3], line 300 corresponds to four production wells placed at the reference terminals of a two-dimensional model with a single injection well located in the middle of the grid. The entries in the RNP for each realization and time step are the permeability averaged on the grid, the variance of the permeability, the pressure of the injector at the bottom of the well, the pressure of the producer at the bottom of the well and the cumulative injection d 'water. The neural network comprises three layers with respectively eighteen, twenty and twenty neural nodes per layer. The rectified linear unit is used as the activation function in this example. The forecasted cumulative oil production values closely follow the actual values, which are obtained from a full physical flow simulator. The following discussion in [fig.4] and [Fig. 5] relate to examples of a drilling module and a logging module for a given oil or gas well system, which can be used to implement techniques based on the use of the overall Kalman filter in conjunction with the deep neural network being able to be applied to drilling and logging scenarios as described above. Petroleum and gaseous hydrocarbons can be naturally present in certain underground formations. In the oil and gas industry, an underground formation containing oil, gas or water is called a reservoir. A tank can be located underground or at sea. The tanks are generally located in a depth range from a few hundred feet (shallow tanks) to a few tens of thousands of feet (ultra-deep tanks). In order to produce oil or gas, a wellbore is drilled in a reservoir or adjacent to a reservoir. The oil, gas or water produced from the wellbore is called a reservoir fluid. An oil or gas well system can be located on land or at sea. [Fig.4] illustrates an example of a drilling module 400 for the implementation of the methods described here. It should be noted that if [Fig. 4] generally illustrates a shore drilling module, the specialist in the field will easily recognize that the principles described here also apply to underwater drilling operations which use platforms and floating or offshore equipment, without leaving the part of the description of the invention. In one or more implementations, the process 200 described above begins after the drilling module 400 has drilled a well 416 penetrating into an underground formation 418. It must be borne in mind, however, that everything treatment performed in Method 200 by any suitable component described here can occur only at the top of the well, only at the bottom of the well, or at least in part of both (i.e., distributed treatment ). As illustrated, the drilling module 400 may include a drilling platform 402 which supports a derrick 404 having a movable block 406 for raising and lowering a drilling column 408. The drilling column 408 may include in particular a drilling rod and a spiral tube, which are generally known to the specialist in the field. A drive rod 410 supports the drill string 408 when lowered via a turntable 412. A drill bit 414 is attached to the distal end of the drill string 408 and is driven by either a downhole motor and / or by rotation of the drill stand 408 from the well surface. When the drill bit 414 rotates, it creates the wellbore 416 which penetrates into various underground formations 418. A pump 420 (for example, a mud pump) circulates the drilling mud 422 through a supply pipe 424 and towards the drive rod 410, which conveys the drilling mud 422 at the bottom of the well through the interior of the drill string 408 and through one or more orifices in the drill bit 414. The drill mud 422 is then recycled to the surface via a ring 426 defined between the drill string borehole 408 and the walls of the borehole 416. At the surface, the recycled or spent drilling mud 422 leaves the ring 426 and can be conveyed to one or more fluid treatment units 428 via a interconnection flow line 430. After passing through the fluid treatment unit (s) 428, a “cleaned” drilling mud 422 is deposited in a retention basin located nearby 432 (that is to say, a mud basin). While they are presented as being arranged at the outlet of the borehole 416 via the ring 426, the specialist in the field will easily understand that the fluid treatment unit or units 428 can be arranged at any other location in the module. drilling 400 to facilitate its proper functioning, without departing from the scope of the description of the invention. Chemicals, fluids, additives and the like can be added to the drilling mud 422 via a mixing hopper 434 communicatively coupled or otherwise in fluid communication with the retention basin 432 Mixing hopper 434 may include, but is not limited to, mixers and associated mixing equipment known to those skilled in the art. However, in other implementations, chemicals, fluids, additives and the like can be added to the drilling mud 422 at any other location in the drilling module 400. In at least one implementation , for example, there can be several retention basins 432, such as several retention basins 432 in series. In addition, the retention basin 432 can be representative of one or more fluid storage installations and / or units where the chemicals, fluids, additives and the like can be stored, reconditioned and / or regulated. until they are added to the drilling mud 422. The processor 438 can be part of the computer hardware used to implement the various blocks, modules, elements, components, methods and illustrative algorithms described here. . The 438 processor can be configured to execute one or more sequences of instructions, programming positions, or code stored on a non-transient computer-readable medium. Processor 438 can be, for example, a general purpose microprocessor, a microcontroller, a digital signal processor, an application-specific integrated circuit, a network of programmable doors in situ, a programmable logic device, a controller, a machine state, gate logic, discrete hardware components, an artificial neural network, or any similar suitable entity capable of performing calculations or other manipulation of data. In some implementations, the hardware may further include such things as, for example, memory (for example, random access memory (RAM), flash memory, read only memory (ROM), programmable read only memory ( PROM), erasable programmable read only memory (EPROM)), registers, hard disks, removable disks, CDROMs, DVDs, or any other suitable device or storage medium. The executable sequences described here can be implemented with one or more code sequences contained in a memory. In some implementations, this code can be read into memory from another machine-readable medium. The execution of instruction sequences contained in the memory can cause a processor 438 to execute the processing steps described here. One or more processors 438 in a multiprocessing arrangement can also be used to execute sequences of instructions in memory. In addition, wired circuits can be used in place of, or in combination with software instructions to implement various applications described here. Thus, the present implementations are not limited to any specific combination of hardware and / or software. As used herein, machine readable media will refer to any media that directly or indirectly provides instructions to processor 438 to execute them. Machine-readable media can take many forms, including, for example, non-volatile media, volatile media, and transmission media. Non-volatile media may include, for example, optical and magnetic disks. Volatile media can include, for example, dynamic memory. Transmission media can include, for example, coaxial cables, wires, optical fibers, and wires that form a bus. Common forms of machine-readable media may include, for example, floppy disks, floppy disks, hard disks, magnetic tapes, other similar magnetic media, CD-ROMs, DVDs, other similar optical media , punch cards, paper strips and similar physical media with patterned holes, random access memory, read only memory, programmable read only memory, erasable programmable read only memory and flash erasable programmable read only memory. The drilling module 400 may also comprise a downhole module (BHA) coupled to the drilling column 408 near the drill bit 414. The BHA may include various tools for measuring the wellbore such as, but not limited to, measurement during drilling (MWD) and logging during drilling (LWD) tools, which can be configured to perform measurements at the bottom of wells and / or at the top of wells of formations surrounding underground 418. Logging during drilling (LWD) or measurement during drilling (MWD) 436 equipment is included along the drill string 408. In one or more implementations, the drilling module 400 involves drilling the wellbore 416 while the logging measurements are performed with the LWD / MWD equipment 436. More generally, the methods described here involve the introduction of a logging tool into the wellbore which is able to determine parameters s drilling wells, including the mechanical properties of the formation. The logging tool can be an LWD logging tool, an MWD logging tool, a wired line logging tool, a smooth cable logging tool, and the like. In addition, it is understood that any processing carried out by the logging tool can only occur at the top of the well, at the bottom of the well, or at least in a part of both (i.e., a distributed processing). According to the present description, the LWD / MWD 436 equipment may include a fixed acoustic sensor and a mobile acoustic sensor used to detect the flow of fluid flowing in and / or adjacent to the wellbore 416. In one example, the fixed acoustic sensor can be arranged around the longitudinal axis of the LWD / MWD equipment 436, and thus of the wellbore 416 at a predetermined fixed location inside the wellbore 416. The mobile acoustic sensor can be arranged around the longitudinal axis of the LWD / MWD equipment 436 and, thus, of the wellbore 416, and is designed to move along the longitudinal axis of the wellbore 416. However, the arrangement of the fixed acoustic sensor and the mobile acoustic sensor is not limited thereto and the acoustic sensors can be arranged in any configuration required by the application and design. The LWD / MWD 436 equipment can transmit the measured data to a processor 438 at the surface on the wired or wireless network. Data transmission is generally illustrated on line 440 to show the communication coupling between the processor 438 and the LWD / MWD equipment 436 and does not necessarily indicate the path to which the communication is carried out. The fixed acoustic sensor and the mobile acoustic sensor can be communicatively coupled to the line 440 used to transfer measurements and signals from the BHA to the processor 438 which processes the acoustic measurements and the signals received by acoustic sensors (for example, sensor fixed acoustic, mobile acoustic sensor) and / or controls the operation of the BHA. In the technology that is the subject of the invention, the LWD / MWD 436 equipment may be capable of recording the analysis of the underground formation 418 near the wellbore 416. In certain implementations, part of the processing can be carried out by a telemetry module (not shown) in combination with the processor 438. For example, the telemetry module can preprocess the signals from individual sensors (for example, by signal conditioning, noise filtering and / or cancellation) and transmitting them to a surface data processing system (e.g., processor 438) for further processing. It should be noted that any processing carried out by the telemetry module can occur only at the top of the well, at the bottom of the well, or at least in part of the two (that is to say, distributed processing). In various implementations, the processed acoustic signals are evaluated in conjunction with the measurements of other sensors (for example, temperature and pressure measurements of surface wells) to evaluate the flow conditions and the overall well integrity. The telemetry module may include any known downhole communication means including, without limitation, a pulsed mud telemetry system, an acoustic telemetry system, a cable communication system, a wireless communication system, or a any combination of these. In some implementations, some or all of the measurements taken by the fixed acoustic sensor and the mobile acoustic sensor can also be stored in a memory associated with the acoustic sensors or with the telemetry module for subsequent recovery at the surface during the retraction of the drill stand 408. [Fig.5] illustrates a logging module 500 having a cable line system suitable for implementing the methods described here. As illustrated, a platform 510 can be equipped with a derrick 512 which supports a winch 514. The drilling of oil and gas wells, for example, is commonly carried out using a drill string connected between them so as to form a drill string which is lowered via a turntable 516 into a wellbore 518. Here, it is assumed that the drill string has been temporarily removed from the wellbore 518 to allow a logging tool 520 (and / or any other suitable wired line tool) to be lowered by wired line 522, the smooth cable, the spiral tubing, the rod, the downhole tractor, the logging cable and / or by any other suitable physical structure or other means of transportation extending downstream from the hole in the wellbore 518. Usually, the logging tool 520 is lowered to a region of interest and then pulled towards e top at a substantially constant speed. During the upward journey, the instruments included in the logging tool 520 can be used to make measurements on the underground formation 524 adjacent to the wellbore 518 during the passage of the logging tool 520. In addition, it it is understood that any processing carried out by the logging tool 520 can only occur at the top of the well, at the bottom of the well, or at least in part of both (that is to say, a distributed processing) . The logging tool 520 can comprise one or more wired line instruments which can be suspended in the wellbore 518 by the wired line 522. The wired line instrument (s) can comprise the fixed acoustic sensor and the sensor mobile acoustics, which can be communicatively coupled to the wired line 522. The wired line 522 may include conductors to transport energy to the wired line instrument and also facilitate communication between the surface and the line instrument cable. The logging tool 520 may include a mechanical component to cause the displacement of the mobile acoustic sensor. In some implementations, the mechanical component may need to be calibrated to provide more precise mechanical movement when the moving acoustic sensor is repositioned along the longitudinal axis of wellbore 518. The acoustic sensors (for example, the fixed acoustic sensor, the mobile acoustic sensor) can include electronic sensors, such as hydrophones, piezoelectric sensors, piezoresistive sensors, electromagnetic sensors, accelerometers or the like. In other implementations, the acoustic sensors may include fiber optic sensors, such as point sensors (for example, fiber Bragg gratings, etc.) distributed at desired or predetermined locations along the length d 'an optical fiber. In still other implementations, the acoustic sensors can include distributed acoustic sensors, which can also use optical fibers and allow a distributed measurement of local acoustics at any given point along the fiber. In still other implementations, the acoustic sensors may include optical accelerometers or optical hydrophones which have fiber optic cabling. In addition or in a variant, in one example (not explicitly illustrated), the acoustic sensors can be fixed or embedded inside one or more rod trains or casing columns covering the wellbore 518 and / or the wall of the wellbore 518 at a predetermined distance axially spaced. A logging installation 528, represented in [fig.5] in the form of a truck, can collect measurements from the acoustic sensors (for example, the fixed acoustic sensor, the mobile acoustic sensor), and can include the processor 438 for controlling, processing, storing and / or viewing the measurements collected by the acoustic sensors. The processor 438 can be communicatively coupled to the wired line instrument or instruments by means of the wired line 522. As a variant, the measurements collected by the logging tool 520 can be transmitted (by cable or by network without wire) or physically delivered to off-site computing facilities where the processes and processes described here can be implemented. [Fig.6] illustrates a block diagram of a set of general components of an example computer device 600. In this example, the computer device 600 includes a processor 602 for executing instructions which can be stored in a device or a memory element 604. The computing device 600 can include many types of memory, data storage or non-transient computer-readable storage medium, such as a first data storage for program instructions to be executed by processor 602, separate storage for images or data, removable memory for sharing information with other devices, etc. The computer device 600 can usually include some type of display element 606, such as a touch screen or a liquid crystal display (LCD). As described, the computing device 600 will include, in many embodiments, at least one input element 610 capable of receiving conventional input from a user. This conventional input may include, for example, a push button, a touchpad, a touchscreen, a scroll wheel, a joystick, a keyboard, a mouse, a numeric keypad or any other device or element by which a user can enter a command. in the device. In some embodiments, however, the computing device 600 may not include buttons and can only be controlled by a combination of visual and audio controls, so that a user can control the computing device 600 without having to be on. contact with the computing device 600. In certain embodiments, the computing device 600 of [fig.6] may include one or more network interface elements 608 for communicating on various networks, such as Wi-Fi communication systems , Bluetooth, RF, wired or wireless. The computer device 600 can communicate in many embodiments with a network, such as the Internet, and may be able to communicate with other computer devices of this type. As described here, different approaches can be implemented in various virions according to the embodiments described. For example, [fig.7] illustrates a block diagram of an example of environment 700 for the implementation of aspects according to various embodiments. As will be understood, although a client-server environment is used for explanatory purposes, different environments can be used, if necessary, to implement various embodiments. The system includes an electronic client device 702, which may include any suitable device that can be used to send and receive requests, messages or information over an appropriate network 704 and return information to a user of the device. Examples of such client devices include personal computers, portable telephones, portable messaging devices, portable computers, set-top boxes, personal data assistants, electronic book readers and the like. Network 704 can include any suitable network, including an intranet, the Internet, a cellular network, a local network, or any other network or combination thereof. The network 704 could be a “push” network, a “pull” network or a combination of these. In a push network, one or more of the servers push data to the client device. In a pull network, one or more of the servers send data to the client device upon request by the client device. The components used for such a system may depend at least in part on the type of network and / or environment selected. The protocols and components for communicating via such a network are well known and will not be described here in detail. Calculation on the 704 network can be activated via wired or wireless connections and combinations thereof. In this example, the network includes the Internet, because the environment includes a server 706 for receiving requests and serving content in response to them, although for other networks, an alternative device for a similar purpose may be used, as will seem obvious to the specialist in the field. The client device 702 can represent the computer device 600 of [fig.6] and the server 706 can represent off-site computer installations in other implementations. Server 706 will usually include an operating system which provides executable program instructions for general administration and operation of that server and will usually include storage instructions for computer readable media which, when executed by a server processor, allow the server to perform its intended functions. Appropriate implementations for the operating system and the general functionality of the servers are known or commercially available and are easily implemented by the specialist in the field, in particular in the light of the present description. The environment in one embodiment is a distributed computing environment using several computer systems and components which are interconnected via computer links, using one or more computer networks or one or more several direct connections. However, the specialist in the field will appreciate that such a system can work equally well in a system having a number of components less or greater than that illustrated in [fig.7]. Thus, the description of the environment 700 at [Lig. 7] should be considered to be illustrative in nature and not to limit the scope of the description of the invention. A storage medium and another non-transient computer-readable medium for containing code, or parts of code, may include any suitable storage medium used in the field, such as, but not limited to, a medium volatile and non-volatile, removable and non-removable implemented in any process or technology for storing information such as computer-readable instructions, data structures, program modules or other data, such as a memory random access memory, read-only memory, electrically erasable programmable read-only memory, flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical storage, magnetic tapes, magnetic tape, storage on magnetic disc or other magnetic storage devices, or any other medium that can be used to store the desired information and accessible by the system device. Based on the description and the lessons provided here, the specialist in the field will have in mind other means and / or procedures for implementing the various implementations. Additional Considerations Various examples of aspects of the invention are described below in the form of clauses for convenience. The procedures of any preceding paragraph, alone or in combination, may further include the following clauses. They are provided by way of example and do not limit the technology which is the subject of the invention. Clause 1. A method comprising: receiving initial input data comprising at least one of the production data or geological data relating to a reservoir; training, using at least the initial input data and a production flow record based on the production data, of a deep neural network (RNP) in order to model a flow simulation tank proxy; carrying out, using an overall Kalman filter, the simulation of proxy reservoir flow based on the RNP formed; assimilation of new data, using the Kalman set filter, by updating a current set to obtain a matching of histories, minimizing the differences between a production output predicted from the proxy flow simulation and production data measured from a deposit; carrying out, using the current updated set, a second simulation of proxy reservoir flow based on the RNP formed; repeating assimilation and realization while new data is available for assimilation; determining a predicted behavior of the reservoir based at least on the proxy flow simulation performed from the reservoir; and providing an indication of the predicted behavior to facilitate the production of fluids from the reservoir. Clause 2. The method of clause 1, in which the realization, using the overall Kalman filter, of the proxy flow simulation of the reservoir comprises the execution of a forecasting operation of the overall Kalman filter. Clause 3. The method of clause 1, further comprising: generating, on the basis at least in part of the initial input data received, of an initial set of input parameters for the simulation of proxy flow, in which the initial set includes a number of realizations of the input parameters. Clause 4. The method of clause 3, in which the initial set of input parameters includes parameters corresponding to the petrophysical properties of porosity and permeability. Clause 5. The method of clause 3, in which the initial set is generated using a geostatistical technique. Clause 6. The method of clause 3, in which the initial set is used as input to the RNP. Clause 7. The method of clause 3, in which the simulation of proxy flow on the basis of RNP is performed for each realization of the initial set in order to predict the production parameters of a previous time step t at a current time step t n . Clause 8. The method of clause 1, in which the simulation of the proxy flow of the reservoir provides output rates and a petrophysical model. Clause 9. The method of clause 8, further comprising: receiving new input data including new production data and new geological data; updating, using the overall Kalman filter, the petrophysical model on the basis of a covariance matrix; and carrying out, using the overall Kalman filter, the simulation of proxy reservoir flow based on the RNP formed and the updated petrophysical model. Clause 10. The method of clause 9, in which the determination of the predicted behavior of the reservoir is based at least in part on the updated petrophysical model, and the predicted behavior of the reservoir predicts the rate of future oil production of the reservoir to facilitate the management of oil production from a production well at the reservoir. Clause 11. A system comprising: a processor; and a memory device comprising instructions which, when executed by the processor, cause the processor to: receive initial input data comprising at least one of the production data or geological data; form, using at least the initial input data and a production flow record based on the production data, a deep neural network (RNP) to model a simulation of a proxy flow of a reservoir ; perform, using an overall Kalman filter, the proxy flow simulation of the reservoir on the basis of the RNP formed; determining a predicted behavior of the reservoir based at least on the simulation of proxy flow performed from the reservoir; and providing an indication of the predicted behavior to facilitate the production of fluids from the reservoir. Clause 12. The system of clause 11, in which the realization, using the overall Kalman filter, of the proxy flow simulation of the reservoir comprises the carrying out of an operation of forecasting the filter of overall Kalman. Clause 13. The system of clause 11, in which the instructions further cause the processor to: generate, on the basis at least in part of the initial input data received, an initial set of input parameters for proxy flow simulation, in which the initial set includes a number of realizations of the input parameters. Clause 14. The system of clause 13, in which the initial set of input parameters includes parameters corresponding to the petrophysical properties of porosity and permeability. Clause 15. The system of clause 13, in which the initial set is generated using a geostatistical technique. Clause 16. The system of clause 13, in which the initial set is used as an entry in the RNP. Clause 17. The system of clause 13, in which the simulation of proxy flow on the basis of RNP is carried out for each realization of the initial set in order to predict the production parameters of a previous time step t at a current time step t n . Clause 18. The system of clause 11, in which to perform the proxy flow simulation of the reservoir, outputs production rates and a petrophysical model. Clause 19. The system of clause 18, in which the instructions further cause the processor to: receive new input data comprising new production data and new geological data; update, using the overall Kalman filter, the petrophysical model based on a covariance matrix; and perform, using the overall Kalman filter, the simulation of proxy reservoir flow based on the RNP formed and the updated petrophysical model. Clause 20. A non-transient computer-readable medium comprising instructions stored therein and which, when executed by at least one computer device, oblige the at least one computer device to perform operations comprising: receiving initial input data comprising at least one of the production data or geological data; training, using at least initial input data, a deep neural network (RNP) to model a simulation of a proxy reservoir flow; carrying out, using an overall Kalman filter, the simulation of proxy reservoir flow based on the RNP formed; assimilating new data, using Kalman's set filter, updating a current set to match historicals, minimizing the differences between a predicted output from the simulation of proxy flow and production data measured from a deposit; carrying out, using the current updated set, a second simulation of proxy reservoir flow based on the RNP formed; repeating assimilation and realization while new data is available for assimilation; determining a predicted behavior of the reservoir based at least on the proxy flow simulation performed from the reservoir; and providing an indication of the predicted behavior to facilitate the production of fluids from the reservoir. [0100] Unless otherwise indicated, a reference to an element in the singular does not mean "one and only", but rather one or more. For example, "a" module can refer to one or more modules. An element preceded by "a", "an", "the" or "said" does not exclude, in the absence of other restrictions, the existence of additional identical elements. The titles and subtitles, if any, are used only for convenience and do not limit the invention. The word "example" is used to refer to an example or illustration. To the extent that the term "understand", "have", or a similar term is used, that term is understood to be inclusive in a manner similar to the interpretation of the term "understand" when used as a transition word in a claim. Relational terms such as "first", "second" and similar terms can be used to distinguish between one entity or action and another, without necessarily imposing or implying a real relationship or order between the said entities or actions. The expressions such as “one aspect”, “the aspect”, “another aspect”, “certain aspects”, “one or more aspects”, “an implementation”, “the implementation”, "Another implementation", "some implementations", "one or more implementations", "an embodiment", "the embodiment", "another embodiment", "certain modes of realization "," one or more embodiments "," a configuration "," the configuration "," another configuration "," certain configurations "," one or more configurations "," the technology object of the invention ", "The description", "the present description", other variants and the like are used for reasons of convenience and do not imply that a description relating to this or these expressions is essential to the technology which is the subject of the invention or that this description applies to all the configurations of the technology object of the invention. A description relating to one or more expressions can apply to all configurations or to one or more configurations. A description relating to one or more expressions can provide one or more examples. An expression such as an aspect or certain aspects can refer to one or more aspects and vice versa, and this applies in a similar way to the other previous expressions. The expression "at least one of" preceding a series of elements, with the terms "and" or "or" to separate any one of the elements, modifies the list as a whole, rather than each part of the list. The expression "at least one of" does not require the selection of at least one element; on the contrary, the expression allows a meaning which includes at least any one of the elements, and / or at least one of any combination of the elements, and / or at least one of each of the elements. By way of example, each of the expressions “at least one of A, B and C” or “at least one of A, B or C” designates only A, only B, or only C; any combination of A, B and C; and / or at least one of each of A, B and C. It is understood that the specific order or hierarchy of the steps, operations or processes described is an illustration of examples of approaches. Unless otherwise indicated, it is understood that the specific order or hierarchy of steps, operations or procedures may be carried out in a different order. Certain steps, certain operations or certain processes can be carried out simultaneously. Associated process claims, if any, present elements of the various steps, operations or processes in a sampling order, and are not intended to be limited to the specific order or hierarchy presented. These can be done in series, linearly, in parallel, or in a different order. It should be understood that the instructions, operations and systems described can generally be integrated together in a single software / hardware product or packaged in several software / hardware products. In one aspect, the term "coupled" or a similar term can refer to direct coupling. In another aspect, the term "coupled" or a similar term may refer to an indirect coupling. Terms such as "top", "bottom", "front", "rear", "side", "horizontal", "vertical" and similar terms refer to an arbitrary frame of reference, rather than to ordinary gravitational reference frame. Thus, this term can also include “upwards”, “downwards”, “diagonally” or “horizontally” in a gravitational reference frame. The description is provided to allow the specialist in the field to practice the various aspects described here. In some cases, well-known structures and components are shown in the form of a block diagram in order to avoid obscuring the concepts of the technology of the invention. The description provides various examples of the technology which is the subject of the invention, and it is not limited to these examples. Various modifications of these aspects will be apparent to those skilled in the art, and the principles described herein can be applied to other aspects. All the structural and functional equivalents of the elements of the various aspects described in the description which are known or which will be known subsequently to the specialist in the field are expressly incorporated here by reference and are intended to be encompassed by the claims. Furthermore, nothing in this description is intended to be disclosed to the public, whether or not this description is explicitly mentioned in the claims. No claim element shall be interpreted in accordance with the provisions of Title 35 USC Section 112, sixth paragraph, unless the element is expressly described using the phrase "means for" or, in the case of a claim process, if the item is described using the phrase "step for". The title, the background, a brief description of the drawings, the abstract and the drawings are incorporated into the description and are provided as examples of the description, and not in the form of restrictive descriptions. It goes without saying that they will not be used to limit the scope or meaning of the claims. In addition, in the detailed description, it can be seen that the description provides illustrative examples and that the various characteristics are grouped together in various implementations in order to simplify the description. The description process should not be interpreted as reflecting an intention that the claimed object requires more features than those expressly stated in each claim. On the contrary, as the claims show, the subject of the invention consists of fewer features than all the features of a single configuration or operation described. The claims are hereby incorporated into the detailed description, each claim alone constituting an object claimed separately. The claims are not intended to be limited to the aspects described here, but should be given the full scope compatible with the language claims and encompass all legal equivalents. However, none of the claims is intended to cover an object which does not meet the requirements of applicable patent law, and should not be interpreted in this way either.
权利要求:
Claims (1) [1" id="c-fr-0001] [Claim 1] [Claim 2] [Claim 3] [Claim 4] claims Process comprising: receipt of initial input data including at least one of the production data or geological data relating to a reservoir; training, using at least the initial input data and a production flow record based on the production data, of a deep neural network (RNP) to model a simulation of proxy reservoir flow; carrying out, using an overall Kalman filter, the simulation of proxy reservoir flow based on the RNP formed; assimilating new data, using Kalman's set filter, updating a current set to match historicals, minimizing the differences between a predicted output from the simulation of proxy flow and production data measured from a deposit; carrying out, using the current updated set, a second simulation of proxy reservoir flow based on the RNP formed; repeating assimilation and realization while new data is available for assimilation; determining a predicted behavior of the reservoir based at least on the proxy flow simulation performed from the reservoir; and providing an indication of the predicted behavior to facilitate the production of fluids from the reservoir. The method of claim 1, wherein performing, using the overall Kalman filter, simulating the proxy flow of the reservoir includes performing a forecast operation of the overall Kalman filter. The method of claim 1 or 2, further comprising: generating, based at least in part on the received initial input data, an initial set of input parameters for the proxy flow simulation, wherein the initial set includes a number of embodiments of the input parameters. The method of claim 3, wherein the initial set of input parameters comprises parameters corresponding to petrophysical properties of porosity and permeability, optionally, in which the initial set is generated using a geostatistical technique, optionally, in which the initial set is used as input to the RNP, and optionally, in which the proxy flow simulation based on RNP is executed for each realization of the initial set in order to predict the production parameters from a previous time step t to a current time step t n . [Claim 5] Method according to any one of the preceding claims, in which the simulation of the proxy flow of the reservoir provides output rates and a petrophysical model. [Claim 6] The method of claim 5, further comprising:receiving new input data including new production data and new geological data;updating, using the overall Kalman filter, of the petrophysical model on the basis of a covariance matrix; andcarrying out, using the overall Kalman filter, the simulation of proxy reservoir flow on the basis of the RNP formed and the updated petrophysical model; [Claim 7] [Claim 7] The method of claim 6, wherein determining the predicted behavior of the reservoir is based at least in part on the updated petrophysical model, and the predicted behavior of the reservoir predicts the future rate of oil production from the reservoir to facilitate management of oil production from a production well at the reservoir. [Claim 8] System comprising:a processor; anda memory device comprising instructions which, when executed by the processor, cause the processor to:receive initial input data including at least one of the production data or geological data;form, using at least the initial input data and a production flow record based on the production data, a deep neural network (RNP) in order to model a simulation of a proxy flow of a reservoir ;run, using an overall Kalman filter, the simulation of proxy reservoir flow based on the RNP formed;determine a predicted behavior of the reservoir based on at least the simulation of proxy flow performed from the reservoir; and providing an indication of the predicted behavior to facilitate the production of fluids from the reservoir. [Claim 9] The system of claim 8, wherein performing, using the overall Kalman filter, the simulation of proxy reservoir flow comprises performing a forecast operation of the overall Kalman filter. [Claim 10] The system of claim 8 or claim 9, wherein the instructions further cause the processor to:generating, based at least in part on the initial input data received, an initial set of input parameters for the proxy flow simulation, the initial set comprising a number of embodiments of the input parameters. [Claim 11] System according to claim 10, in which the initial set of input parameters comprises parameters corresponding to the petrophysical properties of porosity and permeability, optionally, in which the initial set is generated using a geostatistical technique, and optionally, in which the initial set is used as an entry in the RNP. [Claim 12] The system of claim 10, wherein the proxy flow simulation based on the RNP is executed for each realization of the initial set in order to predict the production parameters from a previous time step t to a current time step t n [Claim 13] System according to any one of the preceding claims, in which the simulation of the proxy flow of the reservoir provides, at the output, production rates and a petrophysical model. [Claim 14] The system of claim 13, wherein the instructions further cause the processor to:receive new input data including new production data and new geological data;update, using the overall Kalman filter, the petrophysical model on the basis of a covariance matrix; andrun, using the overall Kalman filter, the proxy reservoir flow simulation based on the RNP formed and the updated petrophysical model. [Claim 15] Non-transient computer-readable media including ins- tructions stored therein which, when executed by at least one computing device, cause the at least one computing device to execute operations comprising: receiving initial input data including at least one of the production data or geological data; training, using at least the initial input data, a deep neural network (RNP) to model a simulation of a proxy reservoir flow; carrying out, using an overall Kalman filter, the simulation of proxy reservoir flow based on the RNP formed; assimilating new data, using Kalman's set filter, updating a current set to match historicals, minimizing the differences between a predicted output from the simulation of proxy flow and production data measured from a deposit; carrying out, using the current updated set, a second simulation of proxy reservoir flow based on the RNP formed; repeating assimilation and realization while new data is available for assimilation; determining a predicted behavior of the reservoir based at least on the proxy flow simulation performed from the reservoir; and providing an indication of the predicted behavior to facilitate the production of fluids from the reservoir.
类似技术:
公开号 | 公开日 | 专利标题 US9074454B2|2015-07-07|Dynamic reservoir engineering US20210166166A1|2021-06-03|Wellsite report system US8024123B2|2011-09-20|Subterranean formation properties prediction US8103493B2|2012-01-24|System and method for performing oilfield operations US8120357B2|2012-02-21|Method and system for fluid characterization of a reservoir FR3070180A1|2019-02-22|NEURON NETWORK MODELS FOR REAL-TIME OPTIMIZATION OF DRILLING PARAMETERS DURING DRILLING OPERATIONS US20200277848A1|2020-09-03|Well planning system FR3085053A1|2020-02-21|PREDICTION OF OIL TANK BEHAVIOR USING A PROXY FLOW MODEL TECHNICAL AREA FR3027944A1|2016-05-06|GENERATING STRUCTURAL ELEMENTS FOR SUBTERRANEAN FORMATION USING STRATIGRAPHIC IMPLICIT FUNCTION US8185311B2|2012-05-22|Multiuser oilfield domain analysis and data management FR3027050A1|2016-04-15|AUTOMATED FRACTURING PLANNING METHODS FOR MULTI-WELL FIELDS FR3034191B1|2019-08-23|DETERMINATION OF TRAINING PRESSURE FR3069929A1|2019-02-08|PREDICTION IN FRONT OF TREPAN USING VERTICAL SEISMIC PROFILE AND GLOBAL INVERSION US8099267B2|2012-01-17|Input deck migrator for simulators FR3086779A1|2020-04-03|MATCHING AN AUTOMATED PRODUCTION HISTORY USING BAYESIAN OPTIMIZATION FR3085404A1|2020-03-06|AUTOMATED OPTIMIZATION OF THE PENETRATION RATE FOR DRILLING US20210285316A1|2021-09-16|Real-Time Well Construction Process Inference Through Probabilistic Data Fusion CN110191999A|2019-08-30|Multilayer groundbed frontier distance | inverting carried out with multiple initial guess US8255816B2|2012-08-28|Modifying a magnified field model Montgomery2015|Characterizing shale gas and tight oil drilling and production performance variability FR3076316A1|2019-07-05|EFFECTIVE REPRESENTATION OF COMPLEX THREE-DIMENSIONAL SIMULATION RESULTS FOR REAL-TIME OPERATIONS FR3092691A1|2020-08-14|PLANNING A WELL CONFIGURATION USING GEOMECHANICAL PARAMETERS
同族专利:
公开号 | 公开日 WO2019221717A1|2019-11-21| GB202013222D0|2020-10-07| US20210027144A1|2021-01-28| CA3090956A1|2019-11-21| GB2585153A|2020-12-30| NO20201012A1|2020-09-15|
引用文献:
公开号 | 申请日 | 公开日 | 申请人 | 专利标题 US7584081B2|2005-11-21|2009-09-01|Chevron U.S.A. Inc.|Method, system and apparatus for real-time reservoir model updating using ensemble kalman filter| EP2391831A4|2009-01-30|2017-05-10|Chevron U.S.A., Inc.|System and method for predicting fluid flow in subterranean reservoirs| EP2599023B1|2010-07-29|2019-10-23|Exxonmobil Upstream Research Company|Methods and systems for machine-learning based simulation of flow| US10545260B2|2011-04-15|2020-01-28|Conocophillips Company|Updating geological facies models using the Ensemble Kalman filter| WO2017188858A1|2016-04-28|2017-11-02|Schlumberger Canada Limited|Reservoir performance system|WO2021127186A1|2019-12-18|2021-06-24|Schlumberger Technology Corporation|Methods and systems for subsurface modeling employing ensemble machine learning prediction trained with data derived from at least one external model| CN111783361B|2020-07-07|2021-03-12|中国人民解放军国防科技大学|Numerical weather forecast mixed data assimilation method based on triple multi-layer perceptron|
法律状态:
优先权:
[返回顶部]
申请号 | 申请日 | 专利标题 PCT/US2018/032816|WO2019221717A1|2018-05-15|2018-05-15|Petroleum reservoir behavior prediction using a proxy flow model| WOPCT/US2018/032816|2018-05-15| 相关专利
Sulfonates, polymers, resist compositions and patterning process
Washing machine
Washing machine
Device for fixture finishing and tension adjusting of membrane
Structure for Equipping Band in a Plane Cathode Ray Tube
Process for preparation of 7 alpha-carboxyl 9, 11-epoxy steroids and intermediates useful therein an
国家/地区
|