CN115375031A - Oil production prediction model establishing method, capacity prediction method and storage medium - Google Patents
Oil production prediction model establishing method, capacity prediction method and storage medium Download PDFInfo
- Publication number
- CN115375031A CN115375031A CN202211061735.6A CN202211061735A CN115375031A CN 115375031 A CN115375031 A CN 115375031A CN 202211061735 A CN202211061735 A CN 202211061735A CN 115375031 A CN115375031 A CN 115375031A
- Authority
- CN
- China
- Prior art keywords
- loss function
- prediction model
- output
- physical
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 67
- 238000004519 manufacturing process Methods 0.000 title claims abstract description 53
- 238000003860 storage Methods 0.000 title claims abstract description 13
- 238000012549 training Methods 0.000 claims abstract description 53
- 238000012360 testing method Methods 0.000 claims abstract description 37
- 238000010276 construction Methods 0.000 claims abstract description 32
- 239000003921 oil Substances 0.000 claims description 50
- 239000003129 oil well Substances 0.000 claims description 19
- 238000004422 calculation algorithm Methods 0.000 claims description 7
- 239000010779 crude oil Substances 0.000 claims description 5
- 238000004590 computer program Methods 0.000 claims description 4
- 230000006835 compression Effects 0.000 claims description 3
- 238000007906 compression Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000035699 permeability Effects 0.000 claims description 3
- 239000000446 fuel Substances 0.000 claims description 2
- 239000003079 shale oil Substances 0.000 abstract description 8
- 238000005293 physical law Methods 0.000 abstract description 6
- 230000006870 function Effects 0.000 description 124
- 238000010586 diagram Methods 0.000 description 10
- 230000015654 memory Effects 0.000 description 10
- 230000008569 process Effects 0.000 description 9
- 238000004891 communication Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 7
- 238000004364 calculation method Methods 0.000 description 6
- 230000007246 mechanism Effects 0.000 description 5
- 239000004215 Carbon black (E152) Substances 0.000 description 4
- 238000013461 design Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 4
- 229930195733 hydrocarbon Natural products 0.000 description 4
- 150000002430 hydrocarbons Chemical class 0.000 description 4
- 238000004140 cleaning Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000004913 activation Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000013434 data augmentation Methods 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000013277 forecasting method Methods 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 235000004257 Cordia myxa Nutrition 0.000 description 1
- 244000157795 Cordia myxa Species 0.000 description 1
- 239000002253 acid Substances 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000005553 drilling Methods 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- Biophysics (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Primary Health Care (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Provided herein are a method for establishing an oil production prediction model, a capacity prediction method, and a storage medium, including: acquiring a training set and a test set which are cleaned, wherein the training set and the test set comprise a plurality of groups of historical data including geological parameters, construction parameters and oil production; inputting geological parameters and construction parameters in the training set into the initialized prediction model to obtain a model output value; determining an error term of the model output value and the real value in the test set by using a loss function, wherein the loss function is determined according to the real loss function and the physical loss function; the weight coefficients of all nodes in the current prediction model are updated according to the loss function to obtain the trained prediction model, the loss function is corrected according to the physical loss function, the physical law of the shale oil and gas reservoir is considered, the weight coefficients of all nodes are updated by determining the loss function through the real loss function and the physical loss function, and the convergence speed and the prediction precision of the model can be improved.
Description
Technical Field
The invention relates to the technical field of geological exploration and development, in particular to an oil production prediction model establishing method, an oil production capacity prediction method and a storage medium.
Background
Currently, methods for predicting the capacity of unconventional oil and gas fields such as shale oil and gas reservoirs based on data mining have been discussed and studied. In summary, researchers are divided into two research ideas:
the first method is to perform fitting of unary nonlinear regression and multiple linear regression according to collected historical drilling data, and then predict the productivity through the unary nonlinear regression and multiple linear regression method after fitting.
The second method is to predict the productivity based on a machine learning technology, for example, predict the productivity through a prediction model, however, in the field of shale hydrocarbon reservoir prediction, the design of a loss function does not consider the physical law of the shale hydrocarbon reservoir, which causes the problems of too slow model convergence or poor model precision.
Disclosure of Invention
In view of the above problems in the prior art, an object of the present disclosure is to provide an oil production prediction model establishing method, an energy production prediction method, and a storage medium, so as to solve the problem that in the prior art, the physical law of a shale oil and gas reservoir is not considered in the design of a loss function, which causes too slow model convergence or poor model accuracy.
In order to solve the technical problems, the specific technical scheme is as follows:
in one aspect, a method for establishing an oil production prediction model is provided herein, including:
acquiring a training set and a test set which are cleaned, wherein the training set and the test set comprise a plurality of groups of historical data, and the historical data comprise geological parameters, construction parameters and corresponding oil production;
inputting the geological parameters and the construction parameters in the training set into a prediction model after initialization to obtain a model output value;
determining an error term of the model output value and a real value in the test set by using a loss function, wherein the loss function is determined according to the real loss function and a physical loss function;
and updating the weight coefficient of each node in the current prediction model according to the loss function to obtain the trained prediction model.
As an embodiment herein, the loss function is determined according to a real loss function and a physical loss function, and further includes:
determining the true loss function from the model output values and the test set;
determining the physical loss function from the model output values and physical constraints of the target well;
and determining the loss function according to a penalty factor, a feature quantity, the real loss function and the physical loss function, wherein the feature quantity is determined according to the group number of the historical data in the training set.
As an embodiment herein, the determining the physical loss function from the model output values and the physical constraints of the target well further comprises:
according to a physical constraint formula:
determining a physical constraint Q of the target well theory Wherein n is F S is a Laplace space variable for the number of fractures in the target oil well pattern,determining in Laplace space according to geological features, wherein the geological features comprise matrix porosity/permeability, reservoir geometric parameters, crude oil viscosity, crude oil compression ratio, production time, bottom hole temperature, initial reservoir pressure and bottom hole flowing pressure;
the physical loss function is E th =(Q NN -Q theory ) 2 Wherein Q is NN The output yield obtained after inputting the test set to the prediction model is obtained.
As one embodiment herein, the determining the true loss function from the model output values and the test set further comprises:
the true loss function is E re =(Q NN -Q real ) 2 Wherein Q is real The true production of the well pattern is concentrated for the test.
As an embodiment herein, the determining the loss function from a penalty factor, a feature quantity, the true loss function, and the physical loss function further comprises:
the loss function is:
wherein n is the number of features and λ is a penalty factor.
As an embodiment herein, after updating the weight coefficients of each node in the current prediction model according to the loss function to obtain the trained prediction model, the method includes:
the first penalty factor lambda is brought into a loss function, and the output yield Q is obtained through prediction of a prediction model with the loss function NN ;
Generating a plurality of second penalty factors lambda 'through a random disturbance algorithm and substituting the second penalty factors lambda' into a loss function, and predicting through a prediction model with the loss function to obtain a plurality of output yields Q NN ′;
Judging output Q NN And output yield Q NN ' size relationship;
if output the output Q NN Greater than output yield Q NN ', the first penalty factor lambda is used as the penalty factor of the current prediction model.
As one embodiment herein, the determining outputs yield Q NN And output yield Q NN ' further comprising:
if output the yield Q NN Less than output yield Q NN Acquiring an initial temperature and a random number, wherein the random number is 0-1;
determining an adoption probability from a formulaWherein T is the initial temperature,. DELTA.E k To output a yield Q NN ' subtract output yield Q NN A difference of (d);
if the adoption probability P k If the number is larger than the random number, taking a second penalty factor lambda' as a penalty factor of the current prediction model;
if the adoption probability P k And if the number is smaller than the random number, taking the first penalty factor lambda as the penalty factor of the current prediction model.
As an embodiment herein, the updating the weight coefficient of each node in the current prediction model according to the loss function to obtain the trained prediction model further includes:
according to gradient descent formula
Updating the weight coefficient omega of each node i+1 Where α is the iteration step coefficient, ω i Is the weight coefficient before updating.
In another aspect, a capacity forecasting method is provided herein, including: and predicting the oil yield of the target oil well according to the geological parameters of the target oil well and the construction parameters of the target oil well by adopting the prediction model established by any oil yield prediction model establishing method.
In another aspect, a computer-readable storage medium is provided, and a computer program is stored in the computer-readable storage medium, and when executed by a processor, the computer program implements any one of the oil production prediction model establishment methods.
By adopting the technical scheme, the oil production prediction model building method realizes the purpose of obtaining a large amount of historical data of oil production by obtaining a training set and a testing set which are cleaned, wherein the training set and the testing set comprise a plurality of groups of historical data, the historical data comprise geological parameters, construction parameters and corresponding oil production, the historical data comprise a plurality of groups of parameters, and each group of parameters comprises a geological parameter, a construction parameter and a corresponding real oil production; obtaining a model output value by inputting the geological parameters and the construction parameters in the training set into a prediction model after initialization, so as to obtain a corresponding prediction value of each pair of geological parameters and construction parameters in the prediction model in each training set, namely a model output value; determining an error term between the model output value and the true value in the test set by using a loss function, wherein the loss function is determined according to the true loss function and a physical loss function, so that the loss function is corrected according to the physical loss function, and the physical loss function takes the physical rule of the shale oil and gas reservoir into consideration; the trained prediction model is obtained by updating the weight coefficient of each node in the current prediction model according to the loss function, so that the effect that the weight coefficient of each node is updated by jointly determining the loss function through the real loss function and the physical loss function is realized, the convergence speed of the model can be increased, and the accuracy is improved.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the embodiments or technical solutions in the prior art, the drawings used in the embodiments or technical solutions in the prior art are briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present disclosure, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is an overall system diagram illustrating a method for modeling an oil production prediction in accordance with an embodiment of the present disclosure;
FIG. 2 is a schematic diagram illustrating steps of a method for modeling fuel production prediction according to an embodiment of the present disclosure;
FIG. 3 is a schematic diagram illustrating a method of generating historical data augmentation for an anti-neural network according to an embodiment of the present disclosure;
FIG. 4 is a schematic diagram illustrating a loss function determination method according to an embodiment herein;
FIG. 5 is a schematic diagram illustrating a penalty factor adjustment method according to an embodiment of the present disclosure;
FIG. 6 illustrates a flow chart of oil production prediction model creation according to an embodiment herein;
FIG. 7 is a flow chart illustrating penalty factor update according to an embodiment herein;
FIG. 8 illustrates an oil production prediction model building apparatus according to an embodiment of the present disclosure;
FIG. 9 is a schematic diagram of a computer device according to an embodiment of the present disclosure.
Description of the symbols of the drawings:
101. a database;
102. an arithmetic server;
801. an acquisition unit;
802. an input unit;
803. a loss function unit;
804. an updating unit;
902. a computer device;
904. a processor;
906. a memory;
908. a drive mechanism;
910. an input/output module;
912. an input device;
914. an output device;
916. a presentation device;
918. a graphical user interface;
920. a network interface;
922. a communication link;
924. a communication bus.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments herein without making any creative effort, shall fall within the scope of protection.
It should be noted that the terms "first," "second," and the like in the description and claims herein and in the above-described drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments herein described are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, apparatus, article, or device that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or device.
As shown in fig. 1, an overall system diagram of a method for establishing an oil production prediction model includes: a database and an arithmetic server;
the database 101 is used for storing historical data in the oil well exploitation history, and in this document, the historical data includes a plurality of sets of parameters corresponding to each other, including geological parameters, construction parameters and oil production corresponding to the geological parameters and the construction parameters, and is shown in a specific table 1.
TABLE 1
Name of oil well | Geological parameters | Construction parameters | Oil production |
108 # oil well in area A | 1,8,9… | 211,78,79,89… | 1000 ton of |
No. 109 oil well in B area | 1,10,9… | 201,79,69,19… | 900 ton of |
110 # oil well in C area | 2,10,10… | 111,98,59,29… | 780 ton (ton) |
In this context, the geological parameters are composed of young's modulus, total organic carbon, maximum principal stress, minimum principal stress, total gas content, porosity, poisson's ratio, fracture pressure, brittleness index, brittle minerals, etc. in order to save space, quantitative substitutions are made in table 1 by several numbers, and the specific geological parameters can be designed by those skilled in the art according to engineering requirements.
In the present text, the construction parameters include hydraulic strength, total acid solution amount, total slippery water amount, total linear gel amount, sand feeding strength, average pump pressure, average displacement, segment length, cluster spacing, cluster number, etc. in order to save space, quantitative substitution is performed by using several numbers in table 1, and specific construction parameters can be designed by those skilled in the art according to engineering requirements.
In this context, the geological and construction parameters in fig. 1 are for illustration, and therefore the oil production is not limited.
The calculation server 102 is configured to run several types of prediction models, such as the oil quantity prediction model herein, and initialization of the oil quantity prediction model may be performed by the calculation server by assigning a random number initial value, or by self-assigning an initial value according to a published document, which is not limited herein. And when the initialized prediction model is operated, the operation server acquires a training set and a test set through the database and carries out loop iteration training.
The operation server 102 is further configured to store a loss function of the initialized prediction model, where the loss function is determined according to a real loss function and a physical loss function, and particularly, the physical loss function is set according to a physical constraint.
For example, there are six sets of historical data in the training set, after the training of the six sets of historical data is completed, the calculation server may update the weight coefficient of each node in the prediction model according to the training result, and after the predetermined number of times of training is completed, the trained prediction model may be obtained.
Preferably, after completing the training of the training set, the first-order loss function may be updated according to the prediction accuracy of the prediction model, as will be described in detail below.
Currently, methods for predicting the capacity of unconventional oil and gas fields such as shale oil and gas reservoirs based on data mining have been discussed and studied. In summary, researchers are divided into two research ideas:
one of the methods is to predict productivity based on a machine learning technology, for example, predict productivity through a prediction model, however, in the field of shale hydrocarbon reservoir prediction, researchers do not consider the physical law of shale hydrocarbon reservoirs in the design of a loss function, which causes the problems of too slow model convergence or poor model precision.
In order to solve the above problems, embodiments herein provide a method for building an oil production prediction model, which can take into account the physical laws of shale oil and gas reservoirs, improve the convergence rate of the prediction model, and improve the prediction accuracy of the prediction model, and fig. 2 is a schematic diagram of steps of a method for building an oil production prediction model provided in embodiments herein, where the present specification provides the method operation steps as described in the embodiments or flowcharts, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps recited in the embodiments is merely one manner of performing the steps in a multitude of orders and does not represent the only order of execution. In the actual implementation of the system or the device product, the method according to the embodiments or shown in the drawings can be executed in sequence or in parallel. Specifically, as shown in fig. 2, the method may include:
And 203, determining an error term between the model output value and the real value in the test set by using a loss function, wherein the loss function is determined according to the real loss function and the physical loss function.
And 204, updating the weight coefficient of each node in the current prediction model according to the loss function to obtain the trained prediction model.
Obtaining a large amount of historical data of oil extraction by obtaining a training set and a test set which are cleaned, wherein the training set and the test set comprise a plurality of groups of historical data, the historical data comprises geological parameters, construction parameters and corresponding oil production, the historical data comprises a plurality of groups of parameters, and each group of parameters comprises a geological parameter, a construction parameter and corresponding real oil production; obtaining a model output value by inputting the geological parameters and the construction parameters in the training set into a prediction model after initialization, so as to obtain a corresponding prediction value of each pair of geological parameters and construction parameters in the prediction model in each training set, namely a model output value; determining an error term between the model output value and the real value in the test set by using a loss function, wherein the loss function is determined according to the real loss function and a physical loss function, so that the loss function is corrected according to the physical loss function, and the physical loss function considers the physical rule of the shale oil and gas reservoir; the trained prediction model is obtained by updating the weight coefficient of each node in the current prediction model according to the loss function, so that the loss function is determined jointly through the real loss function and the physical loss function to update the weight coefficient of each node, the convergence speed of the model can be increased, and the precision can be improved.
As an embodiment herein, in step 201, a training set and a test set of cleaning completion are obtained, wherein the training set and the test set include several sets of historical data, and the historical data includes geological parameters, construction parameters and corresponding oil production amount, and before:
historical data in the database is cleaned, for example, by data alignment or by removing NULL values.
As an embodiment of this document, step 202, inputting both the geological parameters and the construction parameters in the training set into the initialized prediction model to obtain a model output value, further including:
the model output values may also be calculated based on an ANN neural network forward calculation formula (activation function), where the activation function includes SELU, tanh, sigmoid, and linear.
As one embodiment herein, a schematic diagram of a method of generating historical data augmentation for an antineural network is shown in fig. 3.
Initializing a generator G (z): in this context, the model of the generator may be a model of CNN, RNN, LSTM, etc., which is used to capture the distribution characteristics of the original historical data set and introduce random noise to generate new false data, and the selected parameter field may be all parameters of the original sample data set, or may be a combination of partial parameters, including reservoir thickness, formation pressure, bottom hole pressure, production dynamic curve, etc.
Initializing a discriminator model D (x): in this context, the discriminator may be a CNN, RNN, LSTM, or like model that functions to determine the probability that the false history data generated by the generator belongs to the original sample true history data set.
The objective function of the generator is:
the objective function of the discriminator is:
wherein D (x) is an output result of the discriminator, is a real number in the range of 0 to 1, represents a probability of discriminating data as real data, and p data Representing the distribution of real data, p z Indicating the distribution of the generated data, and E indicates the expected value of the distribution specified in the subscript. The goal of the discriminator is to make the objective function V (D, G) as large as possible with both true and spurious data input; the goal of the generator is to make the objective function V (D, G) of the discriminator as small as possible.
In this context, the prediction accuracy of discriminator D (x) is made 50% by updating the generator, i.e. it is difficult to distinguish whether the data set is from a real sample or is falsely distorted.
Specifically, the discriminator gradient is updated as follows
The generator gradient updates as follows:
where m denotes the number of samples, Z denotes the noise or random samples input to the G-network, θ d 、θ g Target update parameters (representing the magnitude of the effect of the features on the predicted values) for the generator and discriminator, respectively.
As shown in fig. 4, as an embodiment herein, the method for determining a loss function is determined according to a real loss function and a physical loss function, and further includes:
In this step, the true loss function is E re =(Q NN -Q real ) 2 Wherein Q is real The true production of the well pattern is concentrated for the test.
Wherein Q NN In order to obtain the output yield after inputting the test set into the prediction model, the real yield and the output yield are subjected to difference making to obtain a difference value E re And the difference E is compared re And the loss function is used for being supplemented to subsequent prediction so as to improve the prediction accuracy of the prediction model.
In this step, according to the physical constraint formula:
determining physical constraints Q of the target well theory Wherein n is F S is a Laplace space variable for the number of fractures in the target oil well pattern,determined in Laplace space according to geological features, wherein the geological features comprise matrix porosity/permeability, reservoir geometric parameters and crude oil viscosityCrude oil compression ratio, production time, bottom hole temperature, initial reservoir pressure, and bottom hole flow pressure.
The physical loss function is E th =(Q NN -Q theory ) 2 Wherein Q is NN The output yield obtained after inputting the test set into the prediction model is obtained.
That is, the theoretical yield Q is obtained according to the physical condition of the region of the target oil well theory In this way, artificially specified yield can be introduced, and then the loss function is adjusted to accelerate the convergence speed of the prediction model.
And step 403, determining the loss function according to a penalty factor, a feature quantity, the real loss function and the physical loss function, wherein the feature quantity is determined according to the group number of the historical data in the training set.
In this step, the loss function is:
wherein n is the number of features and λ is a penalty factor.
In this context, the feature quantity and the feature quantity are determined according to the number of sets of the historical data in the training set, that is, if there are six historical data in the training set, the feature quantity is six, and if there are seven historical data in the training set, the feature quantity is seven. The number of features in this context corresponds to the number of sets of historical data in the training set.
In this document, the loss function is related to a penalty factor λ, so that in order to make the effect of the loss function better and make the generalization performance and robustness of the prediction data more superior, an adjustment method of the penalty factor is provided herein, which can avoid that the global optimum point cannot be found after the loss function falls into the local optimum point.
As an embodiment herein, a schematic diagram of a penalty factor adjustment method as shown in fig. 5 includes:
In this step, the current penalty factor λ may be brought into the loss function to obtain the corresponding Q NN 。
502, generating a plurality of second penalty factors lambda 'through a random disturbance algorithm and bringing the second penalty factors lambda' into a loss function, and predicting through a prediction model with the loss function to obtain a plurality of output yields Q NN ′。
In this step, the second penalty factor λ 'is generated according to the first penalty factor λ and a random number generated by a random perturbation algorithm, specifically, the random perturbation algorithm generates a random number, and then adds the random number to the first penalty factor λ to obtain the second penalty factor λ'.
In order to better adjust the loss function, second penalty factors λ 'can be generated in batches, and after each second penalty factor λ' is led into the loss function in batches, a plurality of adjusted loss functions are obtained. Training the prediction model with the adjusted loss function to obtain a trained prediction model corresponding to each second penalty factor lambda', importing the same group of historical data into different trained prediction models respectively to obtain a plurality of output yields Q NN ′。
In this step, since the prediction model is always subject to an error, it is found in the conventional calculation that the error is likely to be smaller than the actual value of the yield, and therefore, the output yield Q is compared NN And output yield Q NN ' the magnitude relationship, i.e. which one of them is the smallest, demonstrates the smallest deviation from the actual value of production.
In this step, the yield Q is output NN Greater than outputQuantity Q NN ', then the output yield Q is proved NN Deviation from actual yield is minimized, so yield Q will be output NN And taking the corresponding first penalty factor lambda as a penalty factor of the current prediction model. Namely, the prediction model with the loss function is ensured to have stronger fitting effect.
In particular, the output yield Q corresponding to the second penalty factor λ' when randomly generated NN ' the fitting effect is better than the output yield Q corresponding to the first penalty factor lambda NN The loss function with the first penalty factor λ is proved not to be globally optimal. When the current loss function is not the global optimum, a penalty factor corresponding to the loss function in the global optimum is required to be searched.
As one example herein, determining the output yield Q as set forth in step 503 NN And output yield Q NN ' further comprising:
if output the yield Q NN Less than output yield Q NN Acquiring an initial temperature and a random number, wherein the random number is 0-1;
determining the probability of adoption from a formulaWherein T is the initial temperature, Δ E k To output a yield Q NN ' subtract output yield Q NN A difference of (d);
if the adoption probability P k If the number is larger than the random number, taking a second penalty factor lambda' as a penalty factor of the current prediction model;
if the adoption probability P k And if the number is smaller than the random number, taking the first penalty factor lambda as the penalty factor of the current prediction model.
In order to avoid the gradual increase or decrease of the loss function, the local optimum point cannot be skipped, and the curing is performed at the local optimum point. It is therefore determined whether the random number falls within an acceptance probability obtained by the initial temperature, which in this context may be 0.5, i.e. by means of two non-fixed numbers, the loss function can be made to jump extensively to jump out of the local optimum.
As an embodiment herein, the updating the weight coefficient of each node in the current prediction model according to the loss function to obtain the trained prediction model further includes:
according to gradient descent formula
Updating the weight coefficient omega of each node i+1 Where α is the iteration step coefficient, ω i Is the weight coefficient before updating.
In this step, in order to improve the fitting effect of the prediction model, the weight coefficients in the prediction model may be updated by a loss function.
In this context, the weights may also be updated according to an optimization algorithm based on gradients, including but not limited to SGD, RMSprop, adaGrad, adam, and Nadam.
In order to make the training process of the prediction model more clear to those skilled in the art, a flow chart of the oil production prediction model establishment as shown in fig. 6 is given.
And step 604, updating the weight coefficient of each node in the prediction model according to the current loss function. In this step, the weight coefficients of the nodes of the hidden layer in the prediction model may be updated.
And 605, judging whether to update the penalty factor in the loss function, if so, executing the step 606, and if not, finishing the training of the prediction model. In this step, the iterative effect of the loss function can be improved.
And step 606, updating the penalty factor and finishing the training of the prediction model. In this step, the penalty factor update flow chart is shown in fig. 7. Penalty factor updating is accomplished through this figure 7. Fig. 7 includes:
And 707, determining whether the iteration number reaches an iteration threshold, if so, executing step 708, and if not, returning to step 703.
As shown in fig. 8, an oil production prediction model creation apparatus includes:
an obtaining unit 801, configured to obtain a training set and a test set that are completed by cleaning, where the training set and the test set include a plurality of sets of historical data, and the historical data includes geological parameters, construction parameters, and corresponding oil production amounts;
an input unit 802, configured to input the geological parameters and the construction parameters in the training set into a prediction model after initialization to obtain a model output value;
a loss function unit 803, configured to determine an error term between the model output value and the actual value in the test set by using a loss function, where the loss function is determined according to the actual loss function and a physical loss function;
and the updating unit 804 is configured to update the weight coefficient of each node in the current prediction model according to the loss function, so as to obtain the trained prediction model.
The acquisition unit is used for acquiring a large amount of historical data of oil extraction, the historical data comprises a plurality of sets of parameters, and each set of parameters comprises a geological parameter, a construction parameter and a real oil production amount corresponding to the geological parameter and the construction parameter; through an input unit, the corresponding predicted values of each pair of geological parameters and construction parameters in the prediction model in each training set, namely model output values, can be obtained; through the loss function unit, the loss function is corrected according to the physical loss function, and the physical loss function takes the physical law of the shale oil and gas reservoir into consideration; through the updating unit, the loss function is determined jointly through the real loss function and the physical loss function to update the weight coefficient of each node, so that the convergence speed of the model can be increased, and the accuracy can be improved.
Also provided herein is a capacity forecasting method comprising: and predicting the oil yield of the target oil well according to the geological parameters of the target oil well and the construction parameters of the target oil well by adopting the prediction model established by any oil yield prediction model establishing method.
As shown in fig. 9, for a computer device provided for embodiments herein that runs the oil production prediction model building method herein, the computer device 902 may include one or more processors 904, such as one or more Central Processing Units (CPUs), each of which may implement one or more hardware threads. The computer device 902 may also include any memory 906 for storing any kind of information, such as code, settings, data, etc. For example, and without limitation, the memory 906 may include any one or combination of the following: any type of RAM, any type of ROM, flash memory devices, hard disks, optical disks, etc. More generally, any memory may use any technology to store information. Further, any memory may provide volatile or non-volatile retention of information. Further, any memory may represent fixed or removable components of computer device 902. In one case, when the processor 904 executes the associated instructions, which are stored in any memory or combination of memories, the computer device 902 can perform any of the operations of the associated instructions. The computer device 902 also includes one or more drive mechanisms 908, such as a hard disk drive mechanism, an optical disk drive mechanism, etc., for interacting with any memory.
Corresponding to the methods in fig. 2-7, the embodiments herein also provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above-described method.
Embodiments herein also provide computer readable instructions, wherein a program therein causes a processor to perform the method as shown in fig. 2-7 when the instructions are executed by the processor.
It should be understood that, in various embodiments herein, the sequence numbers of the above-mentioned processes do not mean the execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation to the implementation process of the embodiments herein.
It should also be understood that, in the embodiments herein, the term "and/or" is only one kind of association relation describing an associated object, meaning that three kinds of relations may exist. For example, a and/or B, may represent: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship.
Those of ordinary skill in the art will appreciate that the various illustrative components and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the components and steps of the various examples have been described above generally in terms of their functionality in order to clearly illustrate this interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.
It can be clearly understood by those skilled in the art that, for convenience and simplicity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided herein, it should be understood that the disclosed system, apparatus, and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one type of logical functional division, and other divisions may be realized in practice, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may also be an electric, mechanical or other form of connection.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purposes of the embodiments herein.
In addition, functional units in the embodiments herein may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present invention may be implemented in a form of a software product, which is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk, and various media capable of storing program codes.
The principles and embodiments of the present disclosure are explained in detail by using specific embodiments, and the above description of the embodiments is only used to help understanding the method and its core idea; meanwhile, for the general technical personnel in the field, according to the idea of this document, there may be changes in the concrete implementation and the application scope, in summary, this description should not be understood as the limitation of this document.
Claims (10)
1. A method for establishing an oil production prediction model is characterized by comprising the following steps:
acquiring a training set and a test set which are cleaned, wherein the training set and the test set comprise a plurality of groups of historical data, and the historical data comprise geological parameters, construction parameters and corresponding oil production;
inputting the geological parameters and the construction parameters in the training set into a prediction model after initialization to obtain a model output value;
determining an error term of the model output value and a real value in the test set by using a loss function, wherein the loss function is determined according to the real loss function and a physical loss function;
and updating the weight coefficient of each node in the current prediction model according to the loss function to obtain the trained prediction model.
2. The method for establishing the oil production prediction model according to claim 1, wherein the loss function is determined according to a real loss function and a physical loss function, and further comprises:
determining the true loss function from the model output values and the test set;
determining the physical loss function from the model output values and physical constraints of the target well;
and determining the loss function according to a penalty factor, a feature quantity, the real loss function and the physical loss function, wherein the feature quantity is determined according to the group number of the historical data in the training set.
3. The method of claim 2, wherein the determining the physical loss function based on the model output values and the physical constraints of the target well further comprises:
according to a physical constraint formula:
determining a physical constraint Q of the target well theory Wherein n is F S is a Laplace space variable for the number of fractures in the target oil well pattern,determining in Laplace space according to geological features, wherein the geological features comprise matrix porosity/permeability, reservoir geometric parameters, crude oil viscosity, crude oil compression ratio, production time, bottom hole temperature, initial reservoir pressure and bottom hole flowing pressure;
the physical loss function is E th =(Q NN -Q theory ) 2 Wherein Q is NN The output yield obtained after inputting the test set into the prediction model is obtained.
4. The method of building an oil production prediction model of claim 3 wherein said determining said true loss function from said model output values and said test set further comprises:
the true loss function is E re =(Q NN -Q real ) 2 Wherein Q is real The true production of the well pattern is concentrated for the test.
5. The method for building an oil production prediction model according to claim 4, wherein the determining the loss function according to a penalty factor, a feature quantity, the true loss function, and the physical loss function further comprises:
the loss function is:
wherein n is the feature number, and λ is a penalty factor.
6. The method for building the oil production prediction model according to claim 5, wherein after the weight coefficients of the nodes in the current prediction model are updated according to the loss function to obtain the trained prediction model, the method comprises the following steps:
substituting the first penalty factor lambda into the loss function, and predicting through a prediction model with the loss function to obtain the output yield Q NN ;
Generating a plurality of second penalty factors lambda 'through a random disturbance algorithm and substituting the second penalty factors lambda' into a loss function, and predicting through a prediction model with the loss function to obtain a plurality of output yields Q NN ′;
Judging output yield Q NN And output yield Q NN ' size relationship;
if output the yield Q NN Greater than output yield Q NN ', the first penalty factor lambda is used as the penalty factor of the current prediction model.
7. The method for modeling the fuel production according to claim 6, wherein the determination output yield Q is NN And output yield Q NN ' further comprising:
if output the yield Q NN Less than output yield Q NN ', acquiring an initial temperature and a random number, wherein the random number is 0-1;
determining the probability of adoption from a formulaWherein T is the initial temperature, Δ E k To output a yield Q NN ' subtract output yield Q NN A difference of (d);
if the adoption probability P k Greater than the random number, will beTaking the two penalty factors lambda' as the penalty factors of the current prediction model;
if the adoption probability P k And if the number is smaller than the random number, taking the first penalty factor lambda as the penalty factor of the current prediction model.
8. The method for establishing the oil production prediction model according to claim 1, wherein the step of updating the weight coefficient of each node in the current prediction model according to the loss function to obtain the trained prediction model further comprises the steps of:
according to the formula of gradient descent
Updating the weight coefficient omega of each node i+1 Where α is the iteration step coefficient, ω i Is the weight coefficient before update.
9. A productivity prediction method which predicts the oil production of a target oil well with respect to the geological parameters of the target oil well and the construction parameters of the target oil well using a prediction model created by the oil production prediction model creation method according to any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program which, when executed by a processor, implements the oil production prediction model building method of any one of claims 1 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211061735.6A CN115375031A (en) | 2022-08-31 | 2022-08-31 | Oil production prediction model establishing method, capacity prediction method and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211061735.6A CN115375031A (en) | 2022-08-31 | 2022-08-31 | Oil production prediction model establishing method, capacity prediction method and storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115375031A true CN115375031A (en) | 2022-11-22 |
Family
ID=84069541
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211061735.6A Pending CN115375031A (en) | 2022-08-31 | 2022-08-31 | Oil production prediction model establishing method, capacity prediction method and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115375031A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116401935A (en) * | 2023-02-21 | 2023-07-07 | 哈尔滨工业大学 | Building dynamic thermal load neural network prediction method and system |
CN116432534A (en) * | 2023-04-18 | 2023-07-14 | 中海石油(中国)有限公司上海分公司 | Data-driven TOC sample prediction method |
CN116956049A (en) * | 2023-09-19 | 2023-10-27 | 中国联合网络通信集团有限公司 | Training method, device, equipment and storage medium of industrial productivity prediction model |
CN117522173A (en) * | 2024-01-04 | 2024-02-06 | 山东科技大学 | Natural gas hydrate depressurization exploitation productivity prediction method based on deep neural network |
CN117851928A (en) * | 2024-03-07 | 2024-04-09 | 特雷西能源科技股份有限公司 | Method and system for predicting shale oil yield based on CNN-LSTM model |
-
2022
- 2022-08-31 CN CN202211061735.6A patent/CN115375031A/en active Pending
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116401935A (en) * | 2023-02-21 | 2023-07-07 | 哈尔滨工业大学 | Building dynamic thermal load neural network prediction method and system |
CN116432534A (en) * | 2023-04-18 | 2023-07-14 | 中海石油(中国)有限公司上海分公司 | Data-driven TOC sample prediction method |
CN116956049A (en) * | 2023-09-19 | 2023-10-27 | 中国联合网络通信集团有限公司 | Training method, device, equipment and storage medium of industrial productivity prediction model |
CN117522173A (en) * | 2024-01-04 | 2024-02-06 | 山东科技大学 | Natural gas hydrate depressurization exploitation productivity prediction method based on deep neural network |
CN117522173B (en) * | 2024-01-04 | 2024-04-26 | 山东科技大学 | Natural gas hydrate depressurization exploitation productivity prediction method based on deep neural network |
CN117851928A (en) * | 2024-03-07 | 2024-04-09 | 特雷西能源科技股份有限公司 | Method and system for predicting shale oil yield based on CNN-LSTM model |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN115375031A (en) | Oil production prediction model establishing method, capacity prediction method and storage medium | |
Islam et al. | A holistic review on artificial intelligence techniques for well placement optimization problem | |
EP1984860B1 (en) | Methods, systems, and computer-readable media for fast updating of oil and gas field production models with physical and proxy simulators | |
CN107153843B (en) | Ground settlement prediction system and method based on support vector machine | |
Abdollahzadeh et al. | Estimation of distribution algorithms applied to history matching | |
WO2021055954A1 (en) | Machine-learning based system for virtual flow metering | |
CN105452598B (en) | The method for selecting and optimizing the oil field control for yield platform | |
CN104732067A (en) | Industrial process modeling forecasting method oriented at flow object | |
WO2023133213A1 (en) | Method for automated ensemble machine learning using hyperparameter optimization | |
Salehian et al. | A multisolution optimization framework for well placement and control | |
Wang et al. | A novel surrogate-assisted multi-objective optimization method for well control parameters based on tri-training | |
Sankaran et al. | Data Analytics in Reservoir Engineering | |
Korovin et al. | Application of hybrid data mining methods to increase profitability of heavy oil production | |
CN117574755B (en) | Hierarchical multistage optimization method for fracturing parameters of horizontal well of shale reservoir well factory | |
CN113537706A (en) | Oil field production increasing measure optimization method based on intelligent integration | |
CN117076993A (en) | Multi-agent game decision-making system and method based on cloud protogenesis | |
CN112766548A (en) | Order completion time prediction method based on GASA-BP neural network | |
US20230167717A1 (en) | A method of modelling a production well | |
CN112749835A (en) | Reservoir productivity prediction method, device and equipment | |
CN112906760A (en) | Horizontal well fracturing segment segmentation method, system, equipment and storage medium | |
CN116911216B (en) | Reservoir oil well productivity factor assessment and prediction method | |
US20230142526A1 (en) | Systems and methods of predictive decline modeling for a well | |
Liao et al. | Predictive Analytics and Statistical Learning for Waterflooding Operations in Reservoir Simulations | |
CN117236195B (en) | Machine learning offline agent model production optimization method for reducing development risk | |
CN117435904B (en) | Single feature ordering and composite feature extraction method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |