CN114358434A - Drilling machine drilling speed prediction method based on LSTM recurrent neural network model - Google Patents
Drilling machine drilling speed prediction method based on LSTM recurrent neural network model Download PDFInfo
- Publication number
- CN114358434A CN114358434A CN202210022755.6A CN202210022755A CN114358434A CN 114358434 A CN114358434 A CN 114358434A CN 202210022755 A CN202210022755 A CN 202210022755A CN 114358434 A CN114358434 A CN 114358434A
- Authority
- CN
- China
- Prior art keywords
- data
- drilling
- input
- model
- neural network
- 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
- 238000005553 drilling Methods 0.000 title claims abstract description 105
- 230000000306 recurrent effect Effects 0.000 title claims abstract description 41
- 238000000034 method Methods 0.000 title claims abstract description 22
- 238000003062 neural network model Methods 0.000 title claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 28
- 230000035515 penetration Effects 0.000 claims abstract description 24
- 238000013528 artificial neural network Methods 0.000 claims abstract description 18
- 238000010606 normalization Methods 0.000 claims abstract description 9
- 238000012512 characterization method Methods 0.000 claims abstract description 3
- 238000012545 processing Methods 0.000 claims abstract description 3
- 238000012360 testing method Methods 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 19
- 230000015654 memory Effects 0.000 claims description 12
- 230000004913 activation Effects 0.000 claims description 7
- 238000011156 evaluation Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 4
- 239000012530 fluid Substances 0.000 claims description 4
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 238000013139 quantization Methods 0.000 claims description 2
- 238000012546 transfer Methods 0.000 claims description 2
- 230000009466 transformation Effects 0.000 claims description 2
- 230000007547 defect Effects 0.000 abstract description 2
- 230000000875 corresponding effect Effects 0.000 description 5
- 210000002569 neuron Anatomy 0.000 description 5
- 239000011435 rock Substances 0.000 description 5
- 238000012937 correction Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 4
- 230000015572 biosynthetic process Effects 0.000 description 3
- 125000004122 cyclic group Chemical group 0.000 description 2
- 230000008034 disappearance Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004880 explosion Methods 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 239000003208 petroleum Substances 0.000 description 2
- 235000019738 Limestone Nutrition 0.000 description 1
- OEXHQOGQTVQTAT-SSZRJXQFSA-N [(1r,5s)-8-methyl-8-propan-2-yl-8-azoniabicyclo[3.2.1]octan-3-yl] (2r)-3-hydroxy-2-phenylpropanoate Chemical compound C1([C@H](CO)C(=O)OC2C[C@H]3CC[C@@H](C2)[N+]3(C)C(C)C)=CC=CC=C1 OEXHQOGQTVQTAT-SSZRJXQFSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000010459 dolomite Substances 0.000 description 1
- 229910000514 dolomite Inorganic materials 0.000 description 1
- 239000010440 gypsum Substances 0.000 description 1
- 229910052602 gypsum Inorganic materials 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 239000006028 limestone Substances 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
Images
Landscapes
- Earth Drilling (AREA)
Abstract
The invention relates to a drilling machinery drilling speed prediction method based on an LSTM recurrent neural network model, which comprises the following steps: selecting a specific block, and collecting data of different wells in the block as well drilling data; dividing the drilling data into input parameters and output parameters to obtain an initial data set D; processing the initial data set, including quantitative characterization of unstructured data and normalization of structured data; changing the initial data set D into a three-dimensional array required by a recurrent neural network, wherein the three-dimensional array comprises sample numbers, recurrent core expansion steps and input characteristic numbers, and obtaining an input data set D1; using LSTM algorithmsEstablishing a mechanical drilling speed prediction model, and training the model until the model meets the requirements; inputting the input data set D1 into the trained model to obtain the predicted penetration rate, using R2And (5) evaluating the fitting result of the predicted value and the true value of the mechanical drilling speed by indexes. According to the method, the mechanical drilling speed can be accurately predicted by constructing a nonlinear complex relation model, and the defects and shortcomings of the prior art are overcome.
Description
Technical Field
The invention relates to a method for predicting a mechanical drilling speed based on a recurrent neural network in the field of petroleum drilling, in particular to a method for predicting the mechanical drilling speed based on a recurrent neural network model of long-short term memory (LSTM).
Background
Rate of Penetration (ROP) is an indication of the footage of the rock drilled per unit of pure drilling time, and is an indicator of the method of breaking the rock used, the nature of the rock drilled, the drilling process used and the state of the art, and is a direct measure of the time required to drill the well. The mechanical drilling speed is one of important indexes of drilling engineering, factors influencing the mechanical drilling speed are various, the relation is complex, and a convincing and generally applicable mathematical model is not established up to now. The accurate prediction of the mechanical drilling speed can shorten the drilling period and reduce the drilling cost, and has important significance for the optimization of drilling engineering.
To calculate the rate of penetration, Bourgoyne and Young (Bourgoyne A T, Young F S. A Multiple Regression Approach to Optimal Drilling and Abnormal Pressure Detection [ J ]. Society of Petroleum Engineers Journal,1974,14(4): 371) have developed a rate of penetration model that takes into account many factors such as rock strength, weight on bit, Pressure differential, rotational speed, but in 2016 Soares et al (Soares C, Daige H, Gray K. evaluation of PDC bit ROP models and the effect of rock string on coefficients [ J. Journal of Natural Gas Science and Engineering,2016,34: 1225) has revealed limitations of conventional rate of penetration models, including the rate of penetration model and the mechanical Drilling model. Because the factors influencing the mechanical drilling speed are many, the accurate mechanical drilling speed model is difficult to establish,
as neural networks that can address the nonlinear relationships of large amounts of data enter the field of researchers, Jordan and Elman propose a framework of Recurrent neural networks, called Simple Recurrent Networks (SRNs), in 1986 and 1990, respectively, which is considered to be a fundamental version of RNNs that are currently widely prevalent. The problem of gradient explosion and gradient disappearance of the traditional recurrent Neural network is improved by a long-short term memory (LSTM) -based recurrent Neural network (Hochreiter S and Schmidhuber J. Long short-term memory [ J ]. Neural computation,1997,9(8):1735-80), and the recurrent Neural network is the most effective sequence model in practical application nowadays. The invention adopts an LSTM recurrent neural network model to predict the drilling rate, and because the drilling rate is closely related to the well depth, the drilling input parameters are taken as sequence data according to the well depth of the drilling, thereby constructing a nonlinear complex relation sequence model related to the drilling rate and solving the complex prediction problem under the influence of various parameters.
Disclosure of Invention
The invention aims to provide a drilling rate prediction method of a drilling machine based on an LSTM recurrent neural network model, which has reliable principle, can accurately predict the drilling rate by constructing a nonlinear complex relation model, and overcomes the defects and shortcomings of the prior art.
In order to achieve the technical purpose, the invention adopts the following technical scheme.
The cyclic neural network is very effective for nonlinear data with sequence characteristics, and the most important characteristic is that the output of a neuron at a certain moment can be used as input to be input into the neuron again. The invention uses the model based on the LSTM recurrent neural network, the model introduces a memory unit and a gate control memory unit to store historical information and long-term state, and uses gate control to control the flow of information, thereby solving the problems of gradient disappearance and gradient explosion in the general recurrent neural network. Since the rate of penetration is highly correlated with the well depth, the rate of penetration is predicted from a large amount of drilling data by using the drilling data as sequence data according to the well depth.
The drilling machine drilling speed prediction method based on the LSTM recurrent neural network model sequentially comprises the following steps:
step 1: selecting a specific block, collecting data of different wells in the block, respectively arranging the data according to different well numbers to obtain a data sample set, wherein the data sample set comprises logging data and logging data, and uniformly arranging the data sample set into an Excel table or a TXT text to serve as drilling data.
Furthermore, in the step 1, the data acquisition needs to extract the gamma coefficient, the acoustic time difference and the stratum lithology corresponding to the well depth from the logging data; and extracting the bit pressure, the drilling speed, the drilling fluid density, the riser pressure, the discharge capacity, the torque, the bit size and the bit type corresponding to the well depth from the logging data and the well drilling log.
Step 2: dividing the drilling data obtained in the step 1 into input parameters and output parameters, wherein the input parameters comprise well depth, original formation parameters, drilling parameters and drill bit parameters, the original formation parameters comprise lithology, gamma coefficient GR and acoustic time difference AC, the drilling parameters comprise bit pressure, drilling speed, drilling fluid density, riser pressure, displacement and torque, and the drill bit parameters comprise drill bit type and drill bit size; the output parameter is the mechanical drilling speed; and taking 12 input parameters with different characteristic attributes as input variables X and taking the mechanical drilling rate as an output variable Y to obtain an initial data set D.
And step 3: the initial data set is processed, including quantitative characterization of unstructured data and normalization of structured data. The well data obtained by step 1 is not all structured data, e.g., lithology and bit type in the data are unstructured data. The quantization representation of the unstructured data means that a sequential coding (Ordinalencoder) mode is adopted for coding, and the unstructured data are converted into numerical data; the standardization of the structured data refers to that the data fall into a specific interval and are converted into a dimensionless pure numerical type, so that indexes of different units or two levels can be weighted and compared, and the influence of data with overlarge numerical values on model weight is reduced.
Further, the step 3 of using sequential coding means to convert the features of each category into a new integer, i.e. not 0 or 1. Suppose a single unstructured factor is X ═ A1，A2，B1，B2，B3，C1，C2And after sequential coding, the result is X ═ 01，02，11，12，13，21，22Different characteristic attributes of a single factor correspond to different codes.
Further, the normalization of the data in step 3 adopts a data normalization method, and maps the result value between [0, 1] by linear transformation of the original data, and the formula is as follows:
wherein xiRepresenting any point under a certain parameter;
xminrepresents the minimum value under a certain parameter;
xmaxrepresents the maximum value under a certain parameter;
Xirepresenting the normalized value under a certain parameter.
And 4, step 4: changing an initial data set D into a three-dimensional array required by a recurrent neural network, wherein the three-dimensional array comprises the number of samples, the number of recurrent core expansion steps and the number of input characteristics, taking the drilling data per meter as a matrix, the number of samples is the number of matrices formed by the drilling data per meter, predicting the mechanical drilling speed per meter by the drilling data per meter, the number of recurrent core expansion steps is 1, the number of input characteristics is the number of types of input parameters, and the number is 12 in the invention, so that an input data set D1 is obtained.
And 5: and establishing and training a model. An input data set D1 is randomly divided into a training set D2 and a testing set D3, wherein the training set accounts for 80% of the input data set D1, the testing set accounts for 20% of the input data set D1, a mechanical drilling speed prediction model (Yien, Gudeflu, Zhaoshengsu, Liyuanjun, Kuntan, Likayao, deep learning [ M ]. Beijing: the people's post and telecommunications press, 2017) is established by using an LSTM algorithm, the model is trained by the randomly divided training set D2, and the trained model is tested by the testing set D3 until the requirement is met.
Further, in step 5, the drilling rate prediction model is composed of an input layer, a hidden layer and an output layer, various input parameters affecting the drilling rate are used as the input layer, the drilling rate is used as the output layer, a plurality of hidden layers are inserted between the input layer and the output layer, the hidden layers are also called as a cycle core, the number of the cycle memory bodies is set in the cycle core, the cycle memory bodies can store state information obtained after each input, and the state information is input into the next cycle core, so that the output of the model is not only related to the corresponding input, but also related to the state obtained by the previous input.
Further, in step 5, when the model is trained, tanh and ReLU functions are used as the activation functions of grid transfer, an Adam optimizer is used to optimize the weights of the activation functions, and the number of repeated training is the number of iterations. The function used for observing the error between the drilling rate of the machine output by the model and the real drilling rate is called a loss function, the loss function MAE and the MSE are used for recording the error generated by the training set and the testing set in each iteration, and the smaller the loss function of the training set and the testing set is, the better the model is considered.
Further, the Activation Function (Activation Function) refers to a Function running on a neuron of the deep neural network, and is responsible for mapping an input of the neuron to an output. The present invention uses the tanh and ReLU activation functions so that the input and output layers are no longer simply linear relationships.
Furthermore, the Adam optimizer is an optimization algorithm based on random gradient descent (SGD), combines SGDM first-order momentum and RMSProp second-order momentum, adds two correction terms on the basis, and is a comprehensive consideration for first-order moment estimation and second-order moment estimation of the gradient. The learning rate eta of the invention is 0.001, and the attenuation coefficient beta of the first-order matrix10.9, second order matrix attenuation coefficient beta20.999, constant e 10 for value stability-8The initial time series t is 0.
Calculating the gradient gtThe formula of (1) is as follows:
t＝t+1
wherein θ represents a parameter that needs to be updated;
gta gradient representing a decrease of the parameter θ at the time series t;
Updating the first and second moment estimation formulas as follows:
mt＝β1mt-1+(1-β1)gt
wherein m istRepresents the parameter gtThe first moment of (d);
vtrepresents the parameter gtSecond order moment of (a).
The modified first and second moment estimation formulas are as follows:
wherein
A bias correction representing a parameter m;
The parameter θ can be updated by the coefficients obtained by the two bias corrections, that is, an iterative process of the parameter θ, and the formula is as follows:
furthermore, the loss function means that the MSE and MAE methods are used to record the error generated by the training set and the testing set in each iteration. MSE is the mean square error and MAE is the mean absolute error, and the formula is as follows:
wherein n represents the number of data;
yirepresenting the true rate of penetration;
Step 6: inputting the input data set D1 established according to the step 4 into a trained model to obtain the predicted drilling rate, and adopting R2And (5) evaluating the fitting result of the predicted value and the true value of the mechanical drilling speed by indexes.
Further, in step 6, R2Instead of the square of R, an evaluation method, also called the decision coefficient R, is commonly used for regression problems2(R-Square)。R2Is an index reflecting the quality of the fitting evaluation and is most commonly used for evaluating the quality degree, R, of the regression model2The larger (close to 1) the better the regression model is fitted, and the smaller (close to 0) the worse the fitting result. The formula is as follows:
wherein
Mean rate of penetration is indicated.
Compared with the prior art, the invention can obtain the following beneficial effects:
(1) according to the method, a prediction model for the mechanical drilling speed is constructed by utilizing the strong capability of deep learning to carry out regression between nonlinear variables and the learning capability of LSTM to sequence data;
(2) the method adopts the existing deep learning frame to construct and predict the model, namely the deep learning frame tensorflow2.0 in python is used to construct the model, so that the cost required for establishing the model is low, and the method is easy to implement;
(3) the invention takes the well drilling data as the sequence data according to the well depth, and inputs the well drilling parameters of the current meter number and the state obtained by inputting the well drilling parameters before, so that the predicted mechanical drilling speed takes the state of the well drilling before into consideration, and the invention is more in line with the actual situation of the well drilling in engineering.
In conclusion, the invention takes the well drilling data as a logic sequence according to the well depth, inputs the well drilling data into the LSTM recurrent neural network, predicts the mechanical drilling rate, and the model can automatically convert the unstructured data in the well drilling parameters into structured data and add various variables influencing the mechanical drilling rate to achieve higher prediction accuracy.
Drawings
FIG. 1 is a flow chart of a recurrent neural network model training based on LSTM.
FIG. 2 is a diagram of MSE mean square error variation over a training set and a testing set based on iteration number.
FIG. 3 is a graph of the change in mean absolute error of the MAE for the training set versus the test set based on the number of iterations.
FIG. 4 is a graph of the true value and predicted value of the rate of penetration versus the depth of the well.
Detailed Description
The invention is further illustrated below with reference to the figures and examples in order to facilitate the understanding of the invention by a person skilled in the art. It is to be understood that the invention is not limited in scope to the specific embodiments, but is intended to cover various modifications within the spirit and scope of the invention as defined and defined by the appended claims, as would be apparent to one of ordinary skill in the art.
The method tests the data of a certain block in Sichuan, the well history data of the block is relatively complete and comprises the logging data, the well drilling history and the well drilling design data of a plurality of wells, and the flow is shown in figure 1.
Firstly, determining the mechanical drilling rate of each well as a prediction object. Parameters that affect the rate of penetration of the machine, formation parameters (lithology, gamma factor, sonic moveout), drilling parameters (weight on bit, rate of penetration, drilling fluid density, riser pressure, displacement, torque), and bit type and bit size are then determined.
And step two, putting the well drilling data into a data set, collecting the drill bit model and the stratum lithology corresponding to each well, and classifying according to the ascending order of the well depth.
And step three, the lithology and drill bit factors in the stratum factors are character type parameters, belong to non-structural data, and need to be subjected to characteristic engineering to convert the classification characteristics into classification numerical type characteristics. The lithology of the stratum comprises 5 types which are sandstone, mudstone, dolomite, limestone and gypsum respectively, and the lithology is sequentially 0.0, 1.0, 2.0, 3.0 and 4.0 through coding; the bit is divided into 3 kinds, which are PDC bit, roller bit, PDC bit + screw, roller bit + screw and composite bit + screw, and the codes are 0.0, 1.0, 2.0, 3.0 and 4.0 in sequence. Sequential coding may also be used when more complex lithology changes are encountered and when more drilling tool combinations are used. As shown in the following table:
after the coding is finished, normalization processing is carried out on all data by adopting a normalization method to enable the data to be non-dimensionalized, so that the weighting operation of a following algorithm is facilitated:
and step four, changing the initial data set into a three-dimensional array according to the requirement of the recurrent neural network, wherein the sample number of the three-dimensional array is 1532 groups of drilling parameters per meter, the recurrent kernel expansion step number is 1, and the input characteristic number is the type of the input parameters and is 12 in the invention.
And step five, determining the model as an LSTM recurrent neural network model, sending the model into a three-dimensional array according to the requirement of the recurrent neural network when parameters are input, dividing the array into a training set and a test set in a ratio of 8: 2, and designing model parameters of the network, wherein the model is a 4-layer recurrent neural network. Two hidden layers (the number of the loop memories of the first layer is 60, the output state of each time step is sent to the next layer, and the number of the loop memories of the second layer is 80), and one output layer, the number of iterations is 1000. Deep neural network training is performed on the data set based on the CPU. The number of layers of the network and the number of neurons need to be changed based on different data volumes, and in this embodiment, specific parameters are shown in the following table:
| model parameters | Value taking |
| Number of |
1000 |
| Total number of samples | 1532 |
| Number of steps of cyclic nuclear expansion | 1 |
| Number of features | 12 |
| Number of first layer of circular memories | 60 |
| Number of second layer of |
80 |
| Output layer | 1 |
| Activating a function | Tanh，ReLU |
| Optimizer | Adam |
| Loss function | MSE，MAE |
The training model outputs the error between the training set and the test set (see fig. 2 and 3) trained in each step, wherein fig. 2 represents the Mean Square Error (MSE) of each iteration of the training set and the test set, and the smaller the numerical value is, the smaller the deviation of the result is, the better the result is; fig. 3 represents the Mean Absolute Error (MAE) of each iteration of the training set and the test set, and smaller values indicate smaller deviations of the results and better results. The first thousand error results are shown in the following table:
| 1000 times of training | MAE | MSE |
| Training set | 0.0258 | 0.0028 |
| Test set | 0.0263 | 0.0029 |
And step six, predicting the mechanical drilling speed, and inputting data into the stored model. By the use of R2The evaluation method verifies the accuracy, namely, the decision coefficient, reflects the proportion that all the variation of the dependent variable can be explained by the independent variable through the regression relationship, and the value range is [0, 1]]: if the result is 0, the model fitting effect is poor; if the result is 1, the model is error-free, model R2The evaluation score was 0.918 and the prediction was very good (see fig. 4), and fig. 4 represents the results of the real rate of penetration and the predicted rate of penetration, which are almost completely coincident, indicating that the overall accuracy is very high.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.
Claims (7)
1. The drilling machine drilling speed prediction method based on the LSTM recurrent neural network model sequentially comprises the following steps:
step 1: selecting a specific block, collecting data of different wells in the block, and respectively sorting the data according to different well numbers to obtain a data sample set as well drilling data;
step 2: dividing the drilling data obtained in the step 1 into input parameters and output parameters, wherein the input parameters comprise well depth, original stratum parameters, drilling parameters and drill bit parameters, the original stratum parameters comprise lithology, gamma coefficient and acoustic time difference, the drilling parameters comprise bit pressure, drilling speed, drilling fluid density, riser pressure, displacement and torque, and the drill bit parameters comprise drill bit type and drill bit size; the output parameter is the mechanical drilling speed; taking 12 input parameters as input variables X and taking the mechanical drilling speed as an output variable Y to obtain an initial data set D;
and step 3: processing the initial data set, including quantitative characterization of unstructured data and normalization of structured data; the quantization representation of the unstructured data refers to coding in a sequential coding mode, and converting the unstructured data into numerical data; the standardization of the structured data means that the data fall into a specific interval and are converted into a dimensionless pure numerical type;
and 4, step 4: changing the initial data set D into a three-dimensional array required by a recurrent neural network, wherein the three-dimensional array comprises sample numbers, recurrent kernel expansion steps and input characteristic numbers, taking the drilling data per meter as a matrix, the sample numbers are matrix numbers formed by the drilling data per meter, the recurrent kernel expansion steps are 1, the input characteristic numbers are 12, and obtaining an input data set D1;
and 5: randomly dividing an input data set D1 into a training set D2 and a testing set D3, wherein the training set accounts for 80% of the input data set D1, the testing set accounts for 20% of the input data set D1, establishing a mechanical drilling speed prediction model by using an LSTM algorithm, training the model by using the randomly divided training set D2, and testing the trained model by using a testing set D3 until the requirements are met;
step 6: inputting the input data set D1 into the trained model to obtain the predicted penetration rate, using R2And (5) evaluating the fitting result of the predicted value and the true value of the mechanical drilling speed by indexes.
2. The LSTM recurrent neural network model-based drilling rig penetration rate prediction method of claim 1, wherein the sequential coding used in step 3 is to convert each class of features into a new integer, i.e., not 0 or 1, assuming that the single unstructured factor is X ═ { a ═ b { (a) }1，A2，B1，B2，B3，C1，C2And after sequential coding, the result is X ═ 01，02，11，12，13，21，22Different characteristic attributes of a single factor correspond to different codes.
3. The LSTM recurrent neural network model-based drilling rig penetration rate prediction method of claim 1, wherein the normalization of the data in step 3 is performed by data normalization, and the result is mapped between [0, 1] by linear transformation of the raw data:
wherein xiRepresenting any point under a certain parameter;
xminrepresents the minimum value under a certain parameter;
xmaxrepresents the maximum value under a certain parameter;
Xirepresenting the normalized value under a certain parameter.
4. The LSTM recurrent neural network model-based drilling rate prediction method of claim 1, wherein in step 5, the drilling rate prediction model is composed of an input layer, a hidden layer, and an output layer, various input parameters affecting the drilling rate are used as the input layer, the drilling rate is used as the output layer, a plurality of hidden layers are inserted between the input layer and the output layer, the hidden layers are also called recurrent cores, the number of recurrent memories is set in the recurrent cores, the recurrent memories store state information obtained after each input, and input the state information into the next recurrent core, so that the output of the model is not only related to the corresponding input but also related to the state obtained by the previous input.
5. The method of claim 1, wherein in step 5, in the training of the model, functions of tanh and ReLU are used as activation functions of grid transfer, Adam optimizer is used to optimize weights of the activation functions, the number of repeated training is iteration number, and loss functions MAE and MSE are used to record errors generated by the training set and the testing set in each iteration.
6. The LSTM recurrent neural network model-based drilling rig penetration rate prediction method of claim 5, wherein the loss function is the error that each iteration of training and testing sets produces using MSE and MAE, where MSE is the mean square error and MAE is the mean absolute error:
wherein n represents the number of data;
yirepresenting the true rate of penetration;
7. The LSTM recurrent neural network model-based drilling rig penetration rate prediction method of claim 1, wherein R in step 62Is an index reflecting the quality of the fitting evaluation, R2Close to 1, the better the fit, the closer to 0, the worse the fit:
wherein
Mean rate of penetration is indicated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210022755.6A CN114358434A (en) | 2022-01-10 | 2022-01-10 | Drilling machine drilling speed prediction method based on LSTM recurrent neural network model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210022755.6A CN114358434A (en) | 2022-01-10 | 2022-01-10 | Drilling machine drilling speed prediction method based on LSTM recurrent neural network model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114358434A true CN114358434A (en) | 2022-04-15 |
Family
ID=81109728
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210022755.6A Pending CN114358434A (en) | 2022-01-10 | 2022-01-10 | Drilling machine drilling speed prediction method based on LSTM recurrent neural network model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114358434A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114856540A (en) * | 2022-05-11 | 2022-08-05 | 西南石油大学 | Horizontal well mechanical drilling speed while drilling prediction method based on online learning |
| CN115059448A (en) * | 2022-06-01 | 2022-09-16 | 中国石油大学(华东) | Stratum pressure monitoring method based on deep learning algorithm |
| CN115929193A (en) * | 2022-10-21 | 2023-04-07 | 徐州徐工基础工程机械有限公司 | Composite power supply electric rotary drilling rig and power head multi-mode control method thereof |
| CN117349798A (en) * | 2023-12-05 | 2024-01-05 | 西南石油大学 | Unbalanced regression-based mechanical drilling rate prediction method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112901137A (en) * | 2021-03-08 | 2021-06-04 | 西南石油大学 | Deep well drilling mechanical drilling speed prediction method based on deep neural network Sequential model |
| CN113553356A (en) * | 2021-07-23 | 2021-10-26 | 中海石油(中国)有限公司 | Drilling parameter prediction method and system |
-
2022
- 2022-01-10 CN CN202210022755.6A patent/CN114358434A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112901137A (en) * | 2021-03-08 | 2021-06-04 | 西南石油大学 | Deep well drilling mechanical drilling speed prediction method based on deep neural network Sequential model |
| CN113553356A (en) * | 2021-07-23 | 2021-10-26 | 中海石油(中国)有限公司 | Drilling parameter prediction method and system |
Non-Patent Citations (2)
| Title |
|---|
| MAURO A. ENCINAS等: "Downhole data correction for data-driven rate of penetration prediction modeling", 《JOURNAL OF PETROLEUM SCIENCE AND ENGINEERING》 * |
| 闵超 等: "机器学习在油气行业中的应用进展综述", 《西南石油大学学报（自然科学版）》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114856540A (en) * | 2022-05-11 | 2022-08-05 | 西南石油大学 | Horizontal well mechanical drilling speed while drilling prediction method based on online learning |
| CN114856540B (en) * | 2022-05-11 | 2024-05-28 | 西南石油大学 | Horizontal well mechanical drilling speed while drilling prediction method based on online learning |
| CN115059448A (en) * | 2022-06-01 | 2022-09-16 | 中国石油大学(华东) | Stratum pressure monitoring method based on deep learning algorithm |
| CN115929193A (en) * | 2022-10-21 | 2023-04-07 | 徐州徐工基础工程机械有限公司 | Composite power supply electric rotary drilling rig and power head multi-mode control method thereof |
| CN117349798A (en) * | 2023-12-05 | 2024-01-05 | 西南石油大学 | Unbalanced regression-based mechanical drilling rate prediction method and system |
| CN117349798B (en) * | 2023-12-05 | 2024-02-23 | 西南石油大学 | Unbalanced regression-based mechanical drilling rate prediction method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112901137B (en) | Deep well drilling mechanical drilling speed prediction method based on deep neural network Sequential model | |
| CN114358434A (en) | Drilling machine drilling speed prediction method based on LSTM recurrent neural network model | |
| CN112989708B (en) | Well logging lithology identification method and system based on LSTM neural network | |
| Wu et al. | Machine learning-based method for automated well-log processing and interpretation | |
| CN108897975B (en) | Method for predicting gas content of coal bed gas logging based on deep belief network | |
| CN113553780B (en) | Stratum pore pressure prediction method based on machine learning | |
| CN110807557A (en) | Drilling rate prediction method based on BP neural network and drilling rate optimization method based on BP neural network and particle swarm optimization | |
| CN111797364B (en) | Landslide multilayer safety evaluation method based on composite cloud model | |
| CN112948932A (en) | Surrounding rock grade prediction method based on TSP forecast data and XGboost algorithm | |
| CN111027882A (en) | Method for evaluating brittleness index by utilizing conventional logging data based on high-order neural network | |
| CN111048163B (en) | Shale oil hydrocarbon retention amount (S1) evaluation method based on high-order neural network | |
| CN112712025A (en) | Complex lithology identification method based on long-term and short-term memory neural network | |
| CN116822580A (en) | Method, system, equipment and medium for predicting mechanical drilling speed time sequence characteristics based on transducer model | |
| CN111058840A (en) | Organic carbon content (TOC) evaluation method based on high-order neural network | |
| CN114548482A (en) | Creep type landslide kinetic energy change rate face-slip early warning method | |
| CN114991757A (en) | Formation pressure measurement and recording integrated monitoring method and system while drilling | |
| CN117251802B (en) | Heterogeneous reservoir parameter prediction method and system based on transfer learning | |
| CN110671092A (en) | Oil gas productivity detection method and system | |
| CN117035197B (en) | Intelligent lost circulation prediction method with minimized cost | |
| CN114114414A (en) | Artificial intelligence prediction method for 'dessert' information of shale reservoir | |
| CN113408190A (en) | Bayes-LSTM model-based surrounding rock deformation prediction method during construction of highway tunnel | |
| CN111751878A (en) | Method and device for predicting transverse wave velocity | |
| CN115049173B (en) | Deep learning and Eaton method coupling driving stratum pore pressure prediction method | |
| CN110032777B (en) | Drilling fluid density safety window estimation method based on uncertainty analysis | |
| CN115293462B (en) | Method for predicting size range of leakage channel based on deep learning |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220415 |
|
| RJ01 | Rejection of invention patent application after publication |