CN114357887A - A prediction method of mud loss before drilling in complex well conditions based on BP neural network - Google Patents

Abstract

The invention discloses a method for predicting the slurry leakage before drilling under complex well conditions based on a BP neural network, which comprises the following steps: s1, collecting historical seismic attribute data of the target area and engineering and geological data corresponding to the missing case; s2, carrying out standardization processing on the collected data, and carrying out conversion processing on all data for the convenience of implementing a method; s3, taking the preprocessed historical seismic data as input, the loss condition as output and the real loss state as a standard value, and carrying out supervision training and optimization to obtain a loss prediction neural network model; and S4, inputting earthquake real-time seismic data of the target area, and automatically judging the leakage condition corresponding to each depth of the area by the model. The invention solves the problems that the condition in the well is difficult to accurately predict before drilling under the existing complex well condition and mud leakage is easy to occur.

Description

A prediction method of mud loss before drilling in complex well conditions based on BP neural network

technical field

The invention relates to the technical field of drilling construction, in particular to a method for predicting mud loss before drilling in complex well conditions based on a BP neural network.

Background technique

During the drilling and production process, lost circulation often occurs when drilling into fracture-developed zones and/or karst-cavity-developed zones. Drilling in formations prone to lost circulation is a relatively complex drilling work, and drilling in the formations above can easily induce major downhole accidents such as blowouts and pipe sticking. More importantly, drilling engineering accidents directly affect the discovery and protection of oil and gas formations, resulting in a decline in the effectiveness of exploration and development. For example, the commonly used technical means in the north slope area of Tazhong in the Tarim Basin are mainly to test and analyze the lost circulation through well logging, identify the location of the leakage layer, and determine the channel properties of the leakage layer. Knowing that the drilling formation is the drilling lost interval. This method cannot predict the lost circulation in the formation before drilling, so it cannot provide guidance for drilling and avoid drilling leakage intervals. Therefore, the drilling success rate of this method is low.

Chinese Patent Application No. CN 110941010 A, the application publication date is March 31, 2020, the invention-creation title is a method for predicting drilling loss by utilizing seismic data, comprising the following steps: The present invention provides a method for predicting drilling loss by utilizing seismic data The method includes the following steps: S1: perform noise reduction filtering processing on the original seismic data volume to obtain a filtered data volume; S2: perform coherent processing on the filtered data volume to obtain a coherent volume; S3: combine the coherent volume with the The original seismic data volume is fused to obtain a fusion data volume; S4: Determine the distribution area of the coherent volume according to the section of the fusion data volume, and predict the position of the fracture development zone and/or the karst cave development zone according to the distribution area of the coherent volume; S5 : Predict the drilling lost interval according to the position of the fracture development zone and/or the cave development zone. This application proposes a technical solution for predicting drilling loss by using seismic data, but its shortcomings are: first, the processing and fusion of seismic data volumes in the method described in this application has great inaccuracy, and will It affects the final prediction of the location results of the fracture development zone and/or the cave development zone; secondly, this method can only determine the location of the fracture development zone and/or the cave development zone preliminarily, and cannot rely solely on the fracture development zone and/or the cave development zone. The location determines the missing location, and there are many factors that affect the missing, so it has great limitations and uncertainties. Therefore, there is a need for a method for predicting mud loss before drilling for complex well conditions to improve safety.

SUMMARY OF THE INVENTION

To this end, the present invention provides a method for predicting mud loss before drilling in complex well conditions based on BP neural network, so as to solve the problems of difficulty in accurately predicting well conditions and easy mud loss before drilling in complex well conditions.

In order to achieve the above object, the present invention provides the following technical solutions:

The invention discloses a method for predicting mud loss before drilling in complex well conditions based on BP neural network. The method is as follows:

S1. Collect historical seismic attribute data of the target area, engineering and geological data corresponding to missing cases;

S2. Standardize the collected data, and in order to facilitate the implementation of the method, convert all data;

S3. Take the preprocessed historical seismic data as the input, take the leakage situation as the output, and take the real leakage state as the standard value, supervise training and optimize to obtain the leakage prediction neural network model;

S4. Input the real-time seismic data of the target area, and the model automatically judges the leakage situation corresponding to each depth in this area.

Further, in the step S1, through the historical drilling leakage geological parameters and engineering parameters of the target block, the geological parameters mainly include the geological horizon at the time of the leakage and the lithology of the leakage section, and the engineering parameters include the mud leakage rate, the cumulative mud leakage. Quantity, mud density, and construction conditions when mud is lost are collected, sorted and verified, and a variable set of five input parameters is determined: original amplitude, AFE, RMS amplitude, instantaneous frequency, response phase and other seismic attribute data. The collected real value of the missing situation in this block is set as the value of 1 for missing, and the value of 0 for no missing, as the standard value for model training.

Further, in the step S2, the standardization processing method is to use the histogram method; determine the relative layer; draw the histogram; obtain the standard value of each data in the study area; obtain the correction value of the single well curve.

Further, in the determination of the relative layer, the top surface of the Silurian in the study area will be used as the relative standard layer, and the seismic identification features are obvious; the histogram is drawn, the standard layer data of each single well is extracted, and the Gaussian normal distribution function is used to analyze the data. The seismic response frequency distribution histogram is fitted to obtain peak readings for each seismic attribute.

Further, the standard value of each data in the study area is obtained, and the seismic attribute response value of each single well is used to draw the frequency distribution histogram of the original amplitude, AFE, RMS amplitude, instantaneous frequency, and response phase of the whole area, so as to obtain the frequency distribution histogram of the whole area. The peak value of the histogram, that is, the standard value of each attribute.

Further, in the acquisition of the correction value of the single well curve, the correction value = the standard value of the whole area - the peak value of the single well.

Further, in the step S2, the data is converted to normalize the data, and the data normalization is to limit the preprocessed data to the range of [-1, 1], so that all the features of the data are mapped to the same scale.

Further, in the step S3, the input parameter InputData of the training set is set, and the collected 5 seismic attribute parameters are used as the data of neural network training; the output standard value OutputTarget of the training set is set, and the standard value is the actual measured depth corresponding to this block. establish the BP neural network, set the number of nodes in the input layer, hidden layer and output layer, set the transfer function, and use the batch gradient descent algorithm to perform unconstrained nonlinear optimization of the model.

Further, the method for obtaining the dropout prediction neural network model by the supervision training and optimization is as follows: initialize the original variable parameters, set the initial weight, threshold and learning rate, and then input a given sample to calculate the input value of each layer through the sigmoid function function. and output value, the neuron state of each layer only affects the neuron state of the next layer, and gradually propagates to the hidden layer. The input of the hidden layer is equal to the weighted sum of the signals of the input layer, and the output layer is equal to the value of the previous hidden layer. The output value mapped by the excitation function is output, and the output signal is generated at the output end. If the ideal output value cannot be obtained at the output layer, the system transfers to the error signal back propagation process, and the error signal propagates from the output layer to the input layer along the way. The connection weights between layers and the bias values of neurons in each layer are adjusted, and the gradient descent strategy of the error function is implemented to continuously reduce the error signal, and the weights are continuously adjusted to make the network error function reach the minimum value.

Further, in the step S4, based on the difference between the predicted value of the missing situation and the measured value of the missing situation, the back-propagation algorithm will update the weights of the neural network parameters accordingly, and modify the neural network according to the weights. The predicted value is continuously modified to approach the target value until a certain accuracy rate or training times is reached.

The present invention has the following advantages:

The invention discloses a method for predicting mud leakage before drilling in complex well conditions based on BP neural network. The leakage situation obtained through the prediction is used as the judgment basis for well position deployment and development, and provides drilling technicians and construction personnel with more accurate and effective methods. Therefore, the probability of avoiding complex well conditions is increased, the implementation efficiency of drilling operations is increased, repeated operations are avoided, and engineering time is saved. It can quickly and accurately predict the leakage situation before drilling, so as to provide decision support for drilling and improve the probability of avoiding leakage and reducing drilling time.

Description of drawings

In order to illustrate the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the accompanying drawings that are required to be used in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only exemplary, and for those of ordinary skill in the art, other implementation drawings can also be derived from the provided drawings without any creative effort.

The structures, proportions, sizes, etc. shown in this specification are only used to cooperate with the contents disclosed in the specification, so as to be understood and read by those who are familiar with the technology, and are not used to limit the conditions for the implementation of the present invention, so there is no technical The substantive meaning above, any modification of the structure, the change of the proportional relationship or the adjustment of the size should still fall within the technical content disclosed in the present invention without affecting the effect and the purpose that the present invention can produce. within the range that can be covered.

1 is a flowchart of a method for predicting mud loss before drilling in complex well conditions based on a BP neural network according to an embodiment of the present invention;

2 is a schematic diagram of a BP network operation structure provided by an embodiment of the present invention;

3 is a flowchart of a BP network algorithm provided by an embodiment of the present invention;

Fig. 4 is a model training optimization curve diagram provided by an embodiment of the present invention;

FIG. 5 is a comparison diagram of a validation set dropout prediction result provided by an embodiment of the present invention.

Detailed ways

The embodiments of the present invention are described below by specific specific embodiments. Those who are familiar with the technology can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the present invention. , not all examples. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Example

Referring to FIG. 1 , the present embodiment discloses a method for predicting mud loss before drilling in complex well conditions based on BP neural network. The method is as follows:

S1. Collect historical seismic attribute data of the target area, engineering and geological data corresponding to missing cases;

S2. Standardize the collected data, and in order to facilitate the implementation of the method, convert all data;

S3. Take the preprocessed historical seismic data as the input, take the leakage situation as the output, and take the real leakage state as the standard value, supervise training and optimize to obtain the leakage prediction neural network model;

S4. Input the real-time seismic data of the target area, and the model automatically judges the leakage situation corresponding to each depth in this area.

Referring to Fig. 2, it is a schematic diagram of the operation structure of the BP network in the technical solution of the present invention. It can be seen from Fig. 2 that a typical BP neural network mainly refers to a forward network with three or more layers of network structures. The first layer and the tail layer of the structure are generally It is called the input layer and the output layer, and each layer in the middle is called the hidden layer or the middle layer. There is no feedback between the networks, and the network layers are not connected to each other. It mainly depends on the complexity of the system to dynamically adjust the connection relationship and connection strength of the internal nodes of the network, so as to better reflect the structure of the system.

Fig. 3 is the flow chart of the BP network algorithm in the technical solution of the present invention. The learning process is generally as follows: first, standardize, convert and initialize the original variable parameters in turn, set the initial weight, threshold and learning rate, etc., and then input a given sample to pass The sigmoid function calculates the input value and output value of each layer. The neuron state of each layer only affects the neuron state of the next layer, and gradually propagates to the hidden layer. The input of the hidden layer is equal to the weighted sum of the signals of the input layer, and the output The layer is equal to the output value after the output of the previous hidden layer is mapped by the excitation function, and an output signal is generated at the output end. If the ideal output value cannot be obtained at the output layer, the system transfers to the process of back propagation of the error signal, and the error signal propagates from the output layer to the input layer and adjusts the connection weights between layers and the bias values of neurons in each layer along the way. Execute the error function gradient descent strategy to continuously reduce the error signal, and adjust the weights to make the network error function reach the minimum value; the learning rate of the neural network is set to 0.01, the minimum target error is 0.001, the maximum number of iterations is 1000, and other parameters remain default , the number of neurons in the hidden layer is set to 12, the input layer is 5, and the output layer is 1. The following formula is used to calculate the number of hidden layer nodes k,

In the formula, m is the number of hidden layer nodes; n is the number of input layer nodes; l is the number of output layer nodes; Finally, the number of hidden layer nodes to get the best accuracy is 12.

In step S1, the geological parameters and engineering parameters of historical drilling leakage in the target block are determined. The geological parameters mainly include the geological horizon at the time of leakage and the lithology of the leakage section. The collection, sorting and verification of construction conditions when mud is lost has determined a variable set of 5 input parameters: original amplitude, AFE, RMS amplitude, instantaneous frequency, response phase and other seismic attribute data, which will be collected in this area. The true value of the block missing condition, set the missing value as 1 and the non-missing value as 0, as the standard value for model training.

Because the seismic data of the block where each well is located is different in time, depth, downhole environment, instrument series, and the operation mode of the standard calibration in the later stage is not necessarily the same, which may lead to errors between the seismic attributes of different blocks. Therefore, when applying these seismic attribute data, they must be standardized, and in order to facilitate the implementation of the method, all data will be transformed.

The standardization method is to use the histogram method; to determine the relative layers; to draw the histogram; to obtain the standard value of each data in the study area; to obtain the correction value of the single well curve.

In the relative layers, the top surface of the Silurian in the study area will be used as the relative standard layer, and the seismic identification features are obvious; the histogram is drawn, the standard layer data of each single well is extracted, and the seismic response frequency distribution histogram is calculated by the Gaussian normal distribution function. Figures are fitted to obtain peak readings for each seismic attribute. Obtain the standard value of each data in the study area, and take the seismic attribute response value of each single well to draw the frequency distribution histogram of the original amplitude, AFE, RMS amplitude, instantaneous frequency, and response phase of the whole area, and obtain the peak value of the histogram in the whole area, namely The standard value of each attribute is obtained in the calibration value of the single well curve, the calibration value = the standard value of the whole area - the peak value of the single well.

Transformation processing is data normalization. Data normalization is to limit the preprocessed data to the range of [-1, 1], so that all the features of the data are mapped to the same scale, which can avoid due to the dimension The difference makes some characteristics of the data form a dominant role.

In step S3, the input parameter InputData of the training set is set, and the collected 5 seismic attribute parameters are used as the data for neural network training; the output standard value OutputTarget of the training set is set, and the standard value is the actual measurement of the missing situation of the corresponding depth of the block; Build a BP neural network, set the number of nodes in the input layer, hidden layer and output layer, set the transfer function, and use the batch gradient descent algorithm to perform unconstrained nonlinear optimization of the model.

The specific process of supervising, training and optimizing the neural network model for pre-drilling mud loss prediction is as follows. First, initialize the original variable parameters, set initial weights, thresholds, and learning rates, as shown in Figure 3, and then input a given sample to calculate through the sigmoid action function. The input value and output value of each layer, the neuron state of each layer only affects the neuron state of the next layer, and gradually propagates to the hidden layer. The input of the hidden layer is equal to the weighted sum of the signals of the input layer, and the output layer is equal to the previous layer. The output of the hidden layer of the layer is mapped through the output value of the excitation function, and an output signal is generated at the output end. If the ideal output value cannot be obtained in the output layer, the system transfers to the process of error signal back propagation, the error signal propagates from the output layer to the input layer and adjusts the connection weights between layers and the bias values of neurons in each layer along the way. The error function gradient descent strategy is implemented to continuously reduce the error signal, and the weights are continuously adjusted to make the network error function reach the minimum value.

In order to ensure the automatic initialization of weights and thresholds when the model is running, the newff function is used as the generation function of the network model - the newff function, which means that the first step in training the feedforward network is to establish a network object, which is essentially the parameter of the newff function. . This command creates the network object and initializes the network weights and biases, so the network can be trained.

The batch gradient descent method of momentum is triggered by the training function trainingdm;

The hyperbolic tangent function tansig is used as the transfer function between the input layer and the hidden layer; logsigz is used as the transfer function between the hidden layer and the output layer;

The trainingdx function is selected as the training function of the "error inverse propagation" process.

In step S4, the method of supervised training using data such as the predicted value of the missing situation and the standard value is: based on the difference between the predicted value of the missing situation and the measured value of the missing situation, the back-propagation algorithm will update the weights of the neural network parameters accordingly. , and modify the neural network according to the weights. The predicted value of the model is constantly modified to approach the target value until it reaches a certain accuracy rate or training times.

Referring to Figure 4 is the model training optimization curve. It can be seen from the figure that the result is less than the error of 0.001 for 6 consecutive times after 9 iterations, which meets the training requirements and successfully establishes the neural network training model.

Referring to Figure 5, it is a comparison chart of the prediction results of the validation set. It can be seen from the figure that the coincidence rate of the overall prediction has reached more than 80%, which proves the effectiveness of comprehensively using this information to predict the leakage, and also shows that the BP neural network is suitable for the judgment of the leakage in this area , the main reason for the prediction deviation may be due to the geological conditions in different regions, the multi-solution of seismic attributes or the lack of representative data.

The present embodiment discloses a method for predicting mud leakage before drilling in complex well conditions based on BP neural network. The leakage situation obtained through the prediction is used as the judgment basis for well location deployment and development, and provides drilling technicians and construction personnel with more accurate, Effective decision-making basis, thereby increasing the probability of avoiding complex well conditions, increasing the implementation efficiency of drilling operations, avoiding repetitive operations, and saving engineering time. It can quickly and accurately predict the leakage situation before drilling, so as to provide decision support for drilling and improve the probability of avoiding leakage and reducing drilling time.

Although the present invention has been described in detail above with general description and specific embodiments, some modifications or improvements can be made on the basis of the present invention, which will be obvious to those skilled in the art. Therefore, these modifications or improvements made without departing from the spirit of the present invention fall within the scope of the claimed protection of the present invention.

Claims (10)

1. a method for predicting mud leakage before drilling in complex well conditions based on BP neural network, is characterized in that, described method is: S1. Collect historical seismic attribute data of the target area, engineering and geological data corresponding to missing cases; S2. Standardize the collected data, and in order to facilitate the implementation of the method, convert all data; S3. Take the preprocessed historical seismic data as the input, take the leakage situation as the output, and take the real leakage state as the standard value, supervise training and optimize to obtain the leakage prediction neural network model; S4. Input the real-time seismic data of the target area, and the model automatically judges the leakage situation corresponding to each depth in this area. 2. a kind of mud loss prediction method based on BP neural network before drilling in complex well condition as claimed in claim 1, is characterized in that, in described S1 step, by the historical drilling loss geological parameter and engineering parameter of target block, The geological parameters mainly include the geological horizon at the time of leakage and the lithology of the missing section. The engineering parameters include the rate of mud leakage, the cumulative amount of mud leakage, the density of mud, and the collection, arrangement and verification of the construction conditions when the mud is lost. Five parameters have been identified. Variable set of input parameters: seismic attribute data such as original amplitude, AFE, root mean square amplitude, instantaneous frequency, response phase, etc., set the value of the missing value as 1 for the collected real value of the missing situation in this block, and the value of 0 for non-missing , as the standard value for model training. 3. The method for predicting mud loss before drilling in complex well conditions based on BP neural network as claimed in claim 1, characterized in that, in the step S2, the method of standardization is to adopt a histogram method; determine the relative layer; Draw histogram; obtain the standard value of each data in the study area; obtain the calibration value of the single well curve. 4. a kind of mud loss prediction method based on BP neural network before drilling in complex well conditions as claimed in claim 3, is characterized in that, in the described determination relative layer, the top surface of Silurian in the study area will be used as the relative standard layer, The seismic identification features are obvious; the histogram is drawn, the standard layer data of each single well is extracted, and the seismic response frequency distribution histogram is fitted by the Gaussian normal distribution function to obtain the peak readings of each seismic attribute. 5. A method for predicting mud loss before drilling in a complex well condition based on BP neural network as claimed in claim 3, characterized in that, obtaining the standard values of each data in the study area, and drawing the seismic attribute response values of each single well The frequency distribution histogram of the original amplitude, AFE, RMS amplitude, instantaneous frequency, and response phase of the whole area, and the peak value of the histogram of the whole area is obtained, that is, the standard value of each attribute. 6 . The method for predicting mud loss before drilling in complex well conditions based on BP neural network according to claim 3 , wherein, in the obtained single-well curve correction value, the correction value=the standard value of the whole area-single well. 7 . peak. 7. The method for predicting mud loss before drilling in complex well conditions based on BP neural network as claimed in claim 1, characterized in that, in the step S2, the data is converted and processed to normalize the data, and the data is normalized. The transformation is to limit the preprocessed data to the range of [-1, 1], so that all the features of the data are mapped to the same scale. 8. a kind of mud loss prediction method based on BP neural network before drilling in complex well conditions as claimed in claim 1, is characterized in that, in described S3 step, the input parameter InputData of training set is set, and 5 earthquakes collected and arranged will be collected. The attribute parameters are used as the data for neural network training; set the output standard value OutputTarget of the training set, and the standard value is the leakage of the depth corresponding to the measured block; build a BP neural network, and set the number of nodes in the input layer, hidden layer, and output layer , set the transfer function, and use the batch gradient descent algorithm to perform unconstrained nonlinear optimization of the model. 9. The method for predicting mud loss before drilling in complex well conditions based on BP neural network as claimed in claim 1, wherein the method for obtaining the loss prediction neural network model by supervising training and optimizing is: Initialize, set the initial weight, threshold and learning rate, and then input a given sample to calculate the input and output values of each layer through the sigmoid function. Containing layer propagation, the input of the hidden layer is equal to the weighted sum of the signal of the input layer, and the output layer is equal to the output value of the output of the previous hidden layer mapped by the excitation function, and the output signal is generated at the output end. If the ideal output value is obtained, the system transfers to the error signal back propagation process, the error signal propagates from the output layer to the input layer and adjusts the connection weights between layers and the bias values of neurons in each layer along the way, and executes the error function gradient descent strategy. In order to make the error signal continue to decrease, the weights are constantly adjusted to make the network error function reach the minimum value. 10. A method for predicting mud loss before drilling in complex well conditions based on BP neural network according to claim 1, wherein in the step S4, based on the difference between the predicted value of the leakage situation and the measured value of the leakage situation , the back-propagation algorithm will update the weights of the neural network parameters accordingly, and modify the neural network according to the weights. The model prediction value is constantly modified to approach the target value until a certain accuracy rate or training times is reached.

Images