CN107133670A - A kind of Complex lithologic identification method and system based on decision tree data mining algorithm - Google Patents

Abstract

The invention relates to a complex lithology identification method and system based on a decision tree data mining algorithm. It relates to the technical field of oil exploration reservoir evaluation. The method includes: establishing a lithology profile of a sealed core section, forming a lithology identification data set according to the lithology profile; performing data preprocessing on the lithology identification data set; using different logging parameters to correspond to lithology Identify different lithologies in the data set; perform data mining on the lithology in the preprocessed lithology identification data set to form a tree identification model, so that complex Lithology is identified. The present invention forms the lithology identification data set through the lithology profile, performs unified lithology identification, and corresponds to different logging parameters according to the characteristics of lithology, and forms a clear and clear tree after data mining on the lithology identification data set The identification model can perform accurate lithology identification under complex lithology conditions.

Description

A complex lithology identification method and system based on decision tree data mining algorithm

technical field

The invention relates to the technical field of petroleum exploration reservoir evaluation, in particular to a complex lithology identification method and system based on a decision tree data mining algorithm.

Background technique

With the expansion of the breadth and depth of oil exploration and development, on the one hand, oil data is continuously accumulating at a rapid rate. If the massive oil data is used as a carrier of wealth, if it is fully utilized, it can bring additional benefits to oil exploration and development. Traditional Geophysical exploration methods based on petrophysics, mathematics, statistics, and petroleum exploration theory are somewhat powerless to deepen the utilization of these massive data; on the other hand, various unconventional oil and gas reservoirs are becoming the main exploration targets. However, conventional methods such as crossplot, linear regression, and multivariate discriminant analysis cannot effectively solve difficult problems such as lithology identification and reservoir parameter calculation. Therefore, it is necessary to introduce new methods in other research fields such as artificial intelligence, machine learning, and pattern recognition in petroleum exploration and development. As a "data-driven" solution, data mining can well solve problems encountered in complex oil and gas reservoir evaluation. of various problems.

Data mining is the process of extracting implicit, previously unknown, but potentially useful information and knowledge from a large number of incomplete, noisy, fuzzy, and random data. Data mining tasks are divided into two categories: description and prediction. The former derives a generalized pattern of potential relationships in the data, and the latter infers the current data to make predictions. The main methods are neural networks, support vector machines, Bayesian networks, and decision-making. Among them, neural networks have been widely used in lithology identification, sedimentary facies division, permeability prediction, oil, gas and water layer identification, etc. Practice has proved that when dealing with complex petroleum geological problems affected by various factors, nonlinear neural networks Network methods are superior to linear statistical analysis techniques. But one of the biggest disadvantages of neural network is that it is easy to "over-learn" the training samples. There are too many variable parameters in the method. If the algorithm is given enough time, it can "remember" almost anything, and the model established in this way will be out of control. The geological background has no practical application value; in addition, neural networks, support vector machines, and Bayesian networks have a common shortcoming that the predicted models are all "black boxes", and how the sample data and attributes are related to each other Relationship fitting is invisible.

Contents of the invention

In order to solve the above technical problems, the present invention provides a complex lithology identification method and system based on a decision tree data mining algorithm.

The technical solution of the present invention to solve the above technical problems is as follows: a complex lithology identification method based on decision tree data mining algorithm, said method comprising:

Establish the lithology profile of the sealed coring section, and form the lithology identification data set according to the lithology profile;

Performing data preprocessing on the lithology identification data set;

Different logging parameters correspond to different lithologies in the lithology identification data set;

Data mining is performed on the lithology in the preprocessed lithology identification data set to form a tree identification model, so as to identify complex lithology according to the lithology corresponding to the tree identification model and logging parameters.

The beneficial effects of the present invention are: the present invention forms the lithology identification data set through the lithology profile, performs unified lithology identification, and corresponds to different logging parameters according to the characteristics of the lithology, after data mining the lithology identification data set , forming a clear and clear tree identification model, which can carry out accurate lithology identification under complex lithology conditions.

On the basis of the above technical solutions, the present invention can also be improved as follows.

Further, after establishing the lithology section of the closed core section, the logging curve response values of different lithology sections were read according to the logging reading principles of thick layers, thin interbeds, and lithology transition zones, and the conglomerate rock section was established. According to the corresponding relationship between lithology and logging parameters, a lithology identification data set is formed.

Further, the process of data preprocessing on the lithology identification data set includes filling missing values, overall standardization and eliminating abnormal points.

The beneficial effect of adopting the above-mentioned further scheme is that the existence of thin interlayers and the influence of the logging response on the lithological transition zone will cause data anomalies. Therefore, it is necessary to eliminate the anomalies for accurate identification.

Further, the decision tree algorithm is used to carry out data mining on the lithology in the preprocessed lithology identification data set.

The beneficial effect of adopting the above further solution is that the decision tree algorithm belongs to the "white box" storage model, and it is possible to clearly understand how the classifier works and the importance of the logging parameters. For oil and gas reservoirs with severe heterogeneity and complex and changeable lithology, it is difficult for traditional mathematical statistical methods to accurately reflect the nonlinear mapping relationship between logging curves and lithology. The decision tree method of capability and nonlinear modeling can well solve this problem, provide high-precision lithology identification results for reservoir evaluation, and ensure the reasonable and efficient development of complex oil and gas reservoirs.

Further, the process of data mining the lithology in the preprocessed lithology identification data set is as follows: calculate the weight of each logging parameter in lithology identification, obtain sensitive parameters for modeling, and finally use the characterization method from top to bottom Next, a tree-like identification model is established, in which each branch of the tree represents a type of lithology identification rule, and leaf nodes represent the logging parameters constituting the identification rule and the value range of each logging parameter.

The beneficial effect of adopting the above-mentioned further scheme is: by analyzing the weights of different logging parameters, the lithology sensitive parameters are automatically selected, and finally a complex lithology identification model that takes into account the model identification accuracy and the sample generalization ability is established, providing comprehensive information for reservoirs The evaluation provides important geological basis.

In order to solve the above technical problems, the present invention also proposes a complex lithology identification system based on decision tree data mining algorithm, said system comprising:

The data set establishment module is used to establish the lithology profile of the sealed core section, and form the lithology identification data set according to the lithology profile;

A preprocessing module, configured to perform data preprocessing on the lithology identification data set;

The identification module is used to identify different lithologies in the lithology data set with different logging parameters respectively;

The data mining module is used to perform data mining on the lithology in the preprocessed lithology identification data set to form a tree identification model, so as to analyze the complex lithology according to the lithology corresponding to the tree identification model and logging parameters. to identify.

The beneficial effects of the present invention are: the present invention forms the lithology identification data set through the lithology profile, performs unified lithology identification, and corresponds to different logging parameters according to the characteristics of the lithology, after data mining the lithology identification data set , forming a clear and clear tree identification model, which can carry out accurate lithology identification under complex lithology conditions.

Further, the data set establishment module is also used to read the logging data of different lithology sections according to the logging reading principles of thick layers, thin interlayers, and lithology transition zones after establishing the lithology profile of the sealed core section. The response value of the well curve is used to establish the corresponding relationship between the lithology of the conglomerate and the logging parameters to form a lithology identification data set.

Further, the preprocessing module is specifically used to preprocess the lithology identification data set by filling missing values, overall standardization, and eliminating abnormal points.

The beneficial effect of adopting the above-mentioned further scheme is that the existence of thin interlayers and the influence of the logging response on the lithological transition zone will cause data anomalies. Therefore, it is necessary to eliminate the anomalies for accurate identification.

Further, the data mining module is specifically used to perform data mining on the lithology in the preprocessed lithology identification data set by using a decision tree algorithm.

The beneficial effect of adopting the above further solution is that the decision tree algorithm belongs to the "white box" storage model, and it is possible to clearly understand how the classifier works and the importance of the logging parameters. For oil and gas reservoirs with severe heterogeneity and complex and changeable lithology, it is difficult for traditional mathematical statistical methods to accurately reflect the nonlinear mapping relationship between logging curves and lithology. The decision tree method of capability and nonlinear modeling can well solve this problem, provide high-precision lithology identification results for reservoir evaluation, and ensure the reasonable and efficient development of complex oil and gas reservoirs.

Further, the data mining module is also used to calculate the weight of each logging parameter in lithology identification, obtain sensitive parameters for modeling, and finally establish a tree-like identification model from top to bottom in a representational manner, wherein each tree A branch represents a class of lithology identification rules, and leaf nodes represent the logging parameters constituting the identification rules and the value range of each logging parameter.

The beneficial effect of adopting the above-mentioned further scheme is: by analyzing the weights of different logging parameters, the lithology sensitive parameters are automatically selected, and finally a complex lithology identification model that takes into account the model identification accuracy and the sample generalization ability is established, providing comprehensive information for reservoirs The evaluation provides important geological basis.

Description of drawings

Fig. 1 is the flow chart of the complex lithology identification method described in the embodiment of the present invention;

FIG. 2 is a schematic diagram of a tree recognition model described in an embodiment of the present invention;

Fig. 3 is a schematic diagram of the complex lithology identification system described in the embodiment of the present invention.

detailed description

The principles and features of the present invention are described below in conjunction with the accompanying drawings, and the examples given are only used to explain the present invention, and are not intended to limit the scope of the present invention.

As shown in Figure 1, in order to determine the response relationship between different lithologies and logging curves, establish a high-precision, non-linear lithology identification model, and improve the identification accuracy of complex lithologies, this embodiment proposes a data mining method based on decision trees. An algorithmic complex lithology identification method, the method comprising:

Establish the lithology profile of the sealed coring section, and form the lithology identification data set according to the lithology profile;

Performing data preprocessing on the lithology identification data set;

Different logging parameters correspond to different lithologies in the lithology identification data set;

Data mining is performed on the lithology in the preprocessed lithology identification data set to form a tree identification model, so as to identify complex lithology according to the lithology corresponding to the tree identification model and logging parameters.

Preferably, before forming the lithology identification data set, combined with the geological background and reservoir characteristics of the research area, and based on the overall requirements of reservoir evaluation, the core sample description of the sealed coring well and the naming of the cast thin section are determined. The name of lithology, because of the strong heterogeneity of complex reservoirs and the drastic changes in lithology, the local cast thin slices of cores cannot accurately reflect the real lithology of formations; and the description of cores lacks quantitative scales for different mineral compositions. The effective combination of methods can improve the accuracy of naming complex lithology.

Taking the Kexia Formation conglomerate reservoir in the sixth middle area of Karamay Oilfield as an example, the conglomerate reservoir has the characteristics of strong heterogeneity and complex lithology due to gold source, multiple water systems and rapidly changing depositional environment. Accurate identification of nature has become a difficult point in the adjustment and development of this type of reservoir. According to the description of core samples from 8 closed coring wells in the study area and the identification results of cast thin sections, it is determined that the lithology of conglomerate reservoirs mainly includes conglomerate, glutenite, sandy conglomerate, glutenite, pebble-bearing coarse sandstone, and conglomerate. There are more than a dozen lithologies such as pebble sandstone, medium sandstone, fine sandstone, pebble-bearing mudstone, silty mudstone, and mudstone.

Specifically, based on sequence stratigraphy, the lithology profiles of 8 sealed coring well sections were established, and the response values of logging curves of different lithology sections were read. A total of 327 sections of conglomerate lithology and logging The corresponding relationship of the data forms a sample data set for conglomerate lithology identification.

Among them, the principles of logging readings are as follows: ① read the average value of thick layers as the basic data for lithology identification; The basic data for lithology identification; ③ The transition zone between different lithologies is read as the average value, or it can be used as an abnormal point in data preprocessing for backup processing.

Data preprocessing for the established lithology identification sample data set mainly includes filling missing values, overall standardization, eliminating abnormal points, etc. The most important preprocessing procedure is the deletion of abnormal points. For reservoirs with complex lithology , there are two main reasons for the abnormality of data samples. One is the existence of thin interlayers. Due to the limitation of the longitudinal resolution of logging tools, the logging response value cannot reflect the real information of formation lithology; The upper logging response is affected by the surrounding rocks on both sides, and it is difficult for the logging data to accurately calibrate the real lithology information. According to the reasons for the distortion of the log curve values, a total of 18 abnormal point data were deleted, which were mainly caused by thin interlayers and lithological transition zones.

Determine the complex lithology categories that need to be identified based on lithology naming, and according to the requirements of reservoir evaluation, some similar lithologies can be merged, and some important lithologies can also be subdivided; in addition, the mining field of the decision tree algorithm is determined, That is, which well logs are selected to indicate lithology information.

Specifically, according to the overall requirements of conglomerate reservoir evaluation, the lithology of the conglomerate in the Kexia Formation of the Liuzhong District is finally determined to be conglomerate, sandy conglomerate, glutenite, pebble-bearing coarse sandstone, fine sandstone, and conglomerate mudstone 8 types, including silty mudstone and mudstone; mining field selects undisturbed formation resistivity (Rt), natural gamma ray (GR), spontaneous potential (SP), borehole diameter (CAL), neutron porosity (CNL), acoustic wave A total of seven logging parameters, time difference (AC) and compensated density (DEN), indicate lithology.

In this embodiment, a decision tree algorithm is used to perform data mining on the lithology in the preprocessed lithology identification data set.

Specifically, the decision tree algorithm uses C5.0 for data mining of conglomerate lithology. The algorithm uses the information gain ratio (Information Gain Ratio) as the division measure, and selects the attribute with the largest information gain ratio as the split node; C5.0 adopts the pessimistic The method of error pruning (Pessimistic Error Pruning) is to use the continuous correction in the binomial distribution to correct the re-substitution error to obtain a more realistic error rate.

The selection of split attributes, that is, how to select the best split attribute from many attributes, is the core of the algorithm. The selection criteria can be roughly divided into two categories: ① The selection strategy of independent attributes; ② The selection strategy of interrelated attributes. Both methods have their advantages and disadvantages, and are optimized based on actual sample data. The pruning strategy includes two types: pre-pruning and post-pruning. The former adopts an early stop tree growth strategy; the latter first generates an initial decision tree according to the largest scale, and then performs pruning. more successful in the field of hierarchical evaluation.

On the basis of determining the split attribute and pruning strategy, the decision tree algorithm is used to mine the established lithology identification data set. The decision tree algorithm can give the weight of each logging parameter in lithology identification, and then optimize the sensitive The parameters are modeled, and the parameters with smaller weight and lower sensitivity are not involved in the establishment of the decision tree model. After determining the sensitive parameters for modeling, the decision tree algorithm establishes a tree-like identification model from top to bottom "divide and conquer". Each branch of the tree represents a type of lithology identification rule, and the leaf nodes represent the attribute parameters that constitute the identification rule and The range of values for each parameter.

Specifically, among the three sensitive parameters of conglomerate lithology selected from the seven excavated fields, the parameter weight of the undisturbed formation resistivity is the largest, and the sensitivity is the highest; Lithology identification has a small weight and low sensitivity, so it does not participate in the establishment of the decision tree model. Based on the optimization of lithology-sensitive parameters, the decision tree algorithm establishes a tree-like identification model. On the whole, the lithology identification model starts to test data samples from the root node, which is the original formation resistivity, and divides and conquers from top to bottom. Four levels, each branch of the tree represents a type of lithology identification rules, as shown in Figure 2.

The "branch rule" established by using the decision tree can accurately identify the lithology in the sample data set. There are 309 sample data of conglomerate lithology, and the decision tree model can accurately identify 296, and the comprehensive identification accuracy rate reaches 95.79%. Since the high-precision and non-linear recognition model is established, the recognition accuracy is greatly improved compared with the conventional crossplot method. In addition, the generalization ability of the model on other data sets is relatively good, and it can be applied to the complex lithology identification of other wells in the study area.

The decision tree data mining algorithm automatically selects the sensitive parameters of conglomerate lithology identification according to the given mining fields, and then establishes a tree-like identification model of nonlinear mapping. The comprehensive identification accuracy of the model is relatively high. In order to provide geophysicists with a more intuitive and easy-to-apply identification model, among the three sensitive parameters of conglomerate lithology selected by the algorithm, first construct the product of acoustic time difference and natural gamma ray (AC*GR), and then combine the original Formation resistivity (Rt) is used to make a conglomerate lithology intersection chart, and the change law of the reservoir is analyzed according to the change information of the intersection chart, which provides geological basis for the adjustment of the development plan of the conglomerate reservoir and the avoidance of strong water-flooded layers.

Using the lithology identification model established in this example to evaluate the complex lithology of conglomerate reservoirs in Karamay Oilfield has achieved good geological application results, and provides a new solution for reservoir evaluation of complex oil and gas reservoirs. From the example of conglomerate reservoir lithology identification, it can be seen that data mining can automatically find lithological sensitive parameters in a large amount of unknown data, and establish a nonlinear decision tree identification model through the analysis of parameter weights. The process and results provide ideas and methods for the study of using geophysical knowledge to establish lithological maps. The two methods are interdependent and mutually promoting. Data mining can provide people with some analysis rules of massive data in the early stage and help geophysics research to extract useful information; geophysicists use their background knowledge to select some important information and Constructing some new parameters can help data mining methods determine more reasonable and effective mining fields, and provide accurate, quantitative, and subdivided results for the establishment of geophysical models.

As shown in Figure 3, correspondingly, this embodiment also proposes a complex lithology identification system based on a decision tree data mining algorithm, and the system includes:

The data set establishment module is used to establish the lithology profile of the sealed core section, and form the lithology identification data set according to the lithology profile;

A preprocessing module, configured to perform data preprocessing on the lithology identification data set;

The identification module is used to identify different lithologies in the lithology data set with different logging parameters respectively;

The data mining module is used to perform data mining on the lithology in the preprocessed lithology identification data set to form a tree identification model, so as to analyze the complex lithology according to the lithology corresponding to the tree identification model and logging parameters. to identify.

Before forming the lithology identification data set, combined with the geological background and reservoir characteristics of the study area, and based on the overall requirements of reservoir evaluation, the lithology name of the study area is determined from two aspects: the description of the core samples of the sealed coring well and the naming of the casting thin sections , because of the strong heterogeneity of complex reservoirs and drastic changes in lithology, the cast thin slices of cores cannot accurately reflect the real lithology of formations; and the description of cores lacks quantitative scales for different mineral components. The two methods are effective Combining can improve the accuracy of complex lithological naming.

On the basis of lithology nomenclature, based on sequence stratigraphy, the lithology section of the closed core section is established, and the logging curve response values of different lithology sections are read. The principles of logging reading values are as follows: ①Thick layer Read the average value as the basic data for lithology identification; ② read the maximum or minimum value of the logging curve for thin interlayers such as mudstone, glutenite, volcanic rock, and metamorphic rock as the basic data for lithology identification; ③ between different lithologies Read the average value of the transition zone, or use it as an abnormal point in data preprocessing for backup processing.

Data preprocessing for the established lithology identification sample data set mainly includes filling missing values, overall standardization, eliminating abnormal points, etc. The most important preprocessing procedure is the deletion of abnormal points. For reservoirs with complex lithology , there are two main reasons for the abnormality of data samples. One is the existence of thin interlayers. Due to the limitation of the longitudinal resolution of logging tools, the logging response value cannot reflect the real information of formation lithology; The upper logging response is affected by the surrounding rocks on both sides, and it is difficult for the logging data to accurately calibrate the real lithology information.

Determine the complex lithology categories that need to be identified based on lithology naming, and according to the requirements of reservoir evaluation, some similar lithologies can be merged, and some important lithologies can also be subdivided; in addition, the mining field of the decision tree algorithm is determined, That is, which well logs are selected to indicate lithology information.

In this embodiment, a decision tree algorithm is used to perform data mining on the lithology in the preprocessed lithology identification data set.

Specifically, the decision tree algorithm uses C5.0 for data mining of conglomerate lithology. The algorithm uses the information gain ratio (Information Gain Ratio) as the division measure, and selects the attribute with the largest information gain ratio as the split node; C5.0 adopts the pessimistic The method of error pruning (Pessimistic Error Pruning) is to use the continuous correction in the binomial distribution to correct the re-substitution error to obtain a more realistic error rate.

The selection of split attributes, that is, how to select the best split attribute from many attributes, is the core of the algorithm. The selection criteria can be roughly divided into two categories: ① The selection strategy of independent attributes; ② The selection strategy of interrelated attributes. Both methods have their advantages and disadvantages, and are optimized based on actual sample data. The pruning strategy includes two types: pre-pruning and post-pruning. The former adopts an early stop tree growth strategy; the latter first generates an initial decision tree according to the largest scale, and then performs pruning. more successful in the field of hierarchical evaluation.

On the basis of determining the split attribute and pruning strategy, the decision tree algorithm is used to mine the established lithology identification data set. The decision tree algorithm can give the weight of each logging parameter in lithology identification, and then optimize the sensitive The parameters are modeled, and the parameters with smaller weight and lower sensitivity are not involved in the establishment of the decision tree model. After determining the sensitive parameters for modeling, the decision tree algorithm establishes a tree-like identification model from top to bottom "divide and conquer". Each branch of the tree represents a type of lithology identification rule, and the leaf nodes represent the attribute parameters that constitute the identification rule and The range of values for each parameter.

The complex lithology identification method based on the decision tree data mining algorithm described in the present invention, its core technology is the optimization of splitting attribute values and pruning strategies, the algorithm determines each The weight of logging curves in lithology identification, and then optimize the lithology sensitive parameters to build a tree model, the accuracy and generalization ability of the model are relatively good. Compared with crossplot and other data mining algorithms, the decision tree lithology identification method can provide a high-precision, non-linear identification model. For the evaluation of reservoirs with complex lithology, its identification accuracy is higher than that of ordinary crossplot methods; another On the one hand, the decision tree algorithm belongs to the "white box" storage model, which can clearly understand how the classifier works and the relative importance of various parameters, and it also has a good guiding role in the research of geophysics. Therefore, the invention has important application value and good market demand in reservoir evaluation of complex lithology.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection of the present invention. within range.

Claims (10)

1. a kind of Complex lithologic identification method based on decision tree data mining algorithm, it is characterised in that methods described includes：

The lithological profile of sealing core drilling well section is set up, Lithology Discrimination data set is formed according to lithological profile；

Data prediction is carried out to the Lithology Discrimination data set；

Correspond to the different lithology in Lithology Discrimination data set respectively with different log parameters；

Data mining is carried out to the lithology in the Lithology Discrimination data set after pretreatment, tree-shaped identification model is formed, so that root Complex lithology is identified according to the lithology corresponding to the tree-shaped identification model and log parameter.

2. a kind of Complex lithologic identification method based on decision tree data mining algorithm according to claim 1, its feature It is, after the lithological profile of sealing core drilling well section is set up, the well logging readings principle according to thick-layer, short lap and lithology intermediate zone The log response of different lithology section is read respectively, the corresponding relation of conglomerate lithology and log parameter is set up, and forms lithology Identification data collection.

3. a kind of Complex lithologic identification method based on decision tree data mining algorithm according to claim 2, its feature It is, the process that data prediction is carried out to the Lithology Discrimination data set is standardized and disappeared including polishing missing values, entirety Except abnormity point.

4. a kind of Complex lithologic identification method based on decision tree data mining algorithm according to claim 3, its feature It is, data mining is carried out to the lithology in the Lithology Discrimination data set after pretreatment using decision Tree algorithms.

5. a kind of Complex lithologic identification side based on decision tree data mining algorithm according to any one of Claims 1-4 Method, it is characterised in that be to the process that the lithology in the Lithology Discrimination data set after pretreatment carries out data mining：Calculate Each weight of the log parameter in Lithology Discrimination, is obtained sensitive parameter and is modeled, finally built from top to bottom with characteristic manner Tree-shaped identification model is found, wherein each branch set represents the recognition rule of a class lithology, leaf node represents to constitute the identification The log parameter of rule and the numerical intervals of each log parameter.

6. a kind of Complex lithologic identification system based on decision tree data mining algorithm, it is characterised in that the system includes：

Data set sets up module, the lithological profile for setting up sealing core drilling well section, and Lithology Discrimination number is formed according to lithological profile According to collection；

Pretreatment module, for carrying out data prediction to the Lithology Discrimination data set；

Mark module, for corresponding to the different lithology in Lithology Discrimination data set respectively with different log parameters；

Data-mining module, for carrying out data mining to the lithology in the Lithology Discrimination data set after pretreatment, forms tree Shape identification model, so that the lithology according to corresponding to the tree-shaped identification model and log parameter is known to complex lithology Not.

7. a kind of Complex lithologic identification system based on decision tree data mining algorithm according to claim 6, its feature Be, the data set is set up module and is additionally operable to after the lithological profile of sealing core drilling well section is set up, according to thick-layer, short lap and The well logging readings principle of lithology intermediate zone reads the log response of different lithology section respectively, sets up conglomerate lithology and well logging The corresponding relation of parameter, forms Lithology Discrimination data set.

8. a kind of Complex lithologic identification system based on decision tree data mining algorithm according to claim 7, its feature Be, the pretreatment module specifically for the Lithology Discrimination data set is carried out polishing missing values, overall standardization and Eliminate the pretreatment of abnormity point.

9. a kind of Complex lithologic identification system based on decision tree data mining algorithm according to claim 8, its feature It is, the data-mining module is specifically for using decision Tree algorithms to the rock in the Lithology Discrimination data set after pretreatment Property carry out data mining.

10. a kind of Complex lithologic identification system based on decision tree data mining algorithm according to any one of claim 6 to 9 System, it is characterised in that the data-mining module is additionally operable to calculate weight of each log parameter in Lithology Discrimination, is obtained Sensitive parameter is modeled, and finally sets up tree-shaped identification model from top to bottom with characteristic manner, wherein each the branch's generation set The recognition rule of the class lithology of table one, leaf node represents to constitute the log parameter of the recognition rule and the numerical value of each log parameter It is interval.