Drilling fluid performance analysis and early warning method integrating knowledge base and mode mining

Abstract

The invention relates to the field of artificial intelligence, and discloses a drilling fluid performance analysis and early warning method fusing a knowledge base and mode mining, which comprises the following steps: acquiring time sequence historical data such as drilling fluid density, funnel viscosity, plastic viscosity and sand content, and constructing a knowledge base module based on the acquired data; and mining a strong association rule of drilling fluid performance change by using a sequence pattern mining module of prefix projection. And finally, decomposing the data into a trend item, a periodic item and a random item through a time sequence decomposition technology, constructing an analysis and early warning module in combination with a long-short-term memory network, and sending out an early warning signal in time when monitoring that the performance parameters deviate from a normal mode or predicting that the performance is about to be abnormal. According to the method, a new effective method is provided for drilling fluid performance analysis and pollution early warning, the knowledge base and pattern mining are fused, the time sequence decomposition technology and time sequence long-term prediction are utilized, and the problems of overfitting phenomenon and abnormal value and noise tolerance are well solved.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence, and in particular relates to a drilling fluid performance analysis and early warning method integrating a knowledge base and pattern mining. Background Art

[0002] As a key branch of artificial intelligence, machine learning has rapidly risen with the rapid development of Internet technology and data storage technology, and has been widely used in many fields. In the field of drilling fluid performance analysis and early warning, machine learning technology has also gradually gained attention. As the core medium in the drilling process, the performance of drilling fluid directly and critically affects the efficiency and safety of drilling operations. Traditional drilling fluid performance analysis methods mainly rely on expert experience and simple data analysis methods, which are difficult to meet the increasingly stringent requirements of modern drilling projects for drilling fluid performance.

[0003] Among the current methods of applying machine learning to drilling fluid performance analysis, traditional prediction models such as BP neural network, support vector machine and random forest are mostly used. These models have improved the accuracy and efficiency of drilling fluid performance analysis to a certain extent, but there are still many limitations. For example, although BP neural network has strong nonlinear mapping capabilities, it is prone to overfitting when processing complex data sets, especially in small sample and high-noise drilling fluid data scenarios; support vector machine performs well in processing small sample and high-dimensional data, but its computational efficiency is significantly reduced when dealing with large-scale data; although random forest has high accuracy and robustness, it is still sensitive to outliers and noise data.

[0004] In addition, most existing methods fail to fully consider the time series characteristics inherent in drilling fluid performance data. The performance of drilling fluid will change dynamically over time, and these changes are usually closely related to various factors in the drilling process. If time series data can be effectively used, it will help to more accurately predict the changing trend of drilling fluid performance, and then take targeted adjustment and optimization measures in a timely manner.

[0005] In view of the above problems, the present invention proposes a drilling fluid performance analysis and early warning method that integrates knowledge base and pattern mining. This method aims to deeply integrate drilling fluid performance parameters, geological information and engineering data by building a knowledge base to build a comprehensive and systematic knowledge system. At the same time, the sequence pattern mining technology is used to mine the strong correlation rules of the evolution of drilling fluid performance from historical data. On this basis, the time series decomposition technology and long-short-term memory network are combined to monitor and warn the drilling fluid performance in real time, so as to timely discover abnormal performance conditions and take corresponding response strategies. Summary of the invention

[0006] In view of the limitations of the above-mentioned prior art, the present invention aims to provide a method integrating knowledge base and pattern mining to achieve accurate analysis of drilling fluid performance and pollution warning.

[0007] Step 1: A drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining, characterized in that it includes the following steps: Step S10, collecting multivariate time series data related to drilling fluid performance, and preprocessing the time series data, wherein the time series data includes drilling fluid performance parameters, geological information and engineering data; the drilling fluid performance parameters include drilling fluid density, funnel viscosity, plastic viscosity, pH value, shear force 10 seconds, shear force 10 minutes, API fluid loss, API filter cake thickness, mud cake friction coefficient, yield point, sand content, oil content and water content; the geological information includes sampling depth, sampling location, formation and pore pressure; the engineering data includes drilling speed, drill bit speed, pump pressure and outlet temperature; Step S20: Based on the knowledge graph technology, the preprocessed multivariate time series data entities and their relationships are stored in a graph structure, a knowledge base module is constructed, and the weights are calculated based on the correlation analysis of historical data and the scores of domain experts; Step S30, the pre-processed multivariate time series data is divided into a transaction database through a sliding window, and a sequence pattern mining module is used to set a minimum support threshold and a minimum confidence threshold, traverse the transaction database, and the candidate sequence in the evolution process of drilling fluid performance is mined through the pattern mining method of prefix projection, and the rules of high-weight parameter combinations are screened out as strong association rules. In the screening process, the confidence is updated according to the weight of the knowledge base in step S20, and the knowledge base weight is iteratively optimized according to the confidence; Step S40, the multi-parameter statistical feature set is analyzed using time series decomposition technology and long short-term memory network model. First, the multi-parameter statistical feature set is decomposed into trend items that change with engineering data, periodic items affected by the alternation of day and night, and random items caused by accidental factors. Secondly, the knowledge base and strong association rules are integrated, and an analysis and early warning module is established through the long short-term memory network. When the drilling fluid performance parameters are monitored to deviate from the normal mode or it is predicted that the performance is about to become abnormal, a warning signal is issued in time.

[0008] Step 2: According to the drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining described in step 1, it is characterized in that in step S20, the knowledge base module is defined as: in, Represents the knowledge base, stored in the form of a graph structure, It is a node set, including drilling fluid performance parameter nodes, geological information nodes and engineering data nodes. is a set of edges, representing the relationship between nodes. is a node attribute function, defined as , which means mapping each node to a set of attributes , by Stored in tuple form, is the edge type function, defined as , which means mapping each edge to a set of predefined semantic types , Contains adaptation, causality, timing dependencies, and composite relationships, is the edge weight, and its initial value is determined based on the correlation analysis of historical data and the scores of domain experts, and is defined as , For Node and The Pearson correlation coefficient between is the expert score, with a value range of [0,1], It is a balancing factor, which is determined by domain experts. Through the graph structure, effective organization and structured storage of drilling fluid related knowledge can be achieved.

[0009] Step 3: According to the drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining described in step 1, it is characterized in that in step S30, the sequence pattern mining module is defined as: in, represents the sequential pattern mining module, is the association rule representation, It is a set of pre-parameters that trigger the association rules. is with the collection The associated post-parameter set, where , ,and , represents the empty set, Representing a collection and collection There are no common elements. It is a pattern mining method based on prefix projection, and the output is a candidate sequence. It is a transactional database generated from multivariate time series data through sliding windows. is the minimum support threshold, is the minimum confidence threshold, is the weight, and the support calculation is defined as , affairs Is a transactional database A subset of represents a drilling fluid performance parameter sequence, namely , the confidence calculation is defined as and , Represents a transactional database Also contains the collection and collection The proportion of Represents a transactional database Contains the collection The confidence is updated in the process of selecting high-weight parameter rules from the candidate sequence. The update method is defined as , and iteratively optimize the weight of the knowledge base , the optimization method is defined as , Indicates that the knowledge base module and the rule in step S20 are obtained Related edge set The highest weight ,in, and Separately and collection The corresponding knowledge base node.

[0010] Step 4: According to the drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining described in step 1, it is characterized in that in step S40, the analysis and early warning module is defined as: in, It represents the analysis and early warning module. is the warning function, is a prediction function based on long short-term memory network, It refers to the use of the time series decomposition technique Holt-Winters triple exponential smoothing method to Decomposition into trend terms , Periodic Item and random items , It is a set of multi-parameter statistical features, composed of transaction database Calculated, including mean, standard deviation, trend slope, maximum and minimum values, is the knowledge base constructed in the fusion step S20, is the strong association rule mined in the fusion step S30, is the warning threshold, defined by expert experience; trend item , periodic term , random item , is the present moment, It was the moment before. is a fixed cycle length, yes The horizontal term at time , defined as , are the smoothing coefficients of the horizontal term, trend term, and periodic term, respectively. yes The horizontal term at time, yes Trend items at the moment, yes The periodic term of time, yes The actual observed value at time; prediction function Taking trend term, cycle term and random term as the input of long short-term memory network, Extracting transactions Related nodes and weights are used as auxiliary inputs, combined with strong association rules Adjust the predicted value to get the predicted drilling fluid performance value ; Warning function The judgment standard is that when the deviation between the predicted value and the normal range exceeds the set threshold, that is, The early warning mechanism is triggered. is the normal performance range value, defined by expert experience.

[0011] The beneficial effects of the present invention are as follows: This invention deeply integrates multi-source heterogeneous drilling fluid related data to form a comprehensive knowledge system by constructing a knowledge base, effectively overcoming the problems of data isolation and lack of effective association in existing methods, and using sequence pattern mining technology to mine the potential laws and patterns of drilling fluid performance evolution from historical data, so as to more deeply understand the internal mechanism of drilling fluid performance evolution. Compared with traditional methods that only rely on simple data analysis, the method of the present invention can discover more complex and hidden performance evolution patterns. Further combined with the time series analysis method, the present invention fully considers the time series characteristics of drilling fluid performance data, can monitor the trend of drilling fluid performance changes in real time, and timely discover performance anomalies. The time series analysis method that integrates knowledge base and pattern mining proposed by the present invention can more accurately analyze drilling fluid performance and realize pollution warning, which helps drilling engineers take measures in advance to adjust drilling fluid performance, ensure the efficient and safe drilling operation, and ultimately improve the overall efficiency of drilling operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 For the algorithm architecture; DETAILED DESCRIPTION

[0013] The embodiments of the present invention are described in detail below with reference to the accompanying drawings, but the present invention can be implemented in many different ways as defined and covered by the claims.

[0014] The present invention aims to accurately analyze the performance of drilling fluid and provide pollution warning by integrating knowledge base and sequence pattern mining method and combining time series analysis technology.

[0015] First, multivariate time series data related to drilling fluid performance are collected, and the collected data are preprocessed, including missing value processing, outlier detection and processing, data smoothing, and data normalization. Secondly, the preprocessed drilling fluid performance parameters, geological information, and engineering data are deeply integrated using the knowledge graph construction tool, stored in the form of a graph structure, and a knowledge base module is constructed. Then, the minimum support threshold and the minimum confidence threshold are set using the sequence pattern mining module, and the transaction database is traversed to mine strong association rules through the pattern mining method of prefix projection. Finally, the multi-parameter statistical feature set is decomposed into trend items, periodic items, and random items using the time series decomposition technology, and the change law of each component is analyzed. The drilling fluid performance parameter value in the future period is used as the prediction target, and the long-term and short-term memory network model is trained. The knowledge base is used for feature enhancement, and the strong association rules are combined for rule constraints. The analysis and early warning module is constructed. When the performance parameters are monitored to deviate from the normal mode or the performance is predicted to be abnormal, the early warning signal is issued in time.

[0016] Step 1: A drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining, characterized in that it includes the following steps: Step S10, collecting multivariate time series data related to drilling fluid performance, and preprocessing the time series data, wherein the time series data includes drilling fluid performance parameters, geological information and engineering data; the drilling fluid performance parameters include drilling fluid density, funnel viscosity, plastic viscosity, pH value, shear force 10 seconds, shear force 10 minutes, API fluid loss, API filter cake thickness, mud cake friction coefficient, yield point, sand content, oil content and water content; the geological information includes sampling depth, sampling location, formation and pore pressure; the engineering data includes drilling speed, drill bit speed, pump pressure and outlet temperature; Step S20: Based on the knowledge graph technology, the preprocessed multivariate time series data entities and their relationships are stored in a graph structure, a knowledge base module is constructed, and the weights are calculated based on the correlation analysis of historical data and the scores of domain experts; Step S30, the pre-processed multivariate time series data is divided into a transaction database through a sliding window, and a sequence pattern mining module is used to set a minimum support threshold and a minimum confidence threshold, traverse the transaction database, and the candidate sequence in the evolution process of drilling fluid performance is mined through the pattern mining method of prefix projection, and the rules of high-weight parameter combinations are screened out as strong association rules. In the screening process, the confidence is updated according to the weight of the knowledge base in step S20, and the knowledge base weight is iteratively optimized according to the confidence; Step S40, the multi-parameter statistical feature set is analyzed using time series decomposition technology and long short-term memory network model. First, the multi-parameter statistical feature set is decomposed into trend items that change with engineering data, periodic items affected by the alternation of day and night, and random items caused by accidental factors. Secondly, the knowledge base and strong association rules are integrated, and an analysis and early warning module is established through the long short-term memory network. When the drilling fluid performance parameters are monitored to deviate from the normal mode or it is predicted that the performance is about to become abnormal, a warning signal is issued in time.

[0017] Step 2: According to the drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining described in step 1, it is characterized in that in step S20, the knowledge base module is defined as: in, Represents the knowledge base, stored in the form of a graph structure, It is a node set, including drilling fluid performance parameter nodes, geological information nodes and engineering data nodes. is a set of edges, representing the relationship between nodes. is a node attribute function, defined as , which means mapping each node to a set of attributes , by Stored in tuple form, is the edge type function, defined as , which means mapping each edge to a set of predefined semantic types , Contains adaptation, causality, timing dependencies, and composite relationships, is the edge weight, and its initial value is determined based on the correlation analysis of historical data and the scores of domain experts, and is defined as , For Node and The Pearson correlation coefficient between is the expert score, with a value range of [0,1], It is a balancing factor, which is determined by domain experts. Through the graph structure, effective organization and structured storage of drilling fluid related knowledge can be achieved.

[0018] Step 3: According to the drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining described in step 1, it is characterized in that in step S30, the sequence pattern mining module is defined as: in, represents the sequential pattern mining module, is the association rule representation, It is a set of pre-parameters that trigger the association rules. is with the collection The associated post-parameter set, where , ,and , represents the empty set, Representing a collection and collection There are no common elements. It is a pattern mining method based on prefix projection, and the output is a candidate sequence. It is a transactional database generated from multivariate time series data through sliding windows. is the minimum support threshold, is the minimum confidence threshold, is the weight, and the support calculation is defined as , affairs Is a transactional database A subset of represents a drilling fluid performance parameter sequence, namely , the confidence calculation is defined as and , Represents a transactional database Also contains the collection and collection The proportion of Represents a transactional database Contains the collection The confidence is updated in the process of selecting high-weight parameter rules from the candidate sequence. The update method is defined as , and iteratively optimize the weight of the knowledge base , the optimization method is defined as , Indicates that the knowledge base module and the rule in step S20 are obtained Related edge set The highest weight ,in, and Separately and collection The corresponding knowledge base node.

[0019] Step 4: According to the drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining described in step 1, it is characterized in that in step S40, the analysis and early warning module is defined as: in, It represents the analysis and early warning module. is the warning function, is a prediction function based on long short-term memory network, It refers to the use of the time series decomposition technique Holt-Winters triple exponential smoothing method to Decomposition into trend terms , Periodic Item and random items , It is a set of multi-parameter statistical features, composed of transaction database Calculated, including mean, standard deviation, trend slope, maximum and minimum values, is the knowledge base constructed in the fusion step S20, is the strong association rule mined in the fusion step S30, is the warning threshold, defined by expert experience; trend item , periodic term , random item , is the present moment, It was the moment before. is a fixed cycle length, yes The horizontal term at time , defined as , are the smoothing coefficients of the horizontal term, trend term, and periodic term, respectively. yes The horizontal term at time, yes Trend items at the moment, yes The periodic term of time, yes The actual observed value at time; prediction function Taking trend term, cycle term and random term as the input of long short-term memory network, Extracting transactions Related nodes and weights are used as auxiliary inputs, combined with strong association rules Adjust the predicted value to get the predicted drilling fluid performance value ; Warning function The judgment standard is that when the deviation between the predicted value and the normal range exceeds the set threshold, that is, The early warning mechanism is triggered. is the normal performance range value, defined by expert experience.

[0020] The above contents are only specific embodiments of the present invention, and do not limit the protection scope of the invention. Any modification, replacement, improvement or equivalent replacement that does not deviate from the basic technical concept of the present invention and the contents described in the drawings is included in the protection scope of the present invention.

Claims

1. A drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining, characterized in that: The following steps are involved: Step S10, collecting multivariate time series data related to drilling fluid performance, and preprocessing the time series data, wherein the time series data includes drilling fluid performance parameters, geological information and engineering data; the drilling fluid performance parameters include drilling fluid density, funnel viscosity, plastic viscosity, pH value, shear force 10 seconds, shear force 10 minutes, API fluid loss, API filter cake thickness, mud cake friction coefficient, yield point, sand content, oil content and water content; the geological information includes sampling depth, sampling location, formation and pore pressure; the engineering data includes drilling speed, drill bit speed, pump pressure and outlet temperature; Step S20: Based on the knowledge graph technology, the preprocessed multivariate time series data entities and their relationships are stored in a graph structure, a knowledge base module is constructed, and the weights are calculated based on the correlation analysis of historical data and the scores of domain experts; Step S30, the pre-processed multivariate time series data is divided into a transaction database through a sliding window, and a sequence pattern mining module is used to set a minimum support threshold and a minimum confidence threshold, traverse the transaction database, and the candidate sequence in the evolution process of drilling fluid performance is mined through the pattern mining method of prefix projection, and the rules of high-weight parameter combinations are screened out as strong association rules. In the screening process, the confidence is updated according to the weight of the knowledge base in step S20, and the knowledge base weight is iteratively optimized according to the confidence; Step S40, the multi-parameter statistical feature set is analyzed using time series decomposition technology and long short-term memory network model. First, the multi-parameter statistical feature set is decomposed into trend items that change with engineering data, periodic items affected by the alternation of day and night, and random items caused by accidental factors. Secondly, the knowledge base and strong association rules are integrated, and an analysis and early warning module is established through the long short-term memory network. When the drilling fluid performance parameters are monitored to deviate from the normal mode or it is predicted that the performance is about to become abnormal, a warning signal is issued in time.

2. The drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining according to claim 1 is characterized in that: In step S20, the knowledge base module is defined as: in, Represents the knowledge base, stored in the form of a graph structure, It is a node set, including drilling fluid performance parameter nodes, geological information nodes and engineering data nodes. is a set of edges, representing the relationship between nodes. is a node attribute function, defined as , which means mapping each node to a set of attributes , by Stored in tuple form, is the edge type function, defined as , which means mapping each edge to a set of predefined semantic types , Contains adaptation, causality, timing dependencies, and composite relationships, is the edge weight, and its initial value is determined based on the correlation analysis of historical data and the scores of domain experts, and is defined as , For Node and The Pearson correlation coefficient between is the expert score, with a value range of [0,1], It is a balancing factor, which is determined by domain experts. Through the graph structure, effective organization and structured storage of drilling fluid related knowledge can be achieved.

3. The drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining according to claim 1 is characterized in that: In step S30, the sequence pattern mining module is defined as: in, represents the sequential pattern mining module, is the association rule representation, It is a set of pre-parameters that trigger the association rules. is with the collection The associated post-parameter set, where , ,and , represents the empty set, Representing a collection and collection There are no common elements. It is a pattern mining method based on prefix projection, and the output is a candidate sequence. It is a transactional database generated from multivariate time series data through sliding windows. is the minimum support threshold, is the minimum confidence threshold, is the weight, and the support calculation is defined as , affairs Is a transactional database A subset of represents a drilling fluid performance parameter sequence, namely , the confidence calculation is defined as and , Represents a transactional database Also contains the collection and collection The proportion of Represents a transactional database Contains the collection The confidence is updated in the process of selecting high-weight parameter rules from the candidate sequence. The update method is defined as , and iteratively optimize the weight of the knowledge base , the optimization method is defined as , Indicates that the knowledge base module and the rule in step S20 are obtained Related edge set The highest weight ,in, and Separately and collection The corresponding knowledge base node.

4. The drilling fluid performance analysis and early warning method integrating knowledge base and pattern mining according to claim 1 is characterized in that: In step S40, the analysis and warning module is defined as: in, It represents the analysis and early warning module. is the warning function, is a prediction function based on long short-term memory network, It refers to the use of the time series decomposition technique Holt-Winters triple exponential smoothing method to Decomposition into trend terms , Periodic Item and random items , It is a set of multi-parameter statistical features, composed of transaction database Calculated, including mean, standard deviation, trend slope, maximum and minimum values, is the knowledge base constructed in the fusion step S20, is the strong association rule mined in the fusion step S30, is the warning threshold, defined by expert experience; trend item , periodic term , random item , is the present moment, It was the moment before. is a fixed cycle length, yes The horizontal term at time , defined as , are the smoothing coefficients of the horizontal term, trend term, and periodic term, respectively. yes The horizontal term at time, yes Trend items at the moment, yes The periodic term of time, yes The actual observed value at time; prediction function Taking trend term, cycle term and random term as the input of long short-term memory network, Extracting transactions Related nodes and weights are used as auxiliary inputs, combined with strong association rules Adjust the predicted value to get the predicted drilling fluid performance value ; Warning function The judgment standard is that when the deviation between the predicted value and the normal range exceeds the set threshold, that is, The early warning mechanism is triggered. is the normal performance range value, defined by expert experience.

Images