Drilling Fluid Performance Analysis and Early Warning Method Integrating Knowledge Base and Pattern Mining

Abstract

The present invention relates to the field of artificial intelligence, and discloses a method for analyzing and warning the performance of drilling fluid by integrating a knowledge base and pattern mining. By collecting time-series historical data such as drilling fluid density, funnel viscosity, plastic viscosity, sand content, etc., a knowledge base module is constructed based on the collected data, and a strong association rule for the change of drilling fluid performance is mined by using a prefix projection sequence pattern mining module. Finally, the data is decomposed into a trend term, a periodic term, and a random term through time series decomposition technology, and combined with a long short-term memory network to construct an analysis and warning module. When it is monitored that the performance parameter deviates from the normal mode or it is predicted that the performance is about to be abnormal, a warning signal is sent in time. The present invention provides a new and effective method for analyzing the performance of drilling fluid and pollution warning. This method integrates a knowledge base and pattern mining, and uses time series decomposition technology and long-term time series prediction to well solve the problems of overfitting phenomenon and tolerance of outliers and noise.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence, and particularly relates to a method for analyzing and warning the performance of drilling fluid by integrating a knowledge base and pattern mining. Background Art

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

[0003] In the current methods of applying machine learning to analyze the performance of drilling fluid, 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 analyzing the performance of drilling fluid to a certain extent. However, there are still many limitations. For example, although the BP neural network has strong non-linear mapping ability, it is prone to overfitting when dealing with complex data sets, especially in the scenarios of small samples and high-noise drilling fluid data; the support vector machine performs excellently when dealing with small samples and high-dimensional data, but its computational efficiency is significantly reduced when dealing with large-scale data; the random forest has high accuracy and robustness, but it is still relatively sensitive to outliers and noise data.

[0004] In addition, most of the existing methods do not fully consider the inherent time series characteristics of drilling fluid performance data. The performance of drilling fluid changes dynamically over time, and these changes are usually closely related to various factors in the drilling process. If time series data can be effectively utilized, 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 method for analyzing and warning the performance of drilling fluid by integrating a knowledge base and pattern mining. The purpose of this method is to deeply integrate drilling fluid performance parameters, geological information, and engineering data by constructing a knowledge base to build a comprehensive and systematic knowledge system. At the same time, by using sequence pattern mining technology, strong association rules for the evolution of drilling fluid performance are mined from historical data. On this basis, combined with time series decomposition technology and long short-term memory network, the performance of drilling fluid is monitored and warned in real time, so as to timely detect abnormal performance conditions and take corresponding countermeasures. Summary of the Invention

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

[0007] Step 1: A method for analyzing and warning the performance of drilling fluid by integrating a knowledge base and pattern mining, which is characterized by including the following steps:

[0008] Step S10: Collect multivariate time-series data related to the performance of drilling fluid, and preprocess the time-series data. Among them, 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 at 10 seconds, shear force at 10 minutes, API filtration 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, bit rotation speed, pump pressure, and outlet temperature;

[0009] Step S20: Based on knowledge graph technology, store the preprocessed multivariate time-series data entities and their relationships in the form of a graph structure, construct a knowledge base module, and calculate weights based on the correlation analysis of historical data and domain expert scoring;

[0010] Step S30: Divide the preprocessed multivariate time-series data through a sliding window to obtain a transaction database. Use the sequence pattern mining module to set the minimum support threshold and the minimum confidence threshold, traverse the transaction database, and use the prefix projection pattern mining method to mine the candidate sequences in the evolution process of drilling fluid performance. Screen out the rules of high-weight parameter combinations as strong association rules, update the confidence according to the weights of the knowledge base in step S20 during the screening process, and iteratively optimize the knowledge base weights according to the confidence;

[0011] Step S40: Analyze the multi-parameter statistical feature set using time series decomposition technology and a long short-term memory network model. First, decompose the multi-parameter statistical feature set into a trend term that changes with engineering data, a periodic term affected by day and night alternation, and a random term caused by accidental factors. Secondly, integrate the knowledge base and strong association rules, and establish an analysis and warning module through a long short-term memory network. When it is detected that the drilling fluid performance parameters deviate from the normal mode or it is predicted that the performance will be abnormal, a warning signal is sent in a timely manner.

[0012] Step 2: According to the method for analyzing and warning the performance of drilling fluid by integrating a knowledge base and pattern mining described in Step 1, it is characterized in that in Step S20, the knowledge base module is defined as:

[0013]

[0014] Among them, Denotes a knowledge base, stored in the form of a graph structure, is a set of nodes, including nodes of drilling fluid performance parameters, geological information nodes, and engineering data nodes, is a set of edges, representing the association relationships between nodes, is a node attribute function, defined as , indicating mapping each node to a set of attributes , in tuple form, is an edge type function, defined as , indicating mapping each edge to a set of predefined semantic types , including adaptation, causality, temporal dependence, and composite relationships, is the edge weight, with the initial value determined based on the correlation analysis of historical data and the scoring by domain experts, defined as , is the Pearson correlation coefficient between nodes and , is the expert score, with the value range [0, 1], is the balance factor, determined by domain experts, and through the graph structure, realizes the effective organization and structured storage of drilling fluid-related knowledge.

[0015] Step 3: According to the drilling fluid performance analysis and warning method integrating the knowledge base and pattern mining described in Step 1, it is characterized in that in the step S30, the sequential pattern mining module is defined as:

[0016]

[0017] Among them, represents the sequential pattern mining module, is the association rule representation, is the set of preconditions for triggering the association rule, is the set of postconditions associated with the set , where , , and , represents the empty set, represents the set and the set have no common elements, is the pattern mining method of prefix projection, and the output is candidate sequences, is the transaction database, generated from multivariate time series data through a sliding window, is the minimum support threshold, is the minimum confidence threshold, is the weight, and the support calculation is defined as , the transaction is a subset of the transaction database , representing a sequence of drilling fluid performance parameters, that is , the confidence calculation is defined as and , represents the proportion of the transaction database that contains both the set and the set . represents the proportion of the transaction database that contains the set . In the process of screening out high-weight parameter rules from candidate sequences, the confidence is updated, and the update method is defined as , and the weight of the knowledge base is iteratively optimized , and the optimization method is defined as . represents taking the highest weight of the relevant edge set in the knowledge base module in step S20 corresponding to the rule , where and are the knowledge base nodes corresponding to the set and the set respectively.

[0018] Step 4: According to the drilling fluid performance analysis and early warning method that fuses the 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:

[0019]

[0020] Among them, represents the analysis and early warning module, is the early warning function, is the prediction function based on the long short-term memory network, refers to using the Holt-Winters triple exponential smoothing method of time series decomposition technology to decompose into a trend term , a periodic term and a random term , is a multi-parameter statistical feature set, calculated from the transaction database , and includes the mean, standard deviation, trend slope, maximum value, and minimum value. is the knowledge base constructed by fusing step S20, is the strong association rule mined by fusing step S30, is the early warning threshold, defined by expert experience; the trend term , the periodic term , the random term , is the current moment, is the previous moment, is the fixed cycle length, is the horizontal term at moment, defined as , are the smoothing coefficients of the horizontal term, trend term, and periodic term respectively, is the horizontal term at moment, is the trend term at moment, is the periodic term at moment, is the actual observed value at moment; the prediction function takes the trend term, periodic term, and random term as the input of the long short-term memory network, extracts the relevant nodes and weights of the transaction from the knowledge base as the auxiliary input, and then combines the strong association rules to adjust the predicted value to obtain the predicted drilling fluid performance value ; the judgment criterion of the early warning function is that when the deviation between the predicted value and the normal range exceeds the set threshold, that is when the early warning mechanism is triggered, is the normal performance range value, defined by expert experience.

[0021] The beneficial effects of the present invention are as follows:

[0022] Through the construction of a knowledge base, this invention deeply integrates multi-source heterogeneous drilling fluid-related data to form a comprehensive knowledge system, effectively overcoming the problems of data isolation and lack of effective association in existing methods. Using sequence pattern mining technology, it can mine the potential laws and patterns of the evolution of drilling fluid performance from historical data, thereby more deeply understanding the internal mechanism of the evolution of drilling fluid performance. Compared with traditional methods that only rely on simple data analysis, the method of this invention can discover more complex and hidden performance evolution patterns. Further combining time series analysis methods, this invention fully considers the time series characteristics of drilling fluid performance data, can real-time monitor the change trend of drilling fluid performance, and timely detect performance anomalies. The time series analysis method integrating the knowledge base and pattern mining proposed by this invention can more accurately analyze the drilling fluid performance and achieve pollution early warning, which helps drilling engineers take measures in advance to adjust the drilling fluid performance, ensure the efficient and safe progress of drilling operations, and ultimately improve the overall efficiency of drilling operations. Description of the Drawings

[0023] Figure 1 is the algorithm architecture; Specific implementation manners

[0024] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, the present invention can be implemented in many different ways defined and covered by the claims.

[0025] The present invention aims to accurately analyze the performance of drilling fluid and give pollution warnings by integrating a knowledge base and a sequence pattern mining method, and combining time series analysis techniques.

[0026] First, collect multivariate time series data related to the performance of drilling fluid, and preprocess the collected data, including missing value processing, outlier detection and processing, data smoothing, and data normalization. Second, use a knowledge graph construction tool to deeply integrate the preprocessed drilling fluid performance parameters, geological information, and engineering data, store them in the form of a graph structure, and construct a knowledge base module. Then, use the sequence pattern mining module to set the minimum support threshold and the minimum confidence threshold, traverse the transaction database, and mine strong association rules through the prefix projection pattern mining method. Finally, use the time series decomposition technology to decompose the multi-parameter statistical feature set into a trend term, a periodic term, and a random term, analyze the variation laws of each component, take the drilling fluid performance parameter values in a future period as the prediction target, train a long short-term memory network model, use the knowledge base for feature enhancement, combine strong association rules for rule constraints, construct an analysis and warning module, and send out a warning signal in time when it is monitored that the performance parameters deviate from the normal mode or it is predicted that the performance is about to be abnormal.

[0027] Step 1: A method for analyzing and warning the performance of drilling fluid by integrating a knowledge base and pattern mining, characterized by including the following steps:

[0028] Step S10: Collect multivariate time series data related to the performance of drilling fluid, and preprocess 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 at 10 seconds, shear force at 10 minutes, API filtration 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, bit rotation speed, pump pressure, and outlet temperature;

[0029] Step S20: Based on knowledge graph technology, store the preprocessed multivariate time series data entities and their relationships in the form of a graph structure, construct a knowledge base module, and calculate weights based on the correlation analysis of historical data and domain expert scoring;

[0030] Step S30: Divide the preprocessed multi-parameter time series data through a sliding window to obtain a transaction database. Using the sequence pattern mining module, set the minimum support threshold and the minimum confidence threshold, traverse the transaction database, and mine the candidate sequences in the evolution process of the drilling fluid performance through the pattern mining method of prefix projection. Select the rules of high-weight parameter combinations as strong association rules. During the screening process, update the confidence according to the weights in the knowledge base in step S20, and iteratively optimize the knowledge base weights according to the confidence;

[0031] Step S40: Analyze the multi-parameter statistical feature set using time series decomposition technology and long short-term memory network model. First, decompose the multi-parameter statistical feature set into a trend term that changes with engineering data, a periodic term affected by day and night alternation, and a random term caused by accidental factors. Secondly, fuse the knowledge base and strong association rules, and establish an analysis and early warning module through a long short-term memory network. When it is monitored that the drilling fluid performance parameters deviate from the normal mode or it is predicted that the performance will be abnormal, an early warning signal is sent in time.

[0032] Step 2: According to the drilling fluid performance analysis and early warning method that fuses the knowledge base and pattern mining described in step 1, it is characterized in that in step S20, the knowledge base module is defined as:

[0033]

[0034] Among them, represents the knowledge base, which is stored in the form of a graph structure, is the node set, including drilling fluid performance parameter nodes, geological information nodes, and engineering data nodes, is the edge set, representing the association relationship between nodes, is the node attribute function, defined as , indicating mapping each node to the attribute set , in the tuple form, is the edge type function, defined as , indicating mapping each edge to the predefined semantic type set , including adaptation, causality, temporal dependence, and composite relationships, is the edge weight, and its initial value is determined based on the correlation analysis of historical data and the scoring of domain experts, defined as , is the Pearson correlation coefficient between nodes and , is the expert score, and its value range is [0, 1], The balance factor is determined by domain experts and realizes the effective organization and structured storage of drilling fluid-related knowledge through the graph structure.

[0035] Step 3: According to the drilling fluid performance analysis and warning method integrating the knowledge base and pattern mining described in Step 1, it is characterized in that in the step S30, the sequential pattern mining module is defined as:

[0036]

[0037] Wherein, represents the sequential pattern mining module, is the association rule representation, is the set of precondition parameters triggering the association rule, is the set of postcondition parameters associated with the set where , , and , represents the empty set, represents the set and the set have no common elements, is the pattern mining method of prefix projection, and the output is candidate sequences, is the transaction database, which is generated from multivariate time series data through a sliding window, is the minimum support threshold, is the minimum confidence threshold, is the weight, and the support calculation is defined as , the transaction is a subset of the transaction database and represents a sequence of drilling fluid performance parameters, that is, , and the confidence calculation is defined as and , represents the proportion of the transaction database that simultaneously contains the set and the set , represents the proportion of the transaction database that contains the set . The confidence is updated in the process of screening out high-weight parameter rules from the candidate sequences, and the update method is defined as , and the weight of the knowledge base is iteratively optimized, and the optimization method is defined as , represents taking the highest weight of the edge set related to the rule in the knowledge base module in Step S20, wherein, and are the knowledge base nodes corresponding to the sets and the set respectively.

[0038] Step 4: According to the drilling fluid performance analysis and early warning method for integrating the knowledge base and pattern mining described in Step 1, it is characterized in that in the step S40, the analysis and early warning module is defined as:

[0039]

[0040] wherein, represents the analysis and early warning module, is the early warning function, is the prediction function based on the long short-term memory network, refers to using the Holt-Winters triple exponential smoothing method of time series decomposition technology to decompose into a trend term , a periodic term and a random term , is a multi-parameter statistical feature set, calculated from the transaction database , including the mean, standard deviation, trend slope, maximum value and minimum value, is the knowledge base constructed by integrating Step S20, is the strong association rule mined by integrating Step S30, is the early warning threshold, defined by expert experience; the trend term , the periodic term , the random term , is the current time, is the previous time, is the fixed cycle length, is the level term at time, defined as , are the smoothing coefficients of the level term, trend term, and periodic term respectively, is the level term at time, is the trend term at time, is the periodic term at time, is the actual observed value at time; the prediction function uses the trend term, periodic term, and random term as the input of the long short-term memory network, extracts the transaction related nodes and weights from the knowledge base as auxiliary inputs, and then combines the strong association rule Adjust the predicted value to obtain the predicted drilling fluid performance value ; Early warning function The judgment criterion of is that when the deviation between the predicted value and the normal range exceeds the set threshold, that is is the normal performance range value, which is defined by expert experience.

[0041] The above content is only the specific implementation manner of the present invention, rather than limiting the protection scope of the invention. Any modification, replacement, improvement or equivalent replacement without departing from the basic technical idea described in the present invention and the drawings is included in the protection scope of the present invention.

Claims

1. A method for analyzing and warning the performance of drilling fluid by integrating a knowledge base and pattern mining, characterized in that It includes the following steps: Step S10: Collect multivariate time series data related to the performance of drilling fluid, and preprocess the time series data. Among them, 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 at 10 seconds, shear force at 10 minutes, API filtration 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, bit rotation speed, pump pressure, and outlet temperature; Step S20: Based on knowledge graph technology, store the preprocessed multivariate time series data entities and their relationships in the form of a graph structure, construct a knowledge base module, and calculate weights based on the correlation analysis of historical data and domain expert scoring; Step S30: Divide the preprocessed multivariate time series data through a sliding window to obtain a transaction database. Use the sequence pattern mining module to set the minimum support threshold and the minimum confidence threshold, traverse the transaction database, and use the prefix projection pattern mining method to mine the candidate sequences in the evolution process of drilling fluid performance. Select the rules of high-weight parameter combinations as strong association rules, update the confidence according to the weights of the knowledge base in step S20 during the screening process, and iteratively optimize the knowledge base weights according to the confidence; Step S40: Analyze the multi-parameter statistical feature set using time series decomposition technology and long short-term memory network model. First, decompose the multi-parameter statistical feature set into a trend term that changes with engineering data, a periodic term affected by day and night alternation, and a random term caused by accidental factors. Secondly, fuse the knowledge base and strong association rules, and establish an analysis and warning module through a long short-term memory network. When it is monitored that the drilling fluid performance parameters deviate from the normal mode or it is predicted that the performance will be abnormal, a warning signal is sent in time.

2. The drilling fluid performance analysis and early warning method integrating a knowledge base and pattern mining according to claim 1, wherein In step S20, the knowledge base module is defined as: Among them, represents the knowledge base, which is stored in the form of a graph structure, is a set of nodes, including nodes of drilling fluid performance parameters, geological information nodes, and engineering data nodes, is a set of edges, representing the association relationships between nodes, is a node attribute function, defined as , indicating that each node is mapped to the attribute set , in tuple form, is an edge type function, defined as , indicating that each edge is mapped to the predefined semantic type set , including adaptation, causality, temporal dependence, and composite relationships, is the edge weight, and the initial value is determined based on the correlation analysis of historical data and the scoring of domain experts, defined as , is the Pearson correlation coefficient between nodes and , is the expert score, and the value range is [0, 1], is the balance factor, determined by domain experts, and through the graph structure, the effective organization and structured storage of drilling fluid-related knowledge are realized.

3. The method for analyzing and warning the performance of drilling fluid by integrating a knowledge base and pattern mining according to claim 1, characterized in that, In step S30, the sequence pattern mining module is defined as: Among them, represents the sequence pattern mining module, is the association rule representation, is the set of pre - parameters for triggering the association rule, is the set of post - parameters associated with the set where , , and , represents the empty set, represents that the set and the set have no common elements, is the pattern mining method of prefix projection, and the output is candidate sequences, is the transaction database, which is generated from multivariate time - series data through a sliding window, is the minimum support threshold, is the minimum confidence threshold, is the weight, and the support calculation is defined as , and the transaction is a subset of the transaction database , representing a sequence of drilling fluid performance parameters, that is , and the confidence calculation is defined as and , represents the proportion that the transaction database simultaneously contains the set and the set , represents the proportion that the transaction database contains the set . In the process of screening out high - weight parameter rules from candidate sequences, the confidence is updated, and the update method is defined as , and the weight of the knowledge base is iteratively optimized, and the optimization method is defined as , represents taking the highest weight of the edge set related to the rule in the knowledge base module in step S20, where and are the knowledge base nodes corresponding to the sets and the set respectively.

4. The method for analyzing and warning the performance of drilling fluid by integrating knowledge base and pattern mining according to claim 1, wherein In step S40, the analysis and warning module is defined as: Among them, represents the analysis and early warning module, is the early warning function, is a prediction function based on the long short-term memory network, refers to using the Holt-Winters triple exponential smoothing method of time series decomposition technology to decompose into a trend term , a periodic term and a random term , is a multi-parameter statistical feature set, calculated from the transaction database , and includes the mean, standard deviation, trend slope, maximum value and minimum value. is the knowledge base constructed in fusion step S20, is the strong association rule mined in fusion step S30, is the early warning threshold, defined by expert experience; the trend term , the periodic term , the random term , is the current moment, is the previous moment, is the fixed cycle length, is the level term at the moment, defined as , are the smoothing coefficients of the level term, trend term, and periodic term respectively, is the level term at the moment, is the trend term at the moment, is the periodic term at the moment, is the actual observed value at the moment; the prediction function uses the trend term, periodic term, and random term as the input of the long short-term memory network, extracts the relevant nodes and weights of the transaction from the knowledge base as the auxiliary input, and then combines the strong association rule to adjust the predicted value to obtain the predicted drilling fluid performance value ; the judgment criterion of the early warning function 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