Multi-index competitive underwater system lifetime prediction method and system

Abstract

This invention relates to the field of underwater production system life prediction technology, specifically disclosing a multi-index competitive method and system for underwater system life prediction. By identifying the main failure factors, normalizing multi-source monitoring indicators, determining weight coefficients based on Bayesian network analysis, constructing a real-time multi-index competition mechanism to dynamically screen key indicators, establishing a performance degradation prediction model, and introducing an objective function to correct model parameters, finally calculating the remaining service life based on the failure threshold. This invention overcomes the limitations of single-index prediction, achieves accurate characterization of the degradation process under the coupled effects of multiple factors, significantly improves the accuracy and robustness of life prediction, and provides a reliable basis for underwater system maintenance decisions.

Description

Technical Field

[0001] This invention relates to the field of deepwater oil and gas field development technology, and proposes a method and system for predicting the lifetime of underwater systems with multiple competing indicators. Background Technology

[0002] In deep-sea oil and gas field development, subsea production systems, as core equipment for oil and gas transportation, distribution, and control, operate for extended periods in complex environments characterized by high pressure, high temperature, and high corrosion. Key components (such as wellhouse trees and valves) are prone to performance degradation and even failure due to structural fatigue, wear, and corrosion, severely impacting development efficiency and operational safety. Currently, most methods for predicting remaining service life rely on independent modeling and analysis of single monitoring indicators (such as pressure, temperature, or vibration data), making it difficult to comprehensively reflect the equipment degradation process under the coupled effects of multiple factors. Due to the uncertainty of the deep-sea environment, the diversity of degradation mechanisms, and the nonlinear correlations between indicators, these methods often fail to dynamically identify the importance of each indicator under different operating conditions, lack a unified multi-indicator competition mechanism, resulting in insufficient prediction accuracy, poor robustness, and difficulty in adapting to the actual needs of multi-stage and multi-condition operations.

[0003] Therefore, existing technologies have not effectively solved the problems of dynamic weight allocation and real-time competition driven by multi-source data, which limits the accuracy and reliability of remaining lifetime prediction.

[0004] Chinese patent CN113432857A discloses a "Method and System for Predicting the Remaining Service Life of an Underwater Production Tree System Based on Digital Twins." This scheme establishes a digital twin model, constructs a dynamic Bayesian network degradation model by combining multiple index factors (such as temperature, pressure, and flow rate), and uses physical entity data acquisition and model calibration to achieve service life prediction. Its main technical features include: identifying the main failure modes (such as corrosion), establishing multi-index static and dynamic Bayesian networks, performing model calibration through interaction between the digital twin and physical entity data, and determining the failure threshold based on the limit state method. However, this scheme still has shortcomings in multi-index processing: firstly, the determination of index weight coefficients relies on static sensitivity analysis, lacking a real-time dynamic adjustment mechanism, making it difficult to adapt to changes in operating conditions; secondly, it does not introduce a multi-index competition mechanism, making it impossible to dynamically select key indicators based on real-time data, leading to reduced prediction efficiency when indicators are redundant or conflicting; thirdly, model calibration is mainly based on historical data and expert experience, limiting its responsiveness to sudden changes in operating conditions, affecting the real-time performance and accuracy of the prediction.

[0005] Another Chinese patent, CN118536433A, discloses a "Method and System for Intelligent Operation and Maintenance and Production Optimization of Subsea Oil Production Systems." This solution focuses on system health status assessment and operation and maintenance optimization. It calculates static health indices using data from surface sensors and subsea production tree sensors, and estimates dynamic health indices using a Kalman filter algorithm. The optimal maintenance strategy is then formulated with the goals of maximizing resilience and minimizing maintenance costs. Its main technical features include leak detection and location, health index calculation, multi-objective optimization function construction, and solving using a genetic-particle swarm optimization algorithm. However, this solution emphasizes operation and maintenance decision-making and production optimization rather than real-time prediction of remaining lifetime: First, although the health index calculation involves multiple indicators, a real-time competition mechanism between these indicators is not established, making it impossible to dynamically identify the dominant degradation factors under different operating conditions. Second, as an auxiliary link in operation and maintenance decision-making, remaining lifetime prediction lacks a dedicated multi-indicator dynamic modeling and correction mechanism, resulting in insufficient adaptability of the prediction results in complex degradation scenarios. Third, the solution does not fully consider the nonlinear degradation process under the coupling effect of multiple indicators, limiting the accuracy and robustness of the prediction model in variable environments.

[0006] In summary, the existing technologies and the aforementioned patent documents all share the following shortcomings: First, the multi-index processing mechanism is relatively static, unable to achieve real-time dynamic allocation and competitive selection of index weights, making it difficult to accurately capture key degradation factors; second, model correction and optimization rely heavily on historical data or fixed parameters, lacking the ability to adapt to real-time changes in operating conditions; finally, the existing methods need to improve prediction accuracy and robustness when dealing with complex scenarios such as multi-factor coupling and nonlinear degradation.

[0007] Therefore, there is an urgent need for a new method for predicting remaining lifetime that can integrate multi-source data, dynamically identify the contribution of indicators, and introduce a real-time competition mechanism, so as to comprehensively improve the accuracy and reliability of subsea production system condition monitoring and maintenance decision-making, and provide technical support for the safe and efficient development of deep-sea oil and gas fields. Summary of the Invention

[0008] The purpose of this invention is to address the shortcomings of existing technologies by proposing a multi-index competition method and system for predicting the remaining service life of underwater systems, and to solve the following technical problems: Current methods for predicting the remaining service life of underwater production systems are mostly based on single monitoring index modeling, which makes it difficult to comprehensively characterize the nonlinear degradation process under the coupling effect of multiple factors, and lacks dynamic weight allocation and real-time competition mechanisms, resulting in low prediction accuracy and insufficient robustness, and failing to meet the real-time evaluation requirements driven by multi-source data under complex operating conditions.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: a multi-index competitive underwater system lifetime prediction method, comprising the following steps: a multi-index real-time competitive prediction model construction step, wherein a multi-index real-time competitive prediction model is constructed based on multiple performance indicators of the underwater production system, wherein the construction includes normalizing the multiple performance indicators, setting a competition threshold, and constructing a Bayesian network degradation model based on indicators exceeding the competition threshold; a model calibration step, wherein an objective function is constructed and key parameters are selected and calibrated based on sensitivity analysis to optimize the multi-index real-time competitive prediction model; and a remaining lifetime calculation step, wherein a failure threshold of the underwater production system is determined, and the time from the current moment to reaching the failure threshold is calculated based on the multi-index real-time competitive prediction model as the remaining lifetime.

[0010] Preferably, the multi-index real-time competitive prediction model construction step includes determining the main failure factors of the underwater production system.

[0011] Preferably, the multi-indicator real-time competitive prediction model construction step includes normalizing the multiple performance indicators, wherein the normalization process is based on the fluctuation range matrix of the indicators.

[0012] Preferably, the multi-indicator real-time competitive prediction model construction step includes analyzing the sensitivity of each indicator based on a Bayesian network degradation model, and determining the weight coefficients based on the sensitivity.

[0013] Preferably, the multi-indicator real-time competitive prediction model construction step includes setting a competition threshold. When a normalized indicator exceeds the competition threshold, the indicator is used to construct the prediction model.

[0014] Preferably, the multi-index real-time competitive prediction model construction step includes constructing a performance degradation state model, wherein the performance degradation state is calculated based on a Bayesian network degradation model of an index that exceeds a competitive threshold.

[0015] Preferably, the model calibration step includes constructing an objective function to measure the deviation between the model output and the actual observations, and selecting key parameters for calibration through sensitivity analysis.

[0016] Preferably, the remaining useful life calculation step includes determining the failure threshold using the limit state method and calculating the time from the current moment until the performance degradation reaches the failure threshold as the remaining useful life.

[0017] Preferably, the multi-index real-time competitive prediction model construction step, the model calibration step, and the remaining service life calculation step are performed by a subsystem installed on the water control module.

[0018] It also includes a multi-index competitive underwater system lifetime prediction system, comprising: a multi-index real-time competitive prediction model construction subsystem, installed on the surface control module, for constructing a multi-index real-time competitive prediction model based on multiple performance indicators of the underwater production system; a model calibration subsystem, installed on the surface control module, for constructing an objective function and calibrating key parameters based on sensitivity analysis to optimize the prediction model; and a remaining service life calculation subsystem, installed on the surface control module, for determining the failure threshold of the underwater production system and calculating the remaining service life.

[0019] Preferably, the multi-indicator real-time competitive prediction model construction subsystem includes a module for determining the main failure factors of the underwater production system, a module for index normalization processing, a module for determining weight coefficients, a multi-indicator real-time competition module, and a prediction model construction module.

[0020] Preferably, the model calibration subsystem includes an objective function construction module and a key parameter selection and calibration module.

[0021] Preferably, the remaining service life calculation subsystem includes an underwater production system failure threshold determination module and a threshold time acquisition module.

[0022] The beneficial effects of this invention are as follows: This invention dynamically selects key degradation indicators through a multi-indicator real-time competition mechanism, integrates multi-source data to construct a Bayesian network degradation model, and realizes dynamic modeling and real-time accurate prediction of equipment degradation process; it introduces a model correction mechanism, dynamically corrects key parameters based on objective function optimization and sensitivity analysis, and significantly improves the model's adaptability under multiple operating conditions and stages; through a systematic failure threshold identification and life estimation strategy, it enhances the reliability of remaining life assessment and effectively overcomes the limitations of traditional single-indicator evaluation. Attached Figure Description

[0023] Figure 1 It is a Bayesian network degradation model under the influence of multiple index factors; Figure 2 This is a schematic diagram of an underwater production system; Figure 3 This is a schematic diagram of an underwater system lifetime prediction system with multiple competing indicators; Figure 4 This is a schematic diagram of the corrosion depth prediction results; Figure 5 A schematic diagram showing the probability distribution of corrosion depth for different indicators; Figure 6 This is a schematic diagram illustrating the real-time competition of multiple indicators for corrosion depth at the initial moment. Figure 7 This is a diagram illustrating the competitive trends of multiple indicators. Figure 8This is a schematic diagram of the numerical results of four methods (traditional monitoring model, physical model, cumulative model and the present invention) on the same test set.

[0024] In the diagram, 101 is the water control module, 102 is the uninterruptible power supply, 103 is the main control station, 104 is the power unit, 105 is the communication unit, 106 is the hydraulic power unit, 107 is the first hydraulic module of the hydraulic power unit, 108 is the second hydraulic module of the hydraulic power unit, 109 is the electronic control module of the hydraulic power unit, 110 is the third hydraulic module of the hydraulic power unit, 111 is the fourth hydraulic module of the hydraulic power unit, 112 is the electric power unit, and 113 is the second communication control unit. 114. Demodulator / Demodulator, 115. First Communication Modem / Modem, 116. Second Filter, 117. First Filter, 118. Second Power Coupler, 119. First Power Coupler, 120. Underwater Control Module, 121. Underwater Control Module Reversing Valve Assembly, 122. First Reversing Valve, 123. Third Reversing Valve, 124. Fourth Reversing Valve, 125. Fifth Reversing Valve, 126. Underwater Control Module Electronic Module Assembly, 127. Third Underwater Electronic Module Assembly Submodules: 128. Second Underwater Electronic Module; 129. First Underwater Electronic Module; 130. Underwater Christmas Tree; 131. Underwater Christmas Tree Hydraulic Valve Assembly; 132. First Hydraulic Valve; 133. Second Hydraulic Valve; 134. Third Hydraulic Valve; 135. Fourth Hydraulic Valve; 136. Fifth Hydraulic Valve; 137. Underwater Christmas Tree Mechanical Components Assembly; 138. Christmas Tree Cap; 139. Christmas Tree Body; 201. Multi-Indicator Real-Time Competitive Prediction Model Construction Subsystem; 202. The following modules are included: 203. Subsystem for determining the main failure factors of the underwater production system; 204. Subsystem for normalizing indicators; 205. Subsystem for determining weight coefficients; 206. Subsystem for real-time competition of multiple indicators; 207. Subsystem for building a prediction model; 208. Subsystem for model calibration; 209. Subsystem for building an objective function; 210. Subsystem for selecting and calibrating key parameters; 211. Subsystem for calculating remaining useful life; 212. Subsystem for determining the failure threshold of the underwater production system; 213. Subsystem for obtaining the time to reach the threshold. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0026] This invention provides a method and system for predicting the remaining lifetime of underwater systems using a multi-indicator competitive model. It aims to address the problem that existing prediction methods based on a single monitoring indicator are insufficient to comprehensively depict the degradation process under the coupled effects of multiple factors. By integrating multiple key performance indicators, a real-time competitive mechanism is constructed, and model calibration is introduced to achieve more accurate and robust remaining lifetime prediction. The method of this invention mainly includes three steps: constructing a real-time competitive prediction model based on multiple indicators, model calibration, and calculating the remaining lifetime. The system includes a subsystem for constructing a real-time competitive prediction model based on multiple indicators, a model calibration subsystem, and a remaining lifetime calculation subsystem. These subsystems are installed in the surface control module 101. The specific embodiments of this invention will be described in detail below with reference to the accompanying drawings. Example

[0027] This embodiment uses a typical subsea production system as an example to describe in detail the implementation of the present invention. The subsea production system includes a surface control module 101, a hydraulic power unit 106, an electric power unit 112, a subsea control module 119, and a subsea wellhead 130, as shown in the attached diagram. Figure 2 As shown. The system is in a complex environment of high pressure, high temperature and high corrosion for a long time. Key components (such as the tree trunk 139, hydraulic valves, etc.) are prone to degradation such as structural fatigue, wear and corrosion.

[0028] The system structure includes: a surface control module 101 located within the surface control platform, comprising an uninterruptible power supply 102, a main control station 103, a power unit 104, and a communication unit 105. The uninterruptible power supply 102, power unit 104, and communication unit 105 are connected to the main control station 103 via cables for power and communication transmission to the main control station 103. A hydraulic power unit 106 located within the surface control platform includes a first hydraulic module 107, a second hydraulic module 108, an electrical control module 109, a third hydraulic module 110, and a fourth hydraulic module 111. The electrical control module 109 is connected to each hydraulic module via cables for controlling the operating status of the hydraulic modules. An electrical power unit 112 located within the surface control platform includes a first communication modem 114, a second communication modem 113, a first filter 116, a second filter 115, a first power coupler 118, and a second power coupler 117. The first communication modem 114 is connected to the first filter 116 via a cable to filter out unwanted frequency electrical signals. The first communication modem 114 and the first filter 116 are connected to the first power coupler 118 via a cable for coupling electrical signals and communication information. Similarly, the second communication modem 113 is connected to the second filter 115 via a cable, and the second communication modem 113 and the second filter 115 are connected to the second power coupler 117 via a cable. The underwater control module 119 is located within the underwater control compartment and includes an underwater control module reversing valve assembly 120 and an underwater control module electronic module assembly 126. The underwater control module reversing valve assembly 120 includes a first reversing valve 121, a second reversing valve 122, a third reversing valve 123, a fourth reversing valve 124, and a fifth reversing valve 125. The underwater control module electronic module assembly 126 includes a first underwater electronic module 129, a second underwater electronic module 128, and a third underwater electronic module 127. The subsea production tree 130 is located in the subsea oilfield and includes a subsea production tree hydraulic valve assembly 131 and a subsea production tree mechanical assembly 137. The subsea production tree hydraulic valve assembly 131 includes a first hydraulic valve 132, a second hydraulic valve 133, a third hydraulic valve 134, a fourth hydraulic valve 135, and a fifth hydraulic valve 136. The subsea production tree mechanical assembly 137 includes a production tree cap 138 and a production tree body 139.

[0029] The components are connected by cables: the uninterruptible power supply 102 is connected to the hydraulic power unit 106 and the electric power unit 112 via cables, providing the power required for hydraulic and signal transmission; the main control station 103 is connected to the hydraulic power unit 106 and the electric power unit 112 via cables, controlling the hydraulic and signal transmission in oil and gas production; the first hydraulic module 107, the second hydraulic module 108, the third hydraulic module 110, and the fourth hydraulic module 111 of the hydraulic power unit are connected by cables. The cable is connected to the first reversing valve 121, the second reversing valve 122, the third reversing valve 123, the fourth reversing valve 124, and the fifth reversing valve 125 for the transmission of hydraulic fluid; the first power coupler 118 and the second power coupler 117 are connected to the first underwater electronic module 129, the second underwater electronic module 128, and the third underwater electronic module 127 via cables for the transmission of control signals; the underwater control module electronic module group 126 is connected to the underwater control module reversing valve group 120 via cables for controlling the flow direction of the hydraulic circuit.

[0030] The method of the present invention is executed by a subsystem installed in the water control module 101, as shown in the appendix. Figure 3 As shown. The multi-indicator real-time competitive prediction model construction subsystem 201 includes a module 202 for determining the main failure factors of the underwater production system, a module 203 for index normalization processing, a module 204 for determining weight coefficients, a multi-indicator real-time competition module 205, and a prediction model construction module 206. The model calibration subsystem 207 includes a module 208 for constructing the objective function and a module 209 for selecting and calibrating key parameters. The remaining service life calculation subsystem 210 includes a module 211 for determining the failure threshold of the underwater production system and a module 212 for obtaining the time to reach the threshold.

[0031] S101: Identify the main failure factors of the underwater production system: The subsea production system's main failure factor identification module 202 is connected to the main control station 103 via cables, collecting degradation information of each component of the subsea production system. Based on historical fault data, expert knowledge, and on-site monitoring information, it identifies multiple failure factors that significantly impact system performance degradation. These factors reflect multiple degradation dimensions, including structural integrity, motion performance, and environmental adaptability. For example, for the subsea production tree 130, key failure factors include the corrosion rate of the production tree body 139 (unit: mm / year), the sealing leakage rate of hydraulic valves (unit: mL / min), and the valve response time (unit: ms). Specifically, the corrosion rate is monitored by a corrosion sensor, the leakage rate is monitored by a flow sensor, and the response time is calculated by the control signal feedback time. These indicators are collected and transmitted to the main control station 103 via subsea electronic modules (127, 128, 129).

[0032] S102: Indicator Normalization Process The index normalization module 203 is connected to the underwater production system main failure factor determination module 202 via a cable, and performs normalization processing on each index to eliminate the influence of dimensions. Let the index set be X=[x1,x2,...,x...]. n ], where xᵢ represents the i-th indicator. Each indicator has a fluctuation range matrix W, where the element x_{mn} represents the m-th fluctuation value of the n-th indicator. The normalization formula is: X mn '=X mn / max(X mn ); where x_{mn}' is the normalized value. After normalization, the element values ​​of matrix W' are between 0 and 1, making different indicators comparable. For example, if the original value of the corrosion rate is 0.1 mm / year and the maximum allowable value is 0.5 mm / year, then the normalized value is 0.2; if the original value of the leakage rate is 5 mL / min and the maximum allowable value is 20 mL / min, then the normalized value is 0.25.

[0033] S103: Determination of Weighting Coefficients The weight coefficient determination module 204 is connected to the index normalization processing module 203 via a cable, based on the Bayesian network degradation model (as shown in the attached diagram). Figure 1 (As shown) Analyze the sensitivity of each index to the corrosion degradation results. Sensitivity set Y=[y1,y2,...,y... n ], where yᵢ represents the sensitivity of degradation to the i-th indicator, obtained by training a Bayesian network using historical data. The weight coefficients yᵢ' are calculated as follows: yᵢ'=yᵢ / a Where 'a' is the maximum value of set Y. For example, if the sensitivity set is [0.8, 0.6, 0.9], then a = 0.9, and the weighting coefficients are 0.89, 0.67, and 1.0, respectively. The weighting coefficients reflect the relative importance of each indicator in the overall degradation and are used for subsequent multi-indicator competition.

[0034] S104: Real-time Competition of Multiple Indicators The multi-indicator real-time competition module 205 is connected to the weight coefficient determination module 204 via a cable, and sets the competition threshold θ for each indicator. i θ i Based on historical data or expert experience, for example, a value of 0.7 is used for the normalized indicator X. i ' If X i ' ≥θ iIf the index passes the competition, its Bayesian network degradation model will be used for prediction; otherwise, the index is ignored. For example, if the normalized corrosion rate is 0.8 (≥0.7), it passes the competition; if the normalized leakage rate is 0.6 (<0.7), it does not pass the competition.

[0035] S105: Predictive Model Construction The prediction model building module 206 is connected to the multi-index real-time competition module 205 via a cable to construct the performance degradation state V: ;in, f(x i ) It is a Bayesian network degradation model for the i-th indicator, trained using historical data, capable of predicting the future degradation trend of that indicator; i It is an indicator function, when X i '≥θ i At that time, l i =1, otherwise l i =0; For example, if only corrosion rate and response time compete, then V=f(corrosion rate)+f(response time). The Bayesian network degradation model uses a dynamic Bayesian network (DBN) to process time series data, and its conditional probability table is trained based on maximum likelihood estimation.

[0036] Model calibration steps: S201: Constructing the objective function The objective function construction module 208 is connected to the prediction model construction module 206 and the main control station 103 via cables, and constructs the objective function H'=g(H,S) to measure the deviation between the model output and the actual observations. The objective function uses the mean squared error (MSE). ;in, V j This is the performance degradation state predicted by the model. These are actual observed values ​​(calculated from sensor data). N That is the number of samples.

[0037] S202: Select and calibrate key parameters The key parameter selection and calibration module 209 is connected to the objective function construction module 208 via a cable. Based on sensitivity analysis, it selects key parameters that significantly affect the model output. Let the parameter set be P (e.g., the conditional probability of a Bayesian network), and the sensitivity S be calculated as follows: Where Y represents the state after the parameter changes. Y 0P is the baseline state, P is the changed parameter, P0 is the baseline parameter. A positive sensitivity indicates that the variable and parameter change in the same direction, while a negative sensitivity indicates the opposite direction. Based on the sensitivity results, the gradient descent method is used to adjust the parameter to minimize the objective function H'. For example, if the sensitivity analysis shows that the conditional probability of corrosion rate has the greatest impact on the output, then this parameter should be adjusted first.

[0038] Remaining useful life calculation steps: S301: Determine the failure threshold for underwater production systems The underwater production system failure threshold determination module 211 is connected to the key parameter selection and calibration module 209 via a cable, and determines the failure threshold A based on the limit state method. th When the system performance degradation A ≥ A th At that time, the system failed, A th According to system design specifications or safety standards, for example, regarding the overall performance degradation, A th Setting it to 0.8 (normalized value) corresponds to a system performance drop to 80% of the initial value.

[0039] S302: Obtain the time when the threshold is reached The threshold acquisition module 212 is connected to the underwater production system failure threshold determination module 211 via a cable, and calculates the time from the current moment until the performance degradation A reaches A. th The remaining useful life, T: T = inf{t: A(t) ≥ A th}; where A(t) is the performance degradation at time t, obtained through the prediction model. For example, if the current A=0.4, and the prediction model shows that A increases by 0.1 per year, then T=(0.8-0.4) / 0.1=4 years.

[0040] This embodiment improves the accuracy and robustness of the prediction model by dynamically selecting important indicators through a multi-indicator real-time competition mechanism. The model calibration mechanism ensures the adaptability of the model under different working conditions. The remaining service life calculation is based on the failure threshold, providing a scientific basis for maintenance decisions. For example, in the deep-sea high-pressure environment, this method can effectively identify key indicators such as corrosion and leakage, avoiding the limitations of single-indicator prediction.

[0041] Example 2: A multi-index competitive underwater system lifetime prediction method comprises three steps: multi-index real-time competitive prediction model construction, model calibration, and remaining lifetime calculation.

[0042] The specific steps for constructing a multi-index real-time competitive prediction model are as follows: S101: Determine the main failure factors of the subsea production system. Based on the typical failure modes of key equipment such as subsea wellheads during long-term service, and combined with historical failure data, expert knowledge and on-site monitoring information, identify multiple failure factors that have a significant impact on system performance degradation, reflecting multiple degradation dimensions such as structural integrity, motion performance, and environmental adaptability.

[0043] S102: Indicator normalization processing, setting the fluctuation range for each indicator, and presetting the maximum allowable corrosion rate, [x] m1 , ..., x mn [ ] represents the extreme values ​​of different indices under the maximum allowable degradation rate.

[0044] Where X is the set of indicators, X n Represents different indicators; W is the fluctuation range matrix of each indicator, x mn This represents the m-th fluctuation value of the n-th indicator, and W is normalized.

[0045] Where W' is the normalized fluctuation range matrix of each indicator, x mn ' represents the m-th fluctuation value of the n-th indicator after normalization, [x m1 ',...,x mn Each element is 1.

[0046] S103: Determining weighting coefficients; such as Figure 1 As shown, based on the Bayesian network degradation model under the influence of multiple index factors, the sensitivity of each index to the corrosion degradation result is analyzed, and the weight coefficients of each index are determined according to the sensitivity analysis results.

[0047] Where Y is the sensitivity set, y n This indicates the sensitivity of degradation to different indicators, y n ' represents the weight of different indicators, and a is the maximum value of set Y.

[0048] S104: Real-time competition among multiple indicators; based on the data collection results of each indicator, normalization processing is performed on each indicator, and competition thresholds for different indicators are set: Th={T h1 ,T h2 ,K,T hn When a normalized index exceeds the competition threshold, the index passes the competition. In a prediction model with real-time competition among multiple indices, the Bayesian network degradation model of that index will be effectively used.

[0049] S105: Predictive model construction: Based on the collected data, establish a predictive model with real-time competition among multiple indicators, and implement a competition method based on multiple indicators to filter out indicators that exceed their respective thresholds in real time.

[0050] Performance degradation state V: ;where f(X) i Let be the Bayesian network degradation model for the i-th index. When the normalized result of this index exceeds the competition threshold, l i If it is 1, then I i It is 0.

[0051] The specific steps of model calibration are as follows: S201: Construct the objective function and define the objective function of model calibration to measure the degree of deviation between the model output and the actual physical observation. The principle of constructing the objective function is to make the output of the model after parameter calibration as close as possible to the real physical result. Model calibration can ensure the accuracy of the model and better adapt to different application needs, conditions and scenarios.

[0052] S202: Select and calibrate key parameters; based on sensitivity analysis, determine the key parameters that significantly affect the model output. Let the objective function be H´=g(H, S), select the parameter H to be calibrated, and calculate the corresponding sensitivity S: In the formula, S is the sensitivity, Y is the state after the change, Y0 is the state before the change, P is the model parameter after the change, and P0 is the baseline model parameter. A positive sensitivity indicates that the variable and the parameter change in the same direction, that is, the increase or decrease of the model parameter causes the variable to increase or decrease as well. A negative sensitivity indicates that the variable and the parameter change in opposite directions.

[0053] The specific steps for calculating the remaining useful life are as follows: S301: Determine the failure threshold of the underwater production system. When the predicted object degrades for a period of time under internal factors and external influences, its performance gradually degrades. The failure threshold is determined by the limit state method. When the performance is lower than the failure threshold, the system will not be able to complete normal operation.

[0054] S302: Obtain the time to reach the threshold. Based on the overall performance of the underwater production system, calculate the time period from the detection point to the failure point to obtain the remaining service life of the predicted object. The remaining service life is determined by the failure threshold Ath, i.e., the critical value. Let the performance degradation of the underwater production system be A, that is, when A≥Ath, the system fails. That is, the remaining service life T of the underwater production system is defined as: T=inf{t:A(t)≥Ath} th |A(0)>A th};like Figure 2As shown, the underwater production system includes a surface control module 101, a hydraulic power unit 106, an electric power unit 112, an underwater control module 119, and an underwater wellhead 130. The surface control module 101, located within the surface control platform, includes an uninterruptible power supply (UPS) 102, a main control station 103, an electric power unit 104, and a communication unit 105. The UPS 102, electric power unit 104, and communication unit 105 are connected to the main control station 103 via cables for power and communication transmission. The hydraulic power unit 106, also located within the surface control platform, includes a first hydraulic module 107, a second hydraulic module 108, and an electric control module. Block 109, the third hydraulic module 110 and the fourth hydraulic module 111 of the hydraulic power unit; the hydraulic power unit electrical control module 109 is connected to the first hydraulic module 107, the second hydraulic module 108, the third hydraulic module 110 and the fourth hydraulic module 111 of the hydraulic power unit via cables, and is used to control the four hydraulic modules; the electric power unit 112 is located in the water control platform and includes: a first communication modem 114, a second communication modem 113, a first filter 116, a second filter 115, a first power coupler 118 and a second power coupler 117; the first communication modem 114 is connected to the first hydraulic module 107, the second hydraulic module 108, the third hydraulic module 110 and the fourth hydraulic module 111 of the hydraulic power unit via cables, and is used to control the four hydraulic modules; A cable is connected to the first filter 116 to filter out unwanted frequency electrical signals; the first communication modem 114 and the first filter 116 are connected to the first power coupler 118 via a cable for coupling electrical signals and communication information; the second communication modem 113 is connected to the second filter 115 via a cable for filtering out unwanted frequency electrical signals; the second communication modem 113 and the second filter 115 are connected to the second power coupler 117 via a cable for coupling electrical signals and communication information; the underwater control module 119 is located in the underwater control compartment and includes: an underwater control module reversing valve assembly 120 and an underwater control module electronic module assembly 126; the underwater control module reversing valve assembly 120 includes... Includes: a first reversing valve 121, a second reversing valve 122, a third reversing valve 123, a fourth reversing valve 124, and a fifth reversing valve 125; an underwater control module electronic module group 126 includes: a first underwater electronic module 129, a second underwater electronic module 128, and a third underwater electronic module 127; an underwater production tree 130 located in an underwater oilfield includes: an underwater production tree hydraulic valve group 131 and an underwater production tree mechanical part group 137; the underwater production tree hydraulic valve group 131 includes: a first hydraulic valve 132, a second hydraulic valve 133, a third hydraulic valve 134, a fourth hydraulic valve 135, and a fifth hydraulic valve 136; the underwater production tree mechanical part group 137 includes: a production tree cap 138 and a production tree body 139;An uninterruptible power supply 102 is connected to the hydraulic power unit 106 and the electric power unit 112 via cables to provide the power required for hydraulic and signal transmission. A main control station 103 is connected to the hydraulic power unit 106 and the electric power unit 112 via cables to control hydraulic and signal transmission in oil and gas production. The first hydraulic module 107, the second hydraulic module 108, the third hydraulic module 110, and the fourth hydraulic module 111 of the hydraulic power unit are connected to the first directional valve 121, the second directional valve 122, the third directional valve 123, the fourth directional valve 124, and the fifth directional valve 125 via cables for hydraulic fluid transmission. The first electric coupler 118 and the second electric coupler 117 are connected to the first underwater electronic module 129, the second underwater electronic module 128, and the third underwater electronic module 127 via cables for control signal transmission. The underwater control module electronic module group 126 is connected to the underwater control module directional valve group 120 via cables for controlling the flow direction of the hydraulic circuit. ;

[0055] like Figure 3 As shown, the underwater system lifetime prediction system with multi-index competition includes a multi-index real-time competitive prediction model construction subsystem 201 installed on the surface control module 101, a model correction subsystem 207 installed on the surface control module 101, and a remaining lifetime calculation subsystem 210 installed on the surface control module 101.

[0056] The multi-index real-time competitive prediction model construction subsystem 201 includes a main failure factor determination module 202 for the underwater production system, an index normalization processing module 203, a weight coefficient determination module 204, a multi-index real-time competition module 205, and a prediction model construction module 206. The main failure factor determination module 202 is connected to the main control station 103 via a cable and is used to collect degradation information of each component of the underwater production system. The index normalization processing module 203 is connected to the main failure factor determination module 202 via a cable and is used to normalize the degradation indices of the underwater production system. The weight coefficient determination module 204 is connected to the index normalization processing module 203 via a cable and is used to determine the weight coefficients of each degradation index. The weight coefficient determination module 204 is connected to the multi-index real-time competition module 205 via a cable and is used to determine whether each degradation index of the underwater production system contributes to the overall degradation. The prediction model construction module 206 is connected to the multi-index real-time competition module 205 via a cable and is used to construct a performance prediction model for the underwater production system.

[0057] The model calibration subsystem 207 includes an objective function construction module 208 and a key parameter selection and calibration module 209. The objective function construction module 208 is connected to the prediction model construction module 206 and the main control station 103 via cables, and is used to construct the objective function for parameter calibration of the underwater production system degradation model. The key parameter selection and calibration module 209 is connected to the objective function construction module 208 via cables, and is used for dynamic updating and calibration of degradation parameters.

[0058] The remaining service life calculation subsystem 210 includes an underwater production system failure threshold determination module 211 and a threshold attainment time acquisition module 212. The underwater production system failure threshold determination module 211 is connected to the key parameter selection and correction module 209 via a cable and is used to acquire the failure threshold of the underwater production system. The threshold attainment time acquisition module 212 is connected to the underwater production system failure threshold determination module 211 via a cable and is used to calculate the remaining service life of the underwater production system.

[0059] The following is a description of the technical effects of this invention: Taking the indicators (failure factors) of competing failures as fluid flow rate, carbon dioxide partial pressure, external corrosion and H2S corrosion as examples.

[0060] The sensitivity analysis results of this invention are shown in the table below: Where Ti refers to the degree of influence of temperature on the H2S corrosion rate; pco2 refers to the influence of carbon dioxide partial pressure on the corrosion rate; vo refers to the influence of internal fluid flow rate on the corrosion rate; To refers to the influence of external ambient temperature on the external corrosion rate; and pdo refers to the influence of dissolved oxygen partial pressure in the external environment on the corrosion rate.

[0061] Figure 4 For corrosion depth prediction, the failure threshold is 20 mm. Considering the cumulative corrosion depth of each model, the degradation rate is significantly accelerated, reaching the failure threshold in the 12th year. In this invention, the corrosion of each model, from highest to lowest, is influenced by fluid velocity, carbon dioxide partial pressure, external corrosion, and H2S corrosion, exhibiting a basically linear change. In this invention, the H2S content is fixed, and its impact on overall corrosion mainly depends on temperature.

[0062] Figure 5The probability distributions of corrosion depth for different indices represent the probability distributions of corrosion depth under single-model degradation modeling. The probability distribution of H2S corrosion is read from the upper coordinate (H2S corrosion depth) and the right coordinate (H2S corrosion probability), while the probability distributions of other factors are read from the left coordinate (corrosion probability) and the lower coordinate (corrosion depth). The probability distributions of the effects of temperature and external corrosion both follow a normal distribution, with relatively small distribution areas for corrosion depth. The probability distribution of the effect of temperature on corrosion depth is more concentrated, and the corrosion depth of external corrosion is distributed in the range of 0.15~0.265, because the indices affecting both follow a normal distribution. Compared to the external environment, the internal environment is more complex, resulting in deeper corrosion. The effect of carbon dioxide partial pressure approximates a normal distribution centered at 0.45, with a distribution range of 0~1mm. At a corrosion depth of 0.92mm, the probability distribution of the effect of fluid velocity is slightly wider than that of the effect of external corrosion, increasing slowly from 0 to 0.8mm, but rapidly from 0.8 to 0.92mm. Therefore, in a single corrosion degradation model, the influence of internal corrosion is greater than that of external corrosion, and among the factors affecting internal corrosion, the influence of fluid velocity deserves special attention and control.

[0063] Figure 6 The corrosion depth is determined by real-time competition among multiple indicators at the initial moment. The corrosion depth at the initial moment is read from the right coordinate (corrosion depth at the initial moment) and the top coordinate (probability distribution at the initial moment), while the predicted corrosion depth is read from the left coordinate (corrosion depth) and the bottom coordinate (time).

[0064] Figure 7 To determine the multi-indicator competition trend, a real-time multi-indicator competition algorithm was adopted for each indicator (Ti, pco2, vo, To, pdo). Data for each indicator was monitored and extracted in real time at different stages, and normalization was performed to obtain the real-time competition trend. Blue represents the results of the first monitoring calculation, where all indicators are within the 0.3 range, with seawater flow velocity having a significant impact. Green represents the results of the second monitoring calculation, where the flow velocities of both internal and external media are significantly higher than other indicators, with the internal fluid flow velocity approaching its corresponding threshold, indicating that the real-time multi-indicator competition algorithm needs updating. Yellow represents the results of the third monitoring calculation, where the marine environmental temperature and internal fluid flow velocity exceed their corresponding thresholds, and the dissolved oxygen concentration in seawater changes significantly, but remains within a safe range.

[0065] Figure 8 Numerical results from four methods (traditional monitoring model, physical model, cumulative model, and the present invention) on the same test set are presented. In underwater production systems, a model that is accurate for one month can bring huge production benefits and reduce maintenance costs. This model can perform degradation prediction of multiple indicators and improve accuracy by more than one year, which is more advanced than existing technologies.

[0066] The technical advantages of this application are also reflected in the following aspects: This method can maintain stable prediction accuracy under complex operating conditions such as incomplete monitoring data, sensor drift, and noise enhancement, which is difficult to achieve simultaneously by traditional monitoring models, physical models, and cumulative models; The model correction mechanism of this application can update the degradation rate, probability distribution parameters, and uncertainty propagation law, so that the prediction can still maintain high stability when the operating conditions change in the later stage, while the comparison methods are mostly one-time modeling with fixed parameters, which is difficult to adapt to changes in operating conditions; This application provides the model construction process in the specification, including the meaning of key parameters, calculation methods, update conditions, and data sources, which can be adapted by those skilled in the art to model according to different scenarios and different equipment.

[0067] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting the lifetime of an underwater system through multi-index competition, characterized in that, Includes the following steps: The steps for constructing a multi-indicator real-time competitive prediction model are as follows: Based on multiple performance indicators of the underwater production system, a multi-indicator real-time competitive prediction model is constructed. The construction includes normalizing the multiple performance indicators, setting a competition threshold, and constructing a Bayesian network degradation model based on indicators that exceed the competition threshold. The model calibration step involves constructing an objective function and selecting and calibrating key parameters based on sensitivity analysis to optimize the multi-indicator real-time competitive prediction model. The remaining service life calculation step involves determining the failure threshold of the underwater production system and calculating the time from the current moment to reaching the failure threshold based on the multi-index real-time competitive prediction model, which is then used as the remaining service life.

2. The underwater system lifetime prediction method with multi-index competition according to claim 1, characterized in that, The steps for constructing the multi-index real-time competitive prediction model include identifying the main failure factors of the underwater production system.

3. The underwater system lifetime prediction method with multi-index competition according to claim 2, characterized in that, The steps for constructing the multi-indicator real-time competitive prediction model include normalizing the multiple performance indicators, wherein the normalization process is based on the fluctuation range matrix of the indicators.

4. The underwater system lifetime prediction method with multi-index competition according to claim 3, characterized in that, The steps for constructing the multi-indicator real-time competitive prediction model include analyzing the sensitivity of each indicator based on a Bayesian network degradation model and determining the weight coefficients based on the sensitivity.

5. The underwater system lifetime prediction method with multi-index competition according to claim 4, characterized in that, The steps for constructing the multi-indicator real-time competitive prediction model include setting a competition threshold. When a normalized indicator exceeds the competition threshold, that indicator is used to construct the prediction model.

6. The underwater system lifetime prediction method with multi-index competition according to claim 5, characterized in that, The steps for constructing the multi-index real-time competition prediction model include constructing a performance degradation state model, wherein the performance degradation state is calculated based on a Bayesian network degradation model of an index that exceeds a competition threshold.

7. The underwater system lifetime prediction method with multi-index competition according to claim 1, characterized in that, The model calibration steps include constructing an objective function to measure the deviation between the model output and the actual observations, and selecting key parameters for calibration through sensitivity analysis.

8. The underwater system lifetime prediction method with multi-index competition according to claim 1, characterized in that, The remaining useful life calculation step includes determining the failure threshold using the limit state method and calculating the time from the current moment until the performance degradation reaches the failure threshold as the remaining useful life.

9. The underwater system lifetime prediction method with multi-index competition according to claim 1, characterized in that, The steps of constructing the multi-index real-time competitive prediction model, the model calibration step, and the remaining service life calculation step are executed by the subsystem installed on the water control module.

10. A multi-index competitive underwater system lifetime prediction system, characterized in that, include: A multi-indicator real-time competitive prediction model construction subsystem is installed on the surface control module and is used to construct a multi-indicator real-time competitive prediction model based on multiple performance indicators of the underwater production system. The model calibration subsystem, installed on the surface control module, is used to construct the objective function and correct key parameters based on sensitivity analysis to optimize the prediction model; the remaining service life calculation subsystem, installed on the surface control module, is used to determine the failure threshold of the underwater production system and calculate the remaining service life.

11. The underwater system lifetime prediction system with multi-index competition according to claim 10, characterized in that, The multi-indicator real-time competitive prediction model construction subsystem includes a module for determining the main failure factors of underwater production systems, a module for index normalization processing, a module for determining weight coefficients, a multi-indicator real-time competition module, and a prediction model construction module.

12. The underwater system lifetime prediction system with multi-index competition according to claim 10, characterized in that, The model calibration subsystem includes an objective function construction module and a key parameter selection and calibration module.

13. The underwater system lifetime prediction system with multi-index competition according to claim 10, characterized in that, The remaining service life calculation subsystem includes a subsea production system failure threshold determination module and a threshold attainment time acquisition module.

Images