Multi-stage electric heating intelligent plugging device and method for complex freezing plugging of CO2 flooding waxy crude oil gathering pipeline

Abstract

The application relates to a multi-stage electric heating intelligent unblocking device and method for complex frozen blockage of a CO2 flooding waxy crude oil gathering pipeline, wherein the multi-stage electric heating intelligent unblocking device for complex frozen blockage of the CO2 flooding waxy crude oil gathering pipeline comprises a dredging system, an outer framework and an intelligent control system; the outer framework is a bionic flexible guide pipe which is composed of multiple body segment units connected through expansion joints; the dredging system integrates a first-stage global heating system, a second-stage local heating system and a dosing system; a flexible electric heating pipe is embedded in the bionic flexible guide pipe to form a uniform heating body; the bionic flexible guide pipe is provided with a temperature sensor; the second-stage local heating system is an intelligent composite drill bit which can autonomously heat; the intelligent composite drill bit comprises a motor system, a drill bit shell, an electric heating rod, a drill bit and a miniature night vision camera; the dosing system comprises an inhibitor storage tank and multiple nozzles. The application can quickly break the frozen blockage area and realize intelligent operation.

Description

Technical Field

[0001] This invention relates to the fields of oil and gas storage and transportation engineering and CO2-driven enhanced oil recovery (CCUS-EOR), specifically to an intelligent unblocking device and method for complex blockages formed by the coexistence of hydrates, ice and gelled waxy crude oil and silt in CO2-driven waxy crude oil gathering and transportation pipelines. It is particularly suitable for solving the blockage problem of CO2-driven waxy crude oil gathering and transportation pipelines under different conditions. Background Technology

[0002] CO2 enhanced oil recovery (EOR) technology has been widely applied in many oilfields, but the resulting pipeline blockage problem severely restricts production efficiency. The blockages formed in the pipelines are complex, resulting from the combined effects of hydrate freezing, wax and asphaltenes precipitation, and mechanical impurities. Under low temperature and high pressure, CO2 easily forms hydrate crystals, while the extraction of light components from crude oil by CO2 leads to the deposition of heavy organic matter. Combined with the adhesion of corrosion products and other impurities, this causes the pipe diameter to narrow or even completely blockage, resulting in increased injection pressure, decreased production, and even production interruptions and safety risks.

[0003] Existing technologies for unclogging CO2-driven waxy crude oil gathering and transportation pipelines have significant limitations. Traditional single-method unclogging approaches are insufficient to efficiently address complex blockages. While chemical unclogging agents have some effect, they face limitations such as high cost, environmental pressure, and limited specificity. Thermal methods are primarily preventative, lacking the ability to resolve severe existing blockages and consuming high energy. Traditional methods such as mechanical cleaning involve complex procedures, long cycles, high costs, and potential safety risks. Current technologies generally lack intelligence, failing to dynamically adjust unclogging strategies based on the composition and severity of the blockage. Therefore, the industry urgently needs an intelligent unclogging solution capable of adaptively handling complex blockages. Summary of the Invention

[0004] One objective of this invention is to provide a multi-stage electric heating intelligent unblocking device for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines. This multi-stage electric heating intelligent unblocking device is used to solve the problem that traditional single unblocking methods are difficult to efficiently deal with complex blockages. Another objective of this invention is to provide an unblocking method for the multi-stage electric heating intelligent unblocking device for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines.

[0005] The technical solution adopted by this invention to solve its technical problem is as follows: This multi-stage electric heating intelligent unblocking device for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines includes a dredging system, an outer frame, and an intelligent control system; the outer frame is a biomimetic flexible guide tube, composed of multiple segment units connected by expansion joints, which achieve forward movement, backward movement, and fixation through contraction and expansion at the segments; the dredging system integrates a primary global heating system, a secondary local heating system, and a chemical dosing system, with the secondary local heating system installed at the head of the primary global heating system; the heat source of the primary global heating system is a flexible electric heating tube, which is embedded in the biomimetic... A uniform heating element is formed in the flexible guide tube, and thermally conductive material is filled between the flexible electric heating tube and the bionic flexible guide tube. A temperature sensor is installed in the bionic flexible guide tube. The secondary local heating system is an intelligent composite drill bit that can generate its own heat. The intelligent composite drill bit includes a motor system, a drill bit shell, an electric heating rod, a drill bit, and a miniature night vision camera. The electric heating rod runs through the drill bit shell, and thermally conductive material is filled between the electric heating rod and the drill bit shell. The motor system is connected to the drill bit through a drive shaft, and a torque sensor is installed on the drive shaft. A temperature sensor is installed on the drill bit. The dosing system includes an inhibitor storage tank and multiple nozzles evenly distributed at the tail of the intelligent composite drill bit.

[0006] In the above scheme, the biomimetic flexible guide tube is a flexible metal explosion-proof hose. The flexible electric heating tube and heat-conducting material are directly embedded or tightly attached to the tube wall of the biomimetic flexible guide tube, forming a biomimetic flexible guiding and thermally enhanced structural tube. This makes the entire biomimetic flexible guide tube a uniform heat source while in motion, efficiently transferring its heat to the entire inner wall of the pipe, simultaneously moving, preheating, and melting ice. The intelligent composite drill bit performs rotary cutting and impact crushing on the frozen blockage while melting the frozen area. It also feeds back the blockage data to the intelligent control system through actual measurement, enabling the determination of the main components of the blockage and adjustment of the unblocking strategy. After unblocking, the dosing system releases hydrate inhibitors into the gathering and transportation pipeline to prevent the secondary formation of hydrate blockage. The intelligent control system adjusts the heating temperature, drill bit speed, and travel strategy in real time and has self-learning capabilities to optimize the unblocking process.

[0007] In the above scheme, each segment unit of the biomimetic flexible guide tube imitates a segment of an earthworm and is the basic unit for generating driving force. The head segment unit expands and radially locks the tube wall. The subsequent segments contract axially in sequence, pushing the middle segments forward. The middle segments begin to expand and anchor in sequence, while the head segment releases its anchorage. The rear segment contracts axially, generating a backward pulling force that pulls the tail body forward. This contraction wave is transmitted from head to tail and repeats, achieving smooth movement in the tube.

[0008] In the above scheme, auxiliary wheels are installed on both sides of the intelligent composite drill bit to enhance the passage performance of the entire multi-stage electric heating intelligent unblocking device inside the gathering and transportation pipeline; the electric heating rod is a high-temperature ceramic electric heating rod, which makes the surface temperature of the drill bit reach up to 120°C; a heat insulation layer is set at the connection between the intelligent composite drill bit and the first-stage global heating system to prevent heat from being transferred backward, protect the equipment body and focus the heat energy at the front end of the operation to achieve efficient energy utilization.

[0009] The above-mentioned multi-stage electric heating intelligent unblocking device for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines employs a multi-modal collaborative unblocking mechanism. Based on machine learning and intelligent algorithms, it achieves the synergistic effect of mechanical crushing, multi-stage thermal melting, and chemical inhibition, enabling efficient and precise removal of complex frozen blockages in CO2-driven waxy crude oil gathering and transportation pipelines. Simultaneously, it effectively prevents the secondary formation of frozen blockages in a short period. For simple frozen blockage structures, an unblocking mode of simultaneous travel, preheating, and de-icing is adopted. For complex frozen blockage structures, simultaneous travel, preheating, and de-icing are performed, while an intelligent composite drill bit simultaneously performs rotary cutting and impact crushing of the frozen blockage, melting the frozen area. The heating temperature, drill bit speed, and travel strategy are adjusted in real time to optimize the unblocking process. After unblocking, hydrate inhibitors are released into the gathering and transportation pipeline to prevent the secondary formation of hydrate frozen blockages.

[0010] The unblocking method described in the above scheme specifically includes the following steps:

[0011] Step 1: First, determine whether the pipeline needs to be opened based on the location of the blockage: If the blockage is located near the tree or metering room, no opening is required; if the blockage occurs in the gathering and transportation section, especially in the middle and later parts of the section, the pipeline needs to be opened based on the actual location of the blockage.

[0012] Step 2: Connect the unblocking device to a power source. After heating to the target temperature, insert the biomimetic flexible guide tube into the gathering and transportation pipeline to begin unblocking.

[0013] Step 3: The bionic flexible guide tube performs mechanical unblocking. When it cannot proceed further, it indicates that the blockage has been reached. The bionic flexible guide tube is then fixed, and the type of blockage is identified based on feedback from the temperature and torque sensors. If it is a simple blockage, an attempt is made to unblock it at the initial target temperature X0. If the blockage is unblocked, this temperature is maintained to continue proceeding. If the blockage is unblocked, the primary heating target temperature X is adjusted. tTo clear the blockage, primary heating continues continuously, and the power is intelligently adjusted based on the composition and severity of the blockage reported from the front end. For complex blockages, an intelligent learning and adjustment mechanism is activated to determine whether a secondary local heating system needs to be activated. If secondary heating is not required, a new optimal target temperature X for primary heating is determined based on historical blockage clearing experience and the real-time situation of the blockage ahead. t And continue to unblock; if it is necessary to open the secondary heating system, try to unblock based on the power of the intelligent composite drill bit and heating according to the self-built strategy library. If the blockage can be unblocked in a short time, the power of the entire unblocking device is intelligently adjusted in real time according to the data of each sensor of the intelligent composite drill bit.

[0014] Step 4: If the solution based on the historical policy library cannot resolve the issue, activate the dynamic policy optimization module based on reinforcement learning to construct a real-time state-action space, reinforce learning agent interaction and environmental feedback, learn online and adjust in real time, and simultaneously perform security intervention.

[0015] Step 5: After the blockage is cleared, hydrate inhibitors are automatically sprayed into the gathering and transportation pipeline. The minimum effective inhibitor concentration is intelligently calculated based on the composition of the blockage and environmental parameters.

[0016] Step 6: After the blockage is completely removed, turn off the secondary heating and the intelligent composite drill bit, adjust the primary heating target temperature to the initial target temperature X0, then slowly remove the entire unblocking device, turn off the primary heating again, and finally cut off the power.

[0017] Step 7: After the blockage is cleared, the blockage clearing device will learn from the experience and form a strategy library.

[0018] In step three of the above scheme, an intelligent learning and adjustment mechanism is activated to unblock the blockage until the freezing is relieved. This intelligent learning and adjustment mechanism provides the system's recommended optimal unblocking combination based on the results of the device's intelligent learning in the early stages. Specifically, the system first analyzes the data detected by the intelligent drill bit's front end to determine the main components and degree of blockage, and then determines whether the secondary local heating module needs to be activated. If the secondary heating module does not need to be activated, the system will provide a new optimal target temperature X for primary heating based on historical unblocking experience and the real-time situation of the blockage. t The system will continue to unblock the blockage. If it is necessary to open the secondary heating module, the system will attempt to unblock the blockage by providing the secondary heating drill bit and heating power based on its self-built strategy library. If the blockage can be unblocked in a short time, the system will intelligently adjust the power of the entire system based on the data from the front-end sensors.

[0019] Optimal target temperature X for primary heating t The calculation method is as follows:

[0020] ;

[0021] In the diagram, X0 represents the optimal primary heating target temperature calculated by the system at the current moment, in °C; T amb’ t represents the real-time ambient temperature, in °C; ∆T offset α is the empirical offset determined based on the type of blockage; α∈[0.1,0.8] is the real-time comprehensive smoothing coefficient given by the system based on historical experience and actual conditions.

[0022] The specific process of the reinforcement learning-based dynamic policy optimization module in step four of the above scheme is as follows:

[0023] ① Constructing a real-time state-action space:

[0024] State Awareness t The system integrates multi-dimensional sensor data to construct a comprehensive state vector, including: S1: the current power of the primary heating module; S2: the micro-torque value of the secondary drill bit and its spectral characteristics; S3: the temperature gradient of the drill bit tip and the surrounding pipe wall; S4: the identification of the type of blockage ahead based on the strategy library; and S5: the contact pressure distribution between the biomimetic structure section and the pipe wall.

[0025] Action Decision (Action A) t The control system will select and combine actions from a series of refined movements.

[0026] ;

[0027] Where: A1: Adjusts the primary global heating power ∆P 1 A2: Adjust the power of the secondary local heating rod. ∆P 2 A3: Adjust the speed of the mechanical drill bit ∆RPM A4: Controlled dosing system, injecting specific concentrations of inhibitors in a pulsed manner. C β ;

[0028] ② Reinforcement learning agent interaction and environmental feedback:

[0029] The agent, based on the current state S t It selects an optimal action combination A through its policy network. t And after the action is executed, the blockage in the pipe will change, and the sensor will read new data S. t+1 ,

[0030] The reward value R for this action is calculated based on the preset reward function. t ;

[0031] The reward function is designed as follows:

[0032] ;

[0033] R1= ∆V / V max R2 is the rate of change of the blockage volume; ∆τ The reduction in torque is a positive bonus; R3= ∆P / P max , representing the change in total power; R4 = max(F Current_stress -F Safe_Stress_Threshold ), representing the stress exceeding the safety threshold; R5=10, representing the fixed time penalty for each time step; w1, w2, w3, w4, w5 are weighting coefficients that need to balance deblocking efficiency, energy consumption, and safety during training;

[0034] +R1 indicates that the blockage has been largely cleared; +R2 indicates that the torque value at the drill bit has decreased; -R3 indicates that the overall energy consumption of the device has increased, including heating power, drill bit rotation power, and chemical dosage; -R4 indicates that the overall stress of the device is too high; -R5 indicates that the unblocking time is long.

[0035] ③ Online learning and real-time adjustments;

[0036] ④ Failure avoidance and safety system intervention: Through the safety monitoring module, hard safety boundaries are set, including maximum torque, maximum temperature threshold, and maximum heating time. Once the system detects that the device's action may lead to exceeding the boundaries, the module will immediately reject and forcibly intervene, selecting a conservative safety strategy. At the same time, after each operation, the intelligent learning results are continuously updated in the database to generate a smarter and more comprehensive global model.

[0037] Step seven of the above scheme is as follows: After the blockage is cleared, the blockage clearing device will learn from the experience and form a strategy library. Then, before each blockage clearing operation, the data of the current blockage clearing is compared with the data of existing schemes in the strategy library. If the similarity is higher than a threshold... C k It automatically executes the historical best strategy to unblock the blockage; if the similarity comparison is below a threshold... C k Online reinforcement learning is initiated to provide recommended congestion relief strategies; the initial threshold C0 is given based on the results of indoor experiments, and then adjusted based on actual congestion relief experience. C k The value was adjusted;

[0038] The threshold C is updated online using an F1-score-based feedback controller, which dynamically adjusts the similarity threshold by comparing the historical strategy matching success rate with the target success rate to balance the weight of strategy reuse and online learning.

[0039] ;

[0040] Among them, F1 history The F1 score after the historical strategy was executed; F1 target η is the target F1 score, 0.85; η is the learning rate, 0.01; C complexity T represents the complexity level of the current blockage (level 1-5, with level 5 being the most complex). env The real-time ambient temperature is used, and α / β / γ are weighting coefficients. α+β+γ=1, which enables adaptive adjustment of the threshold as the complexity and environment increase, avoiding policy reuse errors caused by excessively high thresholds in complex and congested scenarios.

[0041] Beneficial effects:

[0042] 1. Compared to traditional single unblocking devices and methods for CO2-driven waxy crude oil gathering and transportation pipelines, the unblocking device of this invention integrates the functions of "heating and melting + mechanical breaking + chemical prevention + intelligent learning." This not only improves unblocking efficiency but also enables intelligent switching and coordinated operation based on the impedance characteristics of the blockage. Simultaneously, it effectively suppresses the secondary generation of freezing blockages in a short period. The integration of machine learning and intelligent algorithms allows for the development of corresponding unblocking strategies based on actual conditions, while also significantly reducing energy consumption.

[0043] 2. The dynamic multi-level thermal management mechanism designed in this invention can realize the switching of multiple unblocking modes. Combined with intelligent algorithms, through the cooperation of first-level global heating, second-level local heating and intelligent drill bit, it can not only quickly break the frozen blockage area, but also save energy and reduce consumption. It greatly avoids energy waste, realizes "heating on demand", and significantly reduces operating costs. It is especially suitable for field operation environments where there is a lack of stable power supply.

[0044] 3. This invention enables autonomous learning from unblocking experiences, allowing for real-time adjustments to unblocking solutions based on actual conditions. It also matches or optimizes the most efficient unblocking combination based on site conditions. Based on sensor data such as torque, impedance, and temperature, it adjusts drill bit speed and heating power in real time. After each operation, it learns and builds a "strategy library," allowing for direct recall of historically optimal solutions for similar situations. For new situations, it autonomously explores and optimizes unblocking strategies. This significantly reduces reliance on human experience, improves the success rate and efficiency of unblocking, and makes the device increasingly "intelligent," achieving intelligent operation.

[0045] 4. Addressing the challenge of multiple blockages coexisting in CO2-enhanced crude oil gathering and transportation pipelines, this invention develops a deblocking device and method integrating mechanical breaking, thermal melting, and chemical inhibition, along with intelligent sensing and decision-making capabilities. This is of significant practical importance for ensuring the efficient and safe application of CO2 enhanced oil recovery technology. The deblocking technology based on multi-stage dynamic thermal management proposed in this invention is specifically designed to solve this key technical challenge.

[0046] 5. This invention breaks through the limitations of existing technologies and traditional single-method unblocking, proposing an intelligent unblocking device for CO2-driven waxy crude oil gathering and transportation pipelines based on multi-level dynamic thermal management. This device integrates a "mechanical impact + heating and melting + chemical prevention" mode. The designed intelligent drill bit not only breaks the blockage through its own rotational impact, but also incorporates an intelligent heating device, integrating machine learning and intelligent algorithms. It can adjust the power of mechanical impact and heating and melting in real time based on the blockage status feedback from front-end sensors. Furthermore, the system learns from each unblocking experience and establishes a corresponding strategy library. When unblocking is attempted again, the system compares and decides on the most suitable combination of unblocking methods. This unblocking method breaks through the barriers of existing single-method unblocking technologies for CO2-driven waxy crude oil gathering and transportation pipelines, providing a safe, low-carbon, and efficient new paradigm for on-site unblocking of CO2-driven waxy crude oil gathering and transportation pipelines in the context of "dual carbon" (carbon dioxide, carbon dioxide, and carbon emissions).

[0047] 6. Addressing the blockage problem in CO2-driven waxy crude oil gathering and transportation pipelines, this invention employs a multimodal synergistic unblocking mechanism. Based on machine learning and intelligent algorithms, it achieves the synergistic effect of mechanical fracturing, multi-stage thermal melting, and chemical inhibition, realizing efficient and precise removal of complex blockages caused by frozen materials in CO2-driven waxy crude oil gathering and transportation pipelines. Simultaneously, it effectively prevents the secondary formation of frozen materials within a short period. Its intelligent features significantly reduce reliance on manual experience and energy consumption, making it particularly suitable for the safe and economical unblocking of complex blockages such as hydrates, wax, and silt in gathering and transportation pipelines caused by CO2-driven produced fluids under various conditions. Attached Figure Description

[0048] Figure 1 This is a flowchart of the intelligent unblocking method in this invention.

[0049] Figure 2 This is a schematic diagram of a multi-stage electric heating intelligent unblocking device for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines.

[0050] Figure 3 This is a diagram demonstrating the effect of relieving congestion.

[0051] Figure 4 This is a diagram of the intelligent decision-making process for congestion relief strategies.

[0052] Figure 5 This is a diagram showing the fabrication of the frozen plug structure.

[0053] In the diagram: 1 Flexible electric heating tube, 2 Bionic flexible guide tube, 3 Expansion joint, 4 Drill bit housing, 5 Motor system, 6 Electric heating rod, 7 Drill bit, 8 Miniature night vision camera, 9 Nozzle, 10 Drive shaft, 11 Heat insulation plate, 12 Thermal conductive material, 13 Primary heating power supply main switch, 14 Secondary heating power supply main switch, 15 Power main switch, 16 Emergency brake switch, 17 Inhibitor storage tank, 18 Intelligent control system. Detailed Implementation

[0054] The present invention will be further described below with reference to the accompanying drawings:

[0055] See Figures 1-5 This multi-stage electrically heated intelligent unblocking device for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines includes a dredging system, an outer frame, and an intelligent control system 18. The dredging system integrates a primary global heating system, a secondary local heating system, and a chemical dosing system; the outer frame is a biomimetic flexible guiding and thermally reinforced pipe structure with excellent throughput and structural strength; the intelligent control system adjusts the heating temperature, drill bit speed, and travel strategy in real time based on sensor feedback, and has self-learning capabilities to optimize the unblocking process. Details are as follows:

[0056] The exoskeleton is a biomimetic flexible guide tube designed to mimic the movement of earthworms. It expands and contracts at its segmental units to move within the pipe. This biomimetic flexible guide tube possesses both strength and toughness, protecting the internal flexible electric heating tube 1 while ensuring excellent flow performance within the pipe. A thermally conductive material 12 is coated on the inner wall of the biomimetic flexible guide tube 2, creating a biomimetic flexible guide and thermally reinforced structure. This allows the heat generated by the internal electric heating to be transferred to the outer surface more quickly and evenly, effectively melting frozen structures. A temperature sensor is installed on the outside of this structure to monitor temperature changes within the pipe in real time.

[0057] The biomimetic flexible guiding and thermally reinforced structural tube is the core motion and functional unit of this device, inspired by the efficient peristaltic movement of earthworms. It is not simply a flexible connection, but a high-performance actuator integrating biomimetic motion, powerful propulsion, heat transfer, and directional stability, designed to achieve efficient and reliable movement in complex, narrow, and winding pipe environments. The entire biomimetic flexible guiding and thermally reinforced structural tube is composed of a flexible, explosion-proof metal hose. This hose possesses both strength and flexibility, and is explosion-proof. The hose is composed of multiple independent semi-rigid "segment units." Each segment unit is equivalent to a segment of an earthworm, serving as the basic unit for generating driving force. The forward, backward, and fixed movement of the entire structural tube within the pipe is achieved through the contraction and expansion of the expansion joints 3 at the segments. The heating element (flexible electric heating tube 1) of the primary global heating system and the heat-conducting material are directly embedded or tightly bonded to the wall of the flexible metal hose. This allows the entire structural tube to act as a uniform heat source while in motion, efficiently transferring its heat to the entire inner wall of the pipe. This achieves a synergistic effect of "moving, preheating, and de-icing simultaneously," significantly reducing the workload of subsequent secondary local heating modules. The head segment expands, radially locking the pipe wall (anchoring), while subsequent segments contract axially in sequence, pushing the middle section forward. The middle segments begin to expand and anchor sequentially, while the head segment releases its anchorage. The rear segment contracts axially, generating a backward pulling force that pulls the tail section forward. This contraction wave is transmitted from head to tail, repeating continuously, enabling the entire device to move smoothly within the pipe.

[0058] The unblocking system mainly includes a heating system and a chemical dosing system. The heating system is further divided into a primary global heating module and a secondary local heating module. The heat source for the primary global heating module is a flexible electric heating tube embedded within a biomimetic flexible guiding and thermally reinforced structural tube. High-performance thermally conductive material is filled between the heating tube and the biomimetic flexible guiding and thermally reinforced structural tube. The nesting of the flexible electric heating tube and the outer frame ensures both the heating performance of the device and the overall permeability of the device. The primary heating system mainly increases the overall temperature of the pipeline, effectively inhibiting the further aggravation of freezing blockages and reducing the difficulty of subsequent unblocking. Simultaneously, the outer frame itself has a certain mechanical strength. Therefore, the primary global heating system can dissolve simple freezing structures without activating the secondary local heating system, based on "mechanical unblocking + low-temperature heating and melting," resulting in lower overall energy consumption during the unblocking process. The secondary local heating module is essentially a self-heating drill bit, constructed entirely of high-strength copper alloy. It can rotate to target the frozen area. Mechanical removal is performed using a drill bit equipped with a micro-torque sensor and a temperature sensor. These sensors can transmit blockage data to the terminal to determine the main components of the blockage and adjust the unblocking strategy accordingly. The drill bit also uses an embedded electric heating rod to melt the frozen blockage area. A secondary local heating module is installed at the head of the biomimetic flexible guide and thermally enhanced structure pipe, with auxiliary wheels on both sides to enhance the device's passage performance within the pipeline. The embedded high-temperature ceramic electric heating rod allows the drill bit surface temperature to reach up to 120°C. High-performance thermally conductive material is filled between the electric heating rod 6 and the drill bit shell 4 to achieve rapid heat transfer. A chemical dosing system is installed at the tail of the intelligent drill bit, including several micro-nozzles 9, an integrated flow channel, a rear chemical delivery hose interface, and an inhibitor storage tank 17. This system releases inhibitors into the pipeline after the blockage is cleared, preventing recurrence of blockages in a short time and reducing the frequency of blockages, thus reducing the number of unblocking operations and lowering costs. After or during successful unblocking, hydrate inhibitors (such as methanol and ethylene glycol) are sprayed onto the cleared pipe walls in a targeted and quantitative manner, significantly reducing the risk of hydrate freezing and blockage in the area within a short period. The heating system is equipped with a main power switch 15, an emergency stop switch 16, a primary heating power supply main switch 13, and a secondary heating power supply main switch 14.

[0059] The heating system works in two ways: first, it melts hydrates to unclog the pipe; second, it raises the temperature inside the pipe to inhibit the secondary formation of hydrates within a short period. The primary global heating system is distributed throughout the entire unblocking pipe, consisting of flexible electric heating tubes embedded in a biomimetic flexible guiding and thermally reinforced structure pipe. Both the heating tubes and the biomimetic flexible guiding and thermally reinforced structure pipe are filled with high-performance thermally conductive material. This system primarily thaws areas with minor blockages and raises the average temperature inside the pipe, inhibiting the re-formation of blockages within a short time. The secondary regional heating system is a composite drill bit mounted at the head, primarily targeting severely blocked areas. This drill bit is made of high-strength copper alloy and can independently heat while rotating to mechanically break up the blocked area. It is equipped with a micro-torque sensor, a temperature sensor, and a miniature night vision camera, which can feed back measured blockage data to the terminal to determine the main components of the blockage and adjust the unblocking strategy accordingly. The drill bit also melts the blocked area using its embedded high-temperature ceramic electric heating rod. The secondary local heating module is mainly installed at the head of the biomimetic flexible guiding and thermally reinforced structure pipe.

[0060] This intelligent composite drill bit, capable of generating heat, combines mechanical impact, heating and melting, and chemical prevention functions. Forged from a single piece of high-strength copper alloy, it possesses exceptional mechanical strength under complex stress, and its excellent thermal conductivity ensures efficient utilization of the electric heating element's heat energy. This enables the fastest possible thawing of severely frozen areas in pipelines and prevents secondary freezing, fundamentally improving the efficiency and long-term effectiveness of unblocking operations. The intelligent composite drill bit consists of two main functional modules: ① Intelligent sensing and mechanical crushing module: its core comprises a high-torque micro-motor drive system, the drill bit body, a high-precision torque sensor, a temperature sensor, and a miniature night vision camera. The motor system 5 is connected to the drill bit 7 via a drive shaft 10, which is equipped with a torque sensor. The drill bit, through its unique geometric design (such as spiral grooves and carbide teeth), performs rotary cutting and impact crushing on the frozen material. The torque sensor integrated into the drive chain can monitor the resistance torque on the drill bit in real time, providing the control system with the most direct operating data. Simultaneously, the miniature night vision camera 8 can observe the freezing blockage inside the pipeline in real time. Based on the combined analysis of the torque sensor, temperature sensor, and real-time image, it can determine the composition and severity of the blockage. The system can then intelligently adjust the real-time operating power of each component according to the blockage composition, achieving precise and rapid blockage removal. ② Directional thermal melting module: This module mainly consists of a high-power ceramic heating rod, an embedded thermocouple temperature sensor, a high thermal conductivity copper alloy substrate, and a heat insulation layer (heat insulation plate 11). The thermally conductive material filling the space between the electric heating rod and the drill bit shell ensures that the heating rod can heat the drill bit tip to a high temperature of 120℃ in a short time. Simultaneously, a heat insulation layer is installed at the connection point between the drill bit and the main body of the equipment, effectively preventing heat transfer backward, protecting the equipment body, and focusing heat energy at the working front end for efficient energy utilization. The intelligence of this drill bit is reflected in its adaptive decision-making loop based on real-time feedback. Its working logic is as follows: When the system detects an increase in the device's travel resistance, the drill bit contacts the blockage. The control system first determines the structure and composition of the blockage based on torque sensor signals (T), temperature sensors, and images of the blockage. Then, the system adjusts the blockage-breaking method according to the judgment. If the system determines it is a minor blockage, it only activates the mechanical crushing mode, clearing the blockage at the optimal speed to save energy. If the system determines it is a severely frozen core area, the control system immediately activates the directional thermal melting module. This is not a simple activation, but a dynamic and coordinated adjustment of heating power (temperature) and drill bit speed. When the torque is extremely high, high-power heating (prioritizing melting the bonded parts) is used combined with low speed and high torque (continuous crushing) to avoid equipment overload. When the torque is moderate but continuous, medium-power heating is used combined with the optimal cutting speed to achieve a highly efficient composite unblocking method of "mechanical crushing + thermal melting." After the frozen blockage is broken, the system opens the dosing nozzle to spray inhibitors to prevent the pipe from freezing again in a short time.

[0061] The intelligent control system primarily controls the heating mode of the entire system and implements dynamic thermal management based on machine learning and intelligent algorithms. The system learns through three stages: "indoor," "semi-on-site," and "actual on-site," continuously updating its database. The intelligent control system intelligently adjusts the temperature, frequency of the mechanical unblocking drill bit, and dosage of chemicals in the heating system. It also intelligently learns from each unblocking process, building a strategy library so that it can quickly provide the optimal unblocking strategy when unblocking occurs again.

[0062] This invention presents a multi-stage electric heating intelligent unblocking strategy for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines. The core of this intelligent unblocking strategy is a multi-stage dynamic thermal management mechanism. This mechanism is based on a multi-stage dynamic thermal management system that integrates spatial and temperature coordination, global and local linkage, and intelligent learning and optimization. This system aims to efficiently and energy-savingly solve the freezing blockage problem in CO2-driven gathering and transportation pipelines. The system employs a three-stage progressive learning mechanism, gradually transitioning from a laboratory environment to actual industrial settings, continuously optimizing the unblocking model.

[0063] The aforementioned intelligent unblocking method using a multi-stage electric heating system for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines (i.e., the multi-stage electric heating intelligent unblocking method for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines) integrates machine learning algorithms with the entire unblocking process, achieving an upgrade from "experience-driven fixed logic" to "data-driven adaptive decision-making," making the unblocking method for CO2-driven waxy crude oil gathering and transportation pipelines simpler and more intelligent. The process involves three stages of learning, progressing from laboratory to semi-field testing and then to full-field testing, continuously refining the model. The three specific stages are as follows:

[0064] ① First Stage: Indoor Experimental Learning Stage. This stage mainly involves creating different types of frozen structures using transparent control based on on-site conditions. By controlling the percentage content of six components—water, oil, ice, hydrates, sediment, and gas—in the frozen material, different types of frozen structures are classified, categorizing them as simple or complex. Unblocking experiments are conducted using the aforementioned intelligent unblocking device, recording data from temperature and torque sensors during the unblocking process. Then, a relationship is established between the percentage content of each component and the corresponding temperature and torque values. Finally, the frozen structures are classified into three categories: simple, moderate, and complex. The specific implementation steps are as follows:

[0065] a. Control the content of six components—water, oil, ice, hydrates, sediment, and gas—to generate representative frozen blockage samples. Cover frozen blockage types from "simple" to "complex," with at least 10 samples for each type, and record the precise component proportions for each sample.

[0066] Construct the component vector Z = [W, O, I, H, S, G] (unit: %)

[0067] Wherein, W, O, I, H, S, and G represent water, oil, ice, hydrates, sediment, and gas, respectively;

[0068] b. Using the aforementioned intelligent unblocking device, an unblocking experiment was conducted, simultaneously collecting time-series data from the temperature sensor and torque sensor, and extracting key features:

[0069] Each sample group corresponds to a feature vector X (extracted from sensor data). The extracted feature vectors are associated with the corresponding known contents of frozen blockage components. Machine learning algorithms are used for training to establish a mapping relationship from component contents to sensor response feature vectors.

[0070]

[0071] Establish a mapping relationship between the signal and the sensor feedback signal during the unblocking process, thereby enabling rapid and accurate identification of the type of freezing blockage (simple and complex).

[0072] c. Conduct unblocking experiments to identify the type of freezing blockage.

[0073] The above-mentioned device was used to conduct unblocking experiments. Based on the feedback results from the temperature sensor and torque sensor, the type of freezing blockage was identified according to the above mapping relationship. If it was identified as a simple freezing blockage, first-level global heating was used first to melt and clear the blockage. If it was identified as a complex freezing blockage, the dynamic strategy optimization module based on reinforcement learning was directly activated to coordinate and control the first and second-level heating power, drill bit speed and chemical agent injection strategy.

[0074] ② Second stage: On-site visualization experiment and learning stage. Since there are differences between indoor experiments and actual on-site conditions, this stage uses low-yield old wells as the test objects. Using the same research methods as indoor experiments, the surface gathering and transportation pipeline section is replaced with a transparent pipeline. By controlling the well opening and closing time, different types of frozen blockage structures are generated. Then, unblocking operations are carried out, thereby revising the relationship between the strength of the frozen blockage structure formed in the first stage and the temperature and strength of the frozen blockage structure to better reflect the actual on-site conditions.

[0075] ③ Third Stage: This stage primarily involves on-site unblocking. The system further modifies the aforementioned relationships based on the on-site unblocking situation. Simultaneously, the system can achieve dynamic strategy optimization based on reinforcement learning during unblocking. After unblocking is completed, the system intelligently learns from the unblocking experience, continuously updating its self-built strategy library. After each unblocking operation, the system automatically learns process data and updates the strategy library, forming a continuously evolving intelligent unblocking capability. Furthermore, the system's intelligence is reflected in its closed-loop control of dynamic perception, decision execution, and experience learning. The system incorporates multiple sets of temperature and torque sensors, as well as a miniature night vision camera, to monitor the temperature of each heating unit, the device's travel resistance, motor load changes, and blockage status in real time. Based on this real-time data, the control system dynamically adjusts the output ratio of the primary and secondary heating power and the device's travel speed. For example, when resistance increases, the secondary heating power is automatically increased and the speed slightly reduced; when the pipe is clear, the primary power is reduced to save energy. The system incorporates a machine learning-based algorithm model. Its initial database comes from a large amount of indoor unblocking experimental data under various operating conditions (such as temperature, humidity, and media type). In actual operation, the system continuously records the successful parameters of each unblocking process and uses algorithms for self-learning and optimization. This allows it to automatically predict and set the optimal initial heating parameters (such as the starting temperature X0) when encountering similar conditions again, thereby maximizing unblocking efficiency and minimizing overall energy consumption. For example, in the initial stage of unblocking, a primary global preheating is performed, followed by switching to a secondary local heating + mechanical removal based on impedance feedback. When the impedance results indicate high throughput, the system will adjust or turn off the secondary local heating based on the actual situation. If the throughput is poor, the secondary heating will continue to be activated, and the power will be adjusted based on feedback results. This cycle continues until the blockage is completely cleared. This intelligent adjustment mechanism not only improves the automation and accuracy of the unblocking process but also effectively avoids secondary problems caused by excessive energy consumption and overheating, significantly improving the performance and economy of the entire unblocking system.

[0076] In actual unblocking, the electric heating unblocking device is first connected to the power supply, and the primary global heating element is confirmed to be ready. The system automatically sets the primary heating target temperature X0 based on the ambient temperature. Then, the primary heating is automatically activated, raising the surface temperature of the device to the target temperature X0 under the current ambient temperature conditions. Next, the preheated device is sent into the blocked section. First, the type of blockage is identified. If it is a simple blockage, an attempt is made to unblock it at temperature X0. If the blockage is unblocked, this temperature is maintained to continue. If the blockage is unblocked, the system will adjust the primary heating target temperature X0. tThe system performs unblocking operations continuously using primary heating, intelligently adjusting power based on the composition and severity of the blockage reported from the front end. In cases of complex freezing blockages, the system activates an intelligent learning and adjustment mechanism to unblock the blockage until it is resolved. After the blockage is resolved, the system releases hydrate inhibitors. Once the blockage is completely resolved, the secondary heating element is shut off first, then the device is safely removed from the pipe. Finally, the primary heating element is shut off, and the unblocking device is cleaned for reuse. After unblocking is completed, the system summarizes and analyzes the unblocking process and updates the system strategy library.

[0077] The specific steps for implementing this intelligent congestion resolution strategy are as follows:

[0078] Step 1: First, determine whether the pipeline needs to be opened based on the location of the blockage. If the blockage is located near the tree or metering room, no opening is required. If the blockage occurs in the gathering and transportation section, especially in the middle or later part of the section, the pipeline needs to be opened based on the actual location of the blockage.

[0079] Step 2: Connect the intelligent unblocking device to the power supply and confirm that the heating system is fully ready. The system first sets the initial target temperature X0 for the first-level heating based on the current real-time ambient temperature. Then, the first-level global heating is started, and the control panel automatically calculates the remaining heating time. After the entire device reaches the target temperature, the biomimetic flexible guide and thermally enhanced structure pipe is sent into the gathering and transportation pipeline from the opening to start unblocking.

[0080] Step 3: After inserting the biomimetic flexible guide and thermally reinforced structure tube into the pipeline, it is first allowed to move forward within the pipeline. At this stage, the biomimetic flexible guide and thermally reinforced structure tube provides simple mechanical unblocking. When the tube can no longer move forward, it indicates that a blockage has been reached. The tube is then fixed in place. The system will first identify the type of blockage based on feedback from the temperature and torque sensors. If it is a simple blockage, an initial target temperature of X0 is used to attempt to unblock it. If this temperature is achieved, the tube is maintained and the system continues to move forward. If this temperature is not achieved, the system will adjust the primary heating target temperature X0. t To clear the blockage, primary heating is continuously applied, and the power is intelligently adjusted based on the composition and severity of the blockage reported from the front end. For complex blockages, the system activates an intelligent learning and adjustment mechanism to clear the blockage until it is resolved. This intelligent learning and adjustment mechanism provides the optimal blockage clearing combination based on the results of previous intelligent learning by the device. Specifically, the system first analyzes the data detected by the intelligent drill bit to determine the main components and degree of blockage, and then determines whether to activate the secondary local heating module. If secondary heating is not required, the system will provide a new optimal target temperature X for primary heating based on historical blockage clearing experience and the real-time situation of the blockage.t The system will continue to unblock the blockage. If it is necessary to open the secondary heating module, the system will attempt to unblock the blockage by providing the secondary heating drill bit and heating power based on its self-built strategy library. If the blockage can be unblocked in a short time, the system will intelligently adjust the power of the entire system based on the data from the front-end sensors.

[0081] The above-mentioned optimal target temperature X for primary heating t The calculation method is as follows:

[0082]

[0083] Where X0: the optimal primary heating target temperature calculated by the system at the current moment, in °C; T amb’ t: Real-time ambient temperature, °C; ∆T offset : Empirical offset determined by the type of blockage; α∈[0.1,0.8]: Real-time comprehensive smoothing coefficient given by the system based on historical experience and actual situation;

[0084] Step 4: If the solution based on the historical policy library fails to resolve the issue, the system will initiate a dynamic policy optimization module based on reinforcement learning. The specific process is as follows:

[0085] ① Constructing a real-time "state-action" space: State awareness (State S t The system will fuse multi-dimensional sensor data to construct a comprehensive state vector. This includes, but is not limited to:

[0086] S1: Current power of the primary heating module; S2: Micro-torque value of the secondary drill bit and its spectral characteristics; S3: Temperature gradient of the drill bit tip and surrounding pipe wall; S4: Identification of the type of blockage ahead based on the strategy library; S5: Contact pressure distribution between the biomimetic structure section and the pipe wall.

[0087] Action decision (Action A) t The control system will select and combine actions from a series of refined "movements":

[0088]

[0089] Where: A1: Adjust the first-level global heating power ( ∆P 1 A2: Adjust the power of the secondary local heating rod ( ∆P 2 A3: Adjust the speed of the mechanical drill bit ( ∆RPM A4: Controlled dosing system, injecting specific concentrations of inhibitors in a pulsed manner ( C β ).

[0090] ② Reinforcement learning agent interaction and environmental feedback:

[0091] The agent determines the current state S based on the current state S. t It selects an optimal action combination A through its policy network. t And execute. After the action is executed, the environment (i.e., the blocked pipe) will change, and the sensor will read new data S. t+1 .

[0092] The system calculates the reward value R for this action based on a preset reward function. t .

[0093] The reward function is designed as follows:

[0094]

[0095] R1= ∆V / V max R2 is the rate of change of the blockage volume; ∆τ The reduction in torque is a positive bonus; R3= ∆P / P max , representing the change in total power; R4 = max(F Current_stress -F Safe_Stress_Threshold R1 = 10, representing stress exceeding the safety threshold; R2 = 10, a fixed time penalty for each time step; w1, w2, w3, w4, w5 are weighting coefficients that need to balance deblocking efficiency, energy consumption, and safety during training; +R1: Large-area deblocking (indicating good deblocking effect, main reward); +R2: Reduced torque value at the drill bit (indicating reduced resistance); -R3: Increased overall energy consumption of the device, including heating power, drill bit rotation power, and chemical dosage (cost penalty); -R4: Excessive overall stress on the device (safety penalty); -R5: Long deblocking time (efficiency penalty).

[0096] The goal of an intelligent agent is to learn a strategy that maximizes long-term cumulative rewards, namely, to unblock the blockage in the shortest time, with the lowest energy consumption, and in the safest way.

[0097] ③ Online learning and real-time adjustments:

[0098] After each action, the intelligent system will determine based on (S) t A t R t S t+1The tuple updates its policy network. This means the device can learn the unique properties of the blockage in real time during the unblocking operation and dynamically adjust its strategy. For example, it might discover that for the current blockage, which is a dense hydrate with a high wax content in crude oil, a combined strategy of "high-frequency drill bit fracturing + medium-power secondary heating + primary heating continuous heat preservation" is far more effective than any preset fixed mode.

[0099] ④ Failure avoidance and security system intervention:

[0100] To prevent the reinforcement learning agent from making dangerous decisions during exploration, the system integrates a safety monitoring module. This module sets hard safety boundaries (including maximum torque, maximum temperature threshold, and maximum heating time) based on extensive safety knowledge and historical data. If the system detects that the device's actions may exceed these boundaries, the module immediately rejects and intervenes, selecting a conservative safety strategy. Simultaneously, after each operation, the device's intelligent learning results are continuously updated in the database, generating a smarter and more comprehensive global model.

[0101] Step 5: After the blockage is cleared, the system will automatically spray hydrate inhibitors (such as methanol, ethylene glycol, etc.) into the pipeline as needed. The system will intelligently calculate the minimum effective inhibitor concentration based on the composition of the blockage and environmental parameters. For example, in areas where the blockage clearing process takes a long time and consumes a lot of energy, a larger amount of inhibitor should be sprayed. Conversely, in areas where the blockage clearing process takes a short time and consumes less energy, the system will spray a small amount of hydrate inhibitor.

[0102] Step Six: After the freezing blockage is completely relieved, the system first shuts down the secondary zone heating and the smart drill bit, adjusts the primary heating target temperature to the initial target temperature X0, then slowly removes the entire device, shuts down the primary heating again, and finally cuts off the power.

[0103] Step 7: After each congestion resolution, the intelligent congestion resolution device learns from the experience and forms a strategy library. Before each subsequent resolution, the system compares the current resolution data with existing solutions in the strategy library. If the similarity exceeds a threshold... C k The system will automatically execute the best historical strategy to unblock the blockage. If the similarity is below the threshold... C k At this point, the system initiates online reinforcement learning, providing recommended congestion resolution strategies. The initial threshold C0 is determined based on the results of indoor experiments, and then adjusted according to actual congestion resolution experience. C k The value is adjusted.

[0104] The threshold C is updated online using an F1-score-based feedback controller, which dynamically adjusts the similarity threshold by comparing the historical strategy matching success rate with the target success rate to balance the weight of strategy reuse and online learning.

[0105]

[0106] F1 history The F1 score after the historical strategy was executed; F1 target η is the target F1 score, 0.85; η is the learning rate, 0.01; C complexity T represents the complexity level of the current blockage (level 1-5, with level 5 being the most complex). env The real-time ambient temperature is used, and α / β / γ are weighting coefficients. α+β+γ=1, which enables adaptive adjustment of the threshold to "the higher the complexity and the more extreme the environment, the lower the threshold", avoiding policy reuse errors caused by excessively high thresholds in complex and congested scenarios. Example

[0107] First, a simulated frozen blockage structure was prepared, followed by indoor deblocking experiments. The simulated frozen blockage structure was prepared by mixing oil and water in a specific ratio and then injecting the mixture into a transparent pipe section and freezing it for 24 hours to form a stable frozen blockage structure. Figure 5 As shown in the figure. Subsequently, a deblocking experiment was conducted using a deblocking device.

[0108] Since the frozen area in the indoor experiment is close to the pipe opening, there is no need to consider opening it.

[0109] Step 1: Connect the intelligent unblocking device to the power supply and confirm that the heating system is fully ready. The system first sets the initial target temperature X0 for the first-level heating based on the current real-time ambient temperature. Then, the first-level global heating is started, and the control panel automatically calculates the remaining heating time. After the entire device reaches the target temperature, the biomimetic flexible guide and thermally enhanced structure pipe is sent into the gathering and transportation pipeline from the opening to start unblocking.

[0110] Step Two: After inserting the biomimetic flexible guide and thermally reinforced structure tube into the pipeline, it is first allowed to move forward within the pipeline. At this stage, the biomimetic flexible guide and thermally reinforced structure tube provides simple mechanical unblocking. When the tube can no longer move forward, it indicates that a blockage has been reached. The tube is then fixed in place. The system will first identify the type of blockage based on feedback from the temperature and torque sensors. If it is a simple blockage, an initial target temperature of X0 is used to attempt to unblock it. If this temperature is achieved, the system continues to move forward. If the blockage cannot be unblocked, the system will adjust the primary heating target temperature X0. tTo clear the blockage, primary heating is continuously applied, and the power is intelligently adjusted based on the composition and severity of the blockage reported from the front end. For complex blockages, the system activates an intelligent learning and adjustment mechanism to clear the blockage until it is resolved. This intelligent learning and adjustment mechanism provides the optimal blockage clearing combination based on the results of previous intelligent learning by the device. Specifically, the system first analyzes the data detected by the intelligent drill bit to determine the main components and degree of blockage, and then determines whether to activate the secondary local heating module. If secondary heating is not required, the system will provide a new optimal target temperature X for primary heating based on historical blockage clearing experience and the real-time situation of the blockage. t The system will continue to unblock the blockage. If it is necessary to open the secondary heating module, the system will attempt to unblock the blockage by providing the secondary heating drill bit and heating power based on its self-built strategy library. If the blockage can be unblocked in a short time, the system will intelligently adjust the power of the entire system based on the data from the front-end sensors.

[0111] The above-mentioned optimal target temperature X for primary heating t The calculation method is as follows:

[0112]

[0113] Where X0: the optimal primary heating target temperature calculated by the system at the current moment, in °C;

[0114] T amb’ t: Real-time ambient temperature, °C;

[0115] ∆T offset : Empirical offset determined by the type of blockage;

[0116] α∈[0.1,0.8]: Real-time comprehensive smoothing coefficient given by the system based on historical experience and actual situation;

[0117] Step 3: If the solution based on the historical policy library fails to resolve the issue, the system will initiate a dynamic policy optimization module based on reinforcement learning. The specific process is as follows:

[0118] ① Constructing a real-time "state-action" space: State awareness (State S t The system will fuse multi-dimensional sensor data to construct a comprehensive state vector. This includes, but is not limited to:

[0119] S1: Current power of the primary heating module; S2: Micro-torque value of the secondary drill bit and its spectral characteristics; S3: Temperature gradient of the drill bit tip and surrounding pipe wall; S4: Identification of the type of blockage ahead based on the strategy library; S5: Contact pressure distribution between the biomimetic structure section and the pipe wall.

[0120] Action decision (Action A) t The control system will select and combine actions from a series of refined "movements":

[0121]

[0122] Where: A1: Adjust the first-level global heating power ( ∆P 1 A2: Adjust the power of the secondary local heating rod ( ∆P 2 A3: Adjust the speed of the mechanical drill bit ( ∆RPM A4: Controlled dosing system, injecting specific concentrations of inhibitors in a pulsed manner ( C β ).

[0123] ② Reinforcement learning agent interaction and environmental feedback:

[0124] The agent determines the current state S based on the current state S. t It selects an optimal action combination A through its policy network. t And execute. After the action is executed, the environment (i.e., the blocked pipe) will change, and the sensor will read new data S. t+1 .

[0125] The system calculates the reward value R for this action based on a preset reward function. t .

[0126] The reward function is designed as follows:

[0127]

[0128] R1= ∆V / V max R2 is the rate of change of the blockage volume; ∆τ The reduction in torque is a positive bonus; R3= ∆P / P max , representing the change in total power; R4 = max(F Current_stress -F Safe_Stress_ThresholdR1 = 10, representing stress exceeding the safety threshold; R2 = 10, a fixed time penalty for each time step; w1, w2, w3, w4, w5 are weighting coefficients that need to balance deblocking efficiency, energy consumption, and safety during training; +R1: Large-area deblocking (indicating good deblocking effect, main reward); +R2: Reduced torque value at the drill bit (indicating reduced resistance); -R3: Increased overall energy consumption of the device, including heating power, drill bit rotation power, and chemical dosage (cost penalty); -R4: Excessive overall stress on the device (safety penalty); -R5: Long deblocking time (efficiency penalty).

[0129] The goal of an intelligent agent is to learn a strategy that maximizes long-term cumulative rewards, namely, to unblock the blockage in the shortest time, with the lowest energy consumption, and in the safest way.

[0130] ③ Online learning and real-time adjustments:

[0131] After each action, the intelligent system will determine based on (S) t A t R t S t+1 The tuple updates its policy network. This means the device can learn the unique properties of the blockage in real time during the unblocking operation and dynamically adjust its strategy. For example, it might discover that for the current blockage, which is a dense hydrate with a high wax content in crude oil, a combined strategy of "high-frequency drill bit fracturing + medium-power secondary heating + primary heating continuous heat preservation" is far more effective than any preset fixed mode.

[0132] ④ Failure avoidance and security system intervention:

[0133] To prevent the reinforcement learning agent from making dangerous decisions during exploration, the system integrates a safety monitoring module. This module sets hard safety boundaries (including maximum torque, maximum temperature threshold, and maximum heating time) based on extensive safety knowledge and historical data. If the system detects that the device's actions may exceed these boundaries, the module immediately rejects and intervenes, selecting a conservative safety strategy. Simultaneously, after each operation, the device's intelligent learning results are continuously updated in the database, generating a smarter and more comprehensive global model.

[0134] Step 4: After the blockage is cleared, the system will automatically spray hydrate inhibitors (such as methanol, ethylene glycol, etc.) into the pipeline as needed. The system will intelligently calculate the minimum effective inhibitor concentration based on the composition of the blockage and environmental parameters. For example, in areas where the blockage clearing process takes a long time and consumes a lot of energy, a larger amount of inhibitor should be sprayed. Conversely, in areas where the blockage clearing process takes a short time and consumes less energy, the system will spray a small amount of hydrate inhibitor.

[0135] Step 5: After the freezing blockage is completely relieved, the system first shuts down the secondary zone heating and the smart drill bit, adjusts the primary heating target temperature to the initial target temperature X0, then slowly removes the entire device, shuts down the primary heating again, and finally cuts off the power.

[0136] Step Six: After each congestion resolution, the intelligent congestion resolution device learns from the experience and builds a strategy library. Before each subsequent resolution, the system compares the current resolution data with existing solutions in the strategy library. If the similarity exceeds a threshold... C k The system will automatically execute the best historical strategy to unblock the blockage. If the similarity is below the threshold... C k At this point, the system initiates online reinforcement learning, providing recommended congestion resolution strategies. The initial threshold C0 is determined based on the results of indoor experiments, and then adjusted according to actual congestion resolution experience. C k The value is adjusted.

[0137] The threshold C is updated online using an F1-score-based feedback controller, which dynamically adjusts the similarity threshold by comparing the historical strategy matching success rate with the target success rate to balance the weight of strategy reuse and online learning.

[0138]

[0139] Among them, F1 history The F1 score after the historical strategy was executed; F1 target η is the target F1 score, 0.85; η is the learning rate, 0.01; C complexity T represents the complexity level of the current blockage (level 1-5, with level 5 being the most complex). env The real-time ambient temperature is used, and α / β / γ are weighting coefficients. α+β+γ=1, which enables adaptive adjustment of the threshold to "the higher the complexity and the more extreme the environment, the lower the threshold", avoiding policy reuse errors caused by excessively high thresholds in complex and congested scenarios.

Claims

1. A multi-stage electric heating intelligent unblocking device for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines, characterized in that: This multi-stage electric heating intelligent unblocking device for complex freezing blockages in CO2-driven waxy crude oil gathering and transportation pipelines includes a dredging system, an outer frame, and an intelligent control system. The outer frame is a biomimetic flexible guide tube, composed of multiple segmental units connected by expansion joints. Forward, backward, and fixed movement are achieved through contraction and expansion at the segments. The dredging system integrates a primary global heating system, a secondary local heating system, and a chemical dosing system. The secondary local heating system is installed at the head of the primary global heating system. The heat source for the primary global heating system is a flexible electric heating tube, which is embedded within the biomimetic flexible guide tube to form a uniform heating effect. The heating element consists of a flexible electric heating tube and a biomimetic flexible guide tube, with thermally conductive material filling the space between them. A temperature sensor is installed on the biomimetic flexible guide tube. The secondary local heating system is an intelligent composite drill bit that can generate its own heat. The intelligent composite drill bit includes a motor system, a drill bit shell, an electric heating rod, a drill bit, and a miniature night vision camera. The electric heating rod runs through the drill bit shell, and thermally conductive material is filled between the electric heating rod and the drill bit shell. The motor system is connected to the drill bit via a drive shaft, which is equipped with a torque sensor. A temperature sensor is installed on the drill bit. The dosing system includes an inhibitor storage tank and multiple nozzles evenly distributed at the tail of the intelligent composite drill bit.

2. The multi-stage electric heating intelligent unblocking device for complex freezing blockage in CO2-driven waxy crude oil gathering and transportation pipelines according to claim 1, characterized in that: The biomimetic flexible guide tube is a flexible metal explosion-proof hose. The flexible electric heating tube and heat-conducting material are directly embedded or tightly attached to the tube wall, forming a biomimetic flexible guide and thermally enhanced structure tube. This makes the entire biomimetic flexible guide tube a uniform heat source while in motion, efficiently transferring its heat to the entire inner wall of the pipe, simultaneously moving, preheating, and melting ice. The intelligent composite drill bit performs rotary cutting and impact crushing on the frozen blockage while melting the frozen area. It also feeds back the blockage data to the intelligent control system through actual measurement, enabling the determination of the main components of the blockage and adjustment of the unblocking strategy. After unblocking, the dosing system releases hydrate inhibitors into the gathering and transportation pipeline to prevent the secondary formation of hydrate blockage. The intelligent control system adjusts the heating temperature, drill bit speed, and travel strategy in real time and has self-learning capabilities to optimize the unblocking process.

3. The multi-stage electric heating intelligent unblocking device for complex freezing blockage in CO2-driven waxy crude oil gathering and transportation pipelines according to claim 2, characterized in that: Each segment of the biomimetic flexible guide tube mimics a segment of an earthworm and serves as the basic unit for generating driving force. The head segment expands and radially locks the tube wall, while subsequent segments contract axially in sequence, pushing the middle segments forward. The middle segments begin to expand and anchor sequentially, while the head segment releases its anchorage. The rear segment contracts axially, generating a backward pulling force that pulls the tail body forward. This contraction wave is transmitted from head to tail, repeating continuously, enabling smooth movement within the tube.

4. The multi-stage electric heating intelligent unblocking device for complex freezing blockage in CO2-driven waxy crude oil gathering and transportation pipelines according to claim 3, characterized in that: The intelligent composite drill bit is equipped with auxiliary wheels on both sides to enhance the passage performance of the entire multi-stage electric heating intelligent unblocking device inside the gathering and transportation pipeline; the electric heating rod is a high-temperature ceramic electric heating rod, which can make the surface temperature of the drill bit reach up to 120°C; a heat insulation layer is set at the connection between the intelligent composite drill bit and the first-stage global heating system to prevent heat from being transferred backward, protect the equipment body and focus the heat energy at the front end of the operation, so as to achieve efficient energy utilization.

5. The unblocking method of the multi-stage electric heating intelligent unblocking device for complex freezing blockage in CO2-driven waxy crude oil gathering and transportation pipelines according to claim 4, characterized in that: A multimodal collaborative unblocking mechanism is adopted, based on machine learning and intelligent algorithms to achieve the synergistic effect of mechanical crushing, multi-stage thermal melting, and chemical inhibition. This enables efficient and precise removal of complex blockages caused by freezing in CO2-driven waxy crude oil gathering and transportation pipelines, while effectively preventing the secondary formation of freezing blockages in a short period of time. For simple frozen blockage structures, an unblocking mode of simultaneous travel, preheating, and de-icing is adopted. For complex frozen blockage structures, simultaneous travel, preheating, and de-icing are carried out, while an intelligent composite drill bit performs rotary cutting and impact crushing of the frozen blockage while melting the frozen area. The heating temperature, drill bit speed, and travel strategy are adjusted in real time to optimize the unblocking process. After unblocking, hydrate inhibitors are released into the gathering and transportation pipeline to prevent the secondary formation of hydrate freezing blockages.

6. The unblocking method of the multi-stage electric heating intelligent unblocking device for complex freezing blockage in CO2-driven waxy crude oil gathering and transportation pipelines according to claim 5, characterized in that... Includes the following steps: Step 1: First, determine whether the pipeline needs to be opened based on the location of the blockage: If the blockage is located near the tree or metering room, no opening is required; if the blockage occurs in the gathering and transportation section, especially in the middle and later parts of the section, the pipeline needs to be opened based on the actual location of the blockage. Step 2: Connect the unblocking device to a power source. After heating to the target temperature, insert the biomimetic flexible guide tube into the gathering and transportation pipeline to begin unblocking. Step 3: The bionic flexible guide tube performs mechanical unblocking. When it cannot proceed further, it indicates that the blockage has been reached. The bionic flexible guide tube is then fixed, and the type of blockage is identified based on feedback from the temperature and torque sensors. If it is a simple blockage, an attempt is made to unblock it at the initial target temperature X0. If the blockage is unblocked, this temperature is maintained to continue proceeding. If the blockage is unblocked, the primary heating target temperature X is adjusted. t To clear the blockage, primary heating continues continuously, and the power is intelligently adjusted based on the composition and severity of the blockage reported from the front end. For complex blockages, an intelligent learning and adjustment mechanism is activated to determine whether a secondary local heating system needs to be activated. If secondary heating is not required, a new optimal target temperature X for primary heating is determined based on historical blockage clearing experience and the real-time situation of the blockage ahead. t And continue to unblock; if it is necessary to open the secondary heating system, try to unblock based on the power of the intelligent composite drill bit and heating according to the self-built strategy library. If the blockage can be unblocked in a short time, the power of the entire unblocking device is intelligently adjusted in real time according to the data of each sensor of the intelligent composite drill bit. Step 4: If the solution based on the historical policy library cannot resolve the issue, activate the dynamic policy optimization module based on reinforcement learning to construct a real-time state-action space, reinforce learning agent interaction and environmental feedback, learn online and adjust in real time, and simultaneously perform security intervention. Step 5: After the blockage is cleared, hydrate inhibitors are automatically sprayed into the gathering and transportation pipeline. The minimum effective inhibitor concentration is intelligently calculated based on the composition of the blockage and environmental parameters. Step 6: After the blockage is completely removed, turn off the secondary heating and the intelligent composite drill bit, adjust the primary heating target temperature to the initial target temperature X0, then slowly remove the entire unblocking device, turn off the primary heating again, and finally cut off the power. Step 7: After the blockage is cleared, the blockage clearing device will learn from the experience and form a strategy library.

7. The unblocking method of the multi-stage electric heating intelligent unblocking device for complex freezing blockage in CO2-driven waxy crude oil gathering and transportation pipelines according to claim 6, characterized in that: In step three, an intelligent learning and adjustment mechanism is activated to unblock the blockage until it is cleared. This intelligent learning and adjustment mechanism provides the optimal unblocking combination recommended by the system based on the results of the device's previous intelligent learning. Specifically, the system first analyzes the data detected by the intelligent drill bit to determine the main components and degree of blockage, and then determines whether the secondary local heating module needs to be activated. If the secondary heating module does not need to be activated, the system will provide a new optimal target temperature X for primary heating based on historical unblocking experience and the real-time situation of the blockage. t The system will continue to unblock the blockage. If it is necessary to open the secondary heating module, the system will attempt to unblock the blockage by providing the secondary heating drill bit and heating power based on its self-built strategy library. If the blockage can be unblocked in a short time, the system will intelligently adjust the power of the entire system based on the data from the front-end sensors. Optimal target temperature X for primary heating t The calculation method is as follows: ; In the diagram, X0 represents the optimal primary heating target temperature calculated by the system at the current moment, in °C; T amb’ t represents the real-time ambient temperature, in °C; ∆T offset α is the empirical offset determined based on the type of blockage; α∈[0.1,0.8] is the real-time comprehensive smoothing coefficient given by the system based on historical experience and actual conditions.

8. The unblocking method of the multi-stage electric heating intelligent unblocking device for complex freezing blockage in CO2-driven waxy crude oil gathering and transportation pipelines according to claim 6, characterized in that: The specific process of the dynamic policy optimization module based on reinforcement learning in step four is as follows: ① Constructing a real-time state-action space: State Awareness t The system integrates multi-dimensional sensor data to construct a comprehensive state vector, including: S1: the current power of the primary heating module; S2: the micro-torque value of the secondary drill bit and its spectral characteristics; S3: the temperature gradient of the drill bit tip and the surrounding pipe wall; S4: the identification of the type of blockage ahead based on the strategy library; and S5: the contact pressure distribution between the biomimetic structure section and the pipe wall. Action Decision (Action A) t The control system will select and combine actions from a series of refined movements. ; Where: A1: Adjusts the primary global heating power ∆P 1 A2: Adjust the power of the secondary local heating rod. ∆P 2 A3: Adjust the speed of the mechanical drill bit ∆RPM A4: Controlled dosing system, injecting specific concentrations of inhibitors in a pulsed manner. C β ; ② Reinforcement learning agent interaction and environmental feedback: The agent, based on the current state S t It selects an optimal action combination A through its policy network. t And after the action is executed, the blockage in the pipe will change, and the sensor will read new data S. t+1 , The reward value R for this action is calculated based on the preset reward function. t ; The reward function is designed as follows: ; R1= ∆V / V max R2 is the rate of change of the blockage volume; ∆τ The reduction in torque is a positive bonus; R3= ∆P / P max , representing the change in total power; R4 = max(F Current_stress -F Safe_Stress_Threshold ), representing the stress exceeding the safety threshold; R5=10, representing the fixed time penalty for each time step; w1, w2, w3, w4, w5 are weighting coefficients that need to balance deblocking efficiency, energy consumption, and safety during training; +R1 indicates that the blockage has been largely cleared; +R2 indicates that the torque value at the drill bit has decreased; -R3 indicates that the overall energy consumption of the device has increased, including heating power, drill bit rotation power, and chemical dosage; -R4 indicates that the overall stress of the device is too high; -R5 indicates that the unblocking time is long. ③ Online learning and real-time adjustments; ④ Failure avoidance and safety system intervention: Through the safety monitoring module, hard safety boundaries are set, including maximum torque, maximum temperature threshold, and maximum heating time. Once the system detects that the device's action may lead to exceeding the boundaries, the module will immediately reject and forcibly intervene, selecting a conservative safety strategy. At the same time, after each operation, the intelligent learning results are continuously updated in the database to generate a smarter and more comprehensive global model.

Images