A helium intelligent separation and purification system for geothermal well associated gas

Abstract

This invention belongs to the field of helium extraction technology, specifically disclosing an intelligent helium separation and purification system for associated gas from geothermal wells. The system comprises, sequentially arranged along the gas flow direction, a gas-liquid separator, a pressure equalization tank, a pressurization mechanism, a sulfur adsorption tower, a hydrogen removal tower, a gas dryer, a gas dispersion extraction component, an adsorption purification unit, a vacuum pump, and a product gas storage cylinder group, and is equipped with a central controller. The central controller adjusts the speed of the pressurization mechanism based on pressure and flow to stabilize the gas source, adjusts the valve opening of each membrane module according to its operating parameters to achieve balanced separation, and dynamically adjusts the vacuum pump speed and adsorption switching sequence based on the feed gas parameters. This invention, through multi-stage pretreatment, independent control of membrane modules, and adaptive control of the adsorption unit, effectively adapts to the fluctuating operating conditions of associated gas from geothermal wells, improving helium recovery rate and reducing energy consumption.

Description

Technical Field

[0001] This invention belongs to the field of helium extraction technology, and specifically relates to an intelligent helium separation and purification system for associated gas from geothermal wells. Background Technology

[0002] Helium is widely used in aerospace, semiconductor manufacturing, and high-tech fields, and is an extremely important and scarce resource. Traditionally, helium is mainly extracted from helium-rich natural gas. With the development of geothermal resources, associated gas from geothermal wells contains a certain concentration of helium, and separating and purifying it has extremely high economic value and strategic significance.

[0003] Associated gas from geothermal wells is characterized by dispersed gas sources, complex composition, and drastic fluctuations in operating conditions. Associated gas contains not only helium but also impurities such as hydrogen sulfide, hydrogen, and saturated water vapor. These impurities adversely affect conventional membrane separation and pressure swing adsorption (PSA) purification processes. The gas production flow rate and helium concentration of geothermal wells fluctuate significantly with changes in geological conditions and mining stages, making it difficult for conventional separation and purification systems with fixed process parameters to operate stably. In the membrane separation stage, fluctuations in feed gas flow rate or composition cause the actual operating points of each membrane module to deviate from their design conditions, leading to decreased separation efficiency. Furthermore, uneven feed gas distribution among multiple parallel membrane modules further exacerbates the imbalance in separation performance, and some membrane modules may experience performance degradation over long-term operation. In PSA purification, traditional processes use fixed timing and desorption pressures. When the feed helium concentration fluctuates, fixed timing settings cannot fully utilize adsorption saturation, and fixed desorption pressures cannot guarantee recovery rates at low concentrations or reduce energy consumption at high concentrations. Summary of the Invention

[0004] The purpose of this invention is to overcome the defects in the existing technology and provide an intelligent helium separation and purification system for associated gas from geothermal wells.

[0005] This invention provides an intelligent helium separation and purification system for associated gas from geothermal wells, comprising: The system comprises, in sequence along the gas flow direction, a gas-liquid separator, a pressure equalization tank, a pressurization mechanism, a sulfur adsorption tower, a hydrogen removal tower, a gas dryer, a gas dispersion and extraction component, an adsorption and purification unit, a vacuum pump, and a product gas storage cylinder group; and a central controller connected to the pressure equalization tank, the pressurization mechanism, the gas dispersion and extraction component, and the adsorption and purification unit. The gas dispersion and extraction component includes: The raw material gas distribution main pipe is connected to the output end of the gas dryer; Several membrane modules are arranged in a matrix. The feed gas inlet of each membrane module is connected to the feed gas distribution manifold through an inlet hose, and a gas flow rate regulating valve is connected in series on each inlet hose. A temperature sensor is installed inside each membrane module. Each membrane module has a first outlet connected to an outlet hose. A permeate gas outlet regulating valve, a pressure sensor, and a helium concentration sensor are sequentially arranged on the outlet hose along the airflow direction. Each outlet hose is connected in parallel and then connected to the permeate gas main manifold. Each membrane module has a second outlet connected to a stagnation gas hose, and each of the stagnation gas hoses is connected in parallel to a common stagnation gas manifold. The outlet end of the common stagnation gas manifold is connected to the input end of the pressure equalization tank, and is used to send the collected stagnation gas back to the pressure equalization tank to mix with the associated gas. The adsorption and purification unit is equipped with at least two molecular sieve packed towers, each of which is equipped with a vacuum pump. The molecular sieve packed towers are used to perform adsorption or desorption operations, and adjacent molecular sieve packed towers alternately perform adsorption and desorption operations. The input end of each adsorption and purification unit is connected to the permeate gas manifold through a programmable switching valve group, and the outlet of each adsorption and purification unit is connected to the vacuum pump. The vacuum pump is connected to the product gas storage cylinder group. The central controller is configured to adjust the rotational speed of the variable frequency drive end of the pressurizing mechanism based on the outlet pressure of the pressure equalization tank and the outlet flow rate of the pressurizing mechanism; adjust the opening degree of the gas inflow regulating valve and the permeate gas outlet regulating valve of the membrane module based on the feed gas inflow rate, membrane back pressure, internal temperature and permeate gas helium concentration of each membrane module; dynamically adjust the rotational speed of the vacuum pump and the switching sequence of the programmable switching valve group based on the gas flow rate and helium concentration on the inlet pipeline of the adsorption purification unit; and control the programmable switching valve group to switch between adsorption and desorption operations based on the internal pressure of each molecular sieve packed tower.

[0006] A further embodiment is that the membrane assembly is arranged in a rectangular grid pattern. The central controller calculates the relative deviation between the current separation coefficient and the initial separation coefficient of each membrane module based on the parameters of the helium concentration sensor, pressure sensor and temperature sensor of the membrane module. When the relative deviation exceeds a preset threshold, it is determined that the performance of the current membrane module has deteriorated. The central controller outputs a first control command to gradually reduce the opening of the gas inflow regulating valve of the current membrane module, and proportionally increase the opening of the gas inflow regulating valve of other membrane modules adjacent to the current membrane module. After the opening of the gas inflow regulating valve of the current membrane module is reduced to zero, the central controller outputs a second control command to close the gas inflow regulating valve and the permeate gas outlet regulating valve of the current membrane module, and outputs a maintenance reminder signal.

[0007] A further solution is that the central controller integrates an opening degree collaborative control model, the input parameters of which include: the total flow rate of the raw gas entering the gas dispersion extraction component, the helium concentration in the raw gas, the raw gas inflow rate of each membrane component, the pressure behind the membrane of each membrane component, the internal temperature of each membrane component, and the helium concentration of the permeate gas of each membrane component. The output parameters of the opening degree collaborative control model are the opening coefficient of the gas inflow regulating valve and the opening coefficient of the permeate outlet regulating valve for each membrane module. The training process of the opening degree collaborative control model is as follows: historical data of the gas dispersion extraction component under different operating conditions are collected to form a training sample set; feature vectors and corresponding actual regulating valve opening coefficients are extracted from the training sample set; human experts optimize and label the regulating valve opening coefficients according to the extracted feature vectors; the labeled feature vectors are used as the input of the opening degree collaborative control model; and the optimized and labeled regulating valve opening coefficients are used as the output of the opening degree collaborative control model for iterative training until the error between the opening coefficients output by the model and the opening coefficients labeled by the human experts converges to a preset range.

[0008] A further solution is that the central controller collects the input parameters under the current operating conditions in real time and converts them into feature vectors, which are then input into the opening degree collaborative control model to obtain the opening degree coefficients of the gas inflow regulating valve and the permeate gas outlet regulating valve of each membrane module. The central controller outputs a third control command based on the opening coefficient to adjust the gas inflow regulating valve and the permeate outlet regulating valve of each membrane module.

[0009] A further solution is that the central controller integrates a state extraction module and a pressure prediction module. The state extraction module is used to continuously collect the instantaneous gas flow rate and helium volume concentration at the output end of the permeable gas manifold, form time series data, extract the temporal coupling characteristics of flow rate and concentration according to a preset time window, and output a comprehensive feature vector that reflects the evolution law of the feed gas in the whole time domain. The pressure prediction module receives the comprehensive feature vector, outputs the optimal desorption pressure setpoint that the adsorption purification unit should maintain in the next time window through regression prediction, and sends a speed adjustment signal to the vacuum pump so that the actual desorption pressure changes dynamically with the optimal desorption pressure setpoint.

[0010] A further embodiment is that the state extraction module includes: Long Short-Term Memory (LSTM) neural networks are used to extract local temporal correlation feature vectors for each time window; The differential mapping unit is used to perform differential calculation on the local temporal correlation feature vectors of adjacent time windows to obtain the temporal interaction correlation feature vector; A multi-head self-attention feature fusion layer is used to weight and fuse the temporal interaction-related feature vectors of all time windows to generate the comprehensive feature vector; The pressure prediction module includes a regression prediction network composed of multiple fully connected layers stacked together. Its input is the comprehensive feature vector, and its output is the optimal desorption pressure setting value.

[0011] A further embodiment is that the adsorption and purification unit is equipped with three molecular sieve packed towers; The central controller monitors the internal pressure of the three molecular sieve packed towers in real time. Once any molecular sieve packed column has completed its adsorption operation, the central controller is configured to perform the following operations: By controlling the programmable switching valve group to close the feed gas inlet of the current molecular sieve packed tower and sequentially performing multi-stage pressure equalization operations; Open the exhaust port of the current molecular sieve packed tower and discharge the residual gas along the direction of the raw material gas inlet; Connect the current molecular sieve packed tower to a vacuum pump and evacuate it to a vacuum state so that the adsorbed impurity gas is desorbed and extracted. Connect the current molecular sieve packed tower to the product gas outlet of other molecular sieve packed towers in the adsorption stage, or connect it to an external product gas source to introduce high-pressure product gas and raise the pressure inside the tower back to the initial pressure value required for adsorption operation.

[0012] A further embodiment is that the multi-stage pressure equalization operation includes a first-stage pressure equalization operation and a second-stage pressure equalization operation; The first-stage pressure equalization operation process is as follows: Connect the current molecular sieve packed tower to another molecular sieve packed tower that is in a low-pressure state, so that the pressure of the two molecular sieve packed towers reaches the first intermediate pressure value. The secondary pressure equalization operation process is as follows: Connect the current molecular sieve packed tower to the third molecular sieve packed tower, which is in a low-pressure state, so that the pressure of the two molecular sieve packed towers reaches a second intermediate pressure value that is lower than the first intermediate pressure value. The initial pressure of the second molecular sieve packed tower is higher than that of the third molecular sieve packed tower.

[0013] A further solution is that the duration of each stage of the multi-stage pressure equalization operation is dynamically adjusted by the central controller based on the real-time helium concentration data collected from the inlet pipeline of the adsorption and purification unit: when the helium concentration in the feed gas is detected to be lower than the preset concentration threshold, the central controller extends the duration of the first-stage and second-stage pressure equalization operations.

[0014] A further embodiment is that the central controller integrates a pressure equalization time mapping module, used to adjust the duration of the primary and secondary pressure equalization operations; the pressure equalization time mapping module includes: The gear setting unit uses the difference between the helium concentration in the feed gas and the preset standard concentration as the first parameter, and the rate of change of the difference relative to the previous moment as the second parameter; the value ranges of the first parameter and the second parameter are divided into multiple consecutive numerical gears, and the current gear of the first parameter has only one corresponding current gear of the second parameter. A mapping query unit is used to construct a mapping query table. Each entry in the mapping query table corresponds to a unique mapping relationship between the first parameter and the second parameter, and a unique time extension coefficient is pre-written in each entry. The calculation unit is used to calculate the current first parameter and second parameter in real time, determine the numerical level of the first parameter and second parameter respectively, and read the corresponding time extension coefficient from the mapping lookup table to adjust the time length of the first-level pressure equalization operation and the second-level pressure equalization operation.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention features a sulfur adsorption tower, a hydrogen removal tower, and a gas dryer arranged sequentially along the gas flow direction, forming a multi-stage impurity removal pretreatment that effectively prevents impurities from contaminating subsequent membrane modules and molecular sieve packed towers. The central controller collects the outlet pressure of the pressure equalization tank and the outlet flow rate of the booster mechanism, adjusting the speed of the variable frequency drive end of the booster mechanism in real time to maintain stable gas pressure. This overcomes the impact of drastic fluctuations in geothermal well gas source pressure and flow rate on the separation and purification process, improving the system's adaptability to complex gas sources.

[0016] This invention employs a matrix arrangement of multiple membrane modules to form a gas dispersion and extraction assembly, enabling independent control of the operating status of each membrane module. Control commands are output from a central controller to adjust the opening of the gas inflow regulating valve and the gas inflow regulating valves of adjacent membrane modules. This allows the system to isolate and maintain faulty membrane modules without shutting down, achieving automatic fault tolerance and dynamic load balancing in the membrane separation process.

[0017] This invention extracts the local temporal correlation features between the instantaneous gas flow rate and helium concentration at the output end of the permeation gas manifold using a long short-term memory neural network. The temporal interaction features are then obtained through a differential mapping unit, and a comprehensive feature vector is generated by a multi-head self-attention feature fusion layer. This accurately captures the full-time evolution of the feed gas. The comprehensive feature vector is used as input to a pressure prediction module to dynamically adjust the vacuum pump speed, allowing the pump to adjust its suction power in advance based on the predicted feed status. This effectively reduces unit energy consumption while ensuring helium recovery rate.

[0018] This invention employs three molecular sieve-filled towers and uses multi-stage pressure equalization to recover residual helium from the molecular sieve-filled towers to the other towers, avoiding helium loss caused by excessive pressure differentials during single-stage pressure equalization. Furthermore, a pressure equalization time mapping module is integrated into the central controller, which extends the time of the first and second stage pressure equalization operations when the feed helium concentration is low, allowing for more complete helium recovery. Attached Figure Description

[0019] The following figures are for illustrative purposes only and are not intended to limit the scope of the invention, wherein: Figure 1 : A schematic diagram of the overall framework connection of the system of the present invention; Figure 2 Schematic diagram of the gas dispersion and extraction component; Figure 3 : Network logic block diagram for central controller internal state extraction and pressure prediction; In the diagram: 1. Gas-liquid separator; 2. Pressure equalization tank; 3. Pressurization mechanism; 4. Sulfur adsorption tower; 5. Hydrogen removal tower; 6. Gas dryer; 7. Gas dispersion and extraction assembly; 8. Permeate gas main manifold; 9. Membrane module; 10. Residual gas main manifold; 11. Adsorption and purification unit; 12. Molecular sieve packed tower; 13. Vacuum pump; 14. Product gas storage cylinder group; 15. Central controller; 16. Helium concentration sensor; 17. Raw material gas distribution main; 18. Gas inflow regulating valve; 19. Permeate gas outlet regulating valve; 20. Pressure sensor. Detailed Implementation

[0020] To make the objectives, technical solutions, design methods, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0021] like Figure 1As shown, this invention provides an intelligent helium separation and purification system for associated gas from geothermal wells. The system includes, sequentially arranged along the gas flow direction, a gas-liquid separator 1, a pressure equalization tank 2, a pressurization mechanism 3, a sulfur adsorption tower 4, a hydrogen removal tower 5, a gas dryer 6, a gas dispersion and extraction assembly 7, an adsorption and purification unit 11, a vacuum pump 13, and a product gas storage cylinder group 14. A central controller 15 is connected to the pressure equalization tank 2, the pressurization mechanism 3, the gas dispersion and extraction assembly 7, and the adsorption and purification unit 11. The central controller 15 uses a programmable PLC and connects to all key sensors and actuators to achieve intelligent control of the system. The mixture of geothermal water and associated gas from the geothermal well first enters the gas-liquid separator 1 to remove free water and droplets carried in the associated gas. The associated gas then enters the pressure equalization tank 2, where short-term fluctuations in the gas source pressure and flow are buffered and absorbed, stabilizing the outlet gas. The stabilized associated gas then enters the pressurization mechanism 3. The central controller 15 adjusts the speed of its variable frequency drive end based on the outlet pressure of the pressure equalization tank 2 and the outlet flow rate of the booster mechanism 3, thus boosting the gas pressure to 1.5–2.5 MPa. The pressurized gas then sequentially passes through the sulfur adsorption tower 4, the hydrogen removal tower 5, and the gas dryer 6. (Continue referring to...) Figure 2 In this embodiment, the gas dispersion extraction component 7 includes a raw gas distribution manifold 17 connected to the output end of the gas dryer 6, and several membrane modules 9 arranged in a rectangular grid matrix. The raw gas inlet of each membrane module 9 is connected to the raw gas distribution manifold 17 via an inlet hose. A servo motor-driven gas inflow regulating valve 18 is connected in series on each inlet hose. A temperature sensor is installed inside each membrane module 9. The first outlet of each membrane module 9 is connected to an outlet hose. A permeate gas outlet regulating valve 19, a pressure sensor 20, and a helium concentration sensor 16 are sequentially arranged on the outlet hose along the airflow direction. The outlet hoses are connected in parallel and then connected to the permeate gas manifold 8. The second outlet of each membrane module 9 is connected to a stagnant gas hose. The stagnant gas hoses are connected in parallel and then connected to the stagnant gas manifold 10. The outlet end of the stagnant gas manifold 10 is connected to the input end of the pressure equalization tank 2. The purified feed gas is distributed to each membrane module 9 via the feed gas distribution manifold 17. Within the membrane module 9, helium preferentially permeates through the membrane into the permeate gas side, and is collected along the outlet hose via the permeate gas outlet regulating valve 19, pressure sensor 20, and helium concentration sensor 16 to the permeate gas main manifold 8. At this point, the helium is initially concentrated to 10%–50%. The gas that fails to permeate the membrane (residual gas) is collected via the residual gas hose to the residual gas main manifold 10, and then sent back to the pressure equalization tank 2 to mix with the fresh associated gas before re-entering the extraction process.

[0022] Continue to refer to Figure 1The adsorption purification unit 11 is equipped with three molecular sieve packed towers 12, each equipped with a vacuum pump 13. The input of the adsorption purification unit 11 is connected to the permeate gas manifold 8 via a programmable switching valve assembly. The outlet of each molecular sieve packed tower 12 is connected to the vacuum pump 13, which in turn is connected to the product gas storage cylinder group 14. Helium-rich permeate gas in the permeate gas manifold 8 is distributed to the molecular sieve packed towers 12 in the adsorption state via the programmable switching valve assembly. Impurity gases are adsorbed by the molecular sieves, and high-purity helium gas enters the product gas storage cylinder group 14 from the outlet via the vacuum pump 13. Adjacent molecular sieve packed towers 12 alternate between adsorption and desorption operations, and the programmable switching valve assembly controls the opening and closing of each valve according to the timing set by the central controller 15. In this embodiment, the central controller 15 collects feedback signals from various sensors and actuators in real time, including the outlet pressure of the pressure equalization tank 2, the outlet flow rate of the booster mechanism 3, the outlet temperature and humidity of the gas dryer 6, the feed gas inflow rate of each membrane module 9, the pressure behind the membrane, the internal temperature and helium concentration of the permeate gas, the gas flow rate and helium concentration of the inlet pipeline of the adsorption purification unit 11, and the internal pressure of each molecular sieve packed tower 12. It also sends control commands to each actuator, including the rotation speed of the variable frequency drive end of the booster mechanism 3, the opening degree of the gas inflow regulating valve 18 and the permeate gas outlet regulating valve 19 of each membrane module 9, the switching sequence of the programmable switching valve group, and the rotation speed of the vacuum pump 13. The central controller 15 adjusts the speed of the variable frequency drive end of the booster mechanism 3 according to the outlet pressure of the pressure equalization tank 2 and the outlet flow rate of the booster mechanism 3 to smooth downstream pressure fluctuations; it adjusts the opening of the gas inflow regulating valve 18 and the permeate outlet regulating valve 19 of each membrane module 9 according to the feed gas inflow, membrane back pressure, internal temperature and permeate helium concentration, so that the separation performance of each membrane module 9 tends to be balanced; it dynamically adjusts the speed of the vacuum pump 13 and the switching sequence of the programmable switching valve group according to the gas flow rate and helium concentration of the inlet pipeline of the adsorption purification unit 11; and it controls the programmable switching valve group to switch between adsorption and desorption operations according to the internal pressure of each molecular sieve packed tower 12.

[0023] During system operation, the central controller 15 continuously reads the parameters of the helium concentration sensor 16, pressure sensor 20, and temperature sensor of each membrane module 9, calculates the ratio of the helium concentration in the permeate gas to the feed helium concentration in real time, and compares it with the initial separation coefficient to calculate the relative deviation. Since the membrane material will plasticize or age with use, when the relative deviation of a certain membrane module 9 exceeds 15%, the central controller 15 determines that the performance of the membrane module 9 has deteriorated, and then outputs a control signal to the gas inflow regulating valve 18 of the membrane module 9 to gradually decrease the opening (e.g., decrease by 10% per minute), and at the same time outputs a control signal to the gas inflow regulating valve 18 of other membrane modules 9 adjacent to the faulty membrane module 9 to increase the opening proportionally, so as to smoothly transfer the load of the faulty membrane module 9 to the adjacent healthy membrane module 9 while keeping the total processed gas volume constant. After the gas inflow regulating valve 18 of the membrane module 9 is reduced to zero, the central controller 15 closes the gas inflow regulating valve 18 and the permeate outlet regulating valve 19 of the membrane module 9, completely physically isolating them, and outputs a maintenance reminder signal through the human-machine interface, so that the operator can maintain or replace the membrane module without shutting down the system.

[0024] An opening degree coordinated control model is integrated inside the central controller 15. In order to avoid the severe oscillation of system pressure caused by valve adjustment, and to overcome the oscillation problem caused by multivariable coupling in traditional PID control, the input parameters of the opening degree coordinated control model include the total flow rate of raw gas entering the gas dispersion extraction component 7, the helium concentration in the raw gas, the raw gas inflow rate of each membrane component 9, the pressure behind the membrane of each membrane component 9, the internal temperature of each membrane component 9, and the helium concentration of the permeate gas of each membrane component 9. The output parameters are the opening coefficient of the gas inflow regulating valve 18 and the opening coefficient of the permeate gas outlet regulating valve 19 of each membrane component 9. The opening degree coordinated control model is trained offline: a massive amount of historical data from the gas dispersion extraction component 7 under different operating conditions is collected to form a training sample set. Feature vectors and corresponding actual regulating valve opening coefficients are extracted from the training sample set. Process experts optimize and label the regulating valve opening coefficients based on the extracted feature vectors. The labeled feature vectors are used as the model input, and the optimized and labeled regulating valve opening coefficients are used as the model output for iterative training until the error between the model output opening coefficients and the opening coefficients labeled by human experts converges to a preset range. In the online control phase, the central controller 15 collects the input parameters under the current operating conditions in real time and converts them into feature vectors, which are then input into the opening degree coordinated control model to obtain the opening coefficients of the gas inflow regulating valve 18 and the permeate outlet regulating valve 19 of each membrane component 9. Based on these opening coefficients, corresponding opening degree adjustment signals are output to each regulating valve to achieve synchronous fine-tuning of multiple valves.

[0025] like Figure 3As shown, the central controller 15 also integrates a state extraction module and a pressure prediction module. Since the flow rate and helium concentration of the permeate gas output from the membrane separation are not constant but exhibit temporal evolution characteristics, the state extraction module continuously collects the instantaneous gas flow rate and helium volume concentration at the output end of the permeate gas manifold 8, forming time-series data. It extracts the temporal coupling characteristics of flow rate and concentration according to a preset sliding time window, outputting a comprehensive feature vector reflecting the full-time evolution of the feed gas. The pressure prediction module receives this comprehensive feature vector, predicts through regression, and outputs the optimal desorption pressure setpoint that the adsorption purification unit 11 should maintain in the next time window. It also sends a speed adjustment signal to the vacuum pump 13, causing the actual desorption pressure to dynamically change with this optimal setpoint. Specifically, the state extraction module consists of a Long Short-Term Memory (LSTM) neural network, a differential mapping unit, and a multi-head self-attention feature fusion layer. LSTM extracts the local temporal correlation feature vector for each time window; the differential mapping unit performs differential calculation on the local temporal correlation feature vectors of adjacent time windows to capture the changing trend of the operating conditions and obtain the temporal interaction correlation feature vector; the multi-head self-attention feature fusion layer weightedly fuses the temporal interaction correlation feature vectors of all time windows to generate a comprehensive feature vector. This structure greatly improves the model's sensitivity to sudden fluctuations in operating conditions. The pressure prediction module consists of a regression prediction network composed of multiple fully connected layers stacked together. Its input is the comprehensive feature vector, and its output is the optimal desorption pressure setpoint. For example, when the feed concentration becomes high and stable, there is no need for an excessively high vacuum degree, and the system predicts the optimal desorption pressure to be 20 kPa; when the feed concentration deteriorates, the predicted optimal value is 10 kPa. The controller adjusts the speed of the vacuum pump 13 accordingly. Compared with the traditional method of maintaining an extremely low pressure, this mechanism saves the power of the vacuum pump 13.

[0026] The adsorption purification unit 11 is equipped with three molecular sieve packed towers 12. The central controller 15 monitors the internal pressure of the three molecular sieve packed towers 12 in real time. When any molecular sieve packed tower 12 completes the adsorption operation, the central controller 15 closes the raw material gas inlet of the current molecular sieve packed tower 12 by controlling the programmable switching valve group and sequentially performs multi-stage pressure equalization operation; then, the exhaust port of the current molecular sieve packed tower 12 is opened to discharge residual gas along the raw material gas inlet direction; then, the current molecular sieve packed tower 12 is connected to the vacuum pump 13 and evacuated to a vacuum state, so that the adsorbed impurity gas is desorbed and extracted by the vacuum pump 13; finally, the current molecular sieve packed tower 12 is connected to the product gas outlet of other molecular sieve packed towers 12 in the adsorption stage, or connected to an external product gas source, to introduce high-pressure product gas to raise the pressure inside the tower back to the initial pressure value required for the adsorption operation. In this embodiment, the multi-stage pressure equalization operation includes a first-stage pressure equalization operation and a second-stage pressure equalization operation. After the first molecular sieve-filled tower 12 completes the high-pressure adsorption stage, a large amount of unadsorbed helium remains. The first-stage pressure equalization operation involves connecting the currently adsorbed molecular sieve-filled tower 12 with the first target molecular sieve-filled tower 12, which is in a low-pressure state, to bring their pressures to a first intermediate pressure value. The second-stage pressure equalization operation involves connecting the currently adsorbed molecular sieve-filled tower 12 with the third molecular sieve-filled tower 12, which is in a low-pressure state, to bring their pressures to a second intermediate pressure value, which is lower than the first intermediate pressure value. Through two stages of successive depressurization, the high helium concentration gas remaining in the first molecular sieve-filled tower 12 is gradually recovered into the target molecular sieve-filled tower 12 and the third molecular sieve-filled tower 12, avoiding helium loss caused by excessive pressure difference during single-stage pressure equalization. The central controller 15 dynamically adjusts the duration of each stage of the multi-stage pressure equalization operation based on real-time helium concentration data collected from the inlet pipeline of the adsorption and purification unit 11. Since even slight fluctuations in the helium content of associated geothermal gas can significantly impact the final recovery rate, when the helium concentration in the feed gas is detected to be below a preset concentration threshold, the central controller 15 increases the allocated time for both the primary and secondary pressure equalization operations. This allows sufficient time for residual helium in the bed voids to migrate towards the low-pressure tower, preventing a large amount of helium from being discharged during the venting phase and thus improving the helium recovery rate under low-concentration conditions. The central controller 15 integrates a pressure equalization time mapping module to adjust the time length of the primary and secondary pressure equalization operations. This module includes a level setting unit, a mapping query unit, and a calculation unit. The level setting unit uses the difference between the helium concentration in the feed gas and the preset standard concentration as the first parameter, and the rate of change of this difference relative to the previous moment as the second parameter. The value ranges of the first and second parameters are divided into multiple consecutive numerical levels, and the current level of the first parameter corresponds to only one current level of the second parameter.The mapping query unit constructs a mapping query table, where each entry corresponds to a unique mapping relationship between the first and second parameters, and a unique time extension coefficient is pre-written into each entry. The calculation unit calculates the current first and second parameters in real time, determines their respective numerical levels, reads the corresponding time extension coefficient from the mapping query table, and multiplies the base time lengths of the first and second pressure equalization operations by this coefficient to obtain the actual execution time. The extended pressure equalization time allows for a more thorough transfer of high-purity helium gas from the tower to other towers requiring pressurization, effectively protecting the overall helium recovery rate of the system under extremely low feed concentration conditions.

[0027] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A smart helium separation and purification system for associated gas from geothermal wells, characterized in that, include: The components arranged sequentially along the gas flow direction include a gas-liquid separator, a pressure equalization tank, a pressurization mechanism, a sulfur adsorption tower, a hydrogen removal tower, a gas dryer, a gas dispersion and extraction assembly, an adsorption and purification unit, a vacuum pump, and a product gas storage cylinder assembly. And a central controller that is connected to the pressure equalization tank, the pressurization mechanism, the gas dispersion and extraction component, and the adsorption and purification unit respectively; The gas dispersion and extraction component includes: The raw material gas distribution main pipe is connected to the output end of the gas dryer; Several membrane modules are arranged in a matrix. The feed gas inlet of each membrane module is connected to the feed gas distribution manifold through an inlet hose, and a gas flow rate regulating valve is connected in series on each inlet hose. A temperature sensor is installed inside each membrane module. Each membrane module has a first outlet connected to an outlet hose. A permeate gas outlet regulating valve, a pressure sensor, and a helium concentration sensor are sequentially arranged on the outlet hose along the airflow direction. Each outlet hose is connected in parallel and then connected to the permeate gas main manifold. Each membrane module has a second outlet connected to a stagnation gas hose, and each of the stagnation gas hoses is connected in parallel to a common stagnation gas manifold. The outlet end of the common stagnation gas manifold is connected to the input end of the pressure equalization tank, and is used to send the collected stagnation gas back to the pressure equalization tank to mix with the associated gas. The adsorption and purification unit is equipped with at least two molecular sieve packed towers, each of which is equipped with a vacuum pump. The molecular sieve packed towers are used to perform adsorption or desorption operations, and adjacent molecular sieve packed towers alternately perform adsorption and desorption operations. The input end of each adsorption and purification unit is connected to the permeate gas manifold through a programmable switching valve group, and the outlet of each adsorption and purification unit is connected to the vacuum pump. The vacuum pump is connected to the product gas storage cylinder group. The central controller is configured to adjust the rotational speed of the variable frequency drive end of the pressurizing mechanism based on the outlet pressure of the pressure equalization tank and the outlet flow rate of the pressurizing mechanism; adjust the opening degree of the gas inflow regulating valve and the permeate gas outlet regulating valve of the membrane module based on the feed gas inflow rate, membrane back pressure, internal temperature and permeate gas helium concentration of each membrane module; dynamically adjust the rotational speed of the vacuum pump and the switching sequence of the programmable switching valve group based on the gas flow rate and helium concentration on the inlet pipeline of the adsorption purification unit; and control the programmable switching valve group to switch between adsorption and desorption operations based on the internal pressure of each molecular sieve packed tower.

2. The intelligent helium separation and purification system for associated gas from geothermal wells according to claim 1, characterized in that, The membrane module is arranged in a rectangular grid. The central controller calculates the relative deviation between the current separation coefficient and the initial separation coefficient of each membrane module based on the parameters of the helium concentration sensor, pressure sensor and temperature sensor of the membrane module. When the relative deviation exceeds a preset threshold, it is determined that the performance of the current membrane module has deteriorated. The central controller outputs a first control command to gradually reduce the opening of the gas inflow regulating valve of the current membrane module, and proportionally increase the opening of the gas inflow regulating valve of other membrane modules adjacent to the current membrane module. After the opening of the gas inflow regulating valve of the current membrane module is reduced to zero, the central controller outputs a second control command to close the gas inflow regulating valve and the permeate gas outlet regulating valve of the current membrane module, and outputs a maintenance reminder signal.

3. The intelligent helium separation and purification system for associated gas from geothermal wells according to claim 2, characterized in that, The central controller integrates an opening degree collaborative control model. The input parameters of the opening degree collaborative control model include: the total flow rate of the raw gas entering the gas dispersion extraction component, the helium concentration in the raw gas, the raw gas inflow rate of each membrane module, the pressure behind the membrane of each membrane module, the internal temperature of each membrane module, and the helium concentration of the permeate gas of each membrane module. The output parameters of the opening degree collaborative control model are the opening coefficient of the gas inflow regulating valve and the opening coefficient of the permeate outlet regulating valve for each membrane module. The training process of the opening degree collaborative control model is as follows: historical data of the gas dispersion extraction component under different operating conditions are collected to form a training sample set; feature vectors and corresponding actual regulating valve opening coefficients are extracted from the training sample set; human experts optimize and label the regulating valve opening coefficients according to the extracted feature vectors; the labeled feature vectors are used as the input of the opening degree collaborative control model; and the optimized and labeled regulating valve opening coefficients are used as the output of the opening degree collaborative control model for iterative training until the error between the opening coefficients output by the model and the opening coefficients labeled by the human experts converges to a preset range.

4. The intelligent helium separation and purification system for associated gas from geothermal wells according to claim 3, characterized in that, The central controller collects input parameters under the current operating conditions in real time and converts them into feature vectors, which are then input into the opening degree collaborative control model to obtain the opening degree coefficients of the gas inflow regulating valve and the permeate outlet regulating valve for each membrane module. The central controller outputs a third control command based on the opening coefficient to adjust the gas inflow regulating valve and the permeate outlet regulating valve of each membrane module.

5. The intelligent helium separation and purification system for associated gas from geothermal wells according to claim 4, characterized in that, The central controller integrates a state extraction module and a pressure prediction module. The state extraction module is used to continuously collect the instantaneous gas flow rate and helium volume concentration at the output end of the permeable gas manifold, form time series data, extract the temporal coupling characteristics of flow rate and concentration according to a preset time window, and output a comprehensive feature vector that reflects the evolution law of the feed gas in the whole time domain. The pressure prediction module receives the comprehensive feature vector, outputs the optimal desorption pressure setpoint that the adsorption purification unit should maintain in the next time window through regression prediction, and sends a speed adjustment signal to the vacuum pump so that the actual desorption pressure changes dynamically with the optimal desorption pressure setpoint.

6. The intelligent helium separation and purification system for associated gas from geothermal wells according to claim 5, characterized in that, The state extraction module includes: Long Short-Term Memory (LSTM) neural networks are used to extract local temporal correlation feature vectors for each time window; The differential mapping unit is used to perform differential calculation on the local temporal correlation feature vectors of adjacent time windows to obtain the temporal interaction correlation feature vector; A multi-head self-attention feature fusion layer is used to weight and fuse the temporal interaction-related feature vectors of all time windows to generate the comprehensive feature vector; The pressure prediction module includes a regression prediction network composed of multiple fully connected layers stacked together. Its input is the comprehensive feature vector, and its output is the optimal desorption pressure setting value.

7. The intelligent helium separation and purification system for associated gas from geothermal wells according to claim 6, characterized in that, The adsorption and purification unit is equipped with three molecular sieve packed towers; The central controller monitors the internal pressure of the three molecular sieve packed towers in real time. Once any molecular sieve packed column has completed its adsorption operation, the central controller is configured to perform the following operations: By controlling the programmable switching valve group to close the feed gas inlet of the current molecular sieve packed tower and sequentially performing multi-stage pressure equalization operations; Open the exhaust port of the current molecular sieve packed tower and discharge the residual gas along the direction of the raw material gas inlet; Connect the current molecular sieve packed tower to a vacuum pump and evacuate it to a vacuum state so that the adsorbed impurity gas is desorbed and extracted. Connect the current molecular sieve packed tower to the product gas outlet of other molecular sieve packed towers in the adsorption stage, or connect it to an external product gas source to introduce high-pressure product gas and raise the pressure inside the tower back to the initial pressure value required for adsorption operation.

8. The intelligent helium separation and purification system for associated gas from geothermal wells according to claim 7, characterized in that, The multi-stage pressure equalization operation includes a first-stage pressure equalization operation and a second-stage pressure equalization operation. The first-stage pressure equalization operation process is as follows: Connect the current molecular sieve packed tower to another molecular sieve packed tower that is in a low-pressure state, so that the pressure of the two molecular sieve packed towers reaches the first intermediate pressure value. The secondary pressure equalization operation process is as follows: Connect the current molecular sieve packed tower to the third molecular sieve packed tower, which is in a low-pressure state, so that the pressure of the two molecular sieve packed towers reaches a second intermediate pressure value that is lower than the first intermediate pressure value. The initial pressure of the second molecular sieve packed tower is higher than that of the third molecular sieve packed tower.

9. A smart helium separation and purification system for associated gas from geothermal wells according to claim 8, characterized in that, The duration of each stage of the multi-stage pressure equalization operation is dynamically adjusted by the central controller based on real-time helium concentration data collected from the inlet pipeline of the adsorption and purification unit: when the helium concentration in the feed gas is detected to be lower than the preset concentration threshold, the central controller extends the duration of the first-stage and second-stage pressure equalization operations.

10. A smart helium separation and purification system for associated gas from geothermal wells according to claim 9, characterized in that, The central controller integrates a pressure equalization time mapping module, which is used to adjust the duration of the first-level pressure equalization operation and the second-level pressure equalization operation. The equalization time mapping module includes: The gear setting unit uses the difference between the helium concentration in the feed gas and the preset standard concentration as the first parameter, and the rate of change of the difference relative to the previous moment as the second parameter; the value ranges of the first parameter and the second parameter are divided into multiple consecutive numerical gears, and the current gear of the first parameter has only one corresponding current gear of the second parameter. A mapping query unit is used to construct a mapping query table. Each entry in the mapping query table corresponds to a unique mapping relationship between the first parameter and the second parameter, and a unique time extension coefficient is pre-written in each entry. The calculation unit is used to calculate the current first parameter and second parameter in real time, determine the numerical level of the first parameter and second parameter respectively, and read the corresponding time extension coefficient from the mapping lookup table to adjust the time length of the first-level pressure equalization operation and the second-level pressure equalization operation.

Images