An inertia optimization method for coordinating the electro-thermal-gas energy transport channel

Abstract

The present application provides an inertia optimization method for coordinating the electro-thermal-gas energy transport channels, which relates to the technical field of energy transport channels and includes: establishing a mapping relationship model between the fluid flow state and the eddy current intensity to predict the possible eddy current situations under different working conditions; based on the prediction results, using a genetic algorithm to optimize the sensor layout scheme to obtain the most comprehensive flow state information with the fewest sensors. The present invention takes into account the safety of pipeline vibration and sets up a vibration monitoring and valve adjustment limit mechanism. In addition, the present invention also constructs a trade-off model between energy loss and transport efficiency, uses a multi-objective optimization algorithm to solve the optimal adjustment strategy. Finally, an eddy current suppression effect evaluation model is established to realize the quantitative evaluation of the control effect and the optimization of the strategy. Through the comprehensive application of this series of innovative methods, the present invention realizes the precise suppression of eddy currents and the significant improvement of transport efficiency during the fluid transport process, while ensuring the safe operation of the pipeline system.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to an inertia optimization method for coordinating the electro-thermal-gas energy transport channel. Background Art

[0002] During the high-speed flow of the fluid in the electro-thermal-gas energy transport channel, due to the viscous and inertial effects of the fluid, eddy currents are likely to be generated in areas such as the bends, valves, and variable diameters of the pipeline. The generation of eddy currents will lead to an increase in energy loss during the energy transport process, reducing the transport efficiency. At the same time, the existence of eddy currents will also exacerbate the wear of the inner wall of the pipeline, shortening the service life of the pipeline. Eddy currents may also cause vibrations and noises in the pipeline, posing a threat to the safe and stable operation of the pipeline.

[0003] In order to suppress the generation and development of eddy currents, it is necessary to arrange sensors in the pipeline to monitor the flow state of the fluid in real time, and dynamically adjust the valve opening or take other measures according to the monitoring data to change the flow characteristics of the fluid. However, how to select the arrangement position and quantity of sensors to obtain sufficient flow state information without affecting normal transport, although adjusting the valve opening can change the flow characteristics of the fluid, how to balance the relationship between the adjustment amplitude and the adjustment frequency to minimize the interference to the transport process while suppressing eddy currents, and how to adaptively adjust the control strategy according to the changes in parameters such as the type, temperature, and pressure of the fluid to ensure the control effect, these are all key technical problems to be solved urgently.

[0004] There is a contradiction between eddy current suppression and transport efficiency improvement. Too frequent or excessive adjustment can effectively suppress eddy currents, but may change the transport parameters too much, resulting in a decrease in transport efficiency. How to find the best balance point between eddy current suppression and efficiency improvement is a challenging optimization problem. In addition, actively suppressing eddy currents may introduce new energy consumption. How to select appropriate control mechanisms and control strategies to achieve the optimal control effect at the lowest cost is also a problem worthy of in-depth study. Summary of the Invention

[0005] The present invention provides an inertia optimization for coordinating the electro-thermal-gas energy transport channel, mainly including:

[0006] Establish a mapping relationship model between fluid flow states and vortex intensity to predict possible vortex situations under different working conditions; based on the prediction results, use the genetic algorithm to optimize the sensor layout scheme to obtain the most comprehensive flow state information with the fewest sensors; use the optimized sensor layout scheme to collect real-time flow state data, and apply the reinforcement learning algorithm to optimize the valve adjustment strategy to suppress vortices and improve the transport efficiency; according to different fluid types, temperature and pressure conditions, use the strategy model to generate the optimal valve adjustment scheme in real time; monitor the pipeline vibration caused by valve adjustment in real time, and trigger the safety mechanism to limit the valve adjustment amplitude when the vibration amplitude exceeds the preset threshold; construct a trade-off model between energy loss and transport efficiency, and use the multi-objective optimization algorithm to solve the model to obtain the optimal adjustment strategy that balances vortex suppression and transport efficiency improvement; establish a vortex suppression effect evaluation model to quantitatively evaluate the vortex suppression effect, and if the effect does not meet the expectation, start the secondary optimization of the adjustment strategy.

[0007] Further, the establishment of the mapping relationship model between fluid flow states and vortex intensity includes: obtaining physical property data such as the density, viscosity, and specific heat capacity of the fluid; using computational fluid dynamics software for numerical simulation to obtain flow state data such as the velocity distribution and pressure distribution of the flow field, as well as vortex characteristic data such as vorticity distribution and turbulence intensity; preprocessing the data and extracting key features as the model input;

[0008] Adopt the support vector regression algorithm or the multi-layer perceptron neural network algorithm to train and establish the mapping relationship model between flow states and vortex intensity.

[0009] Further, the optimization of the sensor layout scheme using the genetic algorithm includes: establishing a numerical simulation model of the pipeline flow state to obtain the flow state distribution information inside the pipeline; using the clustering algorithm to divide different flow state regions and determine the key monitoring regions; constructing the encoding of the sensor layout scheme; designing the fitness function, comprehensively considering the number of sensors, measurement coverage quality, cost, and vibration sensor layout; using the genetic algorithm to iteratively search for the optimal layout scheme.

[0010] Further, the optimization of the valve adjustment strategy using the reinforcement learning algorithm includes: constructing a simulation environment to simulate the fluid transport process; setting the optimization goal and quantifying it as a reward function; using the deep reinforcement learning algorithm to construct an agent model; the agent continuously interacts with the environment in the simulation environment to learn the optimal control strategy; deploying the optimal control strategy to the actual system and dynamically adjusting the valve opening according to real-time data.

[0011] Further, the real-time monitoring of the pipeline vibration caused by valve regulation includes: obtaining the vibration signals collected by vibration sensors; extracting features from the vibration data and calculating the vibration amplitude; comparing the vibration amplitude with a preset threshold; if the vibration amplitude exceeds the threshold, triggering a safety protection mechanism to limit the valve regulation amplitude, and reducing the pipeline vibration by optimizing the valve regulation strategy.

[0012] Further, the construction of the trade-off model between energy loss and transport efficiency includes: establishing a mathematical model for the trade-off optimization between eddy current suppression and transport efficiency; using a particle swarm optimization algorithm or a genetic algorithm to solve the model to obtain the Pareto front or the Pareto optimal solution set; analyzing the optimization results to obtain the optimal regulation strategy for balancing eddy current suppression and transport efficiency.

[0013] Further, the establishment of the eddy current suppression effect evaluation model includes: obtaining the eddy current-related data collected by sensors; preprocessing the data; calculating the percentage reduction of eddy current intensity, the percentage increase of transport efficiency, and the percentage reduction of energy loss; weighted summing the three indicators to obtain a comprehensive evaluation score; if the score is lower than the preset threshold, triggering the secondary optimization of the regulation strategy and using a particle swarm optimization algorithm to optimize the eddy current control parameters.

[0014] The technical solution provided by the embodiment of the present invention may include the following beneficial effects:

[0015] The present invention discloses an inertia optimization method for the coordination of an electro-thermal-gas energy transport channel. The method first establishes a mapping model between fluid parameters and eddy current intensity, and uses a genetic algorithm to optimize the sensor layout scheme to achieve efficient information collection. Then, based on the collected real-time data, a reinforcement learning algorithm is applied to optimize the valve regulation strategy to dynamically generate control schemes adapted to different working conditions. During this process, the present invention considers the pipeline vibration safety and sets up a vibration monitoring and valve regulation limit mechanism. In addition, the present invention also constructs a trade-off model between energy loss and transport efficiency, and uses a multi-objective optimization algorithm to solve the optimal regulation strategy. Finally, an eddy current suppression effect evaluation model is established to realize the quantitative evaluation of the control effect and the optimization of the strategy. Through the comprehensive application of this series of innovative methods, the present invention realizes the precise suppression of eddy currents and the significant improvement of transport efficiency during the fluid transport process, while ensuring the safe operation of the pipeline system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flowchart of an inertia optimization method for the coordination of an electro-thermal-gas energy transport channel of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] To further understand the content of the present invention, the present invention will be described in detail with reference to the accompanying drawings and embodiments. The following further elaborates on the present application with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the relevant invention and are not intended to limit the invention. Additionally, it should be noted that for ease of description, only the parts related to the invention are shown in the drawings.

[0018] As Figure 1 , a specific inertia optimization method for coordinating the electro-thermal-gas energy transport channel in this embodiment may specifically include:

[0019] S101. Establish a mapping relationship model between the fluid flow state and the eddy current intensity according to the fluid type, temperature, and pressure parameters. This model uses machine learning algorithms such as support vector regression or neural networks. Input the fluid parameters and output the predicted eddy current intensity. The model training data is sourced from experimental measurements and computational fluid dynamics simulations. This model is used to predict the possible eddy current situations under different working conditions, providing a decision-making basis for sensor arrangement and valve regulation.

[0020] According to parameters such as the fluid type, fluid temperature, and fluid pressure, obtain the physical property data of the fluid such as density, viscosity, and specific heat capacity by referring to engineering manuals and material property tables. Use computational fluid dynamics software such as ANSYS Fluent or OpenFOAM to perform numerical simulations on the target fluid under different working conditions to obtain flow state data such as the velocity distribution and pressure distribution of the flow field, as well as eddy current characteristic data such as vorticity distribution and turbulence intensity. Preprocess the obtained flow state and eddy current characteristic data, including removing outliers, interpolating to fill in missing values, normalizing, etc., and extract key features such as the mean vorticity and mean turbulence intensity as model inputs. Adopt the support vector regression algorithm, with the flow state characteristics as inputs and the mean eddy current intensity as outputs, to train and establish a mapping relationship model between the flow state and the eddy current intensity. Support vector regression can well handle small sample and non-linear problems, and the model complexity is low. Evaluate the prediction accuracy of the support vector regression model on the test set. If the root mean square error is greater than a preset threshold such as 5%, then use the multi-layer perceptron neural network algorithm to retrain the model and optimize the network structure and hyperparameters to improve the prediction accuracy. Package the trained mapping relationship model as a standard Python function or RESTful interface and integrate it into the monitoring system of the fluid device to receive fluid parameters in real time and call the model to predict the eddy current intensity. According to the predicted value of the eddy current intensity, combined with the performance parameters and applicable conditions of the eddy current suppression measures, use the rule engine to automatically decide the sensor arrangement scheme and valve regulation strategy, and send the execution instructions to the on-site control system to achieve the optimal control of fluid flow and reduce the adverse effects of eddy currents on the equipment.

[0021] Specifically, the fluid type, temperature, and pressure are key parameters that determine the physical properties of the fluid. For example,

[0022] For water, at a temperature of 20 °C and a pressure of 1 standard atmosphere, the density is approximately 998 kg / m 3 , the viscosity is approximately 0.02 mPa·s, and the specific heat capacity is approximately 1.8 kJ / (kg·K). These data can be obtained by referring to the "Fluid Mechanics Handbook" or online databases. The physical properties of different types of fluids vary greatly. For example, the density of air at the same temperature and pressure is approximately 2 kg / m 3, the viscosity is approximately 0.18 mPa·s. Obtaining accurate physical property data is the basis for fluid dynamics simulation. Using computational fluid dynamics (CFD) software, the flow state of fluids under different working conditions can be simulated. For example, when simulating the flow of water in a pipeline, boundary conditions such as the geometric shape of the pipeline, the inlet velocity of water, and the outlet pressure can be set. By solving with software such as ANSYS Fluent or OpenFOAM, data such as the velocity distribution, pressure distribution, vorticity distribution, and turbulence intensity of the flow field can be obtained. For example, at the pipe elbow, the flow velocity may increase, the pressure may decrease, while the vorticity increases and the turbulence intensity also increases. These data intuitively show the flow state and vortex characteristics of the fluid. Before data analysis, the original data needs to be preprocessed. For example, there may be some outliers in the simulation results that need to be removed. For missing data, interpolation methods can be used to complete it. To eliminate the influence of dimensions between different physical quantities, the data needs to be normalized, for example, scaling the data between 0 and 1. The preprocessed data can be used to establish a prediction model. For example, the mean value of vorticity and the mean value of turbulence intensity can be extracted as the input features of the model. Support vector regression (SVR) is a commonly used machine learning algorithm that can be used to establish the mapping relationship between the flow state and the vortex intensity. For example, taking the mean values of flow velocity, pressure, and vorticity as inputs and the mean value of turbulence intensity as the output, an SVR model is trained. The SVR model can handle non-linear relationships and also has good generalization ability for small sample data. If the root mean square error of the model on the test set is greater than a preset threshold (such as 5%), the multi-layer perceptron (MLP) neural network algorithm can be considered to retrain the model, and the prediction accuracy can be improved by adjusting the network structure and hyperparameters. The MLP neural network has stronger learning ability and can fit more complex non-linear relationships. Encapsulating the trained model as a Python function or a RESTful interface can be conveniently integrated into the monitoring system of fluid devices. For example, sensors collect real-time data of the temperature, pressure, and flow velocity of the fluid, and the monitoring system inputs these data into the model to obtain the predicted value of the vortex intensity. According to the predicted value of the vortex intensity, combined with the performance parameters and applicable conditions of the vortex suppression measures, for example, installing a deflector in the pipeline can reduce the intensity of the vortex, and the rule engine can automatically determine the installation position and angle of the deflector according to the predicted vortex intensity, or adjust the opening of the valve to control the flow velocity, thereby reducing the adverse effects of the vortex on the equipment. For example, the vortex can cause pipeline vibration and energy loss. And the execution instruction is sent to the on-site control system, such as PLC, to achieve the optimal control of fluid flow. For example, in the water inlet pipeline of a water turbine, the vortex will reduce the efficiency of the water turbine and cause vibration. By predicting the vortex intensity, measures can be taken in advance, such as adjusting the angle of the guide vane, to reduce the influence of the vortex, thereby improving the efficiency of the water turbine and extending its service life.Another example is in chemical pipelines, where eddy currents can cause local corrosion and scaling. By predicting the eddy current intensity, the pipeline design can be optimized or fluid parameters such as flow rate and temperature can be adjusted to mitigate the impact of eddy currents, thereby improving the safety and reliability of the pipeline.

[0023] S102. Based on the above model prediction results, use the genetic algorithm to optimize the sensor layout scheme. Construct a fitness function that includes indicators such as the number of sensors, their positions, information acquisition quality, and cost. The specific form of the fitness function is a weighted sum, and the weights of each indicator are determined by expert experience. The algorithm iteratively searches for the optimal layout scheme through crossover and mutation operations to achieve obtaining the most comprehensive flow state information with the fewest sensors. During this process, the layout of vibration sensors is also considered to monitor the pipeline vibration.

[0024] According to the pipeline structure and fluid characteristic data, establish a numerical simulation model of the pipeline flow state to obtain the flow state distribution information inside the pipeline. For the simulated flow state distribution, use the K-means clustering algorithm to divide different flow state regions and determine the key monitoring regions. According to the characteristics of each monitoring region, determine the types and quantities of sensors such as pressure, flow rate, and temperature that need to be arranged, and construct the encoding of the sensor layout scheme. Design a fitness function that comprehensively considers indicators such as the number of sensors, measurement coverage quality, and cost, and uses the form of weighted summation to quantify the advantages and disadvantages of the layout scheme. The weight coefficients can be assigned according to the importance of the indicators. At the same time, according to the vibration characteristics of the pipeline, introduce an evaluation index for the layout of vibration sensors into the fitness function to balance the needs of flow monitoring and vibration monitoring. Use the genetic algorithm to optimize the sensor layout scheme, set appropriate population size, crossover probability, and mutation probability, and iteratively search for the optimum through selection, crossover, and mutation operations. Judge the convergence condition of the genetic algorithm. For example, when the change range of the fitness value is less than a certain threshold for several consecutive iterations, it is considered that the algorithm has converged, and output the optimal sensor layout scheme at this time. Apply the optimized sensor layout scheme to the actual pipeline system, install sensors at the corresponding positions, and connect data acquisition equipment to achieve online monitoring of the pipeline flow state and vibration.

[0025] Specifically, the numerical simulation of the pipeline flow state is the basis for the sensor layout scheme. First, a geometric model needs to be established based on the CAD model of the pipeline. For example, a straight pipe with a diameter of 1 meter and a length of 10 meters. Then, set the fluid property parameters. For example, the density of water is 998 kg / m³ and the viscosity is 0.01 Pa·s. Next, set the boundary conditions. For example, the inlet velocity is 1 m / s and the outlet pressure is 1 standard atmospheric pressure. Using computational fluid dynamics (CFD) software, such as ANSYS Fluent, to solve this model, the velocity field, pressure field, and temperature field distributions inside the pipeline can be obtained. The simulation results show that the flow velocity is higher in the central region of the pipeline and lower in the region near the wall, showing an obvious boundary layer effect. After obtaining the detailed flow state distribution information inside the pipeline, the K-means clustering algorithm can be used to divide the flow state into regions. For example, according to parameters such as flow velocity, pressure, and temperature, the flow state inside the pipeline is divided into three regions: the high-speed region, the transition region, and the low-speed region. The division results can be visually displayed on the visualization interface of the simulation results, with different colors representing different flow state regions. The high-speed region is mainly concentrated in the center of the pipeline, and the low-speed region is near the pipe wall. This division method can help determine the areas that need to be monitored keyly. For example, at the junction of the high-speed region and the transition region, the fluid state changes violently, and unstable phenomena such as eddies are likely to occur, which need to be focused on. After determining the key monitoring areas, the appropriate type and number of sensors need to be selected according to the characteristics of each area. For example, in the high-speed region, an electromagnetic flowmeter can be selected to measure the flow velocity, a pressure sensor to measure the pressure, and a thermocouple to measure the temperature. In the transition region, due to the violent change of the flow state, the number of sensors can be increased, and sensors with a faster response speed can be selected. In the low-speed region, the number of sensors can be appropriately reduced. The sensor layout scheme can be represented by a code. For example, 10110 means that pressure, temperature, flow, temperature, and pressure sensors are arranged at five monitoring points respectively. To evaluate the quality of the sensor layout scheme, a fitness function needs to be designed. This function comprehensively considers indicators such as the number of sensors, the measurement coverage quality, and the cost. For example, the fitness function can be defined as: Fitness = 5 * Coverage - 3 * Cost - 2 * Number. Among them, the coverage rate represents the proportion of the sensor measurement range in the entire pipeline area, the cost represents the cost of the sensor and its installation and maintenance, and the number represents the total number of sensors. The weight coefficients reflect the importance of each indicator and can be adjusted according to actual needs. For example, if more attention is paid to the measurement coverage quality, the weight coefficient of the coverage rate can be increased. Pipeline vibration is also an important monitoring indicator. Therefore, an evaluation indicator for the layout of vibration sensors needs to be introduced into the fitness function. For example, according to the modal analysis results of pipeline vibration, the position and number of vibration sensors can be determined, and the coverage rate of vibration measurement can be used as part of the fitness function. This can balance the needs of flow monitoring and vibration monitoring.The genetic algorithm is a commonly used optimization algorithm that can be used to find the optimal sensor layout scheme. First, appropriate population size, crossover probability, and mutation probability need to be set. For example, the population size is set to 50, the crossover probability is set to 8, and the mutation probability is set to 1. Then, through operations such as selection, crossover, and mutation, the optimal solution is continuously searched iteratively. For example, the selection operation can retain individuals with higher fitness values, the crossover operation can combine the encodings of two individuals, and the mutation operation can randomly change some bits in the individual encoding. When the change range of the fitness value is less than a certain threshold, such as 01, for several consecutive iterations, it is considered that the algorithm has converged, and the optimal sensor layout scheme at this time is output. Finally, the optimized sensor layout scheme is applied to the actual pipeline system. For example, pressure sensors, flow sensors, temperature sensors, and vibration sensors are installed on the pipeline according to the scheme. These sensors are connected to data acquisition devices and transmit data to the monitoring center to achieve online monitoring of the pipeline flow state and vibration. Through real-time monitoring data, abnormal situations in the pipeline, such as eddy currents, leaks, and excessive vibration, can be detected in a timely manner, and corresponding measures can be taken to ensure the safe operation of the pipeline.

[0026] S103. Use the optimized sensor layout scheme to collect real-time flow state data. Combine these data and apply a reinforcement learning algorithm to optimize the valve adjustment strategy. Construct a simulation environment to simulate the fluid transportation process, with the optimization goals of suppressing eddy currents and improving transportation efficiency. The reinforcement learning agent continuously tries different valve opening adjustment schemes in the simulation environment to learn the optimal control strategy.

[0027] According to parameters such as the geometric structure of the pipeline and fluid properties, a three-dimensional simulation model of the fluid transport process is constructed using computational fluid dynamics software such as ANSYS Fluent. A standardized flow state data set is imported into the simulation model for boundary condition setting and initialization. Optimization objectives are set in the simulation environment, including minimizing the pressure loss in the pipeline, maximizing the fluid transport efficiency, etc., which are quantified into a numerical reward function as the optimization basis for the reinforcement learning algorithm. A deep reinforcement learning algorithm, such as the Deep Deterministic Policy Gradient (DDPG) algorithm, is used to construct an agent model, with the input being the flow state data and the output being the valve opening adjustment strategy. The agent continuously interacts with the environment in the simulation environment, obtains the reward feedback according to the optimization objective, uses the neural network approximation value function and policy function, and updates the model parameters through the gradient descent algorithm to gradually learn the optimal control strategy. The optimal control strategy obtained from training is deployed to the control module of the actual fluid transport system, and the valve opening is dynamically adjusted using the trained agent model according to the real-time collected flow state data. During the actual application process, the operating state and performance indicators of the system, such as pipeline pressure, flow rate, transport efficiency, etc., are continuously monitored and compared with the benchmark values before optimization to evaluate the optimization effect. At the same time, the operating data of the actual system are regularly transmitted back to the simulation environment for retraining and optimization of the agent model to achieve continuous iterative update of the strategy and adapt to changes in the actual working conditions. To further improve the optimization effect and system robustness, an online learning mechanism can be considered, that is, during the operation of the actual system, the agent model is fine-tuned and updated according to the real-time collected data so that it can dynamically adapt to changes in working conditions and external disturbances. In addition, safety thresholds and constraint conditions, such as the upper limit of pipeline pressure and flow rate range, can be set to ensure that the execution of the optimization strategy will not affect the safe and stable operation of the system.

[0028] Specifically, the pipe geometry and fluid properties are the basis for building the simulation model. For example, we can envision a straight pipe that is 10 meters long and 5 meters in diameter, transporting water at room temperature, with a water density of 998 kilograms per cubic meter and a viscosity of 0.01 Pascal-seconds. These parameters will be used to establish a three-dimensional pipe model in CFD software such as ANSYS Fluent and define the physical properties of the fluid. A standardized flow state dataset is used to set boundary conditions and initialize the simulation. For example, the inlet boundary condition can be set as a velocity inlet, and the velocity value comes from the dataset; the outlet boundary condition can be set as a pressure outlet, and the pressure value also comes from the dataset. The initial condition can be set as the pipe being filled with stationary fluid, and its temperature, pressure, and other parameters come from the dataset. This can make the simulation closer to the actual working conditions. The optimization goal needs to be quantified into a numerical reward function. Taking the minimization of pressure loss as an example, the pressure difference between the pipe inlet and outlet can be used as the negative value of the reward function. The smaller the pressure difference, the higher the reward value. Taking the maximization of fluid transport efficiency as an example, the mass or volume of fluid passing through the pipe per unit time can be used as the positive value of the reward function. The larger the transport volume, the higher the reward value. The DDPG algorithm is a deep reinforcement learning algorithm suitable for control problems in continuous action spaces, such as the adjustment of valve openings. The input of the agent model can be sensor data such as pressure and flow velocity at multiple positions inside the pipe, and the output is the valve opening value, which continuously varies between 0 and 1. The interaction process between the agent and the simulation environment is as follows: The agent outputs a valve opening control strategy based on the current flow state data; the simulation environment calculates the new flow state based on the valve opening and feeds back the new state and reward value to the agent; the agent updates the neural network parameters based on the reward value and learns a better control strategy. The trained control strategy can be deployed to the actual system. For example, the trained neural network model can be embedded in the control system to collect data such as pressure and flow velocity inside the pipe in real time and calculate the optimal valve opening based on this data. The feedback of the actual system operation data can be used for retraining and optimizing the model. For example, every once in a while, the flow state data and the corresponding control strategy collected by the actual system are uploaded to the server for updating the agent model. This can make the model adapt to the changes in the actual working conditions and improve the control performance. The online learning mechanism can further improve the optimization effect and system robustness. For example, the incremental learning method can be adopted to fine-tune the agent model according to the real-time data during the operation of the actual system, enabling it to quickly adapt to the changes in working conditions and external disturbances. The setting of safety thresholds and constraints can ensure the safe and stable operation of the system. For example, the upper limit of the pipe pressure can be set. If the valve opening calculated by the control strategy will cause the pipe pressure to exceed the upper limit, the opening will be restricted to ensure the safety of the pipe. Similarly, the range of flow velocity can be set to prevent the system from malfunctioning due to too high or too low flow velocity.

[0029] S104. Based on the policy model trained by reinforcement learning, for different fluid types, temperatures, and pressure conditions, a valve adjustment strategy is dynamically generated. During the actual transportation process, according to the current working condition data collected by sensors, the policy model is used to generate an optimal adjustment plan in real time, realizing the adaptive adjustment of the control strategy and improving the accuracy and real-time performance of vortex suppression.

[0030] According to the historical working condition data of the transportation pipeline, a deep reinforcement learning algorithm, such as DQN or DDPG, is used to train the valve adjustment strategy model to obtain the initial policy model. The input of the model is the state data of the fluid in the pipeline, such as temperature and pressure, and the output is the valve opening adjustment value. During the actual transportation process, the state data of the fluid is collected in real time through temperature sensors and pressure sensors installed in the pipeline. The collected real-time state data is input into the trained policy model, and through forward propagation calculation, the optimal valve opening adjustment value under the current state is obtained. According to the valve opening adjustment value generated by the policy model, the valve actuator is controlled to adjust the valve in real time, changing the valve opening, so as to adjust the flow rate of the fluid in the pipeline and achieve the purpose of suppressing vortexes. Before and after the valve adjustment, the state parameters of the fluid in the pipeline, such as temperature and pressure, are recorded respectively. The state parameters before and after the adjustment are compared and analyzed, and according to the preset vortex evaluation indexes, such as temperature standard deviation and pressure pulsation frequency, the suppression effect of the valve adjustment strategy on vortexes is judged. According to the quality of the suppression effect, the valve adjustment strategy is scored as the reward value of reinforcement learning. The new state data and the scoring reward value after the valve adjustment are input into the reinforcement learning algorithm, and through optimization methods such as gradient descent, the parameters of the policy model are updated and optimized to make it adapt to the changes in the actual transportation working conditions. Repeat the above processes of data collection, strategy generation, valve adjustment, effect evaluation, and model update, so that the valve adjustment strategy model can be adaptively adjusted according to the real-time changes in the working conditions of the transportation pipeline, continuously improving the accuracy and real-time performance of vortex suppression and ensuring the safe and stable operation of the transportation pipeline.

[0031] Specifically, deep reinforcement learning can be used to optimize the valve adjustment strategy of the transportation pipeline, suppress eddy currents, and improve transportation efficiency. Deep reinforcement learning algorithms represented by DQN and DDPG can handle complex continuous action spaces, such as the adjustment of valve openings. First, an initial policy model needs to be trained using historical operating condition data. Suppose there is a long-distance pipeline for transporting crude oil, and the historical data includes the crude oil temperature, pressure in the pipeline under different operating conditions, and the corresponding optimal valve openings. These data can be used to train a neural network model, where the input of the model is the crude oil temperature and pressure, and the output is the valve opening. For example, when the crude oil temperature is 50 degrees Celsius and the pressure is 5 MPa, the valve opening output by the model is 7. This initial model provides a starting point for the subsequent reinforcement learning process. Next, during the actual transportation process, the state data of the fluid in the pipeline needs to be collected in real time. Suppose distributed temperature sensors and pressure sensors are installed on the pipeline, and the crude oil temperature and pressure at different positions in the pipeline can be monitored in real time. For example, the sensors collect data once every second and transmit the data to the control system. The real-time data collected will be input into the trained policy model. For example, the currently collected crude oil temperature is 60 degrees Celsius and the pressure is 6 MPa. These two values are input into the policy model, and the model will output a valve opening, such as 8. The control system adjusts the valve actuator to adjust the valve in real time according to the valve opening value output by the policy model. For example, the control system sends an instruction to the valve actuator to adjust the valve opening to 8. Before and after the valve adjustment, the state parameters of the fluid in the pipeline are recorded respectively. For example, before the adjustment, the crude oil temperature is 60 degrees Celsius and the pressure is 6 MPa, and after the adjustment, the crude oil temperature becomes 58 degrees Celsius and the pressure becomes 5 MPa. By comparing the state parameters before and after the adjustment, the suppression effect of the valve adjustment strategy on eddy currents can be evaluated. Eddy current evaluation indicators are preset, such as temperature standard deviation and pressure pulsation frequency. Suppose the temperature standard deviation before the adjustment is 2 degrees Celsius and the pressure pulsation frequency is 10 Hz, and after the adjustment, the temperature standard deviation is 1 degree Celsius and the pressure pulsation frequency is 5 Hz. This indicates that the valve adjustment strategy effectively suppresses eddy currents. According to the quality of the suppression effect, the valve adjustment strategy is scored as the reward value for reinforcement learning. For example, if the eddy current suppression effect is good, a positive reward, such as +1, is given; if the eddy current suppression effect is poor, a negative reward, such as -1, is given. In this example, since the eddy currents are effectively suppressed, a reward of +1 is given. The new state data and the scoring reward value after the valve adjustment are input into the reinforcement learning algorithm to update and optimize the parameters of the policy model. For example, the gradient descent algorithm is used to update the parameters of the neural network model to make the model better adapt to the changes in the actual operating conditions. By repeating the above processes of data collection, policy generation, valve adjustment, effect evaluation, and model update, the policy model will continuously learn and improve, and finally be able to make adaptive adjustments according to the real-time changes in the operating conditions of the transportation pipeline, improving the accuracy and real-time performance of eddy current suppression.Doing so can improve the transportation efficiency, reduce energy consumption, and ensure the safe and stable operation of the transportation pipeline. The reason for adopting deep reinforcement learning is that it can handle complex non-linear relationships and can autonomously learn the optimal control strategy without the need to preset rules.

[0032] S105. Real-time monitor the pipeline vibration caused by valve adjustment. When the vibration amplitude exceeds the preset threshold, trigger the safety mechanism to limit the valve adjustment amplitude and avoid pipeline safety problems caused by overly aggressive adjustment.

[0033] Obtain the pipeline vibration data caused by valve adjustment. Perform Fourier transform on the time-domain vibration signal collected by the vibration sensor to obtain the frequency-domain vibration data. Extract features from the frequency-domain vibration data, calculate the characteristic values such as vibration frequency and amplitude, and obtain the pipeline vibration amplitude value. Compare the pipeline vibration amplitude value with the preset vibration amplitude threshold in real time to determine whether the current pipeline vibration amplitude exceeds the preset threshold. If the pipeline vibration amplitude exceeds the preset threshold, trigger the safety protection mechanism. The safety protection mechanism includes: calculate the limit coefficient of the valve adjustment amplitude according to the degree of exceeding the threshold, and the limit coefficient is proportional to the degree of exceeding the threshold. Multiply the limit coefficient by the current valve adjustment amplitude to obtain the limited valve adjustment amplitude, and use it as the upper limit value of the valve adjustment. In the case of limited valve adjustment, optimize the valve adjustment strategy to reduce the pipeline vibration. The specific method is: perform low-pass filtering on the adjustment signal of the valve to filter out the high-frequency adjustment components and obtain a smooth adjustment signal curve. Generate valve control commands according to the smooth adjustment curve, so that the valve adjusts slowly according to the smooth adjustment curve to avoid sudden increases in pipeline vibration caused by rapid opening and closing of the valve. Continuously monitor the pipeline vibration situation. When the pipeline vibration amplitude is lower than the preset threshold for multiple consecutive times, gradually relax the limit on the valve adjustment amplitude in fixed steps until the limit is completely cancelled. Obtain valve control parameters such as valve opening and adjustment speed, and use them together with the pipeline vibration data at the corresponding time as sample data. Use the support vector machine (SVM) algorithm to train the sample data and establish a non-linear correlation model between the valve control parameters and the pipeline vibration. Use the trained SVM model to predict the pipeline vibration situation under different combinations of valve control parameters according to the current pipeline vibration state. Select the combination of valve control parameters with the smallest predicted vibration amplitude as the current optimal control strategy, and generate valve control commands according to this control strategy. At the same time, use the SVM model to predict the pipeline vibration trend and dynamically adjust the vibration amplitude threshold to make the threshold setting adapt to the current pipeline working conditions.

[0034] Specifically, pipeline vibration is an important safety indicator during the operation of a transportation pipeline. Excessive vibration may lead to pipeline fatigue or even rupture. To ensure pipeline safety, it is necessary to monitor and control the pipeline vibration caused by valve regulation in real time. First, vibration sensors are installed on the pipeline to collect pipeline vibration signals. For example, acceleration sensors are installed on the pipeline near the valve to measure the vibration acceleration of the pipeline in real time. The signals collected by the sensors are time-domain signals, which reflect the change of pipeline vibration over time. To analyze the frequency components of the vibration signals, it is necessary to convert the time-domain signals into frequency-domain signals. The fast Fourier transform (FFT) can be used to convert the time-domain vibration signals into frequency-domain vibration data. The frequency-domain data can show the vibration components at different frequencies and their amplitudes. Feature extraction is performed on the frequency-domain vibration data, such as calculating the main frequency of the vibration and the corresponding amplitude. Suppose that after Fourier transform, it is found that the main frequency component of the pipeline vibration is 50 Hz, and the corresponding amplitude is 5 mm. This amplitude value of 5 mm is the pipeline vibration amplitude value of concern. The extracted vibration amplitude value is compared with a preset vibration amplitude threshold. The preset threshold is determined according to factors such as pipeline material and operating pressure. For example, suppose that according to the pipeline safety specifications, the vibration amplitude threshold for this pipeline is set at 8 mm. Since the current vibration amplitude of 5 mm is less than the threshold of 8 mm, the pipeline vibration is within the safe range. If the pipeline vibration amplitude exceeds the preset threshold, the safety protection mechanism is triggered. Suppose that at a certain moment, due to the rapid opening of the valve, the pipeline vibration amplitude reaches 2 mm, exceeding the preset threshold of 8 mm. At this time, the safety protection mechanism is triggered. The safety protection mechanism first calculates the limit coefficient of the valve regulation amplitude. The limit coefficient is proportional to the degree of exceeding the threshold. For example, suppose that the vibration amplitude exceeds the threshold by 50% ((2 - 8) / 8 = 5), then the limit coefficient is set at 5. Multiply the limit coefficient by the current valve regulation amplitude to obtain the limited valve regulation amplitude. Suppose that the current valve regulation amplitude is 10%, then the limited regulation amplitude is 5 * 10% = 5%. That is to say, the valve opening regulation cannot exceed 5%. In the case of limited valve regulation, it is necessary to optimize the valve regulation strategy to reduce pipeline vibration. The regulation signal of the valve can be processed by low-pass filtering to filter out high-frequency regulation components. This is equivalent to smoothing the valve regulation to avoid rapid opening and closing of the valve. For example, the valve control signal can be passed through a low-pass filter with a cut-off frequency of 1 Hz, which can filter out control signals with frequencies higher than 1 Hz and make the valve regulation smoother. Continuously monitor the pipeline vibration situation. When the pipeline vibration amplitude is lower than the preset threshold for multiple consecutive times, gradually relax the limit on the valve regulation amplitude. For example, if the vibration amplitude is lower than 8 mm for 5 consecutive times, the limit coefficient is reduced from 5 to 4, and so on, until the limit is completely cancelled. To establish a correlation model between valve control parameters and pipeline vibration, it is necessary to collect sample data. The sample data includes valve control parameters such as valve opening and regulation speed, as well as pipeline vibration data at corresponding times.For example, record that the pipeline vibration amplitude is 6 mm when the valve opening is 50% and the adjustment speed is 2% per second. Use the collected sample data to train a Support Vector Machine (SVM) model. The SVM model can learn the non-linear relationship between valve control parameters and pipeline vibration. Using the trained SVM model, the pipeline vibration conditions under different combinations of valve control parameters can be predicted. For example, the pipeline vibration amplitude when the valve opening is 60% and the adjustment speed is 1% per second can be predicted. Select the combination of valve control parameters with the smallest predicted vibration amplitude as the current optimal control strategy. For example, if it is predicted that the vibration amplitude is the smallest when the valve opening is 55% and the adjustment speed is 5% per second, then this parameter combination is used as the current optimal control strategy, and the corresponding valve control instructions are generated. At the same time, the SVM model can also predict the pipeline vibration trend and dynamically adjust the vibration amplitude threshold according to the prediction results. For example, if it is predicted that the pipeline vibration trend is rising, the vibration amplitude threshold can be appropriately increased to take preventive measures in advance.

[0035] S106. Construct a trade-off model between energy loss and transport efficiency. This model considers factors such as the degree of vortex suppression, fluid transport rate, and energy consumption. Apply a multi-objective optimization algorithm, such as NSGA-II (Non-dominated Sorting Genetic Algorithm II), to solve this model. The algorithm iteratively searches for the Pareto optimal solution set by setting parameters such as population size, crossover rate, and mutation rate, and obtains the optimal adjustment strategy for balancing vortex suppression and transport efficiency improvement.

[0036] According to attributes such as energy loss and transport efficiency, a mathematical model for the trade-off optimization between eddy current suppression and transport efficiency is established. In the model, an eddy current suppression factor and a transport efficiency factor are set as optimization objectives, and a mathematical relationship between them is established. The multi-objective optimization model is transformed into a standard form so that it can be solved by an optimization algorithm. The particle swarm optimization algorithm is used to solve the model. The population size is set to 50, the learning factors c1 and c2 are 0 respectively, the inertia weight w is 8, and the maximum number of iterations is 200. The particle positions and velocities are initialized, and the particle positions are updated according to the fitness function to obtain the Pareto front. The genetic algorithm is used to solve the model. The population size is set to 50, the crossover probability is 8, the mutation probability is 1, and the maximum number of generations is 200. Selection, crossover, and mutation operations are performed on the population, and the population evolves according to the fitness function to obtain the Pareto optimal solution set. The optimized Pareto solution set is analyzed. According to the eddy current suppression factor, the membership function method is used to perform a fuzzy comprehensive evaluation of the eddy current suppression degree of the solution set to obtain the fuzzy comprehensive evaluation value. According to the transport efficiency factor, a judgment matrix is constructed using the nine-scale method, and the weight vector is calculated using the sum-product method to obtain the weighted score of the transport efficiency. The fuzzy comprehensive evaluation value and the weighted score are weighted and summed, and the weights can be adjusted according to actual needs. The solution with the highest final score is the optimal adjustment strategy for balancing eddy current suppression and transport efficiency under the current weights. According to the values of the eddy current suppression factor and the transport efficiency factor corresponding to this strategy, the parameters of the actual system are adjusted to achieve the optimization effect.

[0037] Specifically, during the pipeline transportation process, due to the viscosity of the fluid itself and the friction of the pipeline wall, energy losses will occur, and eddy current is one of the main reasons for energy losses. At the same time, the transportation efficiency is also an important indicator of the pipeline transportation system. Therefore, it is necessary to balance and optimize the eddy current suppression and transportation efficiency to achieve the best transportation effect. The generation of eddy current is related to the flow velocity of the fluid and the geometric shape of the pipeline. For example, when the fluid flows through positions with large local resistance such as pipeline elbows and valves, eddy currents are likely to be generated. The intensity of the eddy current can be represented by an eddy current suppression factor, which can be defined as the ratio of the average velocity in the eddy current region to the average velocity of the pipeline centerline. The smaller the value of the eddy current suppression factor, the better the eddy current suppression effect. The transportation efficiency can be represented by a transportation efficiency factor, which can be defined as the product of the volume of fluid transported per unit time and the cross-sectional area of the pipeline. The larger the value of the transportation efficiency factor, the higher the transportation efficiency. There is a certain contradictory relationship between eddy current suppression and transportation efficiency. For example, to suppress eddy currents, the flow velocity can be reduced, but this will lead to a decrease in transportation efficiency. To improve the transportation efficiency, the flow velocity can be increased, but this will in turn lead to an increase in eddy currents. Therefore, it is necessary to establish a mathematical model to describe the relationship between the eddy current suppression factor and the transportation efficiency factor and find an optimal balance point. This mathematical model can be expressed as a multi-objective optimization problem, with the goal of simultaneously minimizing the eddy current suppression factor and maximizing the transportation efficiency factor. To simplify the model, it can be assumed that there is a linear relationship between the eddy current suppression factor and the transportation efficiency factor. For example, the eddy current suppression factor increases linearly with the increase of the transportation efficiency factor. To solve this multi-objective optimization problem, the particle swarm optimization algorithm can be used. Assume that the population size is 50, that is, there are 50 particles, and each particle represents a possible solution. The position of each particle represents a set of values of the eddy current suppression factor and the transportation efficiency factor. The particles update their positions based on their own experience and the best experience of the group to find better solutions. The genetic algorithm can also be used to solve this multi-objective optimization problem. The genetic algorithm simulates the evolution process in nature and continuously optimizes the population through operations such as selection, crossover, and mutation. Each individual in the population represents a possible solution. The fitness function is used to evaluate the quality of each individual. A set of Pareto optimal solutions can be obtained through the particle swarm optimization algorithm or the genetic algorithm. The Pareto optimal solution refers to the solution set in which no objective function value can be further improved without reducing the values of other objective functions. To select an optimal solution from the Pareto optimal solution set, it is necessary to balance the eddy current suppression and transportation efficiency according to actual needs. For example, if more attention is paid to eddy current suppression, a solution with a smaller eddy current suppression factor can be selected; if more attention is paid to transportation efficiency, a solution with a larger transportation efficiency factor can be selected. To evaluate the degree of eddy current suppression, the membership function method can be used. The membership function maps the eddy current suppression factor to a value between 0 and 1, indicating the degree of eddy current suppression.For example, the smaller the eddy current suppression factor, the larger the membership function value, indicating a better eddy current suppression effect. To evaluate the transport efficiency, the nine-scale method can be used to construct a judgment matrix. The judgment matrix represents the relative importance between different transport efficiency factors. Then, the sum-product method is used to calculate the weight vector to obtain the weighted score of the transport efficiency. Finally, the weighted sum of the fuzzy comprehensive evaluation value of eddy current suppression and the weighted score of transport efficiency is calculated to obtain the final score. The solution with the highest score is the optimal adjustment strategy that balances eddy current suppression and transport efficiency under the current weights. For example, assume that the eddy current suppression factor of a certain Pareto optimal solution is 2 and the transport efficiency factor is 8. The fuzzy comprehensive evaluation value of eddy current suppression calculated through the membership function is 9. The weighted score of transport efficiency calculated through the sum-product method of the nine-scale method is 7. Assume that the weights of eddy current suppression and transport efficiency are 6 and 4 respectively. Then the final score of this solution is 6*9 + 4*7 = 82. According to the eddy current suppression factor and transport efficiency factor values corresponding to the optimal adjustment strategy, the parameters of the actual system can be adjusted. For example, parameters such as the pipe cross-sectional area, fluid flow rate, and pipe geometry can be adjusted to achieve an optimized effect. Assume that the eddy current suppression factor corresponding to the optimal adjustment strategy is 1 and the transport efficiency factor is 9. Then the pipe cross-sectional area can be appropriately reduced to reduce the eddy current intensity, and at the same time, the fluid flow rate can be appropriately increased to improve the transport efficiency.

[0038] S107. Establish an eddy current suppression effect evaluation model to compare and analyze the changes in the fluid flow state before and after adjustment. This model is based on sensor data and calculates indicators such as the percentage reduction in eddy current intensity, the degree of improvement in transport efficiency, and energy loss. Quantitatively evaluate the eddy current suppression effect. If the effect does not meet the expectation, initiate the secondary optimization of the adjustment strategy to further improve the control effect.

[0039] Obtain the eddy current-related data collected by the sensor, including parameters such as flow rate, pressure, temperature, etc., store the obtained data and construct an eddy current dataset. Preprocess the eddy current dataset, use the moving average filtering method to remove outliers and noise, and use the maximum-minimum normalization method to normalize the data to improve the data quality. According to the data subsets before and after the implementation of the eddy current suppression measures, calculate the average eddy current velocity magnitude before and after the implementation, and obtain the percentage reduction in eddy current velocity as a quantitative index of eddy current suppression intensity. By analyzing the changes in transport efficiency-related parameters such as flow rate and pipeline pressure difference before and after eddy current suppression, calculate the percentage increase in transport efficiency as a quantitative index of transport efficiency. Calculate the fluid kinetic energy loss and thermal energy loss before and after eddy current suppression, and obtain the percentage reduction in energy loss as a quantitative index of energy loss. Assume that the weight coefficients of the percentage reduction in eddy current intensity, the percentage increase in transport efficiency, and the percentage reduction in energy loss are 4, 3, and 3 respectively, and sum the three indicators weighted by the weights to obtain the comprehensive evaluation score of the eddy current suppression effect. If the comprehensive evaluation score is lower than the preset threshold (such as 80 points), then trigger the secondary optimization of the adjustment strategy. Use the particle swarm optimization algorithm, with the goal of improving the comprehensive evaluation score, and optimize the eddy current control parameters, such as the inclination angle of the swirl plate and the layout spacing of the suppression device, through iterative search. After multiple rounds of iteration, obtain the optimal parameter combination that makes the comprehensive evaluation score reach the expected goal.

[0040] Specifically, sensors can collect parameters such as the flow velocity, pressure, and temperature of the fluid in the pipeline for constructing an eddy current dataset. For example, a series of sensors can be arranged in the pipeline, such as ultrasonic flow meters, pressure sensors, and temperature sensors, to monitor the state of the fluid in real time. The data from these sensors can be transmitted to a data acquisition system for storage and preprocessing to form an eddy current dataset. To improve the data quality, it is necessary to preprocess the collected eddy current dataset. For example, the moving average filtering method can be used to remove outliers and noise from the data. Suppose there are some spikes in the collected flow velocity data, which may be caused by momentary errors of the sensors, and they can be smoothed out by moving average filtering. In addition, the maximum-minimum normalization method can be adopted to normalize the data, scaling the data range to between 0 and 1 to avoid the influence of data with different dimensions and facilitate subsequent analysis and modeling. To evaluate the effect of eddy current suppression, it is necessary to calculate a quantitative index of eddy current suppression intensity. For example, data is collected for a period of time before and after implementing the eddy current suppression measures, and the average magnitude of the eddy current velocity is calculated. Suppose the average eddy current velocity before implementation is 1 meter per second, and the average eddy current velocity after implementation is 0.5 meter per second, then the percentage reduction in the eddy current velocity is 50%, indicating that the effect of eddy current suppression is relatively significant. In addition to the eddy current suppression intensity, it is also necessary to evaluate the change in transport efficiency. The transport efficiency can be measured by parameters such as flow rate and pipeline pressure difference. For example, suppose after implementing the eddy current suppression measures, the flow rate increases by 10% and the pipeline pressure difference decreases by 5%, then the percentage increase in the transport efficiency can be comprehensively calculated based on the changes in these two parameters. Energy loss is also an important evaluation index. Eddy current suppression measures can reduce fluid kinetic energy loss and thermal energy loss. For example, suppose after implementing the eddy current suppression measures, the fluid kinetic energy loss is reduced by 8% and the thermal energy loss is reduced by 3%, then the percentage reduction in energy loss can be comprehensively calculated based on the changes in these two parameters. To comprehensively evaluate the effect of eddy current suppression, the percentage reduction in eddy current intensity, the percentage increase in transport efficiency, and the percentage reduction in energy loss can be weighted and summed. For example, if the weight coefficients of the three indicators are set to 4, 3, and 3 respectively, then the comprehensive evaluation score of the eddy current suppression effect can be calculated. Suppose the percentage reduction in eddy current intensity is 50%, the percentage increase in transport efficiency is 10%, and the percentage reduction in energy loss is 5%, then the comprehensive evaluation score is 4×50% + 3×10% + 3×5% = 245 points. If the comprehensive evaluation score is lower than a preset threshold, such as 80 points, then the adjustment strategy needs to be optimized again. For example, the particle swarm optimization algorithm can be adopted, with the goal of improving the comprehensive evaluation score, to search for the optimal eddy current control parameters. These parameters can include the inclination angle of the swirl plate, the arrangement spacing of the suppression device, etc. Through multiple rounds of iteration, the optimal parameter combination that enables the comprehensive evaluation score to reach the expected target can be found. Suppose through optimization, the inclination angle of the swirl plate is adjusted to 30 degrees and the arrangement spacing of the suppression device is adjusted to 0.5 meters, and finally the comprehensive evaluation score reaches 90 points, then it is considered that the secondary optimization is successful.

[0041] In addition, it should be noted that, in the above specific embodiments, the various specific technical features described can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not separately describe various possible combination manners. In addition, any combination can be made among the various different embodiments of the present invention, as long as it does not violate the idea of the present invention, and it should also be regarded as the content disclosed by the present invention.

Claims

1. A method for suppressing eddy current in a fluid transport pipeline, characterized in that: include: Establish a mapping model between fluid flow state and eddy current intensity to predict possible eddy currents under different working conditions; Based on the prediction results, a genetic algorithm is used to optimize the sensor layout scheme to obtain the most comprehensive flow state information with the least number of sensors; The optimized sensor arrangement scheme is used to collect real-time flow state data, and the reinforcement learning algorithm is applied to optimize the valve adjustment strategy to suppress vortex and improve transportation efficiency; Based on different fluid types, temperatures and pressure conditions, the strategy model is used to generate the optimal valve adjustment scheme in real time; Real-time monitoring of pipeline vibration caused by valve adjustment. When the vibration amplitude exceeds the preset threshold, the safety mechanism is triggered to limit the valve adjustment amplitude. A trade-off model between energy loss and transport efficiency is constructed, and a multi-objective optimization algorithm is applied to solve the model to obtain the optimal regulation strategy that balances eddy current suppression and transport efficiency improvement. A vortex suppression effect evaluation model is established to quantitatively evaluate the vortex suppression effect. If the effect does not meet expectations, the secondary optimization of the adjustment strategy is initiated.

2. The method for suppressing eddy current in a fluid transport pipeline according to claim 1, characterized in that: The establishing of the mapping relationship model between the fluid flow state and the eddy current intensity comprises: Obtain density, viscosity, and specific heat capacity data of fluids; Use computational fluid dynamics software to perform numerical simulations to obtain velocity distribution, pressure distribution data, vortex distribution, and turbulence intensity data of the flow field; Preprocess the data and extract key features as model input; A support vector regression algorithm or a multi-layer perceptron neural network algorithm is used to train and establish a mapping relationship model between flow state and eddy current intensity.

3. The method for suppressing eddy current in a fluid transport pipeline according to claim 1, characterized in that: The method of optimizing the sensor arrangement scheme by using a genetic algorithm comprises: Establish a numerical simulation model of pipeline flow state and obtain the flow state distribution information inside the pipeline; Clustering algorithms are used to divide different flow state areas and determine key monitoring areas; Construct the code for the sensor deployment plan; Design a fitness function that takes into account the number of sensors, measurement coverage quality, cost, and vibration sensor placement; Genetic algorithm is used to iteratively search for the optimal layout solution.

4. The method for suppressing eddy current in a fluid transport pipeline according to claim 1, characterized in that: The application of reinforcement learning algorithm to optimize valve regulation strategy includes: Construct a simulation environment to simulate the fluid transport process; Set the optimization goal and quantify it as a reward function; Use deep reinforcement learning algorithms to build intelligent agent models; The agent continuously interacts with the environment in the simulation environment to learn the optimal control strategy; Deploy the optimal control strategy to the actual system and dynamically adjust the valve opening according to real-time data.

5. The method for suppressing eddy current in a fluid transport pipeline according to claim 1, characterized in that: The real-time monitoring of pipeline vibration caused by valve adjustment includes: Acquire a vibration signal collected by a vibration sensor; Extract features from vibration data and calculate vibration amplitude; comparing the vibration amplitude to a preset threshold; If the vibration amplitude exceeds the threshold, the safety protection mechanism is triggered to limit the valve adjustment amplitude and reduce pipeline vibration by optimizing the valve adjustment strategy.

6. The method for suppressing eddy current in a fluid transport pipeline according to claim 1, characterized in that: The energy loss and transport efficiency trade-off model is constructed, including: Establish a mathematical model for the trade-off between eddy current suppression and transport efficiency; Use particle swarm optimization algorithm or genetic algorithm to solve the model and obtain the Pareto frontier or Pareto optimal solution set; The optimization results are analyzed to obtain the optimal regulation strategy that balances eddy current suppression and transport efficiency.

7. The method for suppressing eddy current in a fluid transport pipeline according to claim 1, characterized in that: The method of establishing an eddy current suppression effect evaluation model comprises: Obtain eddy current related data collected by sensors; Preprocess the data; Calculate the percentage reduction of eddy current intensity, the percentage increase of transport efficiency and the percentage reduction of energy loss; The three indicators are weighted and summed to obtain a comprehensive evaluation score; If the score is lower than the preset threshold, the secondary optimization of the regulation strategy is triggered, and the particle swarm optimization algorithm is used to optimize the eddy current control parameters.

8. The method for suppressing eddy current in a fluid transport pipeline according to claim 7, characterized in that: The use of a particle swarm optimization algorithm to optimize eddy current control parameters refers to using a particle swarm optimization algorithm to take improving the comprehensive evaluation score as the optimization goal, optimizing the eddy current control parameters through iterative search, and obtaining the optimal parameter combination that enables the comprehensive evaluation score to reach the expected target after multiple rounds of iterations.

9. The method for suppressing eddy current in a fluid transport pipeline according to claim 6, characterized in that: In the mathematical model for trade-off optimization of eddy current suppression and transport efficiency, an eddy current suppression factor and a transport efficiency factor are set as optimization targets; the multi-objective optimization model is converted into a standard form so that it can be solved by an optimization algorithm.

10. The method for suppressing eddy current in a fluid transport pipeline according to claim 3, characterized in that: The iterative search for the optimal arrangement scheme using a genetic algorithm comprises: Genetic algorithm is used to optimize the sensor layout scheme, and the population size, crossover probability and mutation probability are set. The optimal layout is searched iteratively through selection, crossover and mutation operations. The change amplitude of fitness value in multiple consecutive iterations is less than a certain threshold value as the convergence condition of genetic algorithm, and the optimal sensor layout plan at this time is output.

Images