A load balancing regulation method for a liquefied gas carrier liquid cargo handling system simulator

Abstract

The application provides a load balancing regulation method for a liquefied gas carrier liquid cargo operation system simulator, belonging to the technical field of transportation, comprising: first, establishing a liquid tank pressure response model and a ship state monitoring model to provide support for safe operation; deploying a real-time load monitoring module to collect performance data, and performing preprocessing and load evaluation to comprehensively grasp the system working state. On this basis, dynamic task priority adjustment and load trend prediction are adopted to ensure that critical tasks are processed in time and improve the overall efficiency of the system. Finally, based on load distribution evaluation, a dynamic load balancing strategy is formulated to realize optimal resource allocation, and real-time monitoring and adjustment are performed to respond to load changes, ensuring stable and reliable operation of the system. This scheme combines modeling analysis, real-time monitoring, dynamic scheduling and other technologies, solves the technical problems of complex coupling relationship between subsystems, load distribution imbalance which may lead to local overload or resource waste, and affects the overall performance.

Description

Technical Field

[0001] This invention belongs to the field of transportation technology, and more specifically, relates to a load balancing control method for a liquefied gas carrier cargo operating system simulator. Background Technology

[0002] Liquefied gas carriers are specialized vessels used to transport liquefied natural gas (LNG) or liquefied petroleum gas (LPG) and other flammable liquid gases. During the loading and unloading process, it is crucial to strictly control the pressure and temperature within the liquid tanks to maintain the physical state of the cargo and ensure the vessel's buoyancy and strength. Simultaneously, it is necessary to monitor and regulate the operating status of various system components, rationally allocate system resources, and improve overall operational efficiency. Currently, most LNG carrier cargo control systems employ human-machine interaction for control and monitoring, relying on the experience and judgment of operators. However, with the increasing scale and complexity of the systems, manual intervention is insufficient to meet the requirements of real-time, accurate, and stable operation.

[0003] Some research institutions and companies have developed computer simulation software for liquefied gas carrier cargo operating systems to assist in the design and optimization of loading and unloading operations. These simulation software programs generally adopt a modular structure, including functional units such as cargo tank simulation, pressure and temperature control, and ship stability calculations. However, most of these simulation systems statically simulate the loading and unloading process, lacking real-time monitoring and dynamic control of the system's operating status. In practical applications, complex coupling relationships exist between subsystems, and uneven load distribution may lead to localized overloads or resource waste, affecting overall performance. Furthermore, existing simulation software typically cannot be effectively integrated with real cargo loading and unloading equipment, making it difficult to conduct comprehensive simulation tests on actual systems. Summary of the Invention

[0004] In view of this, the present invention provides a load balancing control method for a liquefied gas carrier cargo operating system simulator, which can solve the technical problem that the complex coupling relationship between subsystems and the uneven load distribution may lead to local overload or resource waste, affecting the overall performance.

[0005] This invention is implemented as follows:

[0006] This invention provides a load balancing control method for a liquefied gas carrier cargo operating system simulator, comprising the following steps:

[0007] S10. Establish a liquid tank pressure response model: Construct a liquid tank pressure response model based on the loading volume and vapor discharge volume; Determine the parameters of the liquid tank pressure response model by experimentally measuring the pressure change data during the loading process; The liquid tank pressure response model considers the gain coefficient, time constant, loading rate, return gas rate and initial pressure, and is used to predict liquid tank pressure changes.

[0008] S20. Establish a ship condition monitoring model: Construct a ship condition monitoring model that includes parameters such as transverse stability, longitudinal stability, heel angle, and draft; the ship condition monitoring model considers the impact of changes in the weight of liquid cargo or ballast water on various parameters of the ship, and evaluates the ship's buoyancy, stability, and strength status through real-time calculation, providing safety assurance for liquid cargo operations;

[0009] S30. Deploy a real-time load monitoring module: Deploy a real-time load monitoring module in the liquefied gas carrier's liquid cargo operating system to collect data on the CPU utilization, memory usage, network bandwidth utilization, and disk read / write speed performance of each device or unit; calculate comprehensive performance indicators through weighted averages to provide basic data support for subsequent load balancing.

[0010] S40. Data Preprocessing and Load Assessment: Preprocess the collected performance data, including data cleaning and normalization; calculate the load value of each device or unit using the load assessment function, taking into account performance indicators, load change rate and random disturbance factors, and comprehensively assess the working status of each component of the liquefied gas carrier liquid cargo operating system.

[0011] S50. Task Priority Calculation and Dynamic Adjustment: Calculate the priority value of each task based on its urgency, importance, resource requirements, and waiting time; dynamically adjust task priorities considering time decay factors and random fluctuations to ensure that critical tasks are processed in a timely manner and improve the overall efficiency of the liquefied gas carrier cargo operating system.

[0012] S60. Load Trend Prediction: Employs a long short-term memory neural network model to predict future load trends based on historical load data; captures the long-term and short-term dependencies of load changes through the calculation of forget gate, input gate, output gate, and cell state, providing forward-looking guidance for load balancing strategy formulation.

[0013] S70. Load balancing strategy formulation: Based on the load balancing degree calculation formula, assess the current load distribution of the liquefied gas carrier liquid cargo operating system; considering the load differences and random disturbance factors of each device or unit, formulate a load balancing strategy to achieve optimal allocation of resources of the liquefied gas carrier liquid cargo operating system and improve overall operating efficiency.

[0014] S80. Real-time scheduling and load adjustment: Using a load change detection function, the load change of the liquefied gas carrier cargo operating system is monitored in real time; by calculating the first-order difference and second-order derivative of the load, the system can quickly respond to load changes, dynamically adjust task allocation and resource configuration, and ensure that the liquefied gas carrier cargo operating system maintains stable and efficient operation under various working conditions.

[0015] The liquefied gas carrier cargo operating system simulator adopts a three-layer architecture design, including a host computer closed loop, a host computer closed loop, and a front-end closed loop. The host computer closed loop is used to display the human-machine interface of the liquefied gas carrier loading and unloading system and to display and respond to simulated data. The host computer closed loop is used to simulate cargo loading and unloading functions and internal unit simulations based on instructions and data issued by the host computer. The front-end closed loop is used for data interaction between the simulator and the actual unit devices of the liquefied gas carrier loading and unloading system. The liquefied gas carrier cargo operating system simulator also includes an architecture interaction mechanism, which includes: a data communication protocol to ensure accurate data transmission between the host computer closed loop, the host computer closed loop, and the front-end closed loop; real-time data processing to ensure that the data processing and response time of each layer meets the system requirements; and a synchronization mechanism to achieve timestamp-based data synchronization and ensure data consistency between the host computer, the host computer, and the front-end. The liquefied gas carrier cargo operating system simulator consists of a liquefied natural gas carrier cargo hold simulation unit, an liquefied petroleum gas carrier simulation unit, a reliquefaction system simulation unit, a nitrogen production system simulation unit, and a communication module unit. It adopts a modular structure, connecting the industrial control host, data transmission module, and power system via a bus. It supports the selection of multiple data transmission modules and communicates with the host computer via an Ethernet interface. Each unit of the liquefied gas carrier cargo operating system simulator uses physical interfaces, and the transmitted signals are all real physical signals. The virtualization unit runs within the software system, receiving real input signals through data-driven or simulation-driven methods and generating response signals at the corresponding interfaces.

[0016] Furthermore, the liquid tank pressure response model is specifically represented as follows:

[0017] [z p ]=[M1 / (T1s+1)M2 / (T2s+1)][v1;v2]+[c p ];

[0018] In the formula, z p M1 represents the tank pressure, in kPa; M2 represents the gain coefficients, dimensionless, and M1 + M2 = 1; T1 and T2 represent the time constants, in seconds, and T1 + T2 = T ε v1 represents the loading rate, in meters per second (m). 3 / h; v2 is the return gas rate, in meters per second (m). 3 / h;c p ρ represents the initial pressure of the liquid tank, in kPa; s is the Laplace operator.

[0019] Parameter acquisition method: The average vapor pressure of the liquid cargo at the start of loading was measured to be 2 kPa, and the average pressure of the gas-liquid mixture in the tank after loading was completed was 12 kPa. The loading process lasted 15 hours. (T) τ=36000s. The experimental steps include: 1) preparing the liquefied gas carrier and shore loading and unloading facilities; 2) starting loading and recording the initial pressure; 3) recording the tank pressure every 30 minutes; 4) recording the final pressure after loading is completed.

[0020] The ship condition monitoring model is specifically represented as follows:

[0021]

[0022]

[0023]

[0024]

[0025]

[0026]

[0027]

[0028]

[0029]

[0030] In the formula, GM1 is the new transverse stability height, in meters (m); GM is the original transverse stability height of the ship, in meters (m); m is the weight change of liquid cargo or ballast water, in tons (t); D is the ship's displacement, in tons (t); d m The average draft of the ship is expressed in meters (m); z m ρ represents the vertical coordinates of the changing position of the liquid cargo or ballast water, in meters (m). x The density of each liquid is given in t / m³. 3 i x GM represents the moment of inertia of each liquid surface area, in m⁴. L1 For new longitudinal stability, the unit is m; GM L The original longitudinal stability of the ship is given by θ, which is in meters (m); θ is the heel angle, which is in rad; φ is the pitch angle, which is in rad; y m x represents the lateral coordinate of the changing position of liquid cargo or ballast water, in meters (m). m x represents the longitudinal coordinate of the changing position of the liquid cargo or ballast water, in meters (m). F d represents the vertical coordinate of the center of gravity, in meters (m). F The original draft of the ship, measured in meters (m); d A The stern draft of the vessel, measured in meters (m); d F1 The draft of a ship at its new bow is measured in meters (m); d A1The unit is: draft at the stern of the vessel, in meters (m); L is the length of the vessel, in meters (m); TPC is the tonnage per centimeter of draft, in tons per centimeter (t / cm); MTC is the initial trim moment per centimeter of the vessel, in tons per centimeter (t·m / cm).

[0031] Parameter acquisition method: Original ship parameters (e.g., GM, GM L , D, d F d A (etc.) Obtained from ship design drawings and stability manuals. Parameters related to liquid cargo or ballast water (such as m, x, etc.) m y m , z m (etc.) are measured in real time through a level gauge and weight calculation system.

[0032] The performance metrics of the real-time load monitoring module are calculated using the following formula:

[0033] P i =αC i +βM i +γN i +δD i +ε i ;

[0034] In the formula, P i C represents the overall performance index of the i-th device or unit, and is dimensionless; i This represents CPU utilization, with a value ranging from 0-100%; M i N represents memory usage, ranging from 0% to 100%. i This represents network bandwidth utilization, with a value ranging from 0 to 100%; D i ε represents disk read / write speed, measured in MB / s; α, β, γ, δ are weighting coefficients, dimensionless, and α + β + γ + δ = 1; i The error term follows a normal distribution N(0,σ). 2 ), where σ is the standard deviation, which is usually taken as 0.1.

[0035] Parameter acquisition method: C i M i N i D i Data is obtained in real time through system monitoring tools (such as top, iostat, etc.). The weighting coefficients α, β, γ, and δ are determined through expert experience or historical data analysis.

[0036] The load evaluation function is specifically represented as follows:

[0037]

[0038] In the formula, L iP represents the load assessment value for the i-th device or unit, dimensionless; ij w is the j-th performance index of the i-th device or unit, dimensionless; j Let be the weight of the j-th performance index, which is dimensionless and n represents the number of performance indicators; λ is the weighting coefficient for the load change rate, which is dimensionless and typically ranges from 0.1 to 0.5. The rate of change of the performance index, in units of 1 / s; μ i The random disturbance term follows a uniform distribution U(-a,a), where a is the disturbance amplitude, typically 0.05.

[0039] Parameter acquisition method: P ij This is obtained through the real-time monitoring module in step S3. j λ and λ were determined through historical data analysis and expert experience. It is obtained by calculating the difference in performance indicators between two adjacent time points and dividing by the time interval.

[0040] The specific formula for calculating task priority is as follows:

[0041]

[0042] In the formula, PR k U represents the priority value of the k-th task, which is dimensionless. k The urgency level of the task is represented by a value ranging from 1 to 10; I k The value for R represents the importance of the task, ranging from 1 to 10. k The resource requirements for the task range from 1 to 10; T k ω1, ω2, ω3, ω4 are weighting coefficients, dimensionless, and ω1 + ω2 + ω3 + ω4 = 1; λ is the time decay coefficient, in units of 1 / s, typically ranging from 0.001 to 0.01; ξ k The term is a random fluctuation term, which follows a normal distribution N(0,σ). 2 ), where σ is the standard deviation, which is usually taken as 0.1.

[0043] Parameter acquisition method: U k I k R k Specifyed by the task submitter or the system default setting. k The values ​​are calculated using the system clock. The weighting coefficients ω1, ω2, ω3, ω4, and λ are determined through historical data analysis and expert experience.

[0044] Parameter acquisition method: By collecting historical load data, an LSTM network is trained using the backpropagation algorithm to optimize the weight matrix and bias terms. The training steps include: 1) data preprocessing; 2) constructing the LSTM network structure; 3) setting the loss function and optimizer; 4) iterative training; 5) model validation and tuning.

[0045] The specific formula for calculating load balancing is as follows:

[0046]

[0047] In the formula, B represents the load balancing degree, which is dimensionless and ranges from 0 to 1, with values ​​closer to 1 indicating a more balanced load; L i The load of the i-th device or unit is dimensionless. The average load is dimensionless; m is the total number of devices or units; η is the random disturbance term, following a normal distribution N(0,σ). 2 ), where σ is the standard deviation, which is usually taken as 0.01.

[0048] Parameter acquisition method: L i It is calculated using the load evaluation function in step S4. This is the arithmetic mean of the loads of all devices or units.

[0049] The load change detection function is specifically represented as follows:

[0050]

[0051] In the formula, D(t) is the load change detection function, which is dimensionless; L i (t) represents the load of the i-th device or unit at time t, which is dimensionless; m represents the total number of devices or units; κ is the second derivative weighting coefficient, which is dimensionless and typically ranges from 0.1 to 0.5. Let be the second derivative of the load, representing the acceleration due to the change in load, with units of 1 / s². 2 ;ρ i The random fluctuation term follows a uniform distribution U(-b,b), where b is the fluctuation amplitude, usually taken as 0.05.

[0052] Parameter acquisition method: L i (t) is calculated in real time using the load evaluation function in step S4. The value is approximated by calculating the load values ​​at three adjacent time points using the central difference method. κ is determined through historical data analysis and expert experience.

[0053] Compared with existing technologies, the beneficial effects of the load balancing control method for a liquefied gas carrier cargo operating system simulator provided by this invention are:

[0054] First, this method constructs a liquid tank pressure response model and a ship condition monitoring model, accurately simulating the pressure changes and ship buoyancy during liquid cargo loading and unloading. Through analysis and modeling of actual loading data, mathematical expressions for the changes in liquid tank pressure with cargo volume, return gas volume, and initial pressure are obtained. Simultaneously, based on changes in ship parameters and liquid cargo position, comprehensive evaluation models for transverse and longitudinal stability, draft changes, etc., are established. These models provide fundamental support for the dynamic simulation and optimized control of the system.

[0055] Secondly, this method deploys a real-time load monitoring module to collect performance indicators such as CPU utilization, memory usage, network bandwidth, and disk I / O of various hardware devices and functional units in real time. A comprehensive performance index is obtained through weighted calculation, providing a basis for load assessment and load balancing. Based on this, a load assessment function is used in conjunction with performance change trends and random disturbances to comprehensively analyze the working status of each component of the system, providing a basis for dynamic load balancing.

[0056] Furthermore, this method integrates functions such as task priority calculation, load trend prediction, and load balancing strategy formulation. By considering factors such as task urgency, importance, and resource requirements, it dynamically adjusts task priorities to ensure timely response to critical tasks. Simultaneously, it employs an LSTM neural network to predict future load change trends, providing forward-looking guidance for load balancing strategy formulation. Based on the load balancing degree calculation formula, it dynamically assesses the current load distribution of the system and formulates optimization strategies to achieve rational allocation of system resources.

[0057] Finally, the method also incorporates a real-time load change detection and dynamic adjustment mechanism. By analyzing the first-order difference and second-order derivative of the load, it quickly identifies sudden load changes and adjusts task allocation and resource configuration in a timely manner to ensure stable and efficient system operation under various working conditions.

[0058] In summary, the load balancing control method for the liquefied gas carrier cargo operating system simulator proposed in this invention integrates functions such as ship status monitoring, load assessment, task priority allocation, trend prediction, and dynamic balancing based on existing simulation technology. This achieves comprehensive perception and intelligent control of the system's operating status and solves the technical problem that in practical applications, complex coupling relationships exist between subsystems, and uneven load distribution may lead to local overload or resource waste, affecting overall performance. Attached Figure Description

[0059] Figure 1 A flowchart of the method provided by the present invention;

[0060] Figure 2 This is an architecture design diagram from an embodiment;

[0061] Figure 3This is a diagram of the simulator system structure in the embodiment;

[0062] Figure 4 The strategy graph is further optimized in the example. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0064] like Figure 1 The diagram shown is a flowchart of a load balancing control method for a liquefied gas carrier cargo operating system simulator provided by the present invention. This method includes the following steps:

[0065] S10. Establish a liquid tank pressure response model: Construct a liquid tank pressure response model based on the loading volume and vapor discharge volume; Determine the parameters of the liquid tank pressure response model by experimentally measuring the pressure change data during the loading process; The liquid tank pressure response model considers the gain coefficient, time constant, loading rate, gas return rate and initial pressure, and is used to predict liquid tank pressure changes.

[0066] S20. Establish a ship condition monitoring model: Construct a ship condition monitoring model that includes parameters such as transverse stability, longitudinal stability, heel angle, and draft; The ship condition monitoring model considers the impact of changes in the weight of liquid cargo or ballast water on various ship parameters, and evaluates the ship's buoyancy, stability, and strength status through real-time calculations to provide safety assurance for liquid cargo operations;

[0067] S30. Deploy a real-time load monitoring module: Deploy a real-time load monitoring module in the liquefied gas carrier's liquid cargo operating system to collect data on the CPU utilization, memory usage, network bandwidth utilization, and disk read / write speed performance of each device or unit; calculate comprehensive performance indicators through weighted averages to provide basic data support for subsequent load balancing.

[0068] S40. Data Preprocessing and Load Assessment: Preprocess the collected performance data, including data cleaning and normalization; use the load assessment function to calculate the load value of each device or unit, taking into account performance indicators, load change rate and random disturbance factors, to comprehensively assess the working status of each component of the liquefied gas carrier liquid cargo operating system.

[0069] S50. Task Priority Calculation and Dynamic Adjustment: Calculate the priority value of each task based on its urgency, importance, resource requirements, and waiting time; dynamically adjust task priorities considering time decay factors and random fluctuations to ensure that critical tasks are processed in a timely manner and improve the overall efficiency of the liquefied gas carrier cargo operating system.

[0070] S60. Load Trend Prediction: Employs a long short-term memory neural network model to predict future load trends based on historical load data; captures the long-term and short-term dependencies of load changes through the calculation of forget gate, input gate, output gate, and cell state, providing forward-looking guidance for load balancing strategy formulation.

[0071] S70. Load balancing strategy formulation: Based on the load balancing degree calculation formula, assess the current load distribution of the liquefied gas carrier's liquid cargo operating system; considering the load differences and random disturbance factors of each device or unit, formulate a load balancing strategy to achieve optimal allocation of resources of the liquefied gas carrier's liquid cargo operating system and improve overall operating efficiency.

[0072] S80. Real-time scheduling and load adjustment: Utilizes a load change detection function to monitor load changes in the liquefied gas carrier's cargo operating system in real time; by calculating the first-order difference and second-order derivative of the load, it quickly responds to sudden load changes, dynamically adjusts task allocation and resource configuration, and ensures that the liquefied gas carrier's cargo operating system maintains stable and efficient operation under various working conditions.

[0073] The specific implementation methods of the above steps are described in detail below:

[0074] The specific implementation of step S10 is as follows: First, a liquid tank pressure response model is constructed. This model considers the influencing factors during the loading process, including the loading volume, vapor discharge, and initial pressure. By experimentally measuring the pressure change data during the loading process, the model parameters are determined, and the change law of liquid tank pressure over time is obtained. The specific mathematical expression of the liquid tank pressure response model is as follows:

[0075] [z p ]=[M1 / (T1s+1)M2 / (T2s+1)][v1;v2]+[c p ];

[0076] Among them, z p M1 represents the tank pressure, in kPa; M2 and M1 are gain coefficients, dimensionless, and M1 + M2 = 1; T1 and T2 are time constants, in seconds, and T1 + T2 = T ε v1 represents the loading rate, in meters per second (m). 3 / h; v2 is the return gas rate, in meters per second (m). 3 / h;c p ρ represents the initial pressure of the liquid tank, in kPa; s is the Laplace operator.

[0077] Experiments showed that the average pressure of the liquid cargo vapor at the start of loading was 2 kPa, and the average pressure of the gas-liquid mixture in the tank after loading was completed was 12 kPa. The loading process lasted 15 hours. (T) τ=36000s. The experimental steps include: 1) preparing the liquefied gas carrier and shore loading and unloading facilities; 2) starting loading and recording the initial pressure; 3) recording the tank pressure every 30 minutes; 4) recording the final pressure after loading is completed.

[0078] Such a liquid tank pressure response model can accurately reflect the changing pattern of liquid tank pressure during loading, providing basic support for subsequent system simulation and optimized control.

[0079] The specific implementation of step S20 is as follows: Constructing a ship condition monitoring model. This model includes key parameters such as the ship's transverse stability, longitudinal stability, heel angle, and draft, and can assess the ship's buoyancy, stability, and strength status in real time. The specific mathematical expression of the ship condition monitoring model is as follows:

[0080]

[0081]

[0082]

[0083]

[0084]

[0085]

[0086]

[0087] Wherein, GM1 is the new transverse stability height, in meters (m); GM is the original transverse stability height of the ship, in meters (m); m is the weight change due to liquid cargo or ballast water, in tons (t); D is the ship's displacement, in tons (t); d m The average draft of the ship is expressed in meters (m); z m ρ represents the vertical coordinates of the changing position of the liquid cargo or ballast water, in meters (m). x The density of each liquid is given in t / m³. 3 i x Here is the moment of inertia of each liquid surface area, in meters. 4 GM L1 For new longitudinal stability, the unit is m; GM L The original longitudinal stability of the ship is given by θ, which is in meters (m); θ is the heel angle, which is in rad; φ is the pitch angle, which is in rad; y m x represents the lateral coordinate of the changing position of liquid cargo or ballast water, in meters (m). m x represents the longitudinal coordinate of the changing position of the liquid cargo or ballast water, in meters (m). F d represents the vertical coordinate of the center of gravity, in meters (m). F The original draft of the ship, measured in meters (m); dA The stern draft of the vessel, measured in meters (m); d F1 The draft of a ship at its new bow is measured in meters (m); d A1 The unit is: draft at the stern of the vessel, in meters (m); L is the length of the vessel, in meters (m); TPC is the tonnage per centimeter of draft, in tons per centimeter (t / cm); MTC is the initial trim moment per centimeter of the vessel, in tons per centimeter (t·m / cm).

[0088] These parameters can be obtained from ship design drawings, stability manuals, and real-time measurement systems. By establishing such a ship condition monitoring model, the ship's buoyancy, stability, and strength status can be monitored in real time, providing safety assurance for liquid cargo operations.

[0089] The specific implementation of step S30 is as follows: Deploy a real-time load monitoring module. This module collects performance index data of various devices or units in the liquefied gas carrier's cargo operating system, including CPU utilization, memory usage, network bandwidth utilization, and disk read / write speed. A comprehensive performance index is derived through weighted calculation, providing basic data support for subsequent load assessment and load balancing.

[0090] Overall performance index P i The calculation formula is as follows:

[0091] P i =αC i +βM i +γN i +δD i +ε i ;

[0092] Among them, C i This represents CPU utilization, with a value ranging from 0-100%; M i N represents memory usage, ranging from 0% to 100%. i This represents network bandwidth utilization, with a value ranging from 0 to 100%; D i ε represents disk read / write speed, measured in MB / s; α, β, γ, and δ are weighting coefficients, dimensionless, and α + β + γ + δ = 1; i The error term follows a normal distribution N(0,σ). 2 ), where σ is the standard deviation, which is usually taken as 0.1.

[0093] By deploying this real-time load monitoring module, performance metrics data of each component of the system can be obtained in real time, providing an important basis for subsequent load assessment and balancing strategy formulation.

[0094] The specific implementation of step S40 is as follows: Preprocessing and load assessment are performed on the collected performance index data. First, the raw data is cleaned and normalized to eliminate dimensional differences and outliers. Then, the load value of each device or unit is calculated using a load assessment function, taking into account performance indicators, load change rate, and random disturbance factors, to comprehensively assess the working status of each component of the system.

[0095] The specific expression for the load evaluation function is as follows:

[0096]

[0097] Among them, L i P represents the load assessment value for the i-th device or unit, dimensionless; ij w is the j-th performance index of the i-th device or unit, dimensionless; j Let be the weight of the j-th performance index, which is dimensionless and n represents the number of performance indicators; λ is the weighting coefficient for the load change rate, which is dimensionless and typically ranges from 0.1 to 0.5. The rate of change of the performance index, in units of 1 / s; μ i The random disturbance term follows a uniform distribution U(-a,a), where a is the disturbance amplitude, typically 0.05.

[0098] Such a load assessment process can comprehensively reflect the working status of each component of the system, providing a basis for the formulation of subsequent load balancing strategies.

[0099] The specific implementation of step S50 is as follows: Calculate the priority value of each task based on its urgency, importance, resource requirements, and waiting time. Considering time decay and random fluctuations, dynamically adjust task priorities to ensure that critical tasks are processed in a timely manner and improve overall system efficiency.

[0100] The formula for calculating task priority is as follows:

[0101]

[0102] Among them, PR k U represents the priority value of the k-th task, which is dimensionless. k The urgency level of the task is represented by a value ranging from 1 to 10; I k The value for R represents the importance of the task, ranging from 1 to 10. k The resource requirements for the task range from 1 to 10; T k ω1, ω2, ω3, and ω4 are weighting coefficients, dimensionless, and ω1 + ω2 + ω3 + ω4 = 1; λ is the time decay coefficient, in units of 1 / s, typically ranging from 0.001 to 0.01; ξk The term is a random fluctuation term, which follows a normal distribution N(0,σ). 2 ), where σ is the standard deviation, which is usually taken as 0.1.

[0103] This task priority calculation and dynamic adjustment mechanism ensures that critical tasks are processed in a timely manner, thereby improving the overall operating efficiency of the system.

[0104] The specific implementation of step S60 is as follows: A Long Short-Term Memory (LSTM) neural network model is used to predict future load trends based on historical load data. The LSTM network can capture the long-term and short-term dependencies of load changes, providing forward-looking guidance for load balancing strategy formulation.

[0105] By collecting historical load data, an LSTM network is trained using the backpropagation algorithm to optimize the weight matrix and bias terms. The training steps include: 1) data preprocessing; 2) constructing the LSTM network structure; 3) setting the loss function and optimizer; 4) iterative training; and 5) model validation and tuning.

[0106] The specific implementation of step S70 is as follows: Based on the load balancing formula, assess the current load distribution of the system. Considering the load differences and random disturbances of each device or unit, formulate a load balancing strategy to achieve optimal allocation of system resources and improve overall operating efficiency.

[0107] The formula for calculating load balancing is as follows:

[0108]

[0109] Where B represents the load balancing degree, which is dimensionless and ranges from 0 to 1, with values ​​closer to 1 indicating a more balanced load; L i The load of the i-th device or unit is dimensionless. The average load is dimensionless; m is the total number of devices or units; η is the random disturbance term, following a normal distribution N(0,σ). 2 ), where σ is the standard deviation, which is usually taken as 0.01.

[0110] By formulating such load balancing assessment and optimization strategies, the rational allocation of system resources can be achieved, thereby improving overall operating efficiency.

[0111] The specific implementation of step S80 is as follows: A load change detection function is used to monitor system load changes in real time. By calculating the first-order difference and second-order derivative of the load, rapid response to load surges is achieved, and task allocation and resource configuration are dynamically adjusted to ensure stable and efficient system operation under various working conditions.

[0112] The specific expression for the load change detection function is as follows:

[0113]

[0114] Where D(t) is the load change detection function, which is dimensionless; L i (t) represents the load of the i-th device or unit at time t, which is dimensionless; m represents the total number of devices or units; κ is the second derivative weighting coefficient, which is dimensionless and typically ranges from 0.1 to 0.5. Let be the second derivative of the load, representing the acceleration due to the change in load, with units of 1 / s². 2 ;ρ i The random fluctuation term follows a uniform distribution U(-b,b), where b is the fluctuation amplitude, usually taken as 0.05.

[0115] Through such a load change detection and dynamic adjustment mechanism, the system can be ensured to maintain stable and efficient operation under various working conditions.

[0116] Specifically, the principle of this invention is to establish a multi-level monitoring and control mechanism to achieve comprehensive perception and dynamic optimization of the system's operating status.

[0117] First, this method constructs a tank pressure response model and a ship condition monitoring model, providing fundamental support for system simulation. The tank pressure response model considers factors such as cargo volume, vapor discharge, and initial pressure. Through analysis and modeling of actual loading data, a mathematical expression for the change of tank pressure over time is obtained. This model can accurately reflect the dynamic change law of tank pressure during the loading process.

[0118] The ship condition monitoring model includes key parameters such as lateral and longitudinal stability, heel angle, and draft. Based on changes in the weight of liquid cargo or ballast water, this model calculates various stability and strength indicators of the ship, providing safety assurance for liquid cargo operations. By monitoring the ship's buoyancy and stability in real time, abnormal situations can be detected promptly, and corresponding control measures can be taken.

[0119] Secondly, this method deploys a real-time load monitoring module in the system to collect performance metrics data from each device or unit in real time, including CPU utilization, memory usage, network bandwidth, and disk I / O. A comprehensive performance index is obtained through weighted calculation, providing a data foundation for subsequent load assessment. This real-time monitoring mechanism can comprehensively perceive the current operating status of the system, providing a basis for dynamic load balancing.

[0120] Based on this, the method employs a load assessment function, combining performance indicators, load change trends, and random disturbances to comprehensively evaluate the workload of each device or unit. This load modeling and assessment process objectively reflects the actual operating status of each component of the system, providing a basis for subsequent load balancing strategy development.

[0121] Furthermore, this method integrates functions such as task priority calculation, load trend prediction, and load balancing strategy formulation. By considering factors such as task urgency, importance, and resource requirements, it dynamically adjusts task priorities to ensure timely response to critical tasks. Simultaneously, it employs an LSTM neural network to predict future load change trends, providing forward-looking guidance for load balancing strategy formulation. Based on the load balancing degree calculation formula, it dynamically assesses the current load distribution of the system and formulates optimization strategies to achieve rational allocation of system resources.

[0122] Finally, this method incorporates a real-time load change detection and dynamic adjustment mechanism. By analyzing the first-order difference and second-order derivative of the load, it quickly identifies sudden load changes and adjusts task allocation and resource configuration in a timely manner to ensure stable and efficient system operation under various working conditions.

[0123] The following is a specific embodiment of the present invention: A large liquefied natural gas (LNG) transportation company owns a large fleet of LNG carriers, undertaking domestic and international natural gas transportation tasks. In order to improve the safety and efficiency of LNG cargo loading and unloading operations, the company decided to develop an LNG carrier cargo operating system simulator based on the present invention to comprehensively simulate and optimize the loading and unloading process.

[0124] like Figure 2 The simulator shown employs a three-layer architecture, comprising a host computer closed loop, a main unit closed loop, and a front-end closed loop. The host computer closed loop is responsible for the human-computer interaction interface and the display and response of simulated data; the main unit closed loop simulates liquid cargo loading and unloading functions and internal unit simulations based on instructions from the host computer; and the front-end closed loop is responsible for data interaction with the actual liquefied gas carrier loading and unloading system equipment. The three closed loops are interconnected through data communication protocols, real-time data processing, and synchronization mechanisms to ensure accurate data transmission and timely processing at each layer, achieving data consistency between the host computer, main unit, and front-end.

[0125] Figure 3 A schematic diagram of the overall structure of the liquefied gas carrier cargo operating system simulator is presented. As shown in the diagram, the simulator consists of a liquefied natural gas carrier cargo hold simulation unit, a liquefied petroleum gas carrier simulation unit, a reliquefaction system simulation unit, a nitrogen production system simulation unit, and a communication module unit. Each unit adopts a modular structure, connecting the industrial control host, data transmission module, and power system via a bus. It supports multiple data transmission modes and communicates with the host computer via an Ethernet interface.

[0126] In the specific implementation process, the liquefied gas carrier cargo operating system simulator mainly performs the following steps:

[0127] The first step was to construct a liquid tank pressure response model. A 15-hour pressure test was conducted on a certain type of LNG carrier during the loading process. Data were recorded, showing an average vapor pressure of 2 kPa at the start of loading and an average pressure of 12 kPa for the gas-liquid mixture inside the tank after loading was completed. Based on the experimental data, the following liquid tank pressure response model was established:

[0128] [z p ]=[0.6 / (36s+1)0.4 / (12s+1)][v1;v2]+[2];

[0129] Where, z p v1 represents the tank pressure in kPa; v2 represents the loading rate in meters per second. 3 / h; v2 is the return gas rate, in meters per second (m). 3 / h. This model takes into account factors such as loading volume, vapor discharge, and initial pressure, and can accurately reflect the change of tank pressure over time.

[0130] The second step is to construct a ship condition monitoring model. This model includes lateral stability high GM1 and longitudinal stability high GM1. L1 Heel angle θ, pitch angle φ, mean draft d m Key parameters, etc. Based on the design parameters of a certain type of LNG carrier, the following ship condition monitoring model was established:

[0131]

[0132]

[0133] Where m is the increased weight of the liquid cargo, taken as 300 tons; D is the displacement, taken as 30,000 tons; d F The draft of the original headwaters is 5.6m; d A For the original tail section, the draft is taken as 5.9m; x F The longitudinal coordinate of the center of buoyancy is taken as 50m; L is the ship's length, taken as 80m; TPC is the draft per centimeter, taken as 2.5 tons / cm; MTC is the trim moment per centimeter, taken as 800 tons·m / cm. By monitoring these parameters in real time, the ship's buoyancy stability and strength can be assessed, providing safety assurance for liquid cargo operations.

[0134] The third step is to deploy a real-time load monitoring module. This module collects performance metrics from various devices or units in the system in real time, including CPU utilization, memory usage, network bandwidth utilization, and disk read / write speed. Based on expert experience, the weighting coefficients for each metric have been determined.

[0135] α=0.4, β=0.3, γ=0.2, δ=0.1;

[0136] Substitute the collected performance data into the comprehensive performance index calculation formula:

[0137] P i =0.4C i +0.3M i +0.2N i +0.1D i +ε i ;

[0138] Where, ε i It follows a normal distribution N(0,0.01). 2 By deploying this module, the operational status of each component of the system can be monitored in real time.

[0139] The fourth step involves preprocessing and load assessment of the collected performance data. First, the raw data is cleaned and normalized to eliminate dimensional differences and outliers. Then, the load value L for each device or unit is calculated using the load assessment function. i Performance metrics, load change rate, and random disturbance factors were taken into account.

[0140]

[0141] Where, P i1 This represents the overall performance index of the i-th device or unit. μ is the rate of change of the performance index. i The load follows a uniform distribution U(-0.05, 0.05). This load assessment process can comprehensively reflect the working status of each component of the system.

[0142] Fifth, calculate the priority value (PR) for each task based on its urgency, importance, resource requirements, and waiting time. k :

[0143]

[0144] Among them, U k For mission urgency levels (1-10), I k For task importance (1-10), R k For task resource requirements (1-10), T k Task waiting time (in seconds), ξ k It follows a normal distribution N(0,0.1). 2 By dynamically adjusting task priorities, it can be ensured that critical tasks receive timely responses.

[0145] The sixth step involves using an LSTM neural network to predict future load change trends. After backpropagation training and optimization, this LSTM model can capture the long-term and short-term dependencies of load changes, providing forward-looking guidance for subsequent load balancing strategy formulation.

[0146] Step 7: Evaluate the current load distribution of the system based on load balancing factor B.

[0147]

[0148] Where, L i For the load of the i-th device or unit, Let m represent the average load and m be the total number of devices or units. Based on load prediction using the LSTM model, an optimization strategy is developed to achieve rational allocation of system resources.

[0149] Step 8: Monitor the real-time changes in system load using the load change detection function d(t):

[0150]

[0151] Once a sudden change in load is detected, task allocation and resource configuration should be adjusted immediately to ensure that the system maintains stable and efficient operation under various working conditions.

[0152] In this embodiment, to further optimize system performance and improve the simulator's response speed and resource utilization, the present invention adopts the following load balancing strategy and performance optimization measures, as shown in the strategy diagram below. Figure 4 As shown:

[0153] (1) Dynamic task allocation strategy

[0154] Real-time monitoring: The front-end closed-loop system monitors key performance indicators of devices or units in real time, including CPU utilization, memory usage, and network bandwidth.

[0155] Random Forest Prediction: This method uses the random forest algorithm to predict the future load trend of devices or units based on historical data and real-time monitoring information.

[0156] Load balancing algorithm: Uses advanced load balancing algorithms (Least Connections) to optimize task allocation and ensure balanced resource usage.

[0157] Dynamic adjustment: Based on prediction results and load balancing algorithms, task allocation is dynamically adjusted, prioritizing new tasks for devices or units with lower loads. This strategy aims to prevent single-point overload and improve overall system stability and response speed.

[0158] (2) Software simulation

[0159] Loading and unloading system simulation: Simulates the loading and unloading process of liquefied cargo on a liquefied vessel, and ensures the safety of loading and unloading by controlling the loading and unloading rate, the high-load compressor return gas rate, and the shore station return gas rate.

[0160] Full-process stability simulation: covering the entire process after the ship is put into operation, including the loading and unloading of liquid cargo and the ship's navigation.

[0161] 3D simulation view on the top of the tank: Apply virtual reality technology to the top of the tank of the liquid cargo ship loading and unloading simulator to realize a 3D visual reality and enhance the physical realism of human-computer interaction.

[0162] (3) Resource Management Optimization

[0163] Resource pool mechanism: The system uses a resource pool mechanism to manage system resources and avoid memory leaks and unnecessary resource consumption.

[0164] Waste recycling mechanism: Implement an efficient waste recycling mechanism to clean up useless resources and ensure the stability and efficient operation of the system.

[0165] Resource monitoring: Monitor system resource usage in real time, and promptly detect and address resource waste or abnormal usage.

[0166] (4) Parallel computing and acceleration technologies

[0167] Parallel computing: In host-based closed-loop simulation computing, parallel computing techniques (such as multithreading or multiprocessing) are used to improve computing speed and reduce task execution time.

[0168] Hardware acceleration: Use the CUDA hardware acceleration library to accelerate computing tasks and further optimize system response time. Select an appropriate acceleration solution for optimization based on computing requirements.

[0169] Algorithm optimization: Apply efficient parallel algorithms to improve computational performance and ensure that the system can handle a large number of concurrent tasks without degrading performance.

[0170] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A load balancing control method for a liquefied gas carrier cargo operating system simulator, characterized in that, include: S10. Establish a liquid tank pressure response model: Construct a liquid tank pressure response model based on the loading volume and vapor discharge volume; Pressure change data during the loading process were obtained through experiments to determine the parameters of the liquid tank pressure response model. The liquid tank pressure response model considers the gain coefficient, time constant, loading rate, return gas rate, and initial pressure to predict liquid tank pressure changes. S20. Establish a ship condition monitoring model: Construct a ship condition monitoring model that includes parameters such as transverse stability, longitudinal stability, heel angle, and draft; The ship condition monitoring model considers the impact of changes in the weight of liquid cargo or ballast water on various ship parameters, and evaluates the ship's buoyancy, stability, and strength status through real-time calculations to provide safety assurance for liquid cargo operations; S30. Deploy a real-time load monitoring module: Deploy a real-time load monitoring module in the liquefied gas carrier's cargo operating system to collect the CPU utilization of each device or unit. Memory usage Network bandwidth utilization and disk read / write speed ; The performance metrics of the real-time load monitoring module are calculated using the following formulas: ; In the formula, For the first The overall performance indicators of an individual device or unit These are the weighting coefficients, and , This is the error term; S40. Data Preprocessing and Load Assessment: Preprocess the collected performance data, including data cleaning and normalization; calculate the load value for each device or unit using the load assessment function, which is specifically expressed as follows: ; In the formula, For the first The load assessment value of an individual device or unit. For the first The first device or unit One performance metric, For the first The weights of each performance metric, and ; The number of performance indicators; The weighting coefficient for the load change rate. The rate of change of performance indicators, For random disturbance terms; S50. Task Priority Calculation and Dynamic Adjustment: Calculate the priority value of each task based on its urgency, importance, resource requirements, and waiting time; dynamically adjust task priorities considering time decay factors and random fluctuations to ensure that critical tasks are processed in a timely manner and improve the overall efficiency of the liquefied gas carrier's liquid cargo operating system. S60. Load Trend Prediction: Employs a long short-term memory neural network model to predict future load trends based on historical load data; captures the long-term and short-term dependencies of load changes through the calculation of forget gate, input gate, output gate, and cell state, providing guidance for load balancing strategy formulation. S70. Load Balancing Strategy Formulation: Based on the load balancing degree calculation formula, assess the current load distribution of the liquefied gas carrier's cargo operating system, formulate a load balancing strategy, and achieve optimal resource allocation for the liquefied gas carrier's cargo operating system. The load balancing degree calculation formula is as follows: ; In the formula, The load balancing score ranges from 0 to 1, with values ​​closer to 1 indicating a more balanced load. For the first The load of a device or unit; Average load; Total number of devices or units; The term represents a random disturbance and follows a normal distribution. S80. Real-time Scheduling and Load Adjustment: Utilizing a load change detection function, the system monitors real-time load changes in the liquefied gas carrier's cargo operating system, rapidly responding to sudden load fluctuations and dynamically adjusting task allocation and resource configuration. The load change detection function is as follows: ; In the formula, This is a load change detection function; For the first Each device or unit in The load at any moment; Total number of devices or units; These are the weighting coefficients of the second derivative; Let be the second derivative of the load, representing the acceleration due to the change in load; The term represents random fluctuations and follows a uniform distribution. ,in This refers to the fluctuation range.

2. The load balancing control method for the liquefied gas carrier cargo operating system simulator according to claim 1, characterized in that, The liquefied gas carrier cargo operating system simulator adopts a three-layer architecture design, including a host computer closed loop, a host computer closed loop, and a front-end closed loop. The host computer closed loop is used to display the human-machine interface of the liquefied gas carrier loading and unloading system and to display and respond to simulated data. The host computer closed loop is used to simulate cargo loading and unloading functions and internal unit simulations based on the instructions and data issued by the host computer. The front-end closed loop is used for data interaction between the simulator and the real unit devices of the liquefied gas carrier loading and unloading system.

3. The load balancing control method for the liquefied gas carrier cargo operating system simulator according to claim 2, characterized in that, The liquefied gas carrier cargo operating system simulator also includes an architecture interaction mechanism, which includes: a data communication protocol to ensure accurate data transmission between the host computer closed loop, the host computer closed loop, and the front-end closed loop; real-time data processing to ensure that the data processing and response time of each layer meets the system requirements; and a synchronization mechanism to achieve timestamp-based data synchronization and ensure data consistency between the host computer, the host computer, and the front-end.

4. The load balancing control method for the liquefied gas carrier cargo operating system simulator according to claim 1, characterized in that, The liquefied gas carrier cargo operating system simulator consists of a liquefied natural gas carrier cargo hold simulation unit, a liquefied petroleum gas carrier simulation unit, a reliquefaction system simulation unit, a nitrogen production system simulation unit, and a communication module unit. It adopts a modular structure, connecting the industrial control host, data transmission module, and power system together via a bus. It supports the selection of multiple data transmission modules and achieves communication with the host computer through an Ethernet interface.

5. The load balancing control method for the liquefied gas carrier cargo operating system simulator according to claim 1, characterized in that, The specific representation of the liquid tank pressure response model is as follows: ; In the formula, This refers to the pressure in the liquid tank, expressed in kPa. The gain coefficient is dimensionless, and ; The time constant is expressed in seconds. This refers to the loading rate, expressed in m³ / h. This refers to the return gas rate, expressed in m³ / h. This represents the initial pressure of the liquid tank, expressed in kPa. For the Laplace operator.

6. The load balancing control method for the liquefied gas carrier cargo operating system simulator according to claim 1, characterized in that, The ship condition monitoring model is specifically represented as follows: ; ; ; ; ; ; ; ; ; In the formula, The new lateral stability is high, and the unit is meters (m). The original transverse stability of the ship is high, and the unit is meters (m). The change in weight for liquid cargo or ballast water, in tons; This refers to the ship's displacement, expressed in tons (t). The average draft of the ship is expressed in meters (m). Vertical coordinates of the changing position of liquid cargo or ballast water, in meters; The density of each liquid is given in t / m³. Here is the moment of inertia of each liquid surface area, in meters. 4 ; The new longitudinal stability is high, and the unit is meters (m). The original longitudinal stability of the ship is high, and the unit is meters (m). The tilt angle is expressed in rad. The pitch angle is expressed in rad. The lateral coordinates of the changing position of the liquid cargo or ballast water, in meters; The longitudinal coordinates of the changing position of the liquid cargo or ballast water, in meters; The vertical coordinate of the center of gravity is in meters. The original draft of the ship, measured in meters (m). The stern draft of the vessel, measured in meters (m). The draft of a ship at its new bow, expressed in meters (m). The draft of the stern of a vessel, measured in meters (m). The length of the ship is measured in meters (m). TPC is the tonnage per centimeter of draft, expressed in t / cm; MTC is the initial trim moment per centimeter of trim, expressed in t·m / cm.

Images