BOG compressor intelligent optimization control method based on LNG storage tank pressure prediction

Abstract

The invention relates to the field of BOG compressor intelligent control methods, in particular to a BOG compressor intelligent optimization control method based on LNG storage tank pressure prediction. Comprising the following steps: S1, generating an energy efficiency optimization decision and a control instruction; s2, prospective pressure abnormity early warning and processing; and S3, performing closed-loop execution and iterative optimization of the control method. And through mixed integer programming optimization and prospective early warning, the problems of response lag, high energy consumption and safety passivity of the BOG compressor are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of BOG compressor intelligent control method, in particular to a BOG compressor intelligent optimization control method based on LNG storage tank pressure prediction. BACKGROUND

[0002] In the operation process of liquefied natural gas (LNG) receiving station, the stable control of the storage tank pressure is the core link to ensure the safety of production and the efficiency of energy utilization. The BOG (boil-off gas) compressor, as a key equipment for regulating the pressure of the storage tank, its running state directly affects the safety boundary, equipment life and operating cost of the receiving station. The traditional control mode commonly used in the industry mainly relies on the start-stop logic triggered by fixed pressure threshold or artificial experience judgment. This mode gradually exposes the following significant defects in actual application:

[0003] (1) Response lag problem

[0004] The traditional control mode only triggers the compressor action when the storage tank pressure reaches the preset threshold, which belongs to the typical "after-response" mechanism. Due to the large thermal inertia and pressure buffering characteristics of the LNG storage tank system, the pressure change often has a delay effect, which makes the control system unable to predict the future pressure trend. For example, under the disturbance conditions such as unloading operation or sudden rise of ambient temperature, when the pressure reaches the threshold, the actual pressure rising trend has formed inertia, at this time, starting the compressor may miss the best control opportunity, causing pressure overshoot or subsequent frequent intervention.

[0005] (2) Equipment wear problem

[0006] Limited by the threshold trigger logic, the traditional system is prone to frequent start-stop of the compressor due to short-term pressure fluctuations. This frequent state switching will cause impact load on mechanical components such as motors and valves, accelerating bearing wear and seal aging. The operation data of a receiving station shows that the monthly average start-stop times of a single compressor under the traditional control mode can reach 30-40 times, which is more than 3 times higher than the intelligent control mode, directly leading to a 40% reduction in equipment maintenance cycle and an annual maintenance cost increase of about 250,000 yuan per unit.

[0007] (3) Poor running economy problem

[0008] The traditional control strategy lacks the ability of global optimization of the running condition, and the compressor often runs in the non-efficient load interval. On the one hand, the fixed threshold control cannot dynamically adjust the load according to the BOG production, leading to the phenomenon of "big horse pulling small cart"; on the other hand, it does not consider the difference between peak and valley electricity prices, and still maintains high load operation during high price period. Statistical data shows that the comprehensive energy efficiency ratio (COP) of the BOG compressor under the traditional control mode is only 2.8-3.2, while the theoretical optimal working condition can reach 3.8-4.2, with an annual electricity waste rate of 15%-20%.

[0009] (4) Passive problem of safety boundary

[0010] Existing systems rely on alarm mechanisms when pressure approaches safety limits, representing a passive defense model. Lacking proactive predictive capabilities, operators often have only 10-15 minutes to respond to emergencies (such as a surge in water vapor caused by rapid LNG unloading), making effective pre-control measures difficult. At one coastal receiving terminal, a sudden rise in ambient temperature during a typhoon failed to trigger an early warning from the traditional system, causing tank pressure to approach safety limits within 20 minutes, forcing an emergency release of bulk gas (BOG) and resulting in economic losses and environmental risks.

[0011] The aforementioned shortcomings collectively hinder the intelligent development of LNG receiving terminals, necessitating a novel control method capable of proactive pressure prediction, optimized equipment scheduling, and proactive safety protection. This invention addresses these industry pain points by proposing an intelligent control scheme based on data-driven and operations-optimized approaches, aiming to overcome the limitations of traditional control models. Summary of the Invention

[0012] To address the aforementioned issues, an intelligent optimization control method for BOG compressors based on LNG tank pressure prediction is provided. Through mixed integer programming optimization and forward-looking early warning, the method solves the problems of BOG compressor response lag, high energy consumption, and passive safety.

[0013] To address the problems of existing technologies, this invention provides a smart optimization control method for BOG compressors based on LNG tank pressure prediction, comprising the following steps:

[0014] Step S1: Energy efficiency optimization decision and control instruction generation.

[0015] A mixed-integer programming optimization model is constructed, involving multiple objective functions and security constraints. The optimal control instruction set for the compressor is generated by solving the model to achieve economical and safe operation.

[0016] Step S2: Proactive warning and handling of abnormal pressure.

[0017] By comparing the predicted pressure curve with the safety threshold, a multi-level early warning mechanism is established to intelligently generate emergency response plans and prevent the risk of abnormal pressure.

[0018] Step S3: Closed-loop execution and iterative optimization of the control method.

[0019] Establish instruction verification and fault self-recovery mechanisms to ensure execution reliability, and construct a high-quality feedback dataset for model iteration through data preprocessing.

[0020] As a specific embodiment of the present invention, step S1 includes:

[0021] Step S1.1: Constructing the mixed-integer programming optimization model.

[0022] Define the compressor start / stop state and throughput variables, establish flow balance constraints and time coupling relationship, and construct a structured optimization model framework.

[0023] As a specific embodiment of the present invention, step S1 further includes:

[0024] Step S1.2: Setting multi-objective functions and security constraints.

[0025] Establish a multi-objective function with energy consumption cost and state switching cost as the core, and set pressure safety window and equipment load boundary constraints.

[0026] As a specific embodiment of the present invention, step S1 further includes:

[0027] Step S1.3: Solving and outputting the optimal control instruction set.

[0028] An efficient solution algorithm is used to generate a compressor operation plan for the next 30 hours. After multiple verifications, the optimized pressure trajectory is output to the execution system to achieve fine control.

[0029] As a specific embodiment of the present invention, step S2 includes:

[0030] Step S2.1: Compare the predicted pressure curve with the safety threshold in real time.

[0031] Real-time comparison of predicted pressure with safety thresholds, definition of instantaneous over-limit and fluctuation warning risk indicators, and realization of quantitative identification of pressure anomalies.

[0032] As a specific embodiment of the present invention, step S2 further includes:

[0033] Step S2.2: Triggering and response to multi-level early warning systems.

[0034] Establish a Level 1 and Level 2 early warning mechanism, and set differentiated response times and multi-channel information push strategies.

[0035] As a specific embodiment of the present invention, step S2 further includes:

[0036] Step S2.3: Intelligent generation of emergency response plan.

[0037] Based on multi-dimensional real-time data, the system generates optimal emergency response plans through rule engines and case reasoning, thereby shortening decision-making and response time.

[0038] As a specific embodiment of the present invention, step S3 includes:

[0039] Step S3.1: Intelligent generation of emergency response plan.

[0040] Establish instruction verification and fault self-recovery mechanisms to ensure execution reliability, and construct a high-quality feedback dataset for model iteration through data preprocessing.

[0041] As a specific embodiment of the present invention, step S3 further includes:

[0042] Step S3.2: Model parameter iteration and algorithm optimization.

[0043] The model iteration process is initiated based on preset trigger conditions, and the algorithm is continuously optimized through data retraining, parameter tuning, offline verification and online deployment.

[0044] The advantages of this invention compared to the prior art are:

[0045] 1. Achieve precise and automated tank pressure control.

[0046] This invention breaks away from the reliance on traditional manual experience, enabling proactive prediction of tank pressure trends. The system can automatically adjust control strategies based on LNG tank pressure forecasts, keeping tank pressure fluctuations within a narrower, safer range, improving the accuracy and automation of pressure management, and reducing the need for manual intervention.

[0047] 2. Improve equipment operation and maintenance efficiency and reduce energy consumption.

[0048] The optimization decision engine based on mixed-integer programming effectively avoids ineffective start-stop operations and inefficient operation by optimizing compressor start-stop times and load allocation. This not only reduces equipment mechanical wear and extends service life but also guides compressors to operate in high-efficiency load ranges. Combined with peak-valley electricity pricing for optimized scheduling, it significantly reduces overall energy consumption and operating costs.

[0049] 3. Strengthen the safety protection capabilities of storage tank operation.

[0050] The proactive pressure anomaly early warning mechanism identifies pressure exceedance risks at an early stage by comparing predicted pressure curves with safety thresholds in real time. The system provides operators with ample response time and clear handling plans, transforming passive defense into proactive early warning, enhancing the safety boundaries of tank operation, and reducing the risk of safety accidents such as overpressure releases or negative pressure intake. Attached Figure Description

[0051] Figure 1 This is a flowchart of the intelligent optimization control method for BOG compressors based on LNG tank pressure prediction according to the present invention. Detailed Implementation

[0052] To further understand the features, technical means, and specific objectives and functions achieved by the present invention, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

[0053] The technical solution of this invention achieves intelligent optimization control of the BOG compressor based on the LNG storage tank pressure prediction results (predicted pressure curve).

[0054] Reference Figure 1 The intelligent optimization control method for BOG compressors based on LNG tank pressure prediction, as shown, is characterized by the following steps:

[0055] Step S1: Energy efficiency optimization decision and control instruction generation.

[0056] A mixed-integer programming optimization model is constructed, involving multiple objective functions and security constraints. The optimal control instruction set for the compressor is generated by solving the model to achieve economical and safe operation.

[0057] Step S2: Proactive warning and handling of abnormal pressure.

[0058] By comparing the predicted pressure curve with the safety threshold, a multi-level early warning mechanism is established to intelligently generate emergency response plans and prevent the risk of abnormal pressure.

[0059] Step S3: Closed-loop execution and iterative optimization of the control method.

[0060] Establish instruction verification and fault self-recovery mechanisms to ensure execution reliability, and construct a high-quality feedback dataset for model iteration through data preprocessing.

[0061] Furthermore, step S1 includes:

[0062] Step S1.1: Constructing the mixed-integer programming optimization model.

[0063] Define the compressor start / stop state and throughput variables, establish flow balance constraints and time coupling relationship, and construct a structured optimization model framework.

[0064] Specifically, the model variable system contains two key elements:

[0065] Integer variables are used to describe the compressor's start-stop status, using 0-1 encoding (1 indicates running, 0 indicates stopping).

[0066] Continuous variables represent the compressor's processing capacity, and their values ​​are strictly limited to the range between the equipment's minimum and maximum loads to ensure that the operation complies with mechanical performance constraints.

[0067] Variable definition principle: Integer variables and continuous variables must satisfy time coupling, that is, the processing volume of continuous variables only has actual physical meaning when the compressor is running (integer variable = 1), to avoid the logical contradiction of "there is still flow when the compressor is stopped".

[0068] The model's time granularity is set to one decision unit per hour, consistent with the BOG production prediction time domain (step 1.3). This satisfies the dynamic response requirements while avoiding a surge in computational complexity due to excessively fine time units. In the constraint construction, the flow balance constraint is derived based on the ideal gas law (PV=nRT), linking BOG production, compressor throughput, and tank pressure changes. Specifically, the change in tank pressure per unit time is equal to the product of the net BOG production (the difference between production and compressor throughput) and the gas compressibility coefficient, ensuring that the model strictly follows the laws of thermodynamics.

[0069] The objective function is constructed based on variable definitions and the time dimension. Subsequent iterations will integrate multiple objective factors such as energy consumption costs and equipment wear and tear to form an optimization objective that balances economic efficiency and safety. Through this mathematical framework, the model can achieve dynamic optimization decisions regarding the compressor's operating state, providing structured input for subsequent solution algorithms.

[0070] Furthermore, step S1 also includes:

[0071] Step S1.2: Setting multi-objective functions and security constraints.

[0072] Establish a multi-objective function with energy consumption cost and state switching cost as the core, and set pressure safety window and equipment load boundary constraints.

[0073] Specifically, the construction of multi-objective functions requires a hierarchical optimization strategy to achieve coordinated control of economic efficiency and operational stability.

[0074] The primary objective focuses on electricity price-driven operational optimization, reducing overall energy costs by guiding equipment to operate more during off-peak hours and avoid peak hours. Specific implementation requires consideration of the time-of-use electricity pricing policy in the LNG receiving terminal's location, clearly defining the time intervals for peak, mid-peak, average, and off-peak periods, and their corresponding electricity price coefficients.

[0075] The secondary objective requires the introduction of a state-switching penalty term to quantify the additional costs incurred during compressor start-up and shutdown. Based on the energy consumption characteristics and lifespan loss model of the equipment start-up and shutdown, a fixed electricity cost loss value is set for each start-up and shutdown operation. By adding this penalty term to the objective function, the economic decline and equipment damage risk caused by frequent start-ups and shutdowns can be effectively avoided.

[0076] The safety constraint system must cover both process parameters and equipment capabilities. Specifically, the safety pressure window is set as the design pressure range of the vapor space in the receiving station's storage tanks. During operation, compressor load regulation must be used to ensure the pressure remains within this range to prevent safety accidents such as overpressure releases or negative pressure intake. Regarding equipment constraints, maximum / minimum processing loads must be set according to the performance curves of different models to ensure that operating parameters match the physical limits of the equipment.

[0077] The core logic of the optimized control is to achieve the economical and safe operation of the BOG compressor by weighted optimization of multiple objective functions within the safe pressure window and equipment load boundary, with the electricity price time-of-use strategy as the main guide and the state switching cost as the correction.

[0078] Furthermore, step S1 also includes:

[0079] Step S1.3: Solving and outputting the optimal control instruction set.

[0080] An efficient solution algorithm is used to generate a compressor operation plan for the next 30 hours. After multiple verifications, the optimized pressure trajectory is output to the execution system to achieve fine control.

[0081] Specifically, as the core execution link of the intelligent optimization control of the BOG compressor, its algorithm selection and logic design directly determine the effectiveness and economy of the control strategy. In algorithm selection, the convergence speed and optimality of the solution to the mixed-integer programming problem must be comprehensively considered to ensure the rapid generation of the globally optimal solution under the complex operating conditions of the LNG receiving terminal. Addressing the energy waste and mechanical losses caused by frequent start-ups and shutdowns of the BOG compressor, the system is designed with a judgment logic to "avoid ineffective start-ups and shutdowns due to brief pressure fluctuations": when the future pressure curve predicted by the model fluctuates within a safe window, the control algorithm will maintain the compressor's current operating state, triggering a state switch only when the predicted pressure exceeds a threshold and the duration reaches a set threshold, effectively reducing the frequency of ineffective adjustments.

[0082] Core control instruction output specifications:

[0083] Time series dimension: Generate the compressor operating status (start / stop) and load value (0%-100%) for each hour in the next 30 hours to form a refined control scheme.

[0084] Execution mechanism: Optimized instructions are transmitted to the PLC system via industrial Ethernet, and after logic conversion in the control cabinet, they drive the compressor actuator.

[0085] Data linkage: The output optimized pressure trajectory is synchronously used as the benchmark comparison curve of the abnormal early warning module in step S2, realizing closed-loop coordination of control and diagnosis.

[0086] In step 1.3, the predicted pressure curve is not only the core input for solving the optimal control command, but its prediction accuracy directly affects the effectiveness of the control strategy. The system updates the predicted pressure curve and re-solves the control command every hour through a rolling optimization mechanism, ensuring that the tank pressure remains within the target range even when the BOG generation changes dynamically, while minimizing compressor operating energy consumption. Before the command is issued, it undergoes multiple verifications, including equipment safety constraint checks (such as minimum start-stop intervals and load adjustment rate limits) and communication integrity verification, ensuring reliable execution of the control command.

[0087] Furthermore, step S2 includes:

[0088] Step S2.1: Compare the predicted pressure curve with the safety threshold in real time.

[0089] Real-time comparison of predicted pressure with safety thresholds, definition of instantaneous over-limit and fluctuation warning risk indicators, and realization of quantitative identification of pressure anomalies.

[0090] Specifically, the core of the real-time comparison between the predicted pressure curve and the safety threshold lies in the quantitative identification and early warning of pressure risks through precise algorithmic logic. This comparison algorithm first decomposes the predicted pressure curve for the next 30 hours into discrete data points along the time axis, and then compares it with the preset safety threshold in real time by scanning each point, ensuring full monitoring of pressure change trends.

[0091] To achieve quantitative risk assessment, the system defines two types of "risk exceeding limits" indicators:

[0092] The first is the risk of instantaneous over-limit, that is, the predicted pressure P(t) at a certain point in time exceeds the maximum safe pressure Pmax by 5% (i.e., P(t)>Pmax+5%).

[0093] Second, there is the risk of fluctuation warning, which refers to the fact that the predicted pressure P(t) at three consecutive time points is within the range of Pmax±3%, thereby capturing potential abnormal pressure fluctuation trends.

[0094] Dual-track verification mechanism for comparison objects: The system performs dual-track comparison analysis on the original predicted pressure curve and the optimized pressure trajectory generated by S1.3. When the pressure deviation between the two exceeds a set threshold at the same time point, a secondary verification process is automatically triggered. Cross-validation eliminates potential missed warnings due to optimization model failure, ensuring the completeness and reliability of risk identification.

[0095] This dual-track comparison and graded early warning mechanism not only enables refined management of pressure risks, but also constructs a redundant safety protection layer through cross-validation of raw data and optimization results, providing a scientific basis for the dynamic control of BOG compressors.

[0096] Furthermore, step S2 also includes:

[0097] Step S2.2: Triggering and response to multi-level early warning systems.

[0098] Establish a Level 1 and Level 2 early warning mechanism, and set differentiated response times and multi-channel information push strategies.

[0099] Specifically, to ensure the safe and stable operation of the BOG compressor system at the LNG receiving terminal, a scientific multi-level early warning mechanism needs to be established to achieve graded risk control and rapid response. Based on system pressure prediction results and the feasibility of optimized control, the early warning levels are divided into two core levels: Level 1 and Level 2. Each level clearly defines the triggering conditions, response time standards, and information transmission paths, forming a closed-loop management system.

[0100] The Level 1 warning (optimizable recovery) is primarily for scenarios where system pressure is close to the safe threshold but can be recovered through optimized control. A Level 1 warning is triggered when the intelligent predictive model indicates that the BOG system pressure will reach 90%–95% of the safe threshold within a certain period, and the optimization algorithm confirms that existing control measures (such as adjusting compressor load or optimizing the number of operating units) can restore the pressure to the normal range. To ensure operators have sufficient time to implement controls, a 60-minute operation window must be reserved for Level 1 warnings to ensure intervention is completed before the pressure exceeds the limit.

[0101] The Level 2 warning (emergency risk) is applicable to emergency situations where there is still a risk of pressure exceeding limits even after optimized control. The system automatically triggers a Level 2 warning when the predicted pressure exceeds 95% of the safe threshold, or when the optimized algorithm determines that existing control measures cannot effectively mitigate the upward pressure trend. Considering the urgency of such scenarios, the standard response time for a Level 2 warning is set at 30 minutes, requiring operators to immediately take emergency measures, such as starting the backup compressor and adjusting the LNG unloading plan, to avoid overpressure accidents.

[0102] Efficient transmission of early warning information is crucial for ensuring timely response. The system employs a multi-channel, interconnected information push mechanism: firstly, the SCADA system issues audible and visual alarms in the central control room, displaying the warning level, predicted pressure curve, and suggested operating procedures on the monitoring interface; simultaneously, the warning information, including real-time pressure data, warning level, and remaining response time, is pushed to operators' mobile apps via a dedicated communication protocol, ensuring that relevant personnel can receive and respond promptly regardless of their location. This dual-path design of "system alarm + mobile push" significantly improves the reach and response efficiency of early warning information, providing a vital guarantee for the safe operation of the BOG compressor system.

[0103] Furthermore, step S2 also includes:

[0104] Step S2.3: Intelligent generation of emergency response plan.

[0105] Based on multi-dimensional real-time data, the system generates optimal emergency response plans through rule engines and case reasoning, thereby shortening decision-making and response time.

[0106] Specifically, the core logic of intelligent emergency response plan generation lies in dynamic decision matching based on multi-dimensional real-time data. The system collects key parameters such as compressor operating status (e.g., speed, inlet and outlet pressure, vibration value), tank liquid level, BOG generation, and external pipeline network demand in real time, constructs a dynamic evaluation model, and quickly selects the optimal response strategy from the preset plan library.

[0107] Example of typical scenario handling logic: When the system predicts that the tank pressure will exceed the upper limit, a three-level response mechanism is triggered:

[0108] (1) Prioritize starting the backup compressor (if it is in a shutdown state) to improve processing capacity through parallel operation.

[0109] (2) Increase the load of the existing compressor to the maximum allowable value (the equipment safety constraints must be met).

[0110] (3) Dynamically adjust the outflow rate and achieve pressure balance by coordinating with the downstream pipeline network.

[0111] This mechanism effectively solves the response lag problem caused by the reliance on manual experience in traditional tank pressure control through a hybrid decision-making model that combines rule engines and case-based reasoning (the average decision-making time is reduced from 5-8 minutes in the traditional method to within 30 seconds). The solution generation process strictly follows the triple verification principle of "clarity-executability-safety". All output strategies are accompanied by specific operating parameters (such as compressor load adjustment step size, valve opening command) and expected effect evaluation (pressure drop rate, energy consumption change trend), assisting operators in making rapid and accurate decisions in emergency situations.

[0112] The system's built-in solution library covers a variety of typical abnormal operating conditions and supports self-iterative optimization based on actual operating data. It continuously improves the accuracy of solution matching through reinforcement learning algorithms.

[0113] Furthermore, step S3 includes:

[0114] Step S3.1: Intelligent generation of emergency response plan.

[0115] Establish instruction verification and fault self-recovery mechanisms to ensure execution reliability, and construct a high-quality feedback dataset for model iteration through data preprocessing.

[0116] Specifically, the reliability of the intelligent generation of emergency response plans directly affects the operational stability and optimized control effect of the compressor unit. This stage ensures the accuracy of command execution and the effectiveness of feedback data through a dual guarantee mechanism, providing high-quality input for model iteration.

[0117] Command execution reliability assurance: Due to the special characteristics of BOG compressor control commands, a multi-level reliability assurance system needs to be established.

[0118] The instruction verification mechanism verifies the legality of control instructions generated by the optimization algorithm through preset logical rules, including parameter boundary verification (such as upper and lower limits of speed and pressure setpoint range), timing logic verification (to avoid frequent start and stop), and equipment status matching verification (to ensure that the instruction is compatible with the current operating conditions), effectively filtering abnormal instructions.

[0119] The fault self-recovery mechanism is activated when the instruction execution is abnormal. It restores the system to a stable state through strategies such as switching redundant control channels, backtracking historical instructions, or switching to a safe mode, thereby reducing the impact of single-point failures on the overall control process.

[0120] Feedback Data Quality Control: To ensure the validity of the feedback data, a rigorous data preprocessing procedure must be implemented. Outlier removal employs a composite judgment method based on the 3σ principle and equipment physical constraints to identify and remove abnormal data points caused by sensor failures, communication interference, or instantaneous fluctuations in operating conditions. Data smoothing uses a combination of moving average filtering and Kalman filtering to suppress high-frequency noise while preserving the true operating condition trend, thereby improving the data signal-to-noise ratio. The processed feedback data includes key operating parameters such as compressor inlet and outlet pressures, flow rates, and shaft temperatures. The collected raw process data (data collected during LNG tank pressure prediction) forms a spatiotemporal complement, jointly constructing a multidimensional dataset for model training.

[0121] Data-Model Iteration Closed Loop: Feedback data, acting as the "sensory nerve endings" of optimized control, directly determines the accuracy of model iteration. By fusing real-time feedback data with historical data, model parameter deviations can be dynamically corrected, enhancing the algorithm's adaptability to complex operating conditions such as BOG load fluctuations and equipment performance degradation, ultimately achieving continuous optimization of the control strategy.

[0122] This step, through the bidirectional reinforcement of command execution and data feedback, constructs a complete control loop from decision-making to execution and then to optimization, providing key technical support for the safe and efficient operation of the BOG compressor.

[0123] Furthermore, step S3 also includes:

[0124] Step S3.2: Model parameter iteration and algorithm optimization.

[0125] The model iteration process is initiated based on preset trigger conditions, and the algorithm is continuously optimized through data retraining, parameter tuning, offline verification and online deployment.

[0126] Specifically, the implementation of model parameter iteration and algorithm optimization must be based on a clear triggering mechanism and a systematic process. When the model prediction accuracy is lower than the preset threshold (e.g., the root mean square error exceeds the allowable range of the process), key process parameters drift significantly (e.g., the BOG output fluctuates by more than ±15%), or the equipment operating status undergoes a phased change (e.g., the compressor enters the break-in period or maintenance cycle), the system will automatically start the iterative optimization program.

[0127] Optimize the process in four stages:

[0128] 1. Data retraining: Integrate the latest 6 months of operational data (including extreme working condition samples) to build a dynamically updated training dataset.

[0129] 2. Parameter tuning: Based on the Bayesian optimization algorithm, the learning rate, tree depth and other hyperparameters of the LGB model are adjusted, and the optimal parameter combination is determined by 5-fold cross-validation.

[0130] 3. Offline verification: Reproduce typical historical operating conditions on the simulation platform to verify whether the control deviation of the optimized model has been reduced to within 5%.

[0131] 4. Online deployment: A canary release strategy is adopted to gradually switch control models and monitor the changes in compressor power consumption and BOG processing efficiency in real time.

[0132] The technical advantage of this optimization mechanism stems from the LGB model's ability to deeply mine multi-dimensional and nonlinear time-series characteristics, effectively capturing the complex coupling relationship between BOG generation and variables such as ambient temperature, tank pressure, and compressor load. Through an iterative strategy combining monthly scheduled execution with triggered instantaneous execution, the adaptability of the control model can be significantly improved, ensuring optimal control performance is maintained throughout different operating cycles of the LNG receiving terminal (such as the high evaporation rate period in summer and the low load period in winter). The digital twin platform used in the offline verification phase can complete the model's effectiveness verification without interfering with actual production, greatly reducing deployment risks.

[0133] The above embodiments only illustrate one or more implementations of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the appended claims.

Claims

1. A method for intelligent optimization control of BOG compressors based on LNG tank pressure prediction, characterized in that, Includes the following steps: Step S1: Energy efficiency optimization decision and control instruction generation. A mixed-integer programming optimization model is constructed, involving multiple objective functions and security constraints. The optimal control instruction set for the compressor is generated by solving the model to achieve economical and safe operation. Step S2: Proactive warning and handling of abnormal pressure. By comparing the predicted pressure curve with the safety threshold, a multi-level early warning mechanism is established to intelligently generate emergency response plans and prevent the risk of abnormal pressure. Step S3: Closed-loop execution and iterative optimization of the control method. Establish instruction verification and fault self-recovery mechanisms to ensure execution reliability, and construct a high-quality feedback dataset for model iteration through data preprocessing.

2. The intelligent optimization control method for BOG compressors based on LNG tank pressure prediction according to claim 1, characterized in that, Step S1 includes: Step S1.1: Constructing the mixed-integer programming optimization model. Define the compressor start / stop state and throughput variables, establish flow balance constraints and time coupling relationship, and construct a structured optimization model framework.

3. The intelligent optimization control method for BOG compressors based on LNG tank pressure prediction according to claim 2, characterized in that, Step S1 further includes: Step S1.2: Setting multi-objective functions and security constraints. Establish a multi-objective function with energy consumption cost and state switching cost as the core, and set pressure safety window and equipment load boundary constraints.

4. The intelligent optimization control method for BOG compressors based on LNG tank pressure prediction according to claim 3, characterized in that, Step S1 further includes: Step S1.3: Solving and outputting the optimal control instruction set. An efficient solution algorithm is used to generate a compressor operation plan for the next 30 hours. After multiple verifications, the optimized pressure trajectory is output to the execution system to achieve fine control.

5. The intelligent optimization control method for BOG compressors based on LNG tank pressure prediction according to claim 4, characterized in that, Step S2 includes: Step S2.1: Compare the predicted pressure curve with the safety threshold in real time. Real-time comparison of predicted pressure with safety thresholds, definition of instantaneous over-limit and fluctuation warning risk indicators, and realization of quantitative identification of pressure anomalies.

6. The intelligent optimization control method for BOG compressors based on LNG tank pressure prediction according to claim 5, characterized in that, Step S2 further includes: Step S2.2: Triggering and response to multi-level early warning systems. Establish a Level 1 and Level 2 early warning mechanism, and set differentiated response times and multi-channel information push strategies.

7. The intelligent optimization control method for BOG compressors based on LNG tank pressure prediction according to claim 6, characterized in that, Step S2 further includes: Step S2.3: Intelligent generation of emergency response plan. Based on multi-dimensional real-time data, the system generates optimal emergency response plans through rule engines and case reasoning, thereby shortening decision-making and response time.

8. The intelligent optimization control method for BOG compressor based on LNG tank pressure prediction according to claim 7, characterized in that, Step S3 includes: Step S3.1: Intelligent generation of emergency response plan. Establish instruction verification and fault self-recovery mechanisms to ensure execution reliability, and construct a high-quality feedback dataset for model iteration through data preprocessing.

9. The intelligent optimization control method for BOG compressors based on LNG tank pressure prediction according to claim 8, characterized in that, Step S3 further includes: Step S3.2: Model parameter iteration and algorithm optimization. The model iteration process is initiated based on preset trigger conditions, and the algorithm is continuously optimized through data retraining, parameter tuning, offline verification and online deployment.

Images