Optimized scheduling method and system for LNG receiving station

Abstract

The invention discloses an optimized scheduling method and system for an LNG receiving station. The method comprises the following steps: constructing a multi-level digital twinborn model of the LNG receiving station; collecting and obtaining target data of the LNG receiving station; inputting the target data into the multi-level digital twinborn model, and optimizing target process parameters of the multi-level digital twinborn model to obtain optimized process parameters; and determining a model decision of the LNG receiving station according to the optimized process parameters. Thus, the operation state of the LNG receiving station can be sensed comprehensively in real time, intelligent optimization and decision support are carried out, and therefore the operation efficiency, safety and economical efficiency of the LNG receiving station are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of liquefied natural gas, and in particular to an optimized scheduling method and system for an LNG receiving station. Background Art

[0002] Liquefied natural gas (LNG) receiving stations are key facilities in the natural gas supply chain, responsible for receiving, storing, gasifying and transporting LNG. With the growth of global energy demand and the advancement of clean energy transformation, LNG is becoming increasingly important in the energy structure. The efficient and safe operation of LNG receiving stations is crucial to ensuring energy supply and improving economic benefits.

[0003] At present, the operation and management of LNG receiving stations mainly rely on traditional monitoring systems and manual experience. These systems usually use static models for optimization, which makes it difficult to reflect the dynamic operation status of the receiving station in real time. At the same time, existing optimization methods often target a single goal and make it difficult to balance multiple key indicators such as energy efficiency, operating costs, and safety. Summary of the invention

[0004] In view of the above problems, the present invention provides an optimization scheduling method and system for an LNG receiving station, which can comprehensively and real-time perceive the operating status of the LNG receiving station, perform intelligent optimization and decision support, thereby significantly improving the operational efficiency, safety and economy of the LNG receiving station.

[0005] According to a first aspect of the present invention, a method for optimizing scheduling of an LNG receiving station is provided, comprising:

[0006] Build a multi-level digital twin model of the LNG receiving station;

[0007] Collect and obtain target data of LNG receiving stations;

[0008] Inputting the target data into the multi-level digital twin model, and optimizing the target process parameters of the multi-level digital twin model to obtain optimized process parameters;

[0009] According to the optimized process parameters, a model decision of the LNG receiving station is obtained.

[0010] Optionally, the multi-level digital twin model of the LNG receiving station is constructed, including:

[0011] A process model of the LNG receiving station is constructed using a system dynamics approach;

[0012] A hybrid modeling approach is used to construct a detailed model of the key equipment and systems of the LNG receiving station;

[0013] Finite element analysis and computational fluid dynamics methods are used to construct a detailed model of the LNG receiving station.

[0014] Optionally, inputting the target data into the multi-level digital twin model includes:

[0015] The target data is preprocessed by a hybrid filtering algorithm of Kalman filter and particle filter:

[0016]

[0017] in, is the state estimate at time k, K k is the Kalman gain, z k is the observed value, H k is the observation matrix, is the predicted value of the system state at time k based on the state estimation at time (k-1);

[0018] The preprocessed target data is input into the multi-level digital twin model.

[0019] Optionally, optimizing the target process parameters of the multi-level digital twin model to obtain optimized process parameters includes:

[0020] From multiple objective functions, determine the optimization objective function that needs to be optimized simultaneously;

[0021] Determining target process parameters to be optimized according to the optimization objective function;

[0022] The target process parameters are optimized by using the genetic algorithm NSGA-III to obtain the optimized process parameters.

[0023] Optionally, the mathematical expression of the NSGA-III algorithm is as follows:

[0024] minF(x)=[f1(x),f2(x),...,f m (x)] T

[0025] stg i (x)≤0,i=1,2,...,p

[0026] h j (x)=0,j=1,2,...,q

[0027] Among them, x is the decision variable vector, representing the target process parameter set that needs to be optimized, F(x) is the optimization objective function vector, representing multiple optimization objective functions that need to be optimized simultaneously, f1(x),f2(x),...,f m (x) represents m different optimization objective functions, m is the number of optimization objective functions, gi (x)≤0 is the inequality constraint, h j (x)=0 is an equality constraint, p is the number of inequality constraints, and q is the number of equality constraints.

[0028] Optionally, constraints are dynamically generated based on the knowledge graph and rule reasoning engine of the LNG receiving station.

[0029] Optionally, optimizing the target process parameters of the multi-level digital twin model includes multiple subtasks that can be calculated in parallel; according to the computational complexity of the multiple subtasks, the multiple subtasks are respectively allocated to edge, fog and cloud computing resources for computational processing.

[0030] Optionally, the method further includes:

[0031] SHAP values ​​are used to interpret the model decisions as operational instructions executed by the workers.

[0032] Optionally, a genetic algorithm NSGA-III is used to optimize the target process parameters, and the optimized process parameters are determined through a Pareto optimal solution.

[0033] According to a second aspect of the present invention, there is provided an LNG receiving station optimization scheduling system, comprising:

[0034] Multi-level digital twin model module, used to build a multi-level digital twin model of the LNG receiving station;

[0035] Multi-source data fusion module, used to collect and obtain target data of LNG receiving stations;

[0036] An intelligent optimization algorithm module, used for inputting the target data into the multi-level digital twin model, and optimizing the target process parameters of the multi-level digital twin model to obtain optimized process parameters;

[0037] A decision support module is used to obtain a model decision of the LNG receiving station according to the optimized process parameters.

[0038] The above one or more technical solutions in the embodiments of this specification have at least the following technical effects:

[0039] The embodiment of this specification provides an optimization scheduling method and system for an LNG receiving station, which constructs a multi-level digital twin model of the LNG receiving station; collects and obtains the target data of the LNG receiving station; inputs the target data into the multi-level digital twin model, and optimizes the target process parameters of the multi-level digital twin model to obtain optimized process parameters; and determines the model decision of the LNG receiving station based on the optimized process parameters. In this way, the operating status of the LNG receiving station can be fully and real-time perceived, and intelligent optimization and decision support can be performed, thereby significantly improving the operational efficiency, safety and economy of the LNG receiving station.

[0040] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Also, throughout the accompanying drawings, the same reference figures are used to represent the same components. In the drawings:

[0042] Figure 1 A flow chart of an LNG receiving station optimization scheduling method in an embodiment of the present invention is shown.

[0043] Figure 2 A block diagram of an LNG receiving station optimization scheduling device in an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0045] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0047] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0048] Liquefied natural gas (LNG) receiving stations are key facilities in the natural gas supply chain, responsible for receiving, storing, gasifying and transporting LNG. With the growth of global energy demand and the advancement of clean energy transformation, LNG is becoming increasingly important in the energy structure. The efficient and safe operation of LNG receiving stations is crucial to ensuring energy supply and improving economic benefits.

[0049] At present, the operation and management of LNG receiving stations mainly rely on traditional monitoring systems and manual experience. These systems usually use static models for optimization, which makes it difficult to reflect the dynamic operation status of the receiving station in real time. At the same time, existing optimization methods often target a single goal and make it difficult to balance multiple key indicators such as energy efficiency, operating costs, and safety.

[0050] Based on the above situation, the embodiment of the present invention provides an optimization scheduling method and system for LNG receiving stations, which realizes accurate modeling and real-time update of the dynamic operation status of LNG receiving stations by constructing a multi-level digital twin model; designs a multi-objective dynamic optimization algorithm to achieve balanced optimization in multiple dimensions; develops an intelligent decision-making mechanism based on reinforcement learning to improve the system's ability to respond quickly to emergencies; and realizes an edge-fog-cloud collaborative computing architecture to improve the computing efficiency of large-scale real-time optimization. This significantly improves the operational efficiency, safety, and economy of LNG receiving stations.

[0051] Specifically, combined with Figure 1 As shown, the present invention provides an LNG receiving station optimization scheduling method, which includes steps 101 to 104:

[0052] Step 101: construct a multi-level digital twin model of the LNG receiving station;

[0053] In this embodiment, the LNG receiving station is a key facility in the natural gas supply chain, responsible for receiving, storing, gasifying and transporting LNG. The LNG receiving station needs to adjust the load of the dynamic equipment in the station according to the needs of users, while making the energy consumption of the LNG receiving station low. Among them, the dynamic equipment may include compressors, boosters, seawater pumps, transport pumps, etc. The dynamic equipment is generally the key equipment of the LNG receiving station.

[0054] This embodiment will establish a mechanism model in advance, but the mechanism model may not be suitable for the LNG receiving station. This embodiment establishes a data model based on historical target data. According to the mechanism model and the data model, a multi-level digital twin model is established through a hybrid modeling method. On the one hand, the adaptability of the model is guaranteed, and on the other hand, the accuracy of the entire model is improved.

[0055] In an optional embodiment, the construction of a multi-level digital twin model of the LNG receiving station includes:

[0056] The process model of the LNG receiving station is constructed by using the system dynamics method; the detailed model of the key equipment and system of the LNG receiving station is constructed by using the hybrid modeling method; the fine model of the LNG receiving station is constructed by using the finite element analysis and computational fluid dynamics method, thereby realizing the dynamic coupling and information interaction of multi-scale models.

[0057] Step 102: Collect and obtain target data of the LNG receiving station;

[0058] In this embodiment, the LNG receiving station is equipped with many sensors, heat exchangers, etc., through which relevant target data such as temperature, pressure, flow, etc. can be collected. The multi-source data fusion module also stores historical target data and receives external data. These data serve as the input of the multi-level digital twin model.

[0059] Step 103: inputting the target data into the multi-level digital twin model, and optimizing the target process parameters of the multi-level digital twin model to obtain optimized process parameters;

[0060] In this embodiment, the collected target data is input into a multi-level digital twin model for model calculation, which ultimately assists in the generation of model decisions.

[0061] It should be noted that the target data needs to be preprocessed before being input into the model, that is, the target data is preprocessed by dynamic adaptive data fusion through a hybrid filtering algorithm of Kalman filter and particle filter, which is used to dynamically adjust the weights of different data sources according to the quality of the target data. The Kalman filter and particle filter are combined in series. First, the Kalman filter is used to quickly process the data to obtain a preliminary estimate, and then the particle filter is used to handle more complex situations. This method can simultaneously utilize the speed of the Kalman filter and the flexibility of the particle filter to make the state estimation more accurate.

[0062] The core formula of the hybrid filtering algorithm based on Kalman filter and particle filter is:

[0063]

[0064] in, It is the state estimation at time k, which is used to guide subsequent actions and provide decision assistance. Data fusion combines information from different sensors for subsequent analysis to more accurately understand and predict the state in a dynamic environment and improve the performance of the system.

[0065] K k is the Kalman gain, z k is the observed value, H k is the observation matrix, is the predicted value of the system state at time k based on the state estimate at time (k-1), which can also be understood as the prior state estimate at time k. k|k-1 represents the prediction of time k at time k-1, which is the best estimate of the state at time k based on all the information at time k-1. It is the prediction before obtaining the observed value at time k.

[0066] The preprocessed target data is then input into the multi-level digital twin model.

[0067] In this embodiment, the purpose of dynamic adaptive data fusion is mainly:

[0068] Data Integration:

[0069] Fusion of information from different sensors and data sources

[0070] In an LNG terminal, this may include temperature, pressure, flow, and other measurements.

[0071] Noise Filtering:

[0072] Reduce random noise in measurements and processes

[0073] Improve data reliability and accuracy

[0074] State Estimation:

[0075] Estimate the current status of the system, such as the actual liquid level of the LNG tank, the working status of the vaporizer, etc.

[0076] Predictive capabilities:

[0077] Predict future states based on current states and historical data

[0078] Helps with advance planning and preventive maintenance

[0079] Adaptability:

[0080] Ability to automatically adjust algorithm parameters based on data quality and system dynamics

[0081] Adapt to different operating conditions and environmental changes of LNG receiving stations

[0082] Real-time:

[0083] Rapidly process large amounts of real-time data streams

[0084] Support real-time monitoring and decision-making of LNG receiving stations

[0085] robustness:

[0086] Maintain stability in the face of anomalous data or sensor failures

[0087] Improve the overall reliability of the system

[0088] Multi-scale fusion:

[0089] Integrate data across time and space

[0090] Comprehensive analysis from micro equipment status to macro operating indicators

[0091] Through this dynamic and adaptive data fusion, the optimization system of the LNG receiving station can: obtain a more accurate and comprehensive understanding of the system status, improve the accuracy and timeliness of decision-making, enhance the system's adaptability to environmental changes and abnormal conditions, provide high-quality input data for upper-level optimization algorithms and decision support, and improve the operational efficiency, safety and economy of the LNG receiving station.

[0092] In an optional embodiment, the step of optimizing the target process parameters of the multi-level digital twin model to obtain the optimized process parameters may include:

[0093] From multiple objective functions, determine the optimization objective function that needs to be optimized simultaneously; the optimization objective function can be one or more, which can be determined according to actual conditions.

[0094] Then, according to the optimization objective function, the target process parameters to be optimized are determined;

[0095] The target process parameters are optimized by using the genetic algorithm NSGA-III to obtain the optimized process parameters.

[0096] The mathematical expression of the NSGA-III algorithm is as follows:

[0097] minF(x)=[f1(x),f2(x),...,f m (x)] T

[0098] stg i (x)≤0,i=1,2,...,p

[0099] h j (x)=0,j=1,2,...,q

[0100] Among them, x is the decision variable vector, which represents the target process parameter set that needs to be optimized. In the LNG receiving station, it can refer to operating parameters such as tank temperature, gasification rate, and delivery pressure. F(x) is the optimization objective function vector, which represents multiple optimization objective functions that need to be optimized simultaneously, f1(x), f2(x),..., f m (x) represents m different optimization objective functions, where m is the number of optimization objective functions. In an LNG receiving station, these optimization objective functions may include: maximizing energy efficiency, minimizing operating costs, minimizing safety risks, etc.

[0101] g i (x)≤0 is the inequality constraint, h j (x)=0 is an equality constraint, p is the number of inequality constraints, and q is the number of equality constraints.

[0102] The inequality constraints may represent operational limitations of the LNG terminal, such as upper tank pressure limits, minimum vaporization rate requirements, standby load limits, etc.

[0103] Equality constraints may represent conditions that must be precisely met in an LNG terminal, such as material balance equations, energy balance equations, specific process parameter requirements, etc.

[0104] The goal of this multi-objective optimization problem is to find a set of decision variables x such that all optimization objective functions f1(x) to f m (x) reaches the minimum value at the same time and satisfies all inequality constraints g i (x) and the equality constraint h j(x). However, in practical applications, since there may be conflicts between optimization objective functions, it is often necessary to find a Pareto optimal solution set (also called a set of non-dominated solutions), that is, a set of solutions that will inevitably harm other objectives when improving a certain objective, and finally determine the optimized process parameters.

[0105] Specifically:

[0106] For any two solutions A and B, if A is not worse than B in all objectives and is better than B in at least one objective, then A is said to dominate B.

[0107] If a solution is not dominated by any other solution, it is called a non-dominated solution or a Pareto optimal solution.

[0108] The collection of all non-dominated solutions constitutes the non-dominated solution set or Pareto front.

[0109] Application in LNG receiving station optimization:

[0110] Consider the multi-objective optimization problem of LNG receiving stations, which may include the following objectives:

[0111] Maximize energy efficiency (f1)

[0112] Minimize operating costs (f2)

[0113] Minimize security risks (f3)

[0114] The meaning of non-dominated solution set in the context of LNG terminal optimization:

[0115] Each non-dominated solution represents an operation plan of an LNG terminal.

[0116] These options strike different balances among the three objectives.

[0117] There is no single solution that can improve all objectives simultaneously without sacrificing at least one of the others.

[0118] For example:

[0119] Suppose we have three solutions:

[0120] Solution A: (f1=0.9, f2=100, f3=0.05)

[0121] Solution B: (f1 = 0.8, f2 = 80, f3 = 0.07)

[0122] Solution C: (f1 = 0.85, f2 = 90, f3 = 0.06)

[0123] In this example:

[0124] A is the best in energy efficiency, but has the highest operating cost.

[0125] B is the best in terms of operating cost, but has lower energy efficiency and higher safety risks.

[0126] C is in the middle in all aspects.

[0127] If no other solution can surpass these three solutions without lowering any objective, then A, B, and C are all non-dominated solutions and together constitute the non-dominated solution set.

[0128] The meaning of non-dominated solution set:

[0129] Provide a variety of options: Operators can choose the most appropriate solution based on the current situation.

[0130] Reveal trade-offs between goals: Help understand the interactions between different goals.

[0131] Support the decision-making process: Provide decision makers with a series of optimal options rather than a single "optimal solution".

[0132] In the actual application of LNG receiving stations, after generating the non-dominated solution set, the system may:

[0133] Visualizing these solutions helps operators understand the characteristics of different solutions intuitively.

[0134] Recommend the most suitable solution based on current operational priorities and external environment.

[0135] Operators are allowed to weigh and choose between these solutions, achieving human-machine collaborative decision-making.

[0136] In this way, the optimization system of the LNG receiving terminal can provide flexible and efficient decision support in complex multi-objective scenarios.

[0137] In this embodiment, constraints can be dynamically generated based on the knowledge graph and rule inference engine of the LNG receiving station, wherein the knowledge graph can be related to equipment requirements, mechanism knowledge, etc., and the rule inference engine can be the formulas and rules used in the reaction.

[0138] In this embodiment, the long short-term memory network (LSTM) and the graph convolutional network (GCN) can also be combined to perform time series and topological data analysis, anomaly detection and fault prediction.

[0139] Step 104: Obtain a model decision of the LNG receiving station according to the optimized process parameters.

[0140] In this embodiment, the optimized process parameters refer to the process parameters obtained after the model optimization calculation, such as the storage tank temperature, gasification rate, and delivery pressure. Based on the optimized process parameters, the model decision of the LNG receiving station can be determined. It is also necessary to use SHAP values ​​to interpret the model decision as an operation instruction executed by the staff. Among them, the model decision mainly refers to the adjustment and control of the moving equipment, for example, adjusting the valve opening to achieve the flow calculated in the optimized process parameters; increasing the speed of the seawater pump to achieve the temperature calculated in the optimized process parameters.

[0141] The basic principle of SHAP value:

[0142] 1. Based on game theory: SHAP value is derived from the concept of Shapley value in cooperative game theory.

[0143] 2. Feature contribution: The SHAP value calculates the contribution of each feature to the model prediction results.

[0144] 3. Additivity: The sum of the SHAP values ​​of all features is equal to the difference between the model prediction and the mean prediction.

[0145] About SHAP value to explain model decision:

[0146] 1. Quantifying feature importance:

[0147] The SHAP value of each feature is calculated to indicate the influence of the feature on the prediction result.

[0148] Positive values ​​indicate that the feature increases the predicted value, and negative values ​​indicate that the feature decreases the predicted value.

[0149] 2. Local explanation:

[0150] Provide an explanation for each specific prediction, showing how each feature influenced that particular decision.

[0151] 3. Global explanation:

[0152] By summarizing the SHAP values ​​of multiple predictions, we can get the overall importance ranking of the features.

[0153] 4. Feature Interaction:

[0154] Identify interactions between features and reveal complex non-linear relationships.

[0155] Application in LNG receiving station optimization decision:

[0156] 1. Decision-making transparency:

[0157] For example, when the system recommends increasing the LNG vaporization rate, the SHAP value can explain:

[0158] Current inventory level (SHAP value + 0.3)

[0159] Forecast demand increases (SHAP value +0.5)

[0160] Weather factor (SHAP value -0.1)

[0161] This suggests that forecast demand increases are the main driver, with weather factors having a slight dampening effect.

[0162] 2. Feature Importance Analysis:

[0163] Identify the factors most critical to LNG terminal operations.

[0164] For example, you might find that market demand forecasts and equipment maintenance status have the greatest impact on cost optimization.

[0165] 3. Anomaly detection explanation:

[0166] When the system detects an anomaly, the SHAP value can indicate:

[0167] Which parameters are abnormal (e.g., the SHAP value of the pressure reading deviates significantly from the normal range)

[0168] Possible causes (such as those related to abnormal device status SHAP values)

[0169] 4. Explanation of multi-objective trade-offs:

[0170] When choosing a Pareto optimal solution, the SHAP value can explain the trade-offs between different objectives:

[0171] Why a solution performs better in terms of energy efficiency (the SHAP value of the relevant feature is higher)

[0172] At the same time, there may be some sacrifice in terms of cost (the SHAP value of cost-related features is negative)

[0173] 5. Human-machine collaborative decision support:

[0174] Show operators the SHAP value of each factor to help them understand the decision logic of the system.

[0175] Operators can make more informed decisions based on their own expertise combined with the insights provided by SHAP values.

[0176] 6. Model improvement guidance:

[0177] By analyzing the SHAP values, possible biases or inconsistencies in the model can be identified.

[0178] Guide data scientists to optimize model structure or feature engineering.

[0179] Steps to implement SHAP value interpretation:

[0180] 1. Model training: Use machine learning algorithms (such as random forest, neural network, etc.) to train the LNG receiving terminal optimization model.

[0181] 2. SHAP value calculation: Calculate the SHAP value for each prediction sample.

[0182] 3. Visualization: Use waterfall charts, force diagrams, etc. to visualize SHAP values ​​and intuitively show the impact of features.

[0183] 4. Result integration: Integrate SHAP value interpretation with decision support systems to provide explanations for each key decision.

[0184] 5. Continuous optimization: Based on the feedback of SHAP value, the model and decision-making process are continuously optimized.

[0185] By using SHAP values, the optimization system of the LNG terminal can not only provide high-quality decision-making recommendations, but also make these recommendations explainable and trustworthy, thereby promoting human-machine collaboration and improving overall operational efficiency and safety.

[0186] In addition, this embodiment also designs an augmented reality algorithm to provide intuitive operation guidance on operation instructions to the staff.

[0187] It should be noted that, considering that the model algorithm may involve a large amount of calculations, in order to improve the system response speed and scalability, this embodiment decomposes the calculation task of optimizing the target process parameters of the multi-level digital twin model into multiple subtasks that can be calculated in parallel; then, according to the calculation amount of the multiple subtasks, the multiple subtasks are respectively assigned to the corresponding edge, fog and cloud computing resources. In order to balance the system load and response time, ensure that the tasks distributed on different computing nodes can work together, and ensure the consistency of the overall optimization goals. For example, the subtasks involving multi-phase calculations in the model are assigned to cloud computing resources, and the subtasks that do not involve multi-phase calculations in the model are assigned to edge computing resources or fog computing resources.

[0188] In summary, the embodiment of this specification provides an optimization scheduling method for an LNG receiving station, which constructs a multi-level digital twin model of the LNG receiving station; collects and obtains the target data of the LNG receiving station; inputs the target data into the multi-level digital twin model, and optimizes the target process parameters of the multi-level digital twin model to obtain optimized process parameters; and determines the model decision of the LNG receiving station based on the optimized process parameters. In this way, the operating status of the LNG receiving station can be fully and real-time perceived, and intelligent optimization and decision support can be performed, thereby significantly improving the operational efficiency, safety and economy of the LNG receiving station.

[0189] Based on the same inventive concept, Figure 2 As shown, the embodiment of the present invention further provides an LNG receiving station optimization scheduling system, including:

[0190] Multi-level digital twin model module, used to build a multi-level digital twin model of the LNG receiving station;

[0191] Multi-source data fusion module, used to collect and obtain target data of LNG receiving stations;

[0192] An intelligent optimization algorithm module, used for inputting the target data into the multi-level digital twin model, and optimizing the target process parameters of the multi-level digital twin model to obtain optimized process parameters;

[0193] A decision support module is used to obtain a model decision of the LNG receiving station according to the optimized process parameters.

[0194] In this embodiment, the multi-level digital twin model module is closely connected with the multi-source data fusion module. The multi-level digital twin model continuously receives real-time data updates through the multi-source data fusion module to maintain synchronization with the physical LNG receiving station. The multi-source data fusion module transmits the processed data to the intelligent optimization algorithm module to provide a basis for optimization calculations. The intelligent optimization algorithm module performs optimization calculations based on the updated digital twin model and pre-processed data, and passes the results to the decision support module. The decision support module receives the optimization results, presents them to the staff through a visual interface, and provides model decisions. This embodiment also includes an edge-fog-cloud collaborative computing architecture module, which runs through the other four modules and provides flexible and efficient computing support for each module.

[0195] Optional, multi-level digital twin model modules also include:

[0196] A macro model submodule, which uses a system dynamics method to construct a process model of the LNG receiving station;

[0197] The meso-model submodule uses a hybrid modeling approach to construct detailed models of key equipment and systems of the LNG receiving station;

[0198] The micro-model sub-module uses finite element analysis and computational fluid dynamics methods to construct a detailed model of the LNG receiving station.

[0199] Model updating submodule: realizes dynamic coupling and information interaction of multi-scale models.

[0200] Optionally, the multi-source data fusion module is also used for:

[0201] The target data is preprocessed by a hybrid filtering algorithm of Kalman filter and particle filter:

[0202]

[0203] Among them, x is the decision variable vector, which represents the target process parameter set that needs to be optimized. In the LNG receiving station, it can refer to operating parameters such as tank temperature, gasification rate, and delivery pressure. F(x) is the optimization objective function vector, which represents multiple optimization objective functions that need to be optimized simultaneously, f1(x), f2(x),..., f m (x) represents m different optimization objective functions, where m is the number of optimization objective functions. In an LNG receiving station, these optimization objective functions may include: maximizing energy efficiency, minimizing operating costs, minimizing safety risks, etc.

[0204] g i (x)≤0 is the inequality constraint, h j (x)=0 is an equality constraint, p is the number of inequality constraints, and q is the number of equality constraints.

[0205] The inequality constraints may represent operational limitations of the LNG terminal, such as upper tank pressure limits, minimum vaporization rate requirements, standby load limits, etc.

[0206] Equality constraints may represent conditions that must be precisely met in an LNG terminal, such as material balance equations, energy balance equations, specific process parameter requirements, etc.

[0207] The goal of this multi-objective optimization problem is to find a set of decision variables x such that all optimization objective functions f1(x) to f m (x) reaches the minimum value at the same time and satisfies all inequality constraints g i (x) and the equality constraint h j (x).

[0208] Optionally, the intelligent optimization algorithm module is also used to:

[0209] From multiple objective functions, determine the optimization objective function that needs to be optimized simultaneously;

[0210] Determining target process parameters to be optimized according to the optimization objective function;

[0211] The target process parameters are optimized by using the genetic algorithm NSGA-III to obtain the optimized process parameters.

[0212] Optionally, the mathematical expression of the NSGA-III algorithm is as follows:

[0213] minF(x)=[f1(x),f2(x),...,f m (x)] T

[0214] stg i (x)≤0,i=1,2,...,p

[0215] h j (x)=0,j=1,2,...,q

[0216] Among them, x is the decision variable vector, representing the target process parameter set that needs to be optimized, F(x) is the optimization objective function vector, representing multiple optimization objective functions that need to be optimized simultaneously, f1(x),f2(x),...,f m (x) represents m different optimization objective functions, m is the number of optimization objective functions, g i (x)≤0 is the inequality constraint, h j (x)=0 is an equality constraint, p is the number of inequality constraints, and q is the number of equality constraints.

[0217] The intelligent optimization algorithm module includes a hierarchical reinforcement learning optimization framework.

[0218] The hierarchical architecture is designed to effectively handle optimization problems at different time scales and decision levels in a complex system such as an LNG terminal. The details are as follows:

[0219] 1. Top layer: learning long-term policies using the Deep Deterministic Policy Gradient (DDPG) algorithm

[0220] The top level focuses on the long-term operational strategy of the LNG terminal, which may span from a few weeks to a few months. The main objectives of this level include:

[0221] Long-term resource scheduling: such as LNG procurement plans, equipment maintenance arrangements, etc.

[0222] Strategic decisions: For example, whether to expand facilities, adjust market strategies, etc.

[0223] Reasons for choosing the DDPG algorithm:

[0224] It can handle continuous action space and is suitable for continuous control problems of LNG receiving stations.

[0225] Combines the expressive power of deep learning with the sample efficiency of deterministic policy gradients.

[0226] You can learn long-term strategies and consider the long-term impact of your decisions.

[0227] Implementation:

[0228] State space: includes long-term market forecasts, equipment conditions, seasonal demand, etc.

[0229] Action space: long-term operating parameters, such as monthly LNG purchase volume, annual maintenance plan, etc.

[0230] Reward function: a comprehensive indicator reflecting long-term economic benefits, safety, and reliability.

[0231] 2. Middle layer: Using hierarchical reinforcement learning algorithm to handle mid-term task planning

[0232] The middle level focuses on medium-term planning of several days to weeks, with the main goals including:

[0233] Task decomposition: Break down the top-level long-term strategy into executable mid-term tasks.

[0234] Resource coordination: Allocate resources among different subsystems, such as gasification volume allocation, storage tank usage plan, etc.

[0235] Advantages of Hierarchical Reinforcement Learning:

[0236] The output of the top-level strategy can be used as a guide.

[0237] Able to handle multiple interrelated subtasks.

[0238] Improve learning efficiency and speed up convergence.

[0239] Implementation:

[0240] State space: includes current inventory levels, near-term demand forecasts, equipment status, etc.

[0241] Action space: Medium-term operation plan, such as daily gasification volume, ship unloading arrangements, etc.

[0242] Reward function: balances short-term benefits and long-term strategy execution.

[0243] 3. Bottom layer: Combining model predictive control (MPC) and reinforcement learning to achieve real-time control optimization

[0244] The bottom layer is responsible for real-time or near real-time control optimization, with a time scale of minutes to hours. The main goals include:

[0245] Real-time control: such as gasifier parameter adjustment, tank pressure control, etc.

[0246] Exception handling: Quickly respond to and handle emergencies.

[0247] Advantages of combining MPC and reinforcement learning:

[0248] MPC provides precise model-based control.

[0249] Reinforcement learning can adapt to environmental changes and continuously optimize control strategies.

[0250] Combining the two can improve the adaptability of the system while ensuring control accuracy.

[0251] Implementation:

[0252] State space: real-time process parameters, equipment status, etc.

[0253] Action space: real-time control parameters, such as valve opening, pump speed, etc.

[0254] Reward function: reflects short-term control performance, such as energy consumption, product quality, etc.

[0255] The three interact through information flow and control instructions:

[0256] The top level provides long-term goals and constraints for the middle level.

[0257] The middle level provides short-term goals and operating scope for the bottom level.

[0258] The underlying execution results and real-time data are fed back to the upper layer for strategy adjustment and optimization.

[0259] Through this layered architecture, the system can take into account long-term strategy, medium-term planning and short-term control at the same time, achieving comprehensive optimization of the LNG receiving station.

[0260] Optionally, constraints are dynamically generated based on the LNG receiving station knowledge graph and rule reasoning engine.

[0261] Optionally, the edge-fog-cloud collaborative computing architecture module is also used to:

[0262] Decompose the optimized computing task into multiple subtasks that can be calculated in parallel;

[0263] According to the computational load of the multiple subtasks, determine to allocate the multiple subtasks to edge, fog and cloud computing resources respectively.

[0264] Edge-Fog-Cloud Collaborative Computing Architecture Module

[0265] Optionally, the edge-fog-cloud collaborative computing architecture module also includes:

[0266] The edge computing sub-module is deployed at the LNG receiving station and is responsible for real-time data processing and rapid response.

[0267] The fog computing submodule performs intermediate computing and temporary storage to reduce the pressure on the cloud.

[0268] The cloud computing submodule performs large-scale data analysis, complex model training, and long-term strategy optimization.

[0269] The resource scheduling submodule dynamically allocates computing tasks to the appropriate computing layer.

[0270] Optionally, SHAP values ​​are used to interpret the model decisions as operational instructions executed by the staff.

[0271] Optionally, a genetic algorithm NSGA-III is used to optimize the target process parameters, and the optimized process parameters are determined through a Pareto optimal solution.

[0272] In summary, the embodiment of this specification provides an LNG receiving station optimization and scheduling system, which constructs a multi-level digital twin model of the LNG receiving station; collects and obtains the target data of the LNG receiving station; inputs the target data into the multi-level digital twin model, and optimizes the target process parameters of the multi-level digital twin model to obtain optimized process parameters; and determines the model decision of the LNG receiving station based on the optimized process parameters. In this way, the operating status of the LNG receiving station can be fully and real-time perceived, and intelligent optimization and decision support can be performed, thereby significantly improving the operational efficiency, safety and economy of the LNG receiving station.

[0273] Technicians in the relevant field can clearly understand that for the convenience and simplicity of description, the specific working process of the LNG receiving station optimization scheduling system described above can refer to the corresponding process in the aforementioned method, and will not be elaborated here.

[0274] The above are only various embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An optimization scheduling method for an LNG receiving station, characterized in that: include: Build a multi-level digital twin model of the LNG receiving station; Collect and obtain target data of LNG receiving stations; Inputting the target data into the multi-level digital twin model, and optimizing the target process parameters of the multi-level digital twin model to obtain optimized process parameters; A model decision of the LNG receiving station is determined based on the optimized process parameters.

2. The method according to claim 1, characterized in that The multi-level digital twin model of the LNG receiving station is constructed, including: A process model of the LNG receiving station is constructed using a system dynamics approach; A hybrid modeling approach is used to construct a detailed model of the key equipment and systems of the LNG receiving station; Finite element analysis and computational fluid dynamics methods are used to construct a detailed model of the LNG receiving station.

3. The method according to claim 1, characterized in that The inputting the target data into the multi-level digital twin model comprises: The target data is preprocessed by a hybrid filtering algorithm of Kalman filter and particle filter: in, is the state estimate at time k, K k is the Kalman gain, z k is the observed value, H k is the observation matrix, is the predicted value of the system state at time k based on the state estimation at time (k-1); The preprocessed target data is input into the multi-level digital twin model.

4. The method according to claim 1, characterized in that: The step of optimizing the target process parameters of the multi-level digital twin model to obtain optimized process parameters includes: From multiple objective functions, determine the optimization objective function that needs to be optimized simultaneously; Determining target process parameters to be optimized according to the optimization objective function; The target process parameters are optimized by using the genetic algorithm NSGA-III to obtain the optimized process parameters.

5. The method according to claim 4, characterized in that The mathematical expression of the genetic algorithm NSGA-III is as follows: minF(x)=[f1(x),f2(x),...,f m (x)] T s.t.g i (x)≤0,i=1,2,...,p h j (x)=0,j=1,2,...,q Among them, x is the decision variable vector, representing the target process parameter set that needs to be optimized, F(x) is the optimization objective function vector, representing multiple optimization objective functions that need to be optimized simultaneously, f1(x),f2(x),...,f m (x) represents m different optimization objective functions, m is the number of optimization objective functions, g i (x)≤0 is the inequality constraint, h j (x)=0 is an equality constraint, p is the number of inequality constraints, and q is the number of equality constraints.

6. The method according to claim 5, characterized in that Constraints are dynamically generated based on the knowledge graph and rule reasoning engine of the LNG receiving station.

7. The method according to claim 4, characterized in that The target process parameters are optimized by using the genetic algorithm NSGA-III to obtain the optimized process parameters, including: The target process parameters are optimized by using the genetic algorithm NSGA-III, and the optimized process parameters are determined by the Pareto optimal solution.

8. The method according to claim 1, characterized in that Optimizing the target process parameters of the multi-level digital twin model includes multiple subtasks that can be calculated in parallel; according to the computational amount of the multiple subtasks, the multiple subtasks are respectively allocated to edge, fog and cloud computing resources for computational processing.

9. The method according to claim 1, characterized in that: The method further comprises: SHAP values ​​are used to interpret the model decisions as operational instructions executed by the workers.

10. An optimized dispatching system for an LNG receiving station, characterized in that: include: Multi-level digital twin model module, used to build a multi-level digital twin model of the LNG receiving station; Multi-source data fusion module, used to collect and obtain target data of LNG receiving stations; An intelligent optimization algorithm module, used for inputting the target data into the multi-level digital twin model, and optimizing the target process parameters of the multi-level digital twin model to obtain optimized process parameters; A decision support module is used to obtain a model decision of the LNG receiving station according to the optimized process parameters.

Images