Digital twin management system and method applied to petrochemical wharf

Abstract

The invention relates to the technical field of digital twinning, in particular to a digital twinning management system and method applied to a petrochemical wharf, and the method comprises the steps: S1, collecting physical state parameters of a harbor district, a pipe district and a tank district in the petrochemical wharf in an operation process, constructing an asynchronous time sequence modeling network, and building a digital twinning mirror image body of the three-domain operation process, outputting a three-domain state evolution frame set; s2, generating a prediction evolution trajectory set through a dynamic compression time window and a prediction sliding evolution mechanism, and identifying evolution offset labels and linkage mismatch risk points among the three domains; and S3, combining the collected real-time state input, comparing a twinborn prediction trajectory set, triggering a difference feedback module, outputting a scheduling correction amount, reversely mapping a corrected action to a three-domain twinborn mirror image body, updating a twinborn evolution path, and forming a three-domain linkage prediction-control closed loop. According to the invention, the agility of three-domain cooperative scheduling and the steady-state operation capability of the whole system are improved.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a digital twin management system and method for use in petrochemical terminals. Background Technology

[0002] In the operation of petrochemical terminals, the berthing of ships in the port area, the transportation of pipelines in the pipeline area, and the storage and reception of tanks constitute three interconnected and time-sensitive key operation domains. There are significant dynamic coupling characteristics among the three domains. For example, the berthing rhythm affects the timing of pipeline pump start-up, and the start-up and shutdown of pump sets affect the tank level receiving capacity. However, due to the multiple sources and asynchronous nature of equipment status, fluid parameters, and scheduling instructions in each domain, current management methods often rely on manual scheduling and coordination of domain systems, making it difficult to achieve full-process information collaboration. This leads to frequent problems such as scheduling delays, redundant equipment operation, and decreased unloading efficiency. In addition, the lack of a predictive mechanism for linkage response under complex operation rhythms also makes it difficult for the system to identify and avoid potential risks in a timely manner.

[0003] Existing digital systems mostly focus on single-domain monitoring or simple parameter linkage, and have not yet formed a holographic mapping model and unified control closed loop for three-domain dynamic processes. On the one hand, existing systems fail to make full use of operation control logs and spatiotemporal identifiers for data alignment modeling, resulting in insufficient accuracy of multi-source asynchronous data fusion. On the other hand, traditional scheduling logic is difficult to adapt to high-frequency state evolution and cannot perform proactive control optimization based on evolution trends. Therefore, it is urgent to build a digital twin management method that can synchronously characterize the state changes of the three domains, dynamically predict evolution trends, and achieve closed-loop scheduling self-optimization, so as to improve the response efficiency and safety level of petrochemical terminals in highly complex operation scenarios. Summary of the Invention

[0004] This invention provides a digital twin management system and method for petrochemical terminals.

[0005] A digital twin management method for petrochemical terminals includes the following steps: S1. Construct an asynchronous twin mirror for dynamic processes in three domains: collect physical state parameters of the port area, pipeline area and tank area in the operation process of the petrochemical terminal, including berthing status, pump operation curve, main pipeline pressure difference change, tank area liquid level and its flow velocity response, and establish the mapping relationship between the three domain state data streams by combining operation control logs and spatiotemporal identifiers. By constructing an asynchronous time-series modeling network, a digital twin mirror image of the three-domain operation process is established. The digital twin mirror image includes a sub-mirror image of the state of key equipment in the three domains, a sub-mirror image of fluid behavior, and a sub-mirror image of scheduling actions. The digital twin mirror image is used to characterize the asynchronous evolution characteristics of the three-domain operation state in the physical world and output a set of three-domain state evolution frames. S2, Construct a twin-like inference model based on state evolution frames: Input the set of state evolution frames of the three domains into the twin-like inference model. The twin-like inference model constructs the coupled evolution path of the port area, pipeline area, and tank area under different operating rhythms, and simulates the fluid propagation lag, equipment response redundancy, and scheduling matching offset in the initial stage of unloading, steady-state transmission period, and terminal buffer period. Through dynamic compression time window and predictive sliding evolution mechanism, a set of predicted evolution trajectories is generated, and the evolution offset labels and linkage mismatch risk points between the three domains are identified. S3, Construct a self-evolutionary scheduling control mechanism based on twin feedback: Based on the evolutionary offset label and linkage mismatch risk point, combined with the real-time state input collected in the current physical world, compare with the twin predicted trajectory set, trigger the difference feedback module, and output the scheduling correction amount; The scheduling correction is used to adjust the berthing rhythm, pump start-up and shutdown sequence, and liquid level receiving window. The corrected actions are then mapped back to the three-domain twin mirror body to update the twin evolution path and realize the self-evolution adjustment mechanism of the digital twin mirror body, so as to form a three-domain linkage prediction-control closed loop.

[0006] Optionally, S1 includes: S11, Three-Domain Physical State Parameter Acquisition and Data Stream Mapping: Collect physical state parameters of three domains: port area, pipeline area and tank area of ​​petrochemical terminal, integrate the operation control log of operation control system, and mark all collected data with a unified spatiotemporal identifier, align and correlate data streams from different domains on the timeline, and establish the mapping relationship between the three domain state data streams. S12, Constructing an asynchronous time-series modeling network: Based on the data flow of the mapping relationship, an asynchronous time-series modeling network is constructed. The asynchronous time-series modeling network is composed of parallel and coupled time-series processing units, which respectively process data flows with different sampling frequencies and time lag characteristics in the three domains. By introducing timestamp embedding and lag compensation mechanisms, the causal and temporal dependencies between port berthing events, pipeline pressure fluctuations and tank level changes are learned and characterized. S13, Establish a digital twin mirror body and output a set of state evolution frames: Using the asynchronous temporal modeling network, establish a digital twin mirror body of the three-domain operation process, including a state sub-mirror of the three-domain key equipment, a fluid behavior sub-mirror, and a scheduling action sub-mirror. At fixed time intervals or triggered by key events, output a set of three-domain state evolution frames that integrates the instantaneous states of the three domains. The three-domain key equipment are physical execution equipment and sensing devices that have a direct impact on safety, controllability, and operation rhythm throughout the entire process of ship berthing, fluid transportation, and storage and reception. The key events include berthing completion events, pump start-up and shutdown actions, liquid level over-limit warnings, pressure difference abnormal changes, scheduling strategy switching, and equipment abnormality or alarm triggering.

[0007] Optionally, S11 includes: S111: Collects the berthing status of ships, including berthing speed, angle and real-time distance to the shore, through shipborne AIS and berthing assistance systems deployed on the quay shoreline; S112: The pump set's operating curves, including start-up and stop times, operating frequency, power and efficiency changes, are collected through the built-in vibration, temperature and current sensors. S113: Calculate and monitor the pressure difference changes in the main pipeline using pressure sensors at both ends of the main pipeline; S114: Collect tank level and flow rate responses through radar level gauges and flow meters in the tank area.

[0008] Optionally, the digital twin mirror image includes: Three-domain key equipment status sub-mirror: dynamically maps the real-time status of key equipment, including ships, oil booms, pumps, and valves; Fluid Behavior Sub-Mirror: Simulating the pressure propagation, flow rate changes, and phase characteristics of petrochemical fluids within a pipeline; Scheduling action sub-image: Records and maps the sequence of instructions issued by the central scheduling system.

[0009] Optionally, S2 includes: S21: Constructing a multi-rhythm coupled evolution path: Input the three-domain state evolution frame set into the twin inference model. Based on historical operation mode and real-time state, the twin inference model constructs the coupled evolution path of port area, pipeline area and tank area under different operation rhythms. The coupled evolution path simulates three key stages in the entire unloading operation process, including the initial unloading stage, steady-state transmission period and end buffer period. S22: Generate a set of predicted evolution trajectories: By using a dynamic time window compression mechanism, focus on the key domain state changes with the fastest evolution rate near the current moment, and use a predictive sliding evolution mechanism to generate a set of predicted evolution trajectories of the three domain states in the future by using the current state evolution frame as the initial condition and extrapolating along the coupled evolution path. S23: Identify evolutionary deviation and linkage mismatch risk: Compare the predicted evolutionary trajectory set with the baseline evolutionary path based on the operating procedure to identify evolutionary deviation labels between the three domains. The evolutionary deviation labels are used to quantify the degree of deviation between the real-time state and the expected state of any domain. Based on the evolutionary deviation labels, by analyzing the causal chain of cross-domain state changes, locate the linkage mismatch risk point caused by the chain reaction triggered by the deviation of a single domain.

[0010] Optionally, S21 includes: S211: Simulates the fluid propagation lag caused by pipeline filling and pressure build-up during the initial stage of unloading; S212: During steady-state transmission, simulate the redundancy of equipment response caused by pump performance fluctuations and valve adjustments; S213: During the end buffer period, simulate the scheduling match offset caused by the time difference between the tank farm receiving preparation and the execution of scheduling instructions.

[0011] Optionally, S3 includes: S31: Triggering Differential Feedback and Generating Scheduling Correction Amount: Based on the evolutionary offset label and linkage mismatch risk point, combined with the berthing status, pump group operation curve, main pipeline pressure difference and tank area liquid level data collected in real time from the physical world, a real-time status input is formed; the real-time status input is compared with the twin prediction trajectory set; when the status deviation of any domain exceeds the preset threshold or a specific risk point is triggered, the differential feedback module is triggered; the differential feedback module calculates and outputs the scheduling correction amount based on the preset control strategy library; S32: Execution scheduling correction and reverse mapping: The scheduling correction amount is sent to the execution system in the physical world to correct the execution action. The corrected action is used as a new boundary condition and input event and reverse mapped to the three-domain twin mirror body. S33: Realize self-evolution adjustment and prediction-control closed loop: Based on the corrected actions of the reverse mapping, update the twin evolution path, so that the inference basis of the digital twin mirror body is synchronized with the actual scheduling state of the physical world, realize the self-evolution adjustment mechanism of the digital twin mirror body, and form a three-domain linkage prediction-control closed loop that continuously performs situation prediction, decision optimization and real-time regulation.

[0012] Optionally, S31 includes: S311: Combining the collected berthing status, pump set operation curve, main pipeline differential pressure and tank area liquid level data, a real-time status input is formed; S312: Compare the real-time state input with the twin predicted trajectory set in real time to calculate the real-time deviation data of the three-domain state; S313, Offset and Risk Judgment: Based on the real-time deviation data, determine whether the evolutionary offset label exceeds the preset state deviation trigger threshold, or confirm whether a specific linkage mismatch risk point has been triggered. When the condition is met, immediately trigger the difference feedback module and output the scheduling correction amount.

[0013] Optionally, S32 includes: S321, Adjust port berthing schedule: Send the scheduling corrections related to the port area to the port area scheduling system to adjust the berthing rhythm of subsequent vessels, specifically including dynamically allocating berths, rearranging berthing time sequences, or updating mooring instructions; S322, Optimize the control of pump groups in the pipeline area: Send the scheduling correction amount involving the pipeline area to the pump control system of the pipeline area to optimize the start-up and shutdown sequence and operating frequency of pump groups in the pipeline area, specifically including delaying or advancing the start-up time of specific pump groups, adjusting the pump group operation combination or modifying the frequency setting value. S323, Replanning Tank Farm Reception Scheme: The scheduling correction amount involving the tank farm is sent to the tank farm management system, and the liquid level reception window of the tank farm is replanned, which specifically includes switching the target storage tank, adjusting the reception order, or modifying the safe liquid level threshold.

[0014] A digital twin management system for petrochemical terminals, used to implement the aforementioned digital twin management method, includes the following modules: The three-domain status acquisition and mapping module collects physical status parameters of the port area, pipeline area and tank area in the operation process, including berthing status, pump group operation curve, main pipeline differential pressure, tank area liquid level and flow rate response, and constructs the mapping relationship between the three-domain status data streams by combining operation control logs and spatiotemporal identifiers. Asynchronous temporal modeling module: Based on the state data stream, an asynchronous temporal modeling network is constructed to learn the asynchronous evolution relationship between the three domains and generate a digital twin mirror body, including a device state sub-mirror, a fluid behavior sub-mirror, and a scheduling action sub-mirror; State evolution simulation module: Receives the three-domain state evolution frame set output by the mirror body, and constructs the coupled evolution path of port area, pipeline area and tank area under different operating rhythms, and simulates the key dynamic processes in the initial unloading stage, steady-state transmission period and end buffer period; Predicted trajectory generation module: Based on the evolution path, it generates a set of predicted evolution trajectories of the three domain states over a future period of time through dynamic compression of time windows and sliding evolution mechanism; Offset identification and risk location module: compares the predicted trajectory with the baseline path to identify three-domain evolution offset labels and linkage mismatch risk points; Difference Feedback and Scheduling Correction Module: Compares the real-time status input with the predicted trajectory set. When a deviation or risk point is detected, a feedback mechanism is triggered, and the scheduling correction amount is output. Correction Execution and Mirror Update Module: Distributes the scheduling correction to the port area scheduling system, pipeline pump control system and tank area management system, and at the same time maps the correction results back to the twin mirror body to update the evolution path; Closed-loop control and self-evolution module: Based on the modified twin mirror image, continuous deduction, real-time decision-making and dynamic control are realized to construct a three-domain linkage predictive-control closed-loop system.

[0015] The beneficial effects of this invention are: This invention constructs an asynchronous temporal modeling network for port areas, pipeline areas, and tank farms, enabling the establishment of digital twin mirrors of the three-domain operation process in petrochemical terminal operations. This effectively captures the asynchronous evolution characteristics of the status of key equipment, fluid behavior, and scheduling actions. Combined with the state evolution frame output and multi-rhythm coupled path deduction mechanism, it improves the prediction accuracy of fluid lag, equipment redundancy, and scheduling deviations throughout the unloading process, providing more timely and accurate support for early warning and control of abnormal states.

[0016] This invention identifies evolutionary offset tags and linkage mismatch risk points, and constructs a feedback control mechanism for scheduling corrections, thus opening up a feedback loop path between the physical world and the digital twin. It can not only dynamically adjust the berthing rhythm, pump operation sequence, and liquid level receiving strategy, but also map the correction actions back to the twin system in real time, realizing the self-evolution and update of the mirror body, improving the agility of three-domain collaborative scheduling and the overall steady-state operation capability of the system. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a logical framework diagram of an embodiment of the present invention; Figure 2 This is a system block diagram of an embodiment of the present invention. Detailed Implementation

[0019] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. For some well-known technologies, those skilled in the art may also use other alternative methods to implement the invention. Moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0020] like Figure 1 As shown, a digital twin management method applied to petrochemical terminals includes the following steps: S1. Construct an asynchronous twin mirror for dynamic processes in three domains: collect physical state parameters of the port area, pipeline area and tank area in the operation process of the petrochemical terminal, including berthing status, pump operation curve, main pipeline pressure difference change, tank area liquid level and its flow velocity response, and establish the mapping relationship between the three domain state data streams by combining operation control logs and spatiotemporal identifiers. By constructing an asynchronous temporal modeling network, a digital twin mirror image of the three-domain operation process is established. The digital twin mirror image includes a sub-mirror image of the state of key equipment in the three domains, a sub-mirror image of fluid behavior, and a sub-mirror image of scheduling actions. The digital twin mirror image is used to characterize the asynchronous evolution characteristics of the three-domain operation state in the physical world and outputs a set of three-domain state evolution frames. S1 specifically includes: S11, Three-Domain Physical State Parameter Acquisition and Data Stream Mapping: Acquiring physical state parameters of three domains: the petrochemical terminal port area, pipeline area, and tank area, specifically including: (1) Port area berthing status data collection: Using the shipborne Automatic Identification System (AIS) and berthing assistance system deployed on the wharf shoreline, dynamic parameters of ship berthing are collected, including: ; ; ; Among them, the key equipment in the three domains includes physical execution devices and sensing devices that directly affect safety, controllability, and operational rhythm throughout the entire process from ship berthing to fluid transport to storage and reception, including berthing speed. The value ranges from 0.1 to 1.5. To ensure the safety and accuracy of ship berthing, excessively high speeds can easily cause collision damage, while excessively low speeds will affect efficiency. The berthing angle... The value range is 0-45°, and the berthing angle is usually controlled within 30°. In special cases (such as crosswinds or narrow berths), a larger angle is allowed, but exceeding 45° will increase the difficulty and risk of berthing. Ship-shore distance The value range is 0-10. The real-time berthing distance is used to judge the berthing status and risk warning. The general buffer distance is 2-6m. If it exceeds 10m, it is considered that the berthing has not been effective. 0m indicates that the berthing has been completed. (2) Pump unit operating parameter acquisition: The pump unit's operating curves are acquired through the pump unit's built-in sensor system, including vibration sensors, temperature sensors, and current sensors. The parameters include: ; ; ; ; The start and stop times are synchronized with a unified clock to record control response delays and scheduling rhythm, and the operating frequency is set according to the operation window. The value range is 10-90. Below 10Hz, the pump's starting force is insufficient; above 90Hz, efficiency decreases and the pump is prone to damage. The value range is 5-500, and the operating efficiency is... The value ranges from 40% to 95%, representing energy consumption and performance status. Pump efficiency is usually above 70%, and if it is below 40%, there may be problems such as blockage, cavitation or mechanical failure.

[0021] (3) Main pipe differential pressure acquisition: The inlet pressure is measured by pressure sensors installed at the beginning and end of the pipeline. Export pressure The pressure difference is expressed as: ; in, It characterizes the transport resistance and flow state. 0 kPa indicates quiescent. The normal operating pressure difference is generally between 50 and 300 kPa. A pressure difference exceeding 500 kPa may indicate blockage or overpressure risk. (4) Tank area liquid level and flow velocity acquisition: The liquid level height and inflow / outflow velocity of the oil in the tank are obtained through radar level gauges and flow meters in the tank area, and are expressed as: ; ; Among them, liquid level height The value ranges from 0 to 20. The liquid level depends on the height of the storage tank. For conventional horizontal tanks, it's 2-5m, and for vertical oil tanks, it's 10-20m. 0m represents an empty tank, and 20m is the safe height limit. Overflow is not recommended. (Inlet and outlet flow rates are also considered.) The value ranges from 0 to 5 and is used to evaluate loading and unloading rhythm and buffering capacity. (5) Time and space identification and log synchronization: Extract scheduling and operation logs from the operation control system. All data is given a unified spatiotemporal identifier, represented as: ; in, For the time of collection, Represents spatial coordinates; (6) Data stream mapping: based on unified spatiotemporal identifier stamps Align and interpolate the parameters of each domain on the timeline to establish a mapping matrix for the state data stream. , represented as: ; S12, Constructing an Asynchronous Temporal Modeling Network: Based on the established mapping relationships of the data streams, an asynchronous temporal modeling network is constructed, consisting of parallel and coupled temporal processing units, which respectively process data streams with different sampling frequencies and time lag characteristics in the three domains; by introducing timestamp embedding and lag compensation mechanisms, the causal and temporal dependencies between port berthing events, pipeline pressure fluctuations, and tank level changes are learned and characterized, specifically including: (1) Constructing an asynchronous temporal modeling network: using aligned three-domain state data streams An asynchronous time-series modeling network is constructed, which includes three time-series processing units. Data streams from the port area, pipeline area, and tank area are processed separately. (2) Temporal embedding mechanism: Each data point embeds its sampling time and the time interval between the sampling and the previous sampling to enhance the model's time awareness, as shown below: ; in, For spatiotemporal identification stamps, This indicates the time interval relative to the previous sample. This involves encoding timestamps and intervals into time embedding vectors that can be learned by the model. The embedding dimension is generally a 16-128 dimensional real vector, which enhances the model's time sensitivity to asynchronous dynamic processes. The dimension is set according to the neural network structure; too low a dimension results in insufficient information, while too high a dimension introduces the risk of overfitting. (3) Lag Compensation Mechanism: To adapt to the natural lag between the three domains, a lag compensation term is introduced to dynamically estimate the response offset between different domains, expressed as: ; in, Different state data channels are used to identify the data flow types involved in temporal coupling and causal modeling. This is a delayed compensation term, representing the data channel. Evolution to channel The estimated response lag time; (4) Temporal-causal relationship learning function: The goal of asynchronous modeling network is to learn the dynamic evolutionary relationship between the three domains, expressed as: ; in, For mirror state prediction, This represents an asynchronous modeling network, i.e., from a data matrix. The prediction mapping function to the twin state; In the actual operation of petrochemical terminals, the berthing operations in the port area, the operation and pressure response of pump units in the pipeline area, and the liquid level changes in the tank area, although belonging to the same operational process, are widely distributed in physical space, have heterogeneous sensor deployment, and have natural sampling asynchrony and response lag in the time dimension. This step designs an asynchronous temporal modeling network and adopts a strategy of domain-based processing + temporal alignment + lag compensation to improve the accuracy and adaptability of twin modeling.

[0022] Specifically: In terms of network structure design: independent time-series processing units are set up for port area, pipeline area and tank area respectively. Each unit is responsible for dynamic modeling within its own domain and allows processing of data streams with different sampling frequencies. This design can avoid high-frequency data being "slowed down" by low-frequency domains, thus improving modeling efficiency, while preserving the independent physical characteristics of each domain.

[0023] The introduction of a timestamp embedding mechanism enhances the network's ability to perceive the time of asynchronous data streams. By taking the sampling time of each sample and its interval with the previous sample as explicit input, the network can identify rhythm differences and timing patterns between different domains, making it particularly suitable for scenarios where sudden events and steady-state flow alternate in dock operations.

[0024] Setting up a lag compensation mechanism to model the delay relationship between cross-domain signals can effectively improve the model's ability to understand causal order and can identify scheduling lags or abnormal situations in advance.

[0025] The final output twin prediction state not only contains the current prediction values ​​of each domain, but also integrates the temporal linkage mode between the three domains.

[0026] In summary, this step provides a clear and logically sound modeling path for typical industrial scenarios involving asynchronous, delayed response, and multi-domain linkage, ensuring the timeliness, accuracy, and generalizability of the digital twin system. It has strong engineering practical value and a solid theoretical foundation.

[0027] S13, Establish a digital twin mirror image and output a set of state evolution frames: using Building a complete digital twin mirror It is used to simulate and map the state evolution characteristics of port areas, pipeline areas, and tank farms during asynchronous evolution processes, including equipment state sub-mirrors. Fluid behavior sub-image and scheduling action sub-images , respectively represented as: S131, Sub-mirror of status of critical equipment in three domains This is used to dynamically map the real-time operating status of key equipment in the three domains of a petrochemical terminal. Key equipment includes vessels, oil booms, pumps, and valves. The state of each type of equipment can be abstracted as the following vector: ; in, Indicates the first Key equipment in time The system monitors the operating status, including start / stop status, pose information, and operating indicators. The status data is derived from various sensor parameters and control log information collected in real time in S11. Total number of devices; The critical equipment status sub-image is represented as follows: ; in, An assembly of equipment, such as ships, pumps, valves, and oil delivery arms. For each device's state vector, An asynchronous update mechanism for sub-mirrors updates device status based on sensor input and control logs; S132, Fluid Behavior Sub-Mirror Used to simulate the nonlinear propagation characteristics of petrochemical fluids in pipeline areas, and to characterize the evolution process of multi-physics field variables, its basic form is expressed as: ; in, It is a location time The local pressure of the pipeline is measured in the range of 0-2500 kPa, taking into account both low-pressure and high-pressure transmission scenarios. It is a location time The local volumetric flux, ranging from 0 to 1.5 m / s, is crucial; excessively high values ​​can easily lead to cavitation or erosion, while excessively low values ​​result in poor efficiency. It is a location time The fluid density ranges from 600 to 950 kg / m³. It is a location time The volume fraction of the gas phase ranges from 0 to 1, where 0 represents pure liquid and 1 represents pure gas. The fluid behavior sub-mirror representation is as follows: ; in, The jurisdictional area includes the main pipeline and its nodes. For position time The fluid state field, It supports hysteresis modeling for the propagation model (Navier-Stokes simplified solution).

[0028] S133, Scheduling Action Sub-image Used to map the operation instructions and scheduling behaviors issued by the central dispatch system. The sequence of scheduling instructions is represented as follows: ; in, It is the first Each scheduling action command (such as starting pump 1, closing valve 2, adjusting the berthing window) enables bidirectional interaction between all scheduling actions and the state sub-mirrors, forming a control closed loop. The scheduling action sub-image is represented as: ; in, The scheduling actions include starting the pump, closing the valve, adjusting the berthing rhythm, and setting the liquid level window. For scheduling action sequences, It is a scheduling logic parser used to parse action structures from control system logs or input rules.

[0029] S134, the obtained key equipment status sub-image, fluid behavior sub-image, and scheduling action sub-image are coupled to form an asynchronously updated twin image, represented as: ; S135, according to the set time window Or key event triggering conditions The output state evolution frame set is represented as: ; in, For the set of state evolution frames, For a moment The twin mirror state, For the output frame time points, For frame index number, For a fixed time window, the value ranges from 5 to 300 seconds. Critical events refer to operational nodes or abnormal states that have phased, sudden, or significant linkage effects during the three-domain operations of a petrochemical terminal. Specifically, these include berthing completion events, pump start-up and shutdown actions, liquid level over-limit warnings, abnormal pressure changes, scheduling strategy switching, and equipment abnormalities or alarm triggering.

[0030] This step addresses the dynamic linkage process of the operational status of the port area, pipeline area, and tank area in a petrochemical terminal. It designs a structured digital twin mirror construction method, which comprehensively depicts the evolution of the physical system in the three dimensions of equipment, fluid, and scheduling by constructing three types of sub-mirrors that are asynchronously updated and coupled in evolution. Based on a fixed time rhythm or key event triggering mechanism, it continuously outputs a set of state evolution frames, providing structured input for subsequent deduction analysis and control feedback.

[0031] The combined structure design of the mirror body adopts an asynchronous coupling method, which allows each sub-mirror to be updated independently at different time steps and to achieve state linkage based on the data flow interface. This design not only preserves the model's responsiveness at multiple time scales, but also ensures the logical consistency of cross-domain behavior.

[0032] The output mechanism of the state evolution frame combines a fixed time window output and a key event triggered output strategy to ensure the continuity of the modeling rhythm and ensure timely response to sudden key events (such as liquid level overflow and pressure difference exceeding limits).

[0033] In summary, this step, based on asynchronously updated structured sub-images, realizes a unified mapping of multi-dimensional states of petrochemical terminals. It also forms a state frame stream through a time + event dual-trigger mechanism, which has good timeliness, structural stability and scalability, providing an executable modeling foundation for twin intelligent management in complex operation scenarios.

[0034] S2, Construct a Siamese inference model based on state evolution frames: Input the three-domain state evolution frame set into the Siamese inference model, and the Siamese inference model constructs the coupled evolution path of the port area, pipeline area, and tank area under different operating rhythms, simulates fluid propagation lag, equipment response redundancy, and scheduling matching offset in the initial unloading period, steady-state transmission period, and terminal buffer period; through dynamic compression time window and predictive sliding evolution mechanism, generate a set of predicted evolution trajectories, and identify evolution offset labels and linkage mismatch risk points between the three domains; S2 specifically includes: S21, Constructing a multi-rhythm coupled evolution path: The state evolution frame set Input to twin inference model The model is based on historical job samples Input with current state The coupled evolution path function of the three domains under different operation rhythms is constructed as follows: ; in, It is a coupled evolution path of three domains (port area, pipeline area, tank farm). For the current time, It is a sample of historical assignments. It is a state mirror frame, and the coupling path includes dynamic mechanisms in three typical stages: (1) Initial stage of unloading: ; in, This indicates the fluid propagation lag time caused by pipe filling / pressure build-up. It represents the response delay between the start of fluid flow and the establishment of stable pressure in the system. A larger value indicates lower filling efficiency or obstruction in the pipe / fluid characteristics, and is an important indicator in the initial stage of unloading. This indicates the time at which the flow rate begins to rise, representing the point in time when the flow rate in the pipeline first shows a significant increase after the unloading operation is started. The value ranges from 0 to 60 seconds after the operation begins, characterizing the instantaneous response of the fluid at the start of flow. It is often affected by pump start-up / shutdown and valve opening delays. This indicates the time when the pressure reaches the set value, with a range of 30-180 seconds after the start of operation. The pressure rise lags behind the flow rate, especially in long-distance pipelines or oil and gas mixed transportation; it takes a relatively long time to establish a stable pressure. (2) Steady-state transmission period: ; in, This indicates equipment response redundancy caused by fluctuations in pump unit performance. This indicates the current pump status, frequency 10-90Hz, power 5-500kW. This indicates the pump reference state under steady-state conditions, with a frequency of 10-90Hz and a power of 5-500kW. It is used as a benchmark for comparing observed states and to determine whether the pump deviates from the normal stable operating curve. (3) End-of-term buffer period: ; in, This represents the scheduling matching offset time caused by instruction execution delay. It indicates the time deviation between the scheduling instruction and the device response, reflecting the control closed-loop delay. The smaller the value, the better the system's real-time performance and coordination; a larger value may lead to mismatch in the three-domain linkage. This indicates the time when the scheduling command is issued, representing the point in time when the scheduling system issues the execution instruction (such as starting / stopping the pump or adjusting the receiving window). This indicates the actual time when the action takes effect, representing the time when the field equipment responds to the dispatch command and actually completes the state switch. It may lag behind the command by 5-60 seconds. The lag is caused by network latency, slow action of the actuator, or queue blockage. Excessive delay will cause matching offset problems. This step introduces a twin simulation model to structurally model the dynamic evolution of the port area, pipeline area, and tank area during operation. The core objective is to construct a multi-rhythm coupled evolution path to accurately reflect the system linkage characteristics and lag patterns at different stages under actual working conditions.

[0035] To align with the actual unloading process at petrochemical terminals, the coupled path modeling is subdivided into three typical stages: Initial stage of unloading: This mainly simulates the physical response lag between the initial change in flow rate and the establishment of pressure during the initial filling of the pipeline. This lag is mostly caused by factors such as pipeline inertia, gas entrainment, and slow valve opening. If the lag time exceeds expectations, it will affect the scheduling rhythm and pump start-up and shutdown logic.

[0036] Steady-state transmission phase: During this phase, the system operation tends to stabilize, but pump performance may still fluctuate, such as frequency jitter and efficiency decline. These fluctuations will manifest as "response redundancy" where the pump status deviates from the reference operating curve, requiring continuous monitoring and evaluation to determine whether scheduling intervention or equipment maintenance is needed.

[0037] End-of-pipe buffering phase: When materials are about to be unloaded or the tank farm's receiving capacity is nearing its limit, it often happens that scheduling instructions are issued prematurely but the equipment has not yet responded in time. This phase focuses on the "scheduling match offset time," which is the delay between the scheduling system's logical time and the equipment's actual execution time. If the offset is too large, it can easily lead to risks such as tank farm overflows and sudden pressure changes.

[0038] Through the above three-stage coupled modeling, not only can the historical rhythm be reproduced, but a dynamic path with stage characteristics can also be quickly generated for the current state, providing a high-fidelity modeling foundation for subsequent state prediction, offset identification and scheduling correction.

[0039] S22, Generate the predicted evolution trajectory set: Based on the constructed coupled evolution path, set a dynamically compressed time window. It employs a sliding prediction propagation mechanism to advance the current frame. Along the path Extrapolating forward, it can be represented as: ; in, It is a set of predicted evolutionary trajectories. It is a prediction step window, with a value range of 300-1200s, covering typical operation response lag time and scheduling execution time. This is a dynamically compressed time window used to limit the time period of high evolution rate key state changes that the model prioritizes during prediction. The value ranges from 30 to 180 seconds. If the time is too long, it will introduce irrelevant noise; if it is too short, it may miss sudden events. The coupled evolution path is a joint evolution trajectory function derived from the port area, pipeline area, and tank area under historical operation samples and current state conditions; This step, based on the completion of the three-domain coupled evolution path construction, introduces a sliding prediction propagation mechanism to generate a set of predicted evolution trajectories for future states in real time, thereby achieving advanced perception and risk prediction of the port operation status. In order to improve the real-time performance and sensitivity of the prediction, a dynamic compressed time window is designed to select the key state segments with the most drastic state changes, the most concentrated risks, and the highest prediction value near the current moment, ensuring that computing resources are concentrated on time periods with responsive significance and avoiding ineffective extrapolation of the whole.

[0040] Within this window, the system continuously outputs multiple frames of future state images according to a preset prediction step size (e.g., 5-20 minutes), forming a set of time-series state snapshots, called the predicted evolution trajectory set. This not only preserves the multi-domain coupling characteristics of the original state but also embeds time progression, which can be used to identify state change trends. Early detection of potential risk signals provides a predictive basis for subsequent offset comparison and scheduling correction mechanisms; By designing a sliding prediction mechanism, the function of monitoring the current state and predicting the future state has been expanded, enabling the entire digital twin model to have the ability to make forward-looking decisions and dynamic control, thereby improving the operational safety and scheduling optimization level of complex terminal operations.

[0041] S23, Identify evolutionary drift and linkage mismatch risks: [This involves analyzing the predicted trajectory set...] Compared with the baseline trajectory generated based on the work procedure By comparing the results, the offset label function of each domain is calculated. The cross-domain causal path analysis function is used to identify the chain reaction caused by the offset of a certain domain and extract the risk points of linkage mismatch. , represented as: ; ; in, It is the first Domain in time The offset, or offset label, measures the degree of deviation between the current evolution and the expected state. A larger offset indicates a more severe deviation, which may indicate potential anomalies or rhythm disorders. This serves as the baseline evolution trajectory set, with the same structure as the predicted trajectory, and the number of frames and time steps are consistent. For at any time Next Predicted twin mirror state of the domain, For at any time Next The baseline twin mirror state of the domain, This is a cross-domain causal graph function used to analyze the linkage mismatch logic chain caused by the offset between three domains and output potential risk points. Based on the offset between the predicted state and the baseline state, this function is used to further identify the cross-domain linkage mismatch problem caused by the deviation of a certain subsystem. It is the core module of risk transmission path analysis in the three-domain coupling model. Its main goal is to identify causal chain mismatch risk points, that is, due to the state deviation of a certain work domain (such as the port area), the other two domains (such as the pipeline area or tank area) cannot respond as expected, thereby disrupting the overall scheduling rhythm and work efficiency.

[0042] S3, Construct a self-evolving scheduling control mechanism based on twin feedback: Based on the evolutionary offset label and linkage mismatch risk point, combined with the real-time state input collected in the current physical world, compare with the twin predicted trajectory set, trigger the difference feedback module, and output the scheduling correction amount; The scheduling correction is used to adjust the berthing rhythm, pump start-up and shutdown sequence and liquid level receiving window, and the corrected actions are back-mapped to the three-domain twin mirror body to update the twin evolution path, realize the self-evolution adjustment mechanism of the digital twin mirror body, so as to form a three-domain linkage prediction-control closed loop. S3 specifically includes: S31, triggers difference feedback and generates scheduling correction amount: S311: Evolutionary offset label obtained based on the aforementioned identification Risk points of linkage mismatch Combining the real-time state data collected from the physical system, a unified input vector is formed, represented as: ; in, The real-time berthing speed is set between 0 and 0.5 m / s. Speeds exceeding 0.5 m / s may pose a risk of berthing collisions, while speeds below 0.1 m / s are too slow and negatively impact turnaround efficiency. This refers to the current frequency of the pump unit, ranging from 20-60Hz. Frequency below 20Hz results in poor pump efficiency, while frequency above 60Hz poses an overload risk. This is the differential pressure of the main pipeline, with a value range of 0-3000 kPa, which meets the general differential pressure requirements for petrochemical fluid transportation. Exceeding the upper limit will trigger an overpressure alarm. This refers to the liquid level height in the tank area, with a value range of 0-15m, facilitating segmented liquid level control. It is the real-time physical state input vector; S312: Compare the real-time state input with the corresponding time-series twin predicted state. By performing a difference comparison, a real-time deviation vector is obtained, which calculates the multidimensional difference between the actual state and the predicted mirror state, quantifying the distance between the current operation and the expected behavior, and is expressed as: ; in, This is a multidimensional state deviation; S313: If any state variable exceeds the preset state deviation trigger threshold There may be any risk of mismatch in linkage. ,Right now If this occurs, the difference feedback module is triggered, and the difference feedback module uses a preset scheduling control strategy library. Calculate the correction scheme and output the scheduling correction amount, expressed as: ; in, This is the threshold for triggering state deviations, with a range of 5%-10%. Too low a value will cause frequent triggering, while too high a value will disable the warning function. This is a set of risk points for linkage mismatch. For the control strategy knowledge base, To output the scheduling correction amount, predefine control responses for common discrepancies and risk points to avoid real-time optimization delays and directly output the correction amount.

[0043] The operational processes at petrochemical terminals are inherently asynchronous, and time delays and response redundancies between different domains can easily lead to inconsistencies in state. This step introduces a differential feedback mechanism to dynamically capture potential mismatches during asynchronous evolution; without this feedback mechanism, the digital twin can only observe but cannot intervene. This step enables the system to possess self-sensing and self-adjusting capabilities, forming a complete control closed loop. In high-risk processes such as terminal unloading, any state deviation can trigger significant hidden dangers; therefore, real-time response and dynamic correction are essential to prevent the situation from escalating.

[0044] S32, Perform scheduling correction and reverse mapping: based on the scheduling correction amount The control instructions are issued to the physical execution agencies of the port area, management area, and tank area, including three types of domain-oriented scheduling optimization instructions: (1) Port berthing rhythm optimization: Adjust the vessel entry sequence according to the vessel berthing rhythm deviation, as shown in: ; (2) Pump group regulation in the pipeline area: By modifying the pump group start-up and shutdown logic and the frequency conversion operation scheme, the pump load balance is achieved, which is expressed as: ; (3) Re-division of tank area liquid level receiving window: When an over-level risk or insufficient receiving efficiency is detected, the upper and lower thresholds of the tank capacity are adjusted, as shown below: ; After the physical system execution is complete, these correction instructions are used as new boundary conditions and mapped back to the twin mirror image to update the system state. ; in, For Hong Kong amendments, For district correction items, For tank farm correction items, The expected arrival time window (in seconds) for the next scheduled vessel to berth is set, ranging from -900 to 900 seconds. Fine-tuning within ±15 minutes is generally acceptable, depending on terminal operational tolerances and vessel scheduling flexibility. The scheduling time for pump start-up and shutdown (unit: seconds) indicates when the system plans to switch the pump state. This is the minimum permissible liquid level threshold for the tank area, ranging from 0.5 to 2.0 meters. Low liquid levels can easily lead to pump cavitation or liquid level detection failure; therefore, a bottom liquid level of at least 0.5 meters is typically maintained to ensure operational safety. The maximum allowable liquid level threshold for the tank area is set at 90%–95% of the total height, with a 5%–10% safety buffer space to prevent liquid splashing or overflow.

[0045] Adjusting port berthing scheduling involves sending scheduling corrections related to the port area to the port area scheduling system to adjust the berthing rhythm of subsequent vessels. This includes dynamically allocating berths, rearranging berthing time sequences, or updating mooring instructions. Optimizing pipeline pump control involves sending scheduling corrections related to the pipeline area to the pipeline pump control system to optimize the start-up and shutdown sequence and operating frequency of pipeline pumps. This includes delaying or advancing the start-up time of specific pumps, adjusting pump operation combinations, or modifying frequency settings. Replanning tank farm receiving schemes involves sending scheduling corrections related to tank farms to the tank farm management system to replan the tank farm's liquid level receiving window. This includes switching target tanks, adjusting the receiving order, or modifying safe liquid level thresholds.

[0046] If scheduling corrections remain only at the physical level, the twin's predictions will gradually become distorted, losing their feedback control value. Reverse mapping ensures synchronization between the digital twin and the physical system state; in dynamic systems, scheduling behavior itself is the primary driver of state variable changes; if corrections are not reflected in the twin's evolution path, system predictions will deviate from the actual operating trajectory, leading to future feedback failures. This step performs scheduling corrections and reverse mapping, achieving stable predictions and adaptive scheduling across multiple scenarios through a continuous twin-physical synchronization mechanism, facilitating subsequent expansion to more scheduling factors and nonlinear response behaviors.

[0047] S33, after receiving the corrected scheduling input, the twin inference module will recalculate the subsequent evolution path using the updated mirror image as the initial condition, as shown below: ; Using the latest corrected state as input, update the twin simulation path to achieve resynchronization between the twin system and the physical system. in, The new coupled evolution path generated after correction, For twin path deduction model, These are historical task samples used for path fitting; This is a digital twin state evolution path prediction model for the coordinated operation of port area, management area, and tank farm. Its core objective is to predict potential problems such as state shifts, resource imbalances, and operation timing mismatches in future multi-stage operations based on historical operation evolution patterns and the current state frame set. The model includes the following modules: 1. Historical Pattern Memory Encoder: Using typical unloading operation processes marked in the operation log, clustering and compression encoding are performed on the data of the three stages of unloading initial stage, steady-state transmission and end buffering to extract behavioral evolution trajectory features under multiple rhythms; a multi-stage causal relationship mapping map is established to support the generation of causal structure for subsequent inference paths.

[0048] 2. State Frame Driven Inference Module: Using the current set of state frames output by the twin mirror as the initial input, key state variables (such as critical change points of pressure threshold, inflection points of liquid level, etc.) are extracted through a sliding window strategy; based on the feedforward inference engine, the current state is embedded into the most similar evolution path on the historical trajectory to complete the recursive prediction of the state at the next moment.

[0049] 3. Asynchronous Evolution Prediction Network: Introduces a multi-channel GRU or Transformer variant architecture to perform asynchronous time modeling of state characteristics in different domains (port / pipeline / tank); each channel is driven by a different sampling rate, supporting the alignment and recursion of non-synchronous sampling data; outputs cross-domain evolution prediction trajectory.

[0050] 4. Risk Label Generator: Compares the predicted trajectory with the standard evolution path in the operating procedure and calculates the real-time deviation. State offset label: measures the degree of deviation in the behavior of individual devices or fluids; Linkage mismatch labeling: Identifies potential causal chains in which a change in the state of one domain triggers abnormal responses in other domains; Output the difference label set for the feedback control module to use.

[0051] 5. Sliding update mechanism: The model has the ability to update online. It can dynamically adjust the inference strategy and embedding parameters based on the correction information returned by the difference feedback module to achieve adaptive iterative optimization.

[0052] This process can be continuously iterated. Each round involves feedback on differences, correction execution, and path updates to achieve a self-evolutionary adjustment mechanism for the twin mirror image. Ultimately, it forms a closed-loop system of linkage prediction and control between the three domain states, which has the capabilities of continuous prediction, dynamic correction, and real-time control.

[0053] like Figure 2 As shown, a digital twin management system for petrochemical terminals, used to implement the above method, includes the following modules: The three-domain status acquisition and mapping module collects physical status parameters of the port area, pipeline area and tank area in the operation process, including berthing status, pump group operation curve, main pipeline differential pressure, tank area liquid level and flow rate response, and constructs the mapping relationship between the three-domain status data streams by combining operation control logs and spatiotemporal identifiers. Asynchronous temporal modeling module: Based on the state data stream, an asynchronous temporal modeling network is constructed to learn the asynchronous evolution relationship between the three domains and generate a digital twin mirror body, including a device state sub-mirror, a fluid behavior sub-mirror, and a scheduling action sub-mirror; State evolution simulation module: Receives the three-domain state evolution frame set output by the mirror body, and constructs the coupled evolution path of port area, pipeline area and tank area under different operating rhythms, and simulates the key dynamic processes in the initial unloading stage, steady-state transmission period and end buffer period; Predicted trajectory generation module: Based on the evolution path, it generates a set of predicted evolution trajectories of the three domain states over a future period of time through dynamic compression of time windows and sliding evolution mechanism; Offset identification and risk location module: compares the predicted trajectory with the baseline path to identify three-domain evolution offset labels and linkage mismatch risk points; Difference Feedback and Scheduling Correction Module: Compares the real-time status input with the predicted trajectory set. When a deviation or risk point is detected, a feedback mechanism is triggered, and the scheduling correction amount is output. Correction Execution and Mirror Update Module: Distributes the scheduling correction to the port area scheduling system, pipeline pump control system and tank area management system, and at the same time maps the correction results back to the twin mirror body to update the evolution path; Closed-loop control and self-evolution module: Based on the modified twin mirror image, continuous deduction, real-time decision-making and dynamic control are realized to construct a three-domain linkage predictive-control closed-loop system.

[0054] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0055] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A digital twin management method applied to petrochemical terminals, characterized in that, Includes the following steps: S1: Collect physical state parameters of the port area, pipeline area and tank area in the operation process of the petrochemical terminal, including berthing status, pump operation curve, main pipeline pressure difference change, tank area liquid level and its flow velocity response, and establish the mapping relationship between the three domain status data streams by combining operation control logs and spatiotemporal identifiers. By constructing an asynchronous time-series modeling network, a digital twin mirror image of the three-domain operation process is established. The digital twin mirror image includes a sub-mirror image of the state of key equipment in the three domains, a sub-mirror image of fluid behavior, and a sub-mirror image of scheduling actions. The digital twin mirror image is used to characterize the asynchronous evolution characteristics of the three-domain operation state in the physical world and output a set of three-domain state evolution frames. S2: Input the three-domain state evolution frame set into the twin inference model. The twin inference model constructs the coupled evolution path of the port area, pipeline area and tank area under different operating rhythms, and simulates the fluid propagation lag, equipment response redundancy and scheduling matching offset in the initial unloading period, steady-state transmission period and end buffer period. By dynamically compressing the time window and using a predictive sliding evolution mechanism, a set of predicted evolutionary trajectories is generated, and evolutionary offset labels and linkage mismatch risk points between the three domains are identified. S3: Based on the evolutionary offset label and linkage mismatch risk point, combined with the real-time state input collected in the current physical world, compare with the twin prediction trajectory set, trigger the difference feedback module, and output the scheduling correction amount; The scheduling correction is used to adjust the berthing rhythm, pump start-up and shutdown sequence, and liquid level receiving window. The corrected actions are then mapped back to the three-domain twin mirror body to update the twin evolution path and realize the self-evolution adjustment mechanism of the digital twin mirror body, so as to form a three-domain linkage prediction-control closed loop.

2. The digital twin management method for petrochemical terminals according to claim 1, characterized in that, S1 includes: S11: Collect physical status parameters of the three areas of the petrochemical terminal port area, pipeline area and tank area, integrate the operation control log of the operation control system, and mark all collected data with a unified spatiotemporal identifier. Align and correlate data streams from different areas on the timeline and establish a mapping relationship between the status data streams of the three areas. S12: Based on the data flow of the mapping relationship, an asynchronous time series modeling network is constructed. The asynchronous time series modeling network is composed of parallel and coupled time series processing units, which respectively process data flows with different sampling frequencies and time lag characteristics in the three domains. By introducing timestamp embedding and lag compensation mechanisms, the causal and temporal dependencies between port berthing events, pipeline pressure fluctuations and tank level changes are learned and characterized. S13: Using the asynchronous temporal modeling network, establish a digital twin mirror of the three-domain operation process, including a sub-mirror of the state of key equipment in the three domains, a sub-mirror of fluid behavior, and a sub-mirror of scheduling actions. At fixed time intervals or triggered by key events, output a set of three-domain state evolution frames that integrate the instantaneous states of the three domains. The key equipment in the three domains are physical execution equipment and sensing devices that have a direct impact on safety, controllability, and operation rhythm throughout the entire process of ship berthing, fluid transportation, and storage and reception. The key events include berthing completion events, pump start-up and shutdown actions, liquid level over-limit warnings, abnormal changes in pressure difference, switching of scheduling strategies, and equipment abnormalities or alarm triggers.

3. The digital twin management method for petrochemical terminals according to claim 2, characterized in that, S11 includes: S111: Collects the berthing status of ships, including berthing speed, angle and real-time distance to the shore, through shipborne AIS and berthing assistance systems deployed on the quay shoreline; S112: The pump set's operating curves, including start-up and stop times, operating frequency, power and efficiency changes, are collected through the built-in vibration, temperature and current sensors. S113: Calculate and monitor the pressure difference changes in the main pipeline using pressure sensors at both ends of the main pipeline; S114: Collect tank level and flow rate responses through radar level gauges and flow meters in the tank area.

4. The digital twin management method for petrochemical terminals according to claim 2, characterized in that, The digital twin mirror image includes: Three-domain key equipment status sub-mirror: dynamically maps the real-time status of key equipment, including ships, oil booms, pumps, and valves; Fluid Behavior Sub-Mirror: Simulating the pressure propagation, flow rate changes, and phase characteristics of petrochemical fluids within a pipeline; Scheduling action sub-image: Records and maps the sequence of instructions issued by the central scheduling system.

5. A digital twin management method for petrochemical terminals according to claim 2, characterized in that, S2 includes: S21: Input the three-domain state evolution frame set into the twin inference model. Based on historical operation mode and real-time state, the twin inference model constructs the coupled evolution path of port area, pipeline area and tank area under different operation rhythms. The coupled evolution path simulates three key stages in the entire unloading operation process, including the initial unloading stage, steady-state transmission period and end buffer period. S22: By using a dynamic compression time window mechanism, we focus on the key domain state changes with the fastest evolution rate near the current moment, and adopt a predictive sliding evolution mechanism to generate a set of predicted evolution trajectories of the three domain states in the future by using the current state evolution frame as the initial condition and extrapolating along the coupled evolution path. S23: Compare the predicted evolution trajectory set with the baseline evolution path based on the operating procedure to identify evolutionary offset labels between the three domains. The evolutionary offset labels are used to quantify the degree of deviation between the real-time state and the expected state of any domain. Based on the evolutionary offset labels, by analyzing the causal chain of cross-domain state changes, locate the linkage mismatch risk point caused by the chain reaction triggered by the offset of a single domain.

6. The digital twin management method for petrochemical terminals according to claim 5, characterized in that, S21 includes: S211: Simulates the fluid propagation lag caused by pipeline filling and pressure build-up during the initial stage of unloading; S212: During steady-state transmission, simulate the redundancy of equipment response caused by pump performance fluctuations and valve adjustments; S213: During the end buffer period, simulate the scheduling match offset caused by the time difference between the tank farm receiving preparation and the execution of scheduling instructions.

7. A digital twin management method for petrochemical terminals according to claim 5, characterized in that, S3 includes: S31: Based on the evolutionary offset label and linkage mismatch risk point, combined with the berthing status, pump group operation curve, main pipeline pressure difference and tank area liquid level data collected in real time from the physical world, a real-time status input is formed; the real-time status input is compared with the twin prediction trajectory set; when the status deviation of any domain exceeds the preset threshold or triggers a specific risk point, the difference feedback module is triggered; the difference feedback module calculates and outputs the scheduling correction amount based on the preset control strategy library; S32: The scheduling correction amount is sent to the execution system in the physical world to correct the execution action. The corrected action is used as a new boundary condition and input event and mapped in reverse to the three-domain twin mirror body. S33: Based on the corrected actions of the reverse mapping, update the twin evolution path, so that the inference basis of the digital twin mirror body is synchronized with the actual scheduling state of the physical world, realize the self-evolution adjustment mechanism of the digital twin mirror body, and form a three-domain linkage prediction-control closed loop that continuously performs situation prediction, decision optimization and real-time regulation.

8. A digital twin management method for petrochemical terminals according to claim 7, characterized in that, S31 includes: S311: Combining the collected berthing status, pump set operation curve, main pipeline differential pressure and tank area liquid level data, a real-time status input is formed; S312: Compare the real-time state input with the twin predicted trajectory set in real time to calculate the real-time deviation data of the three-domain state; S313: Offset and Risk Judgment: Based on the real-time deviation data, determine whether the evolutionary offset label exceeds the preset state deviation trigger threshold, or confirm whether a specific linkage mismatch risk point has been triggered. When the conditions are met, immediately trigger the difference feedback module and output the scheduling correction amount.

9. A digital twin management method for petrochemical terminals according to claim 7, characterized in that, S32 includes: S321: The scheduling corrections involving the port area are sent to the port area scheduling system to adjust the berthing rhythm of subsequent vessels, including dynamically allocating berths, rearranging berthing time sequences, or updating mooring instructions. S322: The scheduling corrections for the relevant area are sent to the pump control system of the area to optimize the start-up and shutdown sequence and operating frequency of the pumps in the area. Specifically, this includes delaying or advancing the start-up time of a specific pump, adjusting the pump operation combination, or modifying the frequency setting value. S323: Send the scheduling correction amount involving the tank area to the tank area management system to replan the liquid level receiving window of the tank area, specifically including switching the target storage tank, adjusting the receiving order, or modifying the safe liquid level threshold.

10. A digital twin management system for petrochemical terminals, used to implement the digital twin management method for petrochemical terminals as described in any one of claims 1-9, characterized in that, Includes the following modules: The three-domain status acquisition and mapping module collects physical status parameters of the port area, pipeline area and tank area in the operation process, including berthing status, pump group operation curve, main pipeline differential pressure, tank area liquid level and flow rate response, and constructs the mapping relationship between the three-domain status data streams by combining operation control logs and spatiotemporal identifiers. Asynchronous temporal modeling module: Based on the state data stream, an asynchronous temporal modeling network is constructed to learn the asynchronous evolution relationship between the three domains and generate a digital twin mirror body, including a device state sub-mirror, a fluid behavior sub-mirror, and a scheduling action sub-mirror; State evolution simulation module: Receives the three-domain state evolution frame set output by the mirror body, and constructs the coupled evolution path of port area, pipeline area and tank area under different operating rhythms, and simulates the key dynamic processes in the initial unloading stage, steady-state transmission period and end buffer period; Predicted trajectory generation module: Based on the evolution path, it generates a set of predicted evolution trajectories of the three domain states over a future period of time through dynamic compression of time windows and sliding evolution mechanism; Offset identification and risk location module: compares the predicted trajectory with the baseline path to identify three-domain evolution offset labels and linkage mismatch risk points; Difference Feedback and Scheduling Correction Module: Compares the real-time status input with the predicted trajectory set. When a deviation or risk point is detected, a feedback mechanism is triggered, and the scheduling correction amount is output. Correction Execution and Mirror Update Module: Distributes the scheduling correction to the port area scheduling system, pipeline pump control system and tank area management system, and at the same time maps the correction results back to the twin mirror body to update the evolution path; Closed-loop control and self-evolution module: Based on the modified twin mirror image, continuous deduction, real-time decision-making and dynamic control are realized to construct a three-domain linkage predictive-control closed-loop system.

Images