Crude oil supply chain management method and system

Abstract

The invention relates to the field of supply chain management and optimization, in particular to a crude oil supply chain management method and system, and the method comprises the steps: constructing a supply chain dynamic model through multi-modal data fusion and causal analysis; real-time data and prediction data are generated based on a digital twinning technology, and cross-node association scenes and extreme climate conditions are simulated; a transportation path, an inventory allocation strategy and a delivery sequence are dynamically optimized through a multi-objective optimization algorithm and a flexible emergency adjustment module; an optimization scheme is automatically executed by using an intelligent contract system, feedback adjustment data and improved scheme data are generated based on real-time feedback, and closed-loop optimization is realized. The dynamic adaptive capacity of the crude oil supply chain in a complex environment is effectively improved, the cost, the inventory turnover rate and the delivery efficiency are optimized, and meanwhile the robustness and the emergency response capacity of the system are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of supply chain management and optimization, and particularly to a crude oil supply chain management method and system. Background Art

[0002] The crude oil supply chain covers multiple links from oilfield exploitation, transportation, refining to end-user delivery. Its efficiency and reliability have important impacts on energy supply security, economic development and environmental protection. However, the complexity, volatility of crude oil supply chain management and the unpredictability of emergencies make precise management a very challenging task.

[0003] Existing technologies (Chinese invention patent, publication number: CN118780733B, name: A Cross-border Supply Chain Data Management System and Method) usually optimize the supply chain process through data collection, preprocessing and analysis, and use algorithms such as ASGD to determine the order point and order quantity, aiming to minimize costs and maintain inventory levels. However, these technologies have the following deficiencies:

[0004] Existing methods fail to fully simulate the operating state of the supply chain under extreme conditions (such as climate change or sudden demand fluctuations), resulting in insufficient response capabilities to emergencies. The optimization objectives are mostly static indicators, such as total cost and inventory level, and fail to dynamically adjust the path planning and inventory allocation strategies according to real-time data. After the existing technologies execute the optimization plan, they cannot monitor and dynamically adjust in real time, and it is difficult to adapt to the changes in the supply chain under complex environments, resulting in abnormal problems not being solved in time. Summary of the Invention

[0005] In view of the above-mentioned many problems existing in the prior art, the present invention provides a crude oil supply chain management method and system. The present invention constructs a dynamic supply chain state model through multi-modal data fusion and causal analysis technology; combines digital twin technology to simulate real-time and future states, identify anomalies and generate optimization plans; realizes closed-loop optimization through multi-objective optimization algorithms and smart contracts, and finally improves the operating efficiency and robustness of the supply chain.

[0006] A crude oil supply chain management method includes the following steps:

[0007] Collect multi-source data of each node of the crude oil supply chain, including time series data, image data and text data, generate multi-modal fusion feature data by extracting and fusing all data; analyze the multi-modal fusion feature data by combining causal relationship modeling technology to generate causal association feature data;

[0008] Based on causal association feature data, an improved digital twin technology is used to construct a supply chain status model, generating digital twin real-time data and prediction data; by setting multi-level simulation conditions, including cross-node association scenarios and extreme climate conditions, extreme scenario simulation data is generated;

[0009] According to the extreme scenario simulation data, a multi-objective optimization algorithm is adopted, combined with a global dynamic adjustment mechanism, to generate supply chain optimization plan data; based on the causal association feature data, the transportation path and inventory allocation strategy are dynamically optimized, and at the same time, a flexible emergency adjustment module is introduced to quickly optimize the plan in case of emergencies;

[0010] The optimization plan data is input into the intelligent contract system to automatically execute transportation, inventory allocation, and order delivery operations; based on the real-time feedback data, feedback adjustment data for improving the supply chain operation is generated through anomaly monitoring and dynamic adjustment, and improvement plan data for optimizing the next-stage operation is generated to achieve the closed-loop optimization of the supply chain operation.

[0011] Preferably, the generation of the multi-modal fusion feature data includes the following steps: using a variational autoencoder to perform dimensionality reduction and denoising on time series data to extract time series feature data; performing hierarchical encoding on the semantic features of image data through a vision Transformer model to extract visual feature data; performing in-depth analysis on the semantic and context information of text data through a BERT model to extract text feature data; aligning and deeply fusing the above feature data to generate multi-modal fusion feature data.

[0012] Preferably, the generation of the causal association feature data includes the following steps: jointly modeling the time series features, spatial distribution features, and semantic features in the multi-modal fusion feature data through a structural causal model to analyze the causal relationships between supply chain nodes; deriving the probability distribution of the causal path based on a Bayesian network to generate the causal association weights between each node, and constructing causal association feature data according to the weights to reflect the key influencing factors of the dynamic changes in the supply chain.

[0013] Preferably, the construction of the digital twin status model includes the following steps: modeling the dynamic topological structure of the supply chain through a dynamic graph convolutional network to determine the association strength and change trend between nodes; updating the status of supply chain nodes using the real-time physical parameter data collected by sensors, including the oil flow rate of transportation nodes, the temperature and humidity of warehousing nodes, and the equipment operation parameters of refining nodes; constructing a supply chain status model based on the above data to generate digital twin real-time data and prediction data.

[0014] Preferably, the generation of the extreme scenario simulation data includes the following steps: setting the simulation conditions for the cross-node association scenario according to the causal association feature data, including the resource supply and demand fluctuations between nodes and the risk of transportation path interruption; setting the simulation parameters for the extreme climate conditions, including the impacts of high temperature, low temperature, and high humidity environments on transportation and warehousing; and generating the extreme scenario simulation data for the above scenarios through iterative optimization of the simulation parameters by the inverse reinforcement learning algorithm.

[0015] Preferably, the specific implementation of the multi-objective optimization algorithm includes the following steps: constructing the optimization objective function, including minimizing the total transportation path cost, maximizing the inventory turnover rate, and minimizing the delivery time; searching the solution space through the hybrid quantum particle swarm optimization algorithm and dynamically adjusting the weights of each optimization objective in each iteration; and globally verifying and locally refining the optimization results by combining the genetic algorithm to generate the supply chain optimization scheme data that meets the multi-objective optimization requirements.

[0016] Preferably, the implementation of the flexible emergency adjustment module includes the following steps: based on the real-time feedback data, using the reinforcement learning algorithm to identify emergencies, including abnormal situations such as transportation path delays, warehouse overloading, or order surges; dynamically generating emergency adjustment strategies, including reallocating inventory, adjusting transportation paths, and optimizing the delivery order; and integrating the emergency adjustment strategies into the supply chain optimization scheme data.

[0017] Preferably, the implementation of the smart contract system includes the following steps: building a distributed ledger based on blockchain technology to record the status data of the execution of supply chain tasks, including the completion time of transportation tasks, the inbound and outbound records of warehousing tasks, and the delivery information of orders; automatically verifying the task completion situation through smart contracts and triggering subsequent operations according to the verification results; and generating an immutable operation log in the blockchain system.

[0018] Preferably, the generation of the feedback adjustment data includes the following steps: analyzing the real-time feedback data through a time series anomaly detection model to identify potential anomalies in the operation of the supply chain, including transportation delays, insufficient inventory, or overloading; generating feedback adjustment data according to the anomaly detection results and dynamically updating the optimization scheme data to improve the operation efficiency of the supply chain.

[0019] A system for implementing the crude oil supply chain management method includes:

[0020] A data processing module, configured to collect multi-source data of each node of the crude oil supply chain, including time series data, image data, and text data, and perform feature extraction and fusion on the data to generate multi-modal fusion feature data;

[0021] A digital twin simulation module for constructing a supply chain status model based on multi-modal fusion feature data, generating digital twin real-time data and prediction data, and generating extreme scenario simulation data by setting multi-level simulation conditions;

[0022] An optimization module for generating supply chain optimization plan data by using a multi-objective optimization algorithm and combining a global dynamic adjustment mechanism according to the extreme scenario simulation data, and dynamically optimizing the transportation path and inventory allocation strategy based on causal association feature data;

[0023] An execution and feedback module for inputting the optimization plan data into an intelligent contract system, automatically executing transportation, inventory allocation and order delivery operations, and generating feedback adjustment data and improvement plan data based on real-time feedback data to achieve closed-loop optimization of supply chain operation.

[0024] Compared with the prior art, the advantages and beneficial effects of the present invention are as follows:

[0025] The present invention realizes accurate modeling of the dynamic association of the supply chain through multi-modal data fusion and causal analysis technology, and overcomes the defect of simple processing of node relationships in the prior art;

[0026] The present invention generates real-time data and prediction data through improved digital twin technology, realizes accurate simulation of the supply chain state under complex scenarios, and solves the problem of insufficient adaptability to extreme conditions in the prior art;

[0027] The present invention realizes dynamic optimization and real-time emergency adjustment of supply chain operation through a multi-objective optimization algorithm combined with a flexible emergency adjustment module, and effectively solves the problem of lack of flexibility in the optimization strategy in the prior art;

[0028] The present invention ensures the automation and real-time dynamic adjustment of supply chain management through an intelligent contract and a closed-loop optimization mechanism, and avoids the defect of imperfect feedback and adjustment mechanisms in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a flow schematic diagram of the method of the present invention;

[0030] Figure 2 It is a flow schematic diagram of the digital twin model construction in the present invention;

[0031] Figure 3 It is a flow schematic diagram of extreme scenario simulation and optimization in the present invention;

[0032] Figure 4 It is a flow schematic diagram of flexible emergency adjustment and closed-loop optimization in the present invention;

[0033] Figure 5 It is a structural block diagram of the system of the present invention. Detailed implementation manners

[0034] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, for the sake of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, obviously, one or more embodiments can also be implemented without these specific details. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present disclosure.

[0035] The terms used herein are merely for describing specific embodiments and are not intended to limit the present disclosure. The terms "including", "comprising", etc. used herein indicate the presence of the described features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0036] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those of ordinary skill in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0037] As Figure 1 shown, a crude oil supply chain management method includes the following steps:

[0038] Collect multi-source data of each node in the crude oil supply chain, including time series data, image data, and text data. Through feature extraction and fusion of all data, generate multi-modal fusion feature data; analyze the multi-modal fusion feature data by combining causal relationship modeling technology to generate causal association feature data;

[0039] Preferably, the generation of the multi-modal fusion feature data includes the following steps: Use a variational autoencoder to reduce the dimension and denoise the time series data, and extract time series feature data; perform hierarchical encoding on the semantic features of the image data through a vision Transformer model to extract visual feature data; perform in-depth analysis on the semantic and context information of the text data through a BERT model to extract text feature data; align and deeply fuse the above feature data to generate multi-modal fusion feature data.

[0040] Time series data is an important reflection of the operating status of each node in the supply chain, including dynamic data such as oilfield production, inventory changes, and transportation speed. Since the original time series data usually has high dimensions, noise interference, and redundant information, it needs to be processed by a variational autoencoder (VAE).

[0041] VAE learns the latent distribution of data through a probabilistic model, maps high-dimensional data to a low-dimensional latent space, and reduces noise by reconstructing the data.

[0042] For example, analyze the time series data of a transportation node, extract the main fluctuation patterns of daily transportation volume and the characteristics of abnormal transportation events, and provide support for subsequent route planning.

[0043] Image data includes oilfield monitoring footage, real-time images of the storage environment, and on-site photos of transportation routes. These image data provide intuitive information about the physical state of supply chain nodes, such as the liquid level of oil tanks, temperature and humidity conditions in the storage environment, and the status of transportation equipment.

[0044] The Vision Transformer model (ViT) uses the self-attention mechanism to encode image patches and extract multi-level semantic features. First, the input image is segmented into fixed-size image patches, and each image patch is flattened into a vector, and then an embedding vector is generated through a linear transformation. All image patch embeddings perform feature extraction through the self-attention mechanism.

[0045] For example: Process the real-time monitoring images of a storage node, extract the visual feature data of the temperature and humidity changes in the storage area, and provide a basis for the dynamic adjustment of the storage environment.

[0046] Text data mainly includes international oil price dynamic reports, logistics event records, and policy and regulation interpretations. These text data are important reference bases for supply chain planning and decision-making.

[0047] The BERT model is based on a bidirectional Transformer structure and can consider context information simultaneously, thereby capturing deep semantic associations in the text. For the input text sequence, BERT splits it into a sequence of tokens through a tokenizer and generates an embedding vector for each token. Subsequently, multi-layer Transformers are used for semantic feature extraction.

[0048] For example: Analyze the international oil price dynamic report, extract key semantic features (such as price change trends and market risk predictions), and provide support for the optimization of the crude oil procurement plan.

[0049] After the extraction of time series, visual, and text feature data is completed, it is necessary to generate multi-modal fusion feature data through feature alignment and deep fusion.

[0050] Use the shared attention mechanism to achieve cross-modal feature alignment, map time series, visual, and text features to the same feature space to capture the correlation between different modalities.

[0051] For example, multi-modal data of a certain supply chain node is fused to generate comprehensive features for identifying potential risk scenarios (such as delays and inventory shortages that may be caused by abnormal transportation equipment).

[0052] Preferably, the generation of the causal association feature data includes the following steps: jointly model the time series features, spatial distribution features, and semantic features in the multi-modal fusion feature data through a structural causal model to analyze the causal relationships between supply chain nodes; deduce the probability distribution of the causal paths based on a Bayesian network, generate the causal association weights between each node, and construct the causal association feature data according to the weights to reflect the key influencing factors of the dynamic changes in the supply chain.

[0053] In the present invention, the dynamic association between supply chain nodes is the key to optimization decisions. Therefore, it is necessary to reveal the causal relationships between nodes through causal association feature data. To this end, the present invention generates causal association feature data through the following steps:

[0054] Jointly model the time series features, spatial distribution features, and semantic features in the multi-modal fusion feature data through a structural causal model (SCM) to capture the causal relationships between nodes in the supply chain. The structural causal model is a graph model that can represent the causal relationships between variables with a directed acyclic graph (DAG). In the present invention, each node in the DAG represents a key variable in the supply chain (such as inventory level, transportation time, equipment status, etc.), and the edges represent causal relationships. The goal of joint modeling is to infer these causal edges and their strengths through multi-modal feature associations.

[0055] Based on the DAG structure, use a Bayesian network to deduce the probability distribution of the causal paths. The Bayesian network reflects the dependence relationships between variables through a joint probability distribution, and the conditional probability calculation formula is:

[0056]

[0057] where X represents the set of key variables in the supply chain, X i is the i-th variable, and Pa(X i ) represents the set of parent nodes of X i . By learning these conditional probabilities, deduce the strengths of each causal path, and then generate the causal association weights between nodes.

[0058] The causal association weights are further quantified into a feature matrix to construct causal association feature data. Each element of the feature matrix represents the degree of causal influence between two supply chain nodes. For example, how the change in the inventory level of the warehousing node is affected by the delay of the transportation node. This feature data is used to reflect the key influencing factors of the dynamic changes in the supply chain and provide data support for subsequent optimization decisions.

[0059] Through the generation of causal association feature data, the causal relationships between various nodes in the crude oil supply chain can be accurately identified, the association strength between key variables can be quantified, and potential systemic risks can be revealed. This data can provide an important basis for optimizing transportation routes, dynamic inventory management, and emergency scheduling, significantly improving the scientific nature and robustness of supply chain management.

[0060] Embodiment: In a certain crude oil transportation link, the terminal inventory is tight due to increased seasonal demand. Through the causal association feature data generated by the present invention, the following causal path is analyzed: The delay of the transportation node (variable A) affects the replenishment cycle of the warehousing node (variable B), which ultimately leads to abnormal fluctuations in the inventory level of the terminal node (variable C). Using the Bayesian network to calculate the weight of the causal path, it is found that the influence weight of the delay time on the replenishment cycle is 0.8, and the indirect influence weight on the terminal inventory level is 0.6. This indicates that transportation delays are the main factors causing inventory tightness in the supply chain. Based on this analysis result, the transportation route and allocation plan can be preferentially adjusted during the optimization process, thereby effectively alleviating the inventory pressure and improving the emergency response ability of the supply chain.

[0061] As Figure 2 shown, based on the causal association feature data, an improved digital twin technology is used to construct a supply chain state model, generating digital twin real-time data and prediction data; by setting multi-level simulation conditions, including cross-node association scenarios and extreme climate conditions, extreme scenario simulation data is generated;

[0062] The present invention constructs a supply chain state model through causal association feature data, further generating digital twin real-time data and prediction data to simulate and optimize the dynamic operation state of the crude oil supply chain. This process mainly includes the following two core parts:

[0063] (1) The digital twin technology realizes the dynamic mapping of the system operation state through the two-way data flow between the virtual model and the actual supply chain. Based on the causal association feature data, first reconstruct the dynamic topological structure of the supply chain in the virtual space, including the associations between transportation nodes, warehousing nodes, and terminal nodes.

[0064] Data Input: Input the causal association feature data into the Dynamic Graph Convolutional Network (DGCN) to capture the real-time association strength and changing trends among nodes. Generate the dynamic state information of the supply chain through this model, including inventory levels, transportation flows, and equipment operating status, etc.

[0065] Data Output: Output the real-time data of the digital twin for real-time monitoring of the current state of the supply chain. Meanwhile, generate the future states of the key nodes of the supply chain through a time series prediction model (such as LSTM) to form prediction data. These prediction data can help identify potential risks and optimize the operation strategies.

[0066] (2) Based on the supply chain state model, further generate extreme scenario simulation data by setting multi-level simulation conditions. The simulation conditions include but are not limited to: simulating the impact of resource allocation between transportation nodes and warehousing nodes on the overall operation of the system, such as how transportation disruptions lead to insufficient inventory; simulating the interference of extreme weather such as high temperature and heavy rain on transportation efficiency and warehousing environment.

[0067] Use the Inverse Reinforcement Learning (IRL) algorithm to optimize the simulation conditions and deduce the optimal operation path of the supply chain under different scenarios. IRL evaluates the effects of actions taken by each node in a specific state by calculating the reward function. Generate extreme scenario simulation data for different scenarios by repeatedly adjusting the simulation parameters. These data can be used to analyze the operation bottlenecks of the system in special situations and propose improvement strategies.

[0068] Through the digital twin technology driven by causal association feature data and multi-level simulation conditions, the present invention can comprehensively reflect the dynamic operation state of the crude oil supply chain, accurately predict future changes, and simulate possible extreme scenarios. This method can not only help identify the weak links in the supply chain, but also provide a scientific basis for formulating optimization plans and emergency strategies, thereby significantly enhancing the robustness and adaptability of the supply chain.

[0069] Embodiment: In a certain crude oil supply chain, due to sudden heavy rain, the transportation route was interrupted and the terminal inventory was about to run out. Through the digital twin technology, the system real-time generates the dynamic states of transportation nodes and warehousing nodes, and predicts that the inventory of the warehousing node will be exhausted within two hours. To verify the feasibility of the emergency plan, set cross-node association scenarios and extreme climate conditions for simulation. The simulation results show that if the transportation route is re-planned and inventory is preferentially allocated to the terminal nodes with the most urgent demand, the problem of insufficient inventory can be effectively alleviated. Based on this, the system generates an optimized supply chain scheduling plan to ensure the continuous and efficient operation of the supply chain.

[0070] Preferably, the construction of the digital twin state model includes the following steps: modeling the dynamic topology of the supply chain through a dynamic graph convolutional network to determine the association strength and change trend between nodes; updating the state of supply chain nodes using real-time physical parameter data collected by sensors, including the oil flow rate of transportation nodes, the temperature and humidity of warehousing nodes, and the equipment operation parameters of refining nodes; constructing a supply chain state model based on the above data to generate digital twin real-time data and prediction data.

[0071] In the present invention, reflecting the dynamic changes of the supply chain in real time and predicting future states are the keys to optimizing decisions. To this end, the present invention constructs a digital twin state model through the following steps to generate digital twin real-time data and prediction data:

[0072] Use a dynamic graph convolutional network (Dynamic Graph Convolutional Network, DGCN) to model the dynamic topology of the supply chain, aiming to capture the association strength and change trend between nodes. Supply chain nodes include transportation nodes, warehousing nodes, and refining nodes. The relationships between nodes are represented by a directed graph, and the edge weights in the graph reflect the association strength between nodes. DGCN updates node features through the following formula:

[0073] H (l+1) =σ(D -1 / 2 AD -1 / 2 H (l) W (l) )

[0074] Where, H (l) represents the node feature matrix of the l-th layer, A is the adjacency matrix, D is the degree matrix, W (l) is the weight matrix of the l-th layer, and σ is the activation function. Through multi-layer convolution operations, the node features are dynamically updated and the dynamic topology of the supply chain is generated.

[0075] Collect real-time physical parameter data through sensors to dynamically update the state information of supply chain nodes. These data reflect the current operating state of each node, including the oil flow rate of transportation nodes, the temperature and humidity of warehousing nodes, and the equipment operation parameters of refining nodes. Each node collects data through Internet of Things (IoT) sensors to form a feature vector; for example, the transportation node feature vector includes parameters such as oil flow rate and vehicle location. Through data cleaning and standardization, the data quality of the input model is ensured. Input the real-time physical parameter data into the dynamic topology model to update the feature weights of the nodes to reflect the current operating state of the supply chain.

[0076] Based on the above dynamic topology and real-time physical parameters, a supply chain state model is constructed to generate digital twin real-time data and prediction data. The digital twin real-time data is a virtual reflection of the current supply chain state, including inventory levels, flows between transportation nodes, equipment operating status, etc. Through a time series prediction model (such as LSTM), the node states in the future for a period of time are predicted based on the real-time data. For example, predicting the future flow peak of transportation nodes, the inventory shortage risk of warehousing nodes, etc.

[0077] By constructing the digital twin state model, the dynamic operating state of the supply chain can be accurately monitored and short-term predictions can be made, providing data support for optimization scheduling, resource allocation, and emergency adjustment. Especially in the crude oil supply chain, this model can effectively identify potential risks such as transportation delays, warehousing overloading, or refinery equipment failures, providing optimization suggestions for decision-makers, thereby improving the efficiency and stability of the supply chain.

[0078] Example: In a certain crude oil supply chain, due to seasonal demand fluctuations, the inventory pressure of the warehousing node increases, and the transportation node faces transportation bottlenecks. Through the digital twin state model, the real-time oil flow rate of the transportation node and the temperature and humidity data of the warehousing node are collected to generate a dynamic topology structure, and the correlation strength between each node is analyzed. The model prediction results show that the inventory of the warehousing node will exceed the safety threshold within the next 24 hours, and there is a risk of delay in a certain section of the transportation path. Based on the data output by the model, the system generates an optimized scheduling plan, adjusts the priorities of the transportation nodes, and optimizes the inventory allocation, thus effectively alleviating the inventory pressure and ensuring the normal operation of the supply chain.

[0079] Preferably, as Figure 3 shown, the generation of the extreme scenario simulation data includes the following steps: setting the simulation conditions for the cross-node association scenario according to the causal association feature data, including resource supply and demand fluctuations between nodes and the risk of transportation path interruption; setting the simulation parameters for extreme climate conditions, including the impact of high temperature, low temperature, and high humidity environments on transportation and warehousing; and iteratively optimizing the simulation parameters through the inverse reinforcement learning algorithm to generate extreme scenario simulation data for the above scenarios.

[0080] In the present invention, the generation of extreme scenario simulation data is of great significance for evaluating the robustness of the supply chain and formulating emergency strategies. The present invention generates simulation data for cross-node association scenarios and extreme climate conditions through causal association feature data in combination with the inverse reinforcement learning algorithm (Inverse Reinforcement Learning, IRL), and the specific steps are as follows:

[0081] (1) Set the simulation conditions for cross - node association scenarios. The resource supply - demand relationship between different nodes in the supply chain has dynamic association characteristics. To simulate cross - node association scenarios, first, according to the causal association feature data, identify the dependency relationships and resource flow paths between key nodes.

[0082] Set the simulation conditions, including the amplitude and frequency of resource supply - demand fluctuations between nodes. For example, simulate the scenario where the inventory of the warehouse node is insufficient due to the delay of the transportation node, and quantify the chain reaction of transportation interruption on the delivery of the terminal node. Through causal path analysis, set the supply - demand adjustment coefficients of different nodes, and define the influence propagation path and time delay between nodes.

[0083] (2) Set the simulation parameters for extreme climate conditions. Extreme climate has a significant impact on the operation of the crude oil supply chain. For example, high temperature may cause transportation equipment failures, and high humidity may affect the stability of the storage environment.

[0084] Set the simulation parameters, including the threshold ranges of temperature (high temperature, low temperature) and humidity (high humidity), to constrain the states of transportation and storage nodes. By establishing a mapping function between climate parameters and node operation efficiency, quantify the direct and indirect impacts of climate change on nodes. For example, the impact of high temperature on the transportation path may be reflected by the failure rate of transportation equipment.

[0085] (3) Optimize the simulation parameters through the inverse reinforcement learning algorithm. Inverse reinforcement learning derives the optimal scenario simulation strategy by learning historical operation data. Its core lies in observing the operation results of the supply chain, reverse - deriving the reward function of the system in different states, and using this function to optimize the scenario simulation parameters. Algorithm process:

[0086] Input the historical operation data and simulation conditions of the supply chain;

[0087] Use inverse reinforcement learning to derive the reward function R(s,a), where s represents the current state of the supply chain and a represents the operation actions of nodes;

[0088] By maximizing the cumulative reward Optimize the simulation parameters to generate extreme scenario simulation data that conforms to the actual situation.

[0089] The extreme scenario simulation data generated through the above steps can help supply chain managers identify potential risks in the system under different extreme conditions, providing a scientific basis for optimizing resource allocation, improving operation strategies, and formulating emergency plans. Such simulation data not only improves the robustness of the supply chain but also significantly enhances its adaptability in complex environments.

[0090] Embodiment: In a certain crude oil supply chain, due to the risk of interruption in the transportation path, the terminal node may not be able to obtain crude oil replenishment in a timely manner. Through the present invention, the simulation conditions for cross-node associated scenarios are set to simulate the resource supply and demand fluctuations between the transportation node and the storage node. At the same time, the influence parameter of high temperature on the failure rate of transportation equipment is set to construct an operation model under extreme climate conditions. The inverse reinforcement learning algorithm is used to optimize the simulation parameters to generate simulation data. The results show that the influence coefficient of transportation interruption on the delivery of the terminal node is 0.85, and the safety inventory of the storage node will be exhausted within 3 hours. Based on this simulation data, the system recommends reallocating inventory and adjusting the transportation path, thus effectively avoiding the risk of supply chain interruption.

[0091] According to the extreme scenario simulation data, a multi-objective optimization algorithm is adopted, combined with a global dynamic adjustment mechanism, to generate supply chain optimization plan data. Based on the causal association feature data, the transportation path and inventory allocation strategy are dynamically optimized. At the same time, a flexible emergency adjustment module is introduced to quickly optimize the plan in case of emergencies.

[0092] In the present invention, the goal of supply chain optimization is to ensure the efficient operation of the transportation, inventory, and production links under various scenarios. Especially under extreme conditions, dynamic adjustment is required to ensure robustness and flexibility. The present invention generates an optimization plan based on the extreme scenario simulation data, combined with a multi-objective optimization algorithm and a global dynamic adjustment mechanism. At the same time, a flexible emergency adjustment module is introduced to quickly optimize the supply chain operation strategy in case of emergencies.

[0093] The multi-objective optimization algorithm is used to find the best balance point among multiple conflicting objectives, such as minimizing transportation costs, maximizing inventory turnover rate, minimizing delivery time, etc. A method combining hybrid particle swarm optimization (PSO) and genetic algorithm (GA) is adopted. Particle swarm optimization is used for global search of the optimal solution, and the genetic algorithm improves the diversity of solutions and local optimization ability through crossover and mutation operations. According to the extreme scenario simulation data, supply chain optimization plan data is dynamically generated, including the transportation path, inventory allocation ratio, and delivery sequence of each node.

[0094] The causal association feature data provides the causal relationship and key influencing factors between supply chain nodes, providing data support for dynamic optimization. Through real-time adjustment of the causal path, the transportation path and inventory allocation strategy are dynamically optimized. Combining the extreme scenario simulation data, the delay risk of each transportation path is dynamically evaluated, and the optimal path is re-planned to ensure the efficiency and stability of logistics. According to the resource supply and demand relationship between nodes, the inventory allocation strategy is adjusted to give priority to meeting the terminal nodes with urgent needs, while ensuring the overall inventory turnover rate.

[0095] The flexible emergency adjustment module rapidly generates emergency adjustment strategies through reinforcement learning algorithms for optimization in emergency situations. For example, in the event of transportation disruptions caused by sudden weather or equipment failures, the flexible emergency adjustment module can, based on real-time feedback, reallocate inventory and adjust transportation routes. The reinforcement learning algorithm continuously learns and optimizes the adjustment strategy according to emergency events and feedback data, ensuring the robustness and flexibility of the supply chain in emergency situations. By directly integrating the emergency adjustment strategy into the optimization plan, the impact of emergency situations on the operation of the supply chain is reduced, and the supply chain recovery ability is enhanced.

[0096] Through the above steps, the present invention can generate efficient and flexible supply chain optimization plans in extreme scenarios, effectively reducing transportation costs, increasing inventory turnover rates, and shortening delivery times. At the same time, the flexible emergency adjustment module significantly improves the supply chain's ability to respond to emergency situations, ensuring the stable operation of the supply chain in various complex environments.

[0097] Preferably, the specific implementation of the multi-objective optimization algorithm includes the following steps: constructing an optimization objective function, including minimizing the total cost of the transportation route, maximizing the inventory turnover rate, and minimizing the delivery time; searching the solution space through a hybrid quantum particle swarm optimization algorithm, dynamically adjusting the weights of each optimization objective in each iteration; and globally verifying and locally refining the optimization results in combination with a genetic algorithm to generate supply chain optimization plan data that meets the requirements of multi-objective optimization.

[0098] In the present invention, the optimization plan needs to seek the best balance among multiple conflicting objectives. The present invention adopts a multi-objective optimization algorithm, combining a hybrid quantum particle swarm optimization algorithm and a genetic algorithm, to generate efficient supply chain optimization plan data. The specific implementation steps are as follows:

[0099] The design of the optimization objective function is the core of multi-objective optimization and is used to quantify the overall operating efficiency of the supply chain. The optimization objectives of the present invention include the following three items:

[0100] Minimization of the total cost of the transportation route: Define the total cost of the transportation route as C transport , including fuel costs, tolls, and equipment depreciation costs.

[0101] Maximization of the inventory turnover rate: Maximize the turnover rate to improve inventory management efficiency, represented by the turnover rate index R inventory .

[0102] Minimization of the delivery time: The delivery time T delivery includes the time of the entire transportation process from the oil field to the end user.

[0103] The optimization objective function is defined as:

[0104] MinimizeZ=w1C transport-w2R inventory +w3T delivery

[0105] Among them, Z is the total objective function, and w1, w2, and w3 are dynamically adjusted weights, representing the importance of different optimization objectives.

[0106] The Hybrid Quantum Particle Swarm Optimization (HQPSO) is used to search the solution space to obtain the global optimal solution.

[0107] HQPSO expands the diversity of the search space by introducing quantum superposition states, and at the same time uses the iterative search mechanism of traditional Particle Swarm Optimization (PSO) to ensure convergence. The position update formula of the particle is:

[0108]

[0109] Among them, represents the position of the i-th particle at the t-th iteration, p i is the individual optimal position of the particle, g is the global optimal position, and α and β are learning factors.

[0110] In each round of iteration, the objective weights are dynamically adjusted according to the scenario requirements to ensure the balance of each objective in different optimization stages. For example, in the case of tight transportation, increase the weight of the transportation cost objective to prioritize the optimization of the path planning.

[0111] To improve the diversity of the solution and the local accuracy of the optimization result, the genetic algorithm (GA) is combined to globally verify and locally refine the optimization result of HQPSO.

[0112] The core operations of the genetic algorithm include selection, crossover, and mutation. The optimal solution is screened through the fitness function and new solutions are generated by crossover. At the same time, the search space is expanded through random mutation operations to avoid falling into local optima.

[0113] GA conducts a global fitness evaluation on the candidate solutions generated by HQPSO, excluding the solutions that do not meet the actual supply chain constraint conditions; then locally optimizes the retained solutions to improve the accuracy and feasibility of the final plan.

[0114] Through the above steps, the supply chain optimization plan generated by the multi-objective optimization algorithm can effectively reduce transportation costs, improve inventory turnover rates, and shorten delivery times. At the same time, the combination of the hybrid quantum particle swarm optimization algorithm and the genetic algorithm not only ensures the acquisition of the global optimal solution but also further improves the execution feasibility of the plan through local refinement. In addition, the dynamic weight adjustment mechanism makes the plan highly flexible in different scenarios and can adapt to complex and changing supply chain environments.

[0115] Embodiment. In a certain crude oil supply chain, due to sudden weather, the transportation route was interrupted, and at the same time, the demand at the terminal node surged. Based on the hybrid quantum particle swarm optimization algorithm, the system generated multiple optimization plans, one of which included diverting to an alternate route and increasing the inventory replenishment frequency at the storage nodes. Through the global verification of the genetic algorithm, the route planning was further optimized. The finally generated optimization plan ensured the delivery timeliness at the terminal node while maintaining the lowest transportation cost. The optimization results showed that this plan reduced the transportation cost by 12%, increased the inventory turnover rate by 15%, and shortened the delivery time by 20%.

[0116] Preferably, as Figure 4 shown, the implementation of the flexible emergency adjustment module includes the following steps: Based on real-time feedback data, use the reinforcement learning algorithm to identify emergencies, including abnormal situations such as transportation route delays, warehouse overloading, or order surges; Dynamically generate emergency adjustment strategies, including reallocating inventory, adjusting transportation routes, and optimizing delivery sequences; Incorporate the emergency adjustment strategies into the optimization plan data of the supply chain.

[0117] In the present invention, emergencies such as transportation route delays, warehouse overloading, or order surges have a significant impact on the normal operation of the supply chain. To cope with such emergencies, the present invention designs a flexible emergency adjustment module that dynamically generates emergency adjustment strategies through the reinforcement learning algorithm and incorporates them into the optimization plan, specifically including the following steps:

[0118] The flexible emergency adjustment module first analyzes the real-time feedback data to identify emergencies. Data input: The real-time feedback data includes the delay time of the transportation route, the inventory level of the storage node, order demand fluctuations, etc. Through a time series anomaly detection model (such as an anomaly detection network based on LSTM), abnormal situations such as transportation delays, warehouse overloading, or order surges are identified from the feedback data. Based on the state representation in reinforcement learning, define the current state of the system as s t , and the emergency is input into the reinforcement learning algorithm as part of the state change for processing.

[0119] For emergencies, the flexible emergency adjustment module uses the reinforcement learning algorithm (such as Deep Q-Learning, DQL) to dynamically generate emergency adjustment strategies.

[0120] Reinforcement learning optimizes the decision-making process through a reward mechanism, and its goal is to maximize the cumulative reward. Define the state s t , the action a t and the reward function R(s t ,a t ), and the optimization goal is:

[0121]

[0122] Among them, γ is the discount factor, indicating the importance of future rewards.

[0123] Emergency strategy generation: Combining real-time feedback data, the DQL algorithm dynamically selects the optimal action a according to the state s in each iteration t and generates an emergency adjustment strategy including the following: t Inventory reallocation: Adjust the inventory flow according to node demands, and give priority to meeting the demands of nodes with urgent needs.

[0124] Transport path adjustment: Reduce the delay risk by using alternative paths or temporarily dispatching additional transport resources.

[0125] Delivery order optimization: Reorder the terminal delivery tasks to ensure the priority satisfaction of the needs of key customers.

[0126] The generated emergency adjustment strategy needs to be further combined with the data of the optimization plan of the supply chain to ensure the coherence of global optimization. Taking the emergency adjustment strategy as an additional constraint condition, recalculate the optimization objective function and update the data of the supply chain optimization plan. By simulating the operation of the supply chain under the influence of emergencies, verify the feasibility and optimization effect of the emergency adjustment strategy, and ensure that the adjusted optimization plan has high robustness and stability under extreme conditions.

[0127] The flexible emergency adjustment module realizes the rapid identification and dynamic response to emergencies through the reinforcement learning algorithm. The generated emergency adjustment strategy can effectively reduce the negative impact of emergencies on the operation of the supply chain. Combining real-time feedback data and global optimization strategies, this module significantly improves the flexibility and adaptability of the supply chain, and ensures the efficient operation of the crude oil supply chain under extreme conditions.

[0128] In an embodiment, in a certain crude oil transportation link, due to sudden bad weather, the main transportation path is interrupted, and at the same time, the order demand at the terminal node surges. The flexible emergency adjustment module first identifies the abnormal situations of transportation delays and order surges through feedback data, and then uses the reinforcement learning algorithm to generate an emergency adjustment strategy. The strategy includes reallocating inventory, preferentially transferring reserve inventory to the terminal node; enabling alternative transportation paths and dispatching additional transport resources; and readjusting the delivery order to ensure that the customers with the most urgent needs are given priority to obtain crude oil supply. Finally, integrating the emergency adjustment strategy into the optimization plan, through simulation evaluation, it is found that this strategy effectively reduces the risk brought by transportation delays and ensures the on-time delivery of orders. The entire emergency response time is shortened by 30%, and the balance of inventory consumption is increased by 20%.

[0129]

[0130] ​Input the optimized solution data into the smart contract system to automatically execute transportation, inventory allocation, and order delivery operations; based on real-time feedback data, generate feedback adjustment data for improving the supply chain operation through anomaly monitoring and dynamic adjustment, and generate improved solution data for optimizing the operations in the next stage, realizing the closed-loop optimization of the supply chain operation.

[0131] In the present invention, in order to achieve efficient and automated transportation, inventory allocation, and order delivery, the present invention executes the optimized solution data through the smart contract system, and at the same time, based on real-time feedback data, dynamically adjusts the optimized solution to achieve closed-loop optimization. The implementation process includes the following key steps:

[0132] (1) Input of optimized solution data and execution of smart contract. The optimized solution data includes transportation route planning, inventory allocation strategy, and delivery task priority.

[0133] Based on blockchain technology, deploy smart contracts on the distributed ledger. Each smart contract corresponds to a specific operation task in the supply chain, such as path selection between transportation nodes, inventory transfer of warehousing nodes, and order delivery of terminal nodes.

[0134] Input the optimized solution data into the smart contract system to trigger the corresponding tasks. The smart contract automatically executes after receiving the data, including scheduling transportation vehicles, adjusting inventory allocation, and completing orders according to the set delivery sequence. The execution process of the smart contract ensures the immutability and efficiency of the operations.

[0135] (2) Anomaly monitoring of real-time feedback data. During the task execution process, collect feedback data in real time through Internet of Things (IoT) devices, including transportation status (such as transportation time and vehicle location), inventory level (such as safety inventory and inbound / outbound records), and order completion rate, etc.

[0136] Monitor the feedback data through an anomaly detection model based on time series. For example, use LSTM to detect transportation route delays or inventory anomalies at warehousing nodes.

[0137] Classify the feedback data into normal data and abnormal data, and the abnormal data is used to trigger the dynamic adjustment strategy.

[0138] (3) Dynamic adjustment and generation of improvement data. Based on the feedback data, dynamically adjust the current supply chain operation plan, and generate feedback adjustment data and improved solution data.

[0139] By analyzing the abnormal feedback data, the system dynamically optimizes the transportation route, reallocates inventory, or adjusts the delivery sequence. For example, if a transportation route is delayed, dynamically adjust the route planning to ensure the delivery timeliness.

[0140] Combining the current adjustment results and historical operation data, the system uses reinforcement learning algorithms to generate optimization scheme data for the next stage, further improving the overall operation efficiency of the supply chain.

[0141] Through the combination of the intelligent contract system and real-time feedback data, the supply chain can automatically and efficiently execute the optimization scheme, and at the same time quickly respond and adjust the operation strategy in case of emergencies. The realization of closed-loop optimization has greatly improved the robustness and flexibility of the supply chain, reduced the operation cost, and improved the resource utilization efficiency.

[0142] Preferably, the implementation of the intelligent contract system includes the following steps: constructing a distributed ledger based on blockchain technology to record the status data of supply chain task execution, including the completion time of transportation tasks, the inbound and outbound records of warehousing tasks, and the delivery information of orders; automatically verifying the task completion situation through intelligent contracts and triggering subsequent operations according to the verification results; generating an immutable operation log in the blockchain system.

[0143] In the present invention, the execution status of supply chain tasks needs to be accurately recorded and verified in real time to ensure the transparency and traceability of operations. The present invention constructs an intelligent contract system based on blockchain technology to form a distributed ledger, and realizes the efficiency and transparency of supply chain management by automatically verifying the task completion situation and triggering subsequent operations.

[0144] The key data during the execution of supply chain tasks is recorded on the distributed ledger through blockchain technology to ensure the immutability and transparency of the data. The status data recorded in the distributed ledger includes the completion time of transportation tasks, the inbound and outbound records of warehousing nodes, and the delivery information of orders. Each task record is stored in the form of a block, and the block contains information such as the task unique identifier, operation timestamp, execution node, and verification status. The blocks are chained together through a hash function to ensure the integrity and anti-tampering of the data.

[0145] The intelligent contract automatically verifies the completion situation of supply chain tasks according to the pre-set rules and triggers subsequent operations according to the verification results. The task completion status is uploaded to the blockchain by relevant sensors or system nodes, and the intelligent contract checks the task completion conditions according to the verification rules. For example, the verification conditions for transportation tasks include whether the goods arrival time is within the set time range and whether the operation status of the transportation equipment is normal. After the task completion verification passes, the intelligent contract automatically triggers the next stage of operations. For example, the warehousing node adjusts the inbound and outbound operations according to the inventory level after the transportation task is completed, and the terminal node triggers the invoice generation and payment request after receiving the order.

[0146] The status data and verification results of all supply chain operations automatically generate operation logs through the blockchain system. The operation logs include detailed records of task execution, such as transportation routes, inventory changes, order completion times, etc. These logs have timestamps and unique identifiers to ensure the accuracy and reliability of the data. Since the logs are stored on the blockchain and are immutable, they can provide a trustworthy basis for subsequent operation analysis and risk tracing.

[0147] The smart contract system realizes the efficient and transparent management of the crude oil supply chain by automatically executing task verification, triggering subsequent operations, and recording operation logs. The application of the distributed ledger ensures the integrity and security of data, preventing data fraud and tampering; smart contracts significantly improve the automation level of operations, reduce manual intervention, and improve task execution efficiency. Overall, the present invention enhances the reliability, traceability, and operation efficiency of the crude oil supply chain, providing strong technical support for the intelligent management of the supply chain.

[0148] Embodiment, in a certain crude oil supply chain, the transportation node is responsible for transporting crude oil from the oilfield to the refinery, and the warehousing node needs to record the inbound and outbound quantities of crude oil. Through the smart contract system constructed by the present invention, after each transportation task is completed, the sensors of the transportation node automatically upload the goods arrival time and transportation status data to the blockchain. The smart contract automatically verifies the task completion situation according to preset conditions (such as whether it arrives on time, whether the transportation equipment status is normal). After the verification passes, the system automatically triggers the inbound operation of the warehousing node, and at the same time records the task completion time, transportation route, and equipment status, and generates corresponding operation logs. Finally, these immutable log data provide a reliable basis for the evaluation of the supply chain operation efficiency and risk management, and at the same time improve the operation efficiency of the entire supply chain by reducing human intervention and errors.

[0149] Preferably, the generation of the feedback adjustment data includes the following steps: analyzing the real-time feedback data through a time series anomaly detection model to identify potential anomalies in the supply chain operation, including transportation delays, insufficient inventory, or overloading; generating feedback adjustment data according to the anomaly detection results, and dynamically updating and optimizing the scheme data to improve the supply chain operation efficiency.

[0150] In the present invention, real-time identification and response to operation anomalies are crucial for improving the supply chain efficiency. The present invention timely identifies anomalies and dynamically optimizes the operation plan by generating feedback adjustment data. Specifically, it includes the following steps:

[0151] (1) The generation of feedback adjustment data first requires analyzing real-time feedback data through a time-series anomaly detection model to identify potential anomalies in the supply chain operation. The real-time feedback data includes the arrival time of transportation routes, the inventory levels of warehousing nodes, the delivery status of orders, etc. These data have time-series characteristics. A time-series anomaly detection model based on Long Short-Term Memory (LSTM) is adopted to identify anomaly points by predicting the normal change range of the time series. The model learns the time-series pattern of historical operation data, calculates the deviation between the current state and the predicted value, and if the deviation exceeds the set threshold, it is determined as an anomaly. The anomaly detection results include the following categories: transportation delays, where the arrival time of the route exceeds the expected range; insufficient or overloaded inventory, where the inventory level of the warehousing node is lower than the safety threshold or exceeds the maximum capacity.

[0152] (3) According to the anomaly detection results, feedback adjustment data is generated for dynamically updating the optimization plan. Based on the output of the anomaly detection model, the anomalies are classified, such as the nodes with transportation delays, the warehousing locations with inventory anomalies, etc. Specific adjustment suggestions are generated based on the anomaly type. For example, if a transportation delay is detected, the feedback adjustment data may include suggestions to change the route or increase transportation resources; if insufficient inventory is detected, an adjusted inventory allocation plan is generated. By interacting with the optimization algorithm, the feedback adjustment data is used as an additional constraint condition to dynamically update the optimization plan to solve the anomaly problem.

[0153] (3) After the feedback adjustment data is generated, it is integrated into the supply chain optimization process to dynamically update the optimization plan. The feedback adjustment data is input into a multi-objective optimization algorithm, combined with the current operation data, to recalculate the optimal transportation route, inventory allocation ratio, and delivery priority. The optimization results include the adjusted node routes, inventory reallocation strategy, and the order sequence for priority delivery, ensuring the efficient operation of the supply chain under abnormal conditions.

[0154] By generating feedback adjustment data, the present invention can identify potential anomalies in the supply chain operation in real time and make dynamic adjustments, effectively avoiding the impact of problems such as transportation delays and inventory imbalances on the supply chain operation. The update of the dynamic optimization plan significantly improves the flexibility and adaptability of the supply chain, while reducing the time and cost of anomaly handling. Overall, this method improves the operation efficiency and service level of the supply chain, providing an important guarantee for the stability of the supply chain.

[0155] Example: In a certain crude oil supply chain, a transportation node was delayed due to sudden weather, and at the same time, the inventory at the warehousing node was approaching the lower limit. The time series anomaly detection model detected the delay risk by analyzing the arrival time data of the transportation route and simultaneously identified the anomaly of insufficient inventory at the warehousing node. The system generated feedback adjustment data and proposed two optimization suggestions: First, adjust the transportation route and enable an alternative route to bypass the weather-affected area; Second, reallocate the inventory at the warehousing node and transfer the surplus inventory from other warehousing nodes to the risk node. Based on the feedback adjustment data, the optimization algorithm recalculated the route and inventory allocation strategy to ensure that the transportation arrived on time and avoid inventory shortages. Finally, the transportation delay time was shortened by 25%, the inventory risk was completely resolved, and the operating efficiency of the supply chain was increased by 15%.

[0156] As Figure 5 shown, a system for implementing the above crude oil supply chain management method includes:

[0157] A data processing module, configured to collect multi-source data of each node in the crude oil supply chain, including time series data, image data, and text data, and perform feature extraction and fusion on the data to generate multi-modal fusion feature data; collect multi-source data (time series, image, text) of each node in the crude oil supply chain, and generate multi-modal fusion feature data through multi-modal feature extraction and fusion, accurately reflecting the dynamic state of each node in the supply chain.

[0158] A digital twin simulation module, configured to construct a supply chain state model based on the multi-modal fusion feature data, generate digital twin real-time data and prediction data, and generate extreme scenario simulation data by setting multi-level simulation conditions; based on the multi-modal fusion feature data, construct a supply chain state model, generate digital twin real-time data and prediction data. By setting multi-level simulation conditions (such as cross-node association and extreme climate scenarios), generate extreme scenario simulation data to accurately simulate the operation of the supply chain and potential risks.

[0159] An optimization module, configured to generate supply chain optimization plan data by using a multi-objective optimization algorithm in combination with a global dynamic adjustment mechanism according to the extreme scenario simulation data, and dynamically optimize the transportation route and inventory allocation strategy based on the causal association feature data; utilize the extreme scenario simulation data to generate an optimization plan through a multi-objective optimization algorithm, dynamically adjust the transportation route and inventory allocation strategy to achieve global optimality. The causal association feature data further supports dynamic optimization, improving the accuracy and adaptability of the plan.

[0160] An execution and feedback module is used to input optimization solution data into the intelligent contract system, automatically execute transportation, inventory allocation, and order delivery operations, and generate feedback adjustment data and improvement solution data based on real-time feedback data to achieve closed-loop optimization of the supply chain operation. The optimization solution is automatically executed through the intelligent contract system to achieve efficient management of transportation, inventory allocation, and order delivery. At the same time, based on real-time feedback data, operation anomalies are identified, and feedback adjustment data and improvement solutions are generated to achieve closed-loop optimization of the supply chain.

[0161] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects.

[0162] The above are only the embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A method for crude oil supply chain management, characterized in that, It includes the following steps: Collect multi-source data of each node in the crude oil supply chain, including time series data, image data, and text data. Through feature extraction and fusion of all data, generate multi-modal fusion feature data; Analyze the multi-modal fusion feature data by combining causal relationship modeling techniques to generate causal association feature data; Based on the causal association feature data, use improved digital twin technology to construct a supply chain state model, generate digital twin real-time data and prediction data; by setting multi-level simulation conditions, including cross-node association scenarios and extreme climate conditions, generate extreme scenario simulation data; According to the extreme scenario simulation data, adopt a multi-objective optimization algorithm, combined with a global dynamic adjustment mechanism, to generate supply chain optimization scheme data; based on the causal association feature data, dynamically optimize the transportation route and inventory allocation strategy, and at the same time introduce a flexible emergency adjustment module to quickly optimize the scheme in case of emergencies; Input the optimization scheme data into the intelligent contract system to automatically execute transportation, inventory allocation, and order delivery operations; Based on the real-time feedback data, generate feedback adjustment data for improving the supply chain operation through anomaly monitoring and dynamic adjustment, and generate improvement scheme data for optimizing the next-stage operation, realizing the closed-loop optimization of the supply chain operation.

2. The crude oil supply chain management method according to claim 1, wherein The generation of the multi-modal fusion feature data includes the following steps: Use a variational autoencoder to reduce the dimension and denoise the time series data, and extract time series feature data; Through a vision Transformer model, hierarchically encode the semantic features of the image data to extract visual feature data; Through the BERT model, deeply analyze the semantic and context information of the text data to extract text feature data; Align and deeply fuse the above feature data to generate multi-modal fusion feature data.

3. The crude oil supply chain management method according to claim 2, wherein, The generation of the causal association feature data includes the following steps: Jointly model the time series features, spatial distribution features, and semantic features in the multi-modal fusion feature data through a structural causal model to analyze the causal relationship between supply chain nodes; Based on the Bayesian network, deduce the probability distribution of the causal path, generate the causal association weights between each node, and construct causal association feature data according to the weights to reflect the key influencing factors of the dynamic changes in the supply chain.

4. The crude oil supply chain management method according to claim 1, characterized in that The construction of the digital twin state model includes the following steps: Model the dynamic topological structure of the supply chain through a dynamic graph convolutional network to determine the association strength and change trend between nodes; Use the real-time physical parameter data collected by sensors to update the state of supply chain nodes, including the oil flow rate of transportation nodes, the temperature and humidity of storage nodes, and the equipment operation parameters of refining nodes; Based on the above data, construct a supply chain state model to generate digital twin real-time data and prediction data.

5. The crude oil supply chain management method according to claim 1, characterized in that The generation of the extreme scenario simulation data includes the following steps: Set the simulation conditions for the cross-node association scenario according to the causal association feature data, including the resource supply and demand fluctuations between nodes and the risk of transportation route interruption; Set the simulation parameters for extreme climate conditions, including the impacts of high temperature, low temperature, and high humidity environments on transportation and warehousing; Iteratively optimize the simulation parameters through the inverse reinforcement learning algorithm to generate the extreme scenario simulation data for the above scenarios.

6. The crude oil supply chain management method according to claim 1, wherein The specific implementation of the multi-objective optimization algorithm includes the following steps: Construct the optimization objective function, including the minimization of the total transportation route cost, the maximization of the inventory turnover rate, and the minimization of the delivery time; Search the solution space through the hybrid quantum particle swarm optimization algorithm, and dynamically adjust the weights of each optimization objective in each iteration; Combine the genetic algorithm to globally verify and locally refine the optimization results to generate the supply chain optimization scheme data that meets the multi-objective optimization requirements.

7. The crude oil supply chain management method according to claim 6, wherein The implementation of the flexible emergency adjustment module includes the following steps: Based on the real-time feedback data, use the reinforcement learning algorithm to identify emergencies, including abnormal situations such as transportation route delays, warehouse overloading, or order surges; Dynamically generate emergency adjustment strategies, including reallocating inventory, adjusting transportation routes, and optimizing the delivery sequence; Incorporate the emergency adjustment strategies into the supply chain optimization scheme data.

8. The crude oil supply chain management method according to claim 1, characterized in that, The implementation of the intelligent contract system includes the following steps: Build a distributed ledger based on blockchain technology to record the status data of the execution of supply chain tasks, including the completion time of transportation tasks, the inbound and outbound records of warehousing tasks, and the delivery information of orders; Automatically verify the task completion situation through intelligent contracts and trigger subsequent operations according to the verification results; Generate an immutable operation log in the blockchain system.

9. The crude oil supply chain management method according to claim 8, characterized in that, The generation of the feedback adjustment data includes the following steps: Analyze the real-time feedback data through the time series anomaly detection model to identify potential anomalies in the operation of the supply chain, including transportation delays, insufficient inventory, or overloading; Generate feedback adjustment data according to the anomaly detection results and dynamically update the optimization scheme data to improve the operation efficiency of the supply chain.

10. A system for implementing the crude oil supply chain management method according to any one of claims 1-9, characterized in that, Including: A data processing module, which is used to collect multi-source data of each node in the crude oil supply chain, including time series data, image data, and text data, and perform feature extraction and fusion on the data to generate multi-modal fusion feature data; A digital twin simulation module, which is used to build a supply chain status model based on the multi-modal fusion feature data, generate digital twin real-time data and prediction data, and generate extreme scenario simulation data by setting multi-level simulation conditions; An optimization module, which is used to generate supply chain optimization scheme data by using a multi-objective optimization algorithm and combining a global dynamic adjustment mechanism based on the extreme scenario simulation data, and dynamically optimize the transportation route and inventory allocation strategy based on the causal association feature data; An execution and feedback module, which is used to input the optimization scheme data into the intelligent contract system, automatically execute transportation, inventory allocation, and order delivery operations, and generate feedback adjustment data and improvement scheme data based on the real-time feedback data to achieve the closed-loop optimization of the supply chain operation.

Images