Oil product resource intelligent scheduling method and device

Abstract

The invention discloses an oil product resource intelligent scheduling method and device, and the method comprises the steps: obtaining oil depot data, vehicle data and real-time operation data of a plurality of gas stations based on a streaming processing architecture; inputting the oil depot data, the vehicle data and the real-time operation data of the plurality of gas stations into a trained intelligent recommendation model, and outputting prediction information of the plurality of business scene identifiers and distribution plan recommendation information of each gas station; the business scene identifier reflects one of a sales volume condition, an inventory condition and a distribution condition; the distribution plan recommendation information comprises oil depot information, vehicle information, distribution time information and distribution oil product information; the intelligent recommendation model is obtained by training an LSTM model by using historical oil depot data, vehicle data, operation data of a gas station, and time information and description information of a business scene identifier in advance. According to the method, more intelligent and automatic support is provided for scheduling plan decision making, and the oil product resource scheduling effect is improved.

Description

Technical Field

[0001] This invention relates to the field of refined oil resource scheduling technology, and in particular to a method and apparatus for intelligent scheduling of refined oil resources. Background Technology

[0002] This section is intended to provide background or context for the embodiments of the invention set forth in the claims. The description herein is not an admission that it is prior art simply because it is included in this section.

[0003] Oil resource allocation refers to a series of organizational and coordination activities conducted during the transportation and distribution of refined oil products (such as gasoline, diesel, and aviation kerosene) to ensure their efficient and safe distribution to various gas stations or end users. This requires comprehensive consideration of factors such as the use of vehicles, personnel, oil depots, and the optimization of transportation routes. Several concepts are involved:

[0004] Oil depot: A storage facility for refined oil products, serving as a provider of oil products during the distribution process.

[0005] Oil collection agencies: These are mainly the various oil tanks under the jurisdiction of gas stations, which sell refined oil products and are the demanders of oil products in the dispatching process.

[0006] Vehicles: Responsible for transporting oil products from oil depots to oil receiving agencies. Different vehicles have different carrying capacities and oil product types.

[0007] Oil products: Different oil properties and types.

[0008] The business scenario for the refined oil resource scheduling of this invention is the transportation of oil products from oil depots to oil receiving agencies.

[0009] In existing business scenarios, operations are carried out manually based on the existing oil resource scheduling system. However, manually formulating oil resource scheduling plans requires a lot of time and effort, and the quality of the scheduling plan largely depends on the experience and skill of the dispatcher. Existing technologies have emerged that combine machine learning with linear programming, but because the existing oil resource scheduling system cannot fully and accurately grasp the dynamic information of oil resources, the scheduling decisions lack a scientific basis, the overall optimization capability is poor, and the problem of untimely oil supply is still likely to occur, resulting in poor oil resource scheduling effect. Summary of the Invention

[0010] This invention provides an intelligent oil resource scheduling method to improve oil resource scheduling efficiency based on comprehensive and accurate dynamic information on oil resources. The method includes:

[0011] Based on a streaming processing architecture, the system acquires real-time operational data from oil depots, vehicles, and multiple gas stations. Oil depot data includes information on oil depot type and storage capacity, while vehicle data includes information on vehicle qualifications, fuel transport attributes, and driving status. Real-time operational data includes sales data, inventory data, and fuel delivery data.

[0012] The system inputs oil depot data, vehicle data, and real-time operational data from multiple gas stations into a trained intelligent recommendation model. The model outputs predictive information and delivery plan recommendations for each gas station based on multiple business scenario identifiers. These business scenario identifiers reflect one of three factors: sales volume, inventory status, or delivery status. The predictive information includes time and description information. The delivery plan recommendations include oil depot information, vehicle information, delivery time information, and delivered fuel information. The intelligent recommendation model is pre-trained using historical oil depot data, vehicle data, gas station operational data, and the time and description information of the business scenario identifiers to train a Long Short-Term Memory (LSTM) network model.

[0013] In one embodiment, based on a streaming processing architecture, oil depot data, vehicle data, and real-time operational data of multiple gas stations are acquired, including: subscribing to oil depot data, vehicle data, and real-time operational data of multiple gas stations from a pre-configured distributed publish-subscribe messaging tool using a streaming processing tool; the distributed publish-subscribe messaging tool is built based on a Kafka cluster.

[0014] In one embodiment, before inputting the oil depot data, vehicle data, and real-time operational data of multiple gas stations into the trained intelligent recommendation model, the method further includes: performing data cleaning, normalization, and label encoding on the oil depot data, vehicle data, and real-time operational data of multiple gas stations to obtain multiple label encoding matrix data; and performing dimensionality reduction processing on the multiple label encoding matrix data to obtain dimensionality-reduced data.

[0015] The data includes oil depot data, vehicle data, and real-time operational data from multiple gas stations. These data are then fed into the trained intelligent recommendation model, including inputting the dimensionality-reduced principal component data into the trained intelligent recommendation model.

[0016] In one embodiment, dimensionality reduction processing is performed on multiple label coding matrix data to obtain dimensionality-reduced data, including: applying matrix decomposition to the multiple label coding matrix data to obtain multiple decomposed matrices; performing feature extraction based on the multiple decomposed matrices to obtain low-dimensional feature data; and performing principal component analysis (PCA) dimensionality reduction processing on the low-dimensional feature data to obtain dimensionality-reduced data.

[0017] In one embodiment, the intelligent recommendation model is trained as follows:

[0018] Collect historical data: oil depot data, vehicle data, and gas station operational data, as well as time and description information of business scenario identifiers, to form training and testing sets;

[0019] Construct an LSTM model; the LSTM model includes an input gate, a forget gate, and an output gate;

[0020] The LSTM model architecture is trained using the training set and tested using the test set to obtain a well-trained intelligent recommendation model.

[0021] In one embodiment, the business scenario identifier includes one or any combination of the following:

[0022] Sales are about to reach their peak; large customer groups are about to have large-scale tank and drum sales; inventory is expected to reach the average shipment volume; current inventory stability is poor; delivery volume is low; and delivery frequency is low.

[0023] In one embodiment, the delivery plan recommendation information is displayed in the form of a drop-down list, sorted from high to low recommendation level. Each list includes oil depot information, vehicle information, delivery time information, and delivery oil information.

[0024] In one embodiment, after outputting the prediction information and delivery plan recommendation information of multiple business scenario identifiers for each gas station, the method further includes: receiving the user's selection instruction for the delivery plan recommendation information, and / or modification instruction, to obtain the delivery plan recommendation information determined by the user; and distributing the delivery plan recommendation information determined by the user to the oil depot terminal, the gas station terminal, and the vehicle terminal.

[0025] This invention also provides an intelligent oil resource scheduling device for scheduling oil resources based on comprehensive and accurate dynamic information, thereby improving the efficiency of oil resource scheduling. The device includes:

[0026] The data acquisition module is used to acquire oil depot data, vehicle data, and real-time operational data from multiple gas stations based on a streaming processing architecture. Oil depot data includes data reflecting the type of oil depot and its storage capacity; vehicle data includes data reflecting vehicle qualifications, oil transportation attributes, and driving status; real-time operational data includes sales data, inventory data, and oil delivery data.

[0027] The intelligent recommendation module is used to input oil depot data, vehicle data, and real-time operational data from multiple gas stations into a trained intelligent recommendation model, and output prediction information and delivery plan recommendation information for each gas station based on multiple business scenario identifiers. The business scenario identifiers reflect one of the following: sales volume, inventory status, and delivery status. The prediction information includes time information and descriptive information. The delivery plan recommendation information includes oil depot information, vehicle information, delivery time information, and delivered fuel information. The intelligent recommendation model is trained in advance using historical oil depot data, vehicle data, gas station operational data, and the time and descriptive information of the business scenario identifiers to train an LSTM model.

[0028] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described intelligent scheduling method for oil resources.

[0029] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent scheduling method for oil resources.

[0030] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described intelligent scheduling method for oil resources.

[0031] This invention, based on a streaming processing architecture, acquires oil depot data, vehicle data, and real-time operational data from multiple gas stations, obtaining comprehensive and accurate dynamic information on oil resources. Then, it utilizes artificial intelligence algorithms, specifically an LSTM model, to predict the business scenario identifiers for each gas station and recommend delivery plans. This method, based on big data and intelligent models, can monitor and predict key characteristics of gas stations, avoiding manual scheduling plans and providing intelligent data recommendations for dispatchers. This provides more intelligent and automated support for scheduling plan decisions, significantly improving the effectiveness of oil resource scheduling. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present 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 some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0033] Figure 1 This is a flowchart illustrating the intelligent scheduling method for oil resources in an embodiment of the present invention.

[0034] Figure 2 This is a specific example diagram of the intelligent scheduling method for oil resources in an embodiment of the present invention;

[0035] Figure 3 This is a schematic diagram of data classification in an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram illustrating the analysis of the correlation between oil products and vehicle dispatch distribution in an embodiment of the present invention;

[0037] Figure 5 This is a schematic diagram of an intelligent oil resource scheduling device in an embodiment of the present invention. Detailed Implementation

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0039] The acquisition, transmission, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0040] Figure 1 This is a flowchart illustrating the intelligent scheduling method for oil resources in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes:

[0041] Step 101: Based on the streaming processing architecture, acquire oil depot data, vehicle data, and real-time operational data from multiple gas stations; oil depot data includes data reflecting the type of oil depot and the amount of oil stored in the depot; vehicle data includes data reflecting vehicle qualifications, oil transportation attributes, and driving status; real-time operational data includes sales data, inventory data, and oil delivery data.

[0042] Step 102: Input the oil depot data, vehicle data, and real-time operational data of multiple gas stations into the trained intelligent recommendation model, and output the following for each gas station: prediction information and delivery plan recommendation information for multiple business scenario identifiers; the business scenario identifiers reflect one of the following: sales volume, inventory status, and delivery status; the prediction information includes time information and descriptive information; the delivery plan recommendation information includes oil depot information, vehicle information, delivery time information, and delivered fuel information; the intelligent recommendation model is trained in advance using historical oil depot data, vehicle data, and gas station operational data, as well as the time and descriptive information of the business scenario identifiers, to train the LSTM model.

[0043] from Figure 1As shown in the flowchart, this embodiment of the invention uses a streaming processing architecture to acquire oil depot data, vehicle data, and real-time operational data from multiple gas stations, obtaining comprehensive and accurate dynamic information on oil resources. Then, it utilizes artificial intelligence algorithms, specifically an LSTM model, to predict the business scenario identifiers of each gas station and recommend delivery plans. This method, based on big data and intelligent models, can monitor and predict key characteristics of gas stations, avoiding manual scheduling plans and providing intelligent data recommendations for dispatchers. This provides more intelligent and automated support for scheduling plan decisions, significantly improving the effectiveness of oil resource scheduling.

[0044] Figure 2 This is a specific example diagram of the intelligent scheduling method for oil resources in an embodiment of the present invention, as shown below. Figure 2 As shown, this embodiment of the invention, based on a distribution center, deeply mines and analyzes data such as sales, inventory, and delivery plans of gas stations to better understand gas station market trends and customer needs. Based on key features and tag data of gas stations, it performs monitoring, prediction, and early warning, realizing priority recommendation of oil depots and vehicles. It provides dispatchers with timely delivery plans, offers recommendations of available transportation resources and available oil depots, improves the efficiency of dispatch plan formulation, and provides more intelligent and automated support for decision-making through data-driven approaches.

[0045] First, based on a streaming processing architecture, data from oil depots, vehicles, and real-time operational data from multiple gas stations are acquired.

[0046] Oil depot data includes information reflecting the type of oil depot and its storage capacity, such as the type of reserve oil depot, transfer oil depot, supply oil depot, distance from the oil depot to the gas station, and the actual amount of oil stored in the oil depot.

[0047] Vehicle data includes information on vehicle qualifications, oil transport attributes, and driving status, such as basic vehicle qualification information, types of oil that can be transported, and transportation plans.

[0048] Real-time operational data includes, but is not limited to, sales data, inventory data, and oil delivery data, such as hourly sales data, daily sales data, peak and minimum sales periods, daily dispatch volume, inventory levels, historical oil depots, and historical transportation capacity for gas stations.

[0049] In one embodiment, acquiring oil depot data, vehicle data, and real-time operational data from multiple gas stations based on a streaming processing architecture may include:

[0050] The streaming processing tool subscribes to oil depot data, vehicle data, and real-time operational data from multiple gas stations from a pre-configured distributed publish-subscribe messaging tool; the distributed publish-subscribe messaging tool is built on a Kafka cluster.

[0051] For example, log collection tools can be pre-configured on multiple gas station servers, and then connected to a distributed publish-subscribe messaging tool. Real-time data from the gas stations can be distributed to the distributed publish-subscribe messaging tool, and then a streaming processing tool can consume the data from the distributed publish-subscribe messaging tool. An example of a streaming processing tool is Apache Flink.

[0052] Once the data is received, it can be stored in various data storage systems, such as relational databases, NoSQL databases, and data warehouses.

[0053] This invention enables the understanding of real-time dynamic changes in oil product transportation and scheduling. The distributed publish-subscribe messaging tool, as a high-throughput distributed publish-subscribe messaging system, can capture data generated in various stages of oil product transportation and sales in real time and transmit this data quickly and reliably to downstream systems. Flink, an open-source stream processing framework, can process data from the distributed publish-subscribe messaging tool in real time. It supports true stream processing (event-time semantics), accurately handling out-of-order data and ensuring the accuracy of processing results. The combination of the distributed publish-subscribe messaging tool and Flink ensures extremely low latency throughout the entire process from data generation to processing to result output. This is crucial for gas station inventory management, which requires rapid response to market changes and customer demands. Both the distributed publish-subscribe messaging tool and Flink are designed as high-throughput systems capable of handling large amounts of concurrent data. As business volume increases, the system can be easily expanded by adding nodes to meet greater data processing demands. The distributed publish-subscribe messaging tool ensures data reliability through a multi-replica mechanism, ensuring that even if some nodes fail, normal data production and consumption will not be affected. Flink offers robust state management and fault tolerance mechanisms, ensuring that tasks can automatically recover and resume execution from the point of failure when a node fails, guaranteeing the continuity and accuracy of data processing. Dynamic scheduling and optimization: Flink supports complex window operations and state management, enabling dynamic adjustments to scheduling strategies based on real-time data to optimize resource allocation and achieve more refined oil transportation and inventory management.

[0054] In this example, the Kafka+Flink streaming architecture demonstrates significant advantages in addressing the real-time dynamic changes in oil product allocation and scheduling. It can process large amounts of data efficiently and in real-time, providing more refined and intelligent support for gas station inventory management.

[0055] In one embodiment, before inputting oil depot data, vehicle data, and real-time operational data from multiple gas stations into the trained intelligent recommendation model, the process may further include:

[0056] Data cleaning, normalization, and tag coding were performed on oil depot data, vehicle data, and real-time operational data from multiple gas stations to obtain multiple tag coding matrix data.

[0057] Dimensionality reduction is performed on multiple label encoding matrix data to obtain the dimensionality-reduced data;

[0058] Inputting oil depot data, vehicle data, and real-time operational data from multiple gas stations into a trained intelligent recommendation model can include:

[0059] The dimensionality-reduced principal component data is input into the trained intelligent recommendation model.

[0060] For example, the gas station code (unique identifier of the gas station) is a string, the tank number and oil product number are integers, the sales volume is a floating-point number, and the date is a time type. Different types of data need to be standardized and normalized.

[0061] Figure 3 This is a schematic diagram of data classification in an embodiment of the present invention, for reference only. Figure 3 The acquired feature data can be divided into hourly inventory data, hourly sales data, and hourly sales ratio data. Specifically, it is divided into inventory features, oil receiving features, planning features, and shipment volume, including information such as gas station code, oil tank number, oil product number, and date.

[0062] Different types of data are preprocessed during implementation.

[0063] Label encoding can be used, for example, converting category labels to integers or one-hot encodings, or using one-hot encoding to convert category labels to vectors, or directly representing categories using numbers. In implementation, existing models or rules can be used to generate labels. Training with a small amount of labeled data and a large amount of unlabeled data can generate more accurate labels.

[0064] Furthermore, to improve data quality, in one embodiment, dimensionality reduction processing is performed on multiple label encoding matrix data to obtain dimensionality-reduced data, which may include:

[0065] Matrix decomposition is applied to multiple tag-encoded matrix data to obtain multiple decomposed matrices;

[0066] Feature extraction is performed based on multiple decomposed matrices to obtain low-dimensional feature data;

[0067] Principal component analysis (PCA) is used to reduce the dimensionality of the low-dimensional feature data, resulting in dimensionality-reduced data.

[0068] For example, a matrix factorization method (such as singular value decomposition or nonnegative matrix factorization) is used to decompose a data matrix into the product of two or more matrices. Features are extracted from the decomposed matrix to achieve the first dimensionality reduction. Then, PCA is performed on the extracted features to further reduce the feature dimensionality.

[0069] In this embodiment of the invention, the intelligent recommendation model is trained on an LSTM model by pre-using historical data on oil depots, vehicles, and gas stations, as well as time and description information of business scenario identifiers.

[0070] In one embodiment, the intelligent recommendation model can be trained as follows:

[0071] Collect historical data: oil depot data, vehicle data, and gas station operational data, as well as time and description information of business scenario identifiers, to form training and testing sets;

[0072] Construct an LSTM model; the LSTM model includes an input gate, a forget gate, and an output gate;

[0073] The LSTM model architecture is trained using the training set and tested using the test set to obtain a well-trained intelligent recommendation model.

[0074] When constructing the training and test sets, the aforementioned data dimensionality reduction methods are also needed to process the data, enabling the combined training of matrix factorization, PCA, and LSTM models. This combined approach can reduce data dimensionality while retaining the most important features, thereby improving the training efficiency and prediction accuracy of the model.

[0075] The LSTM model consists of three parts: input gate, forget gate, and output gate. It controls the transmission state through gating state, remembering messages that need to be remembered for a long time while discarding unimportant information.

[0076] LSTM can effectively capture nonlinear features. At the same time, due to its unique design structure, LSTM is also suitable for processing and predicting important events with very long intervals and delays in time series. In other words, LSTM can "memorize" features that are long away from the current time series, such as the impact of information such as holidays and historical temperatures on gas station sales.

[0077] LSTM is a special type of RNN that avoids long-term dependency issues through deliberate design. All recurrent neural networks have chain-like repeating modules.

[0078] (1) Forgotten Gate

[0079] The forgetting gate formula is f t =σ(W f [h t-1 xt ]+b f In the formula, f t For the output at time step t, σ() is the sigmoid function, W f The weight matrix for the forget gate, h, is used to connect historical information and the current input. t-1 The hidden state of the previous time step represents past historical information, x. t As the current time feature input, b f As a bias term, the first step of LSTM is to decide which information to discard from the cell state. This decision is made by the sigmoid network layer known as the "forget gate layer." It receives h... t-1 and x t The output value for each number in the cell state is between 0 and 1. 1 means "accept this completely" and 0 means "ignore this completely".

[0080] (2) Input Gate

[0081] The next step is to determine what new information needs to be preserved in the cell state. The input gate formula is i t =σ(W i [h t-1 x t ]+b i ), In the formula, i t To determine the input gate output at time step t, W determines which new information will be added to the cell state. i Let b be the weight matrix of the input gate. i W is the bias vector of the input gate. c b is the weight matrix used to generate candidate cell states. c To generate the bias vector for candidate cell states, this is divided into two parts. The first part uses an sigmoid network layer, known as the "input gate layer," to determine which information needs updating. The second part uses a tanh-shaped network layer to create a new vector of candidate values. This can be used to add to the cell state. The next step combines the two parts above to produce an update to the state.

[0082] (3) Cell state renewal

[0083] Update old cell state C t-1 Update to C t ,in The previous steps have already determined what needs to be done; all that's needed is to follow them.

[0084] Multiply the old state by a factor to forget what we decided to forget. Then add... These are the new candidate values, scaled proportionally to the updated values ​​determined for each state.

[0085] (4) Output gate

[0086] Finally, the output value needs to be determined. The output depends on the cell state, but it will be a "filtered" version. First, a sigmoid network layer is run to determine which parts of the cell state can be output. Then, the cell state is input to tanh (adjusted to a value between -1 and 1) and multiplied by the output value of the sigmoid network layer. This allows us to output the desired score.

[0087] In practice, the LSTM model can be improved, for example, by adding an LSTM model with a "peephole connection." A popular LSTM variant adds a "peephole connection" to the LSTM, which means that the gate network layer can be fed the cell state as input. Those skilled in the art can improve the LSTM model according to the actual situation.

[0088] When training an LSTM model, optimization is performed using training data.

[0089] Use a trained LSTM model to make predictions on new data.

[0090] This invention combines LSTM with matrix factorization and PCA. While maintaining the powerful time series modeling capabilities of the LSTM model, PCA can effectively reduce data dimensionality, improve computational efficiency, and reduce the risk of overfitting.

[0091] In this embodiment of the invention, oil depot data, vehicle data, and real-time operational data of multiple gas stations are input into a trained intelligent recommendation model, which outputs prediction information and delivery plan recommendation information for each gas station based on multiple business scenario identifiers.

[0092] The business scenario identifier reflects one of the following: sales volume, inventory status, and delivery status. The forecast information includes time information and descriptive information. In one embodiment, the business scenario identifier includes one or any combination of the following:

[0093] Sales are about to reach their peak; large customer groups are about to have large-scale tank and drum sales; inventory is expected to reach the average shipment volume; current inventory stability is poor; delivery volume is low; and delivery frequency is low.

[0094] In this embodiment of the invention, data characteristics such as historical inventory changes, sales trends, and delivery plans of each gas station are extracted to form a standardized tagging system. An intelligent recommendation model is established to achieve recommendations for multiple scenarios under the three dimensions of sales volume, inventory, and plans. Key data of the scenarios are monitored and compared with the extracted feature indicators to form information warnings for key scenarios, triggering the formulation and adjustment of the next scheduling plan. When formulating the scheduling plan, the available oil depots and available oil delivery vehicles are recommended.

[0095] Among them, the predictive information of multiple business scenarios can be visualized in the distribution center, and real-time reminders can be set.

[0096] (1) Sales are about to reach their peak.

[0097] Business scenario description: Each gas station has different time period preferences for the sale of individual oil products, which are considered in hourly units. When an individual oil product is about to reach its peak sales period, the dispatcher should be notified.

[0098] Business Scenario Function: To monitor the sales trends of each fuel type at each gas station in real time, and to provide alerts when sales are about to reach their peak, so as to adjust transportation strategies and avoid fuel shortages at gas stations. Table 1 shows an example of recommended business scenario 1.

[0099] Table 1

[0100]

[0101] (2) Sales of large customer groups of cans and drums are about to occur.

[0102] Business Scenario Description: Oil product sales are divided into retail and wholesale. Tank sales volume represents the wholesale sales volume to large customer groups. Wholesale sales are characterized by large volume and low frequency, making them crucial for gas station sales. Accurately identifying the timing of tank sales and providing alerts to dispatchers is essential.

[0103] Business Scenario Function: This involves statistically modeling tank sales data in real-time, providing alerts when sales are imminent, and adjusting transportation strategies to ensure a smooth supply of oil products. Table 2 shows an example of recommended business scenario 2.

[0104] Table 2

[0105]

[0106]

[0107] (3) Inventory is expected to reach the average shipment volume.

[0108] Scenario Description: The average shipment volume is the average inventory level at which oil transportation is required. If the inventory level is expected to reach the average shipment volume within 24 hours, indicating that oil transportation is necessary, the interface should provide a prompt.

[0109] Scenario Function: Real-time monitoring of inventory levels. An alert is issued when inventory is too low and the remaining time is less than or equal to 24 hours, indicating that oil transportation will be needed soon. This allows dispatchers to take notice and prevent gas stations from running out of inventory. Table 3 shows an example of recommended business scenario 3.

[0110] Table 3

[0111]

[0112] (4) Current inventory stability is poor.

[0113] Business Scenario Description: The inventory level of a specific oil product at a single gas station typically remains stable, meaning the inventory level fluctuates around a specific value relative to the tank capacity. If the current stable inventory level is lower than the usual stable level and has remained below this level for some time, dispatchers should pay close attention.

[0114] Business Scenario Function: Real-time monitoring of gas station inventory levels. If the current stability is poor, a notification is issued, along with a series of indicators of inventory stability and the duration of the poor stability. This facilitates timely adjustments to fuel delivery strategies to prevent fuel shortages at gas stations. Table 4 shows an example of recommended business scenario 4.

[0115] Table 4

[0116]

[0117] (5) The delivery volume is low.

[0118] Business scenario description: The delivery volume of a specific type of fuel at a single gas station typically remains stable. If today's delivery volume is lower than the usual stable level, dispatchers should pay attention.

[0119] Functional purpose: Real-time monitoring of daily delivery volume at gas stations; alerts are issued if the current level is low. This facilitates timely adjustments to fuel delivery strategies and prevents fuel shortages at gas stations. Table 5 shows an example of recommended business scenario 5.

[0120] Table 5

[0121]

[0122] (6) The number of deliveries is low.

[0123] Business scenario description: The number of deliveries of a specific type of fuel at a single gas station typically remains stable. If the number of deliveries today is lower than the usual stable level, dispatchers should pay attention.

[0124] Business Scenario Function: Real-time monitoring of the daily delivery frequency of gas stations; if the current level is low, a notification is issued. This facilitates timely adjustments to the fuel delivery strategy and prevents fuel shortages at gas stations. Table 6 shows an example of recommended business scenario 6.

[0125] Table 6

[0126]

[0127] (7) The next delivery cycle is expected to begin tomorrow.

[0128] Business scenario description: The delivery cycle is defined as the time or spatial interval between completing a round of deliveries when the relative sizes of deliveries repeat in the same order. A notification will be given if the next delivery cycle is about to begin tomorrow.

[0129] Functional purpose of this business scenario: To enable overall control of the gas station's sales rhythm and facilitate timely adjustments to the fuel transportation strategy. Table 7 shows an example of recommended business scenario 7.

[0130] Table 7

[0131]

[0132]

[0133] In this embodiment of the invention, the delivery plan recommendation information includes oil depot information, vehicle information, delivery time information, and delivery fuel information. In one embodiment, the delivery plan recommendation information is displayed in a drop-down list format, sorted from highest to lowest recommendation level, with each list including oil depot information, vehicle information, delivery time information, and delivery fuel information.

[0134] Drop-down menus are a common control in interface design. In this embodiment of the invention, a drop-down list is used to display multiple content tags, allowing users to select one or more items from a predefined list. Using drop-down menus offers several advantages:

[0135] Provide the best option: Provide users with the system's best option, which can be selected by default.

[0136] Diminishing alternative options: Since the dropdown list hides other available options, it effectively minimizes the use of alternatives. This is advantageous when the default values ​​satisfy most users, and the alternatives would be risky for non-expert users.

[0137] Space saving: Drop-down menus are very useful if the interface can only provide a small amount of space for the user. Although drop-down menus are small, they can contain a lot of information in a small space.

[0138] Predictable Input: Collecting user information through text input fields is unpredictable; users may make mistakes, misspell words, or misunderstand the input. Drop-down menus, by providing options, can predict user input.

[0139] Flexibility: The biggest advantage of drop-down menus is that a single drop-down list can contain all options without needing to change the design based on the number of options.

[0140] The drop-down menu recommendation function needs to include two parts: judging the feasibility of oil depots and vehicles for oil pickup, because due to limitations such as the characteristics and storage capacity of oil depots, as well as the carrying capacity and compatibility of vehicles with oil properties, not all oil depots and vehicles can meet business needs, so screening is required; and scoring and ranking the recommendation degree of oil depots and vehicles for oil pickup.

[0141] The UI provides drop-down menu options for the oil receiving agencies and oil products that require them. After the user makes a selection, the recommendation system can provide the top recommended options based on the recommendation algorithm, from highest to lowest, for dispatchers to refer to.

[0142] The vehicle data displayed in the drop-down list is shown in Table 8.

[0143] Table 8

[0144] license plate number Gasoline and diesel attributes state Cargo compartment Cargo capacity xxx diesel fuel Open 22300，22300 44600 xxx diesel fuel Open 22300，22300 44600

[0145] The information needed for the vehicle includes: license plate number, gasoline / diesel attributes, status, and cargo compartment.

[0146] Data processing: The capacities of each compartment in the vehicle are added together to obtain the total capacity / carrying capacity of the vehicle.

[0147] Data feature function: Vehicles that do not meet the required fuel properties will not be arranged for transportation, vehicles whose status is not activated will not be arranged for transportation, and vehicles whose carrying capacity is lower than the required amount will not be arranged for transportation.

[0148] The drop-down list displays information such as oil depots and delivery services, as shown in Table 9.

[0149] Table 9

[0150]

[0151] The following delivery information required in past delivery plans is recommended for display: delivery tracking number, delivery plan status, oil depot ERP code, pipeline code, company code, license plate number, oil product code, tank number, etc.

[0152] In one embodiment, after outputting the prediction information and delivery plan recommendation information for each gas station based on multiple business scenario identifiers, it may further include:

[0153] Receive user's selection instruction for delivery plan recommendation information, and / or modification instruction, and obtain the delivery plan recommendation information determined by the user;

[0154] The delivery plan recommendation information determined by the user is sent to the oil depot terminal, gas station terminal and vehicle terminal.

[0155] Gas stations typically have pre-planned scheduling schedules. Therefore, the recommendations displayed on the UI usually include options such as modifying existing scheduling plans, deleting existing scheduling plans, and adding new scheduling plans. This example allows users to adjust and modify the generated scheduling plans, enabling customized configuration and significantly improving the efficiency and time saved by users in creating plans.

[0156] The scheduling plan set by the user needs to be reviewed by the next higher level. Finally, only the delivery information with the delivery plan status of "approved" is retained.

[0157] In one embodiment, the scheduling plans of each gas station and each oil depot are periodically analyzed to obtain feature analysis results, which are then used to update and iterate the intelligent recommendation model. The feature analysis results include, but are not limited to, the cumulative number of oil depot scheduling, the number of gas station (i.e., oil receiving agency) scheduling, the number of vehicle scheduling, and the correlation between various feature data. Feature data includes oil products, vehicles, oil depots, etc.

[0158] For example, observing the distribution of cumulative dispatch times at oil depots, depot 095A accounts for the vast majority, while depots 090A, 092A, 099A, and 125A also have a significant number of dispatches. Observing the cumulative dispatch times for oil products, the main dispatch demands are for No. 0 diesel, No. 92B gasoline, and No. 95B gasoline.

[0159] The number of times the oil collection agencies were dispatched was statistically analyzed and a distribution map was constructed. The distribution map shows that the vast majority of oil collection agencies dispatched less than 20 times.

[0160] Statistical analysis of vehicle dispatch frequency was conducted, and a distribution map was constructed. This map reveals that, unlike the refueling mechanism, vehicle dispatch frequency is significantly divided into three parts. The first part follows an approximately normal distribution centered at 110, consistent with random dispatch. The second part consists of vehicles dispatched more than 300 times, especially 400 times; these vehicles significantly exceed the normal frequency, indicating a small number of vehicles undertaking an excessive dispatch workload. The third part comprises vehicles dispatched less than 20 times; these vehicles undertake too few dispatch tasks, resulting in significant idle time.

[0161] Observe the correlation between different items within the different characteristic data of the scheduling plan. For example, the Pearson correlation coefficient. The larger the absolute value, the stronger the correlation. The closer to 0, the weaker the correlation. Positive numbers represent positive correlation, and negative numbers represent negative correlation.

[0162] Calculating the correlation of oil product distribution at oil depots reveals a very high positive correlation between No. 0 diesel, No. 92 gasoline, and No. 95 gasoline. This is because these oil products have large dispatch volumes and are mainly handled by oil depot 095A.

[0163] Figure 4 This is a schematic diagram illustrating the analysis of the correlation between oil products and vehicle dispatch distribution in an embodiment of the present invention, such as... Figure 4 As shown, the correlation of fuel distribution in vehicles was calculated: it was found that three types of diesel formed one correlation group, and five types of gasoline formed another correlation group. Furthermore, the correlation within the 92-octane (VIA) and 95-octane (VIA) gasoline grades was high, while the correlation between the 92-octane (VIA) and 95-octane (VIA) gasoline grades was moderate. Firstly, the gasoline / diesel properties of the vehicles determined the negative correlation between gasoline and diesel. Additionally, significant differences were observed in the dispatching choices of 92-octane (VIA) and 95-octane (VIA) gasoline grades for additional vehicles.

[0164] If we combine the oil depot and the vehicle into a unified element, i.e., which vehicle is dispatched from which oil depot, and calculate the correlation of the oil products on it, we can find that the above differences become more significant.

[0165] In one embodiment, oil depot data, vehicle data, and real-time operational data from multiple gas stations are input into a trained intelligent recommendation model, which outputs predictive information and delivery plan recommendation information for each gas station based on multiple business scenario identifiers. This may include:

[0166] The system inputs oil depot data, vehicle data, and real-time operational data from multiple gas stations into a trained intelligent recommendation model, which then updates and outputs predictive information and delivery plan recommendations for each gas station at set time intervals, based on multiple business scenario identifiers.

[0167] For example, the interface is refreshed every 15 minutes so that users can see the latest scheduling plan changes in real time.

[0168] In summary, this invention provides intelligent recommendations for gas station dispatch plans based on big data and artificial intelligence technologies, focusing on the business needs of gas station dispatch plan formulation within the refined oil dispatch center. When formulating daily gas station delivery plans, it identifies abnormal information affecting the stable operation of gas station oil dispatch based on historical gas station sales changes, inventory trends, and dispatch plan data characteristics. It then provides intelligent data recommendations to dispatch personnel, offering more intelligent and automated support for planning decisions and comprehensive support and optimization for dispatch management. The effects of this invention include:

[0169] 1. By using a streaming processing architecture, the system can access real-time dynamic changes in oil transportation and scheduling, preventing excessive inventory or supply shortages at gas stations and ensuring their normal operation.

[0170] 2. Utilize big data and machine learning technologies to monitor and predict key features and tag data of gas stations, identify abnormal states in the dispatching process, and quickly provide behavioral recommendations for dispatching personnel;

[0171] 3. During the scheduling plan formulation process, based on historical scheduling behavior characteristics, provide more decision support for dispatching personnel, realize the recommendation of available / frequent transportation capacity and optimize oil depots, ensure the stable supply of refined oil products to gas stations, and improve operational efficiency.

[0172] 4. Through multi-dimensional data visualization, the changes in various indicators can be displayed more intuitively, assisting in the assessment of the historical distribution of oil resources, analyzing historical patterns, and providing clearer decision-making conditions.

[0173] 5. Establish an intelligent recommendation algorithm for oil product scheduling to optimize delivery plans, reduce the manual workload of plan formulation, and improve the intelligence, scientific feasibility, and executability of plan formulation.

[0174] This invention also provides an intelligent oil resource scheduling device, as described in the following embodiments. Since the principle by which this device solves the problem is similar to that of the intelligent oil resource scheduling method, the implementation of this device can refer to the implementation of the intelligent oil resource scheduling method; repeated details will not be elaborated further.

[0175] Figure 5 This is a schematic diagram of an intelligent oil resource scheduling device in an embodiment of the present invention, such as... Figure 5 As shown, the device 500 includes:

[0176] The data acquisition module 501 is used to acquire oil depot data, vehicle data, and real-time operational data of multiple gas stations based on a streaming processing architecture. The oil depot data includes data reflecting the type of oil depot and the amount of oil stored in the oil depot. The vehicle data includes data reflecting the vehicle's qualifications, oil transportation attributes, and driving status. The real-time operational data includes sales data, inventory data, and oil product delivery data.

[0177] The intelligent recommendation module 502 is used to input oil depot data, vehicle data, and real-time operational data of multiple gas stations into a trained intelligent recommendation model, and output prediction information and delivery plan recommendation information for each gas station based on multiple business scenario identifiers. The business scenario identifiers reflect one of sales volume, inventory status, and delivery status. The prediction information includes time information and descriptive information. The delivery plan recommendation information includes oil depot information, vehicle information, delivery time information, and delivered fuel information. The intelligent recommendation model is trained in advance using historical oil depot data, vehicle data, gas station operational data, and the time and descriptive information of the business scenario identifiers to train an LSTM model.

[0178] In one embodiment, the data acquisition module 501 is specifically used for:

[0179] The streaming processing tool subscribes to oil depot data, vehicle data, and real-time operational data from multiple gas stations from a pre-configured distributed publish-subscribe messaging tool; the distributed publish-subscribe messaging tool is built on a Kafka cluster.

[0180] In one embodiment, the device 500 further includes: a data preprocessing module; the data preprocessing module is used for:

[0181] Before the intelligent recommendation module 502 inputs the oil depot data, vehicle data, and real-time operation data of multiple gas stations into the trained intelligent recommendation model, it performs data cleaning, normalization, and label encoding on the oil depot data, vehicle data, and real-time operation data of multiple gas stations to obtain multiple label encoding matrix data.

[0182] Dimensionality reduction is performed on multiple label encoding matrix data to obtain the dimensionality-reduced data;

[0183] The intelligent recommendation module 502 is specifically used for:

[0184] The dimensionality-reduced principal component data is input into the trained intelligent recommendation model.

[0185] In one embodiment, the data preprocessing module is specifically used to: apply matrix decomposition to multiple label coding matrix data to obtain multiple decomposed matrices; extract features based on the multiple decomposed matrices to obtain low-dimensional feature data; and perform principal component analysis (PCA) dimensionality reduction on the low-dimensional feature data to obtain dimensionality-reduced data.

[0186] In one embodiment, the intelligent recommendation model is trained as follows:

[0187] Collect historical data: oil depot data, vehicle data, and gas station operational data, as well as time and description information of business scenario identifiers, to form training and testing sets;

[0188] Construct an LSTM model; the LSTM model includes an input gate, a forget gate, and an output gate;

[0189] The LSTM model architecture is trained using the training set and tested using the test set to obtain a well-trained intelligent recommendation model.

[0190] In one embodiment, the business scenario identifier includes one or any combination of the following:

[0191] Sales are about to reach their peak; large customer groups are about to have large-scale tank and drum sales; inventory is expected to reach the average shipment volume; current inventory stability is poor; delivery volume is low; and delivery frequency is low.

[0192] In one embodiment, the delivery plan recommendation information is displayed in the form of a drop-down list, sorted from high to low recommendation level. Each list includes oil depot information, vehicle information, delivery time information, and delivery oil information.

[0193] In one embodiment, the device 500 further includes: a user-defined module;

[0194] The user-defined module is used to receive user selection instructions and / or modification instructions for delivery plan recommendations after the intelligent recommendation module 502 outputs prediction information and delivery plan recommendation information for each gas station with multiple business scenario identifiers, and to obtain the delivery plan recommendation information determined by the user; and to send the delivery plan recommendation information determined by the user to the oil depot terminal, gas station terminal and vehicle terminal.

[0195] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described intelligent scheduling method for oil resources.

[0196] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent scheduling method for oil resources.

[0197] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described intelligent scheduling method for oil resources.

[0198] This invention, based on a streaming processing architecture, acquires oil depot data, vehicle data, and real-time operational data from multiple gas stations, obtaining comprehensive and accurate dynamic information on oil resources. Then, it utilizes artificial intelligence algorithms, specifically an LSTM model, to predict the business scenario identifiers for each gas station and recommend delivery plans. This method, based on big data and intelligent models, can monitor and predict key characteristics of gas stations, avoiding manual scheduling plans and providing intelligent data recommendations for dispatchers. This provides more intelligent and automated support for scheduling plan decisions, significantly improving the effectiveness of oil resource scheduling.

[0199] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0200] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0201] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0202] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0203] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent scheduling of oil resources, characterized in that, include: Based on a streaming processing architecture, the system acquires real-time operational data from oil depots, vehicles, and multiple gas stations. Oil depot data includes information on oil depot type and storage capacity, while vehicle data includes information on vehicle qualifications, fuel transport attributes, and driving status. Real-time operational data includes sales data, inventory data, and fuel delivery data. The system inputs oil depot data, vehicle data, and real-time operational data from multiple gas stations into a trained intelligent recommendation model. The model outputs predictive information and delivery plan recommendations for each gas station based on multiple business scenario identifiers. These business scenario identifiers reflect one of three factors: sales volume, inventory status, or delivery status. The predictive information includes time and description information. The delivery plan recommendations include oil depot information, vehicle information, delivery time information, and delivered fuel information. The intelligent recommendation model is pre-trained using historical oil depot data, vehicle data, gas station operational data, and the time and description information of the business scenario identifiers to train a Long Short-Term Memory (LSTM) network model.

2. The intelligent scheduling method for oil resources as described in claim 1, characterized in that, Based on a streaming processing architecture, it acquires real-time operational data from oil depots, vehicles, and multiple gas stations, including: Based on streaming processing tools, data on oil depots, vehicles, and real-time operational data from multiple gas stations are subscribed to from a pre-configured distributed publish-subscribe messaging tool.

3. The intelligent scheduling method for oil resources as described in claim 1, characterized in that, Before inputting oil depot data, vehicle data, and real-time operational data from multiple gas stations into the trained intelligent recommendation model, the following steps are also included: Data cleaning, normalization, and tag coding were performed on oil depot data, vehicle data, and real-time operational data from multiple gas stations to obtain multiple tag coding matrix data. Dimensionality reduction is performed on multiple label encoding matrix data to obtain the dimensionality-reduced data; The data from oil depots, vehicles, and real-time operational data from multiple gas stations are input into a pre-trained intelligent recommendation model, including: The dimensionality-reduced principal component data is input into the trained intelligent recommendation model.

4. The intelligent scheduling method for oil resources as described in claim 3, characterized in that, Dimensionality reduction is performed on multiple label encoding matrix data to obtain the dimensionality-reduced data, including: Matrix decomposition is applied to multiple tag-encoded matrix data to obtain multiple decomposed matrices; Feature extraction is performed based on multiple decomposed matrices to obtain low-dimensional feature data; Principal component analysis (PCA) is used to reduce the dimensionality of the low-dimensional feature data, resulting in dimensionality-reduced data.

5. The intelligent scheduling method for oil resources as described in claim 1, characterized in that, The intelligent recommendation model is trained as follows: Collect historical data: oil depot data, vehicle data, and gas station operational data, as well as time and description information of business scenario identifiers, to form training and testing sets; Construct an LSTM model; the LSTM model includes an input gate, a forget gate, and an output gate; The LSTM model architecture is trained using the training set and tested using the test set to obtain a well-trained intelligent recommendation model.

6. The intelligent scheduling method for oil resources as described in claim 1, characterized in that, The business scenario identifier includes one or any combination of the following: Sales are about to reach their peak; large customer groups are about to have large-scale tank and drum sales; inventory is expected to reach the average shipment volume; current inventory stability is poor; delivery volume is low; and delivery frequency is low.

7. The intelligent scheduling method for oil resources as described in claim 1, characterized in that, The delivery plan recommendation information is displayed in a drop-down list, sorted from highest to lowest recommendation level. Each list includes information on the oil depot, vehicle, delivery time, and delivered fuel.

8. The intelligent scheduling method for oil resources as described in claim 7, characterized in that, After outputting the predicted information and delivery plan recommendation information for each gas station based on multiple business scenario identifiers, it also includes: Receive user's selection instruction for delivery plan recommendation information, and / or modification instruction, and obtain the delivery plan recommendation information determined by the user; The delivery plan recommendation information determined by the user is sent to the oil depot terminal, gas station terminal and vehicle terminal.

9. An intelligent oil resource scheduling device, characterized in that, include: The data acquisition module is used to acquire oil depot data, vehicle data, and real-time operational data from multiple gas stations based on a streaming processing architecture. Oil depot data includes data reflecting the type of oil depot and its storage capacity; vehicle data includes data reflecting vehicle qualifications, oil transportation attributes, and driving status; real-time operational data includes sales data, inventory data, and oil delivery data. The intelligent recommendation module is used to input oil depot data, vehicle data, and real-time operational data from multiple gas stations into a trained intelligent recommendation model, and output prediction information and delivery plan recommendation information for each gas station based on multiple business scenario identifiers. The business scenario identifiers reflect one of the following: sales volume, inventory status, and delivery status. The prediction information includes time information and descriptive information. The delivery plan recommendation information includes oil depot information, vehicle information, delivery time information, and delivered fuel information. The intelligent recommendation model is trained in advance using historical oil depot data, vehicle data, gas station operational data, and the time and descriptive information of the business scenario identifiers to train an LSTM model.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the intelligent scheduling method for oil resources as described in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the intelligent scheduling method for oil resources according to any one of claims 1 to 8.

12. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the intelligent scheduling method for oil resources as described in any one of claims 1 to 8.

Images