Artificial intelligence-based production resource dynamic allocation management method and system

Abstract

The embodiments of the present invention relate to the field of artificial intelligence technology, and specifically to a method and system for dynamic allocation management of production resources based on artificial intelligence. The embodiments of the present invention propose a production resource scheduling method that combines prior learning annotations, business operation data mining networks, and long short-term memory networks. This method not only makes full use of prior knowledge to accelerate and optimize the learning process of machine learning models, but also captures the timing characteristics of business operation data through long short-term memory networks, and realizes accurate calculation and dynamic update of production resource scheduling perspectives. This method significantly improves the accuracy and efficiency of production scheduling of LNG receiving stations, and provides strong support for the operation and management of receiving stations.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and more specifically, to an artificial intelligence-based production resource dynamic allocation management method and system. Background Art

[0002] In the business operation of LNG receiving stations, effective scheduling of production resources is the key to ensure efficient and stable operation of the receiving stations. However, due to the complexity and variability of the operating environment of the receiving stations, traditional scheduling methods based on rules or experience are often difficult to adapt to various emergencies and changes in operational needs. Therefore, a production resource scheduling method that can automatically learn and adapt to the dynamic changes in the operation of the receiving station is needed.

[0003] In recent years, with the rapid development of machine learning technology, production scheduling using data mining and machine learning methods has become a research hotspot. However, most existing methods only focus on extracting patterns from a large amount of historical operation data, while ignoring the importance of prior knowledge. In practical applications, prior knowledge can often provide valuable initial guidance for machine learning models, accelerate their learning process and improve accuracy.

[0004] In addition, since most of the operational data of LNG receiving stations have time series characteristics, traditional machine learning models often have difficulty capturing the long-term dependencies in these data. This leads to large errors in the model when predicting future operational needs and production resource scheduling. Therefore, a data mining method that can fully consider time series characteristics is needed to improve the accuracy and adaptability of production scheduling. Summary of the invention

[0005] In view of this, the present invention provides a method and system for dynamic allocation management of production resources based on artificial intelligence.

[0006] An embodiment of the present invention provides a production resource dynamic allocation management method based on artificial intelligence, which is applied to a production resource dynamic allocation management system, and the method includes: obtaining a first business operation data example and a second business operation data example, wherein the first business operation data example is a receiving station business operation data example carrying a priori learning annotation; inputting the first business operation data example into a business operation data mining network to obtain a first production resource scheduling viewpoint of the first business operation data example; determining a first resource scheduling decision error based on the first production resource scheduling viewpoint and the priori learning annotation; inputting the second business operation data example into the business operation data mining network to obtain a second production resource scheduling viewpoint of the second business operation data example; updating the second production resource scheduling viewpoint through a long short-term memory network to obtain an updated second production resource scheduling viewpoint, wherein the long short-term memory network has a data mining function for derived business operation data; determining a second resource scheduling decision error based on the updated second production resource scheduling viewpoint; and debugging the business operation data mining network based on the first resource scheduling decision error and the second resource scheduling decision error.

[0007] In some technical schemes, the second production resource scheduling viewpoint is updated through the long short-term memory network to obtain an updated second production resource scheduling viewpoint, including: taking the second business operation data example as the business operation data of the receiving station to be mined, and taking the second production resource scheduling viewpoint as a scheduling update guide, and outputting the updated second production resource scheduling viewpoint through the long short-term memory network.

[0008] In some technical solutions, the second business operation data example is ship loading and unloading business operation data, the second business operation data example is used as the receiving station business operation data to be mined, and the second production resource scheduling viewpoint is used as the scheduling update guide, and the updated second production resource scheduling viewpoint is output through the long short-term memory network, including:

[0009] Extracting a plurality of local business operation data of three attention dimensions from the second business operation data example, wherein the three attention dimensions correspond to three resource categories of the ship loading and unloading business operation data;

[0010] Taking the multiple local business operation data of the three attention dimensions as the business operation data of the receiving station to be mined, and taking the second production resource scheduling viewpoint as the scheduling update guide, identifying the derived scheduling viewpoints under the three attention dimensions through the long short-term memory network, wherein the derived scheduling viewpoint under each attention dimension includes the phased production resource scheduling viewpoints of the multiple local business operation data belonging to the same attention dimension;

[0011] The derived scheduling opinions under the three attention dimensions are aggregated using an opinion updating thread to output the updated second production resource scheduling opinion.

[0012] In some technical solutions, the adopting viewpoint update thread aggregates the derived scheduling viewpoints under the three attention dimensions and outputs the updated second production resource scheduling viewpoint, including:

[0013] For any one of the business operation messages in the second business operation data example, on the basis that the derived scheduling viewpoints under the three attention dimensions all determine that the business operation message belongs to the first scheduling category, determining that the business operation message belongs to the first scheduling category; on the basis that there is a derived scheduling viewpoint under at least one attention dimension that believes that the business operation message does not belong to the first scheduling category, determining that the business operation message belongs to the second scheduling category;

[0014] The updated second production resource scheduling viewpoint is outputted according to the scheduling category determination information of each business operation message in the second business operation data example.

[0015] In some technical solutions, the adopting viewpoint update thread aggregates the derived scheduling viewpoints under the three attention dimensions and outputs the updated second production resource scheduling viewpoint, including:

[0016] For any one of the business operation messages in the second business operation data example, on the basis that the derived scheduling viewpoints under at least two of the three attention dimensions determine that the business operation message belongs to the first scheduling category, it is determined that the business operation message belongs to the first scheduling category; on the basis that the derived scheduling viewpoints under at least two of the three attention dimensions believe that the business operation message belongs to the second scheduling category, it is determined that the business operation message belongs to the second scheduling category;

[0017] The updated second production resource scheduling viewpoint is outputted according to the scheduling category determination information of each business operation message in the second business operation data example.

[0018] In some technical solutions, the LNG storage business operation data of the three attention dimensions is used as the business operation data of the receiving station to be mined, and the second production resource scheduling viewpoint is used as the scheduling update guide, and the derivative scheduling viewpoints under the three attention dimensions are identified through the long short-term memory network, including:

[0019] Determining, based on the second production resource scheduling viewpoint, a local resource allocation viewpoint of the plurality of local business operation data of the three attention dimensions, the three attention dimensions comprising a first attention dimension, a second attention dimension, and a third attention dimension;

[0020] Using the multiple local business operation data of the first attention dimension as the business operation data of the receiving station to be mined, and using the local resource allocation viewpoints of the multiple local business operation data of the first attention dimension as the scheduling update guide, identifying the derived scheduling viewpoints under the first attention dimension through the long short-term memory network;

[0021] Using the multiple local business operation data of the second attention dimension as the business operation data of the receiving station to be mined, and using the local resource allocation viewpoints of the multiple local business operation data of the second attention dimension as the scheduling update guide, identifying the derived scheduling viewpoints under the second attention dimension through the long short-term memory network;

[0022] Taking multiple local business operation data of the third attention dimension as the business operation data of the receiving station to be mined, and taking the local resource allocation viewpoints of the multiple local business operation data of the third attention dimension as the scheduling update guide, the derived scheduling viewpoints under the third attention dimension are identified through the long short-term memory network.

[0023] In some technical solutions, the method further includes: performing disturbance processing on the local resource allocation viewpoint.

[0024] In some technical solutions, the business operation data mining network includes a mid-stage side neural network and an edge side neural network; the first production resource scheduling viewpoint is a production resource scheduling viewpoint generated by the edge side neural network;

[0025] Determining a first resource scheduling decision error based on the first production resource scheduling viewpoint and the prior learning annotation includes: determining the first resource scheduling decision error of the edge-side neural network based on the first production resource scheduling viewpoint and the prior learning annotation.

[0026] In some technical solutions, the business operation data mining network includes a middle platform side neural network and an edge side neural network; the second production resource scheduling viewpoint includes a global production resource scheduling viewpoint generated by the middle platform side neural network and an edge production resource scheduling viewpoint generated by the edge side neural network, and the updated second production resource scheduling viewpoint includes an updated global production resource scheduling viewpoint;

[0027] Determining the second resource scheduling decision error based on the updated second production resource scheduling viewpoint includes: determining the second resource scheduling decision error of the edge side neural network based on the edge production resource scheduling viewpoint and the updated global production resource scheduling viewpoint.

[0028] In some technical solutions, determining the second resource scheduling decision error of the edge-side neural network based on the edge production resource scheduling viewpoint and the updated global production resource scheduling viewpoint includes:

[0029] Determining a resource scheduling decision error of a derived scheduling viewpoint based on the edge production resource scheduling viewpoint and the updated global production resource scheduling viewpoint; and determining a resource scheduling joint decision error based on the edge production resource scheduling viewpoint and the global production resource scheduling viewpoint;

[0030] The fusion result between the resource scheduling decision error of the derived scheduling viewpoint and the resource scheduling joint decision error is determined as the second resource scheduling decision error of the edge-side neural network.

[0031] In some technical solutions, the step of determining the fusion result between the resource scheduling decision error of the derived scheduling viewpoint and the resource scheduling joint decision error as the second resource scheduling decision error of the edge-side neural network includes:

[0032] Determine a first confidence coefficient and a second confidence coefficient based on a current number of debugging cycles and a maximum number of debugging cycles of the debugging process;

[0033] Determine a first error-weighted result of the first confidence coefficient and the resource scheduling decision error of the derived scheduling viewpoint, and a second error-weighted result of the second confidence coefficient and the resource scheduling joint decision error;

[0034] The second resource scheduling decision error of the edge-side neural network is determined by combining the first error weighted result and the second error weighted result.

[0035] In some technical solutions, the debugging process includes multi-cycle debugging, and the business operation data mining network includes a mid-stage side neural network and an edge side neural network;

[0036] The debugging of the business operation data mining network according to the first resource scheduling decision error and the second resource scheduling decision error includes:

[0037] According to the first resource scheduling decision error and the second resource scheduling decision error, optimizing the algorithm weight of the edge-side neural network in the u-th debugging cycle, where u is a positive integer;

[0038] The algorithm weight of the neural network on the middle platform side is optimized according to the algorithm weight of the neural network on the edge side before the u-th debugging cycle and the algorithm weight of the u-th debugging cycle.

[0039] In some technical solutions, the debugging process includes multi-cycle debugging, and the method further includes:

[0040] Based on the second production resource scheduling viewpoint of the second business operation data example in the u-th debugging cycle, filtering the business operation data example to be processed from the second business operation data example, where u is a positive integer;

[0041] Obtain the prior learning annotation configured for the business operation data example to be processed, obtain the first business operation data example directly screened and output in the u-th debugging cycle, and add the first business operation data example to the first business operation data example set used in the u+1-th debugging cycle.

[0042] In some technical solutions, the second production resource scheduling viewpoint based on the second business operation data example in the u-th debugging cycle, screening the business operation data example to be processed from the second business operation data example, includes:

[0043] According to the second production resource scheduling viewpoint of the second business operation data example in the u-th debugging cycle, the second business operation data example that meets the set requirements is selected as the business operation data example to be processed;

[0044] The second business operation data example that meets the set requirements includes at least one of the following:

[0045] first X1 second business operation data examples in descending order of the credibility scores of the viewpoints that rely on the second production resource scheduling viewpoint;

[0046] The first X2 second business operation data examples arranged in descending order of a global timing error depending on the second production resource scheduling perspective, the global timing error being a response delay error of all business operation messages in the second production resource scheduling perspective;

[0047] First X3 second business operation data examples arranged in descending order based on the frequent item coefficient of the second production resource scheduling viewpoint, wherein the frequent item coefficient is the proportion of business operation messages greater than a set discrimination value in the second production resource scheduling viewpoint to all business operation messages;

[0048] Among them, X1, X2 and X3 are all positive integers.

[0049] In some technical schemes, the first business operation data example is ship loading and unloading business operation data, and the prior learning annotation of the first business operation data example is configured using linkage annotation rules; the linkage annotation rules are dynamic rules that annotate the local resource allocation perspectives of the LNG storage business operation data in the three attention dimensions in the ship loading and unloading business operation data respectively; wherein the LNG storage business operation data in each of the three attention dimensions is a set of local business operation data among multiple sets of local business operation data belonging to the current attention dimension.

[0050] The present invention also provides a production resource dynamic allocation management system, comprising: a memory for storing program instructions and data; a processor for coupling with the memory and executing instructions in the memory to implement the above method.

[0051] The present invention also provides a computer storage medium, comprising instructions, and when the instructions are executed on a processor, the above method is implemented.

[0052] The embodiment of the present invention proposes a production resource scheduling method that combines prior learning annotations, business operation data mining networks and long short-term memory networks. This method not only makes full use of prior knowledge to accelerate and optimize the learning process of machine learning models, but also captures the time series characteristics in business operation data through long short-term memory networks, and realizes accurate calculation and dynamic update of production resource scheduling perspectives. This method significantly improves the accuracy and efficiency of production scheduling of LNG receiving stations, and provides strong support for the operation and management of receiving stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0054] Figure 1 A schematic flowchart of the steps of a method for dynamic allocation management of production resources based on artificial intelligence is provided in an embodiment of the present invention.

[0055] Figure 2 A structural block diagram of a production resource dynamic allocation and management system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The technical solution of the present invention will be described below in conjunction with the accompanying drawings.

[0057] When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention. It should be noted that the terms "first", "second", etc. in the specification of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0058] The method embodiments provided in the embodiments of the present invention can be executed in a production resource dynamic allocation management system, a computer device or a similar computing device. Taking running on a production resource dynamic allocation management system as an example, the production resource dynamic allocation management system may include one or more processors (the processor may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory for storing data. Optionally, the above-mentioned production resource dynamic allocation management system may also include a transmission device for communication functions. It can be understood by those of ordinary skill in the art that the above-mentioned structure is only illustrative and does not limit the structure of the above-mentioned production resource dynamic allocation management system. For example, the production resource dynamic allocation management system may also include more or fewer components than those shown above, or have a configuration different from that shown above.

[0059] The memory can be used to store computer programs, for example, software programs and modules of application software, such as a computer program corresponding to a method for dynamic allocation and management of production resources based on artificial intelligence in an embodiment of the present invention. The processor executes various functional applications and data processing by running the computer program stored in the memory, that is, implementing the above method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the dynamic allocation and management system of production resources via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0060] The transmission device is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the production resource dynamic allocation management system. In one example, the transmission device includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device can be a radio frequency (Radio Frequency, referred to as RF) module, which is used to communicate with the Internet wirelessly.

[0061] In conjunction with the above, please refer to Figure 1 , Figure 1 It is a flow chart of a production resource dynamic allocation management method based on artificial intelligence provided by an embodiment of the present invention. The method is applied to a production resource dynamic allocation management system and may further include steps 110 to 140.

[0062] In the LNG receiving station production scheduling management system construction project, the production resource dynamic allocation management system, as the core execution entity, undertakes the important task of realizing intelligent operation management. The following will introduce the implementation process of the following steps in detail through a specific application scenario example.

[0063] Step 110: Obtain a first business operation data example and a second business operation data example, wherein the first business operation data example is a receiving station business operation data example carrying a priori learning annotations.

[0064] First, the dynamic allocation and management system of production resources needs to obtain two sets of business operation data examples: the first business operation data example and the second business operation data example. The first business operation data example and the second business operation data example are both derived from the actual operation process of the LNG receiving station, and contain detailed information on the entire process of ship loading and unloading, LNG storage, and gas-liquid transmission at the receiving station. In particular, the first business operation data example carries prior learning annotations, which means that the first business operation data example has undergone certain preprocessing and has been annotated with specific information or patterns so that subsequent data mining and machine learning processes can use this prior knowledge for more effective learning. In other words, the first business operation data example can be understood as a labeled example, and the second business operation data example can be understood as an unlabeled example.

[0065] Step 120: input the first business operation data example into a business operation data mining network to obtain a first production resource scheduling viewpoint of the first business operation data example; and determine a first resource scheduling decision error based on the first production resource scheduling viewpoint and the prior learning annotation.

[0066] Next, the production resource dynamic allocation management system inputs the first business operation data example into the business operation data mining network. The business operation data mining network is a pre-built machine learning model that is specifically used to mine valuable information and patterns from business operation data to guide the scheduling of production resources. Through the calculation of the business operation data mining network, the production resource dynamic allocation management system can obtain the first production resource scheduling view of the first business operation data example, that is, how the business operation data mining network believes that production resources should be scheduled to optimize the operation of the receiving station.

[0067] Then, the production resource dynamic allocation management system will determine the first resource scheduling decision error based on the first production resource scheduling viewpoint and the prior learning annotation. The first resource scheduling decision error reflects the difference between the scheduling viewpoint given by the network and the actual annotation (i.e., prior knowledge), and is an important indicator for measuring the learning effect of the network.

[0068] Step 130: input the second business operation data example into the business operation data mining network to obtain a second production resource scheduling viewpoint of the second business operation data example; update the second production resource scheduling viewpoint through the long short-term memory network to obtain an updated second production resource scheduling viewpoint, the long short-term memory network having a data mining function for derived business operation data; determine a second resource scheduling decision error based on the updated second production resource scheduling viewpoint.

[0069] Subsequently, the production resource dynamic allocation management system also inputs the second business operation data example into the business operation data mining network to obtain the second production resource scheduling viewpoint. However, unlike the first business operation data example, this time the production resource dynamic allocation management system will also update the second production resource scheduling viewpoint through the long short-term memory network. The long short-term memory network is a special recurrent neural network with data mining capabilities for time series data (i.e., derived business operation data). It can capture long-term dependencies in the data, thereby helping the production resource dynamic allocation management system to better understand the dynamic changes in the receiving station operation and update the scheduling viewpoint of production resources accordingly.

[0070] The updated second production resource scheduling viewpoint is closer to the actual operation demand, and the production resource dynamic allocation management system will determine the second resource scheduling decision error based on the updated viewpoint. The second resource scheduling decision error also reflects the learning effect of the business operation data mining network on the new data examples.

[0071] Step 140: Debug the business operation data mining network according to the first resource scheduling decision error and the second resource scheduling decision error.

[0072] Finally, the production resource dynamic allocation management system will debug the business operation data mining network based on the first resource scheduling decision error and the second resource scheduling decision error. The purpose of debugging is to optimize the parameters and structure of the business operation data mining network so that it can better give accurate production resource scheduling views on future data. For example, the production resource dynamic allocation management system can reduce the decision error and improve the performance of the business operation data mining network by adjusting the weight of the business operation data mining network, changing the structure of the business operation data mining network, or adding more training data.

[0073] During this process, the production resource dynamic allocation management system will continue to collect new data from the actual operation of the LNG receiving station and use this data as feedback to further optimize the data mining network. Through this continuous learning and optimization, the production resource dynamic allocation management system will be able to gradually adapt to the various complex situations of the receiving station operation and achieve more intelligent and efficient production scheduling.

[0074] In order to illustrate the implementation process of the above technical solution in more detail, an actual application scenario is introduced below, which involves ship loading and unloading and storage capacity management at an LNG receiving station.

[0075] In this application scenario, the first business operation data example contains the loading and unloading records of multiple LNG ships in the past period of time, including the arrival time, unloading volume, unloading speed and other information of the ships, as well as the ideal storage capacity balance state and scheduling strategy marked in the prior learning annotations. When these data are input into the business operation data mining network, the business operation data mining network will try to learn how to predict the future ship loading and unloading needs from these data, and formulate a reasonable storage capacity management strategy accordingly.

[0076] The second example of business operation data may be the actual operation data in the recent period, which may contain some new patterns or abnormal situations, such as delayed arrival of ships, fluctuations in unloading speed, etc. When these data are input into the network, the long short-term memory network will help the production resource dynamic allocation management system capture these new patterns and update the scheduling view of production resources to adapt to changes in actual operations.

[0077] Through continuous learning and optimization, the production resource dynamic allocation management system will be able to gradually master the operating rules of the LNG receiving station and achieve more accurate and efficient ship loading and unloading and storage capacity management. This will help improve the overall operating efficiency and management level of the receiving station, reduce operating costs, and provide customers with better services.

[0078] It is worth mentioning that step 130, as the core invention of the embodiment of the present invention, introduces the process of how to use the business operation data mining network and the long short-term memory network to work together to process and update the production resource scheduling viewpoint. This process not only reflects the application of advanced data mining and machine learning technology, but also fully demonstrates its role in optimizing the production scheduling of LNG receiving stations.

[0079] First, the second business operation data example is input into the business operation data mining network. This step is the basis of data-driven decision-making and relies on a large amount of business data collected from the actual operation of the LNG receiving terminal. These data examples cover all aspects of the receiving terminal, including detailed information on the entire process of ship loading and unloading, LNG storage, and gas and liquid transmission. The business operation data mining network is a carefully designed and trained machine learning model that can extract valuable information and patterns from these complex data. When the second business operation data example is input into the network, the network begins its calculation process, analyzing various features and associations in the data to form a preliminary view on the scheduling of production resources. This preliminary view, the second production resource scheduling view, is derived by the network based on the current input data and its learned knowledge. It represents how the network believes that production resources should be scheduled to optimize the operation of the receiving terminal under the current operating conditions. However, this view is only based on the current input data and the network's existing knowledge, and may not fully take into account the dynamic changes in the operation of the receiving terminal.

[0080] Therefore, the second production resource scheduling viewpoint needs to be updated through the long short-term memory network. The long short-term memory network is a special recurrent neural network that has a strong ability to mine time series data. In the operation of LNG receiving stations, a lot of data is time series data, which records the changes in the operating status of the receiving station over time. The long short-term memory network can capture the long-term dependencies in these data, that is, the association between current data points and past data points. This association is crucial to understanding the dynamic changes in the operation of the receiving station.

[0081] The second production resource scheduling viewpoint is further optimized through the update of the LSTM network. The LSTM network takes into account more historical information and time series data to more accurately predict future operational needs and production resource scheduling. This updated viewpoint is closer to actual operational needs and can better cope with various complex situations in receiving station operations.

[0082] The production resource dynamic allocation management system then determines the second resource scheduling decision error based on the updated second production resource scheduling viewpoint. This error is an important indicator for measuring the learning effect of the network on new data examples. By calculating this error, the production resource dynamic allocation management system can evaluate the difference between the scheduling viewpoint given by the network and the actual operational needs. This difference may be caused by a variety of reasons, including data noise, network imperfections, changes in the operating environment, etc. Regardless of the reason, the error provides valuable feedback information for further adjusting and optimizing the parameters and structure of the network.

[0083] For example, the production resource dynamic allocation management system will use this error to debug the business operation data mining network and the long short-term memory network. The purpose of debugging is to reduce the error and improve the network's ability to give accurate production resource scheduling views on future data. This may involve a variety of strategies such as adjusting the network's weights, changing the network's structure, and adding more training data. Through continuous debugging and optimization, the network will gradually adapt to the various complex situations of the receiving station operation and achieve more intelligent and efficient production scheduling.

[0084] Therefore, step 130 realizes the accurate calculation and dynamic update of the production resource scheduling viewpoint by combining the business operation data mining network and the long short-term memory network. This method not only improves the accuracy and efficiency of production scheduling, but also enhances the adaptability of the production resource dynamic allocation management system to the dynamic changes in the operation of the receiving station. In future development, this method is expected to be further applied to a wider range of fields, providing strong support for the optimization of various complex production resource dynamic allocation management systems. In the production scheduling management of LNG receiving stations, the implementation of this technical solution is expected to significantly improve the overall operating efficiency and management level of the receiving station, and provide customers with a better service experience.

[0085] In summary, the embodiment of the present invention provides an initial learning benchmark for the business operation data mining network by introducing the first business operation data example with prior learning annotations, so that the business operation data mining network can start learning from data with clear guiding significance, and then form a preliminary production resource scheduling viewpoint. This design not only accelerates the training process of the business operation data mining network, but also improves the accuracy of its understanding of business operation data. By calculating the first resource scheduling decision error, the performance of the network in the initial learning stage can be quantified, providing a clear basis for subsequent network debugging and optimization.

[0086] Furthermore, the embodiment of the present invention uses the long short-term memory network to update the production resource scheduling viewpoint of the second business operation data example, fully considering the time series characteristics of the business operation data, so that the updated production resource scheduling viewpoint is closer to the actual operation needs, and the accuracy and adaptability of the scheduling are improved. By determining the second resource scheduling decision error, the learning effect of the network on the new data is quantified again, providing strong support for further network debugging.

[0087] Finally, the embodiment of the present invention debugs the business operation data mining network based on two resource scheduling decision errors. This two-stage error feedback mechanism ensures that the network can achieve good performance on different data examples, improving the generalization ability and stability of the business operation data mining network. In summary, by combining prior learning annotations, business operation data mining networks and long short-term memory networks, accurate calculation and dynamic update of production resource scheduling views are achieved, significantly improving the accuracy and efficiency of production scheduling.

[0088] In some optional embodiments, the updating of the second production resource scheduling viewpoint through the long short-term memory network to obtain an updated second production resource scheduling viewpoint includes: taking the second business operation data example as the receiving station business operation data to be mined, and taking the second production resource scheduling viewpoint as a scheduling update guide, and outputting the updated second production resource scheduling viewpoint through the long short-term memory network.

[0089] In this embodiment, the second production resource scheduling viewpoint is updated through the long short-term memory network to obtain the updated second production resource scheduling viewpoint, and this process involves the time series data processing technology of deep learning and the specific application of business operation data. As a special recurrent neural network (RNN), the long short-term memory network (LSTM) has shown excellent performance in processing time series data and is particularly suitable for mining business operation data with time series characteristics such as LNG receiving stations.

[0090] First of all, it should be clear that the second business operation data example plays the role of the business operation data of the receiving station to be mined in the embodiment of the present invention. These data are collected from the actual operation of the LNG receiving station, and contain detailed information on the entire process of ship loading and unloading, LNG storage, gas and liquid transmission, etc. These data are time-series, which means that they record the changes in the operating status of the receiving station over time. In order to update the second production resource scheduling viewpoint, the production resource dynamic allocation management system needs to input these time-series data into the long short-term memory network to allow the network to learn the patterns and associations therein.

[0091] Secondly, the second production resource scheduling viewpoint serves as a scheduling update guide, providing an initial scheduling viewpoint for the LSTM network. This viewpoint is based on the preliminary analysis of the second business operation data example by the business operation data mining network. However, since it is only based on current data and the existing knowledge of the network, it may not fully take into account the dynamic changes in the receiving station operation. Therefore, the production resource dynamic allocation management system needs to further update and optimize this viewpoint through the LSTM network.

[0092] During the LSTM update process, the LSTM receives as input a second business operation data example and a second production resource scheduling viewpoint. The network analyzes the temporal characteristics of this data, that is, the association between current data points and past data points. This association is critical to understanding the dynamic changes in receiving station operations. By capturing these long-term dependencies, the network is able to better predict future operational needs and production resource scheduling.

[0093] For example, the long short-term memory network uses its internal memory units and gating mechanisms to process input time series data. The memory units can store and update the state information of the network, while the gating mechanism is responsible for controlling the inflow and outflow of information. These mechanisms enable the network to maintain long-term dependencies when processing time series data and effectively avoid the problem of gradient disappearance or explosion.

[0094] During the processing, the network will gradually update its internal state and generate a new production resource scheduling view based on the current state and input data. This new view is derived after considering more historical information and time series data, so it is closer to actual operational needs. By comparing with the second production resource scheduling view, the production resource dynamic allocation management system can obtain the updated second production resource scheduling view.

[0095] Finally, the updated second production resource scheduling viewpoint will be used as the output of the long short-term memory network. This viewpoint is obtained after fully considering the time series characteristics and historical information, so it is more accurate and reliable. It can serve as an important reference for LNG receiving station production scheduling and help operators make more informed decisions.

[0096] In this way, the process of updating the second production resource scheduling viewpoint through the long short-term memory network is a complex process that combines time series data processing and business operation data mining. It fully utilizes the advantages of the long short-term memory network in processing time series data and the ability of the business operation data mining network in extracting valuable information and patterns. Through this process, the production resource dynamic allocation management system can obtain a more accurate and reliable production resource scheduling viewpoint, providing strong support for the operation management of the LNG receiving station.

[0097] In the next step, the second business operation data example is the ship loading and unloading business operation data, and the second business operation data example is used as the receiving station business operation data to be mined, and the second production resource scheduling viewpoint is used as the scheduling update guide, and the updated second production resource scheduling viewpoint is output through the long short-term memory network, including: extracting multiple local business operation data of three attention dimensions from the second business operation data example, and the three attention dimensions correspond to the three resource categories of the ship loading and unloading business operation data; using the multiple local business operation data of the three attention dimensions as the receiving station business operation data to be mined, and using the second production resource scheduling viewpoint as the scheduling update guide, identifying the derived scheduling viewpoints under the three attention dimensions through the long short-term memory network, and the derived scheduling viewpoints under each attention dimension include the phased production resource scheduling viewpoints of multiple local business operation data belonging to the same attention dimension; using the viewpoint update thread to aggregate the derived scheduling viewpoints under the three attention dimensions, and outputting the updated second production resource scheduling viewpoint.

[0098] Based on this embodiment, it involves integrating the second business operation data example, i.e., ship loading and unloading business operation data, and the second production resource scheduling viewpoint into the long short-term memory network to output the updated second production resource scheduling viewpoint. This process not only involves data processing and network training, but also includes the introduction of attention mechanism and scheduling viewpoint update strategy.

[0099] First, the production resource dynamic allocation management system extracts multiple local business operation data of three attention dimensions from the second business operation data example. These three attention dimensions closely correspond to the three resource categories of ship loading and unloading business operation data, such as representing different resource types, different operation links or different time windows. Through the attention mechanism, the production resource dynamic allocation management system can focus more on the key information in these data, thereby more accurately capturing the complexity and dynamics of ship loading and unloading business operation data.

[0100] The extracted local business operation data not only contains rich detailed information, but also reflects the operation status of ship loading and unloading business under different resource categories. These data, as the business operation data of the receiving station to be mined, provide a solid foundation for the subsequent update of scheduling viewpoints.

[0101] Secondly, the production resource dynamic allocation management system takes multiple local business operation data of these three attention dimensions and the second production resource scheduling view as input, and conducts deep mining through the long short-term memory network. With its excellent time series data processing capabilities, the long short-term memory network can capture the long-term dependencies in these data, thereby more accurately predicting future operation needs and production resource scheduling.

[0102] During the training process of the network, the production resource dynamic allocation management system uses the second production resource scheduling viewpoint as a scheduling update guide to help the network better understand and learn the scheduling rules of ship loading and unloading business. This guidance not only provides an initial scheduling viewpoint for the network, but also provides a valuable reference for its subsequent learning process.

[0103] Through deep mining of the long short-term memory network, the production resource dynamic allocation management system can identify derived scheduling views under three attention dimensions. The derived scheduling views under each attention dimension contain the staged production resource scheduling views of multiple local business operation data belonging to the same attention dimension. These derived scheduling views not only reflect the operational requirements under different resource categories, but also show how these requirements change over time.

[0104] Finally, the production resource dynamic allocation management system uses the viewpoint update thread to aggregate the derived scheduling views under these three attention dimensions, thereby outputting an updated second production resource scheduling viewpoint. The viewpoint update thread is an effective scheduling viewpoint integration strategy that can comprehensively consider the scheduling views under different attention dimensions and make reasonable trade-offs. Through this strategy, the production resource dynamic allocation management system can obtain a more comprehensive and accurate updated production resource scheduling viewpoint.

[0105] In this way, by introducing the attention mechanism and long short-term memory network, the deep mining and effective utilization of ship loading and unloading business operation data are realized. By extracting local business operation data, identifying derived scheduling viewpoints and aggregating updated scheduling viewpoints, the production resource dynamic allocation management system can more accurately predict and schedule production resources, thereby improving the overall operation efficiency and management level of the LNG receiving station.

[0106] In other optional embodiments, the viewpoint update thread is used to aggregate the derived scheduling viewpoints under the three attention dimensions to output the updated second production resource scheduling viewpoint, including: for any business operation message in the second business operation data example, on the basis that the derived scheduling viewpoints under the three attention dimensions all determine that the business operation message belongs to the first scheduling category, determining that the business operation message belongs to the first scheduling category; on the basis that there is at least one derived scheduling viewpoint under the attention dimension that believes that the business operation message does not belong to the first scheduling category, determining that the business operation message belongs to the second scheduling category; outputting the updated second production resource scheduling viewpoint based on the scheduling category determination information of each business operation message in the second business operation data example.

[0107] In this embodiment, a viewpoint update thread is used to aggregate the derived scheduling viewpoints under the three attention dimensions, and finally output the updated second production resource scheduling viewpoint involving the detailed classification and judgment of business operation messages, and forming a comprehensive scheduling viewpoint based on these judgment results.

[0108] First, the core task of the opinion update thread is to integrate the derived scheduling opinions under different attention dimensions, which are mined through the long short-term memory network based on the second business operation data example, that is, the ship loading and unloading business operation data. Each attention dimension corresponds to a specific resource category in the ship loading and unloading business, and the derived scheduling opinions reflect the staged production resource scheduling requirements under these resource categories.

[0109] During the aggregation process, the production resource dynamic allocation management system first considers any business operation message in the second business operation data example. For such a message, if the derived scheduling viewpoints under the three attention dimensions all determine that it belongs to the first scheduling category, then the production resource dynamic allocation management system can determine that the message does belong to the first scheduling category. The "first scheduling category" in the embodiment of the present invention can represent a high priority or urgent scheduling requirement that requires immediate attention and processing.

[0110] However, in actual applications, the dynamic allocation and management system of production resources often encounter more complex situations. If there is a derivative scheduling view under at least one attention dimension that believes that a business operation message does not belong to the first scheduling category, then the dynamic allocation and management system of production resources cannot simply classify it as the first scheduling category. At this time, the dynamic allocation and management system of production resources needs to consider the scheduling requirements of the message more carefully and determine it as the second scheduling category. The "second scheduling category" in the embodiment of the present invention can represent a scheduling requirement with a relatively low priority or that can be processed later.

[0111] Through such a judgment process, the production resource dynamic allocation management system can more accurately understand the scheduling requirements of each business operation message and classify it into the appropriate scheduling category. This provides an important basis for the production resource dynamic allocation management system to form a comprehensive scheduling viewpoint in the future.

[0112] Next, the production resource dynamic allocation management system needs to output the updated second production resource scheduling viewpoint based on the scheduling category determination information of each business operation message in the second business operation data example. This process involves the integration and analysis of a large number of business operation messages, as well as the weighing and selection of different scheduling categories.

[0113] For example, the production resource dynamic allocation management system will count the number of business operation messages, priority and other information under each scheduling category, and form a comprehensive scheduling view based on this information. For example, if the number of business operation messages under the first scheduling category is large and the priority is high, then the production resource dynamic allocation management system needs to pay more attention to the scheduling needs of these messages when forming a scheduling view to ensure that they can be processed in a timely manner.

[0114] At the same time, the production resource dynamic allocation management system will not ignore the business operation messages under the second scheduling category. Although their priority may be relatively low, they also need to be reasonably scheduled and processed. The production resource dynamic allocation management system will arrange the scheduling order and time nodes of these messages according to the actual situation to ensure that they can be properly processed without affecting the processing of high-priority messages.

[0115] Finally, through the aggregation and analysis of the viewpoint update thread, the production resource dynamic allocation management system can output an updated second production resource scheduling viewpoint. This viewpoint comprehensively considers the derived scheduling viewpoints under different attention dimensions, the scheduling category determination information of business operation messages, and the actual operation needs and resource conditions. It can provide more accurate and comprehensive production resource scheduling guidance for the production resource dynamic allocation management system, helping LNG receiving stations achieve more efficient and stable operations.

[0116] In this way, using the viewpoint update thread to aggregate the derived scheduling viewpoints under the three attention dimensions and output the updated second production resource scheduling viewpoint is a complex and meticulous process. It requires the production resource dynamic allocation management system to deeply understand the operation rules and scheduling requirements of the ship loading and unloading business, accurately grasp the relationship and differences between the derived scheduling viewpoints under different attention dimensions, and flexibly respond to challenges and changes in various actual operation scenarios. Through such a technical solution, the production resource dynamic allocation management system can better mine and utilize the valuable information in the ship loading and unloading business operation data, and provide more powerful and intelligent support for the production resource scheduling of the LNG receiving station.

[0117] Under some other optional design ideas, the viewpoint update thread aggregates the derived scheduling viewpoints under the three attention dimensions and outputs the updated second production resource scheduling viewpoint, including: for any business operation message in the second business operation data example, on the basis that the derived scheduling viewpoints under at least two of the three attention dimensions determine that the business operation message belongs to the first scheduling category, determining that the business operation message belongs to the first scheduling category; on the basis that the derived scheduling viewpoints under at least two of the three attention dimensions believe that the business operation message belongs to the second scheduling category, determining that the business operation message belongs to the second scheduling category; outputting the updated second production resource scheduling viewpoint based on the scheduling category determination information of each business operation message in the second business operation data example.

[0118] Based on this design idea, it involves using the viewpoint update thread to aggregate the derived scheduling viewpoints under the three attention dimensions, and finally output the updated second production resource scheduling viewpoint. This design idea emphasizes the importance of comprehensive judgment of business operation messages under multiple attention dimensions to ensure the accuracy and comprehensiveness of scheduling viewpoints.

[0119] First, the core task of the opinion update thread is to integrate the derived scheduling opinions under different attention dimensions. These opinions are mined through the long short-term memory network based on the second business operation data example, that is, the ship loading and unloading business operation data. Each attention dimension corresponds to a specific resource category in the ship loading and unloading business, and the derived scheduling opinions reflect the staged production resource scheduling requirements under these resource categories.

[0120] During the aggregation process, the production resource dynamic allocation management system first considers any business operation message in the second business operation data example. For such a message, if the derived scheduling viewpoints under at least two of the three attention dimensions determine that it belongs to the first scheduling category, then the production resource dynamic allocation management system can determine that the message does belong to the first scheduling category. The "first scheduling category" here may represent a high priority or urgent scheduling requirement that requires immediate attention and processing. This judgment method reflects the application of the majority principle, that is, among the judgment results of multiple attention dimensions, if at least two dimensions believe that a business operation message belongs to the first scheduling category, then the production resource dynamic allocation management system should classify it as the first scheduling category.

[0121] However, in actual applications, the dynamic allocation and management system of production resources will also encounter more complicated situations. If the derived scheduling viewpoints under at least two of the three attention dimensions believe that a business operation message belongs to the second scheduling category, then the dynamic allocation and management system of production resources needs to determine it as the second scheduling category. The "second scheduling category" here may represent a scheduling requirement with a relatively low priority or that can be processed later. This judgment method also reflects the application of the majority principle, that is, in the judgment results of multiple attention dimensions, if at least two dimensions believe that a business operation message belongs to the second scheduling category, then the dynamic allocation and management system of production resources should classify it as the second scheduling category.

[0122] Through such a judgment process, the production resource dynamic allocation management system can more accurately understand the scheduling requirements of each business operation message and classify it into the appropriate scheduling category. This provides an important basis for the production resource dynamic allocation management system to form a comprehensive scheduling viewpoint in the future.

[0123] Next, the production resource dynamic allocation management system needs to output the updated second production resource scheduling viewpoint based on the scheduling category determination information of each business operation message in the second business operation data example. This process involves the integration and analysis of a large number of business operation messages, as well as the weighing and selection of different scheduling categories.

[0124] For example, the dynamic allocation and management system of production resources will count the number and priority of business operation messages under each scheduling category, and form a comprehensive scheduling view based on this information. For example, if the number of business operation messages under the first scheduling category is large and the priority is high, then the dynamic allocation and management system of production resources needs to pay more attention to the scheduling needs of these messages when forming a scheduling view to ensure that they can be processed in a timely manner. At the same time, the dynamic allocation and management system of production resources will also consider the business operation messages under the second scheduling category. Although their priority may be relatively low, they also need to be reasonably scheduled and processed. The dynamic allocation and management system of production resources will arrange the scheduling order and time nodes of these messages according to the actual situation to ensure that they can be properly processed without affecting the processing of high-priority messages.

[0125] When forming a comprehensive scheduling view, the production resource dynamic allocation management system also needs to consider the weight relationship between different attention dimensions. Since each attention dimension corresponds to a specific resource category in the ship loading and unloading business, the weights between different dimensions may not be the same. For example, some resource categories may be more critical or scarce, so their corresponding attention dimensions should have higher weights when forming a scheduling view. The production resource dynamic allocation management system will adjust the weight relationship between different attention dimensions according to actual conditions to ensure that the formed scheduling view is more in line with actual operational needs and resource conditions.

[0126] Finally, through the aggregation and analysis of the viewpoint update thread, the production resource dynamic allocation management system can output an updated second production resource scheduling viewpoint. This viewpoint comprehensively considers the derived scheduling viewpoints under different attention dimensions, the scheduling category determination information of business operation messages, and the actual operation needs and resource conditions. It can provide more accurate and comprehensive production resource scheduling guidance for the production resource dynamic allocation management system, helping LNG receiving stations achieve more efficient and stable operations.

[0127] In this way, it is a complex and meticulous process to use the viewpoint update thread to aggregate the derived scheduling viewpoints under the three attention dimensions and output the updated second production resource scheduling viewpoint. It requires the production resource dynamic allocation management system to have a deep understanding of the operating rules and scheduling requirements of the ship loading and unloading business, accurately grasp the relationship and difference between the derived scheduling viewpoints under different attention dimensions, and flexibly respond to the challenges and changes in various actual operation scenarios. At the same time, the production resource dynamic allocation management system also needs to consider the weight relationship between different attention dimensions to ensure that the formed scheduling viewpoint is more in line with the actual operation needs and resource conditions. Through such a technical solution, the production resource dynamic allocation management system can better mine and utilize the valuable information in the ship loading and unloading business operation data, and provide more powerful and intelligent support for the production resource scheduling of the LNG receiving station.

[0128] In a preferred embodiment, the LNG storage business operation data of the three attention dimensions are used as the business operation data of the receiving station to be mined, and the second production resource scheduling viewpoint is used as the scheduling update guide, and the derivative scheduling viewpoints under the three attention dimensions are identified through the long short-term memory network, including: according to the second production resource scheduling viewpoint, determining the local resource allocation viewpoints of multiple local business operation data of the three attention dimensions, the three attention dimensions include the first attention dimension, the second attention dimension and the third attention dimension; using the multiple local business operation data of the first attention dimension as the business operation data of the receiving station to be mined, and using the local resource allocation viewpoints of multiple local business operation data of the first attention dimension as The scheduling update guide identifies the derived scheduling viewpoints under the first attention dimension through the long short-term memory network; uses the multiple local business operation data of the second attention dimension as the business operation data of the receiving station to be mined, and the local resource allocation viewpoints of the multiple local business operation data of the second attention dimension as the scheduling update guide, and identifies the derived scheduling viewpoints under the second attention dimension through the long short-term memory network; uses the multiple local business operation data of the third attention dimension as the business operation data of the receiving station to be mined, and the local resource allocation viewpoints of the multiple local business operation data of the third attention dimension as the scheduling update guide, and identifies the derived scheduling viewpoints under the third attention dimension through the long short-term memory network.

[0129] It can be understood that the core goal of this embodiment is to identify and mine the derived scheduling viewpoints in the LNG storage business operation data under three different attention dimensions through the long short-term memory network (LSTM). This process requires not only the production resource dynamic allocation management system to process the original data in a refined manner, but also to be able to guide the generation of new scheduling strategies based on the existing production resource scheduling viewpoints.

[0130] First, the three attention dimensions are defined as the first attention dimension, the second attention dimension, and the third attention dimension. These three dimensions represent different perspectives or focuses for analyzing the LNG storage business operation data, which may cover multiple aspects such as time series analysis, spatial distribution characteristics, and specific links of business operations. Through such multi-dimensional division, the production resource dynamic allocation management system can understand and mine the hidden information in the business operation data more comprehensively and deeply.

[0131] Next, the embodiment requires determining the local resource allocation viewpoints of multiple local business operation data under these three attention dimensions based on the second production resource scheduling viewpoint. This step is a key prerequisite for realizing the identification of subsequent derived scheduling viewpoints. As a priori knowledge or guiding principle, the second production resource scheduling viewpoint can help the production resource dynamic allocation management system better understand and interpret business operation data, especially on the core issue of resource allocation. By applying it to local business operation data in three attention dimensions, the production resource dynamic allocation management system can form a more specific and targeted set of local resource allocation viewpoints, providing strong support for subsequent data mining and scheduling strategy updates.

[0132] Subsequently, the embodiment entered the core data processing and mining stage. It requires the production resource dynamic allocation management system to use multiple local business operation data under each attention dimension as the receiving station business operation data to be mined, and combine the local resource allocation view under this dimension as the scheduling update guide, and identify the corresponding derivative scheduling view through the long short-term memory network.

[0133] Specifically, when processing the first attention dimension, the production resource dynamic allocation management system regards the local business operation data under this dimension as the target to be mined, and uses its corresponding local resource allocation viewpoint as a guide to perform deep learning and analysis through the LSTM network. As a special recurrent neural network (RNN), LSTM is good at processing sequence data and can capture long-term dependencies. Therefore, it is very suitable for mining the temporal dynamics and potential laws in business operation data, thereby helping the production resource dynamic allocation management system to identify the derived scheduling viewpoints under the first attention dimension.

[0134] Similarly, the production resource dynamic allocation management system also adopts the same processing method for the second and third attention dimensions. That is, the local business operation data under these two dimensions are used as the receiving station business operation data to be mined, and combined with the local resource allocation viewpoints under each dimension, data mining and identification of derived scheduling viewpoints are performed through the LSTM network.

[0135] It is worth noting that in this process, the role of the LSTM network is not only data analysis and pattern recognition, but also a powerful tool that can help the production resource dynamic allocation management system combine the prior production resource scheduling view with the actual business operation data, thereby mining richer and more in-depth derivative scheduling view. These derivative viewpoints not only reflect the inherent laws of business operation data, but also incorporate the production resource dynamic allocation management system's understanding and expectations of production resource scheduling, so they have higher practical value and guiding significance.

[0136] In this way, through refined data processing and in-depth data mining, a comprehensive analysis of LNG storage business operation data and the identification of derived scheduling viewpoints under three different attention dimensions are achieved. It not only improves the production resource dynamic allocation management system's ability to understand and utilize business operation data, but also provides new ideas and methods for the optimization and innovation of production resource scheduling. In future practice, the production resource dynamic allocation management system can further explore and optimize the application scenarios and algorithm models of this embodiment, in order to achieve more accurate and efficient data-driven decision-making in a wider range of business fields.

[0137] Based on the above content, the method also includes: performing disturbance processing on the local resource allocation viewpoint.

[0138] It can be understood that the introduction of perturbation processing on the local resource allocation viewpoint brings new possibilities and diversity to the generation of derived scheduling viewpoints. Perturbation processing can be regarded as a kind of "fine-tuning" or "variation" of the local resource allocation viewpoint, which explores more scheduling strategy space by introducing small random changes or deviations on the basis of the original viewpoint. Such processing not only helps the production resource dynamic allocation management system to discover scheduling strategies that may be overlooked by traditional methods, but also improves the robustness and adaptability of scheduling strategies. In the face of complex and changing business operation environments, such perturbation processing can make the scheduling strategies of the production resource dynamic allocation management system more flexible and diverse, so as to better cope with various challenges and uncertainties.

[0139] Specifically, disturbance handling can be implemented in a variety of ways. For example, the production resource dynamic allocation management system can make small random adjustments to certain parameters in the local resource allocation view, or introduce some new constraints to change the way resources are allocated. These disturbances can be directed or undirected, depending on the characteristics and scope of the scheduling strategy space that the production resource dynamic allocation management system wants to explore. By continuously performing disturbance handling and observing its impact on the derived scheduling view, the production resource dynamic allocation management system can gradually discover more optimized and effective scheduling strategies.

[0140] In addition, disturbance processing can be combined with the training process of the LSTM network to form an iterative optimization method. The production resource dynamic allocation management system can make small disturbances to the local resource allocation viewpoints in each iteration and identify new derivative scheduling views through the LSTM network. Through such continuous iteration and optimization, the production resource dynamic allocation management system can gradually approach the optimal scheduling strategy while maintaining the diversity and robustness of the strategy.

[0141] In this way, through refined data processing, in-depth data mining and disturbance processing of local resource allocation views, a comprehensive analysis of LNG storage business operation data and identification of derived scheduling views under three different attention dimensions are achieved. It not only improves the production resource dynamic allocation management system's ability to understand and utilize business operation data, but also provides new ideas and methods for the optimization and innovation of production resource scheduling.

[0142] In some other possible embodiments, the business operation data mining network includes a middle-side neural network and an edge-side neural network; the first production resource scheduling viewpoint is a production resource scheduling viewpoint generated by the edge-side neural network; and determining a first resource scheduling decision error based on the first production resource scheduling viewpoint and the prior learning annotations includes: determining the first resource scheduling decision error of the edge-side neural network based on the first production resource scheduling viewpoint and the prior learning annotations.

[0143] Based on this embodiment, we first need to have a clear understanding of the overall business operation data mining network. The business operation data mining network is composed of the middle platform side neural network and the edge side neural network, each of which has different roles and tasks. The middle platform side neural network may focus more on global data processing and analysis, while the edge side neural network is closer to the actual business operation environment and is responsible for generating production resource scheduling views that are closely related to specific business scenarios.

[0144] Furthermore, the first production resource scheduling viewpoint is a production resource scheduling viewpoint generated by the edge side neural network. This means that the viewpoint is formed based on the analysis and understanding of the actual business operation data by the edge side neural network, so it is more real-time and targeted.

[0145] Next, “prior learning annotations” can be understood as a pre-existing resource scheduling strategy or rule based on historical experience or expert knowledge. They serve as a reference standard to evaluate the quality and accuracy of production resource scheduling insights generated by edge-side neural networks.

[0146] First of all, the production resource dynamic allocation management system needs to clarify the definition of the first resource scheduling decision error. Simply put, it is the difference or deviation between the production resource scheduling viewpoint generated by the edge neural network and the prior learning annotation. This difference may come from many aspects, such as inaccuracy in data processing, insufficient model training, complexity of the actual business environment, etc. By quantifying this difference, the production resource dynamic allocation management system can more intuitively understand the performance of the edge neural network in practical applications and further guide its optimization and improvement. The determination of this difference involves a reasonable evaluation mechanism. In this mechanism, the production resource dynamic allocation management system first compares the first production resource scheduling viewpoint generated by the edge neural network with the prior learning annotation in detail. This comparison may involve multiple dimensions, such as the efficiency of resource allocation, the timeliness of scheduling, and the adaptability to business fluctuations. Through comparison, the production resource dynamic allocation management system can find the differences between the two and try to analyze the causes of these differences.

[0147] Secondly, the dynamic allocation management system of production resources needs to quantify the differences. This can be achieved by designing some specific quantitative indicators, such as error rate, degree of deviation, etc. These indicators can help the dynamic allocation management system of production resources more accurately measure the difference between the first production resource scheduling view and the prior learning annotation, and provide clear guidance for subsequent optimization work.

[0148] In the process of quantifying differences, the production resource dynamic allocation management system also needs to consider some practical factors. For example, the complexity of the business operation environment may make some differences inevitable; or, in certain specific cases, the production resource scheduling views generated by the edge neural network may be more reasonable and effective than the prior learning annotations. Therefore, when quantifying differences, the production resource dynamic allocation management system needs to consider a variety of factors to ensure the objectivity and accuracy of the evaluation results.

[0149] Finally, based on the quantitative difference results, the production resource dynamic allocation management system can further analyze the performance of the edge neural network in actual applications and try to put forward suggestions for optimization and improvement. This optimization may involve the improvement of data processing algorithms, the adjustment of model structure, the increase of training data, and other aspects. Through continuous optimization and improvement, the production resource dynamic allocation management system can expect the edge neural network to generate more accurate and effective production resource scheduling views, thereby providing more powerful support for the business operations of enterprises.

[0150] In this way, the production resource dynamic allocation management system can better understand the performance of edge-side neural networks in practical applications and provide clear guidance for their optimization and improvement. In future practice, the production resource dynamic allocation management system can further explore and optimize the application scenarios and algorithm models of this embodiment, in order to achieve more accurate and efficient data-driven decision-making in a wider range of business areas.

[0151] In some other possible embodiments, the business operation data mining network includes a middle-end neural network and an edge-end neural network; the second production resource scheduling viewpoint includes a global production resource scheduling viewpoint generated by the middle-end neural network, and an edge production resource scheduling viewpoint generated by the edge-end neural network, and the updated second production resource scheduling viewpoint includes an updated global production resource scheduling viewpoint; determining a second resource scheduling decision error based on the updated second production resource scheduling viewpoint includes: determining the second resource scheduling decision error of the edge-end neural network based on the edge production resource scheduling viewpoint and the updated global production resource scheduling viewpoint.

[0152] Furthermore, the second resource scheduling decision error of the edge side neural network is determined based on the edge production resource scheduling viewpoint and the updated global production resource scheduling viewpoint, including: determining the resource scheduling decision error of the derived scheduling viewpoint based on the edge production resource scheduling viewpoint and the updated global production resource scheduling viewpoint; and determining the resource scheduling joint decision error based on the edge production resource scheduling viewpoint and the global production resource scheduling viewpoint; and determining the fusion result between the resource scheduling decision error of the derived scheduling viewpoint and the resource scheduling joint decision error as the second resource scheduling decision error of the edge side neural network.

[0153] Furthermore, the fusion result between the resource scheduling decision error of the derived scheduling viewpoint and the resource scheduling joint decision error is determined as the second resource scheduling decision error of the edge side neural network, including: determining a first confidence coefficient and a second confidence coefficient based on the current number of debugging cycles and the maximum number of debugging cycles of the debugging process; determining a first error weighted result of the first confidence coefficient and the resource scheduling decision error of the derived scheduling viewpoint, and a second error weighted result of the second confidence coefficient and the resource scheduling joint decision error; and determining the second resource scheduling decision error of the edge side neural network by combining the first error weighted result and the second error weighted result.

[0154] In this embodiment, the business operation data mining network is composed of the middle platform side neural network and the edge side neural network, each of which undertakes different tasks. The middle platform side neural network focuses more on global data processing and analysis, generating global production resource scheduling views; while the edge side neural network is closer to the actual business operation environment, generating edge production resource scheduling views.

[0155] Next, the updated second production resource scheduling viewpoint includes the updated global production resource scheduling viewpoint. This means that the global production resource scheduling viewpoint has been updated or optimized in some form to better reflect the current business operation environment and needs. The second resource scheduling decision error can be understood as the difference or deviation between the edge production resource scheduling viewpoint generated by the edge-side neural network and the updated global production resource scheduling viewpoint.

[0156] Then, according to the description of the embodiment, it mainly includes two parts of the second resource scheduling decision error: one is the resource scheduling decision error of the derived scheduling viewpoint, and the other is the resource scheduling joint decision error. The derived scheduling viewpoint can be understood as a new production resource scheduling viewpoint generated by the edge side neural network under some disturbance or change, while the resource scheduling joint decision error reflects the difference between the edge production resource scheduling viewpoint and the global production resource scheduling viewpoint.

[0157] In order to determine the second resource scheduling decision error, the production resource dynamic allocation management system first needs to calculate these two parts of the error separately. For the resource scheduling decision error of the derived scheduling viewpoint, the production resource dynamic allocation management system can calculate it by comparing the difference between the derived scheduling viewpoint and the updated global production resource scheduling viewpoint. This difference may involve multiple aspects such as the efficiency of resource allocation and the timeliness of scheduling. For the resource scheduling joint decision error, the production resource dynamic allocation management system needs to compare the difference between the marginal production resource scheduling viewpoint and the global production resource scheduling viewpoint, which may also involve multiple dimensions.

[0158] After calculating these two parts of the error, the production resource dynamic allocation management system needs to fuse them to determine the second resource scheduling decision error of the edge side neural network. The "fusion" here can be understood as a weighted summation process, in which the weight (i.e., the confidence coefficient) is determined based on the current number of debugging cycles and the maximum number of debugging cycles of the debugging process. Specifically, the production resource dynamic allocation management system can dynamically adjust the weights of these two parts of the error according to the progress of debugging. In the early stage of debugging, the production resource dynamic allocation management system may pay more attention to the resource scheduling decision error of the derived scheduling point of view, because it reflects the adaptability and learning ability of the edge side neural network to the new environment; in the later stage of debugging, the production resource dynamic allocation management system may pay more attention to the resource scheduling joint decision error, because it reflects the consistency and synergy between the edge side neural network and the global scheduling strategy.

[0159] Through such a fusion process, the production resource dynamic allocation management system can obtain a second resource scheduling decision error that comprehensively considers the derived scheduling viewpoint and the resource scheduling joint decision. This error not only reflects the performance of the edge neural network in the current business operation environment, but also provides a direction for further optimization and improvement of the production resource dynamic allocation management system. For example, if the resource scheduling decision error of the derived scheduling viewpoint is large, the production resource dynamic allocation management system may need to adjust the training strategy or model structure of the edge neural network to improve its adaptability to the new environment; if the resource scheduling joint decision error is large, the production resource dynamic allocation management system needs to consider how to adjust the global scheduling strategy or optimize the coordination mechanism between the edge neural network and the global scheduling strategy.

[0160] It can be understood that determining the second resource scheduling decision error based on the updated second production resource scheduling viewpoint not only requires the production resource dynamic allocation management system to design a reasonable error calculation mechanism and fusion strategy, but also requires the production resource dynamic allocation management system to comprehensively consider a variety of practical factors to ensure the accuracy and effectiveness of the error assessment. With such a design, the production resource dynamic allocation management system can better understand the performance of the edge-side neural network in the current business operation environment, and provide clear guidance for its optimization and improvement. In future practice, the production resource dynamic allocation management system can further explore and optimize the application scenarios and algorithm models of this embodiment, in order to achieve more accurate and efficient data-driven decision-making in a wider range of business fields.

[0161] In some independent embodiments, the debugging process includes multi-cycle debugging, and the business operation data mining network includes a middle-stage neural network and an edge-side neural network; debugging the business operation data mining network based on the first resource scheduling decision error and the second resource scheduling decision error includes: optimizing the algorithm weight of the edge-side neural network in the u-th debugging cycle based on the first resource scheduling decision error and the second resource scheduling decision error, where u is a positive integer; optimizing the algorithm weight of the middle-stage neural network based on the algorithm weight of the edge-side neural network before the u-th debugging cycle and the algorithm weight of the u-th debugging cycle.

[0162] This embodiment involves how to debug the business operation data mining network based on the first resource scheduling decision error and the second resource scheduling decision error. This process involves multi-cycle debugging, which means that the production resource dynamic allocation management system will gradually optimize and improve the entire network through multiple debugging cycles. The business operation data mining network is composed of the middle platform side neural network and the edge side neural network, which play different roles in the network.

[0163] First, the dynamic allocation and management system of production resources focuses on the optimization of the edge-side neural network. In the u-th debugging cycle (u is a positive integer), the dynamic allocation and management system of production resources adjusts the algorithm weights of the edge-side neural network based on the first resource scheduling decision error and the second resource scheduling decision error. These errors reflect the performance of the edge-side neural network in the current business operation environment, as well as its consistency and coordination with the global scheduling strategy. By adjusting the algorithm weights, the dynamic allocation and management system of production resources can improve the accuracy and efficiency of the edge-side neural network in scheduling production resources, so that it can better adapt to current business operation needs.

[0164] Specifically, the production resource dynamic allocation management system may use some advanced optimization algorithms, such as gradient descent method, Newton method, etc., to adjust the weights of the edge side neural network according to the error information. These algorithms can help the production resource dynamic allocation management system find the direction to minimize the error and gradually adjust the weights along this direction, thereby improving the performance of the edge side neural network.

[0165] Secondly, after optimizing the edge-side neural network, the production resource dynamic allocation management system turns its attention to the optimization of the middle-end neural network. The middle-end neural network is responsible for global data processing and analysis, and generates a global production resource scheduling view. In order to maintain the consistency and synergy of the entire business operation data mining network, the production resource dynamic allocation management system needs to optimize the algorithm weight of the middle-end neural network based on the algorithm weight of the edge-side neural network before the u-th debugging cycle and the algorithm weight of the u-th debugging cycle.

[0166] The goal of this optimization process is to ensure that the middle-end neural network can fully consider the changes of the edge-end neural network and adjust its global scheduling strategy accordingly. By optimizing the algorithm weights of the middle-end neural network, the production resource dynamic allocation management system can improve the accuracy and adaptability of the global scheduling strategy, so that it can better coordinate with the scheduling strategy of the edge-end neural network.

[0167] In the process of achieving this optimization, the production resource dynamic allocation management system may adopt some machine learning methods, such as reinforcement learning, transfer learning, etc. These methods can help the production resource dynamic allocation management system dynamically adjust the weights of the neural network on the middle platform side according to the changes of the neural network on the edge side, thereby achieving adaptive optimization of the global scheduling strategy.

[0168] Furthermore, the production resource dynamic allocation management system needs to consider how to continuously perform this optimization in multiple debugging cycles. In multi-cycle debugging, each debugging cycle will generate new first resource scheduling decision errors and second resource scheduling decision errors, and these error information will serve as the basis for optimization in the next debugging cycle. Through continuous iterative optimization, the production resource dynamic allocation management system can gradually reduce errors and improve the performance and accuracy of the entire business operation data mining network.

[0169] At the same time, the dynamic allocation and management system for production resources also needs to pay attention to the stability and convergence during the debugging process. During multiple debugging cycles, the dynamic allocation and management system for production resources needs to ensure that the performance of the entire network can gradually stabilize and converge to an acceptable range. In order to achieve this goal, the dynamic allocation and management system for production resources may use some regularization methods, learning rate adjustment strategies, etc. to control the stability and convergence of the optimization process.

[0170] In this way, through multi-cycle debugging and continuous iterative optimization, the production resource dynamic allocation management system can gradually improve the performance and accuracy of the entire network and provide more powerful support for the company's business operations.

[0171] In some other independent embodiments, the debugging process includes multi-cycle debugging, and the method further includes: based on the second production resource scheduling viewpoint of the second business operation data example in the u-th debugging cycle, filtering the business operation data example to be processed from the second business operation data example, where u is a positive integer; obtaining the prior learning annotation configured for the business operation data example to be processed, obtaining the first business operation data example directly filtered and output in the u-th debugging cycle, and adding the first business operation data example to the first business operation data example set used in the u+1-th debugging cycle.

[0172] Further, the second production resource scheduling viewpoint based on the second business operation data example in the u-th debugging cycle, screening the business operation data examples to be processed from the second business operation data examples, includes: based on the second production resource scheduling viewpoint of the second business operation data example in the u-th debugging cycle, screening the second business operation data examples that meet the set requirements as the business operation data examples to be processed; wherein the second business operation data examples that meet the set requirements include at least one of the following: the first X1 second business operation data examples arranged in descending order of the viewpoint credibility score depending on the second production resource scheduling viewpoint; the first X2 second business operation data examples arranged in descending order of the global timing error depending on the second production resource scheduling viewpoint, the global timing error is the response delay error of all business operation messages in the second production resource scheduling viewpoint; the first X3 second business operation data examples arranged in descending order of the frequent item coefficient depending on the second production resource scheduling viewpoint, the frequent item coefficient is the proportion of business operation messages greater than the set discrimination value in all business operation messages in the second production resource scheduling viewpoint; wherein X1, X2 and X3 are all positive integers.

[0173] Based on this embodiment, it can be seen that the debugging process is a multi-cycle process, which means that the production resource dynamic allocation management system will gradually optimize and improve the entire business operation data mining network through multiple debugging cycles. In each debugging cycle, the production resource dynamic allocation management system will screen the business operation data examples to be processed that need to be focused on in the next cycle based on the second production resource scheduling viewpoint of the second business operation data example of the current cycle.

[0174] Specifically, the process of selecting the business operation data examples to be processed is a process that comprehensively considers multiple factors. The production resource dynamic allocation management system selects the second business operation data examples that meet the set requirements as the business operation data examples to be processed based on the second production resource scheduling viewpoint. These set requirements include multiple dimensions such as viewpoint credibility score, global time series error, and frequent item coefficient.

[0175] The credibility score of an opinion is an important indicator for measuring the credibility of the second production resource scheduling opinion. The production resource dynamic allocation management system will sort the opinions in descending order according to the credibility score and select the first X1 second business operation data examples as the business operation data examples to be processed. In this way, the production resource dynamic allocation management system can ensure that those opinions with higher credibility receive more attention and processing.

[0176] The global timing error is an indicator that reflects the response delay error of all business operation messages in the second production resource scheduling viewpoint. By calculating the global timing error, the production resource dynamic allocation management system can understand the delay of the business operation message during processing. The production resource dynamic allocation management system selects the first X2 second business operation data examples in descending order of global timing error as the business operation data examples to be processed, so as to focus on analyzing and optimizing those messages with large response delays.

[0177] The frequent item coefficient is an indicator to measure the proportion of business operation messages greater than the set discrimination value in the second production resource scheduling viewpoint to all business operation messages. By calculating the frequent item coefficient, the production resource dynamic allocation management system can understand which business operation messages occupy an important position in the second production resource scheduling viewpoint. The production resource dynamic allocation management system selects the first X3 second business operation data examples in descending order of frequent item coefficients as the business operation data examples to be processed, so as to conduct a more in-depth analysis and processing of those frequently appearing messages.

[0178] Through such a screening process, the production resource dynamic allocation management system can obtain a set of targeted business operation data examples to be processed. These data examples not only reflect the important characteristics and problems of business operations in the current debugging cycle, but also provide optimization and improvement directions for the production resource dynamic allocation management system.

[0179] Next, the production resource dynamic allocation management system needs to configure prior learning annotations for these business operation data examples to be processed. Prior learning annotations are a type of learning guidance information based on prior knowledge or experience, which can help the production resource dynamic allocation management system better understand and process the business operation data examples to be processed. By obtaining prior learning annotations, the production resource dynamic allocation management system can obtain the first business operation data examples directly screened and output in the u-th debugging cycle. These first business operation data examples are the focus of the next step of optimization and improvement of the production resource dynamic allocation management system.

[0180] Finally, the production resource dynamic allocation management system adds these first business operation data examples to the first business operation data example set used in the u+1th debugging cycle. In this way, the production resource dynamic allocation management system can continue to conduct in-depth analysis and processing of these important data examples in the next debugging cycle, thereby achieving continuous optimization and improvement of the business operation data mining network.

[0181] This design provides a more targeted and effective debugging method for the production resource dynamic allocation management system through the screening process based on the second production resource scheduling viewpoint and the configuration of prior learning annotations. This method can not only help the production resource dynamic allocation management system better understand and process business operation data examples, but also achieve continuous optimization and improvement of the business operation data mining network. In future practice, the production resource dynamic allocation management system can further explore and optimize the application scenarios and algorithm models of this embodiment, in order to achieve more accurate and efficient data-driven decision-making in a wider range of business fields.

[0182] In some other independent embodiments, the first business operation data example is ship loading and unloading business operation data, and the prior learning annotation of the first business operation data example is configured using linkage annotation rules; the linkage annotation rules are dynamic rules that annotate the local resource allocation perspectives of the LNG storage business operation data of the three attention dimensions in the ship loading and unloading business operation data respectively; wherein the LNG storage business operation data of each of the three attention dimensions is a group of local business operation data among multiple groups of local business operation data belonging to the current attention dimension.

[0183] In this embodiment, the application scenario focuses on the specific scenario of ship loading and unloading business operation data, involving how to configure prior learning annotations for it, and the application of linkage annotation rules in this process. Ship loading and unloading business is a key link in logistics transportation and port operations. The effective processing and analysis of its data is of great significance to improving business efficiency and optimizing resource allocation.

[0184] First, the first business operation data example in the embodiment of the present invention specifically refers to the ship loading and unloading business operation data. This type of data covers information on multiple links such as ship entry, cargo loading and unloading, storage, and transshipment, and is an important basis for reflecting the port's operating status and efficiency. In order to more effectively mine and utilize this type of data, the production resource dynamic allocation management system configures prior learning annotations for it.

[0185] The configuration of prior learning annotations is not random, but based on a set of carefully designed linkage annotation rules. The core of this set of rules is to annotate the local resource allocation viewpoints of the LNG storage business operation data in the three attention dimensions of the ship loading and unloading business operation data. These three attention dimensions are the three aspects that the production resource dynamic allocation management system pays special attention to when analyzing the ship loading and unloading business. They each contain multiple sets of local business operation data, and the LNG storage business operation data is one of these local data.

[0186] Furthermore, the three attention dimensions can be three key aspects closely related to the ship loading and unloading business, such as the ship's arrival time, cargo loading and unloading efficiency, and LNG storage and transshipment. Taking the LNG storage business operation data as an example, it may include information such as LNG storage volume, storage time, and transshipment frequency, which is crucial for evaluating the port's LNG handling capacity.

[0187] The role of linkage annotation rules is to configure annotations of local resource allocation views for these LNG storage business operation data. These annotations may include opinions and evaluations on whether the current storage volume is sufficient, whether the storage time is reasonable, whether the transfer frequency is efficient, etc. Through these annotations, the production resource dynamic allocation management system can more intuitively understand the operating status of the LNG storage business in the ship loading and unloading business, as well as possible resource allocation problems.

[0188] In actual application, the configuration of linkage annotation rules is a dynamic process. Because the ship loading and unloading business is an ongoing process, its data is constantly updated and changed. Therefore, the production resource dynamic allocation management system needs to dynamically adjust and optimize the linkage annotation rules according to the latest business operation data to ensure the accuracy and effectiveness of the annotations.

[0189] As for the dynamic configuration process, the production resource dynamic allocation management system can use machine learning algorithms to automatically identify and classify different attention dimensions and local business operation data in the ship loading and unloading business operation data. Then, based on these classification results and preset annotation templates, the production resource dynamic allocation management system can automatically generate local resource allocation viewpoint annotations for each set of local business operation data.

[0190] In addition, to ensure the accuracy and reliability of the annotations, the production resource dynamic allocation management system also needs to manually review and verify the generated annotations. This step is essential because although machine learning algorithms perform well in data processing and analysis, they may still be affected by factors such as data quality and algorithmic bias and produce errors. Through manual review and verification, the production resource dynamic allocation management system can detect and correct these errors in a timely manner to ensure the quality and reliability of the annotations.

[0191] In this way, by configuring prior learning annotations for the ship loading and unloading business operation data and providing a set of dynamic configuration methods based on linkage annotation rules, a more effective and accurate data processing method is provided for the production resource dynamic allocation management system. This method can not only help the production resource dynamic allocation management system to better understand and analyze the operation status and resource allocation of the ship loading and unloading business, but also provide targeted optimization and improvement suggestions for the production resource dynamic allocation management system. At the same time, the production resource dynamic allocation management system can also try to combine this method with other data analysis technologies, such as data mining, deep learning, etc., to build a more complete and powerful data analysis platform to provide more powerful support for the business development of enterprises.

[0192] Further, Figure 2 The structural block diagram of the production resource dynamic allocation management system 300 is shown, including: a memory 310 for storing program instructions and data; a processor 320 for coupling with the memory 310 and executing the instructions in the memory 310 to implement the above method.

[0193] Furthermore, a computer storage medium is provided, comprising instructions, which implement the above method when executed on a processor.

[0194] In several embodiments provided in the embodiments of the present invention, it should be understood that the disclosed apparatus and method can also be implemented in other ways. The apparatus and method embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the apparatus, method and computer program product according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0195] In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0196] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code. It should be noted that in this article, the term "include", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such process, method, article or device. Without more constraints, an element defined by the phrase "comprising a..." does not exclude the existence of other identical elements in the process, method, article or apparatus comprising the element.

[0197] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A production resource dynamic allocation management method based on artificial intelligence, characterized in that: The method is applied to a production resource dynamic allocation management system, and the method comprises: Obtain a first business operation data example and a second business operation data example, wherein the first business operation data example is a receiving station business operation data example carrying a priori learning annotation; wherein the first business operation data example includes the loading and unloading records of multiple LNG ships in the past period of time, including the arrival time, unloading volume, unloading speed information of the ships, and the ideal storage capacity balance state and scheduling strategy marked in the priori learning annotation; the second business operation data example is the actual operation data in the current period of time, including the delayed arrival of the ship and the fluctuation of the unloading speed; Inputting the first business operation data example into a business operation data mining network to obtain a first production resource scheduling viewpoint for the first business operation data example; determining a first resource scheduling decision error based on the first production resource scheduling viewpoint and the prior learning annotation; wherein the first production resource scheduling viewpoint represents: how to schedule production resources to optimize the operation of the receiving station; Inputting the second business operation data example into the business operation data mining network to obtain a second production resource scheduling viewpoint of the second business operation data example; updating the second production resource scheduling viewpoint through a long short-term memory network to obtain an updated second production resource scheduling viewpoint, wherein the long short-term memory network has a data mining function for derived business operation data; determining a second resource scheduling decision error based on the updated second production resource scheduling viewpoint; wherein, when the second business operation data example is input into the business operation data mining network, the business operation data mining network analyzes various features and associations in the data to form a preliminary viewpoint on production resource scheduling, which is the second production resource scheduling viewpoint; The business operation data mining network is debugged according to the first resource scheduling decision error and the second resource scheduling decision error.

2. The method according to claim 1, characterized in that The method of updating the second production resource scheduling viewpoint through the long short-term memory network to obtain an updated second production resource scheduling viewpoint includes: taking the second business operation data example as the business operation data of the receiving station to be mined, and taking the second production resource scheduling viewpoint as a scheduling update guide, and outputting the updated second production resource scheduling viewpoint through the long short-term memory network.

3. The method according to claim 2, characterized in that The second business operation data example is ship loading and unloading business operation data, the second business operation data example is used as the receiving station business operation data to be mined, and the second production resource scheduling viewpoint is used as the scheduling update guide, and the updated second production resource scheduling viewpoint is output through the long short-term memory network, including: Extracting a plurality of local business operation data of three attention dimensions from the second business operation data example, wherein the three attention dimensions correspond to three resource categories of the ship loading and unloading business operation data; Taking the multiple local business operation data of the three attention dimensions as the business operation data of the receiving station to be mined, and taking the second production resource scheduling viewpoint as the scheduling update guide, identifying the derived scheduling viewpoints under the three attention dimensions through the long short-term memory network, wherein the derived scheduling viewpoint under each attention dimension includes the phased production resource scheduling viewpoints of the multiple local business operation data belonging to the same attention dimension; The derived scheduling opinions under the three attention dimensions are aggregated using an opinion updating thread to output the updated second production resource scheduling opinion.

4. The method according to claim 3, characterized in that The adoption viewpoint update thread aggregates the derived scheduling viewpoints under the three attention dimensions and outputs the updated second production resource scheduling viewpoint, including: For any one of the business operation messages in the second business operation data example, on the basis that the derived scheduling viewpoints under the three attention dimensions all determine that the business operation message belongs to the first scheduling category, determining that the business operation message belongs to the first scheduling category; on the basis that there is a derived scheduling viewpoint under at least one attention dimension that believes that the business operation message does not belong to the first scheduling category, determining that the business operation message belongs to the second scheduling category; The updated second production resource scheduling viewpoint is outputted according to the scheduling category determination information of each business operation message in the second business operation data example.

5. The method according to claim 3, characterized in that: The adoption viewpoint update thread aggregates the derived scheduling viewpoints under the three attention dimensions and outputs the updated second production resource scheduling viewpoint, including: For any one of the business operation messages in the second business operation data example, on the basis that the derived scheduling viewpoints under two of the three attention dimensions determine that the business operation message belongs to the first scheduling category, it is determined that the business operation message belongs to the first scheduling category; on the basis that the derived scheduling viewpoints under two of the three attention dimensions believe that the business operation message belongs to the second scheduling category, it is determined that the business operation message belongs to the second scheduling category; The updated second production resource scheduling viewpoint is outputted according to the scheduling category determination information of each business operation message in the second business operation data example.

6. The method according to claim 3, characterized in that The method uses the LNG storage business operation data of the three attention dimensions as the business operation data of the receiving station to be mined, and uses the second production resource scheduling viewpoint as the scheduling update guide, and identifies the derived scheduling viewpoints under the three attention dimensions through the long short-term memory network, including: Determining, based on the second production resource scheduling viewpoint, a local resource allocation viewpoint of the plurality of local business operation data of the three attention dimensions, the three attention dimensions comprising a first attention dimension, a second attention dimension, and a third attention dimension; Using the multiple local business operation data of the first attention dimension as the business operation data of the receiving station to be mined, and using the local resource allocation viewpoints of the multiple local business operation data of the first attention dimension as the scheduling update guide, identifying the derived scheduling viewpoints under the first attention dimension through the long short-term memory network; Using the multiple local business operation data of the second attention dimension as the business operation data of the receiving station to be mined, and using the local resource allocation viewpoints of the multiple local business operation data of the second attention dimension as the scheduling update guide, identifying the derived scheduling viewpoints under the second attention dimension through the long short-term memory network; Using the multiple local business operation data of the third attention dimension as the business operation data of the receiving station to be mined, and using the local resource allocation viewpoints of the multiple local business operation data of the third attention dimension as the scheduling update guide, identifying the derived scheduling viewpoints under the third attention dimension through the long short-term memory network; The method further includes: performing a disturbance process on the local resource allocation view.

7. The method according to any one of claims 1 to 6, characterized in that: The business operation data mining network includes a middle platform side neural network and an edge side neural network; the first production resource scheduling viewpoint is a production resource scheduling viewpoint generated by the edge side neural network; Determining a first resource scheduling decision error based on the first production resource scheduling viewpoint and the prior learning annotation includes: determining the first resource scheduling decision error of the edge-side neural network based on the first production resource scheduling viewpoint and the prior learning annotation.

8. The method according to any one of claims 1 to 6, characterized in that: The business operation data mining network includes a middle platform side neural network and an edge side neural network; the second production resource scheduling viewpoint includes a global production resource scheduling viewpoint generated by the middle platform side neural network and an edge production resource scheduling viewpoint generated by the edge side neural network, and the updated second production resource scheduling viewpoint includes an updated global production resource scheduling viewpoint; Determining the second resource scheduling decision error based on the updated second production resource scheduling viewpoint includes: determining the second resource scheduling decision error of the edge side neural network based on the edge production resource scheduling viewpoint and the updated global production resource scheduling viewpoint; The determining the second resource scheduling decision error of the edge-side neural network according to the edge production resource scheduling viewpoint and the updated global production resource scheduling viewpoint includes: Determining a resource scheduling decision error of a derived scheduling viewpoint based on the edge production resource scheduling viewpoint and the updated global production resource scheduling viewpoint; and determining a resource scheduling joint decision error based on the edge production resource scheduling viewpoint and the global production resource scheduling viewpoint; Determine the fusion result between the resource scheduling decision error of the derived scheduling viewpoint and the resource scheduling joint decision error as the second resource scheduling decision error of the edge-side neural network; The step of determining the fusion result between the resource scheduling decision error of the derived scheduling viewpoint and the resource scheduling joint decision error as the second resource scheduling decision error of the edge-side neural network includes: Determine a first confidence coefficient and a second confidence coefficient based on a current number of debugging cycles and a maximum number of debugging cycles of the debugging process; Determine a first error-weighted result of the first confidence coefficient and the resource scheduling decision error of the derived scheduling viewpoint, and a second error-weighted result of the second confidence coefficient and the resource scheduling joint decision error; The second resource scheduling decision error of the edge-side neural network is determined by combining the first error weighted result and the second error weighted result.

9. A production resource dynamic allocation management system, characterized in that: include: Memory, used to store program instructions and data; A processor, coupled to a memory, and configured to execute instructions in the memory to implement the method according to any one of claims 1 to 8.

10. A computer storage medium, characterized in that: Contains instructions, which, when executed on a processor, implement the method according to any one of claims 1 to 8.

Images