Large-scale model-driven intelligent analysis methods, systems, and media for industrial services in chemical industrial parks

Abstract

This application provides a large-scale model-driven intelligent analysis method, system, and medium for industrial services in chemical industrial parks, relating to the field of computational analysis technology. The method includes: reading multiple projects from each participating enterprise in the chemical industrial park; dividing the multiple projects into time-aligned segments according to multiple time period nodes, outputting multiple project sets; driving the training of a large-scale collaborative gain model to combine each project set within the multiple project sets, analyzing the project collaborative indicators of the enterprise combination solution, and obtaining multiple project collaborative indicators; evaluating the collaborative gain index of the multiple project collaborative indicators; if the collaborative gain index passes verification using preset gain rules, outputting the enterprise combination solution as the enterprise matching result for the chemical industrial park. This application can solve the technical problem in existing technologies where it is difficult to accurately characterize the collaborative effects between enterprise projects in chemical industrial parks, achieving the technical effect of improving the accuracy of characterizing the collaborative effects between enterprise projects.

Description

Technical Field

[0001] This application relates to the field of computational analysis technology, and in particular to intelligent analysis methods, systems and media for industrial services in chemical industrial parks driven by large models. Background Technology

[0002] With the continuous advancement of the clustering and green transformation of the chemical industry, chemical industrial parks are gradually becoming core carriers for high-end chemical enterprises and promoting coordinated industrial development. In their investment promotion decisions and collaborative planning among enterprises, park management increasingly relies on data-driven and intelligent analytical methods to ensure that project matching among resident enterprises achieves resource sharing, energy efficiency complementarity, and supply chain synergy.

[0003] Currently, while existing methods for analyzing chemical industrial parks have begun to incorporate some machine learning models or big data clustering techniques, problems still exist. Most of these methods rely primarily on the static attributes of enterprises, neglecting the dynamic interactions and cross-cycle correlations between enterprise projects. Secondly, existing models often fail to handle the alignment of collaborative features across multiple time points, resulting in insufficient modeling of the cyclical differences between different enterprise projects, thus hindering the formation of cross-period collaborative evolution patterns.

[0004] In summary, existing technologies suffer from a lack of a mechanism for aligning and dynamically evolving the collaborative characteristics of multiple enterprises and projects within chemical industrial parks at different time periods. This results in an inability to accurately depict the trends in collaborative gains and long-term matching effects among enterprises over continuous periods, further impacting the scientific and sustainable development of chemical industrial parks. Summary of the Invention

[0005] The purpose of this application is to provide a large-scale model-driven intelligent analysis method, system, and medium for industrial services in chemical industrial parks. This aims to address the technical problem in existing technologies where the lack of a modeling mechanism for aligning and dynamically evolving the collaborative characteristics of multiple enterprises and projects within a chemical industrial park at different time periods makes it impossible to accurately depict the trend of collaborative gain changes and long-term matching effects among enterprises over continuous periods. This further affects the scientific and sustainable nature of industrial collaborative development in chemical industrial parks.

[0006] In view of the above problems, this application provides a method, system and medium for intelligent analysis of industrial services in chemical industrial parks driven by large models.

[0007] Firstly, this application provides a large-scale model-driven intelligent analysis method for industrial services in chemical industrial parks, implemented through a large-scale model-driven intelligent analysis system for industrial services in chemical industrial parks. The method includes: collecting a set of participating enterprises in the chemical industrial park; reading multiple projects from each participating enterprise; setting multiple time period nodes; dividing the multiple projects of each participating enterprise into time-aligned segments according to the multiple time period nodes; outputting multiple project sets, where each time period node includes one project set; training a collaborative gain large-scale model; driving the trained collaborative gain large-scale model to combine each project set in the multiple project sets with a preset target number of enterprises; analyzing the project collaboration index of the enterprise combination solution based on the time period nodes; obtaining multiple project collaboration indices corresponding to the multiple time period nodes; evaluating the collaborative gain index of the multiple project collaboration indices based on continuous time period nodes; verifying the collaborative gain index using preset gain rules; if the verification is successful, outputting the enterprise combination solution as the enterprise matching result of the chemical industrial park.

[0008] Preferably, the large model-driven intelligent analysis method for industrial services in chemical industrial parks further includes: obtaining multiple project time periods for each entered enterprise, including project start time and project end time; traversing multiple project time periods for each entered enterprise, dividing the project start time and project end time into time-aligned segments according to the multiple time period nodes, and outputting multiple project sets.

[0009] Preferably, the large-model-driven intelligent analysis method for industrial services in chemical industrial parks further includes: constructing collaborative gain training data, wherein the collaborative gain training data includes sample data of analogous chemical industrial parks, sample data of enterprise projects participating in the analogous chemical industrial parks, and labeled collaborative gain scoring label samples; reading the feature vector sample group of the enterprise project sample data of the analogous chemical industrial parks; and training a neural network according to the collaborative gain training data and the feature vector sample group until the mean square error between the predicted collaborative gain scoring sample and the collaborative gain scoring label sample is less than a preset threshold, thereby obtaining a collaborative gain large model.

[0010] Preferably, the large model-driven intelligent analysis method for industrial services in chemical industrial parks further includes: the neural network includes an enterprise encoder, a combined encoder, and a predictive regression network; the enterprise encoder is used to perform high-dimensional convolution of the collaborative feature vectors on the feature vector sample group; the combined encoder is used to fit the collaborative relationship of the high-dimensional convolutioned collaborative feature vectors based on a graph neural network; and the predictive regression network is used to perform regression prediction on the collaborative feature combination vectors fitted with the collaborative relationship, and output the predicted collaborative gain score sample.

[0011] Preferably, the large model-driven intelligent analysis method for industrial services in chemical industrial parks further includes: selecting a combination of enterprises with the preset target number of enterprises from the input enterprise set to obtain a first candidate enterprise combination solution; identifying multiple project sets of the first candidate enterprise combination solution at multiple time period nodes; inputting each project set in the multiple project sets into the training collaborative gain large model for collaborative gain scoring to obtain the project collaborative index corresponding to each time period node.

[0012] Preferably, the large-scale model-driven intelligent analysis method for industrial services in chemical industrial parks further includes: after obtaining the project synergy index corresponding to each time period node, obtaining multiple project synergy indices corresponding to the multiple time period nodes in sequence, evaluating the synergy gain index of the multiple project synergy indices based on the continuous time period nodes; verifying the synergy gain index of the continuous time period nodes using preset gain rules, and if the identification gain rule in the preset gain rules is satisfied, outputting the verification pass result, and outputting the enterprise combination solution as the enterprise matching result of the chemical industrial park; wherein, the preset gain rules include multiple gain rules, and each gain rule is obtained through contract configuration using a month-on-month gain index or a cumulative gain index.

[0013] Preferably, the large model-driven intelligent analysis method for industrial services in chemical industrial parks further includes: the identification gain rule is one of the plurality of gain rules, which includes at least a continuous growth rule, an accelerated growth rule, a threshold expectation rule, and a smooth growth rule; and the identification gain rule is obtained through feedback from the management target personnel of the chemical industrial park.

[0014] Preferably, the large model-driven intelligent analysis method for industrial services in chemical industrial parks further includes: if the identifier gain rule in the preset gain rule is not satisfied, output a verification failure result; using the identifier gain rule as the constraint target, optimize the first candidate enterprise combination solution from the input enterprise set until the enterprise combination solution that satisfies the identifier gain rule is output as the enterprise matching result of the chemical industrial park.

[0015] Secondly, this application also provides a large-scale model-driven intelligent analysis system for chemical industrial park services, used to execute the large-scale model-driven intelligent analysis method for chemical industrial park services as described in the first aspect, including: a project reading module, used to collect a set of enterprises participating in the chemical industrial park and read multiple projects of each enterprise; a project set output module, used to set multiple time period nodes, perform time alignment division of multiple projects of each enterprise according to the multiple time period nodes, and output multiple sets of project sets, wherein each time period node includes a set of project sets; a project synergy index acquisition module, used to train a large-scale synergy gain model, drive the trained large-scale synergy gain model to combine each set of project sets in the multiple sets of project sets with a preset target number of enterprises, analyze the project synergy index of the enterprise combination solution based on the time period nodes, and obtain multiple project synergy indices corresponding to the multiple time period nodes; and an enterprise matching result output module, used to evaluate the synergy gain index of the multiple project synergy indices based on continuous time period nodes, verify the synergy gain index using preset gain rules, and if the verification is successful, output the enterprise combination solution as the enterprise matching result of the chemical industrial park.

[0016] Thirdly, a computer-readable storage medium storing a computer program, which, when executed, implements the steps of the large-model-driven intelligent analysis method for industrial services in chemical industrial parks as described in any one of the first aspects above.

[0017] The technical solution provided in this application has at least the following technical effects or advantages: by realizing the technical goal of intelligent modeling and dynamic optimization matching of collaborative gains based on time series project characteristics, it achieves the technical effect of enabling cross-cycle collaborative evaluation and continuous gain verification of chemical industrial park enterprise portfolios.

[0018] The above description is merely an overview of the technical solution of this application. To enable a clearer understanding of the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the intelligent analysis method for large-scale model-driven industrial services in chemical industrial parks used in this application.

[0021] Figure 2 This is a schematic diagram of the structure of the large-scale model-driven intelligent analysis system for chemical industrial park services in this application.

[0022] Attached diagram labels: Module 1 for project reading, Module 2 for project collection output, Module 3 for obtaining project collaboration indicators, and Module 4 for outputting enterprise matching results. Detailed Implementation

[0023] This application addresses the technical problem in existing technologies that lack a mechanism for aligning and dynamically evolving the collaborative characteristics of multiple enterprises and projects within a chemical industrial park across different time periods. This lack of a mechanism prevents accurate characterization of the trends in collaborative gains and long-term matching effects among enterprises over continuous periods, thus impacting the scientific rigor and sustainability of industrial collaborative development in chemical industrial parks. The application achieves the technical goal of intelligent modeling and dynamic optimization matching of collaborative gains based on time-series project characteristics, enabling cross-period collaborative assessment and continuous gain verification of enterprise combinations within chemical industrial parks.

[0024] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0025] Example 1, please refer to the appendix. Figure 1 This application provides a large-scale model-driven intelligent analysis method for industrial services in chemical industrial parks, which is applied to a large-scale model-driven intelligent analysis system for industrial services in chemical industrial parks. The method specifically includes the following steps:

[0026] S1: Collect the set of companies participating in the chemical industrial park and read multiple projects for each company.

[0027] Specifically, all registered or participating enterprises within the chemical industrial park will be identified, forming a data set. This data set comprises a collection of information including basic details, business scope, number of projects, and project types for each enterprise within the park. Data collection can be conducted through the park's information system, business registration platform, or through self-reporting by enterprises, ensuring data integrity and timeliness.

[0028] Next, multiple project data for each enterprise are retrieved from the entered enterprises. A project refers to an enterprise's activities within the park, such as production construction, technological upgrading, capacity expansion investment, energy conservation and environmental protection, or new product development. Each project includes information such as start time, end time, project scale, investment amount, production process, production capacity structure, and energy consumption indicators.

[0029] S2: Set multiple time period nodes, and divide the multiple projects of each enterprise into time-aligned segments according to the multiple time period nodes, and output multiple sets of projects, wherein each time period node includes a set of projects.

[0030] Furthermore, this application also includes: obtaining multiple project time periods for each entered enterprise, including project start time and project end time; traversing multiple project time periods for each entered enterprise, dividing the project start time and project end time into time-aligned segments according to the multiple time period nodes, and outputting multiple project sets.

[0031] Specifically, a set of time reference points is established according to chronological order, that is, multiple time period nodes are set. Time period nodes are benchmarks used to align projects of different companies in the time dimension, and can be based on quarters, years, half-years or specific months.

[0032] Next, we obtain the time cycle information for each project in each enterprise, namely the start and end time of the project, which reflects the project's operation cycle or construction phase and is an important parameter for subsequent time alignment.

[0033] Then, the time periods of multiple projects for each entered enterprise are accessed sequentially, reading the time range of all projects one by one and comparing it with pre-defined time period nodes. This allows the determination of the corresponding position of each project within each time node. Next, the start and end times of projects are aligned according to the time period nodes, assigning projects spanning multiple time periods to their respective periods, thus forming continuous and comparable time segments for the project's time-series data. After time alignment, multiple project sets are output, each corresponding to a time period node. Each time period node contains a set of projects that are in progress within that time segment. For a given enterprise, the project set within a specific time period includes all projects that the enterprise is active within that time period node. This allows the park to easily analyze the activity distribution, project density, and development trends of enterprises across different time periods.

[0034] S3: Train the large collaborative gain model, drive the trained large collaborative gain model to combine each of the multiple project sets in the multiple project sets with a preset enterprise target number, analyze the enterprise combination solution based on the project collaborative index of time period nodes, and obtain multiple project collaborative indices corresponding to the multiple time period nodes.

[0035] Furthermore, this application also includes: constructing collaborative gain training data, wherein the collaborative gain training data includes sample data of analogous chemical parks of the chemical industrial park, sample data of enterprise projects participating in the analogous chemical parks, and labeled collaborative gain scoring label samples; reading the feature vector sample group of the enterprise project sample data of the analogous chemical parks; and training a neural network according to the collaborative gain training data and the feature vector sample group until the mean square error of the predicted collaborative gain scoring sample and the collaborative gain scoring label sample is less than a preset threshold, thereby obtaining a large collaborative gain model.

[0036] Furthermore, this application also includes: the neural network includes an enterprise encoder, a combined encoder, and a predictive regression network; the enterprise encoder is used to perform high-dimensional convolution of the collaborative feature vectors on the feature vector sample group; the combined encoder is used to fit the collaborative relationship of the high-dimensional convolutioned collaborative feature vectors based on a graph neural network; and the predictive regression network is used to perform regression prediction on the collaborative feature combination vector fitted with the collaborative relationship, and output a predicted collaborative gain score sample.

[0037] Furthermore, this application also includes: selecting a combination of enterprises with the preset target number of enterprises from the input enterprise set to obtain a first candidate enterprise combination solution; identifying multiple project sets of the first candidate enterprise combination solution at multiple time period nodes; inputting each project set in the multiple project sets into the training collaborative gain large model for collaborative gain scoring to obtain the project collaborative index corresponding to each time period node.

[0038] Furthermore, this application also includes: after obtaining the project collaboration index corresponding to each time period node, obtaining multiple project collaboration indices corresponding to the multiple time period nodes in sequence, evaluating the collaboration gain index of the multiple project collaboration indices based on the continuous time period nodes; verifying the collaboration gain index of the continuous time period nodes using preset gain rules, and if the identification gain rule in the preset gain rules is satisfied, outputting the verification pass result, and outputting the enterprise combination solution as the enterprise matching result of the chemical industrial park; wherein, the preset gain rules include multiple gain rules, and each gain rule is obtained by contract configuration through the month-on-month gain index or the cumulative gain index.

[0039] Furthermore, this application also includes: if the identifier gain rule in the preset gain rule is not satisfied, output a verification failure result; with the identifier gain rule as the constraint target, optimize the first candidate enterprise combination solution from the input enterprise set until the enterprise combination solution that satisfies the identifier gain rule is output as the enterprise matching result of the chemical industrial park.

[0040] Furthermore, this application also includes: the identification gain rule is one of the plurality of gain rules, the plurality of gain rules including at least a continuous growth rule, an accelerated growth rule, a threshold expectation rule, and a smooth growth rule; and the identification gain rule is obtained through feedback from the management target personnel of the chemical industrial park.

[0041] Specifically, a collaborative gain training dataset is constructed to train a collaborative gain relationship model among enterprises in the chemical industrial park. This training dataset is a comprehensive sample set containing analogous data from chemical industrial parks similar to the target park—parks comparable to the target park in terms of industrial structure, enterprise size, energy consumption, and project types. The training dataset includes analogous chemical industrial park sample data, which provides experience in collaborative development, enabling the training dataset to learn the regular characteristics of enterprise collaboration during the training process. Furthermore, the training dataset also includes sample data of enterprise projects participating in the analogous chemical industrial parks—specific project information implemented by enterprises within the park, such as production projects, investment projects, or environmental upgrade projects. Each project includes attributes such as time period, capacity indicators, material consumption, and energy structure. Finally, the training dataset also includes labeled collaborative gain scoring tags, which are quantitative evaluation values ​​used to characterize the collaborative effects between enterprises in project collaboration, resource sharing, or industrial complementarity.

[0042] Next, the feature vector sample set is read from the sample data of enterprise projects in the analogous chemical industrial park. The feature vector sample set is a set of vectors formed by numerically processing the data features of each enterprise project, including elements such as the project's input resources, output products, investment intensity, energy consumption level, and time period distribution. Each feature vector represents the multi-dimensional attribute spatial location of a project, allowing the model training to learn the similarities or complementarities between different projects in mathematical space.

[0043] Subsequently, the constructed synergistic gain training data and extracted feature vector sample sets are used to train the neural network. A neural network is a computational model that mimics the connection patterns of neurons in the human brain, achieving complex nonlinear mapping relationships through multi-layer parameter adjustments. The training process involves continuously adjusting the weight parameters of the neural network based on the differences between the input enterprise project features and the target synergistic gain score, minimizing the error between the predicted value and the true label value. This continues until the mean squared error between the predicted synergistic gain score sample and the synergistic gain score label sample is less than a preset threshold, resulting in a large synergistic gain model. The mean squared error is measured by the average of the squared differences between the predicted synergistic gain score sample and the synergistic gain score label sample to measure the accuracy of the neural network training. When the mean squared error between the predicted synergistic gain score sample and the true synergistic gain score label sample is less than the preset threshold, it indicates that the predictive performance of the neural network training has met the requirements, meaning the neural network model has been able to stably learn the patterns of enterprise synergistic relationships. The resulting neural network model is then considered a large synergistic gain model. The preset threshold can be customized by those skilled in the art based on actual conditions.

[0044] Furthermore, the neural network consists of three main functional modules: an enterprise encoder, a combined encoder, and a predictive regression network. A neural network is an algorithmic system that simulates the connection structure of neurons in the human brain, learning complex feature relationships through nonlinear mappings between multiple layers of computational units. It can automatically extract potential patterns from enterprise project data and improve prediction accuracy by continuously training and adjusting parameter weights.

[0045] The role of the enterprise encoder is to perform high-dimensional convolution of the input feature vector sample set with synergistic feature vectors. The feature vector sample set is a dataset formed by feature extraction and numerical processing of multiple enterprise projects. Each feature vector contains multi-dimensional attributes such as the project's time period, investment scale, capacity utilization rate, raw material consumption, and carbon emission intensity. High-dimensional convolution refers to using the mechanism of convolutional neural networks to perform feature fusion and pattern recognition in a high-dimensional feature space. It can capture synergistic trends between different dimensions through sliding windows and weighted calculations. For example, when two enterprises' projects simultaneously increase in raw material utilization rate and waste heat recovery rate, the high-dimensional convolution operation can extract synergy as a significant feature signal, thereby enhancing the model's ability to perceive synergistic relationships.

[0046] The combined encoder uses a graph neural network to fit collaborative relationships to collaborative feature vectors obtained from high-dimensional convolutions. A graph neural network is a neural network model that processes relational data structures, where nodes can represent enterprises or projects, and edges represent cooperation, supply and demand, or supply chain connections between enterprises. Collaborative relationship fitting means learning the interdependencies and influence patterns between different enterprises and projects by continuously propagating and aggregating information in the graph structure. For example, if enterprise A's waste can be used as raw material input for enterprise B, or if enterprises C and D share logistics resources within the same time period, strong edge connections are formed in the graph neural network. After multiple layers of graph convolution, the model can automatically identify collaborative patterns, which are reflected in a higher collaborative gain score potential in the output stage.

[0047] The predictive regression network performs regression prediction on the synergistic feature combination vector fitted with synergistic relationships and outputs a predicted synergistic gain score sample. Regression prediction is a machine learning method with continuous value output, predicting a numerical result based on input features. The synergistic feature combination vector is a comprehensive feature representation extracted and fused by the enterprise encoder and the ensemble encoder, containing information such as the synergistic potential between enterprises, time node distribution characteristics, and interaction relationships between projects. The predictive regression network transforms the features into specific synergistic gain scores through multi-layer fully connected computation. The scores are typically used to measure the potential benefits of enterprise combinations in areas such as resource sharing, energy efficiency complementarity, or supply chain integration. For example, if the synergistic effect between enterprise groups is strong, the predicted score may be close to 1; if the synergistic relationship is weak, the score may be close to 0.

[0048] Furthermore, a group of companies with a preset target number of companies is selected from the input company set to obtain the first candidate company combination solution. The preset target number of companies is a fixed value set according to the planning goals or capacity limitations of the chemical industrial park; for example, it could be set to select ten companies. A company combination refers to a group of companies, used as the unit of analysis. A company combination with the preset target number of companies represents all projects already in production and under construction by all selected companies within a certain time period as a whole. The first candidate company combination solution represents a potentially optimal solution used for subsequent calculations and comparisons.

[0049] Next, we identify multiple project sets across multiple time periods for the first candidate company portfolio, indicating the need to determine the project distribution of this portfolio at different time stages. Multiple time periods refer to pre-defined time intervals, such as annually or quarterly; multiple project sets represent the project portfolios of different companies at each time period, such as all projects under construction and put into operation within a given quarter. By extracting and organizing the project status over time, we ensure the projects are comparable across time dimensions.

[0050] Then, each project set from multiple project sets is input into a large-scale collaborative gain model for collaborative gain scoring. This means inputting the data features of the project sets into a pre-trained large-scale collaborative gain prediction model to calculate the overall collaborative effect of different enterprise projects in cooperative or coexisting scenarios. By learning from project collaboration data, the large-scale collaborative gain model can capture the gain effects generated by projects between different enterprises in terms of resource sharing, energy efficiency complementarity, or technological cooperation. The collaborative gain score is a quantitative indicator reflecting the comprehensive collaborative benefits achieved by the enterprise combination within a given time period.

[0051] Finally, the project collaboration indicators for each time period are obtained, representing the comprehensive collaboration results for each time period, used to measure the performance of the candidate enterprise portfolio at different time stages. Project collaboration indicators may include parameters such as energy efficiency improvement rate, cost collaboration coefficient, or comprehensive utilization rate of production capacity.

[0052] Furthermore, after obtaining the project synergy indicators corresponding to each time period node, multiple project synergy indicators corresponding to multiple time period nodes are obtained sequentially. The synergy gain index of these multiple project synergy indicators is evaluated based on continuous time period nodes. This involves summarizing and analyzing the indicators in chronological order to calculate the changing trend of synergy effects over a continuous time period. Project synergy indicators represent the comprehensive collaborative effect of a business combination within a specific time period, and can be composed of indicators such as energy complementarity, capacity synergy rate, or resource sharing efficiency. Continuous time period nodes indicate the temporal continuity of multiple time period nodes, reflecting dynamic changes. The synergy gain index is a comprehensive calculation result that measures the changing trend of a company's synergy benefits over time. For example, by comparing the differences in synergy indicators between two consecutive quarters, the growth rate or volatility can be obtained. For instance, if the synergy indicator for the first quarter is 0.75 and for the second quarter is 0.9, then the continuous time period synergy gain index could be 0.15, indicating a 20% increase in synergy benefits.

[0053] Furthermore, the synergistic gain index of nodes in continuous time periods is verified using preset gain rules to determine whether the enterprise combination meets the gain objectives of the chemical industrial park during continuous operation. If the preset gain rules meet the identifying gain rules, the verification result is output, and the enterprise combination solution is output as the enterprise matching result of the chemical industrial park. The preset gain rules are a set of pre-defined standards used to determine whether the synergistic benefits have reached the expected level, such as requiring the synergistic gain index to maintain positive growth for three consecutive quarters. The identifying gain rules are the core judgment conditions used to mark whether the synergistic combination meets the qualified or preferred standards. A verification result means that the enterprise combination performs well in synergistic operation and is in line with the park's development direction. The output of the enterprise combination solution as the enterprise matching result indicates that the combination has been selected as the recommended enterprise synergy scheme.

[0054] The preset gain rules include multiple gain rules, each obtained through contract configuration using either a month-on-month gain index or a cumulative gain index. The month-on-month gain index is used to compare the magnitude of change between adjacent time periods, such as using the synergy growth rate of the second quarter compared to the first quarter to assess short-term dynamic benefits; the cumulative gain index represents the overall cumulative value of synergy benefits over multiple time periods, used to measure long-term synergy trends; contract configuration means that multiple gain rules are configured based on target parameters in the park's operation plan or cooperation contract, and are binding and guiding.

[0055] Furthermore, if the identified gain rule in the preset gain rules is not met, a verification failure result will be output. That is, when verifying the synergistic gain performance of the enterprise portfolio, if it is found that its synergistic gain index does not reach the standard set in the identified gain rule, a verification failure judgment will be given directly. The identified gain rule is used to determine whether the enterprise portfolio has achieved the synergistic goals of the chemical industrial park. A verification failure result means that the synergistic effect of the enterprise portfolio does not meet the requirements over a continuous time period, such as the synergistic gain index not reaching the preset threshold, unstable growth trend, or negative fluctuations.

[0056] Next, using the identifier gain rule as the constraint objective, the first candidate enterprise combination solution is optimized from the input enterprise set until an enterprise combination solution that satisfies the identifier gain rule is output as the enterprise matching result for the chemical industrial park. That is, if the verification fails, the identifier gain rule is used as the optimization objective to iteratively optimize and re-select the enterprise combination. The constraint objective refers to the condition that must be met during the optimization process, i.e., the newly generated enterprise combination must meet the requirements of the identifier gain rule. The first candidate enterprise combination solution is the initially selected enterprise combination scheme. Optimization is a process of multi-round optimization of the first candidate enterprise combination solution through algorithms, using methods such as genetic algorithms, simulated annealing, or gradient search to improve the synergy gain index. For example, some enterprise members may be replaced or the time period matching method may be adjusted in each round of optimization to improve the overall synergy benefit of the combination. The optimization process continues until an enterprise combination solution that satisfies the identifier gain rule is output, indicating that the optimization process will continue until the optimal solution that meets the objective conditions is found. The final enterprise matching result is the cooperative enterprise combination scheme recommended by the park.

[0057] Furthermore, the identifying gain rule is one of several gain rules, which include at least the continuous growth rule, the accelerated growth rule, the threshold expectation rule, and the smooth growth rule. The identifying gain rule is a crucial rule used to identify whether the synergy gain performance has met the target, representing the final decision-making basis. Multiple gain rules refer to various judgment criteria defined by the system on different dimensions, used to evaluate the synergy performance of the enterprise portfolio from multiple perspectives. The continuous growth rule is used to determine whether the synergy gain index maintains positive growth over several consecutive time periods. For example, if the synergy benefits of the enterprise portfolio increase by 5%, 7%, and 9% respectively in three consecutive quarters, then the continuous growth condition is met. The accelerated growth rule is used to determine whether the growth rate is increasing, indicating that the synergy benefits are not only continuously improving but also accelerating. For example, a 3% increase in the first quarter compared to the previous quarter and a 6% increase in the second quarter indicate that collaboration between enterprises is rapidly strengthening. The threshold expectation rule is used to verify whether the synergy gain has reached the set minimum target value, for example, requiring a gain index of at least 0.8 to be considered acceptable. The smooth growth rule is used to ensure the stability of the growth process and avoid drastic fluctuations or abnormal growth.

[0058] Next, the identification gain rule is obtained through the feedback of the management target personnel in the chemical industrial park. That is, the basis for finally determining which gain rule is selected as the identification gain rule comes from the management target personnel in the chemical industrial park, that is, the personnel with decision-making power or management responsibilities provide feedback according to the development goals and strategic directions of the park. The management target personnel include the investment promotion person in charge, planning managers or experts in the economic benefit evaluation department of the park, and adjust the identification rules according to the management priorities of the park at different stages. For example, when the park is in the initial stage of development, it pays more attention to the stability of the growth of the number of projects, so it may choose the smooth growth rule as the identification rule; after the park enters the high-quality development stage, it may pay more attention to the speed of benefit improvement, and thus choose the accelerated growth rule as the identification rule.

[0059] S4: Evaluate the collaborative gain index of the multiple project collaboration indicators based on consecutive time period nodes, verify the collaborative gain index using a preset gain rule. If the verification passes, output the enterprise combination solution as the enterprise matching result of the chemical industrial park.

[0060] Specifically, evaluate the collaborative gain index of the multiple project collaboration indicators based on consecutive time period nodes. After obtaining the project collaboration indicators within each time period, comprehensively analyze the indicators in chronological order to calculate a collaborative gain index that reflects the overall trend of collaborative benefits. The project collaboration indicator refers to the collaborative performance value of the projects of different enterprises in aspects such as resource sharing, industrial chain coupling, or production cooperation within each time period, and the consecutive time period nodes mean that the analysis is not a static single moment but dynamically spans multiple time periods. The collaborative gain index is a comprehensive indicator used to measure the collaborative growth level of the enterprise combination in the time dimension and reflects the long-term trend. For example, if the collaborative index of an enterprise combination is 0.75 in the first quarter, 0.8 in the second quarter, and 0.85 in the third quarter, the collaborative gain index can be obtained through time weighting or cumulative growth rate methods to reflect whether the collaborative relationship is continuously improving.

[0061] Next, verify the collaborative gain index using a preset gain rule. That is, after obtaining the collaborative gain index, compare it with a series of preset rules to determine whether the enterprise combination meets the collaborative development goals of the chemical industrial park. The preset gain rule is a set of evaluation criteria formulated according to the strategic goals of the park, such as requiring the collaborative gain index to rise continuously, the average annual growth rate not to be lower than a certain value, or to maintain stable fluctuations within a certain period. The verification process is to compare the calculated collaborative gain index with these rules item by item to determine whether the enterprise combination has continuous and stable collaborative capabilities.

[0062] Finally, if the verification is successful, the enterprise combination solution is output as the enterprise matching result for the chemical industrial park. That is, when the synergy gain index of the enterprise combination meets the gain rule, the enterprise combination is identified as the optimal or effective synergy solution and output as the park's enterprise matching result. The enterprise matching result is an important basis for the park's investment attraction and industrial synergy decisions, indicating that the enterprises in the combination are believed to form a good synergy effect in project operation, resource utilization, and development pace. For example, if a park plans to introduce 5 enterprises, and after verification it is confirmed that the synergy gain index of enterprises A, B, C, D, and E reaches 0.9 and has a stable growth trend, then this combination will be output as a recommended solution to guide investment attraction layout or investment decisions.

[0063] In summary, the large-scale model-driven intelligent analysis method for chemical industrial park services provided in this application has the following technical effects: by achieving the technical goal of intelligent modeling and dynamic optimization matching of synergistic gains based on time-series project characteristics, it can achieve the technical effect of cross-cycle synergistic evaluation and continuous gain verification of the enterprise portfolio in chemical industrial parks.

[0064] Example 2: Based on the same inventive concept as the large-model-driven intelligent analysis method for chemical industrial park services in the foregoing examples, this application also provides a large-model-driven intelligent analysis system for chemical industrial park services. Please refer to the appendix. Figure 2 The system includes: a project reading module 1, used to collect the set of enterprises participating in the chemical industrial park and read multiple projects of each enterprise; a project set output module 2, used to set multiple time period nodes, divide the multiple projects of each enterprise into time-aligned segments according to the multiple time period nodes, and output multiple project sets, wherein each time period node includes a project set; a project collaboration index acquisition module 3, used to train a large collaboration gain model, drive the trained large collaboration gain model to combine each project set in the multiple project sets with a preset target number of enterprises, analyze the project collaboration index of the enterprise combination solution based on the time period nodes, and obtain multiple project collaboration indices corresponding to the multiple time period nodes; and an enterprise matching result output module 4, used to evaluate the collaboration gain index of the multiple project collaboration indices based on continuous time period nodes, verify the collaboration gain index using preset gain rules, and if the verification is successful, output the enterprise combination solution as the enterprise matching result of the chemical industrial park.

[0065] Furthermore, the large model-driven intelligent analysis system for industrial services in chemical industrial parks is also used to: obtain multiple project time periods for each entered enterprise, including project start time and project end time; traverse multiple project time periods for each entered enterprise, align the project start time and project end time according to the multiple time period nodes, and output multiple sets of projects.

[0066] Furthermore, the large-scale model-driven intelligent analysis system for industrial services in chemical industrial parks is also used for: constructing collaborative gain training data, which includes sample data of analogous chemical industrial parks, sample data of enterprise projects participating in the analogous chemical industrial parks, and labeled collaborative gain scoring labels; reading the feature vector sample group of the enterprise project sample data of the analogous chemical industrial parks; and training a neural network according to the collaborative gain training data and the feature vector sample group until the mean square error between the predicted collaborative gain scoring samples and the collaborative gain scoring label samples is less than a preset threshold, thereby obtaining a large-scale collaborative gain model.

[0067] Furthermore, the large model-driven intelligent analysis system for industrial services in chemical industrial parks is also used in the following ways: the neural network includes an enterprise encoder, a combined encoder, and a predictive regression network; the enterprise encoder is used to perform high-dimensional convolution of the collaborative feature vectors on the feature vector sample group; the combined encoder is used to fit the collaborative relationship of the high-dimensional convolutioned collaborative feature vectors based on a graph neural network; and the predictive regression network is used to perform regression prediction on the collaborative feature combination vector fitted with the collaborative relationship, and output the predicted collaborative gain score sample.

[0068] Furthermore, the large model-driven intelligent analysis system for industrial services in chemical industrial parks is also used to: select a combination of enterprises with the preset target number of enterprises from the input enterprise set to obtain a first candidate enterprise combination solution; identify multiple project sets of the first candidate enterprise combination solution at multiple time period nodes, input each project set in the multiple project sets into the training collaborative gain large model for collaborative gain scoring, and obtain the project collaborative index corresponding to each time period node.

[0069] Furthermore, the large-scale model-driven intelligent analysis system for industrial services in chemical industrial parks is also used for: obtaining project collaboration indicators corresponding to each time period node, sequentially obtaining multiple project collaboration indicators corresponding to the multiple time period nodes, evaluating the collaboration gain index of the multiple project collaboration indicators based on continuous time period nodes; verifying the collaboration gain index of continuous time period nodes using preset gain rules, and if the identification gain rule in the preset gain rules is satisfied, outputting the verification pass result, and outputting the enterprise combination solution as the enterprise matching result of the chemical industrial park; wherein, the preset gain rules include multiple gain rules, and each gain rule is obtained through contract configuration using a month-on-month gain index or a cumulative gain index.

[0070] Furthermore, the large-scale model-driven intelligent analysis system for industrial services in chemical industrial parks is also used to: the identification gain rule is one of the plurality of gain rules, which includes at least a continuous growth rule, an accelerated growth rule, a threshold expectation rule, and a smooth growth rule; and the identification gain rule is obtained through feedback from the management target personnel of the chemical industrial park.

[0071] Furthermore, the large model-driven intelligent analysis system for industrial services in chemical industrial parks is also used to: output a verification failure result if the identification gain rule in the preset gain rule is not met; optimize the first candidate enterprise combination solution from the input enterprise set with the identification gain rule as the constraint target until the enterprise combination solution that satisfies the identification gain rule is output as the enterprise matching result of the chemical industrial park.

[0072] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The large-model-driven intelligent analysis method and specific examples of industrial services in chemical industrial parks described in the foregoing embodiment one are also applicable to the large-model-driven intelligent analysis system of industrial services in chemical industrial parks in this embodiment. Through the foregoing detailed description of the large-model-driven intelligent analysis method of industrial services in chemical industrial parks, those skilled in the art can clearly understand the large-model-driven intelligent analysis system of industrial services in chemical industrial parks in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0073] Example 3: Based on the same inventive concept as the large-model-driven intelligent analysis method for chemical industrial park services in the foregoing examples, this application also provides a computer-readable storage medium storing a computer program. When the computer program is executed, it implements the steps of the large-model-driven intelligent analysis method for chemical industrial park services described in any one of Examples 1 above.

[0074] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0075] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A large-scale model-driven intelligent analysis method for industrial services in chemical industrial parks, characterized by: The method includes: Collect the set of companies participating in the chemical industrial park and read multiple projects for each company; Set multiple time period nodes, and divide multiple projects of each enterprise into time-aligned segments according to the multiple time period nodes, and output multiple project sets, wherein each time period node includes a project set. Train a large collaborative gain model, drive the trained large collaborative gain model to combine each of the multiple project sets in the multiple project sets with a preset number of enterprise targets, analyze the enterprise combination solution based on the project collaborative index of time period nodes, and obtain multiple project collaborative indices corresponding to the multiple time period nodes. The synergy index of the multiple projects is evaluated based on the synergy gain index of continuous time period nodes. The synergy gain index is verified using a preset gain rule. If the verification is successful, the enterprise combination solution is output as the enterprise matching result of the chemical industrial park. The method for training the large collaborative gain model includes: Construct collaborative gain training data, which includes sample data of analogous chemical parks to the chemical industrial park, sample data of enterprise projects participating in the analogous chemical industrial park, and labeled collaborative gain scoring tags. Read the feature vector sample group of the enterprise project sample data of the analogous chemical industrial park; The neural network is trained according to the synergistic gain training data and feature vector sample group until the mean square error between the predicted synergistic gain score sample and the synergistic gain score label sample is less than a preset threshold, thus obtaining the synergistic gain large model. The neural network includes an enterprise encoder, a combined encoder, and a predictive regression network; The enterprise encoder is used to perform high-dimensional convolution of the collaborative feature vectors on the feature vector sample group. The combined encoder is based on a graph neural network to fit the collaborative relationship of the high-dimensional convolutioned collaborative feature vectors. The prediction regression network is used to perform regression prediction on the collaborative feature combination vector fitted with the collaborative relationship and output the predicted collaborative gain score sample.

2. The large-scale model-driven intelligent analysis method for industrial services in chemical industrial parks as described in claim 1, characterized in that, The multiple projects of each enterprise are time-aligned and divided according to the multiple time period nodes, and multiple sets of projects are output. The method includes: Obtain multiple project timelines for each entered company, including project start time and project end time; Iterate through multiple project time periods for each entered enterprise, align the project start time and project end time according to the multiple time period nodes, and output multiple project sets.

3. The large-scale model-driven intelligent analysis method for industrial services in chemical industrial parks as described in claim 1, characterized in that, The method for obtaining multiple project collaboration indicators corresponding to the multiple time period nodes includes: Select the preset target number of enterprise combinations from the input enterprise set to obtain the first candidate enterprise combination solution; Identify multiple project sets at the multiple time period nodes of the solution of the first candidate enterprise combination, input each project set in the multiple project sets into the training collaborative gain large model for collaborative gain scoring, and obtain the project collaborative index corresponding to each time period node.

4. The large-scale model-driven intelligent analysis method for industrial services in chemical industrial parks as described in claim 3, characterized in that, After obtaining the project collaboration index corresponding to each time period node, the project collaboration index corresponding to the multiple time period nodes is obtained in sequence, and the collaboration gain index of the multiple project collaboration index based on the continuous time period nodes is evaluated. The collaborative gain index of nodes in a continuous time period is verified using a preset gain rule. If the identification gain rule in the preset gain rule is satisfied, the verification result is output, and the enterprise combination solution is output as the enterprise matching result of the chemical industrial park. The preset gain rules include multiple gain rules, each of which is obtained through contract configuration using a ring-to-ring gain index or a cumulative gain index.

5. The large-scale model-driven intelligent analysis method for industrial services in chemical industrial parks as described in claim 4, characterized in that, The identification gain rule is one of the plurality of gain rules, which includes at least a continuous growth rule, an accelerated growth rule, a threshold expectation rule, and a smooth growth rule. Furthermore, the identification gain rule is obtained through feedback from the management personnel of the chemical industrial park.

6. The large-scale model-driven intelligent analysis method for industrial services in chemical industrial parks as described in claim 5, characterized in that, The method for verifying the cooperative gain exponent of nodes with continuous time periods using a preset gain rule also includes: If the identification gain rule in the preset gain rule is not met, the verification failure result will be output. Using the identification gain rule as the constraint target, the first candidate enterprise combination solution is optimized from the input enterprise set until the enterprise combination solution that satisfies the identification gain rule is output as the enterprise matching result of the chemical industrial park.

7. A large-scale model-driven intelligent analysis system for industrial services in chemical industrial parks, characterized in that: The steps for implementing the large-model-driven intelligent analysis method for industrial services in chemical industrial parks as described in any one of claims 1 to 6 include: The project reading module is used to collect the set of enterprises participating in the chemical industrial park and read multiple projects for each enterprise. The project set output module is used to set multiple time period nodes, and to divide multiple projects of each entered enterprise into time alignment according to the multiple time period nodes, and output multiple project sets, wherein each time period node includes a project set. The project collaboration index acquisition module is used to train a large collaboration gain model, drive the trained large collaboration gain model to combine each of the multiple project sets with a preset number of enterprise target quantities, analyze the enterprise combination solution based on the project collaboration index of time period nodes, and obtain multiple project collaboration indices corresponding to the multiple time period nodes. The enterprise matching result output module is used to evaluate the synergistic gain index of the synergistic indicators of the multiple projects based on continuous time period nodes, and to verify the synergistic gain index using preset gain rules. If the verification is successful, the enterprise combination solution is output as the enterprise matching result of the chemical industrial park.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed, implements the steps of the large-model-driven intelligent analysis method for industrial services in chemical industrial parks as described in any one of claims 1 to 6.

Images