A natural gas purchase and sale personnel collaborative management system

Abstract

The application discloses a natural gas purchase and sale personnel collaborative management system and relates to the intelligent management field.The system comprises a central management module, a data acquisition module and a feature extraction module.The central management module is used for providing data access and instruction editing authority of each function module.The data acquisition module is used for acquiring real-time and historical environment data, natural gas use data and signing data of current and historical natural gas purchase and sale contracts of each place in the current region.The feature extraction module is used for extracting real-time natural gas demand features associated with current and future time periods from the environment data and the use data and extracting purchase and sale features of current and future time periods from the purchase and sale contract signing data.A prediction model constructed by using a deep learning algorithm outputs a risk coefficient of a demand amount of each specific place in a future continuous period, so that a manager can clearly identify a high-risk period and place and automatically generate an optimal supply sequence and supply amount distribution scheme in multiple future periods.

Description

Technical Field

[0001] This invention relates to the field of intelligent management technology, specifically to a natural gas purchase, sales, and personnel collaborative management system. Background Technology

[0002] As a clean fossil fuel, natural gas is playing an increasingly important role in energy consumption, with its applications in urban gas, industrial fuel, and power generation becoming more widespread and in-depth. At the same time, the purchase, sale, and supply chain management of natural gas are becoming unprecedentedly complex. On the one hand, market demand is highly dynamic, easily affected by a confluence of factors such as drastic temperature changes, holiday effects, adjustments to industrial production plans, and sudden public events, making forecasting extremely difficult. On the other hand, the supply side involves the overlapping fulfillment of long-term and short-term purchase and sale contracts, the coordinated dispatch of multiple gas sources, and cross-regional pipeline transmission and distribution, forming a complex system with multiple constraints and strong coupling.

[0003] Existing systems often separate demand forecasting from functional modules such as contract management and inventory management, failing to effectively integrate real-time environmental data and contract performance status information. They cannot build a dynamic profile of future demand, nor can they quantify the risk coefficient of contract performance failure due to demand fluctuations or supply disruptions. As a result, managers' perception of potential risks is lagging and vague, and they can only react passively after problems occur. When faced with the problem of supply sequencing and quota allocation in multiple locations, existing technologies usually make decisions based on fixed priority rules or the experience of managers. They lack systematic decision support, cannot comprehensively consider the predicted risk changes in the future multiple cycles, and are difficult to minimize the overall supply cost and risk under multiple constraints such as inventory and transportation capacity. This can easily lead to unreasonable resource allocation, either over-guaranteeing and causing waste, or under-guaranteeing and causing the risk of supply disruption. Summary of the Invention

[0004] (a) Technical problems to be solved In view of the above-mentioned shortcomings of the existing technology, the present invention provides a natural gas purchase and sales personnel collaborative management system, which can effectively solve the problems of the existing technology.

[0005] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: This invention discloses a natural gas purchase, sales, and personnel collaborative management system, comprising: The central management module provides data access and command editing permissions for each functional module; The data acquisition module is used to acquire real-time and historical environmental data, natural gas usage data, and current and historical natural gas purchase and sale contract signing data for various locations in the current area. The feature extraction module is used to extract real-time natural gas demand features that are related to the current and future time periods from environmental data and usage data, and to extract purchase and sale features for the current and future time periods from purchase and sale contract signing data. The risk prediction module is used to build a purchase and sales prediction model through deep learning algorithms. The model takes demand characteristics and purchase and sales characteristics as inputs and predicts the risk coefficient of the demand for each location in the region in the future for multiple consecutive time periods. The location analysis module is used to store and analyze the attribute information of each location within the region. Based on the preset urgency evaluation rules, it independently sorts the supply urgency of each location and generates an urgency ranking list. The inventory monitoring module is used to monitor and obtain the current total inventory and inventory distribution data of the natural gas system in the region in real time. The supply decision module is used to generate a global optimized supply strategy for the region in the future continuous time period based on the current inventory and distribution status, combined with the risk coefficient and urgency ranking list of each location in the future continuous period, through a preset decision algorithm. The strategy specifically includes the supply order and supply quantity allocation scheme of each location in different periods. The strategy execution module is used to translate supply strategies into corresponding scheduling instructions and drive the associated logistics, warehousing and personnel allocation systems to execute supply tasks.

[0006] Furthermore, the environmental data in the data acquisition module includes temperature, air pressure, humidity, precipitation, and special weather events; the natural gas usage data includes gas consumption at different times of day, peak and valley values, load rate of gas-using equipment, and distribution of user types; and the purchase and sale contract signing data includes the supply quantity, supply cycle, price terms, performance constraints, and historical performance records stipulated in the contract.

[0007] Furthermore, the process of providing natural gas demand features in the feature extraction module includes: Historical natural gas usage data is processed using a time series decomposition algorithm to separate long-term trend components, seasonal components, and residual components. The seasonal components include annual, monthly, weekly, and daily cycles. For real-time usage data, the sliding window technique is used to calculate short-term fluctuation characteristics, including the moving average, standard deviation, rate of change, and peak-to-valley ratio of recent gas consumption, forming a time series feature vector that characterizes the basic laws of gas consumption behavior. Key meteorological factors related to gas consumption were screened from environmental data. The core factors included temperature, humidity, wind speed and solar radiation intensity. The marginal impact coefficient of each meteorological factor on gas consumption in different numerical ranges was quantified. For future cycles, the expected gas consumption adjustment value driven by meteorological factors was obtained. The time-series feature vector is fused with the meteorological-driven forecast gas consumption adjustment value. Combined with special event information reported by various locations or collected by sensors, event impact factors are set. Based on the fused data, a real-time natural gas demand feature vector for future continuous time periods is generated. The vector includes the basic forecast value, as well as quantitative indicators such as trend strength, seasonal coefficient, meteorological sensitivity, and event disturbance.

[0008] Furthermore, the process of extracting purchase and sales features in the feature extraction module includes: For the purchase and sale contract signing data, natural language processing technology is used to extract key information, extract the supplier, demander, contract term, agreed supply volume for each period, benchmark price, price adjustment formula, delivery location, performance bond clause and liability for breach of contract clause, establish the relationship between the above entities, and transform the unstructured contract text into a structured contract element knowledge graph. Based on structured contract elements, the reliability coefficient of contract supply in future periods is calculated by weighting the supply stability characteristics, combined with the supplier's historical performance rate, average delay days, current production capacity, and the severity of default clauses in the contract. Obtain the historical price fluctuation range for each period, and define the price risk value according to a preset scale based on the fluctuation amplitude; The complete lifecycle of each purchase and sale contract is mapped onto a unified timeline; key performance point features of the contract are extracted for future consecutive periods; and the time point feature information is combined with reliability coefficients and price risk values ​​to form a purchase and sale feature vector arranged according to time periods.

[0009] Furthermore, the process of constructing the purchase and sales forecasting model in the risk forecasting module includes: Collect and organize environmental data sequences, natural gas usage data sequences, purchase and sale contract signing data sequences, and actual natural gas supply results data for various locations in the region during historical time periods. Clean the collected raw data, including handling missing values ​​and removing outliers, and standardize or normalize data from different sources and with different dimensions to form a historical training dataset. Based on the historical training dataset, feature extraction is performed through the feature extraction module to generate demand feature vectors and purchase and sales feature vectors at various historical time points, which are used as input features for the model. The deviation between the actual natural gas demand and the planned supply at various locations, or the occurrence and severity of supply interruption events, in one or more subsequent consecutive time periods are quantified into continuous risk values, which are used as target labels for model training. A hybrid deep neural network is constructed as the backbone of the purchase and sales prediction model; the network includes: a temporal feature processing subnetwork, a feature processing subnetwork, a feature fusion layer, and a multi-period prediction output layer; The preprocessed historical dataset is divided into training, validation, and test sets. The initial model is trained using the training set data, and the loss function between the predicted risk coefficient and the true risk label is minimized through backpropagation algorithm and optimizer. During training, the performance of the model is monitored using the validation set, and early stopping is used to prevent overfitting. After training, the final prediction accuracy, recall, and generalization ability of the model are evaluated using an independent test set. The optimal model parameters that have passed training and validation are solidified, and the complete model architecture and parameters are deployed to the risk prediction module to complete the initial construction of the purchase and sales prediction model.

[0010] Furthermore, the temporal feature processing subnetwork is used to process demand feature sequences with strong time dependence. This subnetwork is composed of multiple layers of long short-term memory units or gated recurrent units stacked together. The feature processing subnetwork is used to process structured features related to purchase and sale contracts. This subnetwork is composed of multiple fully connected layers. The feature fusion layer is used to concatenate and interact the high-level temporal abstract features output by the temporal subnetwork with the high-level semantic features output by the structured subnetwork. The multi-period prediction output layer consists of multiple parallel fully connected branches, each branch corresponding to a future prediction period, and outputs the demand achievement risk coefficient for the location corresponding to that period.

[0011] Furthermore, the attribute information in the location analysis module includes: location type, user scale, whether it is critical infrastructure, historical emergency response level, and political or economic importance weight; the urgency evaluation rule is to configure different weights according to each attribute information, dynamically calculate the urgency score of each location through a weighted scoring algorithm, and rank them accordingly.

[0012] Furthermore, the decision-making algorithm in the supply decision-making module is a multi-objective optimization algorithm. Its optimization objective function simultaneously considers the following constraints: minimizing overall supply risk, maximizing inventory utilization efficiency, prioritizing high-urgency locations, and smoothing supply fluctuations. Based on the current inventory status, the future cycle prediction risk coefficient matrix, and the urgency ranking list, the decision-making algorithm iteratively solves the problem and outputs the optimal supply sequence and the supply allocation matrix for each location that satisfy all constraints over multiple future time periods.

[0013] Furthermore, during the operation phase, the strategy execution module decomposes the supply strategy into logistics and distribution instructions, warehouse outbound instructions, and personnel scheduling and task assignment instructions, and sends the instructions to the corresponding automated logistics control system, warehouse management system, and human resource management system. It also receives feedback from each execution system in real time and tracks the execution progress of the supply tasks.

[0014] Furthermore, the central management module is interconnected with the data acquisition module, the location analysis module, and the inventory monitoring module via a wireless network; the feature extraction module is interconnected with the data acquisition module and the risk prediction module via a wireless network; and the supply decision module is interconnected with the location analysis module, the inventory monitoring module, the risk prediction module, and the strategy execution module via a wireless network.

[0015] (III) Beneficial Effects Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects: 1. By deeply integrating dynamic data such as real-time environment, historical gas consumption and purchase and sale contracts, and using a predictive model built with deep learning algorithms, the system automatically learns patterns from complex data and outputs the demand achievement risk coefficient for each specific location in future continuous cycles. As a result, managers can clearly identify high-risk periods and locations, providing reliable data support for the early formulation of intervention measures.

[0016] 2. By taking the predicted risk coefficient, location urgency ranking, and real-time inventory status as inputs, and using a multi-objective optimization algorithm for global calculation, the system can automatically generate the optimal supply sequence and supply allocation scheme for multiple future cycles, while meeting hard constraints such as inventory and transportation. This ensures that limited natural gas resources are prioritized for the most urgent and highest-risk links, thereby improving the overall resilience and security of the supply chain and achieving scientific resource scheduling. Attached Figure Description

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

[0018] Figure 1 This is a schematic diagram of the framework of the present invention.

[0019] The numbers in the diagram represent: 1. Central Management Module; 2. Data Acquisition Module; 3. Feature Extraction Module; 4. Risk Prediction Module; 5. Location Analysis Module; 6. Inventory Monitoring Module; 7. Supply Decision Module; 8. Strategy Execution Module. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] The present invention will be further described below with reference to embodiments.

[0022] This embodiment provides a natural gas purchase, sales, and personnel collaborative management system, such as... Figure 1 As shown, it includes: Central Management Module 1 is used to provide data access and instruction editing permissions for each functional module.

[0023] Data acquisition module 2 is used to acquire real-time and historical environmental data, natural gas usage data, and current and historical natural gas purchase and sale contract signing data for various locations in the current area, and transmit the data to feature extraction module 3. Environmental data includes temperature, air pressure, humidity, precipitation, and special weather events. Natural gas usage data includes gas consumption, peak and valley values, gas-using equipment load rate, and user type distribution for various locations at different times. Purchase and sale contract signing data includes the supply volume, supply cycle, price terms, performance constraints, and historical performance records stipulated in the contract.

[0024] Feature extraction module 3 is used to extract real-time natural gas demand features related to the current and future time periods from environmental data and usage data, and to extract purchase and sale features related to the current and future time periods from purchase and sale contract signing data, and output the demand features and purchase and sale features to risk prediction module 4. The process of providing natural gas demand characteristics includes: Historical natural gas usage data is processed using a time series decomposition algorithm to separate long-term trend components, seasonal components, and residual components. The seasonal components include annual, monthly, weekly, and daily cycles. For real-time usage data, the sliding window technique is used to calculate short-term fluctuation characteristics, including the moving average, standard deviation, rate of change, and peak-to-valley ratio of recent gas consumption, forming a time series feature vector that characterizes the basic laws of gas consumption behavior. Key meteorological factors related to gas consumption were screened from environmental data. The core factors included temperature, humidity, wind speed and solar radiation intensity. The marginal impact coefficient of each meteorological factor on gas consumption in different numerical ranges was quantified. For future cycles, the expected gas consumption adjustment value driven by meteorological factors was obtained. The time-series feature vector is fused with the meteorological-driven forecast gas consumption adjustment value. Combined with special event information reported by various locations or collected by sensors, event impact factors are set. Based on the fused data, a real-time natural gas demand feature vector for future continuous time periods is generated. The vector includes the basic forecast value, as well as quantitative indicators such as trend strength, seasonal coefficient, meteorological sensitivity and event disturbance. The process of extracting purchase and sales characteristics includes: For the purchase and sale contract signing data, natural language processing technology is used to extract key information, extract the supplier, demander, contract term, agreed supply volume for each period, benchmark price, price adjustment formula, delivery location, performance bond clause and liability for breach of contract clause, establish the relationship between the above entities, and transform the unstructured contract text into a structured contract element knowledge graph. Based on structured contract elements, the reliability coefficient of contract supply in future periods is calculated by weighting the supply stability characteristics, combined with the supplier's historical performance rate, average delay days, current production capacity, and the severity of default clauses in the contract. Obtain the historical price fluctuation range for each period, and define the price risk value according to a preset scale based on the fluctuation amplitude; The complete lifecycle of each purchase and sale contract is mapped onto a unified timeline; key performance characteristics of the contract in future consecutive periods are extracted, such as the approaching payment node, the supply ramp-up / decline phase, the price review window, and the contract renewal period; the time-point characteristic information is combined with the reliability coefficient and price risk value to form a purchase and sale characteristic vector arranged by time period. This vector comprehensively reflects the supply guarantee capability, cost pressure, and key performance events from the contract level in each future period.

[0025] Risk prediction module 4 is used to build a purchase and sales prediction model through deep learning algorithms. The model takes demand characteristics and purchase and sales characteristics as inputs and predicts the risk coefficient of the demand volume of each location in the region in the future for multiple consecutive time periods.

[0026] Location analysis module 5 is used to store and analyze the attribute information of each location within the region. Based on preset urgency evaluation rules, it independently sorts the supply urgency of each location and generates an urgency ranking list. The attribute information includes: location type, user scale, whether it is critical infrastructure, historical emergency response level, and political or economic importance weight. The urgency evaluation rules are to configure different weights according to each attribute information, dynamically calculate the urgency score of each location through a weighted scoring algorithm, and rank them accordingly.

[0027] Inventory monitoring module 6 is used to monitor and obtain the current total inventory and inventory distribution data of the natural gas system in the region in real time.

[0028] Supply decision module 7 is used to generate a globally optimized supply strategy for the region in the future continuous time periods based on the current inventory and distribution status, combined with the risk coefficients and urgency ranking lists of various locations in the future continuous time periods, through a preset decision algorithm. The strategy specifically includes the supply sequence and supply quantity allocation scheme of various locations in different periods. The decision algorithm is a multi-objective optimization algorithm, and its optimization objective function considers the following constraints: minimizing overall supply risk, maximizing inventory utilization efficiency, prioritizing high urgency locations, and smoothing supply fluctuations. Based on the current inventory status, the future period predicted risk coefficient matrix, and the urgency ranking list, the decision algorithm iteratively solves and outputs the optimal supply sequence and supply quantity allocation matrix of various locations that satisfy all constraints in multiple future time periods.

[0029] The strategy execution module 8 is used to convert the supply strategy into corresponding scheduling instructions and drive the associated logistics, warehousing and human resources allocation systems to execute supply tasks. During the operation phase, the strategy execution module 8 decomposes the supply strategy into logistics distribution instructions, warehouse outbound instructions and human resources scheduling and task assignment instructions, sends the instructions to the corresponding automated logistics control system, warehouse management system and human resources management system, receives feedback status from each execution system in real time, tracks the execution progress of supply tasks, and reports any deviation from the plan or abnormal situation to the central management module 1 in real time.

[0030] The central management module 1 is interconnected with the data acquisition module 2, the location analysis module 5, and the inventory monitoring module 6 via a wireless network. The feature extraction module 3 is interconnected with the data acquisition module 2 and the risk prediction module 4 via a wireless network. The supply decision module 7 is interconnected with the location analysis module 5, the inventory monitoring module 6, the risk prediction module 4, and the strategy execution module 8 via a wireless network.

[0031] At other levels, this embodiment also provides a process for constructing a purchase and sales forecasting model, including: Collect and organize environmental data sequences, natural gas usage data sequences, purchase and sale contract signing data sequences, and actual natural gas supply results data for various locations in the region during historical time periods. Clean the collected raw data, including handling missing values ​​and removing outliers, and standardize or normalize data from different sources and with different dimensions to form a historical training dataset. Based on the historical training dataset, feature extraction is performed through feature extraction module 3 to generate demand feature vectors and purchase and sales feature vectors for each historical time point, which are used as input features for the model. The deviation between the actual natural gas demand and the planned supply in each location, or the occurrence and severity of supply interruption events, in one or more subsequent consecutive time periods are quantified into continuous risk values, which are used as target labels for model training. A hybrid deep neural network is constructed as the backbone of the sales forecasting model. This network includes: a temporal feature processing subnetwork, a feature processing subnetwork, a feature fusion layer, and a multi-period prediction output layer. The temporal feature processing subnetwork processes demand feature sequences with strong time dependencies; this subnetwork is composed of multiple stacked long short-term memory units or gated recurrent units. The feature processing subnetwork processes structured features related to sales contracts; this subnetwork consists of multiple fully connected layers. The feature fusion layer concatenates and interacts with the high-level temporal abstract features output by the temporal subnetwork and the high-level semantic features output by the structured subnetwork. The multi-period prediction output layer consists of multiple parallel fully connected branches, each corresponding to a future prediction period, outputting the demand achievement risk coefficient for the location corresponding to that period. The preprocessed historical dataset is divided into training, validation, and test sets. The initial model is trained using the training set data, and the loss function between the predicted risk coefficient and the true risk label is minimized through backpropagation algorithm and optimizer. During training, the performance of the model is monitored using the validation set, and early stopping is used to prevent overfitting. After training, the final prediction accuracy, recall, and generalization ability of the model are evaluated using an independent test set. The optimal model parameters that have passed training and validation are solidified, and the complete model architecture and parameters are deployed to the risk prediction module 4 to complete the initial construction of the purchase and sales prediction model, enabling it to receive real-time feature inputs and perform online predictions. Actual supply execution results data and corresponding real-world environment and usage data are collected periodically and compared with the prediction results as new samples to calculate the prediction deviation. Based on the deviation, incremental learning or online learning methods are used to automatically adjust and optimize the parameters of the purchase and sales prediction model to improve the long-term prediction accuracy and adaptability of the model.

[0032] This embodiment uses a hybrid neural network to simultaneously analyze the highly periodic environmental and gas demand sequences, as well as the complex contract text features that characterize commercial risks. It also utilizes a feature fusion layer for interactive analysis, thereby enabling a more comprehensive assessment of supply risks caused by the combined effects of market fluctuations, performance obstacles, and sudden changes in demand. This provides reliable risk coefficient outputs for dynamic supply decisions across multiple periods.

[0033] In summary, this invention achieves a quantitative assessment of the risk of future natural gas demand by deeply integrating environmental, gas consumption, and contract data and using deep learning to build a predictive model. It can dynamically identify potential supply risk points and their evolution trends in different locations within a region, and conduct collaborative intelligent analysis of risk prediction, location attribute urgency ranking, and real-time inventory status. Through multi-objective optimization algorithms, it automatically generates the globally optimal supply sequence and supply allocation strategy, improving the predictability of supply security and the scientific nature of decision-making, and effectively reducing the risk of supply disruptions or resource waste caused by supply-demand imbalances. Automated scheduling significantly reduces the workload of management personnel, improves the overall response speed and collaboration efficiency of the supply chain, and realizes full-link management from risk warning and intelligent decision-making to execution monitoring, providing strong intelligent support for the safe operation of natural gas purchase and sale business.

[0034] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A natural gas purchase, sales, and personnel collaborative management system, characterized in that, include: The central management module provides data access and command editing permissions for each functional module; The data acquisition module is used to acquire real-time and historical environmental data, natural gas usage data, and current and historical natural gas purchase and sale contract signing data for various locations in the current area. The feature extraction module is used to extract real-time natural gas demand features that are related to the current and future time periods from environmental data and usage data, and to extract purchase and sale features for the current and future time periods from purchase and sale contract signing data. The risk prediction module is used to build a purchase and sales prediction model through deep learning algorithms. The model takes demand characteristics and purchase and sales characteristics as inputs and predicts the risk coefficient of the demand for each location in the region in the future for multiple consecutive time periods. The location analysis module is used to store and analyze the attribute information of each location within the region. Based on the preset urgency evaluation rules, it independently sorts the supply urgency of each location and generates an urgency ranking list. The inventory monitoring module is used to monitor and obtain the current total inventory and inventory distribution data of the natural gas system in the region in real time. The supply decision module is used to generate a globally optimized supply strategy for the region in the future continuous time period based on the current inventory and distribution status, combined with the risk coefficient and urgency ranking list of each location in the future continuous period, through a preset decision algorithm. The strategy execution module is used to translate supply strategies into corresponding scheduling instructions and drive the associated logistics, warehousing and personnel allocation systems to execute supply tasks.

2. The natural gas purchase, sales, and personnel collaborative management system according to claim 1, characterized in that, The environmental data in the data acquisition module includes temperature, air pressure, humidity, precipitation, and special weather events; the natural gas usage data includes gas consumption, peak and valley values, load rate of gas-using equipment, and user type distribution at various locations and time periods; the purchase and sale contract signing data includes the supply quantity, supply cycle, price terms, performance constraints, and historical performance records stipulated in the contract.

3. The natural gas purchase, sales, and personnel collaborative management system according to claim 1, characterized in that, The process of providing natural gas demand features in the feature extraction module includes: For historical natural gas usage data, long-term trend components, seasonal components, and residual components are separated; for real-time usage data, short-term fluctuation characteristics are calculated, including the moving average, standard deviation, rate of change, and peak-to-valley ratio of recent gas consumption, forming a time-series feature vector that characterizes the basic laws of gas consumption behavior. Key meteorological factors related to gas consumption are screened from environmental data, and the marginal impact coefficients of each meteorological factor on gas consumption in different numerical ranges are quantified. For future cycles, the expected gas consumption adjustment value driven by meteorological factors is obtained. By fusing the time-series feature vector with the meteorological-driven forecast gas consumption adjustment value, a real-time natural gas demand feature vector for future continuous time periods is generated based on the fused data.

4. The natural gas purchase, sales, and personnel collaborative management system according to claim 1, characterized in that, The process of extracting purchase and sales features in the feature extraction module includes: For the purchase and sale contract signing data, natural language processing technology is used to extract key information, extract the supplier, demander, contract term, agreed supply volume for each period, benchmark price, price adjustment formula, delivery location, performance bond clause and liability for breach of contract clause, establish the relationship between the above entities, and transform the unstructured contract text into a structured contract element knowledge graph. Based on structured contract elements, the reliability coefficient of contract supply in future periods is calculated by weighting the supply stability characteristics, combined with the supplier's historical performance rate, average delay days, current production capacity, and the severity of default clauses in the contract. Obtain the historical price fluctuation range for each period, and define the price risk value according to a preset scale based on the fluctuation amplitude; The complete lifecycle of each purchase and sale contract is mapped onto a unified timeline; key performance point features of the contract are extracted for future consecutive periods; and the time point feature information is combined with reliability coefficients and price risk values ​​to form a purchase and sale feature vector arranged according to time periods.

5. A natural gas purchase, sales, and personnel collaborative management system according to claim 1, characterized in that, The process of constructing the purchase and sales forecasting model in the risk forecasting module includes: Collect and organize the environmental data sequences, natural gas usage data sequences, purchase and sale contract signing data sequences, and actual natural gas supply results data for various locations in the region during historical time periods to form a historical training dataset; Based on the historical training dataset, feature extraction is performed through the feature extraction module to generate demand feature vectors and purchase and sales feature vectors for each historical time point, which are used as input features for the model. The deviation between the actual natural gas demand and the planned supply in each location in one or more subsequent consecutive time periods is quantified into a continuous risk value, which is used as the target label for model training. A hybrid deep neural network is constructed as the backbone of the purchase and sales prediction model; the network includes: a temporal feature processing subnetwork, a feature processing subnetwork, a feature fusion layer, and a multi-period prediction output layer; The preprocessed historical dataset is divided into training, validation, and test sets; the initial model is trained using the training set data; after training, the model's final prediction accuracy, recall, and generalization ability are evaluated using an independent test set. The optimal model parameters that have passed training and validation are solidified, and the complete model architecture and parameters are deployed to the risk prediction module to complete the initial construction of the purchase and sales prediction model.

6. The natural gas purchase, sales, and personnel collaborative management system according to claim 5, characterized in that, The temporal feature processing subnetwork is used to process demand feature sequences with strong time dependence. This subnetwork is composed of multiple layers of long short-term memory units or gated recurrent units stacked together. The feature processing subnetwork is used to process structured features related to purchase and sale contracts. This subnetwork is composed of multiple fully connected layers. The feature fusion layer is used to concatenate and interact the high-level temporal abstract features output by the temporal subnetwork with the high-level semantic features output by the structured subnetwork. The multi-period prediction output layer consists of multiple parallel fully connected branches. Each branch corresponds to a future prediction period and outputs the demand achievement risk coefficient for the location corresponding to that period.

7. The natural gas purchase, sales, and personnel collaborative management system according to claim 1, characterized in that, The attribute information in the location analysis module includes: location type, user scale, whether it is critical infrastructure, historical emergency response level, and political or economic importance weight; the urgency evaluation rule is to configure different weights according to each attribute information, dynamically calculate the urgency score of each location through a weighted scoring algorithm, and rank them accordingly.

8. The natural gas purchase, sales, and personnel collaborative management system according to claim 1, characterized in that, The decision-making algorithm in the supply decision-making module is a multi-objective optimization algorithm. Its optimization objective function simultaneously considers the following constraints: minimizing overall supply risk, maximizing inventory utilization efficiency, prioritizing high-urgency locations, and smoothing supply fluctuations. Based on the current inventory status, the future cycle prediction risk coefficient matrix, and the urgency ranking list, the decision-making algorithm iteratively solves the problem and outputs the optimal supply sequence and the supply allocation matrix for each location that satisfy all constraints over multiple future time periods.

9. A natural gas purchase, sales, and personnel collaborative management system according to claim 1, characterized in that, During the operation phase, the strategy execution module decomposes the supply strategy into logistics and distribution instructions, warehouse outbound instructions, and personnel scheduling and task assignment instructions. It sends the instructions to the corresponding automated logistics control system, warehouse management system, and human resource management system, and receives feedback status from each execution system in real time to track the execution progress of the supply tasks.

10. A natural gas purchase, sales, and personnel collaborative management system according to claim 1, characterized in that, The central management module is interconnected with the data acquisition module, the location analysis module, and the inventory monitoring module via a wireless network. The feature extraction module is interconnected with the data acquisition module and the risk prediction module via a wireless network. The supply decision module is interconnected with the location analysis module, the inventory monitoring module, the risk prediction module, and the strategy execution module via a wireless network.

Images