Coal market data processing method and system based on time series modeling

Abstract

This invention discloses a coal market data processing method and system based on time series modeling, belonging to the field of data processing. The method includes: establishing a coal demand change model based on enterprise coal datasets; constructing a coal procurement knowledge graph and performing supply chain risk propagation simulation to obtain supply risk propagation paths; mapping the impact of procurement rhythm and making procurement control decisions based on the propagation paths and demand change model to determine an initial strategy; performing transportation fluctuation prediction and time lag cumulative effect mapping on the initial strategy to obtain a lag compensation guidance tensor; expanding the safety stock range and mapping the inventory constraint sensitivity of the initial strategy to obtain a safety elasticity guidance tensor; and optimizing the initial strategy based on the compensation guidance tensor and the elasticity guidance tensor. This application solves the technical problem in the prior art that it is difficult to capture the dynamic changes in enterprise coal demand, leading to an imbalance in coal-using enterprise inventory management and affecting the production safety and operational efficiency of coal-using enterprises.

Description

Technical Field

[0001] This invention relates to the field of data processing, specifically to a method and system for processing coal market data based on time series modeling. Background Technology

[0002] As a fundamental commodity in the energy system, the scientific nature of coal procurement and regulation directly determines the continuity of production and operation, supply chain stability, and cost control capabilities of downstream coal-consuming enterprises. Currently, the supply and demand pattern of the coal market is significantly dynamic and uncertain due to multiple factors such as capacity release, transportation capacity fluctuations, and extreme weather. Most existing coal market data processing and procurement regulation solutions rely solely on historical coal consumption data to conduct static, single-dimensional demand forecasts and formulate standardized procurement plans with fixed cycles and fixed batches. This can easily lead to a series of operational risks such as production disruptions, inventory backlogs, and soaring procurement costs, making it difficult to adapt to the procurement management needs of the complex and ever-changing coal market environment. Summary of the Invention

[0003] This application provides a coal market data processing method and system based on time series modeling, aiming to solve the technical problem that existing technologies struggle to capture the dynamic changes in enterprises' coal demand, leading to imbalances in coal-using enterprises' inventory management and affecting their production safety and operational efficiency.

[0004] In view of the above problems, this application provides a coal market data processing method and system based on time series modeling.

[0005] The first aspect disclosed in this application provides a method for processing coal market data based on time series modeling, the method comprising: A time-series model is built based on the target company's coal dataset to establish a coal demand change model. A coal procurement knowledge graph is constructed based on the coal market dataset, and the supply chain risk propagation is simulated for the target company based on the coal procurement knowledge graph to obtain the coal supply risk propagation path. Based on the coal supply risk propagation path and the coal demand change model, the impact of procurement rhythm on the target company is mapped and procurement control decisions are made to determine the initial coal procurement control strategy. The initial coal procurement control strategy is then subjected to transportation fluctuation prediction and time lag cumulative effect mapping to obtain a transportation lag compensation guidance tensor. The initial coal procurement control strategy is further subjected to safety stock range expansion and inventory constraint sensitivity mapping to obtain an inventory safety elasticity guidance tensor. The initial coal procurement control strategy is then subjected to multi-dimensional coupling optimization based on the transportation lag compensation guidance tensor and the inventory safety elasticity guidance tensor.

[0006] Another aspect of this application discloses a coal market data processing system based on time series modeling, the system comprising: The system comprises the following modules: a modeling module for time-series modeling based on the target company's coal dataset and establishing a coal demand change model; a deduction module for constructing a coal procurement knowledge graph based on the coal market dataset and performing supply chain risk propagation deduction on the target company based on the knowledge graph to obtain the coal supply risk propagation path; a strategy module for mapping the impact of procurement rhythm on the target company and making procurement control decisions based on the coal supply risk propagation path and the coal demand change model to determine the initial coal procurement control strategy; a prediction module for predicting transportation fluctuations and mapping the time lag cumulative effect on the initial coal procurement control strategy to obtain a transportation lag compensation guidance tensor; an extension module for expanding the safety stock range and mapping the inventory constraint sensitivity on the initial coal procurement control strategy to obtain an inventory safety elasticity guidance tensor; and an optimization module for performing multi-dimensional coupling optimization of the initial coal procurement control strategy based on the transportation lag compensation guidance tensor and the inventory safety elasticity guidance tensor.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: By conducting time-series modeling on a full-dimensional coal dataset of target enterprises, and through time alignment, supply and demand flow relationship calibration, and supply and demand balance inversion, the implicit demand sequence of enterprises is obtained. Then, through structural decomposition and alignment fusion modeling, a coal demand change model is constructed that can simultaneously and accurately fit the enterprise's steady-state baseline demand and effectively capture sudden fluctuations in demand. Simultaneously, based on a full-dimensional coal market dataset, a coal procurement knowledge graph covering the main entities and business relationships across the entire industry chain is constructed. Based on this, supply chain risk propagation is extrapolated. Through reverse tracing, contribution calculation, priority ranking, and minimum path selection, the path of coal supply risk propagation is accurately located. This is then combined with… The risk propagation path and demand change model completes the mapping of the impact of procurement rhythm and the initial procurement control strategy. Further, based on the initial strategy, transportation fluctuation prediction and time lag cumulative effect mapping are carried out to generate a transportation lag compensation guiding tensor. Safety stock range expansion and inventory constraint sensitivity mapping are carried out to generate an inventory safety elasticity guiding tensor. Finally, based on the two guiding tensors, the multi-constraint, multi-objective, and multi-dimensional coupling optimization of the initial strategy is completed, which improves the scientific nature and risk resistance of coal procurement control for coal-using enterprises. It not only ensures the continuity and stability of coal use in enterprise production and operation, but also effectively reduces the comprehensive operating costs of the entire coal procurement, transportation, and inventory chain.

[0008] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0009] Figure 1 A flowchart illustrating a coal market data processing method based on time series modeling is provided for embodiments of this application. Figure 2 A schematic diagram of the structure of a coal market data processing system based on time series modeling is provided for embodiments of this application.

[0010] Figure labeling: Modeling module 11, Inference module 12, Strategy module 13, Prediction module 14, Extension module 15, Optimization module 16. Detailed Implementation

[0011] To further illustrate the technical means and effects of the present invention in achieving the intended purpose, the following detailed description of the specific implementation methods, structures, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0012] The overall concept of the technical solution provided in this application is as follows: This application provides a coal market data processing method and system based on time series modeling. By performing time series modeling on the full-dimensional coal dataset of the target enterprise, it completes the calibration of the enterprise's coal supply and demand relationship, the inversion of supply and demand balance, and the decomposition of the demand sequence structure. It constructs a coal demand change model that can simultaneously and accurately fit steady-state benchmark demand and capture sudden fluctuations in demand. Simultaneously, it constructs a coal procurement knowledge graph with time-series dynamic characteristics based on the full-dimensional coal market dataset, conducts supply chain risk propagation simulation, and accurately locates the core propagation path of coal supply risk. On this basis, it completes the mapping of the impact of procurement rhythm and the formulation of initial procurement control strategies. Furthermore, it generates a transportation lag compensation guiding tensor by mapping transportation fluctuation prediction and time lag cumulative effect, and generates an inventory safety elasticity guiding tensor by mapping safety inventory range expansion and inventory constraint sensitivity. Finally, based on the two guiding tensors, it completes multi-constraint, multi-objective, and multi-dimensional coupled optimization of procurement strategies, comprehensively ensuring the coal supply and demand balance of coal-using enterprises throughout the entire cycle and improving the risk resistance capability of coal market data processing.

[0013] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0014] Example 1, as Figure 1 As shown in the embodiments of this application, a coal market data processing method based on time series modeling is provided. The method includes: S100: Based on the target company's enterprise coal dataset, perform time series modeling to establish a coal demand change model.

[0015] Specifically, the first step is to perform time alignment and supply-demand relationship labeling operations based on the target company's enterprise coal dataset, which is a collection of time-series business data generated by the target company throughout its entire production and operation cycle, including coal procurement and warehousing, production consumption, inventory changes, production and sales plans, and historical coal consumption ledgers. Time alignment refers to unifying coal business data from different sources and with different sampling frequencies (daily, weekly, and monthly) in the dataset to the same time scale, marking each node in the entire chain of coal flow from warehousing, inventory caching, production requisition to final consumption, clarifying the coal inflow and outflow logic, quantity correspondence, and time connection rules between each node, and finally generating an enterprise coal supply and demand relationship diagram, which is a topological structure diagram that visually presents the quantity flow and time connection relationship between the target company's coal supply side, inventory caching side, and production demand side.

[0016] Based on this, the supply and demand balance inversion of the target enterprise is performed according to the generated enterprise coal supply and demand relationship diagram to obtain the implicit demand sequence. The supply and demand balance inversion is constrained by the coal material conservation formula of beginning inventory + current purchase volume - current consumption volume = ending inventory. It reverse-derives the potential coal demand in the enterprise's production and operation that is not directly reflected in the explicit purchase plan and coal consumption ledger, including unplanned coal demand caused by factors such as production plan adjustment, equipment maintenance, and sudden changes in working conditions. The implicit demand sequence is the enterprise's full-cycle potential coal demand data sequence obtained through supply and demand balance inversion and arranged according to a unified time scale.

[0017] Subsequently, the obtained implicit demand sequence is structurally decomposed to obtain steady-state demand subsequences and fluctuating demand subsequences. Structural decomposition refers to using a time series decomposition algorithm to break down the implicit demand sequence into subsequences with different fluctuation characteristics, separating the trend and periodic deterministic components and the random disturbance and uncertainty components from the demand data. The steady-state demand subsequence is a coal demand sequence decomposed from the implicit demand sequence, driven by deterministic factors such as the enterprise's normalized production and operation plan, fixed production capacity, and benchmark coal consumption quota, and has a stable trend and fixed periodicity. It constitutes the core benchmark part of the enterprise's coal demand. The fluctuating demand subsequence is a coal demand sequence decomposed from the implicit demand sequence, driven by non-deterministic factors such as market order fluctuations, seasonal changes in coal consumption, sudden production adjustments, and equipment failures, and has no fixed period, exhibiting random fluctuation characteristics.

[0018] Finally, time series alignment and fusion modeling is performed on the steady-state demand subsequence and the fluctuating demand subsequence to output the final coal demand change model. The time series alignment and fusion modeling refers to realigning the two types of feature subsequences after decomposition to the same time scale, adapting the corresponding time series prediction algorithms to the fluctuation characteristics of the two types of subsequences respectively, and then integrating the output results of the two types of algorithms through weighted fusion to construct a full-dimensional dynamic prediction model that can accurately fit the normal benchmark coal consumption trend of enterprises and effectively capture sudden fluctuations in demand.

[0019] S200: Construct a coal procurement knowledge graph based on the coal market dataset, and perform supply chain risk propagation simulation on the target enterprise based on the coal procurement knowledge graph to obtain the coal supply risk propagation path.

[0020] Specifically, the first step is to use a coal market dataset, which is a multi-source time-series dataset covering the entire coal industry chain, including quality and price parameters of different coal types, supplier qualifications and historical transaction data, port throughput capacity and scheduling plans, transportation capacity configuration and timeliness data of transportation routes, inventory turnover data of each node, and historical performance anomalies and risk event data of the entire chain. Entity identification is then performed to extract a set of coal market entities. Entity identification refers to the use of natural language processing and structured data extraction technologies to accurately capture the core participants in the entire coal procurement chain from unstructured and structured market data. The set of coal market entities covers six major categories of core entities: coal type, supplier, port, transportation route, inventory node, and coal user. Each entity is set as a graph node, which is the basic unit of the coal procurement knowledge graph, used to store the full static attributes and dynamic time-series data of the corresponding entity.

[0021] Based on this, the procurement, transportation, delivery, and inventory replenishment relationships among all graph nodes are identified using the coal market dataset. These business relationships are set as graph edges, which are the connecting carriers used to represent the business flow logic, data interaction direction, and supply and demand dependence between nodes. At the same time, each graph edge is configured with corresponding edge attributes, which include six quantitative indicators: transportation cycle, delivery deviation, supply fulfillment rate, transportation capacity, inventory buffer capacity, and number of historical anomalies.

[0022] Subsequently, temporal attribute alignment and graph structure integration are performed on all graph nodes, graph edges and corresponding edge attributes. Temporal attribute alignment refers to unifying multi-source heterogeneous data from different nodes and edges to the same time scale, ultimately generating a coal procurement knowledge graph with full-link topology visualization, multi-dimensional parameter quantification and temporal dynamic deduction characteristics. Subsequently, based on this graph, a supply chain risk propagation simulation was conducted for the target enterprise. First, the risk triggering events of the target enterprise were identified, namely, various abnormal events that may cause coal supply disruptions and supply-demand imbalances, including but not limited to abnormal drops in coal inventory below the safety threshold, abnormal performance of core suppliers, transportation line interruptions, and abnormal port scheduling. The target node corresponding to the risk triggering event in the coal procurement knowledge graph was then located, which is the core entity node directly affected by the risk event. Then, based on the topology of the coal procurement knowledge graph, a reverse tracing search was conducted on the target node, that is, along the business dependency logic of the graph edges, tracing back from the target node to all entity nodes with direct or indirect supply-demand relationships, and obtaining multiple related upstream nodes.

[0023] Next, based on the edge attribute data of the coal procurement knowledge graph, the node target contribution of each associated upstream node to the target node is calculated. This involves quantifying parameters such as the upstream node's share of coal supply, contract performance stability, and irreplaceability to obtain the node's impact weight on the target company's coal supply security. Subsequently, based on the calculated node target contributions, multiple potential risk propagation paths from associated upstream nodes to the target node are prioritized and screened. The causation priority is determined by the sum of the target contributions of all nodes on a single risk propagation path. The higher the value, the stronger the causation of the path and the greater its impact on the target company's supply risk. Non-critical paths with extremely low contribution and no substantial impact are eliminated through sorting and screening to obtain the set of critical causation paths. Then, the set of critical causation paths is screened for minimum path length. That is, from the set of highly causative paths, the path with the fewest nodes and the shortest transmission link is selected. This type of path has the fastest risk transmission speed and the most direct impact, ultimately generating the coal supply risk propagation path for the target company.

[0024] S300: Based on the coal supply risk propagation path and the coal demand change model, the target enterprise is mapped to the impact of procurement rhythm and procurement control decisions are made to determine the initial strategy for coal procurement control.

[0025] Specifically, firstly, based on the risk propagation path of coal supply, the impact of procurement rhythm on target enterprises is mapped. This involves quantifying the risk parameters corresponding to each key node and related edge attribute in the risk propagation path, transforming potential supply risks in the supply chain into quantifiable impacts on the core execution elements of the entire coal procurement process, and outputting the first result of the procurement impact mapping: a quantitative matrix of risk impacts on each element of the procurement rhythm covering the entire time cycle under the supply-side risk dimension. Simultaneously, a coal demand change model is used to map the impact of procurement rhythm on target enterprises, outputting the second result of the procurement impact mapping: a quantitative scheme for the procurement rhythm benchmark required to achieve full-cycle coal supply and demand balance under the demand-side dimension, defining the bottom line of demand that cannot be breached in procurement execution.

[0026] Subsequently, the execution time of the first and second results of the procurement impact mapping were aligned, superimposed, and integrated. Time alignment refers to unifying the quantitative results of supply-side risk impact with the quantitative scheme of demand-side procurement benchmarks to a time scale completely consistent with the aforementioned time series modeling. Superimposition and integration means that the quantitative results of the two dimensions are bidirectionally adapted and superimposed, with the supply-demand balance on the demand side as the core bottom-line constraint and the risk hedging on the supply side as the core adjustment target. Invalid content that conflicts between risk adjustment and demand constraints is eliminated, and effective control items that simultaneously meet the requirements of demand protection and risk hedging are retained. Finally, the third result of the procurement impact mapping is output, which is a set of quantitative control targets for procurement rhythm that covers the entire time window and takes into account both rigid demand protection and supply risk hedging. This clarifies the core objectives and adjustment boundaries that need to be achieved in procurement execution within each time period.

[0027] Ultimately, based on the third result of the procurement impact mapping, procurement control decisions are made for the target enterprise. This involves initial decisions based on three key procurement factors: coal procurement frequency, coal procurement volume, and coal replenishment time window. Coal procurement frequency refers to the number of coal procurement orders placed within a fixed period. The initial decision is based on the supply stability and demand fluctuation frequency of the risk path. For high-risk, high-volatility scenarios, procurement cycles are broken down and procurement frequency is increased to reduce the risk of supply-demand imbalance caused by a single supply disruption. For low-risk, steady-state demand scenarios, a fixed benchmark procurement frequency is set to control procurement execution costs. Coal procurement volume refers to the quantity of coal purchased in a single procurement order. The decision-making process is divided into two parts: baseline batch and flexible batch. The baseline batch matches the normal coal consumption requirements of the steady-state demand subsequence, while the flexible batch matches the incremental demand and safety reserve requirements for risk hedging of the fluctuating demand subsequence. The coal replenishment time window refers to the optimal time interval for placing a purchase order, locking in the source of goods, and receiving the goods into the warehouse. Its initial decision is based on the inventory consumption node on the demand side, the transportation cycle and delivery deviation of the risk path on the supply side, reserving a risk buffer time in advance, setting the optimal replenishment trigger and completion interval, and finally integrating the initial decision results of the three factors to generate the initial coal procurement control strategy, that is, the initial coal procurement execution plan that simultaneously matches the enterprise's coal consumption demand and hedges the supply chain risks.

[0028] S400: Perform transportation fluctuation prediction and time lag cumulative effect mapping on the initial strategy for coal procurement control to obtain the transportation lag compensation guidance tensor.

[0029] Specifically, the initial strategy for coal procurement control, including the target enterprise's full-cycle coal procurement frequency, procurement batch, and replenishment time window, is used as the basis for procurement execution, matching the enterprise's coal demand with the core risks of the supply chain. This is combined with the edge attribute data of the corresponding transportation links in the coal procurement knowledge graph, including transportation cycle, arrival deviation, supply fulfillment rate, transportation capacity, and historical anomaly frequency, as well as the core transportation risk nodes in the coal supply risk propagation path. This allows for transportation fluctuation prediction, which involves probabilistically predicting the full-cycle arrival time deviation for each batch of procurement in the initial strategy, corresponding to the transportation route, departure node, arrival node, and planned time. Finally, the coal procurement transportation delay distribution is obtained, which is the set of probability distributions of the arrival delay duration of each procurement batch in different time windows, output by the transportation fluctuation prediction. This quantifies the magnitude, timing, and probability of transportation delay risks faced by different batches of procurement.

[0030] Based on this, time window mapping and batch association calibration are carried out on the obtained coal procurement and transportation delay distribution. Time window mapping refers to accurately mapping the transportation delay distribution of each procurement batch to the replenishment time window, inventory consumption time window and supply and demand balance verification window pre-set in the initial strategy, clarifying the scope and boundaries of the impact of transportation delay in the time dimension. Batch association calibration refers to calibrating the arrival time linkage relationship between adjacent and related procurement batches based on the time sequence connection logic of multiple procurement batches in the initial strategy and the sequential dependence of inventory replenishment, clarifying the transmission relationship of the transportation delay of the preceding batch on the procurement execution and inventory turnover of the subsequent batches. Through these two operations, the time lag transmission relationship is finally obtained, that is, the quantitative logic and impact link of how the transportation delay of a single batch is transmitted to the subsequent procurement cycle and the supply and demand balance link of the whole chain through the overlap of time windows and the inventory connection between batches.

[0031] Subsequently, based on the time lag transmission relationship, a time lag cumulative effect mapping was performed. First, the time lag cumulative effect of the initial coal procurement control strategy was extrapolated. That is, the transportation delay risk of all procurement batches throughout the entire cycle of the initial strategy was simulated in multiple scenarios. The comprehensive impact of single-point and multi-point transportation delays on the target enterprise's coal inventory level, supply and demand balance, and procurement plan execution rate was extrapolated after continuous superposition and amplification in the time dimension. The focus was on capturing the amplification effect of multiple batches and multiple time windows of delays, which is different from the single-point delay effect. Finally, the lag cumulative effect matrix was obtained, which is a two-dimensional quantitative matrix output by the extrapolation. This matrix uses the full cycle time window / procurement batch of the initial strategy as the row dimension and the impact of transportation delays on indicators such as inventory gap, duration of supply and demand imbalance, restocking pressure, and increase in procurement costs as the column dimension. The magnitude of the impact of the cumulative effect of transportation delays at different time nodes and different batches on the entire procurement execution process was quantified.

[0032] Ultimately, compensation allocation is carried out based on the obtained lagged cumulative effect matrix. This follows the principles of prioritizing compensation for high cumulative impacts, focusing on core risk links, and prioritizing the bottom line of supply and demand balance. Procurement adjustment compensation actions are precisely allocated to corresponding time windows, procurement batches, and transportation links, clarifying the compensation direction and magnitude for different links. This generates a transportation lagged compensation guidance tensor, which is a high-dimensional guidance dataset that is ultimately output through compensation allocation and covers multiple parameters. This tensor contains all parameters in the time dimension, procurement batch dimension, transportation link dimension, compensation action dimension, and compensation magnitude dimension. It can guide the targeted adjustment of the initial strategy in the transportation link, clarifying which procurement batch needs to lock in capacity by how much time in advance, which transportation link needs to divert how many procurement batches, which time window needs to supplement with how much safe procurement volume, and which risky transportation route needs to be replaced with an alternative link, etc.

[0033] S500: Expand the safety stock range and map the stock constraint sensitivity for the initial coal procurement control strategy to obtain the stock safety elasticity guidance tensor.

[0034] Specifically, the initial coal procurement control strategy, combined with the demand calculation results of the coal demand change model and the risk quantification data of the coal supply risk propagation path, is used to extract the multi-window projected inventory and multi-window projected consumption of target enterprises. The multi-window refers to a continuous and subdivided rolling time interval that is fully aligned with the previous full-process time series modeling and has a unified time scale. The multi-window projected inventory refers to the theoretical inventory level at the end of each time window, calculated based on the procurement and warehousing plan of the initial strategy and historical inventory turnover patterns. The multi-window projected consumption refers to the benchmark amount and upper limit of coal consumption required for enterprise production and operation within each time window, calculated based on the steady-state demand subsequence and the fluctuating demand subsequence decomposed by the coal demand change model. Together, these two constitute the basic data for inventory dynamic analysis.

[0035] Subsequently, supply and demand fluctuation analysis and time window correlation expansion were conducted on the extracted multi-window projected inventory and multi-window projected consumption to complete the expansion of the safety stock range and prioritize the acquisition of the safety stock elasticity range. The supply and demand fluctuation analysis refers to combining the demand fluctuation characteristics of the coal demand change model, the supply interruption risk of the coal supply risk propagation path, and the probability distribution of arrival delays in the transportation link to quantitatively calculate the deviation range between the projected inventory and the projected consumption in each time window, and clarify the extreme boundaries of inventory gaps and excesses that may be caused by two-way fluctuations on both the supply and demand sides. The time window correlation expansion refers to incorporating the inventory turnover transmission, replenishment action connection, and consumption demand linkage relationship of adjacent and related time windows into the accounting system. The safety stock elasticity range is the dynamic control boundary of the safety stock line. Its lower limit is the minimum inventory red line to ensure that enterprises do not experience production and supply interruption risks, and its upper limit is the highest inventory threshold that takes into account the cost of inventory capital occupation and risk buffering capacity. The fluctuation space between the upper and lower limits is the elastic range of inventory that can be adjusted.

[0036] Based on this, and according to the calculated safety stock elasticity range, an inventory constraint sensitivity mapping is performed on the initial coal procurement control strategy to obtain an inventory constraint sensitivity matrix. The inventory constraint sensitivity mapping refers to using the safety stock elasticity range as the constraint benchmark to quantify the impact of unit changes in the three major procurement elements in the initial strategy—procurement frequency, procurement batch, and replenishment time window—on the deviation of the inventory level from the elasticity range, including the magnitude, direction, and time window. At the same time, combined with external risk variables such as demand fluctuations, supply interruptions, and transportation delays, the sensitivity changes of inventory constraints to changes in procurement parameters under different risk scenarios are calculated to identify the core control levers that have the most significant impact on the inventory safety status and ineffective adjustment items with no substantial impact. The inventory constraint sensitivity matrix is ​​a two-dimensional quantitative matrix obtained through the above mapping. This matrix uses the procurement parameters of the initial strategy as the row dimension and core control indicators such as the magnitude of the deviation of the inventory level from the elasticity range, the probability of supply disruption risk, and the increase in inventory backlog costs as the column dimension, quantifying the full-dimensional impact of different procurement parameter adjustments on the inventory safety status. Finally, based on the obtained inventory constraint sensitivity matrix, inventory risk stratification, replenishment priority calibration, and elastic control weight allocation are carried out sequentially to generate an inventory safety elastic guidance tensor. The inventory safety elastic guidance tensor is the output high-dimensional guidance dataset. This tensor includes all parameters of the time window dimension, procurement parameter dimension, inventory risk dimension, control action dimension, and weight allocation dimension. It can guide the differentiated and elastic adjustment of the initial strategy on the inventory side, clarify the adjustment range of procurement batches in different time windows, the optimization range of replenishment windows, the inventory elasticity space corresponding to different risk levels, and the control priority of each procurement parameter.

[0037] S600: The initial strategy for coal procurement control is optimized through multi-dimensional coupling based on the transportation lag compensation guidance tensor and the inventory safety elasticity guidance tensor.

[0038] Specifically, the first step is to align the temporal dimensions of the two guiding tensors and unify their parameter space mapping. The transportation lag compensation guiding tensor is a high-dimensional dataset including time, procurement batch, transportation link, compensation action, and compensation magnitude dimensions. It guides the initial strategy's pre-emptive compensation adjustments for transportation fluctuations and accumulated time lag risks. The inventory safety elasticity guiding tensor is a high-dimensional dataset including time window, procurement parameter, inventory risk, control action, and weight allocation dimensions. It guides the initial strategy's dynamic elastic control adjustments for inventory safety. Temporal dimension alignment refers to dividing the time scale and rolling time window of the two tensors. This involves fully unifying the system with the benchmark system that is completely consistent with the initial time series modeling and coal procurement control strategy of the previous whole process. This eliminates the deviations of the two tensors in terms of time granularity and window boundary division. Unified parameter space mapping means mapping all the optimization parameters of the two tensors belonging to different business dimensions to the three decision variables of the initial coal procurement control strategy, namely coal procurement frequency, coal procurement batch, and coal replenishment time window, into a unified parameter space. This transforms the compensation actions such as capacity locking, batch diversion, and advance setting on the transportation side, and the control actions such as batch adjustment, replenishment window offset, and elastic range adaptation on the inventory side, into quantifiable adjustment instructions for the three decision variables.

[0039] Based on this, a coupling synergy verification under multiple constraints is conducted. This verification involves using the decision-making criteria of the entire scheme—namely, the rigid bottom line of full-cycle coal supply and demand balance as an inviolable bottom line, the primary goal of pre-emptive prevention and control of core supply chain risks, and the auxiliary constraint of optimal comprehensive cost across the entire procurement and inventory chain—as rules to perform bidirectional cross-verification of the two tensor adjustment instructions mapped to a unified parameter space. This identifies and locates optimization actions with synergistic conflicts, such as the requirement of advance replenishment batches on the transportation side exceeding the upper limit of the inventory safety elasticity range, the conflict between the requirement of multiple batches of small-volume procurement on the inventory side and the scale effect constraint of trunk transportation on the transportation side, and the overcompensation action in high-risk windows exceeding the hard constraints of enterprise procurement funds and warehousing capacity. Simultaneously, it verifies whether all adjustment instructions meet the rigid constraints of full-cycle coal demand in the front-end coal demand change model, the core risk prevention and control requirements of the coal procurement knowledge graph and the supply risk propagation path. Invalid adjustment items that break through core constraints or have irreconcilable synergistic conflicts are eliminated, while valid synergistic adjustment items that simultaneously satisfy the optimization objectives of the two tensors and do not break through the core constraints of the entire process are retained. Finally, the feasible solution domain of the multi-dimensional coupled optimization is defined.

[0040] Subsequently, based on the defined feasible solution domain, a multi-objective, multi-dimensional coupled optimization model was constructed and the optimal solution was obtained. Multi-dimensional coupled optimization refers to the deep integration and linkage of parameters, constraints, and objectives across multiple interrelated and mutually influential business dimensions to achieve synergistic optimization across the entire supply chain and all dimensions. This multi-objective coupled optimization model uses three major decision variables as the optimization objects and sets three priority levels of objective functions: the first priority is to maximize the full-cycle coal supply and demand balance guarantee rate and minimize the risk of supply disruption; the second priority is to maximize the transmission disruption rate of core supply chain risks; and the third priority is to minimize the comprehensive cost of the entire coal procurement chain. Simultaneously, the model incorporates rigid constraints throughout the entire process. The model incorporates constraints on total demand and timing in the coal demand change model, supplier supply capacity and transportation capacity in the coal procurement knowledge graph, time-based compensation constraints in the transportation lag compensation guiding tensor, safety stock elasticity range constraints in the inventory safety elasticity guiding tensor, and hard operational constraints such as the target company's procurement funds and storage capacity limits. Finally, a multi-objective intelligent optimization algorithm is used to solve the model, obtaining a Pareto optimal solution set, which is the equilibrium optimal solution set that cannot optimize other objectives without compromising any priority objective. From this set, the unique optimal solution that meets the target company's operational priorities is selected, and the full-cycle optimal adjustment scheme for the three core decision variables is output.

[0041] Finally, for the optimal adjustment scheme obtained by the solution, dynamic simulation verification and optimization convergence iteration are carried out. That is, the optimized procurement strategy is substituted into the simulation scenario to simulate various risks that may occur in real business, such as demand fluctuations, abnormal supplier performance, transportation line interruptions, and delivery delays. The supply and demand balance guarantee capability, risk hedging effect and comprehensive cost control level of the optimized strategy are deduced. At the same time, the simulation results are compared with the initial strategy to verify the optimization effect. If the simulation results show that there are still uncovered risk gaps or the comprehensive cost exceeds the preset threshold, the simulation deviation results are fed back to the multi-dimensional multi-objective coupled optimization model. The optimization weight allocation of the two guiding tensors is iteratively adjusted, and the solution and simulation verification are completed again until the optimization results are fully converged, that is, simultaneously satisfying the rigid supply and demand bottom line, core risk prevention and control requirements and the comprehensive cost optimization objective. Finally, the optimized final coal procurement control strategy is output.

[0042] Furthermore, in the method provided in the application embodiment, time series modeling is performed based on the target enterprise's coal dataset to establish a coal demand change model, including: performing time alignment and supply-demand flow relationship calibration based on the enterprise coal dataset to obtain an enterprise coal supply-demand relationship diagram; performing supply-demand balance inversion of the target enterprise based on the enterprise coal supply-demand relationship diagram to obtain an implicit demand sequence; performing structural decomposition on the implicit demand sequence to obtain a steady-state demand subsequence and a fluctuating demand subsequence; and performing time series alignment and fusion modeling on the steady-state demand subsequence and the fluctuating demand subsequence to obtain the coal demand change model.

[0043] Specifically, the first step involves time alignment and supply-demand relationship calibration based on the enterprise's coal dataset. Time alignment refers to unifying business data from different sources and with different sampling frequencies, including daily inventory data, weekly procurement data, and monthly production plan data, to the same time scale and granularity, eliminating time-series discrepancies between different data. Supply-demand relationship calibration involves marking each step of the time-aligned data along the entire lifecycle of coal within the enterprise, from procurement to warehousing, inventory buffer turnover, production plan requisition, on-site consumption and write-off, to the end-of-life inventory. The entire node flow logic clarifies the quantity correspondence, time connection rules, business triggering conditions, and historical deviation parameters of coal inflow and outflow between each node, ultimately generating a coal supply and demand relationship diagram for the enterprise, namely a visualized and computable coal supply and demand topology model. This model uses coal warehousing nodes, inventory nodes, requisition nodes, and consumption nodes as core topology nodes, and the flow relationship between nodes as connecting edges. The edge attributes are synchronously configured with quantitative parameters such as flow quantity, time period, write-off rules, and historical performance deviation, fully presenting the dynamic linkage relationship between the enterprise's coal supply side, inventory buffer side, and demand consumption side.

[0044] Based on this, the supply and demand balance inversion of the target enterprise is performed according to the generated enterprise coal supply and demand relationship diagram to obtain the implicit demand sequence. The supply and demand balance inversion is based on the full-link flow logic of the enterprise coal supply and demand relationship diagram. The core rigid constraint is the coal material conservation formula of beginning inventory + current cumulative purchase and warehousing volume - current cumulative actual consumption volume = ending actual inventory. The known quantities are the actual inventory count data, purchase and warehousing data, and consumption and write-off data of the enterprise. The potential coal demand not reflected in the explicit production plan, formal purchase application, and regular coal consumption ledger during the enterprise's production and operation process is deduced in reverse. It covers demand items that are easily missed by traditional accounting, such as sudden increase in production orders, increase in unit consumption caused by changes in equipment operating conditions, fuel structure adjustment caused by environmental regulation, unplanned coal consumption caused by equipment maintenance / failure, and reasonable inventory loss. The implicit demand sequence is the enterprise's full-cycle coal demand data sequence obtained by the supply and demand balance inversion, which is completely matched with the aforementioned unified time scale and arranged in chronological order. It superimposes the regular coal demand recorded in the enterprise's explicit ledger and the implicit potential coal demand obtained by inversion.

[0045] Subsequently, a structural decomposition was performed on the obtained implicit demand sequence, resulting in steady-state and fluctuating demand subsequences. Structural decomposition involves using a time series decomposition algorithm adapted to the characteristics of coal demand to break down the composite time series of implicit demand into independent subsequences with different fluctuation characteristics and driving factors. The core of this process is to separate the deterministic trend components and uncertain disturbance components from the demand data. The steady-state demand subsequence, decomposed from the implicit demand sequence, is a coal demand sequence driven by deterministic and predictable factors such as fixed enterprise production capacity, routine production and operation plans, benchmark product unit consumption quotas, and long-term stable downstream orders. This constitutes the rigid bottom-line constraint for procurement decisions. The fluctuating demand subsequence, decomposed from the implicit demand sequence, is a coal demand sequence driven by uncertain and unpredictable random factors such as sudden market order fluctuations, coal consumption increases or decreases due to seasonal extreme weather, temporary adjustments to production processes, sudden equipment failures, environmental protection production restrictions / increases, and sudden changes in upstream and downstream supply chains.

[0046] Finally, time series alignment and fusion modeling is performed on the steady-state demand subsequence and the fluctuating demand subsequence to obtain the final coal demand change model. The time series alignment and fusion modeling involves first realigning the two decomposed feature subsequences to a pre-defined unified time scale and granularity, ensuring a one-to-one correspondence between the two subsequences in each time window and eliminating potential time series offsets during the decomposition process. Then, based on the differentiated fluctuation characteristics of the two subsequences, corresponding time series prediction algorithms are adapted. For the steady-state demand subsequence, which exhibits stable trends and strong regularities, a prediction algorithm adept at fitting long-term trends and cyclical patterns is adopted. The algorithm, targeting volatile demand subsequences with strong randomness and large fluctuations, employs a prediction algorithm adept at capturing sudden fluctuations and adapting to nonlinear characteristics to obtain accurate prediction results for two subsequences. Finally, through a dynamic weighted fusion algorithm, the two prediction results are fused and calibrated by combining the contribution of the two subsequences to the overall enterprise demand, historical prediction accuracy, and risk impact weights. This constructs a full-dimensional dynamic prediction model that can simultaneously cover the enterprise's steady-state benchmark demand and accurately capture sudden volatile demand. This model can output the enterprise's coal demand benchmark value, upper and lower limits of fluctuation, and probability of demand occurrence for each time window in the future full cycle.

[0047] Furthermore, the method provided in the application embodiment, which constructs a coal procurement knowledge graph based on a coal market dataset, includes: performing entity recognition on the coal market dataset to obtain a coal market entity set, and using the coal market entity set as multiple graph nodes; identifying procurement relationships, transportation relationships, delivery relationships, and inventory replenishment relationships among the multiple graph nodes based on the coal market dataset, and obtaining multiple graph edges; configuring multiple edge attributes corresponding to the multiple graph edges based on the coal market dataset, each edge attribute including transportation cycle, delivery deviation, supply fulfillment rate, transportation capacity, inventory buffer capacity, and historical anomaly count; and performing temporal attribute alignment and graph structure integration on the multiple graph nodes, the multiple graph edges, and the multiple edge attributes to generate the coal procurement knowledge graph.

[0048] Specifically, the first step is to perform entity recognition based on the coal market dataset to extract the coal market entity set. Entity recognition involves accurately identifying the core participants in the entire coal procurement chain from unstructured market announcements, transaction contracts, logistics documents, and structured business ledgers and statistical reports. The coal market entity set includes six major categories of entities: coal type, suppliers, ports, transportation routes, inventory nodes, and coal-consuming entities. Each independent entity is set as a graph node, which is the basic unit of the coal procurement knowledge graph. Each node synchronously stores all static attributes of the corresponding entity, including the calorific value and sulfur content of the coal type, the production capacity and qualification level of the supplier, the designed throughput capacity of the port, the start and end nodes and transportation methods of the transportation route, the maximum storage capacity of the inventory node, and the annual coal consumption scale of the coal-consuming entity, as well as dynamic time-series attributes, including the monthly available production capacity of the supplier, the real-time stockpiling status of the port, the real-time available transport capacity of the transportation route, and the daily inventory level of the inventory node.

[0049] Based on this, using the full business data of the coal market dataset, the procurement, transportation, delivery, and inventory replenishment relationships among all graph nodes are identified. These four core business relationships are set as graph edges, which are directed connections used to represent the business flow logic, supply and demand dependence, data interaction rules, and performance responsibility boundaries between graph nodes. The procurement relationship corresponds to the coal buying and selling supply and demand transaction relationship between suppliers and coal users / traders, clarifying the rights and obligations and transaction rules of both parties. The transportation relationship corresponds to the logistics capacity carrying relationship between coal shipping nodes and receiving nodes, clarifying the realization path and responsible parties for coal spatial displacement. The delivery relationship corresponds to the coal delivery performance relationship between shipping nodes and receiving nodes, clarifying the timeliness, quantity, quality standards, and acceptance rules for coal delivery. The inventory replenishment relationship corresponds to the replenishment flow relationship between upstream and downstream inventory nodes, clarifying the triggering conditions and flow rules for cross-node inventory scheduling, thus realizing the mapping of business relationships in the industrial chain.

[0050] Subsequently, based on historical time-series data and real-time dynamic data from the coal market dataset, corresponding edge attributes were configured for each graph edge. These attributes are the core dataset used to transform the static business relationships between nodes into computable, quantifiable, and predictable dynamic performance capability parameters and risk quantification indicators. Each graph edge's attributes include six dimensions: transportation cycle, arrival deviation, supply fulfillment rate, transportation capacity, inventory buffer capacity, and historical anomaly frequency. The transportation cycle refers to the average and extreme range of coal transportation time for that business link, while the arrival deviation refers to the mean, variance, and maximum deviation between the actual and planned arrival times historically for that link. Deviation value, supply fulfillment rate refers to the proportion of valid orders that are delivered on time, in accordance with quality and quantity, within the corresponding period of the link. Transportation capacity refers to the maximum capacity limit, the capacity occupied within the period, and the scale of real-time available capacity corresponding to the transportation link. Inventory buffer capacity refers to the maximum emergency replenishment inventory capacity and turnover efficiency that the corresponding receiving node of the link can undertake. Historical anomaly count refers to the cumulative number and frequency of fulfillment anomaly events such as supply interruption, transportation delay, substandard quality, and shortage of delivered quantity that have occurred in the historical period of the link. Through the configuration of edge attributes, the strength of association between nodes, fulfillment stability, and probability of risk transmission can be quantified.

[0051] Subsequently, time-series attribute alignment and graph structure integration were performed on all graph nodes, graph edges, and corresponding edge attributes to ultimately generate a coal procurement knowledge graph. Time-series attribute alignment refers to uniformly calibrating multi-source heterogeneous data from different nodes and edge sources to a time scale and granularity completely consistent with the previous coal demand change model, eliminating time-series deviations and sampling frequency differences between different data sources, and ensuring that all static attributes and dynamic parameters in the graph have the ability to perform time-series dynamic extrapolation that is completely matched with the demand-side time series model. Graph structure integration refers to integrating all graph nodes, directed graph edges, and quantified edge attributes that have completed time-series alignment, according to the actual business flow logic and risk transmission path of the coal industry chain, to construct a professional knowledge graph with directed, empowered, and time-series dynamically updated characteristics. This graph can not only visualize the business topology structure of the entire coal procurement industry chain, but also support quantitative analysis operations such as reverse tracing search, node contribution calculation, risk propagation path extrapolation, and multi-scenario dynamic simulation.

[0052] Furthermore, in the method provided in the application embodiment, the process of performing supply chain risk propagation simulation on the target enterprise based on the coal procurement knowledge graph to obtain coal supply risk propagation paths includes: obtaining risk triggering events of the target enterprise and determining the target nodes corresponding to the risk triggering events in the coal procurement knowledge graph; performing reverse tracing search on the target nodes based on the coal procurement knowledge graph to obtain multiple related upstream nodes; calculating the contribution degree of each related upstream node to the target node based on the coal procurement knowledge graph to obtain multiple node target contribution degrees; sorting and filtering the multiple risk propagation paths from the multiple related upstream nodes to the target node according to the multiple node target contribution degrees to obtain a set of key causal paths; and filtering the set of key causal paths by minimum path length to generate the coal supply risk propagation path.

[0053] Specifically, the first step is to acquire the risk-triggered events of the target enterprise and determine the target nodes corresponding to these risk-triggered events in the coal procurement knowledge graph. Risk-triggered events refer to various abnormal events that cause coal supply and demand imbalances, supply interruptions, and inventory drops below the safety threshold for the target enterprise. The core triggering conditions are abnormal coal inventory and coal supply and demand imbalances of the target enterprise. These also include sudden or potential abnormal events that directly affect the enterprise's coal supply security, such as supplier defaults, interruptions of key transportation routes, abnormal scheduling of transit ports, and substandard coal quality. Target nodes are the core graph nodes in the coal procurement knowledge graph that are directly associated with and most directly affected by the risk-triggered event. These include the target enterprise's inventory node corresponding to abnormal inventory events, the core supplier node corresponding to supplier default events, and the transportation route node corresponding to transportation route interruptions. This helps to pinpoint the initial location of the risk in the graph topology.

[0054] Subsequently, based on the directed topology of the coal procurement knowledge graph and the full-link business association logic, a reverse tracing search is carried out on the locked target node to obtain multiple related upstream nodes. The reverse tracing search refers to tracing back along the directed graph edges in the coal procurement knowledge graph that represent procurement relationships, transportation relationships, delivery relationships, and inventory replenishment relationships, against the normal flow direction of coal from the upstream raw coal production end to the downstream coal-using enterprise, starting from the target node and tracing upstream all graph nodes that have direct or indirect business associations, supply and demand dependencies, and replenishment transmission relationships with the target node, forming a set of related upstream nodes.

[0055] Next, based on the static and dynamic attribute data of each node in the coal procurement knowledge graph, as well as the quantified edge attributes of the related edges between nodes, the contribution of each related upstream node to the target node is calculated, and multiple node target contribution degrees are obtained. The node target contribution degree is a core quantitative indicator used to quantify the influence weight of a single upstream node on the coal supply security capability of the target node, and the dependence of the target node on the upstream node. Its calculation uses the edge attributes of the corresponding related edges in the coal procurement knowledge graph, such as transportation cycle, delivery deviation, supply fulfillment rate, transportation capacity, inventory buffer capacity, and number of historical anomalies, as the data source. It is obtained by weighting the upstream node's coal supply ratio to the target node, historical supply stability, and the irreplaceability of the corresponding link. The higher the supply ratio, the stronger the fulfillment stability, and the higher the irreplaceability of the link, the higher the node target contribution degree, which means that the impact of the node's anomaly on the supply security of the target node is greater.

[0056] Subsequently, based on the calculated target contribution of multiple nodes, the multiple risk propagation paths from the multiple associated upstream nodes to the target node are prioritized and screened to obtain a set of key causal paths. The priority ranking and screening uses the sum of the target contribution of all nodes on a single risk propagation path as the core criterion. The higher the total contribution of a single path, the stronger the causal effect of the path on the supply risk of the target enterprise and the greater the impact after the risk is transmitted. It is the core potential link that causes the supply anomaly of the target node. By prioritizing from high to low, non-critical paths with extremely low total contribution and no substantial impact on the target node are eliminated, and only the core risk links with high contribution and high impact are retained, thus forming the set of key causal paths.

[0057] Finally, the set of key causal paths is screened for minimum path length to generate the coal supply risk propagation path. The minimum path length screening refers to selecting the risk path with the fewest nodes and the shortest transmission link from the set of key causal paths with high causality. Such paths have the fewest risk transmission links, the fastest transmission speed, and the most direct impact. They are the core target points for supply chain risk prevention and control and procurement strategy adjustment. At the same time, coal supply early warning instructions can be generated simultaneously based on the finally generated coal supply risk propagation path to achieve advance identification and proactive early warning of risks.

[0058] Furthermore, in the method provided in the application embodiment, the initial coal procurement control strategy is determined by mapping the impact of the coal supply risk propagation path on the target enterprise's procurement rhythm and making procurement control decisions based on the coal supply risk propagation path and the coal demand change model. This includes: mapping the impact of the coal supply risk propagation path on the target enterprise's procurement rhythm to obtain a first result of the procurement impact mapping; mapping the impact of the coal supply risk propagation path on the target enterprise's procurement rhythm to obtain a second result of the procurement impact mapping; performing time-aligned overlay and fusion of the first and second results of the procurement impact mapping to obtain a third result of the procurement impact mapping; and making initial decisions on the target enterprise's coal procurement frequency, coal procurement batch size, and coal replenishment time window based on the third result of the procurement impact mapping to generate the initial coal procurement control strategy.

[0059] Specifically, firstly, based on the coal supply risk propagation path—that is, the risk transmission link that affects the coal supply security of target enterprises most rapidly and to the greatest extent—the impact of procurement rhythm on target enterprises is mapped. This mapping involves quantifying the risk parameters of each key node and the corresponding edge attributes in the risk propagation path, transforming potential supply risks in the supply chain into quantifiable impacts on core execution elements of the entire coal procurement process. Specifically, risk indicators such as supply fulfillment rate, transportation cycle fluctuations, delivery deviations, and historical anomalies at core nodes in the risk path are mapped one by one to the corresponding stages of procurement execution. This clarifies the impact of different risk events on procurement delivery timeliness, supply stability, and batch guarantee capabilities, as well as the impact time window and transmission probability. The final output is the first result of the procurement impact mapping: a quantitative matrix of risk impacts on core elements of the procurement rhythm throughout the entire time cycle under the supply-side risk dimension, clearly defining the measures to hedge supply risks. The core risks of the supply chain require adjustments to the procurement pace and boundaries. Simultaneously, based on the coal demand change model constructed through time series modeling of enterprise coal datasets in the aforementioned steps, a full-dimensional dynamic prediction model can accurately fit the enterprise's normalized steady-state coal consumption trend and effectively capture sudden fluctuations in demand. This model maps the impact of procurement pace on the target enterprise. The mapping uses the steady-state demand subsequence and fluctuating demand subsequence, which are decomposed from the coal demand change model, as the core benchmarks. It transforms the enterprise's coal demand changes throughout the entire cycle into rigid constraints and flexible requirements for procurement execution. The steady-state demand subsequence corresponds to the benchmark constraints for procurement execution, clarifying the fixed procurement scale and cycle requirements required to ensure the enterprise's normalized production and operation. The fluctuating demand subsequence corresponds to the flexible adjustment space for procurement execution, clarifying the procurement flexibility requirements required to cope with unplanned coal consumption and sudden fluctuations in demand. Finally, the second result of the procurement impact mapping is output, namely, under the demand-side dimension, defining the rigid bottom line of coal demand that cannot be breached in procurement execution.

[0060] Subsequently, the execution time of the first and second results of the procurement impact mapping is aligned, superimposed, and integrated. Time alignment refers to unifying the quantitative results of supply-side risk impact and the quantitative scheme of demand-side procurement benchmarks to a time scale and granularity completely consistent with the aforementioned full-process time series modeling. This eliminates deviations between the two in terms of time window division and timing nodes, ensuring that risk adjustment and demand constraints can be accurately matched and superimposed on the same time dimension. Superimposition and integration refers to using the full-cycle coal supply and demand balance on the demand side as a rigid bottom-line constraint and the supply chain risk hedging on the supply side as the core adjustment target. The quantitative results of the two dimensions are bidirectionally adapted and superimposed, eliminating invalid content that conflicts between risk adjustment and demand constraints, and retaining effective control items that simultaneously meet the requirements of rigid demand protection and effective risk hedging. Finally, the third result of the procurement impact mapping is output, which is a set of quantitative control targets for procurement rhythm that covers the entire time window and takes into account both rigid demand protection and supply risk hedging. It clarifies the goals, adjustment boundaries, and constraints that need to be achieved in procurement execution within each time period.

[0061] Ultimately, based on the third result of the procurement impact mapping, initial decisions were made regarding three key procurement factors for the target enterprise: coal procurement frequency, coal procurement volume, and coal replenishment time window. Coal procurement frequency refers to the number of coal procurement orders placed within a fixed period. The core of this initial decision is based on the frequency characteristics of supply stability and demand fluctuations along risk paths. For high-risk, high-demand-fluctuation scenarios, procurement cycles are broken down and procurement frequency is increased to reduce the risk of supply-demand imbalance caused by single supply disruptions. For low-risk, steady-state demand scenarios, a fixed benchmark procurement frequency is set to control transaction and management costs in procurement execution. Coal procurement volume refers to the quantity of coal purchased in a single procurement order. The initial decision is divided into two parts: a benchmark volume and a flexible volume. The benchmark volume is strictly controlled... The system matches the normalized coal consumption requirements of the steady-state demand subsequence to ensure the coal needs of enterprises' basic production and operation. It also flexibly matches the incremental demand of the fluctuating demand subsequence and the safety reserve requirements for risk hedging, reserving buffer space to cope with sudden fluctuations. The coal replenishment time window refers to the optimal time interval between placing an order, locking in the supply, and receiving the goods into storage. Its initial decision is based on the inventory depletion nodes on the demand side, the transportation cycle and delivery deviations of the risk path on the supply side, reserving risk buffer time in advance, setting the optimal replenishment trigger threshold and completion interval, and finally integrating the initial decision results of the three factors to generate the initial coal procurement control strategy. This is a full-cycle coal procurement benchmark execution plan that simultaneously matches the enterprise's full-cycle coal consumption needs and hedges against core supply chain risks. Furthermore, in the method provided in the application embodiment, the transportation fluctuation prediction and time lag cumulative effect mapping of the initial coal procurement control strategy to obtain the transportation lag compensation guidance tensor includes: predicting transportation fluctuations based on the initial coal procurement control strategy to obtain the coal procurement transportation delay distribution; mapping the coal procurement transportation delay distribution to a time window and batch association calibration to obtain the time lag transmission relationship; performing time lag cumulative effect deduction on the initial coal procurement control strategy based on the time lag transmission relationship to obtain the lag cumulative effect matrix; and performing compensation allocation based on the lag cumulative effect matrix to generate the transportation lag compensation guidance tensor.

[0062] Specifically, the initial coal procurement control strategy is combined with the edge attribute data of the corresponding transportation links in the coal procurement knowledge graph, including transportation cycle, arrival deviation, supply fulfillment rate, transportation capacity, number of historical anomalies, etc., and the transportation risk nodes in the coal supply risk propagation path to predict transportation fluctuations. That is, for each batch of procurement in the initial strategy, the transportation route, departure node, arrival node and planned timeliness are combined with historical transportation data, real-time transportation capacity changes, seasonal transportation fluctuations, extreme weather and other influencing factors to make a probabilistic prediction of the arrival timeliness deviation throughout the cycle. Finally, the coal procurement transportation delay distribution is obtained, that is, the set of probability distributions of the arrival delay time of each procurement batch in the initial strategy in different time windows, which are output by the transportation fluctuation prediction.

[0063] Based on this, time window mapping and batch association calibration are performed on the obtained coal procurement and transportation delay distribution to obtain the time lag transmission relationship. Time window mapping refers to accurately mapping the transportation delay distribution of each procurement batch to the replenishment time window, inventory consumption time window, and supply and demand balance verification window pre-set in the initial strategy and fully aligned with the full-process time series modeling. This clarifies the impact range, boundaries, and key impact nodes of a single batch's transportation delay in the time dimension. Batch association calibration refers to calibrating the arrival time linkage relationship between adjacent and related procurement batches based on the time sequence connection logic of multiple procurement batches in the initial strategy, the sequential dependence of inventory replenishment, and the continuous consumption characteristics of coal used for enterprise production. This clarifies the transmission relationship and chain impact of the transportation delay of the preceding batch on the procurement execution, inventory turnover, and supply and demand balance of subsequent batches. The time lag transmission relationship obtained through these two operations is to understand how the obtained single batch transportation delay is affected by the overlap of time windows, the inventory connection between batches, and the continuous consumption of coal used for production.

[0064] Subsequently, based on the time lag transmission relationship, the time lag cumulative effect of the initial coal procurement control strategy was extrapolated to obtain the lag cumulative effect matrix. The time lag cumulative effect extrapolation refers to the multi-scenario superposition simulation of the transportation delay risk of all procurement batches throughout the entire cycle of the initial strategy. It extrapolates the comprehensive impact of single-point and multi-point transportation delays on the target enterprise's coal inventory level, supply-demand balance, procurement plan execution rate, and production and operation continuity after continuous superposition and amplification over time. The focus is on capturing the amplification of delays from multiple batches and multiple time windows, which differs from the single-point delay's single-event impact. The effect, namely the snowballing amplification effect of the continuous inventory gap and replenishment pressure when the delay gap cannot be made up by the normal arrival of subsequent batches, is ultimately obtained as a lagged cumulative effect matrix, which is a two-dimensional quantitative matrix output through deduction. This matrix uses the full cycle time window / procurement batch of the initial strategy as the row dimension and the impact of transportation delay on indicators such as the magnitude of the inventory gap, the duration of supply and demand imbalance, the level of replenishment pressure, the increase in procurement costs, and the probability of production disruption risk as the column dimension. It quantifies the magnitude of the cumulative effect of transportation delay at different time points and for different batches on the entire procurement execution process.

[0065] Finally, compensation allocation is carried out based on the obtained lagged cumulative effect matrix, generating a transportation lag compensation guidance tensor. The compensation allocation follows the principles of prioritizing compensation for high cumulative impacts, focusing on core risk links, prioritizing the bottom line of supply and demand balance, and minimizing the overall cost. It precisely allocates procurement adjustment compensation actions to corresponding time windows, procurement batches, and transportation links, clarifying the compensation direction, compensation magnitude, and optimal execution method for different links. The resulting transportation lag compensation guidance tensor is a high-dimensional guidance dataset with multi-dimensional parameters, which is the final output of the compensation allocation. This tensor contains all parameters in the time dimension, procurement batch dimension, transportation link dimension, compensation action dimension, and compensation magnitude dimension. It can guide the targeted adjustment of the initial strategy in the transportation link, including clarifying how much time in advance to lock in capacity for which procurement batch, how much procurement batch to divert to alternative links for which high-risk transportation links, how much safety procurement volume to supplement in which time window to offset the delay gap, and which type of delay scenario requires adjustment of the replenishment trigger threshold, etc.

[0066] Furthermore, the method provided in the application embodiment, which expands the safety stock range and maps the inventory constraint sensitivity of the initial coal procurement control strategy to obtain an inventory safety elasticity guidance tensor, includes: extracting the multi-window projected inventory and multi-window projected consumption of the target enterprise based on the initial coal procurement control strategy; performing supply and demand fluctuation analysis and time window correlation expansion on the multi-window projected inventory and multi-window projected consumption to obtain a safety stock elastic range; mapping the inventory constraint sensitivity of the initial coal procurement control strategy based on the safety stock elastic range to obtain an inventory constraint sensitivity matrix; and performing inventory risk stratification, replenishment priority calibration, and elastic control weight allocation based on the inventory constraint sensitivity matrix to generate the inventory safety elasticity guidance tensor.

[0067] Specifically, firstly, by combining the initial coal procurement control strategy with the full-cycle demand calculation results of the coal demand change model and the supply risk quantification data of the coal supply risk propagation path, the projected inventory and consumption of target enterprises across multiple windows are extracted. Multiple windows refer to continuously segmented rolling time intervals under a unified time scale fully aligned with the previous full-process time series modeling and the initial coal procurement control strategy, including daily and weekly rolling windows. Projected inventory across multiple windows refers to the theoretical inventory level and dynamic trend at the end of each segmented time window, calculated based on the initial strategy's procurement and warehousing plan, historical inventory turnover patterns, and ending inventory carryover logic. Projected consumption across multiple windows refers to the benchmark coal consumption, upper limit of fluctuating demand, and reasonable loss required for enterprise production and operation within each segmented time window, calculated based on the steady-state demand subsequence and fluctuating demand subsequence decomposed by the coal demand change model. Together, these two constitute the core foundational data for dynamic inventory analysis and flexible management.

[0068] Subsequently, supply and demand fluctuation analysis and time window correlation expansion were performed on the extracted multi-window projected inventory and multi-window projected consumption to obtain the safety stock elasticity range. The supply and demand fluctuation analysis refers to combining the demand fluctuation characteristics output by the coal demand change model, the supply interruption risk with a clear coal supply risk propagation path, and the probability distribution of arrival delays in the transportation link to perform full-scenario quantitative calculation of the deviation range between projected inventory and projected consumption for each time window. This clarifies the limit value of inventory gap and the critical value of inventory surplus that may be caused by two-way fluctuations on both the supply and demand sides, including various risk events such as extreme demand surges, supply interruptions, and long-term transportation delays. The inventory fluctuation boundary and time window correlation extension incorporate the inventory turnover transmission logic, replenishment action connection, and consumption demand linkage characteristics of adjacent and related time windows into the accounting system. This enables full-cycle inventory linkage accounting, where inventory balances in the preceding window are rolled backward and demand fluctuations in the subsequent window are buffered forward. The lower limit of the safety stock elasticity range is the minimum inventory red line that ensures that enterprises do not experience production supply disruptions and meet minimum production needs. The upper limit is the highest inventory threshold that takes into account inventory capital occupation costs, warehousing capacity limitations, and risk buffering capacity. The fluctuation space between the upper and lower limits is the elastic range of dynamically adjustable inventory.

[0069] Based on this, and according to the calculated safety stock elasticity range, an inventory constraint sensitivity mapping is performed on the initial coal procurement control strategy to obtain an inventory constraint sensitivity matrix. This inventory constraint sensitivity mapping refers to quantifying the impact of unit changes in the three major procurement elements—coal procurement frequency, coal procurement batch size, and coal replenishment time window—on the deviation of inventory levels from the elasticity range, taking the safety stock elasticity range as the core constraint benchmark. The mapping assesses the magnitude, direction, and time window of these unit changes on the deviation of inventory levels from the elasticity range. Simultaneously, it considers external risk variables such as demand fluctuations, supply disruptions, and transportation delays to calculate the sensitivity changes of inventory constraints to changes in procurement parameters under different risk scenarios. This identifies the core control levers that have the most significant impact on inventory safety and ineffective adjustment items with no substantial impact. The inventory constraint sensitivity matrix is ​​a two-dimensional quantitative matrix obtained through the above mapping. This matrix uses the three core procurement parameters of the initial strategy as the row dimension and core control indicators such as the magnitude of inventory level deviation from the elasticity range, the probability of supply disruption risk, the increase in inventory backlog costs, and storage capacity occupancy rate as the column dimension, quantifying the full-dimensional impact of different procurement parameter adjustments on inventory safety.

[0070] Finally, based on the obtained inventory constraint sensitivity matrix, inventory risk stratification, replenishment priority labeling, and elastic control weight allocation are performed sequentially to generate an inventory safety elastic guidance tensor. Inventory risk stratification refers to classifying inventory risk into high, medium, and low levels across all time windows of the entire cycle according to the probability of inventory falling below the lower limit of the safe range, the magnitude of the gap, and the degree of impact on enterprise production and operation, thus identifying high-risk control targets. Replenishment priority labeling involves prioritizing replenishment actions across all time windows of the entire cycle based on the risk stratification results, the degree of demand rigidity, and the impact weight of the supply risk chain, prioritizing inventory safety in high-risk windows, rigid production demand windows, and windows corresponding to core risk chains. Elastic control weight allocation refers to... The parameter impact magnitude of the constraint sensitivity matrix is ​​used to assign corresponding control weights to adjustments of different procurement parameters and time windows. Higher weights are assigned to parameters with high sensitivity and high impact, as well as high-risk windows, to ensure optimal inventory safety with minimal procurement adjustments. The inventory safety elasticity guidance tensor is the final output high-dimensional guidance dataset. This tensor includes all parameters in the time window dimension, procurement parameter dimension, inventory risk dimension, control action dimension, and weight allocation dimension. It can guide the differentiated and flexible adjustment of the initial strategy on the inventory side, clarifying the procurement batch adjustment range for different time windows, the optimization range of replenishment windows, the inventory elasticity space corresponding to different risk levels, and the control priority of each procurement parameter.

[0071] Furthermore, in the method provided in the application embodiments, the coal market entity set includes coal type, supplier, port, transportation route, storage node and coal user.

[0072] Specifically, firstly, based on a multi-source time-series coal market dataset—a collection of structured and semi-structured data encompassing all dimensions, including coal category parameters, supplier qualifications and performance data, port throughput and storage data, transportation route capacity data, warehouse node inventory data, and coal-consuming enterprise production and consumption data—named entity recognition technology is employed to automatically identify and extract entities with specific business meanings from the unstructured / semi-structured data. Simultaneously, entity classification and normalization are performed, accurately extracting and standardizing the business objects in the dataset to ultimately form a coal market entity set. Each type of entity in this set is then used to construct a coal procurement knowledge graph. Independent graph nodes; among them, the coal market entity set is specifically divided into six categories of entities, and the extraction and labeling of each category of entity is matched with corresponding business rules and attribute mapping: First, coal type entities, as the core target entities of coal trading and procurement, will have their key coal quality indicators such as calorific value, ash content, sulfur content, and volatile matter, as well as the time-series attributes such as the main producing area and benchmark price range of the corresponding coal type, simultaneously labeled during the extraction process; Second, supplier entities, as the upstream supply entities of the coal supply chain, including coal mining enterprises, coal traders, primary agents, etc., will have their core attributes such as production capacity, range of coal types supplied, historical performance records, and compliance qualifications simultaneously associated during extraction; Third, port entities, as coal... The core hub nodes for coal-water-land intermodal transport and transshipment warehousing encompass ports of origin, transshipment, and unloading. During extraction, their berth capacity, storage capacity, throughput efficiency, historical data on port congestion risk, and collection and distribution systems are simultaneously calibrated. Fourthly, the physical transport routes, serving as the carriers of coal spatial circulation, include all types of transport channels such as main railway lines, shipping routes, inland waterways, and highway branches. During extraction, their transport mileage, rated capacity, transport cycle, historical delay rates, and seasonal capacity fluctuation patterns are simultaneously correlated. Fifthly, the physical inventory nodes, serving as the core nodes for coal storage and buffer regulation, include supplier forward warehouses, port yards, enterprise-owned warehouses, and third-party storage bases. During extraction, attributes such as maximum storage capacity, safety stock threshold, turnover efficiency, and inventory buffer capacity are simultaneously calibrated. Sixth, coal-using entities, as the terminal entities for coal procurement and consumption (i.e., target enterprises and comparable coal-consuming enterprises in the same industry), are simultaneously associated with attributes such as coal consumption patterns, capacity planning, inventory strategies, and historical procurement behavior during extraction. After completing the extraction, attribute calibration, and classification of the six categories of entities, all entities will be uniquely identified and aligned with time-series attributes to ensure that each entity node has a unique identity and that all attribute data is consistent with the time-series dimension of the coal market dataset, ultimately forming a standardized, computable, and associative set of coal market entities.

[0073] Furthermore, in the method provided in the application embodiment, obtaining the coal supply risk propagation path includes: generating a coal supply early warning instruction based on the coal supply risk propagation path.

[0074] Specifically, the first step is to perform final standardized verification, solidification, and computable encapsulation of the coal supply risk propagation path. This path refers to the minimum business link in the coal supply chain capable of risk transmission, from the upstream risk triggering node to the end-user coal-consuming target node. This includes the risk source node, intermediate transmission nodes, the transmission relationships between nodes, and the risk amplification coefficient. In practice, the risk triggering events for the target enterprise are first identified, including abnormal coal inventory, supply-demand imbalance, and abnormal supplier performance. The corresponding target nodes in the knowledge graph are then determined. Through reverse tracing, related upstream nodes are obtained, the target contribution of each node is calculated, and priority ranking and minimum path length selection are performed to obtain an initial risk path set. This initial set then undergoes dual validity verification in both temporal and business dimensions. In the temporal dimension, based on a real-time updated coal market dataset, the current business status of each node in the path and the real-time values ​​of the edge attributes between nodes are verified, eliminating failed links that no longer have transmission capabilities. In the business dimension, the path is verified... The correlation with the target enterprise's coal procurement business and coal consumption demand is used to eliminate redundant paths unrelated to the target enterprise's core procurement chain. Then, the effective paths that pass the verification are quantitatively calibrated for risk transmission intensity. Based on the node target contribution of each node in the path, that is, the proportion of the upstream related node's contribution to the supply and demand guarantee of the target node, the higher the contribution, the greater the impact of abnormal fluctuations of the node on the target enterprise's coal supply. Parameters such as the number of historical anomalies, supply fulfillment rate, transportation buffer capacity, and inventory buffer capacity in the edge attributes are used to calculate the risk transmission coefficient of each path. That is, the amplification or attenuation factor of the risk impact during the transmission of a risk event from the upstream node to the target node along the chain. A coefficient greater than 1 means that the risk will be amplified step by step along the chain, and a coefficient less than 1 means that the risk will be attenuated by the buffer capacity in the chain. At the same time, according to the risk transmission coefficient and the path cause priority, the risk propagation path is divided into four levels: red (extremely high risk), orange (high risk), yellow (medium risk), and blue (low risk), to complete the final standardized acquisition and hierarchical management of coal supply risk propagation path.

[0075] Based on this, a process for generating coal supply early warning instructions based on tiered risk propagation paths is conducted. These instructions are standardized warning signals with clear direction and enforceability, generated from supply chain risk identification results. They include core risk information, scope of impact, duration of impact, and tiered response guidelines. First, the core elements of the standardized risk propagation path are extracted, including the type of risk triggering event, the name and location of the risk source node, core transmission nodes, the complete risk transmission chain, the risk transmission coefficient, the time window of risk impact, and the quantitative impact on the target enterprise's coal procurement rhythm and inventory safety. Then, the corresponding warning level and instruction content are matched according to the risk level of the risk propagation path. For red (extremely high risk) paths, a Level 1 emergency warning is generated, clearly indicating the expected time of the risk of supply disruption, the affected coal types and procurement batches, and the core causal nodes. It also includes core response guidelines for emergency replenishment and switching to alternative suppliers. For orange (high risk) paths, a Level 2 warning is generated, indicating the expected time for the risk to reach the target enterprise, the estimated duration of supply delays, and the affected procurement batches. It also includes response guidelines for advancing procurement schedules and reserving transportation capacity. For yellow (medium risk) and blue (low risk) paths, Level 3 and Level 4 advisory warnings are generated respectively, indicating the fluctuation of risk nodes and the potential scope of impact. They also include routine response guidelines for dynamic inventory monitoring and supplier performance tracking.

[0076] Simultaneously, a dual alignment verification is performed during the generation of early warning instructions. In terms of time dimension, it aligns with the enterprise demand time window output by the coal demand change model, and in terms of business dimension, it aligns with the core decision-making dimension of the initial coal procurement and control strategy. This ensures that the early warning instructions can directly affect the decision optimization of procurement frequency, procurement batch, and replenishment time window, and ultimately generate standardized, hierarchical, and implementable coal supply early warning instructions, which are then simultaneously pushed to the corresponding business modules of the target enterprise, such as procurement management, inventory management, and production scheduling.

[0077] In summary, the coal market data processing method based on time series modeling provided in this application has the following technical effects: By conducting time-series modeling on a full-dimensional coal dataset of target enterprises, and through time alignment, supply and demand flow relationship calibration, and supply and demand balance inversion, the implicit demand sequence of enterprises is obtained. Then, through structural decomposition and alignment fusion modeling, a coal demand change model is constructed that can simultaneously and accurately fit the enterprise's steady-state baseline demand and effectively capture sudden fluctuations in demand. Simultaneously, based on a full-dimensional coal market dataset, a coal procurement knowledge graph covering the main entities and business relationships across the entire industry chain is constructed. Based on this, supply chain risk propagation is extrapolated. Through reverse tracing, contribution calculation, priority ranking, and minimum path selection, the path of coal supply risk propagation is accurately located. This is then combined with… The risk propagation path and demand change model completes the mapping of the impact of procurement rhythm and the initial procurement control strategy. Further, based on the initial strategy, transportation fluctuation prediction and time lag cumulative effect mapping are carried out to generate a transportation lag compensation guiding tensor. Safety stock range expansion and inventory constraint sensitivity mapping are carried out to generate an inventory safety elasticity guiding tensor. Finally, based on the two guiding tensors, the multi-constraint, multi-objective, and multi-dimensional coupling optimization of the initial strategy is completed, which improves the scientific nature and risk resistance of coal procurement control for coal-using enterprises. It not only ensures the continuity and stability of coal use in enterprise production and operation, but also effectively reduces the comprehensive operating costs of the entire coal procurement, transportation, and inventory chain.

[0078] Example 2, based on the same inventive concept as the coal market data processing method based on time series modeling in the previous examples, such as... Figure 2 As shown in the embodiment of this application, a coal market data processing system based on time series modeling is provided. The system includes: Modeling module 11 is used to perform time series modeling based on the target enterprise's coal dataset to establish a coal demand change model; inference module 12 is used to construct a coal procurement knowledge graph based on the coal market dataset, and to perform supply chain risk propagation inference on the target enterprise based on the coal procurement knowledge graph to obtain the coal supply risk propagation path; strategy module 13 is used to map the impact of procurement rhythm on the target enterprise and make procurement control decisions based on the coal supply risk propagation path and the coal demand change model to determine the initial coal procurement control strategy; prediction module 14 is used to perform transportation fluctuation prediction and time lag cumulative effect mapping on the initial coal procurement control strategy to obtain the transportation lag compensation guidance tensor; extension module 15 is used to extend the safety stock range and map the inventory constraint sensitivity of the initial coal procurement control strategy to obtain the inventory safety elasticity guidance tensor; optimization module 16 is used to perform multi-dimensional coupling optimization of the initial coal procurement control strategy based on the transportation lag compensation guidance tensor and the inventory safety elasticity guidance tensor.

[0079] Furthermore, the modeling module 11 is also used to perform the following steps: perform time alignment and supply and demand flow relationship calibration based on the enterprise coal dataset to obtain an enterprise coal supply and demand relationship diagram; perform supply and demand balance inversion of the target enterprise based on the enterprise coal supply and demand relationship diagram to obtain an implicit demand sequence; perform structural decomposition on the implicit demand sequence to obtain a steady-state demand subsequence and a fluctuating demand subsequence; perform time series alignment and fusion modeling on the steady-state demand subsequence and the fluctuating demand subsequence to obtain the coal demand change model.

[0080] Furthermore, the inference module 12 is also used to perform the following steps: perform entity recognition on the coal market dataset to obtain a set of coal market entities, and use the set of coal market entities as multiple graph nodes; identify the procurement relationship, transportation relationship, delivery relationship and inventory replenishment relationship between the multiple graph nodes according to the coal market dataset, and obtain multiple graph edges; configure multiple edge attributes corresponding to the multiple graph edges according to the coal market dataset, each edge attribute including transportation cycle, delivery deviation, supply fulfillment rate, transportation capacity, inventory buffer capacity and historical anomaly count; perform temporal attribute alignment and graph structure integration on the multiple graph nodes, the multiple graph edges and the multiple edge attributes to generate the coal procurement knowledge graph.

[0081] Furthermore, the inference module 12 is also used to perform the following steps: obtaining the risk triggering event of the target enterprise and determining the target node corresponding to the risk triggering event in the coal procurement knowledge graph; performing a reverse tracing search on the target node according to the coal procurement knowledge graph to obtain multiple related upstream nodes; calculating the contribution degree of each related upstream node to the target node based on the coal procurement knowledge graph to obtain multiple node target contribution degrees; sorting and filtering the multiple risk propagation paths from the multiple related upstream nodes to the target node according to the multiple node target contribution degrees to obtain a set of key causal paths; and filtering the set of key causal paths by minimum path length to generate the coal supply risk propagation path.

[0082] Furthermore, the strategy module 13 is also used to perform the following steps: mapping the impact of the procurement rhythm on the target enterprise according to the coal supply risk propagation path, and obtaining a first result of the procurement impact mapping; mapping the impact of the procurement rhythm on the target enterprise according to the coal demand change model, and obtaining a second result of the procurement impact mapping; performing time alignment and superposition fusion on the first result and the second result of the procurement impact mapping, and obtaining a third result of the procurement impact mapping; making initial decisions on the coal procurement frequency, coal procurement batch and coal replenishment time window of the target enterprise according to the third result of the procurement impact mapping, and generating the initial strategy for coal procurement control.

[0083] Furthermore, the prediction module 14 is also used to perform the following steps: predicting transportation fluctuations based on the initial coal procurement control strategy to obtain the coal procurement transportation delay distribution; mapping the coal procurement transportation delay distribution to a time window and calibrating batch associations to obtain the time lag transmission relationship; performing time lag cumulative effect deduction on the initial coal procurement control strategy based on the time lag transmission relationship to obtain the lag cumulative effect matrix; and performing compensation allocation based on the lag cumulative effect matrix to generate the transportation lag compensation guidance tensor.

[0084] Furthermore, the extension module 15 is also used to perform the following steps: extracting the multi-window projected inventory and multi-window projected consumption of the target enterprise according to the initial coal procurement control strategy; performing supply and demand fluctuation analysis and time window correlation extension on the multi-window projected inventory and multi-window projected consumption to obtain the safety stock elasticity range; performing inventory constraint sensitivity mapping on the initial coal procurement control strategy according to the safety stock elasticity range to obtain the inventory constraint sensitivity matrix; and performing inventory risk stratification, replenishment priority calibration, and elastic control weight allocation according to the inventory constraint sensitivity matrix to generate the inventory safety elasticity guidance tensor.

[0085] Furthermore, the simulation module 12 is also used to perform the following steps: the coal market entity set includes coal type, supplier, port, transportation route, inventory node and coal user.

[0086] Furthermore, the inference module 12 is also used to perform the following steps: generating a coal supply early warning instruction based on the coal supply risk propagation path.

[0087] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A coal market data processing method based on time series modeling, characterized in that, The method includes: Time series modeling is performed based on the target company's coal dataset to establish a coal demand change model. A coal procurement knowledge graph is constructed based on a coal market dataset, and a supply chain risk propagation simulation is performed on the target enterprise based on the coal procurement knowledge graph to obtain the coal supply risk propagation path. Based on the coal supply risk propagation path and the coal demand change model, the target enterprise's procurement rhythm impact is mapped and procurement control decisions are made to determine the initial coal procurement control strategy. The initial strategy for coal procurement control is subjected to transportation fluctuation prediction and time lag cumulative effect mapping to obtain the transportation lag compensation guidance tensor. The initial coal procurement control strategy is expanded with a safety stock range and its sensitivity to inventory constraints is mapped to obtain the inventory safety elasticity guidance tensor. The initial strategy for coal procurement control is optimized through multidimensional coupling based on the transportation lag compensation guidance tensor and the inventory safety elasticity guidance tensor.

2. The coal market data processing method based on time series modeling as described in claim 1, characterized in that, Based on the target company's enterprise coal dataset, time series modeling is performed to establish a coal demand variation model, including: Based on the enterprise coal dataset, time alignment and supply and demand flow relationship calibration are performed to obtain the enterprise coal supply and demand relationship diagram; Based on the coal supply and demand relationship diagram of the enterprise, perform the supply and demand balance inversion of the target enterprise to obtain the implicit demand sequence; The implicit demand sequence is structurally decomposed to obtain a steady-state demand subsequence and a fluctuating demand subsequence; The steady-state demand subsequence and the fluctuating demand subsequence are subjected to time series alignment and fusion modeling to obtain the coal demand change model.

3. The coal market data processing method based on time series modeling as described in claim 1, characterized in that, A knowledge graph of coal procurement was constructed based on a coal market dataset, including: Entity recognition is performed on the coal market dataset to obtain a set of coal market entities, and the set of coal market entities is used as multiple graph nodes. Based on the coal market dataset, identify the procurement relationships, transportation relationships, delivery relationships, and inventory replenishment relationships among the multiple graph nodes, and obtain multiple graph edges; Configure multiple edge attributes corresponding to the multiple graph edges according to the coal market dataset. Each edge attribute includes transportation cycle, delivery deviation, supply fulfillment rate, transportation capacity, inventory buffer capacity, and number of historical anomalies. The coal procurement knowledge graph is generated by aligning the multiple graph nodes, multiple graph edges, and multiple edge attributes with temporal attributes and integrating the graph structure.

4. The coal market data processing method based on time series modeling as described in claim 1, characterized in that, Based on the coal procurement knowledge graph, a supply chain risk propagation simulation is performed on the target enterprise to obtain the coal supply risk propagation path, including: Obtain the risk-triggered events of the target enterprise and determine the target nodes corresponding to the risk-triggered events in the coal procurement knowledge graph; Based on the coal procurement knowledge graph, a reverse tracing search is performed on the target node to obtain multiple related upstream nodes; Based on the coal procurement knowledge graph, the contribution of each associated upstream node to the target node is calculated to obtain the target contribution of multiple nodes. Based on the target contribution of the multiple nodes, the multiple risk propagation paths from the multiple associated upstream nodes to the target node are sorted and filtered according to their causal priority to obtain a set of key causal paths; The minimum path length is used to filter the set of key causal paths to generate the coal supply risk propagation path.

5. The coal market data processing method based on time series modeling as described in claim 1, characterized in that, Based on the aforementioned coal supply risk propagation path and the aforementioned coal demand change model, the target enterprise's procurement rhythm impact is mapped and procurement control decisions are made to determine the initial coal procurement control strategy, including: Based on the coal supply risk propagation path, the target enterprise's procurement rhythm is mapped to obtain the first result of the procurement impact mapping. Based on the coal demand change model, the target enterprise's procurement rhythm is mapped to obtain a second result of the procurement impact mapping. The first and second results of the procurement impact mapping are time-aligned, superimposed, and fused to obtain the third result of the procurement impact mapping. Based on the third result of the procurement impact mapping, initial decisions are made on the target enterprise's coal procurement frequency, coal procurement volume, and coal replenishment time window, thereby generating the initial coal procurement control strategy.

6. The coal market data processing method based on time series modeling as described in claim 1, characterized in that, The initial strategy for coal procurement control is used to predict transportation fluctuations and map the cumulative effects of time lags to obtain a transportation lag compensation guidance tensor, including: Based on the initial coal procurement control strategy, transportation fluctuations are predicted to obtain the coal procurement transportation delay distribution. The time window mapping and batch association calibration of the coal procurement and transportation delay distribution are performed to obtain the time lag transmission relationship; Based on the time lag transmission relationship, the time lag cumulative effect of the initial coal procurement control strategy is deduced to obtain the lag cumulative effect matrix; The transportation lag compensation guidance tensor is generated by performing compensation allocation based on the lag cumulative effect matrix.

7. The coal market data processing method based on time series modeling as described in claim 1, characterized in that, The initial coal procurement control strategy is expanded to include a safety stock range and a sensitivity mapping of inventory constraints to obtain a safety stock elasticity tensor, including: Based on the initial coal procurement control strategy, extract the target enterprise's multi-window projected inventory and multi-window projected consumption. Perform supply and demand fluctuation analysis and time window correlation extension on the multi-window projected inventory and the multi-window projected consumption to obtain the safety stock elasticity range; Based on the safety stock elasticity range, the initial coal procurement control strategy is mapped to inventory constraint sensitivity to obtain an inventory constraint sensitivity matrix. Based on the inventory constraint sensitivity matrix, inventory risk is stratified, replenishment priority is calibrated, and elastic adjustment weights are allocated to generate the inventory safety elastic guidance tensor.

8. The coal market data processing method based on time series modeling as described in claim 3, characterized in that, The coal market participants include coal types, suppliers, ports, transportation routes, storage points, and coal users.

9. The coal market data processing method based on time series modeling as described in claim 1, characterized in that, To identify the transmission paths of coal supply risks, including: Based on the aforementioned coal supply risk propagation path, a coal supply early warning instruction is generated.

10. A coal market data processing system based on time series modeling, characterized in that, The system is used to execute the coal market data processing method based on time series modeling as described in any one of claims 1-9, and the system includes: The modeling module is used to perform time series modeling based on the target company's enterprise coal dataset and establish a coal demand change model. The deduction module is used to construct a coal procurement knowledge graph based on the coal market dataset, and to perform supply chain risk propagation deduction for the target enterprise based on the coal procurement knowledge graph to obtain the coal supply risk propagation path. The strategy module is used to map the impact of the coal supply risk propagation path and the coal demand change model on the target enterprise's procurement rhythm and make procurement control decisions, thereby determining the initial strategy for coal procurement control. The prediction module is used to predict transportation fluctuations and map the time lag cumulative effect of the initial coal procurement control strategy, and obtain the transportation lag compensation guidance tensor. The extension module is used to expand the safety stock range and map the stock constraint sensitivity of the initial coal procurement control strategy to obtain the stock safety elasticity guidance tensor. The optimization module is used to perform multi-dimensional coupled optimization of the initial coal procurement control strategy based on the transportation lag compensation guidance tensor and the inventory safety elasticity guidance tensor.

Images