Chemical coal blending method and device and computer equipment

Abstract

The invention relates to a chemical coal blending method and device and computer equipment. The method comprises the following steps: acquiring spectral data of a to-be-distributed coal flow, and performing batch segmentation processing and regression processing on the spectral data to obtain a batch quality vector of the to-be-distributed coal flow; determining a batch quality code of the batch quality vector according to the hearth tolerance envelope; determining a reference time frame, and constructing a time sequence topological graph based on the reference time frame; determining a target cross-layer edge set pointing to a hearth node in the time sequence topological graph, determining a root node state of the search tree based on the target cross-layer edge set, and performing mixed integer tree search in combination with a priority bit of batch quality coding to obtain a candidate scheme set; performing index evaluation and risk priority ranking on candidate schemes in the candidate scheme set to obtain an ordered feasible solution list; and analyzing the first feasible scheme in the ordered feasible solution list into an equipment readable instruction, and carrying out coal flow scheduling according to the equipment readable instruction. According to the invention, a coal blending scheme capable of meeting various requirements can be provided.

Description

Chemical coal blending methods, equipment and computer equipment Technical Field

[0001] This application relates to the field of logistics scheduling technology, and in particular to a chemical coal blending method, apparatus and computer equipment. Background Technology

[0002] Most current coal blending decision-making systems are built on a three-tier architecture: testing, planning, and execution. Enterprises typically install online sampling devices at the plant or storage yard, while maintaining a daytime laboratory. These samples are tested to obtain indicators such as calorific value and sulfur content, and the results are then aggregated into the dispatch server on a shift-by-shift basis. Once the data is generated, the planning layer uses existing ERP (Enterprise Resource Planning) or APS (Advanced Planning and Scheduling) platforms to retrieve static bills of materials. Within a defined timeframe, it provides blending ratios that meet the lower limit of calorific value and the upper limit of emissions. The blending instructions are then sent to the DCS (Distributed Control System) or the duty room, where on-site personnel organize the feeding according to a preset sequence. The entire process is refreshed every half hour to several hours, covering the fuel balance requirements during periods of stable daily load.

[0003] This three-stage coal blending system, from testing to planning to execution, still suffers from significant gaps in the decision-making process in terms of time and space. For example, coal quality information relies heavily on offline testing or online probes with window averaging, resulting in data granularity that doesn't match the furnace feeding cycle and insufficient sensitivity to fluctuations in real-time calorific value and sulfur content. The decision-making level typically uses single-period linear programming or heuristic algorithms to first provide a theoretical blending ratio, followed by a separate logistics verification module that checks throughput capacity and time windows segment by segment, requiring manual iteration when conflicts arise. When disturbances such as logistics delays, port closures, or equipment failures are reported to the system, dispatchers need to re-trigger the model and reissue instructions, a process that often spans from ten minutes to several hours, making it difficult to meet the demands of high-frequency load fluctuations. Because coal quality identification, blending ratio calculation, logistics verification, and execution monitoring are handled by different engines, coal blending schemes often suffer from high costs, frequent rework, or insufficient steady-state duration. Summary of the Invention

[0004] Therefore, it is necessary to provide a chemical coal blending method, apparatus, and computer equipment that can take into account the calorific value and emission requirements, logistical feasibility, and cost of the coal blending scheme, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for blending coal in a chemical process, comprising:

[0006] Acquire the spectral data of the coal flow to be allocated, perform batch segmentation and regression processing on the spectral data, and obtain the batch quality vector of the coal flow to be allocated.

[0007] The batch quality code of the batch quality vector is determined based on the furnace tolerance envelope;

[0008] Determine a baseline time frame, and construct a timing topology diagram including a material layer and an action layer based on the baseline time frame;

[0009] Determine the set of target cross-layer edges pointing to the furnace node in the time-series topology graph, determine the root node state of the search tree based on the set of target cross-layer edges, and perform a mixed integer tree search in combination with the priority bits of the batch quality code to obtain a set of candidate schemes;

[0010] The candidate solutions in the candidate solution set are evaluated by indicators and ranked by risk priority to obtain an ordered list of feasible solutions;

[0011] The first feasible solution in the ordered feasible solution list is parsed into a device-readable instruction, and coal flow scheduling is performed according to the device-readable instruction.

[0012] Secondly, this application also provides a chemical coal blending device, comprising:

[0013] The spectral data processing module is used to acquire the spectral data of the coal flow to be allocated, perform batch segmentation and regression processing on the spectral data, and obtain the batch quality vector of the coal flow to be allocated.

[0014] A quality coding determination module is used to determine the batch quality code of the batch quality vector based on the furnace tolerance envelope;

[0015] The timing topology construction module is used to determine a baseline time frame and construct a timing topology diagram including a material layer and an action layer based on the baseline time frame.

[0016] The candidate solution set determination module is used to determine the target cross-layer edge set pointing to the furnace node in the time-series topology graph, determine the root node state of the search tree based on the target cross-layer edge set, and perform a mixed integer tree search in combination with the priority bits of the batch quality code to obtain the candidate solution set.

[0017] The ordered feasible solution list determination module is used to evaluate the candidate solutions in the candidate solution set and sort them by risk priority to obtain an ordered feasible solution list;

[0018] The feasible solution parsing module is used to parse the first feasible solution in the ordered feasible solution list into a device-readable instruction, and to perform coal flow scheduling based on the device-readable instruction.

[0019] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the steps described above.

[0020] The aforementioned chemical coal blending method, apparatus, and computer equipment acquire spectral data of the coal stream to be allocated. Through batch segmentation and regression processing, batch quality vectors are obtained, and allocation is performed on a batch-by-batch basis. This improves the temporal resolution of the coal stream data, ensuring consistency with the furnace feeding cycle and reducing information latency. The batch quality code of the batch quality vector is determined based on the furnace tolerance envelope, and the quality level of each batch is identified through this code to ensure that the calorific value and emission requirements of the batch coal stream meet the corresponding standards. A temporal topology graph is constructed based on a baseline time frame, transforming the physical constraints of material nodes and coal blending actions into node and edge attributes in the topology graph, thus meeting the logistics requirements for coal blending. During the coal blending decision-making stage, the root node state of the search tree changes from the target layer across layers. Once the edge set is determined, the target cross-layer edge set pointing to the furnace node can represent the action of feeding coal into the furnace for combustion. Based on the root node state and the priority bit of the batch quality code, a mixed integer tree search is performed to obtain a set of candidate solutions that meet the requirements of calorific value, emissions, and logistics feasibility. Then, through index evaluation and risk priority ranking, considering the execution cost and shutdown risk of subsequent solutions, an ordered list of feasible solutions is obtained. The first feasible solution is the optimal solution that can meet all requirements. After being parsed into equipment-readable instructions, it can be sent to relevant coal blending facilities or material nodes for coal flow scheduling. The scheduling status of each batch of coal flow is synchronized to each node. Therefore, while taking into account the calorific value and emission requirements and logistics feasibility requirements of the coal blending scheme, the coal blending cost is also reduced. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 shows the application environment of a chemical coal blending method in one embodiment;

[0023] Figure 2 is a schematic flowchart of a chemical coal blending method in one embodiment;

[0024] Figure 3 is a flowchart illustrating the process of batch segmentation and regression processing of spectral data to obtain the batch quality vector of the coal flow to be allocated in one embodiment.

[0025] Figure 4 is a flowchart illustrating the process of determining the batch quality code of the batch quality vector based on the furnace tolerance envelope in one embodiment.

[0026] Figure 5 is a flowchart illustrating the process of constructing a timing topology diagram containing a material layer and an action layer based on a baseline time frame in one embodiment.

[0027] Figure 6 is a flowchart illustrating the process after coupling the material layer and action layer into a timing topology graph using a set of candidate cross-layer edges in one embodiment.

[0028] Figure 7 is a flowchart illustrating the process of determining the root node state of the search tree based on the target cross-layer edge set in one embodiment, and performing a mixed integer tree search in combination with the priority bits of the batch quality code to obtain a set of candidate solutions.

[0029] Figure 8 is a flowchart illustrating the process of evaluating the indicators and prioritizing the risks of candidate solutions in the candidate solution set to obtain an ordered list of feasible solutions in one embodiment.

[0030] Figure 9 is a schematic diagram of the process of parsing the first feasible solution in the ordered feasible solution list into a device-readable instruction and performing coal flow scheduling according to the device-readable instruction in one embodiment.

[0031] Figure 10 is a flowchart illustrating the chemical coal blending method in another embodiment;

[0032] Figure 11 is a structural block diagram of a chemical coal blending device in one embodiment;

[0033] Figure 12 is an internal structure diagram of a computer device in one embodiment. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0035] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0036] In the coal blending decision-making system established according to the three-layer architecture of laboratory analysis, planning, and execution in related technologies, it usually includes a coal quality perception layer, a decision-making layer, and an execution layer. In the coal quality perception link, the common technical route still relies on off-line laboratory analysis, and some units add near-infrared or gamma-ray ash meters on the belt to improve timeliness. The sensor maps the continuous spectral signal into intermediate features such as ash content and volatile matter, and then estimates the gross calorific value through multiple linear regression and uploads it to the SCADA (Supervisory Control And Data Acquisition) system. Due to the limited model size, the real-time algorithm usually only outputs a single confidence lower limit, without making batch-level distinctions, but takes the average value of continuous readings within a time window to mitigate the impact of noise.

[0037] At the decision-making modeling level, most systems adopt a single-period coal blending model based on linear programming. The inventory is split into several quality grades according to the ore source, and the daily furnace load is converted into the target proportioning quantity, and the single-objective fuel cost is minimized to solve. To balance complexity and solution speed, only two hard constraints, calorific value and sulfur content, are retained in the model, and the calorific value loss is converted by the average moisture content. If cross-period rolling is required, the method is to bring the remaining coal quantity of the previous period into the initial inventory of the next period, and the whole process is manually triggered by the dispatcher. Some academic or commercial software introduces heuristic searches such as genetic algorithms and simulated annealing on this basis to handle additional rules such as FIFO (First Input First Output) and minimum storage days, but still maintains the order of first giving the proportioning and then verifying with the bin simulation.

[0038] The physical constraints of logistics are usually inspected by an independent module, such as a discrete event simulation or a rule engine, after the plan calculation is completed. First, the proportioning result is mapped into a list of operations such as ship loading, stacking, and feeding, and then it is checked segment by segment according to parameters such as equipment throughput capacity, loading and unloading time periods, and yard traffic flow. When a conflict is found, the platform returns a manual prompt, and the dispatcher adjusts the proportioning order or supplements the purchase quantity to re-iterate. Since logistics and proportioning calculations belong to different engines, the two are mostly connected through files or shared tables, and do not support second-level linkage.

[0039] The execution layer writes the final operation instruction into the DCS through the OPC (OLE for Process Control) interface, and at the same time prints a paper work order in the duty room. The system will regularly read the belt scale data and the silo liquid level, and update the inventory according to the tonnage completion, but there is no automatic correction mechanism for real-time coal quality. When sudden weather, port closure, or equipment failure causes the coal arrival to be postponed, the site usually reports through phone calls or instant messaging, and then the dispatcher temporarily inserts coal or reduces the load, and the decision still follows the traditional closed-loop of manual adjustment, re-solving, and manual issuance.

[0040] The aforementioned three-stage coal blending system exhibits significant discontinuities in temporal, spatial, and model coupling aspects. To address these coupling deficiencies and timeliness issues, this application embeds the spectral batch detection of the coal flow to be allocated, the end-to-end logistics constraints from port to stockpile to conveyor belt to furnace, and the furnace tolerance envelope into a temporal topology graph. An executable and low-cost coal blending instruction is output using a graph-constrained mixed-integer search tree, with the instruction output process completed within minutes. In the event of coal quality data revisions or logistics delays, recalculation can be performed through local branching to maintain furnace stability. This achieves synchronous mapping of real-time coal quality information with end-to-end physical constraints such as port, stockpile, transportation, and furnace into the same decision model. It generates a low-cost coal blending scheme that meets both calorific value and emission requirements while maintaining logistical feasibility within minutes, avoiding manual post-processing or multiple rework. Furthermore, it maintains furnace stability and prevents shutdown during disturbances through rapid local recalculation.

[0041] The chemical coal blending method provided in this application embodiment can be applied to the application environment shown in Figure 1. The chemical coal blending platform 102 communicates with the terminal 104 via a network. A data storage system can store the data that the chemical coal blending platform 102 needs to process. The data storage system can be integrated onto the chemical coal blending platform 102, or it can be located in the cloud or on another network server. The chemical coal blending platform 102 acquires the spectral data of the coal flow to be allocated sent by the terminal 104, performs batch segmentation and regression processing on the spectral data to obtain the batch quality vector of the coal flow to be allocated; determines the batch quality code of the batch quality vector based on the furnace tolerance envelope; determines the reference time frame, and constructs a time-series topology graph including the material layer and the action layer based on the reference time frame; determines the target cross-layer edge set pointing to the furnace node in the time-series topology graph, determines the root node state of the search tree based on the target cross-layer edge set, and performs a mixed integer tree search in combination with the priority bit of the batch quality code to obtain a candidate solution set; evaluates the candidate solutions in the candidate solution set by indicators and sorts them by risk priority to obtain an ordered feasible solution list; parses the first feasible solution in the ordered feasible solution list into a device-readable instruction, and performs coal flow scheduling according to the device-readable instruction. The terminal 104 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, sensors, etc. The chemical coal blending platform 102 can be an independent physical server configured with multiple AI algorithms, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0042] In an exemplary embodiment, as shown in FIG2, a chemical coal blending method is provided. Taking the application of this method to the chemical coal blending platform in FIG1 as an example, the method includes the following steps 202 to 206. Wherein:

[0043] Step 202: Obtain the spectral data of the coal flow to be allocated, perform batch segmentation and regression processing on the spectral data, and obtain the batch quality vector of the coal flow to be allocated.

[0044] A coal stream awaiting allocation refers to a coal stream for which an allocation plan has not yet been determined. The spectral data of this coal stream can be considered as image data, which can be acquired using spectral acquisition equipment. Batch segmentation processing refers to the process of segmenting continuous coal stream spectral data into batch-level coal stream spectral data. Regression processing refers to the process of predicting relevant data of the coal stream using linear or nonlinear regression. A batch quality vector is a vector representing the combustion quality of a specific batch of coal stream. Coal streams from various mines can undergo compositional analysis upon entering the plant or before entering the plant, carrying quality data. Through regression processing, the quality data of the coal streams segmented into batch-level streams is represented by batch quality vectors.

[0045] For example, the spectral data of the coal flow to be allocated is acquired by the spectral acquisition device, the spectral data is processed by batch segmentation to obtain spectral segments, and the quality data of the spectral segments is processed by regression to obtain the batch quality vector of the coal flow to be allocated.

[0046] Step 204: Determine the batch quality code of the batch quality vector based on the furnace tolerance envelope.

[0047] The furnace tolerance envelope refers to the envelope formed by three dimensions of coal combustion in the furnace: calorific value, sulfur content, and combustion time. The envelope boundary can be set according to the quality requirements of the coal. Batch quality coding refers to the code obtained by evaluating the combustion quality of each batch of coal stream. It is understandable that coal streams contain a certain amount of sulfur, which, after combustion, is converted into sulfur oxides and emitted into the atmosphere, causing pollution. Therefore, the sulfur content envelope can reflect the emission requirements of the coal blending scheme.

[0048] For example, the batch quality code of the batch quality vector is determined based on the envelope boundary of the furnace tolerance envelope.

[0049] Step 206: Determine the baseline time frame and construct a timing topology diagram including the material layer and the action layer based on the baseline time frame.

[0050] The baseline timeframe describes the impact time of various events involved in the coal blending process and serves as a reference for time nodes in the time-series topology diagram for time alignment. A time-series topology diagram is a topology graph that uses nodes and edges containing time information to represent actual relationships. The material layer is a virtual layer structure in the time-series topology diagram composed of spatial nodes. The action layer is a virtual layer structure in the time-series topology diagram composed of action nodes and cross-layer edges. Both the material and action layers may also include intra-layer edges.

[0051] For example, a baseline time frame is determined based on the start and end times of coal blending-related events. Virtual nodes are created for entity objects within the baseline time frame, and a time-series topology diagram including a material layer and an action layer is constructed based on the virtual nodes. Entity objects may include entity locations and executed actions.

[0052] Step 208: Determine the set of target cross-layer edges pointing to the furnace node in the time-series topology graph, determine the root node state of the search tree based on the set of target cross-layer edges, and perform a mixed integer tree search in combination with the priority bits of the batch quality code to obtain the candidate scheme set.

[0053] A furnace node is a node representing the furnace location in the time-series topology diagram, situated in the material layer. A cross-layer edge is an edge that spans the material layer and the action layer. Connections between nodes in the material layer are established through nodes in the action layer and cross-layer edges. This corresponds to the actual coal flow scheduling between different entity locations, which requires executing corresponding scheduling actions. A cross-layer edge pointing to a furnace node represents a furnace start-up action, and cross-layer edges in the target cross-layer edge set can serve as candidate edges for furnace start-up. The priority bit of the batch quality code refers to the furnace start-up priority order represented by the batch quality code. Hybrid integer tree search refers to a tree search method performed through integer-based solutions. The candidate scheme set refers to the set of candidate coal blending schemes.

[0054] For example, the set of target cross-layer edges pointing to the furnace node in the time-series topology graph is determined, the batch quality vector in the set of target cross-layer edges is determined and written into the root node of the search tree, and a mixed integer tree search is performed in combination with the priority bit of the batch quality code corresponding to the batch quality vector to obtain the candidate scheme set.

[0055] Step 210: Evaluate the candidate solutions in the candidate solution set by indicators and prioritize them by risk to obtain an ordered list of feasible solutions.

[0056] Indicator evaluation refers to assessing the effectiveness of candidate solutions from different indicator dimensions. Risk priority ranking refers to sorting candidate solutions according to the priority of the risk of furnace shutdown that may result from their execution. Ordered feasible solution list refers to storing a list of feasible solutions corresponding to the candidate solutions after being ranked by risk priority.

[0057] For example, the candidate solutions in the candidate solution set are evaluated by indicators, the candidate solutions with excessive risk are eliminated, and the remaining candidate solutions are ranked by risk priority to obtain an ordered list of feasible solutions.

[0058] Step 212: Parse the first feasible solution in the ordered feasible solution list into a device-readable instruction, and perform coal flow scheduling based on the device-readable instruction.

[0059] The first feasible solution refers to the candidate solution that ranks first in the ordered feasible solution list. Equipment-readable instructions refer to instructions that can be read by coal-blending related equipment. For example, the first feasible solution in the ordered feasible solution list is parsed into an equipment-readable instruction, which is then sent to the distributed control node for coal flow scheduling.

[0060] In the aforementioned chemical coal blending method, spectral data of the coal flow to be allocated is obtained. Through batch segmentation and regression processing, batch quality vectors are obtained, and allocation is performed on a batch-by-batch basis. This improves the temporal resolution of the coal flow data, ensuring consistency with the furnace feeding cycle and reducing information latency. The batch quality code of the batch quality vector is determined based on the furnace tolerance envelope, and the quality level of each batch is identified through this code to ensure that the calorific value and emission requirements of the batch coal flow meet the corresponding standards. A temporal topology graph is constructed based on a baseline time frame, transforming the physical constraints of material nodes and coal blending actions into node and edge attributes in the topology graph, thus meeting the logistics requirements for coal blending. During the coal blending decision-making stage, the root node state of the search tree is determined by the target cross-layer edge set. The target cross-layer edge set pointing to the furnace node can represent the action of feeding coal into the furnace for combustion. Based on the root node state and the priority bit of the batch quality code, a mixed integer tree search is performed to obtain a set of candidate solutions that meet the requirements of calorific value, emissions, and logistics feasibility. Then, through index evaluation and risk priority ranking, considering the execution cost and shutdown risk of subsequent solutions, an ordered list of feasible solutions is obtained. The first feasible solution is the optimal solution that can meet all requirements. After being parsed into equipment-readable instructions, it can be issued to relevant coal blending facilities or material nodes for coal flow scheduling. The scheduling status of each batch of coal flow is synchronized to each node. Therefore, while taking into account the calorific value and emission requirements and logistics feasibility requirements of the coal blending scheme, the coal blending cost is also reduced.

[0061] In an exemplary embodiment, as shown in Figure 3, batch segmentation and regression processing are performed on the spectral data to obtain the batch quality vector of the coal stream to be allocated, including steps 302 to 304. Wherein:

[0062] Step 302: Obtain batch identification pulses and divide the spectral data into multiple spectral segments based on the batch identification pulses.

[0063] Batch identification pulses refer to pulses used for batch segmentation. Spectral segments refer to segments obtained after batch segmentation. For example, a near-infrared spectral probe is selected as the spectral acquisition device. The near-infrared spectral probe is installed on fixed brackets at the beginning of the inlet conveyor belt and the material handling port in the stockpile, allowing the field of view of the near-infrared spectral probe to cover the entire coal flow cross-section. Real-time scanning of the continuously flowing coal flow is performed, and the batch identification pulse emitted upstream of the probe is used as a segmentation signal to cut the raw spectral data stream into equal-length spectral segments according to pulse intervals. A source identifier and a time identifier can also be written to each segment in real time. The source identifier can be a source number to trace the compositional differences of coal flows from different sources. The time identifier can be an arrival timestamp to locate the logistics stage of the coal flow.

[0064] Step 304: Based on the composition detection information of the ore source to which the spectral fragment belongs, predict the combustion quality of the coal stream to be allocated and obtain the batch quality vector. The batch quality vector includes the calorific value, sulfur content and moisture content of the coal stream to be allocated.

[0065] Composition detection information refers to the information obtained by analyzing the composition of coal streams, which may include calorific value, sulfur content, and moisture content. For example, a chemical coal blending platform may include an edge computing box, within which a lightweight regression network is pre-installed for regression processing. This lightweight regression network can retain only three convolutional layers and one linear regression head, enabling forward inference to be completed in fragment-level time, outputting a batch quality vector along with the confidence intervals of each component. For instance, the batch quality vector can use {calorific value}... , sulfur content , moisture } represents, and is accompanied by its respective confidence interval. .in, : Batch of combustible high calorific value; Sulfur content of each batch; : Batch moisture content; : The lower and upper confidence values ​​of the corresponding indicators. The batch quality vector generated by the edge computing box can be pushed to the fuel dispatch center of the chemical coal blending platform in real time via an encrypted link, and simultaneously written to the batch and location buffer tables of the data storage system. The completion of the write operation triggers the inventory dual-tag refresh.

[0066] In this embodiment, by segmenting spectral data to obtain spectral fragments and outputting batch quality vectors including calorific value, sulfur content, and moisture content, the temporal resolution of coal quality data perception can be improved and information delay can be reduced.

[0067] In an exemplary embodiment, as shown in FIG4, determining the batch quality code of the batch quality vector based on the furnace tolerance envelope includes steps 402 to 404.

[0068] Step 402: Determine the envelope boundary of the furnace tolerance envelope and calculate the minimum remaining amount between the batch mass vector and the envelope boundary.

[0069] Minimum residual quantity refers to the minimum residual quantity between each component of the batch mass vector and the envelope boundary. The envelope boundary can be determined according to the rules for constructing the furnace withstand envelope.

[0070] For example, the construction of the furnace tolerance envelope can refer to the following process: converting the batch mass vector... The write-determination module synchronously retrieves the load curve of the furnace for future rolling cycles from the scheduling database of the data storage system. Sulfur emission thresholds were read from the load curve. and obtain the confirmed maintenance window. Write the above information into the same memory page and assign a unique identifier. .Will In and Envelope generation is performed, and the minimum calorific value function required to maintain the load is obtained based on the historical combustion calibration curve, as shown in Formula 1 below.

[0071] Formula 1: .

[0072] in, The minimum acceptable calorific value at time t; Furnace calorific value – load ratio coefficient; : Combustion margin constant.

[0073] The three-dimensional furnace tolerance envelope is defined as shown in Formula 2 below.

[0074] Formula 2: .

[0075] in, Furnace load demand at time t; Sulfur emission limit; Maintenance time window; The start time of the maintenance time window; The end time of the maintenance time window; This round of aggregated memory pages; Refresh the unique identifier; : The combined feasible region of calorific value-sulfur content-time period; The coordinates of the calorific value and sulfur content of the batch to be judged; : The current moment.

[0076] Step 404: Determine the batch quality code based on the minimum remaining quantity and the preset safety margin.

[0077] Safety margin refers to the value that maintains a safety limit between the envelope boundary and the components in the batch quality vector; the minimum safety margin of the envelope boundary. It can be pre-configured by the operations department and included. Lock it in until the next furnace parameter change.

[0078] For example, All batches i to be judged )of The determination is carried out sequentially. First, the minimum residual amount at the envelope boundary is calculated, which can be referred to in the following formula 3.

[0079] Formula 3: .

[0080] in, : Safety margin for batch i from exceeding the limit; : Batch i's estimated arrival time at the furnace.

[0081] Subsequently and Relationship execution hierarchy: If Mark batch i as ready to deliver immediately; if Mark batch i as controlled and deployable; if and Batch i was marked as "risk pending". Finally, the batch number, batch quality code, and... Write to list .

[0082] For example, a quality coding list can also be used. For each batch of ore, the logistics tracking is performed using the ore source identifier, and its transit route is retrieved in the order of port, storage yard, conveyor belt, and furnace front. .right Each segment k reads three durations: travel time. queuing Loading and unloading belt replacement Calculate the earliest arrival time of this batch using Formula 4 below.

[0083] Formula 4: .

[0084] in, : Current determination time; : The set of logistics segments for batch i; The travel time of segment k; Queuing time for segment k; : Loading and unloading time of segment k; : The earliest arrival time of batch i in the furnace.

[0085] Will Timing of pre-charging furnace charging trough at fuel dispatch center The adjustable time window is calculated by comparison, as shown in Formula 5 below.

[0086] Formula 5: .

[0087] in, : The time when the empty slot in front of the furnace has been allocated for batch i; : Adjustable time window, which can be regarded as the time difference that batch i can still be advanced or delayed on the logistics side.

[0088] Name of the mineral source Current node , and Write them on the same line, together with their batch quality code, to form a double label. All of them can be done. Batch write records to dynamic inventory table An exclusive lock is applied to the target row before writing to avoid conflicts caused by parallel scheduling. The inventory version number is refreshed immediately after writing. Then an inventory vector is generated. The inventory vector, indexed by version number, declares a unique point in time for this round of data view. For records in the previous two rounds that have not been referenced by any graph model, the current time minus the latest time the record was created is considered to be greater than the operational threshold. The conditions are directly marked as expired.

[0089] For example, priority bit setting and edge weight mapping are performed on batches, in Internally, a mandatory priority bit is written for batches that can be deployed immediately. For controlled, investable, and risk-pending batches, write the data to a delayed location. And set a threshold for the number of records for each type. .Will and The edge weights are mapped to a temporal topology graph, which guides the subsequent tree search process to distinguish between those that must be immediately fed into the furnace and those that can be paired, thus reducing invalid branches.

[0090] In this embodiment, the batch quality code can be determined by judging whether the batch quality vector falls within the range of the furnace tolerance envelope, and different scheduling methods can be implemented for batches of different coal qualities.

[0091] In an exemplary embodiment, as shown in FIG5, a timing topology diagram including a material layer and an action layer is constructed based on a baseline time frame, including steps 502 to 506.

[0092] Step 502: For each entity location in the baseline time frame, create a spatial node, write the spatial node into the first set, bind the quality status child node as a child node of the spatial node, write it into the second set, combine the first set and the second set into a material layer with a parent-child relationship, and obtain the quality status child node by traversing the inventory vector in the dynamic inventory table.

[0093] Entity location refers to the location of entities involved in coal flow scheduling, which may include ports, stockpiles, feeding belts, and furnaces. The first set refers to the set of spatial nodes. Quality status sub-nodes refer to nodes constructed for each batch of coal flow, containing both quality and time information. The second set refers to the set of quality status sub-nodes. By binding quality status sub-nodes as child nodes of spatial nodes, the first and second sets are combined to form a material layer.

[0094] For example, a spatial node is created for each entity location loc in the reference time frame. Write to the first set Traverse the inventory vector and construct a quality status sub-node for each batch of material i, as shown in Formula 6.

[0095] Formula 6: .

[0096] in, Batch unique number; : The spatial node to which the batch currently belongs; The earliest arrival time of the batch at its current node.

[0097] Will Bind to The child nodes are written to the second set. Using the father-son relationship to... and Combined into material layers The inner edge of the middle layer It is used only to maintain the hierarchical structure and does not participate in the flow of logistics. It is regarded as a static anchor point in the subsequent tree search process.

[0098] For example, during the generation of the baseline time frame, the inventory vector can be loaded synchronously with three types of events: port closure, yard maintenance, and furnace shift change, and the start and end times of each event can be read. The following formula is used to define all event boundaries and the end of the rolling cycle. After performing ordered merging, the discrete-time scale sequence is obtained as shown in Formula 7 below.

[0099] Formula 7: .

[0100] in, : The current moment; End of the rolling cycle; The start and end times of the event; Deduplication in ascending order; : Set of discrete-time scales.

[0101] Pairing adjacent tick marks in a discrete-time readable sequence together forms a half-open interval. Based on the events corresponding to each interval, allow or prohibit signs are added to the intervals to close the sections affected by maintenance or navigation closures. Finally, all intervals are spliced ​​together to form a base time frame. The frame is read-only within this round of graph construction, ensuring that the time coordinates of all subsequent nodes and edges are aligned with the same reference.

[0102] Step 504: For each action in the action library, create an action node in the action start and end window, write it into the third set, and structure the third set into an action layer.

[0103] An action library is a database that stores execution actions. An execution action refers to the production actions performed on the coal stream from its entry into the plant to its commissioning, and can include four categories: turning, loading, blending, and commissioning. An action start and end window is a time window consisting of the start and end times of the execution action. An action node is a node constructed based on the execution actions. The third set is a collection of action nodes.

[0104] For example, four types of actions—turning over the pile, loading onto trucks, blending, and feeding into the furnace—are written into the action library and recorded as indexes. For each type of action k, read its set of executable intervals. And in each action start and end window Create action nodes Write to the third set Define the upstream and downstream touchable node sets based on the physical connection range of the action. and Candidate cross-layer edge sets can be generated according to the following formula 8.

[0105] Formula 8: .

[0106] in, Candidate cross-layer edges for action k; : Source batch node; Action node; : Target space node; : Start and end scale of the action window; : The next free tick mark next to the target node; : The set of upstream and downstream positions of action k.

[0107] All Merge into cross-layer edge sets and will Structured to Action Layer , This represents the inner edge of the action layer.

[0108] Step 506: Couple the material layer and the action layer into a time-series topology graph using a set of candidate cross-layer edges.

[0109] Candidate cross-layer edges connect the material layer and the action layer, and the material layer and the action layer are connected through a set of candidate cross-layer edges. Coupled into a complete time-series topology.

[0110] In this embodiment, a two-layer time-series topology diagram is formed by coupling the material layer and the action layer, and the entire logistics chain is mapped to the topology diagram, which can meet the logistics feasibility requirements of the coal blending scheme.

[0111] In an exemplary embodiment, as shown in FIG6, after coupling the material layer and the action layer into a timing topology graph through a set of candidate cross-layer edges, steps 602 to 604 are further included.

[0112] Step 602: Map each cross-layer edge in the candidate cross-layer edge set to the physical facility it passes through, perform throughput verification on cross-layer edges of the same physical facility within the same action window, and block cross-layer edges that fail the throughput verification.

[0113] Throughput verification refers to the operation of checking the capacity of physical facilities for batch coal flow. Different coal blending schemes have different throughput requirements for physical facilities, which need to be adapted to the actual throughput load of physical facilities to avoid overload operation. Therefore, cross-layer edges that fail throughput verification can be shielded.

[0114] For example, during the process of throughput verification and edge parameter fixing, the candidate cross-layer edge set is... Each cross-layer edge is mapped to the physical facility (fac) it passes through (e.g., yard loading / unloading point, conveyor belt section, furnace feeding point). For the same facility within the same action window... The set of all candidate edges within is denoted as The minimum batch requirement within this set can be obtained by summing the results using the following formula (9).

[0115] Formula 9: .

[0116] in, : Fac facilities at the window Total batch usage; Minimum batch requirement for candidate edge e; : The set of candidate edges within the window that point to the same facility.

[0117] like ,in If the maximum rated throughput of the facility is reached, then the minimum batch size for writing to all edges within this set is used. With the duration of the assignment Both parameters are retained; otherwise, the out-of-limit edge is marked as shielded and the reason for shielding is recorded. In this way, logistics bottlenecks can be frozen directly at the graph structure level, and subsequent tree searches can avoid overloaded paths without further capacity checks. Among these, The rated throughput limit of the facility fac; Let e ​​be the duration of the job corresponding to candidate edge e.

[0118] Step 604: Add virtual buffer nodes to the time sequence topology graph, connect the virtual buffer nodes to the key nodes in the material layer through zero-delay edges, and add economic penalties to the cross-layer edges that are retained in the candidate cross-layer edge set when the cross-layer edge crosses the virtual buffer node.

[0119] Virtual buffer nodes are auxiliary nodes used to determine whether economic penalties should be applied to cross-layer edges. Zero-latency edges are edges without any time delay. Key nodes in the material layer include ports, yards, and conveyor belts. Economic penalty is a way to measure the cost of an edge; cross-layer edges with added economic penalties can serve as backup channels during tree search, activated when lower-cost conventional edges break.

[0120] For example, quality label inheritance and backup channel implantation are performed. For each cross-layer edge e already retained in the candidate cross-layer edge set, two write operations are performed: writing the quality label and physical parameters, and copying the quality label from the source batch node. The time will be reached when the edge attribute region is reached. Minimum Batch With the duration of the assignment Three types of physical parameters are written in parallel to ensure that subsequent tree searches can simultaneously perceive coal quality risk and operational cycle time on a single edge. A virtual buffer node is added to the time-series topology graph. And connect the edges with zero latency It connects with three key nodes: ports, storage yards, and conveyor belts, enabling any crossing... cross-layer edge Additional economic penalties will be imposed, referring to the following formula 10.

[0121] Formula 10: .

[0122] in, Edge cost after penalty; Original logistics costs; The penalty coefficient for the backup channel can be set by the operations department. Indicator function, if via Then take 1, otherwise take 0; : The set of edges that pass through all virtual buffer nodes.

[0123] For example, a consistency scan is performed on the two-layer temporal topology graph after nodes and edges have been written. The graph is checked sequentially for isolated nodes, forward loops, and time reversal errors. If any anomalies are found, they are located and corrected. After confirmation, the entire temporal topology graph is encapsulated into a graph object G, with the node field storing type, position, and time, and the edge field storing start point, end point, quality label, and physical parameters. Finally, a graph version number is appended to G. .

[0124] In this embodiment, the cost structure formed after applying economic penalties can encourage subsequent tree searches to prioritize regular edges when normal paths are feasible, and only activate backup channels at a higher cost when regular edges break due to shipping blockages or belt outages, thus ensuring economic priority and fault tolerance from a business perspective.

[0125] In an exemplary embodiment, as shown in FIG7, the root node state of the search tree is determined based on the target cross-layer edge set, and a mixed integer tree search is performed in combination with the priority bits of the batch quality code to obtain a candidate scheme set, including steps 702 to 706.

[0126] Step 702: Extract the quality information and arrival time information of each cross-layer edge in the target cross-layer edge set, add the quality information and arrival time information to the furnace entry candidate table, and inject the furnace entry candidate table and its version number into the root node of the search tree.

[0127] The quality information of cross-layer edges can be obtained from the batch quality vector corresponding to the batch. Arrival time information represents the estimated arrival time of the batch coal flow in the furnace at the current moment, which can be calculated by referring to the baseline time frame. The furnace feed candidate table stores the target cross-layer edges corresponding to the batch coal flows to be fed into the furnace. The version number of the furnace feed candidate table is a number that identifies the update status of the furnace feed candidate table.

[0128] For example, the graph object G generated in the above embodiments is loaded into memory, and all target cross-layer edges pointing to the furnace node are scanned one by one to extract their source batch numbers. Quality Label and the earliest expected arrival time in the furnace The list is compiled into a candidate list for furnace entry. Set the current material level in the furnace. Temperature difference in high-temperature zone Load curve for the next three hours Write each item into the same table, and record the root version number at the beginning of the table. Ultimately, Together This information is injected into the root node of the search tree, aligning the search starting point with the real-time status of the furnace. Batch quality code.

[0129] For example, the tree search process can be performed within a steady-state priority domain. During the generation process, it is possible to according to Perform ascending sorting and call the furnace tolerance envelope locked in the above embodiments. The sorting results are connected into a candidate batch sequence l using the sliding window method, and the position of each sequence within the window is calculated. Internal calorific value trajectory With sulfur trajectory The feasibility of a sequence can be determined by referring to a sequence segmentation function shown in Formula 11 below.

[0130] Formula 11: .

[0131] in, : A feasible indication of sequence l; : The instantaneous heat value corresponding to sequence l; : Instantaneous sulfur content corresponding to sequence l; The minimum calorific value required to maintain the load; Sulfur emission threshold.

[0132] Only keep The sequence is written into the steady-state priority domain. Then put Translate into a restricted matrix As the first-level branch admission condition of the tree generator, it can ensure that the tree search falls into the continuous steady-state region at the beginning.

[0133] Step 704: Using the root node as the starting point, incrementally accumulate leaf nodes based on the batch information and priority of the coal flow to be allocated, and obtain a feasible subtree.

[0134] Batch information for coal flows to be allocated can be recorded in the batch quality vector. The priority bits of the batch quality code can be used during the search process to evaluate the start-up time of the batch coal flow and determine the target cross-layer edge to be saved in the candidate start-up table. Branch expansion starting from the root node can generate leaf nodes, and the breadth and depth ranges of the branch expansion can be preset.

[0135] For example, setting a maximum depth for the tree generator With width and inject the root state. With constraint matrix Starting from the root node, for nodes falling within... Leaf nodes are generated for each batch i. The incremental accumulation generation process can be referred to the following formula 12.

[0136] Formula 12: .

[0137] in, : The heat value trajectory of the current node k; : The sulfur trajectory of the current node k; The heat value trajectory of the parent node of the current node k; : The sulfur trajectory of the parent node of the current node k; : Furnace charging time for batch i; : Time interval indicator function.

[0138] After each node update, it can be immediately... and In contrast, if any point in the trajectory goes out of bounds, backtracking pruning is performed; if the trajectory remains within the furnace tolerance envelope, the node is preserved and the process continues downwards. In this way, a feasible subtree centered on steady-state continuity can be constructed with minimal branching, laying the foundation for a high-quality solution space for integerization and cost reduction.

[0139] Step 706: Solve the integer mapping problem for the target leaf nodes in the feasible subtree to obtain the target leaf node set. Determine the cost index and steady-state continuity index of each leaf node in the target leaf node set. Filter the nodes according to the cost index and steady-state continuity index to obtain the candidate solution set.

[0140] Integerized mapping refers to the operation of approximating the optimal solution stepwise through integer constraints. The cost metric measures the cost of leaf node domination. The steady-state continuity metric measures the continuity of a leaf node within its steady-state priority domain. Leaf nodes that meet both cost and steady-state continuity requirements are selected and included in the candidate solution set.

[0141] For example, during the process of integerizing leaf nodes and mapping costs, the current feasible subtree is... Leaf nodes that have reached the depth limit or have no remaining batches to choose from are extracted one by one, and the leaf node number is denoted as m. For the batch set carried by leaf node m... Perform integer mapping for each selected cross-layer edge on the path. Create binary variables Its value of 1 indicates that the corresponding logistics action is performed along this edge in the batch. All The process is divided into three categories: loading, transportation, and furnace charging. The three hard constraints of capacity, time window, and minimum batch size are translated into a coefficient matrix A and an upper limit vector b, and a comprehensive objective function as shown in Formula 13 is constructed.

[0142] Formula 13: .

[0143] in, The total logistics cost of leaf node m; The unit logistics cost of edge e; : The actual tonnage carried by side e; : Penalty coefficient for backup channels; : Indicator value, which is 1 if e passes through a virtual buffer node, otherwise 0.

[0144] exist Solve equation 13 using integer programming within the constraint framework. If convergence is achieved, then the optimal variable is... Write back to the leaf node and simultaneously calculate the terminal heat margin. with terminal sulfur margin If convergence is not achieved, the leaf node is marked as a model failure and skipped in subsequent traversals.

[0145] For example, rolling branching can also be triggered when the fuel dispatch center pushes coal quality revisions or logistics delay events in real time. First, the affected edge set is located by batch number. Then recursively search upwards for its location. The nearest feasible rendezvous node in To The subtree rooted at the root performs the following processing: reset the state of all disturbed nodes to unevaluated; retain the state of the rooted subtree. The existing upward path remains unchanged; only the disturbed subtree is re-applied with the aforementioned steady-state domain filtering and dynamic pruning logic; if a new subtree generates a leaf node, the integer mapping solution is re-performed. Through the local rollback and in-situ recalculation mechanism, the repair can be completed within a short time window without restarting the entire search tree, ensuring that the steady-state trunk of the furnace is not destroyed and avoiding redundant calculation delays.

[0146] For example, after traversing all leaf nodes and performing integerization in each round, the leaf node set is... Calculate the steady-state continuity fraction for each node. Read the end-effector heat margin at leaf node m. , terminal sulfur content margin And the rolling load demand for the next shift change And refer to the following formula 14 for evaluation calculation.

[0147] Formula 14: .

[0148] in, : Steady-state continuity fraction of leaf node m; Weighting coefficients are preset by the production department. Safety thresholds for calorific value and sulfur content; : The predicted heat load at leaf node m at the start of the next shift; Planned load for the next shift change window.

[0149] For each leaf node, a binary index is formed. The nodes are then added to the initial set, and Pareto front screening is performed to remove nodes that are dominated by higher costs and low sustainability, resulting in a set of candidate solutions. It can also be used to... The optimal solution is written to the execution schedule and resources are locked through the interface. At the same time, the secondary optimal solution is stored in the alternative library according to priority number so that it can be switched directly when new disturbances occur in front of the furnace, ensuring that the dual goals of production continuity and cost control are met in a timely manner.

[0150] In this embodiment, by adding cumulatively generated leaf nodes, a feasible subtree is constructed. The target leaf node is determined by integer mapping and node screening is performed. In the event of coal quality revision or logistics delay, the stability of the furnace can be maintained, and a coal blending scheme that meets production and cost targets can be output.

[0151] In an exemplary embodiment, as shown in FIG8, the candidate solutions in the candidate solution set are evaluated by indicators and ranked by risk priority to obtain an ordered list of feasible solutions, including steps 802 to 806.

[0152] Step 802: Calculate the fuel cost index and production risk index of the candidate scheme. The production risk index includes the probability of furnace shutdown.

[0153] Fuel cost indicators refer to the costs incurred in burning a batch of coal in a furnace, and are related to coal quality and material flow. Production risk indicators refer to indicators related to the normal operation of each production stage. For example, for continuous production lines, shutdown risk is a type of production risk. Shutdown probability refers to the predicted probability of a shutdown event.

[0154] For example, during the calculation of fuel cost and production risk indicators for candidate solutions, the list of feasible solutions can be loaded into the evaluation buffer in batches, and the solution identifiers can be read one by one. Its decision vector. For the calculation of fuel cost indicators, four costs—batch-level purchase price, port handling fees, yard transportation fees, and blending energy consumption—are extracted from the cost dictionary and mapped to the scheme dimension by batch tonnage, then summed to obtain the standardized fuel cost. To calculate production risk indicators, the calorific value trajectory and sulfur margin recorded by the candidate schemes are retrieved and compared with the furnace tolerance envelope on a unified time axis to estimate the probability of furnace shutdown. Multiply by the shutdown cost preset by the operations department. Obtain production risk In addition, it is also possible to... and By fixed weighting coefficients Add to composite index .

[0155] Step 804: Obtain the duration of continuous sulfur content exceeding the limit for candidate schemes, and eliminate schemes based on the shutdown probability and the duration of continuous sulfur content exceeding the limit.

[0156] The duration of continuous sulfur exceeding the limit refers to the continuous period during which the sulfur content emitted after combustion exceeds a certain limit, according to the coal blending method of the candidate scheme. Candidate schemes with a high probability of shutdown and long duration of continuous sulfur exceeding the limit can be eliminated, while schemes with lower remaining production risks and that meet emission requirements can be retained.

[0157] For example, for all candidate solutions in the evaluation buffer, the candidate solutions are checked. If the continuous sulfur content exceeds the limit, and the risk of shutdown or the continuous sulfur content exceeding the limit exceeds the operational threshold, the candidate will be eliminated, and the reason for elimination will be written as exceeding the risk limit in the metadata of the eliminated candidate.

[0158] Step 806: Sort the remaining solutions after elimination according to the fuel cost index and assign them priority numbers to obtain an ordered list of feasible solutions.

[0159] The remaining solutions after elimination can be sorted from low to high according to fuel cost as indicated by the fuel cost index, and assigned priority numbers according to the sorting order to obtain an ordered list of feasible solutions. For example, the remaining solutions are sorted by fuel cost. Arrange in ascending order from lowest to highest cost. If costs are tied, then rank them according to production risk. Sort the list again from lowest to highest priority. After sorting, write a unique priority number for each row of the list. Generate a sorted list of feasible solutions.

[0160] For example, the first number of the ordered feasible solution list can also be... The proposed scheme is written into the execution instruction module of the chemical coal blending platform. The template includes three core fields: batch number, charging sequence, and furnace start-up time, and additionally includes the furnace interface version number. It is then pushed to the field automation system via the scheduling API (Application Programming Interface) to be officially confirmed as the coal blending execution scheme for this round. (Regarding the batch number...) to The plan, along with its composite indicators The criteria for removal and the allowed switching windows are written to the candidate database to generate a fast switching index. When a sudden event occurs at the furnace and triggers the switching logic, the scheduling terminal can directly replace the alternative solution after index retrieval, ensuring that the dual objectives of production continuity and cost control remain undisturbed.

[0161] In this embodiment, the candidate solutions are evaluated for both cost and production risk to obtain an ordered list of feasible solutions. The top solution can be used as the final coal blending solution, while the other solutions with lower priority can be used as alternative solutions, ensuring that the dual objectives of production continuity and cost control are not disturbed.

[0162] In an exemplary embodiment, as shown in FIG9, the first feasible solution in the ordered feasible solution list is parsed into a device-readable instruction, and coal flow scheduling is performed according to the device-readable instruction, including steps 902 to 906.

[0163] Step 902: Map the device parameters in the first feasible solution to device readable instructions, and write the device readable instructions into the scheduling details table, which includes an automatic execution table and a manual review table.

[0164] The first feasible solution refers to the candidate solution ranked first in the ordered feasible solution list. The scheduling details table is a table storing coal flow scheduling-related data. The automatic execution table is a table for DCS (Distributed Control System) used for automated equipment control. The manual verification table is a table for the control room, used for manually controlling equipment status.

[0165] For example, the first feasible solution is parsed, and the stockpile number is determined. Material ratio Turning order belt number Feeding time slice Ship / vehicle serial number Furnace batch number Arrival Time Once the device parameters are mapped item by item into device-readable instructions, write all device-readable instructions into the system. and Complete the dual-table write. : An automatic execution table for DCS, with fields including device address, target status quantity, and execution timestamp. : Manual review form for the duty room, fields maintain consistency with Synchronization, but with additional operational guidelines and safety tips. After writing to both tables, the same version number can be recorded in the table headers. This ensures that the scheduling content on the platform side and the manual side is completely consistent.

[0166] Step 904: Send the equipment readable instructions in the automatic execution table to the furnace front control terminal through the first communication channel, and send the equipment readable instructions in the manual review table to the port terminal, storage yard terminal and transportation terminal through the second communication channel.

[0167] The first communication channel is used to issue equipment-readable instructions from the automatic execution table. The second communication channel is used to issue equipment-readable instructions from the manual review table. Since different surfaces are directed to different objects, different types of transmission methods can be used to issue instructions. The furnace front control terminal refers to the control terminal used to control the furnace. The port end, storage yard end, and transportation end are all communication terminals located at their respective physical locations.

[0168] For example, for The OPC support library is invoked to convert each row record into an OPCTag and push it to the furnace front control terminal. A message queue mechanism is used to simultaneously broadcast to the port, yard, and transportation ends. The system listens for acknowledgment messages from each end, and only after all acknowledgment statuses are marked as ACK (acknowledgment) is the message recorded in the instruction trace table. Write the "issued" status to the internal memory and start a countdown timer. .

[0169] Step 906: Determine the target inventory batch associated with the first feasible option and mark the target inventory batch as occupied in the dynamic inventory table.

[0170] The target inventory batch refers to the batch in the dynamic inventory table associated with the first feasible solution. Once marked as occupied, this batch can be locked to avoid scheduling conflicts. For example, this refers to the set of inventory batches involved in the first feasible solution. In the dynamic inventory table Write the occupancy tag LOCK to lock the quantity and quality fields. In the end-to-end graph G, add the corresponding cross-layer edges. Set to RESERVED to prevent other scheduling instances from using the same logistics channel. Start the real-time monitor after locking is complete. The monitoring system uses batch numbers as indexes to track location, yard flow, and conveyor belt speed. If any indicator deviates beyond a threshold... That is, the DEVIATE signal is published through the event bus to trigger the switching logic of the alternative scheme. .

[0171] In this embodiment, the device-readable instructions are sent to different terminals through different communication mechanisms to execute the coal blending operation, and the batches involved in the coal blending scheme are marked for occupation, which can prevent scheduling conflicts.

[0172] In an exemplary embodiment, as shown in FIG10, the chemical coal blending method includes the following steps:

[0173] S1. Real-time spectral acquisition of the feed conveyor belt and stockpile reclaimer. Lightweight edge computing network infers calorific value, sulfur content, and moisture content, generating batch mass vectors and confidence intervals. Near-infrared spectral probes are deployed at the feed conveyor belt and stockpile reclaimer for real-time spectral acquisition. The continuous coal flow is segmented and scanned in batches. Within the edge computing box, a lightweight network containing only convolutional layers and linear regression heads is used to infer batch-level calorific value, sulfur content, moisture content, and their confidence intervals in real time. The batch number, ore source, and timestamp are pushed to the dispatch center via an encrypted link along with the mass vector.

[0174] S2. Determine if the batch quality falls within the three-dimensional calorific value-sulfur content-time period furnace tolerance envelope. If yes, mark it as "Immediately Ready for Use," write it to the dynamic inventory table, and calculate the earliest arrival time and adjustable time window. If no, mark it as "Controlled for Use" or "Risk to be Discretionary," retain the record, and wait for subsequent scheduling opportunities. The calorific value-sulfur content-time period furnace tolerance envelope can be constructed by reading the future rolling load, emission threshold, and maintenance window of the furnace. A safety margin can be calculated for each batch, and its batch quality code can be marked. Further, the earliest arrival time and adjustable time window can be generated by combining the transit time from the port, storage yard, and conveyor belt to the furnace, and written to the dynamic inventory table along with the priority. Scheduling opportunities may include quality improvement or the arrival of a window.

[0175] S3. Construct a baseline time frame, integrate events such as shutdown / maintenance, and generate a two-layer time-series topology graph containing material and action layers. Perform throughput verification on candidate cross-layer edges, implant virtual buffer nodes, and apply economic penalties. Merge the boundaries of events such as shutdown and maintenance with the end point of the rolling cycle into a baseline time frame. Within this frame, establish material layer spatial nodes and batch sub-nodes, construct action layer nodes for turning, loading, blending, and furnace feeding, generate cross-layer edges according to action windows, and perform throughput verification on candidate edges passing through the same facility. Write quality labels and physical parameters to qualified edges, and simultaneously implant virtual buffer nodes and apply economic penalties, ultimately forming a two-layer time-series topology graph with a version number.

[0176] S4. Steady-state envelope screening combined with incremental pruning expansion generates feasible leaf nodes. Integer programming is used to solve the leaf nodes, outputting the optimal solution for logistics cost and reserve penalty under capacity / time window / minimum batch constraints. It is determined whether a logistics or coal quality disturbance event has been received. If so, rolling insertion local recalculation is performed, the disturbed subtree is reset and quickly repaired, and the recalculation result is output to proceed to the next step; otherwise, proceed directly to the next step. The search tree root state is formed by the set of cross-layer edges pointing to the furnace nodes and the real-time material level, temperature difference, and future load of the furnace. The sliding window method is called to generate the steady-state priority domain within the envelope. The mixed integer tree search is expanded using an incremental accumulation method and prunes immediately when the boundary is exceeded. Binary variables are established for leaf nodes that have reached the depth limit, and integer programming with logistics cost plus reserve penalty as the objective is solved under capacity, time window, and minimum batch constraints. When a disturbance occurs, local recalculation is performed through rolling insertion.

[0177] S5. Combine fuel cost and shutdown risk to conduct a composite index assessment, and generate an ordered feasible solution list after risk filtering. For the convergent blade node, calculate the standardized fuel cost based on purchase price, loading and unloading fees, transportation costs, and blending energy consumption. Combine calorific value and sulfur margin to estimate the shutdown probability and obtain the production risk. Form composite indices with fixed weights, perform risk filtering and cost priority ranking, and output an ordered feasible solution list.

[0178] S6. Parse the first feasible solution into a set of equipment-readable instructions, write it into the automatic execution table and the manual review table, lock the batch and cross-layer edge, and monitor the batch position and flow in real time. If the deviation exceeds the threshold, issue a DEVIATE signal to trigger the switching of the alternative solution. Parse the first sorting solution into equipment-readable instructions for the yard, conveyor belt, furnace front, etc., and write them into the automatic execution table facing the DCS and the manual review table facing the duty room, respectively, keeping them the same version. Send them through OPC and message queue dual channels. Then lock the relevant batch in the dynamic inventory table and mark the corresponding cross-layer edge as reserved. Monitor the batch position and flow deviation in real time. If the threshold is triggered, call the alternative solution to switch.

[0179] In this embodiment, near-infrared spectroscopy probes and lightweight regression networks are deployed at the feed conveyor belt and stockpile inlet in the coal quality sensing stage. This enables batch-level real-time inference and direct output of confidence intervals, ensuring that the temporal resolution and namespace of the coal quality data are consistent with the furnace feeding cycle, avoiding information delays and granularity mismatches. By using a reference time frame and a two-layer temporal topology, physical constraints such as port closures, stockpile maintenance, and conveyor capacity are transformed into node and edge attributes during the modeling stage and simultaneously incorporated into the decision space, rather than performing discrete event verification after the mix design is solved. Therefore, a solution that simultaneously satisfies both quality and logistics conditions can be obtained in one go without manual back-and-forth modifications. This integrates steady-state envelope determination, incremental pruning of hybrid integer tree search, and node-level quality... The combination of tags enables dynamic evaluation of calorific value-sulfur content trajectories and logistics capacity in the same search process. Compared with the serial structure of linear programming plus heuristic scheduling, it is easier to maintain furnace thermal balance and emission compliance. For sudden delays or coal quality revisions, local recalculation within the subtree can be completed by using rolling insertion. This preserves the locked transportation nodes and compresses the response time to the minute level, providing sustainable scheduling capabilities for high-frequency load changes. Overall, the synergy of real-time batch identification, graph-tree integrated solution and local fast recalculation can directly output executable and cost-controlled coal blending instructions, avoiding multi-engine file interaction and repeated manual splicing. It demonstrates structural technical advantages in terms of timeliness and coupling integrity.

[0180] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0181] Based on the same inventive concept, this application also provides a chemical coal blending device for implementing the aforementioned chemical coal blending method. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the chemical coal blending device provided below can be found in the limitations of the chemical coal blending method described above, and will not be repeated here.

[0182] In an exemplary embodiment, as shown in FIG11, a chemical coal blending device 1100 is provided, including: a spectral data processing module 1101, a quality coding determination module 1102, a time-series topology graph construction module 1103, a candidate solution set determination module 1104, an ordered feasible solution list determination module 1105, and a feasible solution parsing module 1106, wherein:

[0183] The spectral data processing module 1101 is used to acquire the spectral data of the coal flow to be allocated, perform batch segmentation and regression processing on the spectral data, and obtain the batch quality vector of the coal flow to be allocated.

[0184] The quality coding determination module 1102 is used to determine the batch quality code of the batch quality vector based on the furnace tolerance envelope.

[0185] The timing topology construction module 1103 is used to determine the baseline time frame and construct a timing topology diagram including the material layer and the action layer based on the baseline time frame.

[0186] The candidate scheme set determination module 1104 is used to determine the target cross-layer edge set pointing to the furnace node in the time-series topology graph, determine the root node state of the search tree based on the target cross-layer edge set, and perform a mixed integer tree search in combination with the priority bit of the batch quality code to obtain the candidate scheme set.

[0187] The ordered feasible solution list determination module 1105 is used to evaluate the indicators and prioritize the risks of the candidate solutions in the candidate solution set to obtain an ordered feasible solution list.

[0188] The feasible solution parsing module 1106 is used to parse the first feasible solution in the ordered feasible solution list into a device-readable instruction, and to perform coal flow scheduling based on the device-readable instruction.

[0189] In an exemplary embodiment, the spectral data processing module 1101 is further configured to: acquire a batch identification pulse; divide the spectral data into multiple spectral segments based on the batch identification pulse; predict the combustion quality of the coal stream to be allocated based on the component detection information of the ore source to which the spectral segments belong; and obtain a batch quality vector, wherein the batch quality vector includes the calorific value, sulfur content, and moisture content of the coal stream to be allocated.

[0190] In an exemplary embodiment, the quality coding determination module 1102 is further configured to: determine the envelope boundary of the furnace tolerance envelope, calculate the minimum remaining amount of the batch quality vector and the envelope boundary, and determine the batch quality code based on the minimum remaining amount and a preset safety margin.

[0191] In an exemplary embodiment, the time-series topology graph construction module 1103 is further configured to: create spatial nodes for each entity position in the baseline time frame, write the spatial nodes into a first set, bind quality status child nodes as child nodes of the spatial nodes, write them into a second set, combine the first set and the second set into a material layer with a parent-child relationship, and obtain the quality status child nodes by traversing the inventory vectors in the dynamic inventory table; create action nodes within the action start and end window for each action in the action library, write them into a third set, and structure the third set into an action layer; and couple the material layer and the action layer into a time-series topology graph through a candidate cross-layer edge set.

[0192] In an exemplary embodiment, the chemical coal blending device 1100 further includes a cross-layer edge processing module, used to: map each cross-layer edge in the candidate cross-layer edge set to the physical facility it passes through; perform throughput verification on cross-layer edges of the same physical facility within the same action window; and block cross-layer edges that fail the throughput verification; add virtual buffer nodes in the time-series topology graph; connect the virtual buffer nodes to key nodes in the material layer through zero-latency edges; and add an economic penalty to the cross-layer edges retained in the candidate cross-layer edge set when the cross-layer edge crosses a virtual buffer node.

[0193] In an exemplary embodiment, the candidate scheme set determination module 1104 is further configured to: extract the quality information and arrival time information of each cross-layer edge in the target cross-layer edge set; add the quality information and arrival time information to the furnace entry candidate table; inject the furnace entry candidate table and its version number into the root node of the search tree; use the root node as the starting point, perform incremental accumulation to generate leaf nodes based on the batch information and priority of the coal flow to be allocated, and obtain a feasible subtree; perform integer mapping to solve the target leaf nodes in the feasible subtree to obtain a target leaf node set; determine the cost index and steady-state continuity index of each leaf node in the target leaf node set; and perform node screening based on the cost index and steady-state continuity index to obtain a candidate scheme set.

[0194] In an exemplary embodiment, the ordered feasible solution list determination module 1105 is further configured to: calculate the fuel cost index and production risk index of the candidate solutions, the production risk index including the shutdown probability; obtain the continuous sulfur content exceeding the limit duration of the candidate solutions, and eliminate the solutions according to the shutdown probability and the continuous sulfur content exceeding the limit duration; sort the remaining solutions after elimination according to the fuel cost index and assign priority numbers to obtain the ordered feasible solution list.

[0195] In an exemplary embodiment, the feasible solution parsing module 1106 is further configured to: map the equipment parameters in the first feasible solution to equipment readable instructions, write the equipment readable instructions into a scheduling detail table, the scheduling detail table including an automatic execution table and a manual review table; send the equipment readable instructions in the automatic execution table to the furnace front control terminal through a first communication channel, and send the equipment readable instructions in the manual review table to the port terminal, the yard terminal, and the transportation terminal through a second communication channel; determine the target inventory batch related to the first feasible solution, and mark the target inventory batch as occupied in the dynamic inventory table.

[0196] Each module in the aforementioned chemical coal blending device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0197] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram is shown in Figure 12. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database of the computer device stores spectral data of the coal flow to be allocated. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a chemical coal blending method.

[0198] Those skilled in the art will understand that the structure shown in Figure 12 is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0199] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0200] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0201] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0202] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0203] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0204] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for blending coal in chemical processes, characterized in that, The method includes: acquiring spectral data of the coal flow to be allocated; performing batch segmentation and regression processing on the spectral data to obtain a batch quality vector of the coal flow to be allocated; determining the batch quality code of the batch quality vector based on the furnace tolerance envelope; determining a baseline time frame; constructing a temporal topology graph including a material layer and an action layer based on the baseline time frame; determining a set of target cross-layer edges pointing to furnace nodes in the temporal topology graph; determining the root node state of the search tree based on the set of target cross-layer edges; and performing a mixed integer tree search in combination with the priority bits of the batch quality code to obtain a candidate solution set; evaluating the candidate solutions in the candidate solution set based on indicators and ranking them by risk priority to obtain an ordered feasible solution list; parsing the first feasible solution in the ordered feasible solution list into a device-readable instruction; and performing coal flow scheduling based on the device-readable instruction.

2. The method according to claim 1, characterized in that, The step of performing batch segmentation and regression processing on the spectral data to obtain the batch quality vector of the coal stream to be allocated includes: acquiring a batch identification pulse; segmenting the spectral data into multiple spectral segments based on the batch identification pulse; predicting the combustion quality of the coal stream to be allocated based on the component detection information of the ore source to which the spectral segments belong; and obtaining the batch quality vector, wherein the batch quality vector includes the calorific value, sulfur content, and moisture content of the coal stream to be allocated.

3. The method according to claim 1, characterized in that, The step of determining the batch quality code of the batch quality vector based on the furnace tolerance envelope includes: determining the envelope boundary of the furnace tolerance envelope, calculating the minimum remaining amount between the batch quality vector and the envelope boundary, and determining the batch quality code based on the minimum remaining amount and a preset safety margin.

4. The method according to claim 1, characterized in that, The construction of a time-series topology graph containing a material layer and an action layer based on the baseline time frame includes: creating a spatial node for each entity position in the baseline time frame, writing the spatial node into a first set, binding quality status child nodes as child nodes of the spatial node, writing them into a second set, combining the first set and the second set into the material layer using parent-child relationships, the quality status child nodes being obtained by traversing the inventory vectors in the dynamic inventory table; creating an action node within the action start and end window for each executed action in the action library, writing it into a third set, structuring the third set into the action layer; and coupling the material layer and the action layer into the time-series topology graph through a candidate cross-layer edge set.

5. The method according to claim 4, characterized in that, After coupling the material layer and the action layer into the timing topology graph using the candidate cross-layer edge set, the method further includes: mapping each cross-layer edge in the candidate cross-layer edge set to the physical facility it passes through; performing throughput verification on the cross-layer edges of the same physical facility within the same action window; and blocking cross-layer edges that fail the throughput verification; adding virtual buffer nodes to the timing topology graph; connecting the virtual buffer nodes to key nodes in the material layer using zero-latency edges; and imposing an economic penalty on the cross-layer edges retained in the candidate cross-layer edge set when the cross-layer edge traverses the virtual buffer node.

6. The method according to claim 1, characterized in that, The process of determining the root node state of the search tree based on the target cross-layer edge set and performing a mixed integer tree search in conjunction with the priority bits of the batch quality code to obtain a candidate solution set includes: extracting the quality information and arrival time information of each cross-layer edge in the target cross-layer edge set; adding the quality information and arrival time information to the furnace entry candidate table; injecting the furnace entry candidate table and its version number into the root node of the search tree; using the root node as the starting point, incrementally accumulating leaf nodes based on the batch information of the coal flow to be allocated and the priority bits to obtain a feasible subtree; performing integer mapping on the target leaf nodes in the feasible subtree to obtain a target leaf node set; determining the cost index and steady-state continuity index of each leaf node in the target leaf node set; and filtering nodes based on the cost index and steady-state continuity index to obtain the candidate solution set.

7. The method according to claim 1, characterized in that, The step of evaluating and prioritizing the candidate solutions in the candidate solution set to obtain an ordered list of feasible solutions includes: calculating the fuel cost index and production risk index of the candidate solutions, wherein the production risk index includes the probability of shutdown; obtaining the duration of continuous sulfur content exceeding the limit for the candidate solutions, and eliminating solutions based on the shutdown probability and the duration of continuous sulfur content exceeding the limit; and sorting the remaining solutions after elimination according to the fuel cost index and assigning them priority numbers to obtain the ordered list of feasible solutions.

8. The method according to any one of claims 1 to 7, characterized in that, The step of parsing the first feasible solution in the ordered feasible solution list into a device-readable instruction and performing coal flow scheduling based on the device-readable instruction includes: mapping the device parameters in the first feasible solution into the device-readable instruction; writing the device-readable instruction into a scheduling detail table, the scheduling detail table including an automatic execution table and a manual review table; sending the device-readable instruction in the automatic execution table to the furnace front control terminal through a first communication channel, and sending the device-readable instruction in the manual review table to the port terminal, the storage yard terminal, and the transportation terminal through a second communication channel; determining the target inventory batch related to the first feasible solution, and marking the target inventory batch as occupied in the dynamic inventory table.

9. A chemical coal blending device, characterized in that, The device includes: a spectral data processing module for acquiring spectral data of the coal flow to be allocated, performing batch segmentation and regression processing on the spectral data to obtain a batch quality vector of the coal flow to be allocated; a quality coding determination module for determining the batch quality code of the batch quality vector based on the furnace tolerance envelope; a temporal topology graph construction module for determining a reference time frame and constructing a temporal topology graph including a material layer and an action layer based on the reference time frame; a candidate solution set determination module for determining the target cross-layer edge set pointing to the furnace node in the temporal topology graph, determining the root node state of the search tree based on the target cross-layer edge set, and performing a mixed integer tree search in combination with the priority bit of the batch quality code to obtain a candidate solution set; an ordered feasible solution list determination module for evaluating the indicators and prioritizing the candidate solutions in the candidate solution set to obtain an ordered feasible solution list; and a feasible solution parsing module for parsing the first feasible solution in the ordered feasible solution list into a device-readable instruction and performing coal flow scheduling based on the device-readable instruction.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 8.

Images