Coal bunker monitoring method, system, device and storage medium

Abstract

The application discloses a coal bunker monitoring method, system, device and storage medium, and belongs to the technical field of digital twinning. The method comprises the following steps: constructing a digital twin body through multi-source heterogeneous data, and forming a dynamic environment perception model; deploying a plurality of detection nodes based on the dynamic environment perception model, optimizing the decision model of each detection node, and establishing a cooperation framework among the plurality of detection nodes through a graph neural network to obtain a monitoring network; simulating the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions based on the dynamic environment perception model, evaluating the material accumulation mode and stress concentration points, iteratively adjusting material management strategy parameters to obtain an optimization result, and feeding back to a coal bunker control system to generate an adaptive adjustment scheme. The embodiment of the application can improve the accuracy of coal bunker state monitoring, predict safety abnormal conditions occurring in the coal bunker, and generate an adaptive adjustment scheme, thereby improving the safety of coal storage and the utilization efficiency of storage space.

Description

Technical Field

[0001] This application relates to the field of digital twin technology, and in particular to a coal bunker monitoring method, system, device and storage medium. Background Technology

[0002] In modern industrial environments, coal bunkers, as core facilities for coal storage and transportation, directly determine the efficiency and safety of the entire production process. To ensure the safe and stable operation of coal bunkers, existing technologies mainly rely on traditional sensor networks and rule-based control systems: data is collected by deploying sensors for physical quantities such as temperature, humidity, and pressure within the coal bunker, and then operations such as ventilation and material discharge are adjusted according to preset thresholds or rule sets; at the same time, basic fluid dynamics simulation software is used to conduct simple material flow simulations.

[0003] However, such software cannot accurately assess stress concentration points under complex working conditions, and traditional control systems have limited data processing capabilities, making it difficult to effectively integrate multi-source heterogeneous data and build a comprehensive and detailed high-fidelity digital model of the coal bunker, resulting in insufficient monitoring accuracy. Summary of the Invention

[0004] The purpose of this application is to provide a coal bunker monitoring method, system, device, and storage medium, which can effectively improve the accuracy of coal bunker status monitoring, predict safety anomalies occurring inside the coal bunker, and generate adaptive adjustment schemes, thereby improving the safety of coal storage and the efficiency of storage space utilization.

[0005] To achieve the above objectives, a first aspect of this application provides a coal bunker monitoring method, comprising: High-fidelity modeling of the coal bunker is performed using multi-source heterogeneous data to construct a digital twin that matches the coal bunker, forming a dynamic environment perception model. Based on the dynamic environment perception model, multiple detection nodes are deployed in the coal bunker. The decision model of each detection node is optimized using machine learning algorithms, and a collaborative framework among the multiple detection nodes is established through a graph neural network to obtain a monitoring network. Based on the dynamic environment perception model, the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions are simulated. The material accumulation pattern and stress concentration points are evaluated by simulation analysis. The material management strategy parameters are iteratively adjusted using optimization algorithms to obtain optimization results. The optimization results are then fed back to the coal bunker control system to generate an adaptive adjustment scheme.

[0006] Compared with existing technologies, the coal bunker monitoring method provided in this application has the following advantages: It constructs a high-fidelity digital twin of the coal bunker using multi-source heterogeneous data, forming a dynamic environmental perception model. This lays a precise data foundation for coal bunker status monitoring and control, overcoming the limitations of traditional single data acquisition and static models. Based on this model, detection nodes are deployed, and the node decision logic is optimized through machine learning, while a collaborative framework is built using graph neural networks. This enables the monitoring network to have adaptive and collaborative decision-making capabilities, effectively avoiding the monitoring blind spots and misjudgments of isolated sensors, and improving the accuracy and response speed of anomaly identification. Simultaneously, based on the dynamic perception model, simulation analysis of material flow characteristics and pressure distribution under multiple operating conditions is conducted to accurately locate material accumulation patterns and stress concentration points. The management strategy is iteratively adjusted through optimized algorithms and fed back to the physical control system to generate adaptive adjustment schemes. This achieves both forward-looking prediction and precise prevention and control of coal bunker operation risks, and optimizes material management efficiency, thereby improving the safety of coal storage and the efficiency of storage space utilization, and comprehensively enhancing the safety, stability, and intelligence level of coal bunker operation.

[0007] In some embodiments, based on the dynamic environment perception model, multiple detection nodes are deployed within the coal bunker, machine learning algorithms are used to optimize the decision models of each detection node, and a collaborative framework among the multiple detection nodes is established through a graph neural network to obtain a monitoring network, including: Based on the dynamic environment perception model, key monitoring locations within the coal bunker are determined, detection nodes with autonomous learning capabilities are deployed, and the parameters of the detection nodes are set through initialization processing. Define a standardized communication protocol between detection nodes and establish an information sharing mechanism; Based on the aforementioned information sharing mechanism, a data interaction network between detection nodes is constructed; A graph neural network framework is constructed based on the data interaction network. The graph neural network framework is used to train and adjust the cooperative relationship between detection nodes, forming a dynamic cooperative framework and a monitoring network.

[0008] In some embodiments, after constructing a graph neural network framework based on the data interaction network, and using the graph neural network framework to train and adjust the cooperative relationships between detection nodes to form a dynamic cooperative framework and a monitoring network, the method further includes: By combining the data interaction network, the dynamic collaboration framework, and the real-time data collected by the detection nodes, the decision-making model of the detection nodes is continuously optimized through machine learning mechanisms to adapt to changing working conditions and data patterns.

[0009] In some embodiments, the step of combining the data interaction network, the dynamic collaboration framework, and the real-time data collected by the detection nodes, and continuously optimizing the decision model of the detection nodes through machine learning mechanisms to adapt to changing operating conditions and data patterns, includes: The decision model of the detection node is incrementally updated using the real-time data collected by the detection node, and a training dataset is obtained through a data filtering algorithm. Based on the training dataset, the decision model of the detection node is trained by combining reinforcement learning mechanism and gradient boosting decision tree algorithm to obtain a local decision tree model. Then, based on the local decision tree model and privacy protection technology, the model is co-trained to obtain a global decision tree model. Based on the global decision tree model, the model parameters are fine-tuned using a parameter optimization algorithm to obtain the behavior criteria of the detection nodes, and then the dynamic collaboration framework is combined to generate consistent node behavior criteria. Based on the aforementioned consistent node behavior guidelines, when a detection node identifies a complex problem, it evaluates the optimal response strategy based on the data interaction network and multi-agent system, and determines the cooperation path and generates a solution through a search algorithm.

[0010] In some embodiments, the step of co-training the model based on the local decision tree model and privacy protection techniques to obtain a global decision tree model includes: Based on the local decision tree model and federated aggregation, a preliminary global model is obtained; Based on the initial global model and personalized fine-tuning, a personalized local model adapted to local characteristics is obtained. A global decision tree model is obtained based on distributed hyperparameter optimization and iterative update loop.

[0011] In some embodiments, based on the dynamic environment perception model, the material flow characteristics and pressure distribution changes within the coal bunker under different operating conditions are simulated. Simulation analysis is used to evaluate material accumulation patterns and stress concentration points. An optimization algorithm is used to iteratively adjust material management strategy parameters to obtain optimization results, which are then fed back to the coal bunker control system to generate an adaptive adjustment scheme, including: Based on the dynamic environment perception model and combined with multi-physics field coupling analysis, high-precision simulations were performed on the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions, and simulation results were obtained. Based on the simulation results, the material accumulation mode and stress concentration points are evaluated by nonlinear material behavior modeling, and the material accumulation and stress distribution map is generated by topology optimization method. Based on the material accumulation and stress distribution map, the material management strategy parameters are iteratively adjusted using a Bayesian optimization algorithm, and the optimal operation strategy is autonomously learned using a reinforcement learning framework. The performance indicators under different parameter settings are evaluated using a random forest regression prediction model to obtain the optimization results. The optimization results are adjusted based on real-time feedback information from the industrial IoT platform to generate control parameters that adapt to the current operating conditions. Digital twin technology is used to achieve synchronous operation of the physical and virtual coal bunkers. Anomaly detection algorithms are used to promptly identify and respond to potential problems, generate adaptive adjustment schemes, and update the configuration of the coal bunker control system based on the adaptive adjustment schemes to guide actual operation and adapt to changes in operating conditions.

[0012] In some embodiments, the method further includes: A secure data synchronization channel is established between a digital twin and a physical coal bunker using blockchain technology, and sensitive data during transmission is protected using homomorphic encryption technology. When a detection node detects a potential risk, a classifier is used to perform preliminary screening of the alarm to filter out false alarms, and a multi-objective optimization algorithm is launched to generate a solution set.

[0013] To achieve the above objectives, a second aspect of this application provides a coal bunker monitoring system, the system comprising: The modeling module is used to perform high-fidelity modeling of the coal bunker using multi-source heterogeneous data, construct a digital twin that matches the coal bunker, and form a dynamic environment perception model. The detection module is used to deploy multiple detection nodes in the coal bunker based on the dynamic environment perception model, optimize the decision model of each detection node using machine learning algorithms, and establish a collaborative framework among the multiple detection nodes through a graph neural network to obtain a monitoring network. The adjustment module is used to simulate the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions based on the dynamic environment perception model. It uses simulation analysis to evaluate the material accumulation pattern and stress concentration points, uses optimization algorithms to iteratively adjust the material management strategy parameters to obtain optimization results, and feeds the optimization results back to the coal bunker control system to generate an adaptive adjustment scheme.

[0014] To achieve the above objectives, a third aspect of this application provides an electronic device, the electronic device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the method described in the first aspect.

[0015] To achieve the above objectives, a fourth aspect of the present application provides a computer-readable storage medium comprising a stored computer program, wherein the computer program, when executed, controls the device containing the computer-readable storage medium to perform the method described in the first aspect.

[0016] To achieve the above objectives, a fifth aspect of the present application provides a computer program product, which includes a computer program or computer instructions, wherein the computer program or computer instructions, when executed by a processor, implement the method described in the first aspect. Attached Figure Description

[0017] Figure 1 This is a flowchart of a coal bunker monitoring method provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of a coal bunker monitoring system provided in an embodiment of this application; Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0019] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0020] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0021] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0022] In modern industry, the operational status of coal bunkers directly affects the efficiency and safety of production processes. Currently, their monitoring mainly relies on traditional sensor networks and rule-based control systems. These systems collect data by deploying sensors for temperature, humidity, and pressure within the bunker, and adjust operations such as ventilation and material discharge based on preset thresholds, while also using basic fluid dynamics simulations to perform a rough simulation of material flow.

[0023] However, this approach has significant limitations: firstly, its data processing capabilities are insufficient, making it difficult to effectively integrate multi-source heterogeneous data and construct a comprehensive and detailed digital model of the coal bunker; secondly, existing simulation tools are mostly applicable to static conditions, lacking sufficient accuracy in assessing complex issues such as stress concentration under dynamic working conditions. Therefore, while traditional methods can achieve basic monitoring, they lack forward-looking predictive analysis capabilities, making it difficult to detect anomalies in a timely manner and respond effectively, and are particularly ineffective in dealing with complex and variable working conditions.

[0024] Based on this, embodiments of this application provide a coal bunker monitoring method, system, device, and storage medium, which can effectively improve the accuracy of coal bunker status monitoring, predict safety anomalies occurring inside the coal bunker, and generate adaptive adjustment schemes, thereby improving the safety of coal storage and the efficiency of storage space utilization.

[0025] Please see Figure 1 , Figure 1 This is an optional flowchart of the coal bunker monitoring method provided in the embodiments of this application. Figure 1 The method may include, but is not limited to, steps S101 to S103.

[0026] Step S101: High-fidelity modeling of the coal bunker is performed using multi-source heterogeneous data to construct a digital twin that matches the coal bunker, forming a dynamic environmental perception model; Step S102: Based on the dynamic environment perception model, multiple detection nodes are deployed in the coal bunker. Machine learning algorithms are used to optimize the decision model of each detection node, and a collaborative framework between multiple detection nodes is established through graph neural networks to obtain the monitoring network. Step S103: Based on the dynamic environment perception model, the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions are simulated. The material accumulation mode and stress concentration points are evaluated by simulation analysis. The material management strategy parameters are iteratively adjusted using optimization algorithms to obtain optimization results. The optimization results are then fed back to the coal bunker control system to generate an adaptive adjustment scheme.

[0027] Steps S101 to S103 of this application embodiment construct a high-fidelity digital twin of the coal bunker using multi-source heterogeneous data, forming a dynamic environmental perception model. This lays a precise data foundation for coal bunker status monitoring and control, breaking through the limitations of traditional single data acquisition and static models. Based on this model, detection nodes are deployed, and machine learning optimizes node decision logic, while graph neural networks build a collaborative framework. This enables the monitoring network to have adaptive and collaborative decision-making capabilities, effectively avoiding the monitoring blind spots and misjudgments of isolated sensors, and improving the accuracy and response speed of anomaly identification. Simultaneously, simulation analysis of material flow characteristics and pressure distribution under multiple operating conditions is conducted based on the dynamic perception model, accurately locating material accumulation patterns and stress concentration points. The management strategy is iteratively adjusted through optimized algorithms and fed back to the physical control system to generate adaptive adjustment schemes. This achieves both forward-looking prediction and precise control of coal bunker operation risks and optimizes material management efficiency, thereby improving the safety of coal storage and the efficiency of storage space utilization, and comprehensively enhancing the safety, stability, and intelligence level of coal bunker operation.

[0028] In step S101 of some embodiments, the multi-source heterogeneous data can be a collection of data from different sources (such as temperature sensors, humidity sensors, pressure sensors, video surveillance systems, production scheduling systems, historical operation records, etc.), with different data types and formats; high-fidelity modeling can be a modeling method that restores the physical structure, operating environment, and material characteristics of the coal bunker with extremely high accuracy, making the model's matching degree with the physical coal bunker close to the real state; the digital twin can be a virtual copy constructed in digital space that is completely mapped to the physical attributes, operating status, and behavioral characteristics of the physical coal bunker, and can achieve real-time synchronous interaction with the physical coal bunker; the dynamic environment perception model can be a model based on the digital twin that can capture and update changes in the coal bunker's operating environment (such as material status and environmental parameters) in real time, and has dynamic perception capabilities.

[0029] First, heterogeneous data from various sources, including sensors, monitoring systems, scheduling systems, and historical operation records, are collected from the coal bunker. After cleaning, fusion, and standardization, a high-precision digital twin is constructed using data-driven techniques, combined with the coal bunker's physical structural parameters, material characteristic parameters, and operating conditions, employing technologies such as Generative Adversarial Networks (GANs). A real-time mapping relationship is established between the digital twin and the physical coal bunker, enabling the digital twin to synchronously reflect the structural state, operating parameters, and environmental changes of the physical coal bunker. Based on the digital twin, a real-time data acquisition and update module is embedded, giving the model the ability to perceive, capture, predict, and update the dynamic operating environment of the coal bunker. This embodiment significantly improves the accuracy of coal bunker status monitoring and anomaly prediction, enabling the coal bunker control system to respond promptly to changes in operating conditions and enhancing the system's adaptability and flexibility.

[0030] In step S102 of some embodiments, the detection node can be a hardware unit deployed inside the coal bunker, capable of data acquisition, preliminary analysis, and decision-making; the machine learning algorithm can refer to an algorithm that can learn patterns and optimize decision logic through data training (such as decision trees, neural networks, support vector machines, etc.); the decision model of the detection node can be a logical model used by the detection node to determine whether the monitoring data is abnormal and whether the monitoring strategy needs to be adjusted, such as a decision tree; the graph neural network can be a neural network specifically for processing data with graph structure, capable of capturing the relationships between nodes; the collaboration framework can be a logical architecture that enables multiple detection nodes to achieve information interaction and collaborative judgment; and the monitoring network can be an overall monitoring system with collaborative decision-making capabilities formed by connecting multiple detection nodes through the collaboration framework.

[0031] Based on the structural characteristics and potential risk areas of the coal bunker reflected by the dynamic environmental perception model, detection nodes are rationally deployed at key locations. Data collected by these nodes is input into machine learning algorithms for training, allowing the decision-making model to autonomously learn abnormal data characteristics and optimize its judgment logic. Graph neural networks are used to capture the spatial and data relationships between the detection nodes, constructing a logical architecture for information interaction and collaborative judgment among nodes. This ensures that the detection nodes can share information and collaboratively handle problems within the coal bunker, thereby generating an efficient monitoring network. Each detection node can not only independently perform local data analysis but also cooperate with other detection nodes to jointly address complex problems. This embodiment of the application, by introducing detection nodes and reinforcement learning algorithms, enables the system to possess self-optimization capabilities, providing early warnings of potential risks. Furthermore, through efficient collaboration among detection nodes, it reduces false alarm and false negative rates, improving overall monitoring efficiency and reliability.

[0032] In some embodiments, step S102 may include, but is not limited to, steps S201 to S204: Step S201: Determine the key monitoring locations in the coal bunker based on the dynamic environment perception model, deploy detection nodes with autonomous learning capabilities, and set the detection node parameters through initialization processing; Step S202: Define a standardized communication protocol between detection nodes and establish an information sharing mechanism; Step S203: Based on the information sharing mechanism, construct a data interaction network between detection nodes; Step S204: Construct a graph neural network framework based on the data interaction network, and use the graph neural network framework to train and adjust the cooperative relationship between detection nodes to form a dynamic cooperative framework and a monitoring network. In step S201 of some embodiments, the key monitoring location can be a specific area within the coal bunker that is prone to material accumulation, stress concentration, and abnormal temperature / pressure parameters, and that has a decisive impact on the safe operation and storage efficiency of the coal bunker, such as the corner of the bunker wall, below the feed inlet, around the discharge outlet, and the stress-sensitive area in the middle of the bunker. The initialization process can be the process of initially calibrating and setting the core operating logic and configuration parameters of the detection node before it is officially put into operation. It is a basic operation to ensure that the node can quickly adapt to monitoring needs. The detection node parameters can be the core configuration data that supports the normal operation of the detection node, including data acquisition frequency, anomaly judgment threshold, data upload interval, communication protocol parameters, and autonomous learning trigger conditions.

[0033] Based on a dynamic environmental perception model and its mapped structural characteristics and operational risk distribution patterns in the coal bunker, key monitoring locations prone to material accumulation, stress concentration, and parameter anomalies are accurately identified. Subsequently, detection nodes are physically deployed at these key monitoring locations. Finally, through initialization processing, combined with the coal bunker's basic operating conditions, historical data, and monitoring requirements, core parameters such as acquisition frequency and anomaly thresholds are scientifically set for each detection node, enabling it to possess initial monitoring capabilities. This step, through the dynamic environmental perception model, accurately locates key monitoring positions, ensuring targeted deployment of detection nodes, comprehensive coverage of high-risk areas, and avoidance of monitoring blind spots, significantly improving the monitoring system's coverage and accuracy of key coal bunker conditions. Simultaneously, the scientific setting of detection node parameters lays the foundation for its self-learning ability, enabling it to quickly adapt to operating conditions after startup, providing accurate and efficient data acquisition support for subsequent anomaly identification and collaborative decision-making.

[0034] In step S202 of some embodiments, the standardized communication protocol can be a unified rule (including data format, transmission timing, interaction logic, etc.) for data exchange between detection nodes, ensuring that different detection nodes follow a consistent standard for communication; the information sharing mechanism can be a process and rule system based on the standardized communication protocol to realize real-time data transmission and status communication between detection nodes.

[0035] For the deployed detection nodes, a unified standardized communication protocol is defined based on their hardware characteristics and data interaction requirements. An information sharing mechanism is established based on this protocol to enable real-time information transmission and status communication between detection nodes, ensuring efficient data exchange according to a unified standard. The standardized communication protocol eliminates communication barriers between detection nodes, guaranteeing the reliability and efficiency of data transmission. The information sharing mechanism enables status coordination and data exchange between nodes, providing efficient information support for subsequent collaborative decision-making and enhancing the overall coordination of the monitoring system.

[0036] In step S203 of some embodiments, the data interaction network can be a network structure that connects detection nodes through an information sharing mechanism to achieve real-time data exchange. Based on the aforementioned information sharing mechanism, each detection node connects according to unified rules and exchanges data in real time, forming a data interaction network between detection nodes. This data interaction network enables real-time information exchange between detection nodes, breaks down local data barriers, provides basic network support for subsequent collaborative decision-making, and improves the system's overall perception and response speed and flexibility regarding the coal bunker's status.

[0037] In step S204 of some embodiments, the graph neural network (GNN) framework is a neural network architecture specifically designed to process graph structure data. It can capture the relationship between nodes and edges and is suitable for learning the collaborative logic between detection nodes. The collaborative relationship between detection nodes can be the collaborative working logic relationship formed by each detection node in the process of data interaction, anomaly judgment, etc. The dynamic collaboration framework refers to a flexible collaboration system that can adjust the node collaboration logic in real time according to the changes in coal bunker working conditions and node data feedback. Based on a data interaction network, a graph neural network framework is constructed—mapping each detection node as a node in a graph, and mapping the logical connections between nodes as edges. Then, historical and real-time data are input to train the network, learning optimal cooperation patterns and adjusting the cooperative relationships between detection nodes. This ultimately forms a dynamic cooperation framework, where multiple detection nodes are connected through this framework to form a monitoring network with collaborative decision-making capabilities. Historical data refers to the accumulated data from past coal bunker operations and node monitoring, while real-time data refers to the current coal bunker status data collected by the detection nodes. The graph neural network framework accurately captures the correlation patterns between detection nodes, enabling intelligent learning and dynamic adjustment of cooperative relationships. This makes the cooperation between detection nodes more adaptable to the complex and ever-changing coal bunker conditions, significantly improving the collaborative decision-making accuracy and operational reliability of the monitoring system.

[0038] In some other embodiments, after step S204, step S205 is also included: combining the data interaction network, the dynamic collaboration framework, and the real-time data collected by the detection node, the decision model of the detection node is continuously optimized through a machine learning mechanism to adapt to changing working conditions and data patterns.

[0039] Specifically, in step S205 of some embodiments, the real-time data collected by the detection node can be the current state data of the coal bunker continuously acquired by the detection node during operation (such as temperature, pressure, material flow parameters, etc.); the machine learning mechanism can be an algorithm system that learns rules and optimizes models through data-driven autonomous learning (which can be understood here as a learning logic that can realize dynamic iteration of the model); the changing working conditions can be the dynamic changes in the operating state of the coal bunker (such as fluctuations in feed rate, changes in environmental parameters, etc.); and the data pattern can be the distribution characteristics and correlation patterns of the real-time data (dynamically adjusted with changes in working conditions).

[0040] Based on the decision-making model of the detection nodes, and combined with the data interaction network, dynamic collaboration framework, and real-time status data of the coal bunker collected by them, the parameters and logic of the decision-making model are dynamically and iteratively optimized using the self-learning capability of machine learning mechanisms. This enables the decision-making model to adapt to changing operating conditions and corresponding new data patterns in the coal bunker. Through continuous optimization of the decision-making model, it is ensured that the detection nodes can still accurately analyze and make efficient decisions when operating conditions and data patterns change, significantly improving the nodes' adaptability and enhancing the robustness and long-term operational stability of the monitoring system.

[0041] In summary, this application's embodiments construct a highly intelligent and adaptive coal bunker monitoring system through a dynamic environment perception model, detection node deployment, reinforcement learning algorithms, graph neural network framework, and continuous optimization mechanisms. This not only improves the accuracy and efficiency of coal bunker monitoring but also enhances its adaptability and reliability, providing a strong guarantee for the safe and stable operation of the coal bunker.

[0042] In some embodiments, step S205 may include, but is not limited to, steps S301 to S304: Step S301: Use the real-time data collected by the detection nodes to incrementally update the decision model of the detection nodes, and obtain the training dataset through the data filtering algorithm. Step S302: Based on the training dataset, the decision model of the detection node is trained by combining reinforcement learning mechanism and gradient boosting decision tree algorithm to obtain a local decision tree model. Then, the model is co-trained based on the local decision tree model and privacy protection technology to obtain a global decision tree model. Step S303: Based on the global decision tree model, apply the parameter optimization algorithm to fine-tune the model parameters, obtain the behavior criteria of the detection nodes, and combine the dynamic collaboration framework to generate consistent node behavior criteria. Step S304: Based on the consistent node behavior criteria, when a detection node identifies a complex problem, it evaluates the optimal response strategy based on the data interaction network and multi-agent system, and determines the cooperation path through a search algorithm to generate a solution.

[0043] In step S301 of some embodiments, the real-time data collected by the detection node can be the current state data of the coal bunker continuously acquired during the operation of the detection node (such as environmental parameters, material status, changes in operating conditions, etc.); incremental update can be an update method that performs local iterative optimization of the decision model based on new data, rather than full retraining; the data filtering algorithm can be an algorithm used to filter effective information from real-time data, such as the sliding window algorithm, which can extract representative latest data; the training dataset can be a high-quality data set used to optimize the decision model after filtering.

[0044] By utilizing real-time data collected from detection nodes, the decision-making model is incrementally updated (local iterative optimization). Simultaneously, a data filtering algorithm extracts effective information from the real-time data to obtain a dataset for training the decision-making model. Incremental updates ensure that the decision-making model adapts to the latest operating conditions in a timely manner, improving its sensitivity and adaptability; the data filtering algorithm ensures the timeliness and representativeness of the training dataset, providing a high-quality data foundation for subsequent model optimization and enhancing training effectiveness.

[0045] In step S302 of some embodiments, the reinforcement learning mechanism can be an algorithmic logic of autonomous learning of the optimal strategy through environment interaction-reward feedback, which can drive the model to adapt to dynamic working conditions; the gradient boosting decision tree algorithm (GBDT) can be an ensemble learning algorithm that improves model performance by iteratively training multiple decision trees and accumulating the results of weak classifiers; privacy protection technology can be a technology that ensures that sensitive information is not leaked during data sharing and training (such as federated learning); model collaborative training can be a process in which multiple detection nodes jointly participate in model optimization and share training results.

[0046] Based on the training dataset, a reinforcement learning mechanism and a gradient boosting decision tree algorithm are combined to train the decision model for each detection node. Cross-validation is introduced to prevent overfitting, resulting in a local decision tree model. Specifically, this process first introduces a deep network from the reinforcement learning mechanism, enabling each detection node to continuously optimize its behavior strategy based on its collected data and environmental feedback. Simultaneously, experience replay technology is used to break down correlations between data, ensuring stable model performance under different scenarios. Next, the gradient boosting decision tree algorithm is applied to iteratively train the decision model for each detection node. To prevent overfitting, cross-validation is introduced to ensure the model maintains good generalization performance during training, ultimately yielding the local decision tree model. Privacy protection technology is also introduced to achieve secure data sharing and collaborative model training among detection nodes while ensuring data security, ultimately resulting in a global decision tree model. The combination of reinforcement learning and gradient boosting decision tree significantly improves the prediction accuracy and generalization ability of the decision model. Privacy protection technology, while ensuring data security, promotes multi-node collaborative training, enhances the model's global adaptability, and ensures the system's efficiency and accuracy under complex conditions.

[0047] In some embodiments, the step of co-training the model based on the local decision tree model and privacy protection technology to obtain the global decision tree model may include, but is not limited to, steps S401 to S403: Step S401: Based on the local decision tree model and federated aggregation, a preliminary global model is obtained; Step S402: Based on the preliminary global model and personalized fine-tuning, a personalized local model adapted to local characteristics is obtained. Step S403: Based on distributed hyperparameter optimization and iterative update loop, a global decision tree model is obtained.

[0048] In step S401 of some embodiments, the local decision tree model can be a decision model obtained by training the decision model of the detection node based on the training dataset, combined with reinforcement learning mechanism and gradient boosting decision tree algorithm; federated aggregation can be a process of summarizing and fusing the model parameters or update weights scattered in each detection node without directly exchanging the original sensitive data of each detection node, relying on the federated learning algorithm, with the core adopting the federated averaging (FedAvg) algorithm; the preliminary global model can be a benchmark model that reflects the global operation trend of the coal bunker and has basic local adaptation capabilities after fusing the model information of each detection node through federated aggregation and adjusting and adapting local characteristics through personalized federated learning (PFL).

[0049] Based on the local decision tree model of each detection node, a federated learning algorithm is used to share model parameters or updated weights with other nodes. Before sharing, differential privacy technology is used to add noise to generate secure model parameter updates to ensure data privacy. Then, secure model parameter updates are used to coordinate all detection nodes to participate in the global model construction. The federated averaging (FedAvg) algorithm is introduced into the model aggregation process to summarize and merge the model updates trained by each detection node based on local data. At the same time, personalized federated learning (PFL) is used to adjust the local model of each detection node to adapt to different local characteristics, and finally a preliminary global model is obtained.

[0050] This step combines federated learning with differential privacy technology to achieve multi-node collaborative training while avoiding the leakage of raw data, thus ensuring data security and privacy. The federated averaging algorithm improves the efficiency of model aggregation, while personalized federated learning allows the model to take into account both global operating trends and local operating condition differences. The preliminary global model integrates the local monitoring experience of each detection node, laying a robust, secure, and adaptable foundation for subsequent model optimization.

[0051] In step S402 of some embodiments, based on the preliminary global model, personalized federated learning is used to adjust the local model of each detection node. Each detection node, while retaining the advantages of the global model, fine-tunes its local model for its specific working conditions and data patterns to adapt to different local characteristics and generate a personalized local model.

[0052] More specifically, in practical applications, the design of personalized federated learning adjustment aims to address the lack of effective information sharing and collaborative working frameworks in existing technologies. This application's embodiments introduce a complex personalized adjustment mechanism that not only considers the historical characteristics and performance of detection nodes but also incorporates regularization terms to prevent overfitting, ensuring that the model can adapt to global trends while also performing well locally. This enables the system to maintain high efficiency and accuracy under complex and changing conditions. Specifically, this application's embodiments will incorporate the personalized adjustment portion... It is represented as a function that depends on the features and historical performance of the detection nodes, and incorporates a regularization term to prevent overfitting.

[0053] set up It is a preliminary global model obtained by aggregation through the federated averaging algorithm. Indicates the first The historical feature vectors of each detection node (including sensor data, environmental parameters, etc.) This represents the historical performance matrix of the detection node (recording past performance and adjustment effects). This application's embodiment defines a personalized adjustment section. as follows: ; in: and These are weighting coefficients used to balance the influence of detection node features and historical performance; It is a feature weight matrix used to weight the features of the detection nodes; It is an error matrix, representing the difference between the current model prediction and the actual data; It is a regularization parameter used to control the strength of the regularization term; It is a regularization function, such as L2 regularization or elastic network regularization, used to prevent overfitting; Indicates the model The gradient.

[0054] Ultimately, personalized local models It can be represented as: ; in, It is a preliminary global model obtained by aggregation through the federated averaging algorithm.

[0055] This application's embodiments, by incorporating the historical characteristics and performance of detection nodes, enable the personalized adjustment part to better capture the unique operating conditions of each detection node, enhancing the system's adaptability and improving the model's generalization performance; combined with the error matrix... The personalized adjustment part can dynamically adjust the model based on the current prediction error, thereby improving the prediction accuracy of the model and reducing the false positive rate and false negative rate. By adding a regularization term, the personalized adjustment part effectively prevents the model from overfitting and ensures the stability and reliability of the model on new data. The personalized adjustment mechanism can allocate computing resources more rationally, avoid unnecessary repeated training, and improve the system's operating efficiency and response speed.

[0056] In one specific embodiment, to ensure that detection nodes whose features are closer to the global trend receive greater weight in personalized adjustments, thereby better adapting to changes in overall operating conditions, and to ensure that detection nodes that have performed well in the past receive greater weight in personalized adjustments, thereby improving the overall performance and reliability of the system, this application embodiment adjusts the weight coefficients in the personalized federated learning adjustment step. and The details have been further elaborated.

[0057] set up It is a preliminary global model obtained by aggregation through the federated averaging algorithm. Indicates the first Historical feature vectors of each detection node This represents the historical performance matrix of the detection node. The weighting coefficients are defined in this embodiment. and as follows: ; ; in: and These are hyperparameters used to control the strength of similarity and performance impact; Indicates the detection node Historical feature vector The average feature of all detection nodes The similarity between them can be calculated using methods such as cosine similarity or Euclidean distance.

[0058] Indicates the detection node Historical performance scores can be measured by the model's accuracy on local data, loss function value, or other evaluation metrics.

[0059] Finally, the personalized adjustment section It can be represented as: ; This application embodiment introduces... It can achieve detection based on nodes Historical feature vector The average feature of all detection nodes Weights are assigned based on the similarity between nodes. This method ensures that detection nodes with features closer to the global trend receive greater weight in personalized adjustments, thus better adapting to changes in overall operating conditions. By introducing... It can achieve detection based on nodes The system uses historical performance scores to assign weights. This method ensures that detection nodes that have performed well in the past are given greater weight in personalized adjustments, thereby improving the overall performance and reliability of the system.

[0060] Furthermore, this embodiment of the application can normalize the weights using a softmax function, ensuring the sum of the weights is 1, and can smoothly adjust the weight differences between different detection nodes, avoiding the problem of some detection nodes having excessively large weights. Through weight allocation based on similarity and historical performance, the personalized adjustment part can better capture the unique working conditions of each detection node, enhancing the system's adaptability and improving the model's generalization performance; combined with the error matrix... The personalized adjustment part can dynamically adjust the model based on the current prediction error, thereby improving the prediction accuracy of the model and reducing the false positive rate and false negative rate. By adding a regularization term, the personalized adjustment part effectively prevents the model from overfitting and ensures the stability and reliability of the model on new data. Based on the personalized adjustment mechanism, computing resources can be allocated more rationally, avoiding unnecessary repeated training and improving the system's operating efficiency and response speed.

[0061] In step S403 of some embodiments, a distributed hyperparameter optimization algorithm is introduced to adjust the key hyperparameters in the federated learning process based on the personalized local model, thereby obtaining a hyperparameter-optimized global model. This step specifically includes: setting the initial hyperparameter configuration; using methods such as Bayesian optimization, random search, or evolutionary algorithms to perform hyperparameter search in parallel on multiple detection nodes; evaluating the importance of each feature based on Shapley value decomposition technology, and selecting the optimal hyperparameter configuration, thereby obtaining a hyperparameter-optimized global model.

[0062] The hyperparameter-optimized global model is redistributed to each detection node for the next round of local training. Differential privacy technology is applied to protect the shared data, generating a secure update loop mechanism. Specifically, differential privacy technology adds noise before the shared model parameters to prevent the leakage of sensitive information. This secure update loop mechanism enables the system to continuously optimize while protecting data privacy, enhancing the system's credibility and reliability.

[0063] Based on the aforementioned secure update loop mechanism, as the number of iterations increases, the construction and optimization process of the global model is continuously repeated, enabling the detection node to continuously improve its decision tree model and obtain the global decision tree model.

[0064] Based on the aforementioned secure update loop mechanism, the construction and optimization process of the global model is repeatedly repeated as the number of iterations increases. In each iteration, the detection node will undergo a new round of training and adjustment based on the latest global model and local data, gradually improving its decision tree model. Through continuous iterative optimization, a fully optimized global model is finally obtained.

[0065] The iterative optimization process enables the detection nodes to continuously improve their decision tree models, gradually enhancing the overall performance of the system. As the number of iterations increases, the model gradually approaches the optimal solution, ensuring the long-term stability and efficiency of the system. This continuous improvement mechanism provides a strong guarantee for the system's adaptive capabilities, enabling it to maintain its best state under constantly changing operating conditions.

[0066] In summary, the embodiments of this application utilize a series of advanced technologies and methods, including personalized federated learning, distributed hyperparameter optimization, differential privacy technology, and iterative optimization mechanisms, to achieve efficient construction and continuous optimization of the global model. This method not only improves the model's prediction accuracy and adaptability but also enhances the system's security and collaborative capabilities, providing a strong guarantee for the safe and stable operation of the coal bunker.

[0067] In step S303 of some embodiments, the parameter optimization algorithm can be an algorithm that adjusts model parameters through precise search to approximate optimal performance (such as the Bayesian optimization algorithm); the model parameters can be the core configuration data that determines the operational logic and output results of the decision model; the behavior guidelines of the detection nodes can be the operational specifications that guide the data collection, anomaly judgment, and collaborative interaction of the detection nodes; the coordinated and consistent node behavior guidelines can be the operational specifications that all detection nodes uniformly follow, which take into account both local monitoring needs and global collaborative goals.

[0068] Based on a global decision tree model, a parameter optimization algorithm is applied to fine-tune the model parameters and determine the independent behavior criteria of the detection nodes. Then, combined with the previously constructed dynamic collaboration framework, inter-node collaboration logic is incorporated to ultimately generate consistent node behavior criteria. The application of the parameter optimization algorithm makes the model parameter configuration more scientific, further improving the accuracy of independent decision-making by the detection nodes. The unified behavior criteria generated by the dynamic collaboration framework ensure efficient collaboration among the detection nodes, balancing local monitoring characteristics with the global optimal goal, significantly enhancing the system's adaptability and operational reliability.

[0069] In step S304 of some embodiments, the complex problem can be a coal bunker anomaly that a single detection node cannot handle independently (such as stress concentration in multiple areas, complex material accumulation, etc.); the multi-agent system can be a management system that coordinates multiple detection nodes (agents) to allocate tasks and work together; the optimal response strategy can be an efficient collaborative solution to deal with complex problems; the search algorithm refers to an algorithm used to find the optimal collaborative path between nodes; the cooperative path can be an efficient interactive process for collaborative problem handling between detection nodes; and the solution can be a specific processing scheme for the aforementioned complex problem.

[0070] Based on consistent node behavior guidelines, when a detection node identifies a complex problem that it cannot handle independently, it leverages real-time information from the data interaction network to evaluate the optimal response strategy. A multi-agent system is introduced to manage task allocation and collaborative work among different detection nodes. A search algorithm determines efficient cooperation paths between nodes, ultimately generating a solution. The combination of the multi-agent system and the search algorithm enhances the system's collaborative problem-solving capabilities, ensuring optimal response strategies and efficient cooperation paths, significantly improving the system's flexibility and intelligence in handling complex operating conditions.

[0071] This application introduces a reinforcement learning mechanism and a gradient boosting decision tree (GBDT) algorithm to jointly train the decision tree model of the detection nodes. It also employs a federated learning algorithm to achieve secure data sharing and collaborative training, addressing the issues of insufficient model optimization capabilities and weak data privacy protection in related technologies. Specifically, through incremental updates and reinforcement learning, the detection nodes can continuously optimize their behavior strategies based on newly collected data and environmental feedback, significantly enhancing the system's adaptability. The GBDT algorithm and cross-validation technology effectively prevent overfitting while improving model prediction accuracy, ensuring the reliability and stability of the model in practical applications. Regarding privacy protection, a federated learning algorithm is used to achieve collaborative training where data is available but not visible, and differential privacy technology is combined to further strengthen data privacy protection, generating secure model parameter update results. Through federated averaging and personalized federated learning, the effective construction of the global model and personalized adjustment of the local models of each detection node are achieved, balancing global trends and local characteristics, improving system flexibility and adaptability. Furthermore, a distributed hyperparameter optimization algorithm and Shapley value decomposition technology are introduced to automatically adjust key hyperparameters and evaluate feature importance, ensuring optimal model configuration and further improving system performance and operating efficiency. In summary, this application effectively addresses the shortcomings in model optimization and data privacy protection in existing coal bunker monitoring systems by employing advanced machine learning and privacy protection technologies, providing strong technical support for the safe and stable operation of coal bunkers.

[0072] In step S103 of some embodiments, different operating conditions can refer to the operating state of the coal bunker under different production loads, material characteristics, and environmental conditions (such as feed rate, humidity differences, etc.); material flow characteristics can refer to the flow velocity, trajectory, uniformity, and other characteristics of coal in the coal bunker; pressure distribution changes can refer to the distribution and dynamic changes of the pressure exerted by coal on the bunker wall at different locations and times; simulation analysis can refer to the analysis method of the material state in the coal bunker through numerical simulation technology (such as fluid mechanics and structural mechanics simulation); material accumulation pattern can refer to the accumulation shape, location, density, and other characteristics of coal in the bunker; stress concentration points can refer to the locations on the bunker wall and structure where stress is abnormally concentrated and prone to fatigue damage; optimization algorithm can refer to an algorithm that finds the optimal parameter combination through iterative calculation (such as Bayesian optimization, etc.); material management strategy parameters can refer to control parameters related to operations such as feeding, discharging, and ventilation (such as feeding speed, discharging frequency, etc.); optimization result can refer to the optimal material management strategy parameter combination obtained through algorithm iteration; coal bunker control system can refer to the hardware and logic system that regulates the operation of the physical coal bunker; and adaptive adjustment scheme can refer to a control scheme that can autonomously adjust the management strategy according to the real-time state of the coal bunker.

[0073] Based on a dynamic environment perception model, the flow characteristics and pressure distribution changes of materials in the coal bunker under different operating conditions are simulated. The material accumulation pattern and stress concentration points are evaluated through simulation analysis. Optimization algorithms are used to iteratively adjust material management strategy parameters to obtain optimization results. These optimization results are then fed back to the coal bunker control system to generate an adaptive adjustment scheme. This step enables accurate prediction of the material state and structural risks in the coal bunker under multiple operating conditions. Through algorithm optimization and closed-loop control, an adjustment scheme adapted to dynamic operating conditions is generated, improving the scientific nature of material management, ensuring the safety and efficiency of coal bunker operation, and reducing operating costs.

[0074] In some embodiments, step S103 may include, but is not limited to, steps S501 to S504: Step S501: Based on the dynamic environment perception model and combined with multi-physics field coupling analysis, high-precision simulation is performed on the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions to obtain simulation results. Step S502: Based on the simulation results, the material accumulation mode and stress concentration points are evaluated by nonlinear material behavior modeling, and the material accumulation and stress distribution map is generated by topology optimization method. Step S503: Based on the material accumulation and stress distribution map, the material management strategy parameters are iteratively adjusted using the Bayesian optimization algorithm, and the optimal operation strategy is learned autonomously using the reinforcement learning framework. The performance index under different parameter settings is evaluated using the random forest regression prediction model to obtain the optimization results. Step S504: Adjust the optimization results through real-time feedback information from the industrial IoT platform, generate control parameters adapted to the current working conditions, use digital twin technology to realize the synchronous operation of the physical coal bunker and the virtual coal bunker, promptly discover and respond to potential problems based on the anomaly detection algorithm, generate an adaptive adjustment scheme, and update the configuration of the coal bunker control system based on the adaptive adjustment scheme to guide actual operation and adapt to changes in working conditions.

[0075] In step S501 of some embodiments, multiphysics coupling analysis can be a comprehensive analysis method that integrates the laws of multiple physical fields such as fluid mechanics and structural mechanics to analyze the interaction between material flow and pressure on the coal bunker wall; high-precision simulation can be a high-fidelity virtual reproduction of physical phenomena in the coal bunker by using advanced numerical simulation technologies such as computational fluid dynamics (CFD) and finite element analysis (FEA); simulation results can be quantitative data and visualization results obtained by high-precision simulation that reflect the changes in material flow characteristics and pressure distribution under different working conditions.

[0076] Based on the coal bunker structure and operating condition information provided by the dynamic environment perception model, and combined with multiphysics coupling analysis methods, high-precision simulations of material flow characteristics and pressure distribution changes within the coal bunker under different operating conditions are performed using technologies such as computational fluid dynamics (CFD) and finite element analysis (FEA). The simulation results ultimately reflect the complex operating conditions inside the coal bunker. This high-precision simulation achieves high-fidelity reproduction of complex physical phenomena within the coal bunker, accurately capturing the dynamic laws of material flow and pressure distribution under different operating conditions. This provides a reliable data foundation for subsequent material accumulation risk assessment, stress concentration point identification, and material management strategy optimization, significantly improving the system's prediction accuracy and the scientific basis of decision-making regarding the coal bunker's operating status.

[0077] In step S502 of some embodiments, nonlinear material behavior modeling can be a modeling method that considers the nonlinear mechanical response (such as plastic deformation and viscous flow) of materials such as coal during stress and flow; topology optimization method can be a design method that optimizes the distribution of structural materials to achieve optimal structural performance under constraints; material accumulation and stress distribution diagram is a visual chart that intuitively presents the material accumulation shape, location and stress concentration area and size of the silo wall / material in the coal bunker.

[0078] Based on simulation results, nonlinear material behavior modeling was used to analyze and evaluate the material accumulation patterns and stress concentration points within the coal bunker. Then, topology optimization was employed to process the evaluation data, ultimately generating detailed material accumulation and stress distribution maps. Nonlinear material behavior modeling improved the accuracy of material accumulation and stress concentration assessments, while topology optimization made the distribution maps more valuable for engineering guidance. The combination of these two methods provides a scientific basis for optimizing the internal structure of the coal bunker and formulating material management safety strategies, thereby improving material storage safety, optimizing the coal bunker structural design, and enhancing the overall system performance.

[0079] In step S503 of some embodiments, the random forest regression prediction model can be a regression analysis model that integrates multiple decision trees to predict system performance indicators under different parameter settings; the performance indicators can be core indicators for measuring the effectiveness of material management (such as stacking risk value, stress safety threshold, operating efficiency, etc.); the optimization result can be the optimal combination of material management strategy parameters obtained after algorithm adjustment, strategy learning and performance evaluation.

[0080] Based on material accumulation and stress distribution maps, a Bayesian optimization algorithm is used to iteratively adjust the material management strategy parameters. Combined with a reinforcement learning framework, the system autonomously learns the optimal operating strategy from changes in operating conditions. Simultaneously, a random forest regression prediction model is employed to evaluate performance indicators under different parameter settings, ultimately yielding the optimized material management strategy parameters. The combination of Bayesian optimization and reinforcement learning enables efficient and intelligent adjustment of material management parameters, allowing the strategy to dynamically adapt to changes in operating conditions. Random forest regression prediction provides accurate performance evaluation data for parameter optimization, forming a closed-loop control of analysis-adjustment-learning-evaluation. This significantly improves the safety and stability of the operating strategy, ensuring the system maintains high efficiency under complex operating conditions.

[0081] In step S504 of some embodiments, the real-time feedback information of the industrial IoT platform can be real-time data such as the physical coal bunker operating status and environmental parameters collected by the industrial IoT platform; the control parameters can be operating parameters (such as feeding speed, discharge frequency, etc.) adapted to the current working conditions and used to regulate the operation of the coal bunker; the configuration update can be a dynamic adjustment operation on the operating logic and parameter settings of the coal bunker control system.

[0082] The optimization results are dynamically adjusted based on real-time feedback from the Industrial Internet of Things (IIoT) platform. Combined with edge computing capabilities, the material management strategy parameters, after Bayesian optimization, are adjusted to generate control parameters adapted to the current operating conditions. A virtual coal bunker is constructed using digital twin technology, enabling synchronized operation between the physical and virtual bunkers. Anomaly detection algorithms promptly identify and respond to potential problems, generating adaptive adjustment schemes (such as control commands). Finally, the coal bunker control system is configured and updated based on these adaptive adjustment schemes to guide actual operation and adapt to changing operating conditions, ensuring the system maintains optimal performance under different conditions. The IIoT platform ensures the real-time and accurate adjustment of control parameters; digital twin technology enables visualization and predictability of the coal bunker's operating status, enhancing system controllability; anomaly detection algorithms improve the identification and response speed of potential risks, ensuring operational safety; and the adaptive adjustment scheme and control system configuration updates form a closed loop, achieving intelligent dynamic adaptation of coal bunker operations and ensuring continuous, efficient, and safe operation of the system under complex conditions.

[0083] In some embodiments, the method further includes: establishing a secure data synchronization channel between the digital twin and the physical coal bunker using blockchain technology, and protecting sensitive data during transmission using homomorphic encryption technology; when a detection node detects a potential risk, using a classifier to perform preliminary screening of alarms to filter false alarms, and initiating a multi-objective optimization algorithm to generate a solution set.

[0084] This application utilizes blockchain technology to establish a two-way data synchronization channel between a digital twin and the physical coal bunker, combined with homomorphic encryption technology to ensure the security of transmitted sensitive data. A support vector machine (SVM) classifier is used to screen risk alarms and filter false alarms. For confirmed risks, a multi-objective optimization solution based on a genetic algorithm is initiated to generate a solution set. When intelligent monitoring nodes detect potential risks (such as abnormal temperature, pressure changes, or material flow characteristics), these alarm messages are first preliminarily screened by a SVM classifier. SVM is a supervised learning model that can distinguish between alarms and false alarms under normal operating conditions by analyzing historical data, thereby effectively reducing the false alarm rate. This approach strengthens data security and privacy protection, improves the accuracy and efficiency of risk assessment and response, and provides strong support for the intelligentization and automation of coal bunker monitoring systems.

[0085] Please see Figure 2 This application also provides a coal bunker monitoring system that can implement the above-described coal bunker monitoring method. The system includes: Modeling module 201 is used to perform high-fidelity modeling of the coal bunker using multi-source heterogeneous data, construct a digital twin that matches the coal bunker, and form a dynamic environment perception model; The detection module 202 is used to deploy multiple detection nodes in the coal bunker based on a dynamic environment perception model, optimize the decision model of each detection node using machine learning algorithms, and establish a collaborative framework between multiple detection nodes through a graph neural network to obtain a monitoring network. The adjustment module 203 is used to simulate the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions based on the dynamic environment perception model. It uses simulation analysis to evaluate the material accumulation mode and stress concentration points, uses optimization algorithms to iteratively adjust the material management strategy parameters to obtain optimization results, and feeds the optimization results back to the coal bunker control system to generate an adaptive adjustment scheme.

[0086] The specific implementation method of this coal bunker monitoring system is basically the same as the specific implementation method of the coal bunker monitoring method described above, and will not be repeated here.

[0087] Thirdly, embodiments of this application provide an electronic device, see [link to relevant documentation]. Figure 3 The diagram shown is a structural schematic of an electronic device provided in this application.

[0088] like Figure 3 As shown, the device includes: Memory 31 is used to store computer programs; Processor 32 is used to execute computer programs; When the processor 32 executes the computer program, it implements the coal bunker monitoring method as described in any of the above embodiments.

[0089] For example, a computer program may be divided into one or more modules / units, one or more of which are stored in memory 31 and executed by processor 32 to complete this application. One or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in an electronic device.

[0090] The processor 32 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0091] The memory 31 can be used to store computer programs and / or modules. The processor 32 implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory 31 and calling the data stored in the memory 31. The memory 31 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 31 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0092] It should be noted that the aforementioned electronic devices include, but are not limited to, processors and memory, as will be understood by those skilled in the art. Figure 3 The structural diagram is merely an example of the electronic device described above and does not constitute a limitation on the electronic device. It may include more components than shown in the diagram, or combine certain components, or use different components.

[0093] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed, implements the coal bunker monitoring method of any of the above embodiments.

[0094] It should be understood that the implementation of all or part of the processes in the above-described coal bunker monitoring method can also be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-described coal bunker monitoring method. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content contained in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the relevant jurisdiction. For example, in some relevant jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0095] Fifthly, embodiments of this application also provide a computer program product, which is stored in a storage medium and executed by at least one processor to implement the coal bunker monitoring method of any of the above embodiments.

[0096] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0097] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.

Claims

1. A coal bunker monitoring method, characterized in that, include: High-fidelity modeling of the coal bunker is performed using multi-source heterogeneous data to construct a digital twin that matches the coal bunker, forming a dynamic environment perception model. Based on the dynamic environment perception model, multiple detection nodes are deployed in the coal bunker. The decision model of each detection node is optimized using machine learning algorithms, and a collaborative framework among the multiple detection nodes is established through a graph neural network to obtain a monitoring network. Based on the dynamic environment perception model, the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions are simulated. The material accumulation pattern and stress concentration points are evaluated by simulation analysis. The material management strategy parameters are iteratively adjusted using optimization algorithms to obtain optimization results. The optimization results are then fed back to the coal bunker control system to generate an adaptive adjustment scheme.

2. The coal bunker monitoring method as described in claim 1, characterized in that, Based on the dynamic environment perception model, multiple detection nodes are deployed within the coal bunker. Machine learning algorithms are used to optimize the decision models of each detection node, and a collaborative framework among the multiple detection nodes is established using a graph neural network to obtain a monitoring network, including: Based on the dynamic environment perception model, key monitoring locations within the coal bunker are determined, detection nodes with autonomous learning capabilities are deployed, and the parameters of the detection nodes are set through initialization processing. Define a standardized communication protocol between detection nodes and establish an information sharing mechanism; Based on the aforementioned information sharing mechanism, a data interaction network between detection nodes is constructed; A graph neural network framework is constructed based on the data interaction network. The graph neural network framework is used to train and adjust the cooperative relationship between detection nodes, forming a dynamic cooperative framework and a monitoring network.

3. The coal bunker monitoring method as described in claim 2, characterized in that, After constructing a graph neural network framework based on the data interaction network, and using the graph neural network framework to train and adjust the cooperative relationships between detection nodes to form a dynamic cooperative framework and a monitoring network, the method further includes: By combining the data interaction network, the dynamic collaboration framework, and the real-time data collected by the detection nodes, the decision-making model of the detection nodes is continuously optimized through machine learning mechanisms to adapt to changing working conditions and data patterns.

4. The coal bunker monitoring method as described in claim 3, characterized in that, The process of combining the data interaction network, the dynamic collaboration framework, and the real-time data collected by the detection nodes, and continuously optimizing the decision model of the detection nodes through machine learning mechanisms to adapt to changing operating conditions and data patterns, includes: The decision model of the detection node is incrementally updated using the real-time data collected by the detection node, and a training dataset is obtained through a data filtering algorithm. Based on the training dataset, the decision model of the detection node is trained by combining reinforcement learning mechanism and gradient boosting decision tree algorithm to obtain a local decision tree model. Then, based on the local decision tree model and privacy protection technology, the model is co-trained to obtain a global decision tree model. Based on the global decision tree model, the model parameters are fine-tuned using a parameter optimization algorithm to obtain the behavior criteria of the detection nodes, and then the dynamic collaboration framework is combined to generate consistent node behavior criteria. Based on the aforementioned consistent node behavior guidelines, when a detection node identifies a complex problem, it evaluates the optimal response strategy based on the data interaction network and multi-agent system, and determines the cooperation path and generates a solution through a search algorithm.

5. The coal bunker monitoring method as described in claim 4, characterized in that, The process of co-training the model based on the local decision tree model and privacy protection technology to obtain a global decision tree model includes: Based on the local decision tree model and federated aggregation, a preliminary global model is obtained; Based on the initial global model and personalized fine-tuning, a personalized local model adapted to local characteristics is obtained. A global decision tree model is obtained based on distributed hyperparameter optimization and iterative update loop.

6. The coal bunker monitoring method as described in claim 1, characterized in that, Based on the dynamic environment perception model, the flow characteristics and pressure distribution changes of materials in the coal bunker under different working conditions are simulated. Simulation analysis is used to evaluate material accumulation patterns and stress concentration points. An optimization algorithm is used to iteratively adjust the material management strategy parameters to obtain optimization results. These optimization results are then fed back to the coal bunker control system to generate an adaptive adjustment scheme, including: Based on the dynamic environment perception model and combined with multi-physics field coupling analysis, high-precision simulations were performed on the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions, and simulation results were obtained. Based on the simulation results, the material accumulation mode and stress concentration points are evaluated by nonlinear material behavior modeling, and the material accumulation and stress distribution map is generated by topology optimization method. Based on the material accumulation and stress distribution map, the material management strategy parameters are iteratively adjusted using a Bayesian optimization algorithm, and the optimal operation strategy is autonomously learned using a reinforcement learning framework. The performance indicators under different parameter settings are evaluated using a random forest regression prediction model to obtain the optimization results. The optimization results are adjusted based on real-time feedback information from the industrial IoT platform to generate control parameters that adapt to the current operating conditions. Digital twin technology is used to achieve synchronous operation of the physical and virtual coal bunkers. Anomaly detection algorithms are used to promptly identify and respond to potential problems, generate adaptive adjustment schemes, and update the configuration of the coal bunker control system based on the adaptive adjustment schemes to guide actual operation and adapt to changes in operating conditions.

7. The coal bunker monitoring method as described in claim 1, characterized in that, The method further includes: A secure data synchronization channel is established between a digital twin and a physical coal bunker using blockchain technology, and sensitive data during transmission is protected using homomorphic encryption technology. When a detection node detects a potential risk, a classifier is used to perform preliminary screening of the alarm to filter out false alarms, and a multi-objective optimization algorithm is launched to generate a solution set.

8. A coal bunker monitoring system, characterized in that, include: The modeling module is used to perform high-fidelity modeling of the coal bunker using multi-source heterogeneous data, construct a digital twin that matches the coal bunker, and form a dynamic environment perception model. The detection module is used to deploy multiple detection nodes in the coal bunker based on the dynamic environment perception model, optimize the decision model of each detection node using machine learning algorithms, and establish a collaborative framework among the multiple detection nodes through a graph neural network to obtain a monitoring network. The adjustment module is used to simulate the material flow characteristics and pressure distribution changes in the coal bunker under different working conditions based on the dynamic environment perception model. It uses simulation analysis to evaluate the material accumulation pattern and stress concentration points, uses optimization algorithms to iteratively adjust the material management strategy parameters to obtain optimization results, and feeds the optimization results back to the coal bunker control system to generate an adaptive adjustment scheme.

9. An electronic device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the coal bunker monitoring method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the coal bunker monitoring method as described in any one of claims 1 to 7.

Images