Coal whole-process intelligent supervision system

Abstract

The invention, which relates to the technical field of coal full-process supervision, discloses a coal full-process intelligent supervision system comprising a data acquisition module, a data processing module, an early warning module, an event recording module and a management evaluation module which are in electrical signal connection. According to the coal full-process intelligent supervision system, potential laws and features of data of coal quantity, quality, transportation and combustion links are deeply mined through a model trained by using a deep learning algorithm; in a coal quantity link, demand and loss conditions can be accurately predicted, and purchase and inventory strategies can be planned in advance; whether coal meets production requirements or not can be accurately judged in the coal quality link, and the risks of equipment damage and environmental pollution are effectively reduced; early warning delay in advance and reasonably evaluating loss in a transportation link, and optimizing transportation arrangement; the combustion efficiency and pollutant emission are predicted in the combustion link, energy conservation and emission reduction measures are guided to be implemented, and the refinement degree and decision-making efficiency of power plant operation management are comprehensively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of coal full-process supervision, and in particular to a coal full-process intelligent supervision system. Background Art

[0002] In the field of full-process coal management, traditional supervision methods mainly rely on manual operation and experience judgment, which has many disadvantages; in the coal procurement process, manual statistics and analysis of procurement data are inefficient, and it is difficult to track market trends and supplier conditions in real time, resulting in a lack of timeliness and accuracy in the formulation of procurement plans, and an inability to effectively respond to price fluctuations and supply shortage risks; in the coal transportation process, there is a lack of effective automated monitoring methods; for transportation losses, post-accounting can only be carried out after the coal arrives at the factory, and it is difficult to detect problems in time during transportation; if the coal is spilled, damp or loses calorific value due to factors such as failure of transportation vehicles or ships, bad weather or improper loading and unloading operations, it is impossible to take timely measures to stop the loss, which not only affects the quality of the coal, but may also cause economic disputes; in terms of coal quality testing , the manual sampling, sample preparation and testing processes are easily interfered by subjective factors, and it is difficult to achieve standardization and normalization of the entire process; differences in technical levels and operating habits of different operators may lead to insufficient sampling representativeness, sample preparation deviations or testing errors, which reduces the reliability of coal quality data; once low-quality coal enters the combustion link, it will affect combustion efficiency, increase pollutant emissions, and even damage combustion equipment, shorten equipment life, and increase equipment maintenance and replacement costs; with the continuous expansion of the scale of the coal industry and the increasing requirements for production management, intelligent, precise and efficient supervision of the entire coal process has become a key issue that needs to be solved urgently, which has prompted the research and development of an intelligent supervision system for the entire coal process to fill the gaps in traditional supervision methods and achieve optimal management and utilization of coal resources. Summary of the invention

[0003] In order to solve the above technical problems, the present invention is implemented through the following technical solutions: a coal full-process intelligent supervision system, including a data acquisition module, a data processing module, an early warning module, an event recording module and a management evaluation module, wherein the modules are connected by electrical signals;

[0004] The data acquisition module is used to obtain data from the fuel full-process management system, SAP settlement system, fuel platform system, and SIS system, and to organize and verify the collected data, remove obviously erroneous or incomplete data, and ensure the reliability of subsequent processing;

[0005] The data processing module stores the sorted and verified data into the database and uses the deep learning algorithm to train the model; including classifying and storing the data when storing it in the database, and establishing corresponding data tables according to the different links and categories of coal quantity, quality, transportation, and combustion; for the deep learning algorithm, a multi-layer neural network structure is used to train according to different working conditions and data characteristics; the potential laws and characteristics in the coal full process data are extracted to complete the analysis and prediction of coal quantity, quality, and combustion effect, thereby obtaining the analysis results;

[0006] The early warning module issues early warnings based on the analysis results of the data processing module and the set thresholds;

[0007] The event record module uses distributed database technology to store event records, dispersing data on multiple nodes to improve data storage capacity and reliability; for each event record, including data changes, warning threshold changes, warning occurrence, warning processing, operator, operation time, IP address of the operation terminal, original value, modified value, and operation type are recorded;

[0008] During the storage process, the AES encryption algorithm is used to encrypt the data to ensure the security and confidentiality of the data and prevent the data from being illegally obtained and tampered with. It also provides a query function, and managers can query event records in a variety of ways, including querying by time range, querying by operator, and querying by event type. For each event record, it includes: recording the modification operations of the manager on the data. When modifying the coal purchase plan quantity, record the modifier, modification time, and values ​​before and after the modification; for the adjustment of the warning threshold, the relevant information of the adjustment is also recorded; and stored in the event log database for subsequent query and tracing, providing a basis for analyzing the system operation and solving problems, ensuring the traceability and stability of the system operation; the management evaluation module summarizes and analyzes statistical data by week, month, quarter or year, calculates the compliance rate, non-compliance rate, output, and consumption of each data, and generates an analysis report based on the analysis results; and displays it in the form of charts. Through these summaries and analyses, it provides comprehensive management decision support for managers, helps managers evaluate the effectiveness of fuel management, discover potential problems and formulate improvement measures;

[0009] It also includes a video AI intelligent analysis subsystem, which is used to connect to the video monitoring equipment of the coal sampling, sample preparation and testing process to obtain real-time video streams; use high-definition cameras and stable video transmission technology to ensure the clarity and smoothness of video images, and can clearly capture each link and operation details in the coal monitoring process; use image recognition technology and deep learning algorithms to analyze video images, identify personnel operation behaviors, equipment operating status and sample processing during the detection process; determine whether the sample preparation equipment is operating normally; and whether the sample processing during the test is standardized;

[0010] When the video AI intelligent analysis subsystem detects a violation, it immediately issues an alarm signal and highlights the alarm information on the system interface, while recording the video clip, time, location and type of the violation; the alarm signal reminds quality supervisors and laboratory managers through sound and pop-up windows; relevant personnel can investigate and handle violations in a timely manner based on the information recorded by the system, ensuring the standardization and accuracy of the coal quality testing process and the reliability of coal quality data, thereby improving the effectiveness of the entire fuel intelligent supervision system.

[0011] Preferably, in the data acquisition module, the process of sorting and verifying the collected data includes: first, clarifying the data types including coal purchase plan data, coal transportation process data, coal entry and combustion related data, and fuel data and raw material data; wherein, the coal purchase plan data includes the monthly purchase plan coal quantity and quality information; the coal transportation process data includes the coal contract for each ship, loading port, unloading port, and transit quantity loss data; the coal entry and combustion related data includes the entry calorific value, the furnace calorific value, the coal storage in the coal yard, the coal loading amount, and the coal consumption; the fuel data and raw material data include limestone consumption, ash production, ash slag, and fuel oil data;

[0012] Then, different verification methods are used for different types of data. For numerical data, numerical range verification and data consistency verification are used, and a calibrated physical range is preset. If the numerical data exceeds the calibrated physical range, it is judged as abnormal data; at the same time, for the numerical data of the same batch of coal in different links, if the interpolation value is greater than the calibrated physical range and there is no reasonable reason, it is also marked as abnormal data; for text data, format verification and keyword matching verification are used, and the text data is verified based on the standard format corresponding to the text data. If it does not meet the standard format, it is marked as abnormal data;

[0013] When abnormal data is found, the data collection module performs data repair, and the process includes: for simple errors, including typos or format errors during data entry, automatic repair algorithms are used to correct them;

[0014] For complex errors or missing data, estimation and supplementation are carried out through data interpolation and regression analysis methods; at the same time, the data verification process and repair results are recorded for management personnel to review and audit, ensuring that the collected data has high accuracy and reliability, providing a solid foundation for subsequent data analysis and supervision.

[0015] Preferably, when storing in a database, the process of classifying and storing the data includes: for coal quantity data, creating a coal quantity purchase plan data table when storing in the database, the table contains fields for purchase plan quantity, quantity at the loading port stage, unloading port stage, factory entry stage, coal loading quantity, coal consumption, and coal yard inventory; each record corresponds to a timestamp or business operation batch, so as to track the change of quantity over time; in the data entry process, the quantity information obtained by the data acquisition module is inserted according to the preset format and data type specification;

[0016] Coal quality data is stored in the coal quality data table, which contains key quality indicators of coal such as calorific value, ash content, sulfur content, and volatile matter, and is associated with sampling time, sampling location, and coal batch number information. To ensure the accuracy and traceability of the data, the quality data is verified and standardized before being stored. The data value range is set according to industry standards for verification. Data out of the range will trigger an abnormal prompt, and the detection equipment will be rechecked or calibrated based on the abnormal prompt. At the same time, a composite index strategy based on coal batch number and sampling time is used to query and retrieve the quality data of a specified batch of coal at different time points, providing efficient data access support for quality analysis and comparison.

[0017] The transportation data is stored in the coal transportation data table, which records the contract information, transportation tool details, transit information and transportation time nodes during the coal transportation process; among them, the contract information includes the contract number, signing date, and supplier information; the transportation tool details include the ship number, train number, transportation starting point and destination; the transit information includes the number of transits, transit location, transit loss and loss ratio; the transportation time nodes include loading time, expected arrival time, and actual arrival time;

[0018] The data of the combustion process is stored in the coal combustion data table, which contains the quality data of the coal entering the furnace, combustion time, unit operating parameters, the amount of ash produced by combustion, and the carbon content of fly ash; among them, the quality data is obtained from the coal quality data table, and the unit operating parameters include the load, temperature, and pressure of the unit operation; through the above classified storage of different links and categories of data on coal quantity, quality, transportation, and combustion, a database with a clear structure, easy management and efficient query is constructed, which provides a solid foundation for the stable operation and data analysis of the coal full-process intelligent supervision system.

[0019] Preferably, the process of training the model using a deep learning algorithm includes:

[0020] In the full-process intelligent supervision system of coal, the analysis results of the model trained by deep learning algorithm in each link are as follows: In the link of coal quantity, a multi-layer neural network including input layer, multiple hidden layers and output layer is adopted. The input layer receives the historical coal purchase quantity, coal quantity records of loading port stage, unloading port stage and factory stage, timestamp, season, and market supply and demand conditions; the hidden layer uses the long short-term memory network LSTM unit to capture the dynamic change law of coal quantity in time series, and combines the fully connected layer for feature fusion and nonlinear transformation. It is trained through a large amount of historical data, uses mean square error MSE as the loss function, and uses stochastic gradient descent SGD or Adam to adjust the network weights to obtain the coal quantity analysis model;

[0021] The coal quantity analysis model is used to predict the demand for coal quantity in the future. By comparing the current inventory and the amount of coal in transit, it is determined whether the amount of coal is sufficient and generates the analysis results of the coal quantity. At the same time, for the loss of coal during transportation and storage, by comparing the changes in the amount of coal at different stages, combined with the transportation distance and storage time, it is analyzed whether the loss is within a reasonable range. If there is abnormal loss, it indicates that there may be transportation leakage or storage management problems.

[0022] In the coal quality link, a neural network with a multi-layer perceptron (MLP) structure is constructed. The input layer includes the quality indicators of coal calorific value, ash content, sulfur content, and volatile matter, coal source information, mining process data, and environmental parameters during sampling and testing. The hidden layer uses ReLU for nonlinear processing to enhance the model's ability to learn complex quality relationships. Based on a large amount of coal sample data with labeled quality levels: high-quality, qualified, and low-quality, the cross-entropy loss function and optimization algorithm are used to train the model to classify and evaluate coal quality, and a coal quality analysis model is obtained through training.

[0023] The coal quality analysis model is used to determine whether the coal quality meets the production requirements of the power plant. The coal quality analysis model determines that the coal quality does not meet the standards, and indicates the possible risks of corrosion to the equipment and environmental pollution, and generates the analysis results of the coal quality.

[0024] Preferably, the process of training the model using a deep learning algorithm also includes:

[0025] In the coal transportation link, a hybrid model combining convolutional neural network (CNN) and recurrent neural network (RNN) is designed; the CNN layer is used to extract the spatial features in the transportation contract text, the image appearance of the transportation vehicle or ship, the load identification features, and the transportation route geographic information, while the RNN layer is used to process the transportation time series data; through a large number of historical transportation records for training, using transportation time deviation and transportation loss rate as loss indicators, the training model optimizes the prediction and evaluation capabilities of the transportation process, and obtains a coal transportation analysis model;

[0026] The coal transportation analysis model is used to predict the arrival time of coal transportation, give early warning of possible transportation delays, and generate analysis results of coal transportation so that power plants can prepare for receiving the goods. For transportation losses, the relationship between the stability of the transportation tools, road or sea conditions, and the degree of standardization of loading and unloading operations during the transportation process and the actual losses can be analyzed to accurately determine whether the losses are reasonable. If the coal transportation analysis model predicts that a certain transportation route may cause coal spillage or increased moisture under certain weather conditions, and the actual transportation results exceed the reasonable range, the supervision system will generate a prompt message, investigate the transportation process based on the prompt message, and make improvements immediately after obtaining the reasons for the investigation to ensure the efficiency and quality of coal transportation.

[0027] In the coal combustion process, a deep neural network architecture is used. The input layer receives the quality data of coal before entering the furnace, the load of the unit operation, combustion temperature, pressure, air flow parameters and performance parameters of the combustion equipment; the hidden layer learns the physical and chemical changes in the coal combustion process through a combination of multi-layer convolution and fully connected layers; a large amount of combustion experimental data and actual production operation data are used for training, and combustion efficiency and pollutant emissions are used as evaluation indicators. The model parameters are continuously adjusted through the back propagation algorithm to build a coal combustion analysis model;

[0028] The coal combustion analysis model is used to predict the combustion efficiency of coal under the current unit operating conditions. By comparing the actual combustion efficiency with the theoretical optimal efficiency or the historical average efficiency, it is determined whether there are any problems in the combustion process and the analysis results of coal combustion are obtained. At the same time, it is also used to predict the pollutants generated during the combustion process: the emissions of sulfur dioxide, nitrogen oxides, and smoke. When the predicted emissions exceed the environmental emission standards, early warnings are issued and emission reduction measures are recommended. The emission reduction measures include adjusting the coal blending ratio and optimizing the combustion air supply, thereby achieving clean coal combustion and environmentally friendly production.

[0029] Based on the analysis results of the four links of coal quantity, quality, transportation and combustion, the analysis results of the whole coal process are obtained; the analysis results include whether the coal quantity is sufficient, whether the quality is up to standard, whether the loss in each link is reasonable, and the utilization efficiency of various fuels and raw materials. The analysis results are output in a standardized data format to provide a decision-making basis for subsequent early warning modules and management evaluation modules, thereby realizing intelligent supervision and effective management of the whole coal process; among them, coal characteristic data include data on the impact of changes in unit operating load on coal consumption and combustion efficiency.

[0030] Preferably, in terms of coal quantity supervision, the data collection module specifically performs the following operations:

[0031] For the monthly purchase plan and execution, the purchase plan coal quantity and actual in-plant quantity information in the business system are collected regularly and saved synchronously in the local supervision system. During the collection process, a scheduled task scheduling mechanism is adopted to ensure the timeliness and accuracy of the data. The latest purchase plan and in-plant quantity data are automatically obtained from the business system at 2 a.m. every day, and the data format is converted and verified, and stored in a special purchase plan data table in the database of the local supervision system. At the same time, managers are allowed to manage the synchronized data. When data errors are found or additional explanations are required, managers can operate in the management background, such as correcting the wrong in-plant quantity or filling in the reasons for the delay, such as the delay of coal transportation due to bad weather during transportation, so that the actual in-plant quantity is lower than the planned quantity. The supervision system is based on the monitoring rules set by the managers. The monitoring rules include setting a third-level warning when the monthly achievement rate is lower than 70%, and a second-level warning when it is lower than 50%. Warning judgments are made and displayed in the form of tables and charts through the dashboard, so that managers can intuitively view the progress of the purchase plan execution and the warning situation.

[0032] In view of the quantity loss after coal loading, unloading and transshipment, the actual loading and unloading port quantities in the business system are monitored and collected, and the transit loss and loss ratio are automatically calculated; in the calculation process, accurate calculations are performed according to mathematical formulas, and the transit loss amount is equal to the loading port quantity minus the unloading port quantity, and the loss ratio is equal to the transit loss amount divided by the loading port quantity multiplied by 100%; the calculation results are saved synchronously in the local supervision system, and managers are allowed to manage the data. For example, when it is found that the manual entry in the business system is biased or the collected data is inaccurate, resulting in an incorrect loss value, the manager can make corrections; the supervision system is based on the set loss threshold between different destinations. For example, for coal shipped to a specific power plant, the loss rate threshold is set to ±0.3%. When the loss threshold is exceeded, an early warning is issued, and the loading port quantity, unloading port quantity, and loss ratio information are displayed in clear tables and charts to facilitate managers to analyze and deal with quantity loss problems.

[0033] Preferably, in terms of coal quality supervision, the data processing module operates in the following manner:

[0034] For the coal quality supervision of the monthly procurement plan and execution, the procurement plan of the business system and the quality information of coal entering the site are collected regularly and stored in the coal quality data table in the database; during the collection, the collected quality data are standardized and the quality indicators of different units or precisions are uniformly converted into a standard format;

[0035] The data is analyzed using a quality assessment model in a deep learning algorithm, which is trained based on a large amount of historical coal quality data and quality standards to identify whether the coal quality meets the requirements of the procurement plan;

[0036] If the quality assessment model determines that the quality is not up to standard, an early warning will be triggered. At the same time, managers are allowed to manage the synchronized data. If the data is found to be incorrect or requires additional explanation, such as possible deviations in the coal sampling process, corrections or annotations can be made. Early warnings are issued based on the set quality thresholds, and the comparison of the planned coal quality, actual coal quality and quality thresholds is displayed in a chart on the dashboard, so that managers can intuitively understand the implementation of coal quality.

[0037] In the quality supervision of the coal loading, unloading and transshipment process, the actual information of the loading and unloading ports, the settlement calorific value difference and the collection business system are monitored to automatically calculate the transit calorific value loss and the loss ratio. When calculating the transit calorific value loss, the physical and chemical change factors of coal in the loading, unloading and transshipment process are comprehensively considered based on the professional calorific value calculation method and the principle of conservation of energy. The transit calorific value loss and the loss ratio are analyzed using the trained quality change monitoring model to determine whether the change in coal quality during the transshipment process is within a reasonable range. The model determines it as abnormal and issues an early warning. Management personnel can manage and correct the data, and the system issues an early warning based on the set calorific value loss threshold, and displays information such as the loading port quantity, unloading port quantity, settlement calorific value difference, and transit calorific value loss in detailed charts, providing strong support for management personnel to analyze the reasons for changes in coal quality.

[0038] Preferably, the warning in the warning module is divided into four levels: level one warning, level two warning, level three warning and level four warning, with different levels corresponding to different severity and processing requirements. The process includes:

[0039] Level 1 warning is the most serious level, which is triggered when the key coal quality indicators are seriously below the standard and may affect the safe and stable operation of the unit, or when there is a serious shortage of coal supply that may cause the unit to shut down. For level 1 warning, the supervision system immediately notifies key personnel including senior management of the power plant, production department heads, and fuel procurement department heads through a variety of emergency notification methods, such as pop-up display on the large screen of the power plant control center, sending emergency text messages and voice calls to relevant management personnel; at the same time, the emergency plan is activated, which includes emergency adjustment of unit operating parameters to reduce energy consumption, activation of backup fuel supply channels, and organization of professional and technical personnel to quickly detect and analyze coal quality problems to ensure the safe and stable operation of the power plant, and requires relevant personnel to feedback the treatment measures and progress within the specified time of 1 hour, and the system tracks and records the entire treatment process;

[0040] Level 2 warnings indicate more serious problems, such as a certain degree of deviation in coal quality but not reaching the level of level 1 warning, or a serious lag in the implementation of the coal procurement plan, which may affect subsequent production arrangements. After triggering the level 2 warning, the system will notify the corresponding managers, including the head of the production department and the head of the fuel management department, through a pop-up window on the system interface and SMS notifications. The managers must formulate and implement response measures within the specified time of 4 hours, including secondary inspection of the coal, negotiation with suppliers to speed up the delivery progress or adjust the procurement plan, and feedback the processing results to the system. The system will record and evaluate the processing situation. If the problem is not effectively resolved, it may be upgraded to level 1 warning.

[0041] Level 3 warnings are for general problems, such as coal quantity loss slightly higher than the normal threshold but within an acceptable range, or limestone consumption approaching the upper limit of the warning. When a level 3 warning occurs, the system displays the warning information on the system interface and sends email notifications to fuel transportation managers and limestone purchasers. The warning situation is investigated and handled within 8 hours, and the handling results are entered into the supervision system. The supervision system monitors and records the handling process. If the problem persists or worsens, it may be upgraded to a level 2 warning.

[0042] Level 4 warning is the lightest level and is used to prompt some potential minor problems. For level 4 warning, the supervision system records it in the system log on the system interface and displays it with blue markings in the corresponding data report. Relevant personnel can pay attention to and handle these warning information in their daily work.

[0043] Preferably, in the management evaluation module, the process of generating the corresponding analysis report includes: summarizing and analyzing the statistical data in the system according to a predetermined time period: week, month, quarter or year; in the summarizing process, extracting the coal quantity, quality, inventory, consumption, limestone consumption, ash production, ash slag, and fuel data from the database, and performing data cleaning and preprocessing to remove abnormal values ​​and duplicate data; using data analysis algorithms and statistical models to calculate the compliance rate, non-compliance rate, average value, standard deviation, and trend change index of each data;

[0044] An analysis report is generated based on the analysis results. The content of the analysis report includes text description, data table and chart display; the text description includes a summary and explanation of the analysis results; the data table is used to list the values ​​and calculation results of key data; the chart display adopts the form of bar graphs, line graphs and pie charts to intuitively present the changing trends and proportional relationships of the data; the analysis report can be viewed and downloaded through the system interface, providing managers with a comprehensive and accurate basis for decision-making, helping managers to promptly identify problems, optimize management strategies and improve fuel management efficiency.

[0045] Preferably, in the video AI intelligent analysis subsystem, image recognition technology and deep learning algorithms are used to analyze video images to identify personnel operating behaviors, equipment operating status and sample processing conditions during the detection process, and to detect whether the sample processing during the test is standardized. The process includes: first, in the image acquisition stage, a high-resolution camera with a wide dynamic range and low-illumination performance is used to ensure that clear and stable video images can be obtained under different lighting conditions and complex environments such as coal sampling, sample preparation, and testing; the camera is installed along the detection process area to ensure that there are no visual blind spots, and the video image is transmitted to the video AI intelligent analysis subsystem in real time; based on the video image, the target detection and behavior recognition model in the deep learning algorithm is used to identify the personnel operating behavior; first, the behavior recognition model is trained through a large number of labeled personnel operation sample images, which is used for the behavior recognition model to recognize the action posture of the sampling personnel and identify whether the sampling operation is performed in accordance with the standard process. , including the correct use of sampling tools, standardized selection of sampling points, and accurate control of sampling depth; obtain behavior recognition results; in the sample preparation process, identify whether the operator operates the sample preparation equipment correctly, that is, identify the equipment operating status, and obtain operation recognition results; operation recognition results include the control of grinding time and whether the sample screening operation is compliant; in the test link, detect whether the test personnel accurately weigh samples, add reagents, and operate instruments to obtain test recognition results; the behavior recognition model is based on the convolutional neural network CNN architecture, which extracts personnel features and action features in the image, gradually reduces the image resolution and extracts key features through multiple convolutional layers and pooling layers, and performs classification and judgment in the fully connected layer, and compares the identified behavior recognition results, operation recognition results, and test recognition results with the preset standard operating procedures. Once a deviation is found, it is determined to be an abnormal behavior; among them, the standard operating procedures include standard sampling operations, standard sample preparation operations, and standard test operations.

[0046] The present invention provides a coal full-process intelligent supervision system, which has the following beneficial effects:

[0047] This full-process intelligent coal supervision system, by using a model trained with a deep learning algorithm, deeply mines the potential patterns and characteristics of data on coal quantity, quality, transportation and combustion, providing a scientific basis for management decisions; in the coal quantity link, it can accurately predict demand and losses, and plan procurement and inventory strategies in advance; in the coal quality link, it can accurately determine whether the coal meets production requirements, effectively reducing the risks of equipment damage and environmental pollution; in the transportation link, it can provide early warning of delays and reasonably assess losses, and optimize transportation arrangements; in the combustion link, it predicts combustion efficiency and pollutant emissions, guides the implementation of energy-saving and emission reduction measures, and comprehensively improves the refinement and decision-making efficiency of power plant operation management. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the framework of the present invention;

[0049] Figure 2 This is a schematic diagram of the framework of the video AI intelligent analysis subsystem in the present invention;

[0050] Figure 3 This is a schematic diagram of the framework in which the data processing module in the present invention uses a deep learning algorithm to train a model to obtain analysis results for each link. DETAILED DESCRIPTION

[0051] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0052] like Figures 1 to 3 As shown, the present invention provides a technical solution: a coal full-process intelligent supervision system, including a data acquisition module, a data processing module, an early warning module, an event recording module and a management evaluation module, wherein the modules are connected by electrical signals;

[0053] The data acquisition module is used to obtain data from the fuel full-process management system, SAP settlement system, fuel platform system, and SIS system, and to organize and verify the collected data, remove obviously erroneous or incomplete data, and ensure the reliability of subsequent processing;

[0054] The data processing module stores the sorted and verified data into the database and uses the deep learning algorithm to train the model; including classifying and storing the data when storing it in the database, and establishing corresponding data tables according to the different links and categories of coal quantity, quality, transportation, and combustion; for the deep learning algorithm, a multi-layer neural network structure is used to train according to different working conditions and data characteristics; the potential laws and characteristics in the coal full-process data are extracted to complete the analysis and prediction of coal quantity, quality, and combustion effects, so as to obtain the analysis results; for example, for coal quality data, the model is trained through a large amount of historical coal quality sample data, so that it can accurately identify the changing trends and abnormal conditions of coal quality; during the training process, the weight and bias parameters of the model are continuously adjusted to improve the accuracy and generalization ability of the model, ensure that the model can adapt to data from different sources and characteristics, and conduct scientific and rigorous analysis of various types of data to meet the supervision needs under different working conditions;

[0055] The early warning module issues early warnings based on the analysis results of the data processing module and the set thresholds;

[0056] The event recording module uses distributed database technology to store event records, and stores data in multiple nodes in a dispersed manner to improve the storage capacity and reliability of data; for each event record, including data changes, warning threshold changes, warning occurrence, warning processing, record the operator, operation time, IP address of the operation terminal, original value, modified value, and operation type; for example, when the manager modified the coal quality threshold through the office computer (IP address is 192.168.1.100) at 14:30 on May 10, 2023, the original calorific value threshold was changed from 5000 kcal to 5100 kcal, and the operation type was edited The system will record this information completely; during the storage process, the AES encryption algorithm is used to encrypt the data to ensure the security and confidentiality of the data and prevent the data from being illegally obtained and tampered with; it also provides a query function, and managers can query event records in a variety of ways; including querying by time range, such as querying all event records from April 1, 2023 to April 30, 2023; querying by operator, such as querying all operation records performed by a specific manager; you can also query by event type, such as only querying warning events or data modification events; the query results are displayed in a table format, including all the above-mentioned relevant information, and support exporting to the common format of Excel, which is convenient for managers to conduct further analysis and report generation, and provide a powerful tool for auditing and tracing problems in system operation; among them, for each event record, including: recording the modification operation of the manager on the data, when modifying the coal purchase plan quantity, record the modifier, modification time and the values ​​before and after the modification; for the adjustment of the warning threshold, the relevant information of the adjustment is also recorded; and stored in the event log database for subsequent query and tracing, providing a basis for analyzing the system operation and solving problems, ensuring the traceability and stability of the system operation; management The management evaluation module summarizes and analyzes statistical data on a weekly, monthly, quarterly or annual basis, calculates the compliance rate, non-compliance rate, output, and consumption of each data, and generates analysis reports based on the analysis results; for example, it collects statistics on the compliance rate of the monthly coal procurement plan, the compliance rate of coal quality, the coal consumption in different time periods, and the limestone consumption data, and displays them in the form of charts, such as a bar chart showing the comparison of the coal procurement plan achievement rate in different months, and a line chart showing the trend of coal quality over time; through these summaries and analyses, it provides comprehensive management decision support for managers, helping them evaluate the effectiveness of fuel management, identify potential problems and formulate improvement measures;

[0057] It also includes a video AI intelligent analysis subsystem, which is used to connect to the video monitoring equipment of the coal sampling, sample preparation and testing process to obtain real-time video streams; it uses high-definition cameras and stable video transmission technology to ensure the clarity and smoothness of video images, and can clearly capture each link and operation details in the coal monitoring process; it uses image recognition technology and deep learning algorithms to analyze video images to identify personnel operation behaviors, equipment operating status and sample processing during the detection process; for example, through the trained model, it can identify whether the sampling personnel are sampling in accordance with standard operating procedures, such as whether the selection of sampling points is reasonable and whether the sampling depth meets the requirements; it can determine whether the sample preparation equipment is operating normally, such as whether the speed of the sample preparation machine is within the normal range and whether the grinding time meets the standard; it can also detect whether the sample handling during the test is standardized, such as whether the sample is weighed accurately and whether the reagent is added correctly;

[0058] When the video AI intelligent analysis subsystem detects a violation, such as the sampling personnel failing to collect samples according to regulations, the sample contamination during the sample preparation process, or abnormal fluctuations in the test data, an alarm signal will be issued immediately, and the alarm information will be highlighted on the system interface. At the same time, the video clip, time, location and type information of the violation will be recorded; the alarm signal will remind quality supervision personnel and laboratory managers through sound and pop-up windows; relevant personnel can investigate and deal with violations in a timely manner based on the information recorded by the system, ensure the standardization and accuracy of the coal quality detection process, and ensure the reliability of coal quality data, thereby improving the effectiveness of the entire fuel intelligent supervision system.

[0059] In the data collection module, the process of sorting and verifying the collected data includes: first, clarifying the data types including coal purchase plan data, coal transportation process data, coal entry and combustion related data, and fuel data and raw material data; among them, coal purchase plan data includes monthly purchase plan coal quantity and quality information; coal transportation process data includes each ship coal contract, loading port, unloading port, and transit quantity loss data; coal entry and combustion related data include entry calorific value, furnace calorific value, coal storage in the coal yard, coal loading, and coal consumption; fuel data and raw material data include limestone consumption, ash production, ash slag, and fuel oil data;

[0060] Then, different verification methods are used for different types of data. For numerical data, numerical range verification and data consistency verification are used, and a calibrated physical range is preset. If the numerical data exceeds the calibrated physical range, it is determined to be abnormal data; at the same time, for the numerical data of the same batch of coal in different links, if the interpolation value is greater than the calibrated physical range and there is no reasonable reason, it is also marked as abnormal data; for example, the calorific value of coal should be within a reasonable physical range, such as between 3000-7000 kcal. If the collected calorific value exceeds this range, it is determined to be abnormal data; at the same time, the calorific value data of the same batch of coal in different links are compared. If there is a large difference and there is no reasonable reason, such as the difference between the calorific value entering the factory and the calorific value of the sampled test is more than 200 kcal, it is marked; for text data, format verification and keyword matching verification are used, and the text data is verified based on the standard format corresponding to the text data. If it does not conform to the standard format, it is marked as abnormal data, such as the name of the coal supplier and the contract number, format verification and keyword matching verification are used to ensure that the data format complies with the regulations, such as the contract number should be in a specific combination of numbers and letters and contain the supplier's key identification information;

[0061] When abnormal data is found, the data collection module performs data repair. The process includes: for simple errors, including typos or format errors during data entry, automatic repair algorithms are used to correct them; for example, if the decimal point position of coal quantity data is wrong, automatic correction is performed based on the reasonable range of the data and historical data trends;

[0062] For complex errors or missing data, estimates and supplements are made through data interpolation and regression analysis. For example, when the coal consumption data for a certain period of time is missing, the value of the missing data is estimated using a linear regression model based on the consumption data for the previous and subsequent periods of time and the unit operating load. At the same time, the data verification process and repair results are recorded for management personnel to review and audit, ensuring that the collected data has high accuracy and reliability, providing a solid foundation for subsequent data analysis and supervision.

[0063] For simple errors, including typos or format errors during data entry, the automatic repair algorithm is used to correct the process, including: For common typos during data entry, such as wrong decimal point position or wrong number of digits in numerical data, the automatic repair algorithm first makes a judgment based on the type and expected range of the data; for example, coal quantity data is usually within a reasonable range A. If the collected data is seriously inconsistent with the order of magnitude of the reasonable range A, and through comparison with data of the same type and analysis of data change trends, it is found that the anomaly is caused by the decimal point position deviation, the algorithm will automatically correct it according to the reasonable range A of the value and the magnitude distribution law of historical data; for example, if the historical coal purchase quantity is usually between hundreds of tons and tens of thousands of tons, and the collected data is thousands of tons but obviously does not conform to the actual purchase situation, and it is judged that the decimal point may be misplaced, the algorithm adjusts the decimal point position according to the order of magnitude corresponding to the reasonable range A to return it to the normal numerical range;

[0064] In terms of text data, if there are spelling errors or irregular formats in the coal supplier name or contract number, the automatic repair algorithm uses the pre-established word library and format templates for processing; for the supplier name, fuzzy matching is performed with the standard supplier name library already in the system to find the closest correct name; if a word in the name is found to be misspelled or there are redundant characters, the algorithm will refer to the correct spelling and common name formats in the word library to correct it; for the contract number, according to the established numbering rules and check digit calculation method, if the collected contract number does not comply with the rules, the algorithm will analyze the error type, such as a possible digit error or missing character, and then repair it according to the rules and related business logic to ensure the accuracy and uniqueness of the contract number and make it meet the system's format requirements for this type of data, so that subsequent data processing and query can proceed smoothly, thereby ensuring the efficiency of the entire data collection process and the reliability of data quality;

[0065] In the coal full-process intelligent supervision system, when encountering complex errors or missing data, data interpolation and regression analysis are effective means of estimation and supplementation. The specific technical contents are as follows:

[0066] Data interpolation is often used when time series data, such as daily coal consumption or coal quality index data at a certain stage, are missing. If the data shows a linear trend, linear interpolation is used. For example, if the coal consumption at time point t1 is x1 and the coal consumption at time point t3 is x3, when the data at time t2 is missing and the intervals between t1, t2, and t3 are uniform, the formula to estimate the coal consumption at time t2; if the data change trend presents a quadratic curve, the spline interpolation method is used; taking the change of coal calorific value over time as an example, the known time points and corresponding calorific value data are used as nodes to construct a spline function, and the missing calorific value data is supplemented by solving the value of the spline function at the missing time point, thereby ensuring the continuity and integrity of the data in the time series and better reflecting the change process of coal calorific value;

[0067] Regression analysis plays a key role in dealing with complex data relationships related to multiple factors. For example, when estimating the relationship between limestone consumption and coal combustion, unit load, and coal sulfur content, a large amount of historical data is first collected, including limestone consumption in different time periods, corresponding coal combustion, unit operating load, and coal quality index data; using the multivariate linear regression model:

[0068] y=β0+β1x1+β2x2+…+β n x n +e, where y is the limestone consumption, x1, x2,…, x n are factors affecting limestone consumption (such as coal combustion, unit load), β0, β1, …, β n is the regression coefficient, and e is the error term. The regression coefficient is estimated by the least square method to establish a regression equation. When the limestone consumption data at a certain moment is missing, the values ​​of other related factors known at that moment are substituted into the regression equation to calculate the estimated value of limestone consumption. At the same time, in the process of regression analysis, the determination coefficient R is calculated. 2 To evaluate the goodness of fit of the model to the data, ensure that the model has high accuracy and reliability, provide effective technical support for the estimation and supplement of complex data, and improve the integrity and accuracy of the system's data processing for the entire coal process.

[0069] It should be further explained that in the specific implementation process, the accuracy and reliability of the coal full-process data were ensured through rigorous data collection, collation and verification processes; in the data collection module, special verification methods were used for different types of data, and abnormal data were effectively repaired, laying a solid foundation for subsequent analysis;

[0070] Among them, in the collection of coal quantity data, the quantity information of each link is accurately recorded, so that the power plant can clearly grasp the coal inventory and flow situation, avoid production interruptions or resource waste due to data errors, and ensure the accurate formulation and implementation of production plans.

[0071] When storing in the database, the process of classifying and storing the data includes: for coal quantity data, a coal quantity purchase plan data table is created when storing in the database, and the table contains fields for purchase plan quantity, quantity at the loading port stage, unloading port stage, and factory entry stage, coal loading quantity, coal consumption, and coal yard inventory; each record corresponds to a timestamp or business operation batch to track the change of quantity over time; during the data entry process, the quantity information obtained through the data acquisition module is inserted according to the preset format and data type specifications; for example, the purchase plan quantity field is stored in integer type to accurately record the planned purchase of coal tons; and for the coal consumption field, according to the measurement accuracy requirements of the power plant, a floating point type may be used to ensure that the actual consumption can be accurately reflected;

[0072] Coal quality data is stored in the coal quality data table, which contains key quality indicators of coal such as calorific value, ash content, sulfur content, and volatile matter, and is associated with sampling time, sampling location, and coal batch number information. To ensure the accuracy and traceability of the data, the quality data is verified and standardized before being stored. For example, calorific value data is uniformly converted into international standard units (such as kcal / kg), and the data value range is set according to industry standards for verification. Data out of the range will trigger an abnormal prompt, and the detection equipment will be rechecked or calibrated based on the abnormal prompt. At the same time, a composite index strategy based on coal batch number and sampling time is used to query and retrieve the quality data of a specified batch of coal at different time points, providing efficient data access support for quality analysis and comparison.

[0073] The transportation data is stored in the coal transportation data table, which records the contract information, transportation tool details, transit information and transportation time nodes during the coal transportation process; among them, the contract information includes the contract number, signing date, and supplier information; the transportation tool details include the ship number, train number, transportation starting point and destination; the transit information includes the number of transits, transit location, transit loss and loss ratio; the transportation time nodes include loading time, expected arrival time, and actual arrival time;

[0074] The data of the combustion process is stored in the coal combustion data table, which contains the quality data of the coal entering the furnace, combustion time, unit operating parameters, the amount of ash produced by combustion, and the carbon content of fly ash; among them, the quality data is obtained from the coal quality data table, and the unit operating parameters include the load, temperature, and pressure of the unit operation; through the above classified storage of different links and categories of data on coal quantity, quality, transportation, and combustion, a database with a clear structure, easy management and efficient query is constructed, which provides a solid foundation for the stable operation and data analysis of the coal full-process intelligent supervision system.

[0075] The process of training the model using the deep learning algorithm includes: in the coal full-process intelligent supervision system, the analysis results of the model training using the deep learning algorithm in each link are as follows: in the coal quantity link, a multi-layer neural network including an input layer, multiple hidden layers and an output layer is used, and the input layer receives the historical coal purchase quantity, the coal quantity records at the loading port stage, the unloading port stage and the factory stage, the timestamp, and the season and market supply and demand conditions; the hidden layer uses the long short-term memory network LSTM unit to capture the dynamic change law of the coal quantity in the time series, and combines the fully connected layer for feature fusion and nonlinear transformation, and trains through a large amount of historical data, using the mean square error (MSE) as the loss function, and using stochastic gradient descent (SGD) or Adam to adjust the network weights to obtain a coal quantity analysis model;

[0076] The coal quantity analysis model is used to predict the demand for coal quantity in the future. By comparing the current inventory and the amount of coal in transit, it is determined whether the amount of coal is sufficient and generates the analysis results of the coal quantity. For example, before the peak season of power generation, based on historical data and current production plans, it is predicted that the demand for coal will increase significantly. If the inventory level is lower than the safety threshold and the procurement plan is not adjusted in time, the coal quantity analysis model will issue an early warning of insufficient coal quantity. At the same time, for the loss of coal during transportation and storage, by comparing the changes in the amount of coal at different stages, combined with the transportation distance and storage time, it is analyzed whether the loss is within a reasonable range. If there is abnormal loss (such as exceeding a certain standard deviation of the historical average loss rate), it indicates that there may be transportation leakage or storage management problems.

[0077] In the coal quality link, a neural network with a multi-layer perceptron (MLP) structure is constructed. The input layer includes the quality indicators of coal calorific value, ash content, sulfur content, and volatile matter, coal source information, mining process data, and environmental parameters during sampling and testing. The hidden layer uses ReLU for nonlinear processing to enhance the model's ability to learn complex quality relationships. Based on a large amount of coal sample data with labeled quality levels: high-quality, qualified, and low-quality, the cross-entropy loss function and optimization algorithm are used to train the model to classify and evaluate coal quality, and a coal quality analysis model is obtained through training.

[0078] The coal quality analysis model is used to determine whether the coal quality meets the production requirements of the power plant. For example, when the sulfur content of the coal is higher than the maximum sulfur content threshold that the power plant boiler can withstand, the coal quality analysis model determines that the coal quality does not meet the standard, and indicates the possible corrosion and environmental pollution risks to the equipment, and generates the analysis results of the coal quality; at the same time, it can predict the changing trend of coal quality during subsequent transportation, storage and combustion. For example, based on the volatile matter of the coal and the current storage environment temperature and humidity, it can predict the risk of spontaneous combustion or the degree of quality deterioration of the coal during storage, providing a decision-making basis for coal storage management.

[0079] The process of training the model using deep learning algorithms also includes: in the coal transportation link, designing a hybrid model combining convolutional neural network CNN and recurrent neural network RNN; the CNN layer is used to extract the spatial features in the transportation contract text, the image appearance of the transportation vehicle or ship, the load identification features, and the transportation route geographic information, and the RNN layer is used to process the transportation time series data, such as the transportation start time, the estimated arrival time, and the actual transportation progress at different stages; through a large number of historical transportation records for training, using transportation time deviation and transportation loss rate as loss indicators, the training model optimizes the prediction and evaluation capabilities of the transportation process, and obtains a coal transportation analysis model; the coal transportation analysis model is used Predict the arrival time of coal transportation, warn of possible transportation delays in advance, and generate analysis results of coal transportation so that power plants can prepare for receiving goods; for transportation losses, accurately judge whether the losses are reasonable by analyzing the stability of the transportation tools, road or sea conditions, and the degree of standardization of loading and unloading operations during the transportation process and the actual losses; if the coal transportation analysis model predicts that a certain transportation route may cause coal spillage or increased moisture under certain weather conditions, and the actual transportation results exceed the reasonable range, the supervision system will generate a prompt message at this time, investigate the transportation process based on the prompt message, and make improvements immediately after obtaining the reasons for the investigation to ensure the efficiency and quality of coal transportation;

[0080] In the coal combustion link, a deep neural network architecture is adopted. The input layer receives the quality data of coal before entering the furnace, the load of the unit operation, combustion temperature, pressure, air flow parameters and performance parameters of the combustion equipment; the hidden layer learns the physical and chemical changes in the coal combustion process through a combination of multi-layer convolution and fully connected layers; a large amount of combustion experimental data and actual production operation data are used for training, and the combustion efficiency and pollutant emissions are used as evaluation indicators. The model parameters are continuously adjusted through the back propagation algorithm to construct a coal combustion analysis model; the coal combustion analysis model is used to predict the combustion efficiency of coal under the current unit operating conditions, and by comparing the actual combustion efficiency with the theoretical optimal efficiency or the historical average efficiency, it is judged whether there is a problem in the combustion process, and the analysis results of coal combustion are obtained; for example, if the model predicts that the combustion efficiency should reach 90% under the current coal quality and unit load, but the actual monitored combustion efficiency is only 80%, it indicates that there may be insufficient combustion. At the same time, it is also used to predict the pollutants generated during the combustion process: emissions of sulfur dioxide, nitrogen oxides, and smoke. When the predicted emissions exceed the environmental emission standards, early warning is given and emission reduction measures are recommended. The emission reduction measures include adjusting the coal blending ratio and optimizing the combustion air supply, so as to achieve clean coal combustion and environmentally friendly production. Based on the analysis results of the four links of coal quantity, quality, transportation and combustion, the analysis results of the entire coal process are obtained. The analysis results include whether the coal quantity is sufficient, whether the quality is up to standard, whether the loss in each link is reasonable, and the utilization efficiency of various fuels and raw materials. The analysis results are output in a standardized data format to provide a decision-making basis for subsequent early warning modules and management evaluation modules, so as to achieve intelligent supervision and effective management of the entire coal process. Among them, coal characteristic data includes data on the impact of changes in the unit operating load on coal consumption and combustion efficiency.

[0081] By using models trained with deep learning algorithms, we can deeply explore the potential patterns and characteristics of data on coal quantity, quality, transportation and combustion, providing a scientific basis for management decisions. In the coal quantity link, we can accurately predict demand and losses, and plan procurement and inventory strategies in advance. In the coal quality link, we can accurately determine whether the coal meets production requirements, effectively reducing the risks of equipment damage and environmental pollution. In the transportation link, we can provide early warning of delays and reasonably assess losses, and optimize transportation arrangements. In the combustion link, we can predict combustion efficiency and pollutant emissions, guide the implementation of energy-saving and emission reduction measures, and comprehensively improve the refinement of power plant operation management and decision-making efficiency.

[0082] In terms of coal quantity supervision, the data collection module specifically performs the following operations: for the monthly procurement plan and execution status, the procurement plan coal quantity and actual in-plant quantity information in the business system are collected regularly and saved synchronously in the local supervision system; during the collection process, a scheduled task scheduling mechanism is adopted to ensure the timeliness and accuracy of the data, and the latest procurement plan and in-plant quantity data are automatically obtained from the business system at 2 a.m. every day, and the data format is converted and verified, and stored in a special procurement plan data table in the database of the local supervision system; at the same time, managers are allowed to manage the synchronized data. When data errors are found or additional explanations are required, managers can operate in the management background, such as correcting the wrong in-plant quantity or filling in the reasons for the delay, such as the delay of coal transportation due to bad weather during transportation, so that the actual in-plant quantity is lower than the planned quantity, etc.; the supervision system is based on the monitoring rules set by the managers. The monitoring rules include setting a third-level warning when the monthly achievement rate is lower than 70%, and a second-level warning when it is lower than 50%, making warning judgments, and displaying them in the form of tables and charts through the dashboard, so that managers can intuitively view the progress of procurement plan execution and warning status;

[0083] In view of the quantity loss after coal loading, unloading and transshipment, the actual loading and unloading port quantities in the business system are monitored and collected, and the transit loss and loss ratio are automatically calculated; in the calculation process, accurate calculations are performed according to mathematical formulas, and the transit loss amount is equal to the loading port quantity minus the unloading port quantity, and the loss ratio is equal to the transit loss amount divided by the loading port quantity multiplied by 100%; the calculation results are saved synchronously in the local supervision system, and managers are allowed to manage the data. For example, when it is found that the manual entry in the business system is biased or the collected data is inaccurate, resulting in an incorrect loss value, the manager can make corrections; the supervision system is based on the set loss threshold between different destinations. For example, for coal shipped to a specific power plant, the loss rate threshold is set to ±0.3%. When the loss threshold is exceeded, an early warning is issued, and the loading port quantity, unloading port quantity, and loss ratio information are displayed in clear tables and charts to facilitate managers to analyze and deal with quantity loss problems.

[0084] In terms of coal quality supervision, the data processing module operates in the following way:

[0085] For the coal quality supervision of the monthly procurement plan and execution, the procurement plan of the business system and the quality information of coal entering the site are collected regularly and stored in the coal quality data table in the database; during the collection, the collected quality data are standardized and the quality indicators of different units or precisions are uniformly converted into a standard format;

[0086] The data is analyzed using the quality assessment model in the deep learning algorithm. The model is trained based on a large amount of historical coal quality data and quality standards to identify whether the coal quality meets the requirements of the procurement plan. For example, when the planned coal calorific value is 5500 kcal and the actual incoming coal calorific value is lower than 5300 kcal, the quality assessment model determines that it does not meet the standard and triggers an early warning. At the same time, managers are allowed to manage the synchronized data. If the data is found to be incorrect or requires additional explanation, such as possible deviations in the coal sampling process, corrections or annotations are made. Early warnings are issued based on the set quality thresholds, and the comparison of the planned coal quality, actual coal quality and quality thresholds is displayed in the form of charts on the dashboard, so that managers can intuitively understand the implementation of coal quality.

[0087] In the quality supervision of coal loading, unloading and transshipment, the actual information of loading and unloading port quantity and settlement calorific value difference in the business system is monitored and collected, and the transit calorific value loss and loss ratio are automatically calculated; when calculating the transit calorific value loss, the physical and chemical change factors of coal in the loading, unloading and transshipment process are comprehensively considered based on professional calorific value calculation methods and the principle of energy conservation; the transit calorific value loss and loss ratio are analyzed using the trained quality change monitoring model to determine whether the change in coal quality during the transshipment process is within a reasonable range; for example, when the settlement calorific value difference exceeds ±10 kcal or the transit calorific value loss exceeds ±20 kcal, the model determines it as abnormal and issues an early warning; managers can manage and correct the data, and the supervision system issues an early warning based on the set calorific value loss threshold, and displays information such as the loading port quantity, unloading port quantity, settlement calorific value difference, and transit calorific value loss in detailed charts, providing strong support for managers to analyze the reasons for changes in coal quality;

[0088] It should be further explained that, in the specific implementation process, in the early warning module, based on the analysis results of the data processing module, the process of issuing early warnings according to the set thresholds includes: the early warning is divided into four levels, and different levels correspond to different severity and processing requirements; for coal quantity supervision, when the execution of the monthly purchase plan coal quantity does not reach the set monthly achievement rate threshold, such as less than 80%, a warning message of the corresponding level is issued; when the quantity loss rate of coal after loading and unloading and transshipment exceeds ±0.5%, a warning message is also triggered; in terms of coal quality supervision, if the quality of coal in the monthly purchase plan does not meet the standards, or the settlement calorific value difference after coal loading and unloading and transshipment is greater than ±10 kcal, the transit calorific value loss is greater than ±20 kcal, and the calorific value difference between entering the factory and the furnace is greater than 100 kcal, a warning message will be issued; the warning information is notified to the management personnel through the system interface, SMS, and email to ensure that the management personnel can obtain the warning information and take corresponding measures in the first time.

[0089] The warnings in the warning module are divided into four levels: level 1 warning, level 2 warning, level 3 warning and level 4 warning. Different levels correspond to different severity and processing requirements. The process includes:

[0090] The first-level warning is the most serious level, which is triggered when the key quality indicators of coal are seriously below the standard and may affect the safe and stable operation of the unit, or when there is a serious shortage of coal supply that may cause the unit to shut down. For example, the calorific value of the coal entering the plant is lower than 80% of the designed minimum calorific value and lasts for more than 24 hours, or the coal inventory in the coal yard is lower than 50% of the minimum threshold for safe operation of the unit and cannot be replenished in the short term. For the first-level warning, the supervision system immediately notifies key personnel including senior management of the power plant, heads of production departments, and heads of fuel procurement departments through a variety of emergency notification methods, such as pop-up display on the large screen of the power plant control center, sending emergency text messages and voice calls to relevant management personnel; at the same time, the emergency plan is activated, which includes emergency adjustment of unit operating parameters to reduce energy consumption, activation of backup fuel supply channels, and organization of professional and technical personnel to quickly detect and analyze coal quality problems to ensure the safe and stable operation of the power plant, and requires relevant personnel to feedback the treatment measures and progress within the specified time of 1 hour, and the system tracks and records the entire treatment process.

[0091] Level 2 warnings indicate more serious problems, such as a certain degree of deviation in coal quality but not reaching the level of level 1 warning, or a serious lag in the implementation of the coal procurement plan, which may affect subsequent production arrangements; for example, the sulfur content of coal exceeds the standard value by 50% but does not reach the level of level 1 warning, or the completion rate of the monthly procurement plan is less than 40%; after triggering the level 2 warning, the system will inform the corresponding management personnel including the production department supervisor and the fuel management department supervisor through a pop-up window on the system interface and SMS notification; the management personnel must formulate and implement response measures within the specified time of 4 hours, including secondary inspection of coal, negotiation with suppliers to speed up the delivery progress or adjust the procurement plan, and feedback the processing results to the system, which will record and evaluate the processing situation. If the problem is not effectively resolved, it may be upgraded to level 1 warning;

[0092] The third-level warning is for general problems, such as coal quantity loss slightly higher than the normal threshold but within the acceptable range, or limestone consumption close to the upper limit of the warning; for example, the coal transfer quantity loss rate reaches ±0.8% but less than ±1.0%, or the limestone consumption reaches 90% of the rated consumption; when the third-level warning occurs, the system displays the warning information on the system interface, and sends email notifications to fuel transportation management personnel and limestone procurement personnel, and investigates and handles the warning situation within 8 hours, such as checking whether there are problems in the coal loading and unloading and transshipment process, verifying the inventory and use of limestone, etc., and entering the processing results into the supervision system, which monitors and records the processing process. If the problem persists or worsens, it may be upgraded to a second-level warning;

[0093] Level 4 warning is the lightest level and is used to indicate some potential minor problems, such as a coal quality indicator that is close to but does not exceed the threshold, or a small fluctuation in the amount of non-production coal. For example, the ash content of coal is close to but does not exceed 95% of the standard value, or the amount of non-production coal fluctuates by more than 10% within a week but does not exceed the set larger fluctuation threshold. For level 4 warnings, the supervision system displays relatively hidden warning prompts in the system interface by recording them in the system log and marking them with blue in the corresponding data report. Relevant personnel can pay attention to and handle these warning information in their daily work, such as regularly checking the trend of coal quality data, analyzing the reasons for fluctuations in the amount of non-production coal, etc., and record the handling situation in the system for subsequent analysis and statistics by the system.

[0094] It should be further explained that, in the specific implementation process, the four-level warning mechanism of the early warning module has achieved accurate graded handling of problems; from the first-level warning that seriously affects the safe and stable operation of the unit to the fourth-level warning for minor potential problems, different levels correspond to different processing procedures and response time requirements; this ensures that managers can quickly focus on key issues and take targeted measures in a timely manner, such as quickly launching emergency plans during the first-level warning, minimizing losses, ensuring the safe and stable operation of the power plant, and avoiding major accidents and economic losses caused by untimely problem handling.

[0095] In the management evaluation module, the process of generating the corresponding analysis report includes: summarizing and analyzing the statistical data in the system according to the predetermined time period: week, month, quarter or year; in the summarizing process, extracting the coal quantity, quality, inventory, consumption, limestone consumption, ash production, ash slag, and fuel data from the database, and performing data cleaning and preprocessing to remove abnormal values ​​and duplicate data; for example, for coal consumption data, using data smoothing algorithms to remove abnormal high or low values ​​caused by temporary equipment failures or measurement errors; using data analysis algorithms and statistical models to calculate the compliance rate, non-compliance rate, average value, standard deviation, and trend change indicators of each data; such as calculating the compliance rate of the monthly coal procurement plan, the average value and standard deviation of coal quality indicators over time, and the average value of limestone consumption in different seasons;

[0096] An analysis report is generated based on the analysis results, and the content of the analysis report includes text description, data tables and charts; the text description includes a summary and explanation of the analysis results, such as the overall trend of the implementation of the coal procurement plan, the causes and impacts of fluctuations in coal quality, and the changing characteristics of the consumption of various types of fuels and raw materials; the data table is used to list the values ​​and calculation results of key data, such as the coal procurement plan achievement rate for each month, and the coal quality index values ​​for different time periods; the chart display adopts the form of bar charts, line charts, and pie charts to intuitively present the changing trends and proportional relationships of the data; for example, a line chart is used to show the curve of the change of coal quality calorific value over time, and a pie chart is used to show the proportion of different fuels in the total cost; the analysis report can be viewed and downloaded through the system interface, providing managers with a comprehensive and accurate basis for decision-making, helping managers to promptly identify problems, optimize management strategies and improve fuel management efficiency.

[0097] The management evaluation module conducts in-depth summary and analysis of statistical data on a periodic basis, and generates detailed reports containing text, tables and charts, which intuitively present the compliance rate and change trend of key indicators in each link of the entire coal process; managers use this to comprehensively evaluate the effectiveness of fuel management, accurately identify potential problems and formulate effective improvement measures, continuously optimize the management strategies of coal procurement, storage and use, improve fuel management efficiency, reduce operating costs, and enhance the overall competitiveness of power plants.

[0098] In the video AI intelligent analysis subsystem, image recognition technology and deep learning algorithms are used to analyze video images to identify personnel operation behaviors, equipment operating status and sample processing during the detection process, and to detect whether the sample processing during the test is standardized. The process includes: First, in the image acquisition stage, a high-resolution camera with a wide dynamic range and low-light performance is used to ensure that clear and stable video images can be obtained under different lighting conditions and complex environments such as coal sampling, sample preparation, and testing; the camera is installed along the detection process area to ensure that there are no visual blind spots and transmit video images to the video AI intelligent analysis subsystem in real time; based on video images, the target detection and behavior recognition model in the deep learning algorithm is used to identify personnel operation behaviors; first, a large number of labeled personnel operation sample images are used to train the behavior recognition model, which is used for the behavior recognition model to identify the action posture of the sampling personnel and whether the sampling operation is performed in accordance with the standard process, including Including the correct use of sampling tools, standardized selection of sampling points, and accurate control of sampling depth; obtain behavior recognition results; in the sample preparation process, identify whether the operator correctly operates the sample preparation equipment, that is, identify the equipment operation status, and obtain operation recognition results; operation recognition results include the control of grinding time and whether the sample screening operation is compliant; in the test link, detect whether the test personnel accurately weigh samples, add reagents, and operate instruments to obtain test recognition results; the behavior recognition model is based on the convolutional neural network CNN architecture, extracts personnel features and action features in the image, gradually reduces the image resolution and extracts key features through multiple convolutional layers and pooling layers, and performs classification and judgment in the fully connected layer, and compares the identified behavior recognition results, operation recognition results, and test recognition results with the preset standard operating procedures. Once a deviation is found, it is determined to be an abnormal behavior; among them, the standard operating procedures include standard sampling operations, standard sample preparation operations, and standard test operations;

[0099] It should be further explained that, in the specific implementation process, in terms of equipment operation status identification, image-based equipment fault diagnosis technology is adopted, and deep learning models are used to learn the appearance characteristics of the equipment in normal operation and different fault states, and the image information of the movement status of the components; for example, for the motor and transmission device of the sample preparation equipment, by analyzing the vibration image characteristics, temperature image characteristics (which can be obtained by infrared thermal imaging technology) and displacement image characteristics of the mechanical components during operation, it is judged whether the equipment has potential faults or has already failed; during the model training process, a large amount of image data of normal and faulty equipment is used for comparative learning, and the generative adversarial network (GAN) is used to enhance the diversity of fault samples and improve the generalization ability of the model; when it is detected that there is a significant difference between the equipment operation image and the normal state image, such as the motion blur image characteristics caused by abnormal motor speed and the abnormal thermal radiation image characteristics caused by equipment overheating, the system will promptly issue an equipment fault warning and provide possible fault causes and related suggestions;

[0100] Image segmentation and feature extraction technology are used to analyze the sample processing situation. The sample area in the video image is segmented and identified through a deep learning model, and the state changes of the sample at each processing stage are analyzed. This includes: during the packaging process of the sample after sampling, check whether the sample is sealed intact and the label is correctly affixed; during the sample preparation process, observe whether the particle size change and color change of the sample are in line with expectations; during the analysis process, detect whether the reaction phenomenon of the sample in the reaction container is normal, such as whether the flame color and combustion speed in the combustion test are within the normal range; the model is based on the U-Net image segmentation network architecture, and classifies and segments the sample image at the pixel level, extracts the key feature information of the sample, and compares it with the standard sample processing process and the expected sample feature change law to ensure the accuracy and standardization of the sample processing process and the reliability of the coal quality inspection results.

[0101] The video AI intelligent analysis subsystem is used to monitor the coal sampling, sample preparation and testing processes in real time, effectively identifying violations of personnel operation regulations, hidden equipment failures and irregular sample handling. Once a violation is detected, an alarm is immediately triggered and detailed information is recorded, prompting relevant personnel to investigate and handle the violation in a timely manner, ensuring the standardization and accuracy of the coal quality testing process, ensuring the reliability of coal quality data, improving the compliance and reliability of the entire fuel management process, and enhancing the power plant's ability to control coal quality.

[0102] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. A coal full-process intelligent supervision system, characterized in that: It includes a data acquisition module, a data processing module, an early warning module, an event recording module and a management and evaluation module, wherein the modules are connected by electrical signals; The data acquisition module is used to obtain data from the fuel full process management system, SAP settlement system, fuel platform system, and SIS system, and to organize and verify the collected data; the data processing module stores the organized and verified data into a database, and uses a deep learning algorithm to train the model; including classifying and storing the data when storing it in the database, and establishing corresponding data tables according to different links and categories of coal quantity, quality, transportation, and combustion; for the deep learning algorithm, a multi-layer neural network structure is used to train according to different working conditions and data characteristics; the potential laws and characteristics in the coal full process data are extracted to complete the analysis and prediction of coal quantity, quality, and combustion effect, thereby obtaining the analysis results; the early warning module issues an early warning according to the set threshold based on the analysis results of the data processing module; It also includes a video AI intelligent analysis subsystem, which is used to connect to the video monitoring equipment of the coal sampling, sample preparation and testing process to obtain real-time video streams; use cameras and video transmission technology to capture each link and operation details in the coal testing process; use image recognition technology and deep learning algorithms to analyze video images, identify the operating behavior of personnel, equipment operating status and sample processing during the testing process, and detect whether the sample processing during the testing process is standardized; When the video AI intelligent analysis subsystem detects a violation, it immediately issues an alarm signal and highlights the alarm information on the system interface, while recording the video clip, time, location and type information of the violation; the alarm signal reminds quality supervisors and laboratory managers through sound and pop-up windows.

2. A coal full-process intelligent supervision system according to claim 1, characterized in that: In the data collection module, the process of sorting and verifying the collected data includes: first, clarifying the data types including coal procurement plan data, coal transportation process data, coal entry and combustion related data, and fuel data and raw material data; among them, the coal procurement plan data includes the monthly procurement plan coal quantity and quality information; the coal transportation process data includes the coal contract for each ship, loading port, unloading port, and transit quantity loss data; the coal entry and combustion related data include the entry calorific value, the furnace calorific value, the coal storage in the coal yard, the coal loading amount, and the coal consumption; the fuel data and raw material data include limestone consumption, ash production, ash slag, and fuel oil data; Then, for numerical data, numerical range check and data consistency check are used, and a calibrated physical range is preset. If the numerical data exceeds the calibrated physical range, it is judged as abnormal data; at the same time, for the numerical data of the same batch of coal in different links, if the interpolation value is greater than the calibrated physical range and there is no reasonable reason, it is also marked as abnormal data; for text data, format check and keyword matching check are used, and the text data is checked based on the standard format corresponding to the text data. If it does not meet the standard format, it is marked as abnormal data; When abnormal data is found, the data collection module performs data repair, and the process includes: for simple errors, including typos or format errors during data entry, automatic repair algorithms are used to correct them; For complex errors or missing data, estimation and supplementation are carried out through data interpolation and regression analysis methods; at the same time, the data verification process and repair results are recorded.

3. A coal full-process intelligent supervision system according to claim 2, characterized in that: When storing in the database, the process of classifying and storing data includes: for coal quantity data, a coal quantity procurement plan data table is created when storing in the database, which contains fields such as procurement plan quantity, quantity at the loading port stage, unloading port stage, and factory stage, coal loading quantity, coal consumption, and coal yard inventory; each record corresponds to a timestamp or business operation batch; coal quality data is stored in a coal quality data table, which contains key quality indicators of coal such as calorific value, ash content, sulfur content, and volatile matter, and is associated with sampling time, sampling location, and coal batch number information; quality data is verified and standardized before being stored; and the data value range is set according to industry standards for verification, and data outside the range is An abnormal prompt will be triggered, and the detection equipment will be re-checked or calibrated based on the abnormal prompt; at the same time, a composite index strategy based on coal batch number and sampling time is adopted to query and retrieve the quality data of a specified batch of coal at different time points; the transportation link data is stored in the coal transportation data table, which records the contract information, transportation tool details, transit information and transportation time nodes during the coal transportation process; the combustion link data is stored in the coal combustion data table, which contains the quality data of the coal entering the furnace, combustion time, unit operating parameters, ash amount generated by combustion, and fly ash carbon content; through the above classified storage of different links and categories of data on coal quantity, quality, transportation, and combustion, a database is constructed.

4. A coal full-process intelligent supervision system according to claim 3, characterized in that: The process of training a model using a deep learning algorithm includes: In the full-process intelligent supervision system of coal, the analysis results of the model trained by deep learning algorithm in each link are as follows: In the link of coal quantity, a multi-layer neural network including input layer, hidden layer and output layer is adopted. The input layer receives the historical coal purchase quantity, coal quantity records of loading port stage, unloading port stage and factory stage, timestamp, season, and market supply and demand conditions; the hidden layer uses the long short-term memory network LSTM unit to capture the dynamic change law of coal quantity in time series, and combines the fully connected layer for feature fusion and nonlinear transformation. It is trained through historical data, uses mean square error MSE as the loss function, and uses stochastic gradient descent SGD or Adam to adjust the network weights to obtain the coal quantity analysis model; The coal quantity analysis model is used to predict the demand for coal quantity in the future. By comparing the current inventory and the amount of coal in transit, it is determined whether the amount of coal is sufficient and generates the analysis results of the coal quantity. At the same time, for the loss of coal during transportation and storage, by comparing the changes in the amount of coal at different stages, combined with the transportation distance and storage time, it is analyzed whether the loss is within a reasonable range. If there is abnormal loss, it indicates that there may be transportation leakage or storage management problems. In the coal quality link, a neural network with a multi-layer perceptron (MLP) structure is constructed. The input layer includes the quality indicators of coal calorific value, ash content, sulfur content, and volatile matter, coal source information, mining process data, and environmental parameters during sampling and testing. The hidden layer uses ReLU for nonlinear processing. Based on a large amount of coal sample data with labeled quality levels: high-quality, qualified, and low-quality, the cross entropy loss function and optimization algorithm training model are used to classify and evaluate coal quality, and the coal quality analysis model is trained. The coal quality analysis model is used to determine whether the coal quality meets the production requirements of the power plant. The coal quality analysis model determines that the coal quality does not meet the standards, and indicates the possible risks of corrosion to the equipment and environmental pollution, and generates the analysis results of the coal quality.

5. A coal full-process intelligent supervision system according to claim 4, characterized in that: The process of training a model using a deep learning algorithm also includes: In the coal transportation link, a hybrid model combining convolutional neural network (CNN) and recurrent neural network (RNN) is designed; the CNN layer is used to extract the spatial features in the transportation contract text, the image appearance of the transportation vehicle or ship, the load identification features, and the transportation route geographic information, while the RNN layer is used to process the transportation time series data; through historical transportation records, the training model is trained to optimize the prediction and evaluation capabilities of the transportation process, using transportation time deviation and transportation loss rate as loss indicators, and a coal transportation analysis model is obtained; The coal transportation analysis model is used to predict the arrival time of coal transportation, give early warning of possible transportation delays, and generate analysis results for coal transportation. For transportation losses, the stability of the transportation tools, road or sea conditions, the degree of standardization of loading and unloading operations, and the actual losses are analyzed to accurately determine whether the losses are reasonable. In the coal combustion process, a deep neural network architecture is used. The input layer receives the quality data of coal before entering the furnace, the load of the unit operation, combustion temperature, pressure, air flow parameters and performance parameters of the combustion equipment; the hidden layer learns the physical and chemical changes in the coal combustion process through a combination of multi-layer convolution and full connection layers; the combustion experimental data and actual production operation data are used for training, and the combustion efficiency and pollutant emissions are used as evaluation indicators. The model parameters are continuously adjusted through the back propagation algorithm to build a coal combustion analysis model; The coal combustion analysis model is used to predict the combustion efficiency of coal under the current unit operating conditions. By comparing the actual combustion efficiency with the theoretical optimal efficiency or the historical average efficiency, it is determined whether there are any problems in the combustion process and the analysis results of coal combustion are obtained. At the same time, it is also used to predict the pollutants generated during the combustion process: sulfur dioxide, nitrogen oxides, and smoke emissions. When the predicted emissions exceed the environmental emission standards, early warnings are issued and emission reduction measures are recommended. The emission reduction measures include adjusting the coal blending ratio and optimizing the combustion air supply, thereby achieving clean coal combustion and environmentally friendly production; Based on the analysis results of the four links of coal quantity, quality, transportation and combustion, the analysis results of the entire coal process are obtained; the analysis results include whether the coal quantity is sufficient, whether the quality is up to standard, whether the loss in each link is reasonable, and the utilization efficiency of various fuels and raw materials.

6. A coal full-process intelligent supervision system according to claim 5, characterized in that: In terms of coal quantity supervision, the data collection module specifically performs the following operations: For the monthly purchase plan and execution, the purchase plan coal quantity and actual quantity entering the factory are collected regularly in the business system and saved synchronously in the local supervision system. During the collection process, a scheduled task scheduling mechanism is adopted to obtain the latest purchase plan and factory quantity data from the business system, convert and verify the data format, and store it in a special purchase plan data table in the database of the local supervision system. Make early warning judgments based on the monitoring rules set by managers, and display them on the dashboard in the form of tables and charts; In view of the quantity loss after coal loading, unloading and transshipment, the actual loading and unloading port quantities in the business system are monitored and collected, and the transit loss amount and loss ratio are automatically calculated; in the calculation process, accurate calculations are performed according to mathematical formulas. The transit loss amount is equal to the loading port quantity minus the unloading port quantity, and the loss ratio is equal to the transit loss amount divided by the loading port quantity multiplied by 100%; the calculation results are synchronously saved in the local supervision system, and management personnel are allowed to manage the data. According to the set loss threshold between different destinations, an early warning is issued when the loss threshold is exceeded, and the loading port quantity, unloading port quantity, and loss ratio information are displayed in tables and charts.

7. A coal full-process intelligent supervision system according to claim 6, characterized in that: In terms of coal quality supervision, the data processing module operates in the following way: For the coal quality supervision of the monthly procurement plan and execution, the procurement plan of the business system and the quality information of coal entering the site are collected regularly and stored in the coal quality data table in the database; during the collection, the collected quality data are standardized and the quality indicators of different units or precisions are uniformly converted into a standard format; The data is analyzed using the quality assessment model in the deep learning algorithm. The model is trained based on a large amount of historical coal quality data and quality standards to identify whether the coal quality meets the requirements of the procurement plan. If the quality assessment model determines that the coal quality does not meet the standards, an early warning is triggered. At the same time, managers are allowed to manage synchronized data. If they find that the data is wrong or needs additional explanation, they can make corrections or add comments. Early warnings are issued based on the set quality thresholds, and the comparison of planned coal quality, actual coal quality and quality thresholds is displayed in charts on the dashboard. In the quality supervision of coal loading, unloading and transshipment, the actual loading and unloading port quantities and settlement calorific value differences in the business system are monitored and collected, and the transit calorific value loss and loss ratio are automatically calculated. The trained quality change monitoring model is used to analyze the transit calorific value loss and loss ratio to determine whether the changes in coal quality during the transshipment process are within a reasonable range. The model determines it as abnormal and issues an early warning.

8. The whole-process intelligent supervision system of coal according to claim 7 is characterized by: The warnings in the warning module are divided into four levels: level 1 warning, level 2 warning, level 3 warning and level 4 warning. Different levels correspond to different severity and processing requirements. The process includes: Level 1 warning is the most serious level. For level 1 warning, a pop-up window will be displayed on the large screen of the power plant control center, and emergency text messages and voice calls will be sent to relevant managers to notify key personnel including senior management of the power plant, heads of production departments, and heads of fuel procurement departments. At the same time, the emergency plan will be activated, which includes emergency adjustment of unit operating parameters to reduce energy consumption, activation of backup fuel supply channels, and organization of professional and technical personnel to conduct rapid detection and analysis of coal quality issues. Relevant personnel are required to feedback on treatment measures and progress within the specified time of 1 hour; Level 2 warning indicates a more serious problem. When a level 2 warning is triggered, the corresponding managers, including the production department manager and the fuel management department manager, will be notified through a pop-up window on the system interface or SMS notification. The managers must formulate and implement response measures within the specified time of 4 hours, including secondary inspection of coal, negotiation with suppliers to speed up supply progress or adjust procurement plans. Level 3 warning is for general problems. When a level 3 warning occurs, the warning information is displayed on the system interface, and an email notification is sent to the fuel transportation management personnel and limestone purchasers. The warning situation is investigated and handled within 8 hours, and the handling results are entered into the supervision system; Level 4 warning is the lightest level and is used to prompt some potential minor problems. For level 4 warning, the warning prompt is recorded in the system log on the system interface and marked in blue in the corresponding data report.

9. A coal full-process intelligent supervision system according to claim 8, characterized in that: In the management evaluation module, the process of generating the corresponding analysis report includes: summarizing and analyzing the statistical data in the system according to the predetermined time period: week, month, quarter or year; in the summarizing process, extracting the coal quantity, quality, inventory, consumption, limestone consumption, ash production, ash slag, and fuel data from the database, and performing data cleaning and preprocessing to remove abnormal values ​​and duplicate data; using data analysis algorithms and statistical models to calculate the compliance rate, non-compliance rate, average value, standard deviation, and trend change indicators of each data; An analysis report is generated based on the analysis results. The content of the analysis report includes text description, data tables and graphical presentations. The text description includes a summary and explanation of the analysis results. The data table is used to list the data values ​​and calculation results, and the graphical presentation uses bar graphs, line graphs and pie charts to present the changing trends and proportional relationships of the data.

10. A coal full-process intelligent supervision system according to claim 9, characterized in that: In the video AI intelligent analysis subsystem, image recognition technology and deep learning algorithms are used to analyze video images to identify personnel operation behaviors, equipment operating status and sample processing during the detection process, and to detect whether the sample processing during the test is standardized. The process includes: First, in the image acquisition stage, a camera is used to obtain video images; the camera is installed along the detection process area, and the video images are transmitted to the video AI intelligent analysis subsystem in real time; based on the video image, the target detection and behavior recognition model in the deep learning algorithm is used to identify personnel operation behaviors; first, the behavior recognition model is trained through the labeled personnel operation sample images, which is used for the behavior recognition model to identify the action posture of the sampling personnel and whether the sampling operation is performed in accordance with the standard process, including the correct use of sampling tools, sampling Standardized selection of sample points and accurate control of sampling depth; obtaining behavior recognition results; in the sample preparation process, identifying whether the operator correctly operates the sample preparation equipment to obtain operation recognition results; in the test link, detecting whether the test personnel accurately weigh samples, add reagents, and operate instruments to obtain test recognition results; the behavior recognition model is based on the convolutional neural network CNN architecture, which extracts personnel features and action features in the image, gradually reduces the image resolution and extracts key features through multiple convolutional layers and pooling layers, performs classification and judgment in the fully connected layer, and compares the identified behavior recognition results, operation recognition results, and test recognition results with the preset standard operating procedures. Once a deviation is found, it is determined to be an abnormal behavior; among them, the standard operating procedures include standard sampling operations, standard sample preparation operations, and standard test operations.

Images