AI-driven environment monitoring and early warning method, medium and equipment

Abstract

The invention discloses an AI-driven environmental monitoring and early warning method, medium and equipment, and relates to the technical field related to environmental monitoring, and the method comprises the steps: constructing an abnormal trend prediction model, and carrying out the abnormal event classification and abnormal trend prediction according to the monitoring data of a furnace flue; taking the exception prediction event as a search engine, and searching in the exception event analysis library to obtain an exception influence factor set; configuring an acquisition strategy of the monitoring acquisition equipment and obtaining abnormal monitoring probability data; and inputting the abnormal monitoring probability data into the abnormal recognition early warning model, performing abnormal event recognition and risk level judgment, and generating early warning information according to an abnormal event recognition result and a risk level. The technical problems that in the prior art, complex and changeable furnace flue environments are difficult to effectively cope with, abnormal events cannot be predicted in advance, and consequently the accuracy and timeliness of environment monitoring and early warning are insufficient are solved, and the technical effects that environment monitoring and early warning are intelligent, and the timeliness and accuracy of environment monitoring and early warning are improved are achieved.

Description

Technical Field

[0001] This application relates to the field of environmental monitoring technology, and in particular to AI-driven environmental monitoring and early warning methods, media and equipment. Background Technology

[0002] Furnaces and kilns are widely used in steel smelting, glass manufacturing, ceramic firing, and other fields. Their operation directly affects production efficiency and product quality. As a key channel for exhaust gas discharge, accurate and real-time monitoring of furnace and kiln flues is crucial. Traditional environmental monitoring methods are significantly inadequate when facing the complex and dynamically changing monitoring needs of furnace and kiln flues. On the one hand, existing monitoring relies on fixed thresholds for judgment. When the furnace and kiln operating status undergoes complex changes due to production process adjustments, equipment aging, etc., it cannot accurately reflect the true situation, easily leading to missed or misjudged abnormal events. On the other hand, regular inspections are not only inefficient, failing to achieve all-time, all-round monitoring of furnace and kiln flues, but also struggle to capture instantaneous anomalies within the flue, making it difficult to detect potential problems in a timely manner. This, in turn, affects the timeliness, accuracy, and comprehensiveness of furnace and kiln flue environmental monitoring.

[0003] At present, the relevant technologies have technical problems such as difficulty in effectively dealing with the complex and ever-changing flue gas environment of furnaces and kilns, and inability to predict abnormal events in advance, resulting in insufficient accuracy and timeliness of environmental monitoring and early warning. Summary of the Invention

[0004] This application provides an AI-driven environmental monitoring and early warning method, medium, and equipment, which solves the technical problems in the prior art that make it difficult to effectively cope with the complex and ever-changing flue gas environment of furnaces and kilns and that it is impossible to predict abnormal events in advance, resulting in insufficient accuracy and timeliness of environmental monitoring and early warning. It realizes the intelligentization of environmental monitoring and early warning and improves the technical effect of timely and accurate environmental monitoring and early warning.

[0005] This application provides an AI-driven environmental monitoring and early warning method, which includes: constructing an abnormal trend prediction model; classifying and predicting abnormal events and trends based on monitoring data from furnace flues using the abnormal trend prediction model to obtain predicted abnormal events and their probabilities; using the predicted abnormal events as a search engine to search an abnormal event analysis library to obtain a set of abnormal impact factors, which includes abnormal impact factors and their probability coefficients; configuring the acquisition strategy of the monitoring and acquisition equipment based on the predicted abnormal events and their probabilities, as well as the abnormal impact factors and their probability coefficients; obtaining abnormal monitoring probability data based on the acquisition strategy of the monitoring and acquisition equipment; inputting the abnormal monitoring probability data into an abnormal identification and early warning model to identify abnormal events and determine their risk levels; and generating early warning information according to the abnormal event identification results and risk levels.

[0006] In one possible implementation, the AI-driven environmental monitoring and early warning method is further configured to perform the following processing: collecting historical sample data of the furnace and kiln, including normal and abnormal states; comparing the furnace and kiln flue monitoring data corresponding to the normal and abnormal states to determine abnormal events and abnormal performance characteristics; establishing a training dataset based on the abnormal events and abnormal performance characteristics, and training a classifier using the training dataset; fitting time-series trend features based on the time-series sample data of the abnormal states, training a time-series network framework to obtain a trend probability prediction model; and integrating and connecting the classifier with the trend probability prediction model to obtain the abnormal trend prediction model.

[0007] In one possible implementation, the AI-driven environmental monitoring and early warning method is further configured to perform the following processing: Based on the monitoring data of the furnace flue in the abnormal state, a time-series monitoring data chain is constructed according to the monitoring time sequence; based on the identification nodes of the abnormal state in the time-series monitoring data chain, preliminary tracing is performed to obtain preliminary data change information; based on the change time period of the preliminary data change information, a comparison time period is determined; based on the comparison time period, periodic data of the normal state is extracted to obtain normal state periodic data; the normal state periodic data is aligned with the comparison time period data of the abnormal state to obtain difference data features, and the difference data features are used as abnormal performance features; abnormal performance features are mapped and associated according to the abnormal events corresponding to the abnormal state to determine the abnormal events and abnormal performance features.

[0008] In one possible implementation, the AI-driven environmental monitoring and early warning method is further used to perform the following processing: deploying monitoring sensors according to the characterization factors of the normal state and the abnormal state, wherein the monitoring sensors include at least monitoring sensors corresponding to temperature, pressure, flow rate, gas concentration, humidity, flame spectrum, and vibration spectrum; and collecting monitoring data of the furnace flue in real time according to the deployed monitoring sensors.

[0009] In one possible implementation, the AI-driven environmental monitoring and early warning method is further configured to perform the following processing: based on the abnormal events and their abnormal characteristics, analyze relevant monitoring data and the influence coefficients of each monitoring data on the abnormal events based on historical sample data; using the abnormal events as the top node and each relevant monitoring data as the branch node, establish connection edges between the nodes and the branch nodes, with the weight of the connection edge being the influence coefficient, to obtain the analytical graph structure of each abnormal event; and combine the analytical graph structures of all abnormal events in parallel to construct the abnormal event analysis library.

[0010] In one possible implementation, the AI-driven environmental monitoring and early warning method is further configured to perform the following processes: obtain the abnormal trend probability of the abnormal prediction event, configure the collection time window, wherein the larger the abnormal trend probability, the shorter the collection time window; set the collection frequency according to the collection time window and the influence probability coefficient of the abnormal influence factor; and configure the collection strategy of the monitoring and collection device corresponding to the abnormal influence factor according to the collection time window and the collection frequency, based on the mapping relationship between the abnormal prediction event and the abnormal influence factor.

[0011] In one possible implementation, the AI-driven environmental monitoring and early warning method is further configured to perform the following processing: when the abnormal influence factors of multiple abnormal prediction events overlap, the corresponding monitoring and acquisition device's acquisition strategy is configured based on the maximum value of the overlapping area of ​​the acquisition time window and acquisition frequency, and a mapping association is established between multiple abnormal prediction events and the acquisition strategies of the monitoring and acquisition device.

[0012] In one possible implementation, the AI-driven environmental monitoring and early warning method is further configured to perform the following processing: cleaning and preprocessing the abnormal monitoring probability data, extracting local temporal mutation features, long-term dependencies, and establishing spatial correlations of the sensor network to obtain the spatiotemporal features of the monitoring data; inputting the spatiotemporal features of the monitoring data into the abnormal identification and early warning model for abnormal feature identification and judgment to obtain abnormal event identification results; determining the risk level based on the abnormal event identification results; and using the abnormal event identification results and risk level as output results.

[0013] This application also provides a computer-readable storage medium, including: a computer program stored thereon, which, when executed by a processor, implements an AI-driven environmental monitoring and early warning method.

[0014] This application also provides an electronic device, including: a memory for storing executable instructions; and a processor for implementing an AI-driven environmental monitoring and early warning method when executing the executable instructions stored in the memory.

[0015] This application proposes an AI-driven environmental monitoring and early warning method, medium, and equipment to construct an anomaly trend prediction model. Based on monitoring data from furnace flues, it classifies abnormal events and predicts anomaly trends. Using the predicted anomaly events as a search engine, it searches an anomaly event analysis library to obtain a set of anomaly impact factors. It configures the acquisition strategy of the monitoring and acquisition equipment and obtains anomaly monitoring probability data. This probability data is then input into the anomaly identification and early warning model for anomaly event identification and risk level determination. Early warning information is generated based on the anomaly event identification results and risk levels. This addresses the technical problems in existing technologies, such as the inability to effectively cope with complex and ever-changing furnace flue environments and the inability to predict anomalies in advance, leading to insufficient accuracy and timeliness in environmental monitoring and early warning. It achieves intelligent environmental monitoring and early warning, improving the timeliness and accuracy of environmental monitoring and early warning. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings of the embodiments of the present invention will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a flowchart illustrating the AI-driven environmental monitoring and early warning method provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the process for constructing an abnormal trend prediction model in the AI-driven environmental monitoring and early warning method provided in the embodiments of this application.

[0019] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0020] Explanation of reference numerals in the attached drawings: Input device 401, processor 402, memory 403, output device 404. Detailed Implementation

[0021] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, product, or server that includes a series of steps is not necessarily limited to those steps explicitly listed, but may include other steps not explicitly listed or inherent to such processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0024] This application provides an AI-driven environmental monitoring and early warning method, such as... Figure 1 As shown, the method includes: Step S100: Construct an abnormal trend prediction model. Based on the monitoring data of the furnace flue, classify abnormal events and predict abnormal trends through the abnormal trend prediction model to obtain the probability of abnormal predicted events and abnormal trends.

[0025] Preferably, historical monitoring data of the furnace flue is acquired, including monitoring results of various sensors on temperature, pressure, flow rate, gas concentration, etc., with both normal and abnormal state data. The historical monitoring data is then preprocessed, including cleaning to remove noise, erroneous values, and missing values ​​to ensure data quality and accuracy. Training and testing sets are established using the features of the abnormal state data. A classification module is constructed based on a decision tree, and the classification module is trained and tested using the training and testing sets. A prediction module is constructed based on a Long Short-Term Memory (LSTM) network, and the prediction module is trained and tested using the temporal features of the abnormal state data. Finally, the classification module and the prediction module are integrated and connected to obtain an abnormal trend prediction model. Real-time monitoring data of the furnace flue is collected by sensors and input into a trained anomaly trend prediction model. The model categorizes anomalies in the monitoring data into different types, such as temperature anomalies, pressure anomalies, and flow rate anomalies. It also predicts the trend of these anomalies, such as whether temperature anomalies will continue to rise, remain stable, or gradually decrease, thus assessing the severity and development of the anomalies. After anomaly classification and trend prediction, the model outputs possible anomalies, i.e., predicted anomaly events. For example, it predicts that the furnace flue temperature will be too high in the future, and generates a corresponding probability for each predicted anomaly event as the anomaly trend probability, reflecting the likelihood of the anomaly event occurring. For example, the probability of a predicted excessively high furnace flue temperature occurring within the next 24 hours is 80%.

[0026] Furthermore, such as Figure 2 As shown, step S100 further includes step S110, collecting historical sample data of the furnace and kiln, including normal and abnormal states; step S120, comparing the furnace and kiln flue monitoring data corresponding to the normal and abnormal states to determine abnormal events and abnormal performance characteristics; step S130, establishing a training dataset based on the abnormal events and abnormal performance characteristics, and training a classifier using the training dataset; step S140, fitting time-series trend features based on the time-series sample data of the abnormal states, training a time-series network framework to obtain a trend probability prediction model; and step S150, integrating and connecting the classifier and the trend probability prediction model to obtain the abnormal trend prediction model.

[0027] Preferably, historical sample data of the furnace is collected, including temperature, pressure, flow rate, gas concentration, humidity, flame spectrum, vibration spectrum, etc., to reflect the operating status of the furnace from different perspectives. Among them, the historical sample data includes normal and abnormal states. Normal state data represents the performance of the furnace when it is running stably and without faults, while abnormal state data is a record of indicators when the furnace has faults, performance degradation, or other abnormal conditions. For example, excessively high temperature may indicate abnormal chemical reactions in the furnace or poor heat dissipation of the equipment. Changes in gas concentration can reflect whether combustion is complete or whether there are leaks. Flame spectrum can provide information about the composition of substances and energy release during combustion, and vibration spectrum can help detect whether the furnace structure is loose or worn.

[0028] Preferably, the monitoring data of the furnace flue gas duct under normal and abnormal conditions are compared from multiple dimensions such as the average value, fluctuation range, and rate of change. For example, the average temperature under normal and abnormal conditions is compared, and the fluctuation range of pressure is assessed to determine whether it exceeds the normal range. Through comparative analysis, the specific events that cause significant differences in the data are identified. For example, if the temperature suddenly rises sharply and the pressure also increases abnormally, the corresponding abnormal event may be local overheating or abnormal combustion in the furnace. If the gas concentration changes abnormally, it may be due to leakage or incomplete combustion. For each abnormal event, the characteristic changes in the monitoring data are extracted as abnormal manifestations. For example, for a local overheating event in the furnace, the abnormal manifestations may include a sharp rise in temperature at a specific location, uneven temperature distribution, and rapid increase in pressure. For a gas leakage event, it may be manifested as a sudden and continuous increase in gas concentration and an imbalance in the proportion of related gases.

[0029] Preferably, abnormal events during furnace operation are used as labels, and the corresponding abnormal performance features are used as input data to establish a training dataset. For example, a sample may be "temperature increases by 30°C, pressure increases by 15%, and specific gas concentration increases by 20%" corresponding to the "local overheating in the furnace" event. Then, a classification module is constructed based on a decision tree, and a classifier is obtained by training it using the training dataset. During the training process, the classifier continuously adjusts its parameters to learn the mapping relationship between abnormal performance features and abnormal events, thereby enabling accurate judgment of possible abnormal events based on new monitoring data.

[0030] Preferably, the monitoring data under abnormal conditions is time-varying sequence data. Sliding windows and differencing techniques are used to fit the time-series trend characteristics from the time-series sample data under abnormal conditions, capturing information such as data change trends (e.g., increases, decreases, fluctuations) and periodicity. For example, by calculating the difference between data at adjacent time points, the rate of change of indicators such as temperature and pressure can be observed, or the average value and standard deviation of data over a certain period can be statistically analyzed using sliding windows to understand the local change characteristics of the data. Then, a time-series network framework is constructed based on a Long Short-Term Memory (LSTM) network, and the time-series network framework is trained using time-series sample data under abnormal conditions. During training, the model learns the development patterns and trends of abnormal events, thereby enabling it to predict the probability and development trend of future abnormal events based on real-time monitoring data.

[0031] Preferably, the trained classifier and trend probability prediction model are integrated. First, the classifier is used to classify the current monitoring data and identify possible abnormal events. Then, for each possible abnormal event, the trend probability prediction model is used to predict its development trend and probability of occurrence. When new monitoring data is input, the integrated abnormal trend prediction model can simultaneously provide the classification results of the abnormal event, as well as the development trend and probability of occurrence of the abnormal event, providing users with comprehensive and accurate abnormal early warning information.

[0032] Furthermore, step S120 also includes step S121, constructing a time-series monitoring data chain based on the monitoring data of the furnace flue in the abnormal state, arranging them in the order of monitoring time, and performing preliminary tracing based on the identification nodes of the abnormal state in the time-series monitoring data chain to obtain preliminary data change information; step S122, determining the comparison time period based on the change time period of the preliminary data change information; step S123, extracting periodic data of the normal state based on the comparison time period to obtain normal state periodic data; step S124, aligning the normal state periodic data with the comparison time period data of the abnormal state to obtain difference data features, and using the difference data features as abnormal performance features; step S125, mapping and associating abnormal performance features according to the abnormal events corresponding to the abnormal state to determine the abnormal events and abnormal performance features.

[0033] Preferably, the monitoring data of the furnace flue in abnormal state are arranged in chronological order of monitoring time to form a continuous, time-series data chain, namely the time-series monitoring data chain. Then, based on the identification nodes of abnormal states in the time-series monitoring data chain, a preliminary tracing is performed. Specifically, the identification nodes of abnormal states are first determined, that is, specific data points that clearly show abnormalities. Then, starting from the identification node, the data changes before the abnormality occurred are traced back to obtain the preliminary data change information. For example, when the temperature suddenly rises to the abnormal value (identification node), the trend and fluctuation of data such as temperature, pressure, and flow rate in the period before this are traced back. Then, based on the time span from the time point when the data in the preliminary data change information begins to change to the identification node of the abnormal state, the comparison time period is determined, which reflects the time process from the beginning of the abnormality to the final formation of a clear abnormality. For example, if it takes 10 hours for the temperature to slowly rise to reach the abnormal high temperature value, then the comparison time period may be set to 10 hours.

[0034] Preferably, based on a determined comparison time period, monitoring data under normal conditions is extracted. Specifically, periodic data corresponding to the comparison time period is extracted from the normal state data, following the same time length and time interval, representing the characteristics of the furnace / kiln under the same time scale during normal operation. Then, the normal state periodic data is aligned with the abnormal state data within the comparison time period, i.e., the two data sequences are precisely matched in time. By comparison, the differences between the normal and abnormal state data within the same time period are determined and used as differential data features. For example, under normal conditions, the temperature fluctuates within a small range over 10 hours, while under abnormal conditions, the temperature rises or falls sharply within the same 10 hours; this temperature change difference is used as a differential data feature. Finally, the differential data features are mapped and associated with the abnormal events corresponding to the abnormal states to determine the abnormal performance characteristics corresponding to different abnormal events, thereby establishing the relationship between abnormal events and abnormal performance characteristics.

[0035] Furthermore, step S100 also includes step S160, which involves deploying monitoring sensors based on the characterization factors of the normal and abnormal states. The monitoring sensors include at least monitoring sensors corresponding to temperature, pressure, flow rate, gas concentration, humidity, flame spectrum, and vibration spectrum. Step S170 involves collecting monitoring data of the furnace flue in real time based on the deployed monitoring sensors.

[0036] Preferably, the characterization factors for normal and abnormal states refer to key factors that can reflect whether the furnace is in normal or abnormal operation. For example, during normal operation, the furnace's temperature, pressure, flow rate, gas concentration, humidity, flame spectrum, and vibration spectrum are all within specific ranges. However, when the furnace experiences abnormalities, such as local overheating, incomplete combustion, or loose equipment components, these factors will change accordingly. Then, considering the furnace's structure, operating characteristics, and possible locations of abnormalities, monitoring sensors are deployed according to the characterization factors of the furnace's operating state. For example, temperature sensors are strategically placed in areas prone to local overheating (such as near the burner); pressure and flow rate sensors are installed at bends and diameter changes in the flue, where airflow changes and blockages are likely to occur; and gas concentration sensors are installed at seals where gas leaks may occur.

[0037] Preferably, the monitoring sensors include at least sensors for temperature, pressure, flow rate, gas concentration, humidity, flame spectrum, and vibration spectrum. Specifically, the temperature sensor measures the temperature inside the furnace and at various points in the flue; temperature changes at different locations directly reflect the heat distribution within the furnace. The pressure sensor is installed in key parts of the flue and furnace to monitor pressure changes; abnormally high pressure indicates flue blockage, while excessively low pressure indicates leakage. The flow rate sensor measures the gas velocity within the flue; a decrease in velocity indicates obstruction by foreign objects or a fan malfunction. The gas concentration sensor detects the main gases involved in combustion within the furnace (such as oxygen, etc.). The system monitors changes in the concentration of carbon monoxide, sulfur dioxide, and other potentially leaking hazardous gases to determine whether combustion is complete and whether gas leaks exist. Humidity sensors monitor the humidity of gases inside the furnace and in the flue. Flame spectrum sensors detect the spectral characteristics of the flame to reflect the composition and combustion state of the substances during combustion; different fuels, combustion conditions, and abnormal situations (such as impurity contamination or unstable combustion) can cause changes in the flame spectrum. Vibration spectrum sensors are installed on key components of the furnace (such as the furnace body and fans) to monitor vibration during equipment operation; wear and loosening of equipment components can alter the vibration spectrum. Finally, based on the deployed monitoring sensors, the data from the furnace and flue are monitored in real time and transmitted to the central control unit via specific transmission methods (such as wired cables, wireless Bluetooth, Wi-Fi, or industrial wireless communication protocols), thus integrating the data into a comprehensive dataset reflecting the operating status of the furnace and flue.

[0038] Step S200: Using the abnormal prediction event as a search engine, search in the abnormal event analysis library to obtain the abnormal impact factor set, which includes abnormal impact factors and impact probability coefficients.

[0039] Preferably, the abnormal prediction event is used as a search engine, i.e., as a search keyword, to search in the abnormal event analysis database. The abnormal event analysis database stores various known furnace and kiln abnormal events, as well as the influencing factors and probability coefficients corresponding to different types of abnormal events. Among them, a large amount of historical data is collected, recording the occurrence of various abnormal events and corresponding influencing factors in the past. The probability coefficient of each abnormal event is determined by frequency statistics. Taking the abnormal rise in furnace and kiln temperature as an example, in the past 100 events of abnormal temperature rise, it was statistically determined that 70 were caused by excessive fuel supply. Therefore, the probability coefficient of the influencing factor "excessive fuel supply" on the abnormal event "abnormal rise in furnace and kiln temperature" is 70 ÷ 100 = 0.7.

[0040] Preferably, the search is performed in the abnormal event analysis database using the abnormal prediction event as the keyword. This involves comparing and matching records corresponding to the abnormal prediction event in the database, thus obtaining the set of abnormal influencing factors related to the abnormal event. Abnormal influencing factors refer to various factors that may lead to the abnormal event. For example, for an abnormal rise in furnace temperature, abnormal influencing factors may include excessive fuel supply, insufficient combustion air, and deterioration of furnace wall insulation. The influence probability coefficient is a quantitative representation of the likelihood that each abnormal influencing factor will cause the abnormal event to occur. It is usually determined based on statistical analysis of historical furnace data. For instance, through data analysis of multiple historical events of abnormal furnace temperature rises, it is found that excessive fuel supply has a high probability of occurrence, and a relatively high influence probability coefficient is calculated, such as 0.7; while the probability of deterioration of furnace wall insulation is relatively low, and a relatively low influence probability coefficient is calculated, such as 0.3. Thus, the set of abnormal influencing factors comprehensively describes the possible influencing factors of each abnormal event and the likelihood of these factors causing the abnormal event to occur.

[0041] Furthermore, step S200 also includes step S201, which involves analyzing relevant monitoring data and the influence coefficient of each monitoring data on the abnormal event based on historical sample data, according to the abnormal event and its abnormal performance characteristics; step S202, which involves establishing connection edges between nodes and branch nodes, with the weight of the connection edge being the influence coefficient, using the abnormal event as the top node and each relevant monitoring data as the branch node, to obtain the analytical graph structure of each abnormal event; and step S203, which involves combining the analytical graph structures of all abnormal events in parallel to construct the abnormal event analysis library.

[0042] Preferably, a large amount of historical sample data is collected, including various monitoring data of the furnace flue under different operating conditions, as well as whether corresponding abnormal events have occurred and the specific types of abnormal events. For each abnormal event, the relevant monitoring data is analyzed. For example, for abnormal rise in furnace temperature, the relevant monitoring data includes fuel flow rate, combustion air flow rate, furnace wall temperature, etc. The influence coefficient of each monitoring data on the occurrence of the abnormal event is calculated through statistical analysis, indicating the relative importance of the monitoring data in predicting the occurrence of the abnormal event. Then, the abnormal event is used as the top node, and the monitoring data related to the abnormal event are used as the branch nodes. For example, for the top node "abnormal rise in furnace temperature", its branch nodes may be "fuel flow rate monitoring data", "combustion air flow rate monitoring data", and "furnace wall temperature monitoring data". Then, the connection edges between the top node and the branch nodes are established using the influence probability coefficient, forming the analytical graph structure of each abnormal event. Finally, the analytical graph structures of all different abnormal events are combined side by side to form an abnormal event analysis library, which contains various possible abnormal events and their relationship with relevant monitoring data. Table 1 below shows the core data of the example abnormal event analysis library: Table 1. Probability Coefficients of Abnormal Events and Influencing Factors in Furnaces and Kilns Furnace abnormality events Abnormal Influencing Factors Influence probability coefficient abnormal rise in furnace temperature Excessive fuel supply 0.6 Insufficient combustion air 0.3 The furnace wall insulation effect deteriorated 0.1 Abnormal increase in flue gas pressure Duct blockage (ash, foreign objects) 0.7 Fan malfunction (abnormal speed) 0.2 The valve was not fully opened. 0.1 Abnormal gas concentration (excessive carbon monoxide). Incomplete combustion (improper air-to-fuel ratio) 0.8 Burner malfunction (reduced combustion efficiency) 0.15 Flue leak (damaged seal) 0.05 abnormally low flow rate Duct blockage (ash, foreign objects) 0.6 Degraded fan performance (blade wear) 0.3 Increased pipe resistance (smaller pipe diameter, too many bends) 0.1 Flame spectrum anomalies (color, intensity changes) Changes in fuel quality (increased impurity content) 0.5 Burner nozzle clogged or damaged 0.3 Abnormal Combustion Air Composition 0.2 Abnormal vibration spectrum (excessive equipment vibration) Equipment components are loose (bolts are loose). 0.4 Component wear (bearing wear) 0.4 Imbalance (rotor imbalance) 0.2 Step S300: Configure the acquisition strategy of the monitoring and acquisition device according to the abnormal prediction event and the probability of abnormal trend, as well as the abnormal impact factor and the impact probability coefficient, and obtain abnormal monitoring probability data based on the acquisition strategy of the monitoring and acquisition device.

[0043] Preferably, the acquisition strategy of monitoring and acquisition equipment is configured based on the probability of abnormal predicted events and abnormal trends, as well as the abnormal influencing factors and their probability coefficients. That is, for abnormal influencing factors that are related to high-probability abnormal predicted events and have a large probability coefficient, the corresponding monitoring and acquisition equipment should be configured with a higher acquisition frequency and accuracy. For example, excessive fuel supply is determined to be an important abnormal influencing factor leading to excessive furnace temperature, and the probability of this abnormal predicted event occurring is relatively high. Therefore, equipment such as flow sensors used to monitor fuel supply should be set with a higher acquisition frequency to obtain relevant data on fuel supply more timely and accurately, and to detect abnormal fuel supply situations in a timely manner. Conversely, for abnormal influencing factors that are related to low-probability abnormal predicted events and have a small probability coefficient, the acquisition frequency and accuracy can be set relatively low to save resources and reduce costs. Then, the monitoring and acquisition equipment collects real-time data on the relevant operating parameters and status of the furnace and kiln according to the configured acquisition strategy. For example, the flow sensor collects fuel flow data at a set frequency, and the temperature sensor collects furnace and kiln temperature data. Combined with the pre-determined abnormal influence factors and influence probability coefficients, the abnormal monitoring probability data of each abnormal prediction event is calculated, which represents the probability of each abnormal prediction event actually occurring under the current data collection conditions. If the calculated probability is high, it means that according to the current monitoring data, the furnace temperature is likely to be high or about to occur, and corresponding measures need to be taken in a timely manner for processing and intervention.

[0044] Furthermore, step S300 also includes step S310, obtaining the abnormal trend probability of the abnormal prediction event and configuring the collection time window, wherein the larger the abnormal trend probability, the shorter the collection time window; step S320, setting the collection frequency according to the collection time window and the influence probability coefficient of the abnormal influence factor; step S330, configuring the collection strategy of the monitoring and collection device corresponding to the abnormal influence factor according to the collection time window and the collection frequency, based on the mapping relationship between the abnormal prediction event and the abnormal influence factor.

[0045] Preferably, based on historical operating data and current operating status of the furnace flue, the corresponding abnormal trend probability is calculated for each abnormal prediction event, reflecting the likelihood of the abnormal event occurring in the future. Then, the data collection time window is determined based on the abnormal trend probability. The higher the abnormal trend probability, the more frequently data needs to be monitored in order to detect abnormalities in a timely manner, and the shorter the data collection time window is configured. For example, for an abnormal rise in furnace temperature with an abnormal trend probability of 70%, the data collection time window may be set to 1 hour, that is, data is collected once per hour. For another abnormal event with an abnormal trend probability of 30%, the data collection time window may be set to 4 hours.

[0046] Preferably, for each predicted abnormal event, the collection frequency is set according to the probability coefficient of the abnormal influencing factor based on the collection time window. Specifically, for abnormal influencing factors with a high probability coefficient, a higher collection frequency is set within the given collection time window. For example, if the furnace temperature rises abnormally and the fuel flow rate is identified as an important abnormal influencing factor with a high probability coefficient, fuel flow rate data may be collected every 10 minutes within a 1-hour collection time window. Conversely, for abnormal influencing factors with a low probability coefficient, the collection frequency is relatively low. For example, the probability coefficient of combustion air humidity affecting the abnormal rise in furnace temperature is low, so combustion air humidity data may be collected every 30 minutes within a 1-hour collection time window.

[0047] Preferably, the mapping relationship between abnormal prediction events and abnormal influencing factors is determined, that is, it is determined which abnormal influencing factors affect each abnormal prediction event. Then, according to the determined collection time window and collection frequency, a specific collection strategy is configured for the monitoring and collection equipment corresponding to each abnormal influencing factor. For example, for fuel flow, the corresponding collection time window is 1 hour and the collection frequency is once every 10 minutes. Then, the flow meter responsible for monitoring fuel flow is configured to collect data. Similarly, other abnormal influencing factors are also set with corresponding collection strategies according to their respective collection time windows and collection frequencies to ensure that various data related to abnormal prediction events can be obtained accurately and in a timely manner.

[0048] Furthermore, step S330 also includes configuring the acquisition strategy of the corresponding monitoring and acquisition device according to the maximum value of the overlapping area of ​​the acquisition time window and acquisition frequency when the abnormal influence factors of multiple abnormal prediction events overlap, and establishing a mapping association between multiple abnormal prediction events and the acquisition strategy of the monitoring and acquisition device.

[0049] Preferably, the abnormal influencing factors of different abnormal prediction events in the furnace flue may overlap, meaning that different abnormal prediction events have the same abnormal influencing factor. For example, the two abnormal prediction events, "excessively high furnace temperature" and "incomplete combustion," may both be affected by the two abnormal influencing factors, "fuel flow rate" and "air flow rate." Then, for each abnormal prediction event, based on its abnormal trend probability and the influence probability coefficient of the abnormal influencing factor, there is its own corresponding collection time window and collection frequency. When multiple abnormal prediction events correspond to the same abnormal influencing factor, the maximum value of the overlap area between the collection time window and the collection frequency is found. For the overlapping abnormal influencing factor "fuel flow rate," the overlap area of ​​the collection time window is 0.5 hours (0.5 hours is the overlap between 0.5 hours and 1 hour), and the collection frequency... In terms of frequency, the maximum overlap between once every 10 minutes and once every 15 minutes is once every 10 minutes (the frequency of once every 10 minutes is higher and can cover the collection points once every 15 minutes). Then, the collection strategy of the corresponding monitoring and acquisition device (such as fuel flow sensor) is configured according to the maximum value of the overlap area. Finally, the mapping relationship between the collection strategy of each monitoring and acquisition device and the abnormal prediction event is clarified. When the sensor collects data according to the set collection strategy, multiple related abnormal prediction events can be analyzed and judged at the same time based on the monitoring data, and possible abnormal situations can be detected in time. At the same time, for monitoring devices without abnormal association, the collection time window is expanded, and they are not focused on monitoring, but the monitoring of normal cycles must be ensured to avoid missing other abnormal states, thereby improving the monitoring and early warning capability of abnormal situations in the furnace flue.

[0050] Step S400: Input the abnormal monitoring probability data into the abnormal identification and early warning model to identify abnormal events and determine risk levels, and generate early warning information according to the abnormal event identification results and risk levels.

[0051] Preferably, the anomaly identification and early warning model is used to identify the existence of abnormal events and determine the risk level of abnormal events based on the input anomaly monitoring probability data. It acquires various monitoring data related to furnace operation, including but not limited to monitoring results from sensors such as temperature, pressure, flow rate, and gas composition, as well as information such as equipment operating status and operation records. The acquired monitoring data is then labeled, clearly defining normal operating status data, abnormal operating status data, and the type and risk level of the anomaly. This data is then preprocessed, including data cleaning and standardization, such as removing noise, missing values, and erroneous data. If the ratio of normal operating status data to abnormal operating status data in the dataset is unbalanced, oversampling or undersampling is used to adjust the data ratio, enabling the model to better learn the characteristics of abnormal data. Then, an early warning model is built based on machine learning models (such as support vector machines, random forests, etc.). The preprocessed data is divided into training and testing sets for training and evaluation of the early warning model, ultimately obtaining the anomaly identification and early warning model, which can receive monitoring data in real time and perform anomaly identification and early warning.

[0052] Preferably, the anomaly identification and early warning model analyzes and processes the input anomaly monitoring probability data, including analyzing the changes in various indicators in the data, such as temperature, pressure, and flow rate. If the parameter changes exceed the normal range or show an abnormal trend, such as a sudden and sharp rise in temperature or excessive pressure fluctuations, the anomaly identification and early warning model identifies the abnormal event and determines the risk level of the abnormal event based on the severity of the anomaly, the duration of the anomaly, and the scope of the anomaly's impact. For example, if the temperature in a small area slightly exceeds the normal range and the duration is short, it is determined to be at a low risk level (Level 1); if the pressure in the entire furnace flue suddenly drops significantly, it is determined to be at a high risk level (Level 3). Finally, based on the anomaly event identification results and risk level, corresponding early warning information is generated, specifically including the early warning level, triggering conditions, and response actions, as shown in Table 2, which provides an example of early warning information. Table 2. Details of Early Warning Classification for Furnaces and Kilns Warning Level Triggering conditions Response Action Level 1 (Yellow) The temperature in a localized area exceeds the preset threshold by 10%. Immediately notify professionals to conduct a thorough inspection of the relevant area. Level 2 (Orange) A gas leak was detected, accompanied by abnormal equipment vibration. Quickly activate the ventilation system to accelerate air circulation and reduce the concentration of harmful gases. Level 3 (Red) Temperature exceeds preset threshold by 30%, or concentration of harmful gases exceeds limit threshold. Immediately execute emergency shutdown procedures and organize personnel evacuation to ensure their safety. Furthermore, step S400 also includes step S410, cleaning and preprocessing the anomaly monitoring probability data, extracting local temporal mutation features, long-period dependencies, and establishing spatial correlation of the sensor network to obtain the spatiotemporal features of the monitoring data; step S420, inputting the spatiotemporal features of the monitoring data into the anomaly identification and early warning model for anomaly feature identification and judgment to obtain anomaly event identification results, determining the risk level based on the anomaly event identification results, and using the anomaly event identification results and risk level as output results.

[0053] Preferably, anomaly monitoring probability data may be affected by various interference factors, resulting in noise, erroneous values, or missing values. These data undergo cleaning and preprocessing to improve quality. For example, noise can be removed using filtering algorithms, and missing values ​​can be filled using interpolation or other methods, making the data more accurate and reliable. Specific algorithms (such as sliding window method, differential method, etc.) are used to detect and extract local temporal abrupt changes, such as sudden increases or decreases in temperature or pressure at a certain moment. Simultaneously, time series analysis methods (such as Fourier transform) are used to mine the long-term dependencies of the anomaly monitoring probability data, providing a more comprehensive understanding of the furnace flue's operating status. Then, the spatial correlation of the sensor network is established, meaning that the data collected by multiple sensors arranged in the furnace flue exhibit spatial correlation. For example, temperature sensor data from adjacent locations may have similarities, or pressure changes at one location may affect flow rates at other locations. Finally, the information such as local temporal abrupt changes, long-term dependencies, and the spatial correlation of the sensor network are integrated to obtain the spatiotemporal characteristics of the monitoring data, reflecting the changes in the data over time and space.

[0054] Preferably, the spatiotemporal characteristics of the acquired monitoring data are input into the anomaly identification and early warning model. The anomaly identification and early warning model compares and analyzes the spatiotemporal characteristics of the monitoring data with the learned normal and abnormal patterns. If the input characteristics differ significantly from the normal pattern and conform to the characteristics of certain abnormal patterns, then an anomaly is identified. Based on the identification of the anomaly characteristics, the model outputs the anomaly event identification result, i.e., whether an anomaly event has occurred and the type of the anomaly event, such as local overheating of the furnace or flue blockage. Then, based on the anomaly event identification result, combined with multiple factors such as the severity, scope of impact, and possible consequences of the anomaly event, the risk level of the anomaly event is further determined. For example, a minor local overheating is identified as a low-risk level, while a serious anomaly event that may lead to equipment damage or safety accidents is identified as a high-risk level. Finally, the anomaly event identification result and risk level are used as the final output to notify relevant personnel to take corresponding measures, such as conducting equipment inspections, adjusting operating parameters, or activating emergency plans, to ensure the safe and stable operation of the furnace.

[0055] Based on the foregoing embodiments, this application also provides an electronic device and a computer-readable storage medium storing a computer program. When the computer program is executed by the processor of the electronic device, it can implement the methods described in any of the preceding embodiments.

[0056] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention, showing a block diagram of an exemplary electronic device suitable for implementing the embodiments of the present invention. Figure 3The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention. This electronic device is in the form of a general-purpose computing device, and its components may include, but are not limited to, an input device 401, a processor 402, a memory 403, and an output device 404. The processor 402 may be one or more; the memory 403 may include a computer-readable medium and at least one program product having a set (at least one) of program modules configured to perform the functions of the embodiments of this application.

[0057] The memory 403 shown in this embodiment of the invention can be any combination of one or more computer-readable media. The computer-readable storage medium can be, but is not limited to, infrared, semiconductor systems, devices or components, or any combination thereof, for storing software programs, computer-executable programs and modules, such as the program instructions / modules corresponding to the AI-driven environmental monitoring and early warning method in this embodiment of the invention. The processor 402 executes various functional applications and data processing of the computer device by running the software programs, instructions and modules stored in the memory 403, thereby realizing the above-mentioned AI-driven environmental monitoring and early warning method.

[0058] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. An AI-driven environmental monitoring and early warning method, characterized in that, include: An abnormal trend prediction model is constructed. Based on the monitoring data of the furnace flue, the abnormal event is classified and the abnormal trend is predicted by the abnormal trend prediction model to obtain the abnormal predicted events and the probability of abnormal trends. Using the aforementioned abnormal prediction event as a search engine, the abnormal event analysis library is searched to obtain a set of abnormal impact factors, which includes abnormal impact factors and impact probability coefficients. Based on the probability of abnormal predicted events and abnormal trends, as well as the abnormal impact factors and impact probability coefficients, the acquisition strategy of the monitoring and acquisition equipment is configured, and abnormal monitoring probability data is obtained based on the acquisition strategy of the monitoring and acquisition equipment. The anomaly monitoring probability data is input into the anomaly identification and early warning model to identify anomalies and determine risk levels. Early warning information is then generated based on the anomaly identification results and risk levels.

2. The AI-driven environmental monitoring and early warning method according to claim 1, characterized in that, Constructing an anomaly trend prediction model, including: Collect historical sample data of the furnace and kiln, including normal and abnormal conditions; By comparing the furnace flue monitoring data corresponding to the normal and abnormal states, the abnormal events and their abnormal characteristics are determined. A training dataset is established based on the abnormal events and abnormal behavior characteristics, and a classifier is trained using the training dataset. By fitting time-series trend features to time-series sample data of abnormal states, a time-series network framework is trained to obtain a trend probability prediction model. The classifier is integrated with the trend probability prediction model to obtain the abnormal trend prediction model.

3. The AI-driven environmental monitoring and early warning method according to claim 2, characterized in that, By comparing the furnace flue monitoring data corresponding to the normal and abnormal states, abnormal events and their abnormal characteristics are determined, including: Based on the monitoring data of the furnace flue in the abnormal state, a time-series monitoring data chain is constructed according to the monitoring time sequence. Based on the identification nodes of the abnormal state in the time-series monitoring data chain, the previous data change information is obtained by tracing back in advance. The comparison time period is determined based on the time period of change in the preceding data. Based on the comparison time period, periodic data of the normal state is extracted to obtain normal state periodic data; Align the normal state periodic data with the abnormal state comparison periodic data to obtain the difference data features, and use the difference data features as abnormal performance features. Based on the abnormal events corresponding to the abnormal states, abnormal behavior features are mapped and associated to determine the abnormal events and their abnormal behavior features.

4. The AI-driven environmental monitoring and early warning method according to claim 2, characterized in that, Based on monitoring data from the furnace flue, abnormal events are classified using an abnormal trend prediction model, including: Based on the characterization factors of the normal and abnormal states, monitoring sensors are deployed, including at least monitoring sensors corresponding to temperature, pressure, flow rate, gas concentration, humidity, flame spectrum, and vibration spectrum. The monitoring data of the furnace flue is collected in real time based on the deployed monitoring sensors.

5. The AI-driven environmental monitoring and early warning method according to claim 4, characterized in that, Using the aforementioned anomaly prediction event as a search engine, a search is conducted in the anomaly event parsing library, prior to which the following is included: Based on the abnormal events and their abnormal characteristics, relevant monitoring data and the influence coefficients of each monitoring data on the abnormal events are analyzed based on historical sample data. Using abnormal events as the top node and relevant monitoring data as the branch nodes, establish connection edges between nodes and branch nodes. The weight of the connection edge is the influence coefficient, thus obtaining the analytical graph structure of each abnormal event. The exception event parsing library is constructed by combining the parsing graph structures of all exception events in parallel.

6. The AI-driven environmental monitoring and early warning method according to claim 1, characterized in that, Based on the predicted abnormal events and the probability of abnormal trends, as well as the abnormal impact factors and probability coefficients, the acquisition strategy of the monitoring and acquisition equipment is configured, including: Obtain the abnormal trend probability of the abnormal prediction event, and configure the collection time window, where the collection time window is shorter when the abnormal trend probability is higher. Based on the collection time window, the collection frequency is set according to the probability coefficient of the abnormal influencing factors. Based on the mapping relationship between the predicted abnormal events and the abnormal influencing factors, the acquisition strategy of the monitoring and acquisition devices corresponding to the abnormal influencing factors is configured according to the acquisition time window and acquisition frequency.

7. The AI-driven environmental monitoring and early warning method according to claim 6, characterized in that, Configure the acquisition strategy of the monitoring and acquisition equipment corresponding to the abnormal influencing factors according to the acquisition time window and acquisition frequency, including: When the abnormal influence factors of multiple abnormal prediction events overlap, the acquisition strategy of the corresponding monitoring and acquisition device is configured according to the maximum value of the overlapping area of ​​the acquisition time window and acquisition frequency, and a mapping association is established between multiple abnormal prediction events and the acquisition strategy of the monitoring and acquisition device.

8. The AI-driven environmental monitoring and early warning method according to claim 1, characterized in that, The anomaly monitoring probability data is input into the anomaly identification and early warning model to identify abnormal events and determine risk levels, including: The abnormal monitoring probability data is cleaned and preprocessed to extract local temporal mutation features, long-period dependencies, and establish spatial correlations of the sensor network to obtain the spatiotemporal characteristics of the monitoring data. The spatiotemporal characteristics of the monitoring data are input into the anomaly identification and early warning model for anomaly feature identification and judgment, anomaly event identification results are obtained, the risk level is determined based on the anomaly event identification results, and the anomaly event identification results and risk level are used as output results.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the AI-driven environmental monitoring and early warning method as described in any one of claims 1-8.

10. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; The processor, when executing executable instructions stored in the memory, implements the AI-driven environmental monitoring and early warning method according to any one of claims 1-8.

Images