Data processing method based on AI environment monitoring and server

Abstract

The invention provides a data processing method based on AI environment monitoring and a server, and relates to the technical field of artificial intelligence, and the method comprises the steps: firstly obtaining an environment data set which is collected by a plurality of sensor nodes in real time in a target monitoring region and comprises air quality, noise intensity and meteorological parameters, and dynamically dividing monitoring sub-regions according to the data space distribution characteristics; a data feature extraction rule is allocated to each sub-region, so that abnormal fluctuation indexes and region association features are extracted, a sub-region feature set is generated, a pre-trained dynamic anomaly analysis model is used for carrying out multi-dimensional association analysis on the abnormal fluctuation indexes, the types and influence ranges of environment abnormal events are determined, and an environment optimization strategy set is generated; and executing the environment optimization strategy set and sending a regulation and control instruction, and updating a model parameter weight according to feedback data, thereby realizing intelligent processing and analysis of environment data, effectively coping with environment abnormity, and improving the environment monitoring and treatment level.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a data processing method and server based on AI environment monitoring. Background Art

[0002] As people pay more and more attention to environmental quality, environmental monitoring plays an increasingly important role in many fields such as environmental protection, urban planning, and public health. Environmental monitoring aims to obtain environmental information in real time and accurately so as to detect environmental problems in a timely manner and take corresponding measures.

[0003] Existing methods mostly use simple threshold judgments or pre-set fixed analysis models to identify environmental anomalies. This processing method lacks in-depth exploration of the complex relationships of environmental data, and it is difficult to accurately distinguish different types of environmental anomalies and their impact ranges. For example, when faced with complex environmental problems caused by the interaction of multiple environmental factors, traditional methods may not be able to accurately determine the specific role and contribution of each factor in the abnormal event, resulting in the inability to formulate effective response strategies.

[0004] For the division of environmental monitoring areas, the existing technology usually adopts a fixed geographical area division method, which does not take into account the spatial distribution characteristics of the environmental data itself. This fixed division method cannot adapt to the diversity and variability of environmental data in different regions, so that the unified rules are used in the process of extracting data features in different regions, resulting in the extracted data features not accurately reflecting the actual environmental conditions of each region, reducing the accuracy and effectiveness of data analysis.

[0005] In addition, traditional environmental monitoring and treatment systems often lack a feedback mechanism for environmental regulation effects. After implementing environmental optimization strategies, it is impossible to adjust and optimize the monitoring and analysis models in real time according to the actual regulation effects, making it difficult for the system to continuously and effectively respond to changing environmental problems. Summary of the invention

[0006] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a data processing method based on AI environment monitoring, the method comprising: Acquire a set of environmental data collected in real time by multiple sensor nodes in a target monitoring area, wherein the set of environmental data includes air quality parameters, noise intensity parameters and meteorological parameters in a time series; According to the spatial distribution characteristics of the environmental data set, the target monitoring area is dynamically divided into a plurality of monitoring sub-areas, and a corresponding data feature extraction rule is assigned to each monitoring sub-area; Based on the data feature extraction rules, abnormal fluctuation indicators and regional correlation features are extracted from the environmental data of each monitoring sub-region to generate a sub-region feature set; Using the pre-trained dynamic anomaly analysis model, a multi-dimensional correlation analysis is performed on the abnormal fluctuation indicators in the sub-region feature set to determine the type and impact range of environmental abnormal events in the target monitoring area, and generate an environmental optimization strategy set; At least one environment optimization strategy in the environment optimization strategy set is executed, a control instruction is sent to a target execution device, and a parameter weight of the dynamic anomaly analysis model is updated based on feedback data of the control instruction.

[0007] On the other hand, an embodiment of the present invention further provides a server, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0008] Based on the above aspects, the embodiment of the present application first obtains the environmental data set containing air quality parameters, noise intensity parameters and meteorological parameters collected in real time by multiple sensor nodes in the target monitoring area, and then dynamically divides the monitoring sub-areas according to the spatial distribution characteristics of the environmental data set, and assigns corresponding data feature extraction rules to each sub-area, which can flexibly adapt to the differences in environmental data in different regions, can accurately capture the unique environmental change laws of each sub-area, and improve the accuracy and effectiveness of data feature extraction. Furthermore, based on targeted data feature extraction rules, abnormal fluctuation indicators and regional correlation features are extracted from the environmental data of each monitoring sub-area and a sub-area feature set is generated, which can not only keenly perceive the abnormal fluctuations in environmental data, but also deeply explore the potential correlation between different regions. Then, the pre-trained dynamic anomaly analysis model is used to perform multi-dimensional correlation analysis on the abnormal fluctuation indicators, which can comprehensively consider the impact of multiple factors on environmental anomalies, not only limited to simple causal judgment, but also analyze the causes and development trends of environmental abnormal events from multiple angles and multiple levels, so as to accurately determine the types and scope of environmental abnormal events in the target monitoring area, and the environmental optimization strategy set generated on this basis improves the pertinence and effectiveness of environmental governance and optimization. Finally, at least one environmental optimization strategy in the environmental optimization strategy set is executed, and the parameter weights of the dynamic anomaly analysis model are updated based on the feedback data of the control instructions. The analysis model can be adjusted in real time according to the actual control effect, and the model's analysis and prediction capabilities for environmental anomalies can be continuously improved, so that the entire environmental monitoring and optimization system can continuously adapt to complex and changeable environmental conditions and achieve long-term stable and efficient operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a schematic diagram of the execution flow of the data processing method based on AI environment monitoring provided by an embodiment of the present invention.

[0010] Figure 2 It is a schematic diagram of the server hardware architecture provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0011] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a data processing method based on AI environment monitoring provided by an embodiment of the present invention. The data processing method based on AI environment monitoring is introduced in detail below.

[0012] Step S110, obtaining a set of environmental data collected in real time by multiple sensor nodes in a target monitoring area, wherein the set of environmental data includes air quality parameters, noise intensity parameters and meteorological parameters in a time series.

[0013] In this embodiment, the target monitoring area is an industrial park as an example, and the industrial park is distributed with many factory workshops, warehouses, office buildings and some green areas, etc. In the industrial park, multiple sensor nodes are set up, and each sensor node can be installed in different positions, for example, near the exhaust port of the factory workshop, at the entrance of the warehouse, on the roof of the office building, and in the middle of the green area, etc.

[0014] Each sensor node is used to collect environmental data in real time. Taking air quality parameters as an example, the sensor node can detect the concentration of various pollutants in the air, such as sulfur dioxide (SO 2 ), nitrogen oxides (NOₓ), particulate matter (PM2.5, PM10), etc., and record data at a certain time interval, such as every 5 minutes, to form a time series. In terms of noise intensity parameters, sensor nodes can measure the noise decibel value of the surrounding environment, and meteorological parameters also change over time, reflecting the impact of production activities and traffic flow in different periods of time in the park on the noise environment. Meteorological parameters include temperature, humidity, wind direction, wind speed, etc. For example, wind direction and wind speed will affect the propagation direction and speed of pollutants emitted by factories in the park. Therefore, through the continuous collection of data by these sensor nodes, an environmental data set containing time series of air quality parameters, noise intensity parameters and meteorological parameters is obtained.

[0015] Step S120: dynamically divide the target monitoring area into a plurality of monitoring sub-areas according to the spatial distribution characteristics of the environmental data set, and assign a corresponding data feature extraction rule to each monitoring sub-area.

[0016] In the above industrial park scenario, the location coordinates and parameter change gradient of each sensor node are first extracted from the collected environmental data set. For example, the air quality parameter change gradient of the sensor node located near the exhaust port of the factory workshop may be large, especially during the peak production period when pollutant emissions increase, while the air quality parameter change gradient of the sensor node in the green area is relatively small. Therefore, the spatial distribution matrix can be constructed based on the above information.

[0017] Then, the density cluster analysis of the spatial distribution matrix is ​​performed to identify the parameter mutation boundary within the industrial park. For example, around a newly put into production high-pollution workshop, the air quality parameters suddenly deteriorate, thus forming a parameter mutation boundary. At the same time, the steady-state area is determined. For example, the air quality parameters in the green area are relatively stable and belong to the steady-state area.

[0018] Based on the geometric shape of the parameter mutation boundary (assuming that the mutation boundary around this new workshop is approximately circular, because the workshop is a relatively regular building and pollutants diffuse outward from the workshop as the center) and the parameter similarity of the steady-state area (the air quality parameters detected by each sensor node in the green area are highly similar), a dynamic grid segmentation scheme covering the entire industrial park is generated. Continuous grid cells with parameter similarity higher than the preset threshold (assuming that this threshold is set to 90%) in the dynamic grid segmentation scheme are merged into the same monitoring sub-area. For example, the grid cells of several adjacent green areas are merged into one monitoring sub-area, marked as "green area monitoring sub-area". For each monitoring sub-area, label the spatial attribute labels, such as "high pollution risk area around the workshop" and "low pollution area in the office area".

[0019] According to the type of environmental parameter corresponding to the spatial attribute label, the noise filtering threshold and feature fusion weight in the data feature extraction rule are matched from the preset rule library. For example, for the "high pollution risk area around the workshop", since the air quality parameters here fluctuate greatly and the noise mainly comes from the workshop equipment, the noise filtering threshold is set relatively high to avoid misjudging the noise generated by the normal operation of the workshop as abnormal; and for the "low pollution area of ​​the office area", the noise filtering threshold can be relatively low, because it is relatively quiet here, and smaller noise changes may need attention. At the same time, according to different types of environmental parameters, the feature fusion weight is determined. For example, in air quality analysis, for areas with high pollutant concentrations, certain key pollutants (such as SO 2 ) may have a higher weight.

[0020] Step S130: based on the data feature extraction rule, extract abnormal fluctuation indicators and regional correlation features from the environmental data of each monitoring sub-region to generate a sub-region feature set.

[0021] Taking the monitoring sub-area of ​​“high pollution risk area around workshop” as an example, the time series environmental data of air quality parameters, noise intensity parameters and meteorological parameters collected by the sensor nodes here are first processed by sliding window standardization to generate a normalized data sequence. 2 The concentration data are,normalized within a sliding window (assumed to be 1 hour) to make the data from,different sensor nodes comparable.

[0022] According to the noise filtering threshold in the data feature extraction rule, the normalized data sequence is subjected to abnormal point removal and data interpolation. Assuming that the noise filtering threshold of this monitoring sub-area is set to a higher value, when there is a short-term high-noise data point (which may be abnormal noise caused by a short-term equipment failure), if the high-noise data point exceeds the normal fluctuation range and does not conform to the noise characteristics of normal operation of the workshop, it will be removed and interpolated based on the previous and next data to generate a smooth data sequence.

[0023] Calculate the parameter change rate of adjacent time windows in the smoothed data series, for example, calculate the SO in two adjacent 1-hour windows 2 The change rate of concentration. When the change rate of this parameter exceeds the upper limit of the historical fluctuation range, the corresponding time window is marked as an abnormal fluctuation interval. If the production process of the workshop changes or the equipment fails during a certain period of time, SO 2 This abnormal fluctuation range will occur when emissions suddenly increase.

[0024] Extract parameter peaks (such as SO 2 The abnormal fluctuation index is generated by analyzing the maximum value of the concentration), duration (how long this abnormally high concentration lasts) and spatial propagation direction (the diffusion direction of pollutants is determined based on changes in wind direction and surrounding sensor node data).

[0025] The propagation direction of all abnormal fluctuation indicators in the same monitoring sub-area is matched with the abnormal fluctuation indicators of the adjacent sub-area for similarity. For example, pollutants in the high-pollution risk area around the workshop may spread to the adjacent office area. By analyzing the propagation direction and change trend of the abnormal fluctuation indicators in the two areas, a cross-regional association path is generated, and its topological structure is used as the regional association feature. If it is found that the pollutants emitted by the workshop have a tendency to spread to the office area, then the cross-regional association path reflects this potential pollution propagation relationship.

[0026] Step S140, using the pre-trained dynamic anomaly analysis model, performs multi-dimensional correlation analysis on the abnormal fluctuation indicators in the sub-region feature set, determines the type and impact range of environmental abnormal events in the target monitoring area, and generates an environmental optimization strategy set.

[0027] In the industrial park scenario, the pre-trained dynamic anomaly analysis model is used to match the parameter peaks of abnormal fluctuation indicators obtained from monitoring sub-areas such as the "high pollution risk area around the workshop" with the historical environmental event database. Assuming that the SO 2 The peak concentration was very high, similar to the pattern of high pollution events caused by equipment failures that had occurred in history. Through matching, the potential pollution source type was identified as workshop equipment failure, as well as the corresponding pollution diffusion model (such as a model centered on the workshop and diffusing according to wind direction and wind speed).

[0028] According to the cross-regional association path in the regional association characteristics, the propagation speed and attenuation coefficient of the pollution diffusion model in the target monitoring area (industrial park) are calculated. For example, if pollutants are found to diffuse from the workshop to the office area, the propagation speed is calculated to be several meters per hour based on the distance between the two areas, wind direction and speed, and changes in pollutant concentration. At the same time, due to the dilution of the air, the pollutant concentration will gradually decay, and the attenuation coefficient is calculated.

[0029] Combined with the environmental parameter types corresponding to the spatial attribute labels of the monitoring sub-areas, the degree of superposition of the pollution diffusion model on the air quality parameters and meteorological parameters is predicted. For the "high pollution risk area around the workshop" and the "low pollution area in the office area", the impact of pollutants emitted by the workshop on the air quality of the office area is predicted by taking into account meteorological parameters such as wind direction and wind speed. 2 How long will the concentration rise and to what extent, and what slight impact it may have on meteorological parameters such as temperature and humidity within the park.

[0030] Based on the superposition impact exceeding the preset safety threshold (such as office area SO 2 The spatial coordinates of the monitoring sub-areas where the concentration exceeds a certain standard are generated to generate an environmental abnormality event type with an emergency control priority. If the pollution level in the office area is about to exceed the safety standard, it is defined as a high-urgency environmental abnormality event, and the event type is the spread of workshop pollution to the office area.

[0031] According to the propagation speed and attenuation coefficient of the pollution diffusion model, the impact range boundary and duration prediction value of the environmental abnormality event are determined. For example, it is predicted which areas of the office area the pollutants will spread to in the next few hours and how long the entire pollution event may last.

[0032] According to the pollution source type (workshop equipment failure) corresponding to the type of environmental abnormal event, the type of target execution equipment and the range of control parameters are selected from the equipment control strategy library. For example, for this kind of workshop pollution event, the target execution equipment may be the exhaust gas treatment equipment in the workshop, and the range of control parameters may be to increase the power of the treatment equipment, adjust the amount of treatment agent, etc. Based on the spatial coordinates of the boundary of the impact range (the affected area of ​​the office area) and the emergency control priority (high urgency), each target execution equipment is assigned a start time sequence (such as starting the exhaust gas treatment equipment immediately) and a parameter adjustment gradient (such as how much power will be increased within how long). Combined with the duration prediction value and the wind direction and speed data in the meteorological parameters, the expected coverage rate of the parameter adjustment gradient of the target execution equipment within the impact range is calculated. If the expected coverage rate is lower than the preset efficiency threshold (such as the processing capacity of the exhaust gas treatment equipment cannot control the pollution in the office area within the specified time within a safe range), add the location deployment suggestions of the auxiliary control equipment (such as adding air purification equipment near the office area) and the collaborative working mode (such as how the purification equipment and the exhaust gas treatment equipment work together) to the target execution equipment. The control parameter ranges, startup time sequences, and auxiliary control device deployment recommendations of all target execution devices are encapsulated into an executable instruction set to generate an environment optimization strategy set.

[0033] Step S150, executing at least one environment optimization strategy in the environment optimization strategy set, sending a control instruction to a target execution device, and updating a parameter weight of the dynamic anomaly analysis model based on feedback data of the control instruction.

[0034] For the previously generated environmental optimization strategy set, take the workshop exhaust gas treatment equipment as an example. According to the type of the target execution equipment (exhaust gas treatment equipment), the executable instruction set is converted into an instruction format compatible with the equipment control protocol. For example, the instructions for regulating the power and dosage of the exhaust gas treatment equipment are converted into a specific format that the equipment can recognize.

[0035] Send an initialization command to the target execution device according to the startup time sequence, and receive a status confirmation signal returned by the target execution device. Assume that after sending a command to start the exhaust gas treatment device, the device returns a signal indicating that it is ready.

[0036] When the status confirmation signal indicates that the target execution device is ready, the control instructions are sent in stages according to the parameter adjustment gradient, and the actual parameter values ​​fed back by the target execution device are monitored in real time. For example, the power of the exhaust gas treatment device is gradually increased according to the set gradient, and the actual power value fed back by the device, the concentration of pollutants after treatment and other actual parameter values ​​are obtained in real time.

[0037] The deviation between the actual parameter value and the parameter target value corresponding to the expected coverage rate is calculated. If the deviation exceeds the fault tolerance range, the coordination instruction of the auxiliary control equipment is triggered. If the actual treatment effect of the exhaust gas treatment equipment fails to meet expectations, that is, the actual reduction in pollutant concentration is less than expected, then the coordination instruction of the air purification equipment near the office area is triggered.

[0038] Record the execution timestamp of the control command, the equipment response delay and the number of parameter calibrations, and generate equipment execution performance reports. For example, record the time delay from when the exhaust gas treatment equipment receives the command to when it starts to adjust the power, and the number of times the parameter is calibrated (such as adjusting the amount of reagent) to achieve a better treatment effect.

[0039] Extract the correlation between the number of parameter calibrations and the expected coverage from the equipment execution performance report, and construct a strategy execution effect evaluation matrix. Calculate the time prediction error coefficient of the dynamic anomaly analysis model based on the difference between the actual duration of the environmental anomaly event and the predicted value. Assuming that the pollution event was originally predicted to last for 3 hours, but actually lasted for 4 hours, calculate the corresponding error coefficient.

[0040] The strategy execution effect evaluation matrix and the time prediction error coefficient are input into the model weight optimizer to generate the parameter weight adjustment gradient. The back propagation algorithm is used to iteratively update the weights of the fully connected layer of the dynamic anomaly analysis model until the time prediction error coefficient drops below the convergence threshold. The updated model parameter weights are synchronized to the dynamic anomaly analysis model copies corresponding to all monitoring sub-areas, so that environmental anomalies can be more accurately predicted and responded to in subsequent environmental monitoring and analysis.

[0041] Based on the above steps, the embodiment of the present application first obtains the environmental data set containing air quality parameters, noise intensity parameters and meteorological parameters collected in real time by multiple sensor nodes in the target monitoring area, and then dynamically divides the monitoring sub-areas according to the spatial distribution characteristics of the environmental data set, and assigns corresponding data feature extraction rules to each sub-area, which can flexibly adapt to the differences in environmental data in different regions, can accurately capture the unique environmental change laws of each sub-area, and improve the accuracy and effectiveness of data feature extraction. Furthermore, based on targeted data feature extraction rules, abnormal fluctuation indicators and regional correlation features are extracted from the environmental data of each monitoring sub-area and a sub-area feature set is generated, which can not only keenly perceive the abnormal fluctuations in environmental data, but also deeply explore the potential correlation between different regions. Then, the pre-trained dynamic anomaly analysis model is used to perform multi-dimensional correlation analysis on the abnormal fluctuation indicators, which can comprehensively consider the impact of multiple factors on environmental anomalies, not only limited to simple causal judgment, but also analyze the causes and development trends of environmental abnormal events from multiple angles and multiple levels, so as to accurately determine the types and scope of environmental abnormal events in the target monitoring area, and the environmental optimization strategy set generated on this basis improves the pertinence and effectiveness of environmental governance and optimization. Finally, at least one environmental optimization strategy in the environmental optimization strategy set is executed, and the parameter weights of the dynamic anomaly analysis model are updated based on the feedback data of the control instructions. The analysis model can be adjusted in real time according to the actual control effect, and the model's analysis and prediction capabilities for environmental anomalies can be continuously improved, so that the entire environmental monitoring and optimization system can continuously adapt to complex and changeable environmental conditions and achieve long-term stable and efficient operation.

[0042] In a possible implementation, step S120 includes: Step S121, extracting the position coordinates and parameter change gradient of each sensor node from the environmental data set, and constructing a spatial distribution matrix.

[0043] In the industrial park, sensor nodes are distributed in various key locations, and each sensor node has its precise location coordinates. For example, a sensor located near the exhaust port of workshop A has coordinates (x1, y1, z1), and another sensor located at the entrance of warehouse B has coordinates (x2, y2, z2), etc. The parameter change gradient reflects the spatial variation trend of environmental parameters. Taking the sulfur dioxide concentration in the air quality parameter as an example, the sulfur dioxide concentration of the sensor near the exhaust port decreases with increasing distance due to its proximity to the pollution source. The rate of change of this concentration with distance is the parameter change gradient. By integrating the location coordinates and parameter change gradients of all sensor nodes, a spatial distribution matrix is ​​constructed, which comprehensively reflects the spatial distribution of environmental parameters in the industrial park.

[0044] Step S122, performing density cluster analysis on the spatial distribution matrix to identify parameter mutation boundaries and steady-state regions within the target monitoring area.

[0045] Due to differences in production activities and environmental conditions in different areas, there are parameter mutation boundaries and steady-state areas. For example, around a newly commissioned high-pollution workshop, air quality parameters will mutate due to concentrated emissions of pollutants. Through density cluster analysis, such areas can be clearly demarcated. The air quality parameters detected by sensors around the workshop are significantly different from those in more distant areas, and the edge of this difference forms a parameter mutation boundary. In the green area, the air quality parameters are relatively stable, and the data detected by each sensor fluctuates less, so this is identified as a steady-state area.

[0046] Step S123, generating a dynamic grid segmentation scheme covering the target monitoring area according to the geometric shape of the parameter mutation boundary and the parameter similarity of the steady-state area.

[0047] Step S124: merge the continuous grid units in the dynamic grid segmentation scheme whose parameter similarity is higher than a preset threshold into the same monitoring sub-region, and annotate each monitoring sub-region with a spatial attribute label.

[0048] Assuming the preset threshold is 90%, if the similarity of the air quality parameters, noise intensity parameters and meteorological parameters of several adjacent grid cells is higher than 90%, and these grid cells are continuous, then they are merged into one monitoring sub-area. For example, in the green area, multiple continuous grid cells with high parameter similarity are merged into one monitoring sub-area, marked as "green area monitoring sub-area". For areas with heavy pollution and similar parameters around workshops, they are merged and marked as "high pollution monitoring sub-area around workshops".

[0049] Step S125 , matching the noise filtering threshold and feature fusion weight in the data feature extraction rule from a preset rule library according to the environmental parameter type corresponding to the spatial attribute tag.

[0050] For example, for the "high-pollution monitoring sub-area around the workshop", since the production activities here generate more noise and the air quality parameters fluctuate greatly, a higher noise filtering threshold is matched from the preset rule base to avoid misjudging the noise generated by normal production activities as abnormal conditions. At the same time, for air quality analysis, the feature fusion weight is determined according to the pollution characteristics of the area. For example, a higher weight is given to the detection data of major pollutants such as sulfur dioxide and nitrogen oxides, because these pollutants are emitted more in the workshop production process and have a greater impact on the environment. For the "green area monitoring sub-area", since the environment is relatively quiet and the air quality is good, a lower noise filtering threshold is matched, and in the meteorological parameter analysis, the weights of parameters such as temperature and humidity may be relatively high, because the microclimate in the green area is greatly affected by these meteorological parameters. As a result, appropriate data feature extraction rules are assigned to each monitoring sub-area to facilitate more accurate analysis and processing of environmental data in the future.

[0051] Wherein, step S123 includes: Step S1231, extracting a spatial coordinate sequence of boundary points from the parameter mutation boundary, and identifying a geometric shape type of the boundary line according to the continuity of the spatial coordinate sequence.

[0052] Step S1232: dividing the boundary line into a plurality of geometric feature segments according to the curvature change feature of the geometric shape type, and marking the curvature change direction for each geometric feature segment.

[0053] Step S1233, based on the curvature change direction, initial grid units are generated on both sides of the boundary line, and the side lengths of the initial grid units are matched with the parameter similarity of the steady-state region to determine the side length adjustment ratio of the initial grid units.

[0054] Step S1234: expanding or contracting the initial grid unit according to the side length adjustment ratio to generate a boundary grid unit that is adapted to the geometric shape of the boundary line.

[0055] Step S1235 , extracting the sensor node group with the highest parameter similarity from the steady-state region, and determining the grid unit division density of the steady-state region according to the spatial distribution density of the sensor node group.

[0056] Step S1236, based on the grid unit division density, generating uniformly distributed steady-state grid units in the steady-state region, and aligning the side lengths of the steady-state grid units with the side lengths of the boundary grid units to generate transitional grid units.

[0057] Step S1237, according to the changing trend of the side length of the transitional grid unit, the boundary grid unit and the steady-state grid unit are seamlessly connected to generate a dynamic grid segmentation solution covering the target monitoring area.

[0058] Assuming that the parameter mutation boundary around a highly polluted workshop has a relatively regular circular feature in its spatial coordinate sequence of boundary points, this is because the structure of the workshop is relatively regular and the pollutants diffuse around the workshop. Then, according to the curvature change characteristics of the geometric shape type of this circle, the boundary line is divided into multiple geometric feature segments, and the curvature change direction is marked for each geometric feature segment. For example, on the circular boundary, the circumference is divided into several segments according to the angle, and each segment is marked with whether the curvature change direction is inward or outward. Based on the curvature change direction, the initial grid cells are generated on both sides of the boundary line. The side length of the initial grid cell needs to be matched with the parameter similarity of the steady-state area to determine the side length adjustment ratio. Assuming that the parameter similarity of the steady-state area is high, it means that the environmental parameters of the area are relatively stable. If the side length of the initial grid cell is too large, some local minor changes may be ignored. Therefore, the side length needs to be adjusted according to the parameter situation of the steady-state area. If the sensor nodes in the steady-state area are densely distributed and the parameter similarity is high, the side length of the initial grid cell may need to be reduced. According to the side length adjustment ratio, the initial grid cell is expanded or contracted to generate a boundary grid cell that matches the geometric shape of the boundary line.

[0059] The sensor node group with the highest parameter similarity is extracted from the steady-state area, and the grid unit division density of the steady-state area is determined according to the spatial distribution density of the sensor node group. In the steady-state area of ​​the green area, the sensor node group with the highest similarity in detecting air quality parameters, noise intensity parameters and meteorological parameters is found. If these sensor nodes are densely distributed, it means that the environmental conditions in the area do not change much within a small range. Then the grid unit division density of the steady-state area can be appropriately increased, which can not only ensure effective monitoring of the environment but also reduce the amount of calculation. Based on this grid unit division density, uniformly distributed steady-state grid cells are generated in the steady-state area, and the side lengths of the steady-state grid cells are aligned with the side lengths of the boundary grid cells to generate transition grid cells. For example, at the junction of the green area and the area surrounding the workshop, the side lengths of the steady-state grid cells and the boundary grid cells are adjusted so that the two can transition smoothly and generate transition grid cells. According to the changing trend of the side length of the transitional grid cells, the boundary grid cells are seamlessly connected with the steady-state grid cells to generate a dynamic grid segmentation scheme covering the entire industrial park. This dynamic grid segmentation scheme can accurately reflect the environmental characteristics of different areas within the industrial park, taking into account both the complex situation of the mutation area and the relative stability of the steady-state area.

[0060] Step S1238, mapping the spatial coordinate sequences of the boundary grid cells, transition grid cells and steady-state grid cells in the dynamic grid segmentation scheme with the parameter similarities of the parameter mutation boundary and the steady-state region to generate grid cell attribute labels.

[0061] For example, since the boundary grid cell is close to the parameter mutation boundary, its attribute label may contain information such as "high pollution risk boundary", and at the same time mark its degree of association with the parameter mutation boundary and the internal parameter similarity range. The attribute label of the transition grid cell reflects its characteristics in the transition between different regions, such as "transition from high pollution area to low pollution area" and the corresponding parameter changes. The attribute label of the steady-state grid cell reflects the environmental parameter characteristics of the steady-state area where it is located, such as "low pollution steady state in green area".

[0062] Step S1239, based on the grid cell attribute labels, the grid cells in the dynamic grid segmentation scheme are clustered according to parameter similarity to generate multiple grid cell groups, and a spatial attribute label is annotated for each grid cell group. Based on the spatial attribute label, the grid cell groups in the dynamic grid segmentation scheme are matched with the spatial distribution characteristics of the target monitoring area to generate a dynamic grid segmentation scheme covering the target monitoring area.

[0063] In this embodiment, grid cells with similar parameter similarity can be grouped together. For example, all grid cells located in high pollution risk areas around workshops and with high parameter similarity can be grouped together and labeled as "high pollution risk group around workshops". The grid cells in this group are similar in terms of the changing trend of environmental parameters, pollution degree, etc. Other groups, such as the low pollution group in the office area, are also labeled according to their respective characteristics.

[0064] Then, each grid unit group can be matched to the actual area in the industrial park to ensure that the grid segmentation scheme can accurately reflect the actual spatial distribution of the industrial park. For example, the grid units of the high pollution risk group around the workshop cover all areas around the workshop that may be affected by pollution, and the grid units of the low pollution group in the office area accurately correspond to the office area, etc. In this way, a complete dynamic grid segmentation scheme that conforms to the environmental characteristics of the industrial park is generated.

[0065] In a possible implementation, step S130 includes: Step S131, performing sliding window normalization processing on the time series environmental data of all sensor nodes in the monitoring sub-area to generate a normalized data sequence.

[0066] Taking the "high pollution monitoring sub-area around the workshop" in the industrial park as an example, each sensor node in this sub-area continuously collects environmental data such as air quality parameters, noise intensity parameters and meteorological parameters. When processing these time series environmental data, the sliding window standardization processing method is adopted. For example, for the sulfur dioxide concentration data in the air quality parameters, a sliding window time length is set to one hour. Within this hour, the sulfur dioxide concentration values ​​detected by each sensor node are collected, and these concentration values ​​of different magnitudes are converted into a normalized data sequence through a specific standardization algorithm. The normalized data sequence makes the data of different sensor nodes comparable, which is convenient for subsequent analysis and processing. Similarly, for noise intensity parameters and meteorological parameters, such as noise decibel value, temperature, humidity and other data, the same sliding window standardization processing is also used to obtain a complete normalized data sequence in the entire monitoring sub-area.

[0067] Step S132, according to the noise filtering threshold in the data feature extraction rule, outlier removal and data interpolation are performed on the normalized data sequence to generate a smoothed data sequence.

[0068] In the "high pollution monitoring sub-area around the workshop", due to the certain regularity of the noise and pollutant emissions generated by the workshop production activities, the corresponding noise filtering threshold is set in the data feature extraction rules. For example, for the noise intensity parameter, if the noise decibel value detected at a certain moment suddenly exceeds the noise range that may be generated by normal production activities, and this exceeded value is greater than the noise filtering threshold, this data point is determined as an abnormal point. In the sulfur dioxide concentration data, if a sudden extremely high value appears and does not conform to the fluctuation range of normal workshop emissions, it is also determined as an abnormal point based on the noise filtering threshold. For these abnormal points, they are eliminated. Then, data interpolation is performed based on the previous and next data to ensure the integrity of the data sequence. As a result, the air quality parameters, noise intensity parameters and meteorological parameters in the entire normalized data sequence are all subjected to abnormal point elimination and data interpolation, and finally a smooth data sequence is generated, which can more accurately reflect the real change trend of environmental parameters in the monitoring sub-area.

[0069] Step S133, calculating the parameter change rate of adjacent time windows in the smoothed data sequence, and marking the time window where the parameter change rate exceeds the upper limit of the historical fluctuation range as an abnormal fluctuation interval.

[0070] When processing the smoothed data sequence of the "high pollution monitoring sub-area around the workshop", for each environmental parameter, the parameter change rate of adjacent time windows is calculated. For example, for sulfur dioxide concentration, compare the concentration change rate in adjacent one-hour sliding windows. Calculate the concentration value of this hour with the concentration value of the previous hour to get the concentration change ratio. Referring to the fluctuation range of sulfur dioxide concentration in this sub-area in historical data, if the parameter change rate of a certain adjacent time window exceeds the upper limit of the historical fluctuation range, the time window is marked as an abnormal fluctuation interval. The same method is used to calculate and mark noise intensity parameters and meteorological parameters. If the production equipment in the workshop fails or the production process is temporarily adjusted, it may cause a sudden change in the pollutant emission rate or a sudden increase in noise intensity. These situations will be reflected in the parameter change rate and thus be marked as an abnormal fluctuation interval.

[0071] Step S134, extracting the parameter peak value, duration and spatial propagation direction within the abnormal fluctuation interval to generate the abnormal fluctuation index.

[0072] In the marked abnormal fluctuation interval, taking sulfur dioxide concentration as an example, extract the maximum value of sulfur dioxide concentration in the abnormal fluctuation interval, which is the parameter peak value. At the same time, determine the duration of this abnormal fluctuation, that is, the time from the concentration starting to exceed the normal fluctuation range to the recovery to the normal range. For the determination of the spatial propagation direction, it is judged in combination with the position of each sensor node in the sub-area and the concentration change trend. If the sensor near the exhaust port of the workshop first detects an increase in sulfur dioxide concentration, and then the downwind sensor also detects an increase in concentration in turn along the wind direction, then it can be determined that the spatial propagation direction of the pollutant is to diffuse from the exhaust port to the downwind direction along the wind direction. For the abnormal fluctuation interval of noise intensity, also extract the maximum value of noise decibels, duration and propagation direction (if the noise source has obvious propagation directionality). These parameter peaks, duration and spatial propagation direction and other information constitute the abnormal fluctuation index.

[0073] Step S135, performing similarity matching on the propagation directions of all abnormal fluctuation indicators in the same monitoring sub-region and the abnormal fluctuation indicators of adjacent sub-regions, generating a cross-region association path, and using the topological structure of the cross-region association path as a regional association feature.

[0074] In an industrial park, pollutants in the "high-pollution monitoring sub-area around the workshop" may spread to the adjacent "low-pollution monitoring sub-area in the office area". The abnormal fluctuation indicators in the "high-pollution monitoring sub-area around the workshop" (such as the propagation direction of pollutants) are matched with the abnormal fluctuation indicators in the "low-pollution monitoring sub-area in the office area" for similarity. If the pollutants emitted by the workshop spread toward the office area along the wind direction, and the pollutant concentration change trend detected in the office area has a certain similarity with the concentration change trend around the workshop in direction and time, then it can be determined that there is a cross-regional association path. The topological structure of the cross-regional association path reflects the propagation relationship of environmental anomalies between different monitoring sub-areas, such as one-way propagation or multi-directional propagation, whether the propagation path is a straight line or a curve, etc. This topological structure is used as a regional association feature.

[0075] In a possible implementation, step S140 includes: Step S141, using the pre-trained dynamic anomaly analysis model, pattern matching is performed on the parameter peaks in the abnormal fluctuation index and the historical environmental event database to identify the potential pollution source type and pollution diffusion model.

[0076] For example, for abnormal fluctuation indicators obtained from sub-areas such as the "high pollution monitoring sub-area around the workshop", the parameter peaks (such as the sulfur dioxide concentration peaks) are input into the pre-trained dynamic anomaly analysis model, which matches the parameter peaks with the data in the historical environmental event database. If the current sulfur dioxide concentration peak is similar to the concentration peak in the high pollution event caused by workshop equipment failure in history, then the potential pollution source type can be identified as workshop equipment failure. At the same time, the corresponding pollution diffusion model is determined based on the pollutant diffusion recorded in historical events. For example, when the workshop equipment fails, the pollutants are centered on the workshop and diffuse to the surrounding area according to a certain wind direction and wind speed. The diffusion pattern is determined as the pollution diffusion model of the current event.

[0077] Step S142, calculating the propagation speed and attenuation coefficient of the pollution diffusion model in the target monitoring area according to the cross-regional association path in the regional association feature.

[0078] For example, based on the previously determined cross-regional association path from the "high-pollution monitoring sub-area around the workshop" to the "low-pollution monitoring sub-area in the office area", the distance between the sub-areas, wind direction and speed, and the changes in pollutant concentrations at different locations are combined to calculate the propagation speed of the pollution diffusion model in the industrial park. For example, the distance from the workshop exhaust port to a certain location in the office area is a certain value. By detecting the time difference in the change of pollutant concentration between the two locations, as well as meteorological parameters such as wind direction and speed, the speed at which pollutants diffuse from the workshop to the office area is calculated. At the same time, due to the dilution and sedimentation of the air, the concentration of pollutants will gradually decay. By analyzing the concentration changes at different locations, the attenuation coefficient is calculated.

[0079] Step S143, combining the environmental parameter type corresponding to the spatial attribute label of the monitoring sub-area, predicting the degree of superposition of the pollution diffusion model on the air quality parameters and meteorological parameters.

[0080] For example, for the "high-pollution monitoring sub-area around the workshop" and the "low-pollution monitoring sub-area in the office area", the types of environmental parameters corresponding to their spatial attribute labels are taken into account. For example, the "high-pollution monitoring sub-area around the workshop" mainly focuses on the impact of high pollution emissions on the surrounding air quality, while the "low-pollution monitoring sub-area in the office area" pays more attention to the air quality and comfort of the office environment. According to the pollution diffusion model, the degree of superimposed influence of pollutants emitted from the workshop on air quality parameters (such as sulfur dioxide, nitrogen oxides, particulate matter concentration, etc.) after they diffuse to the office area is predicted. At the same time, the influence of meteorological parameters (such as temperature, humidity, wind direction and wind speed) is taken into account. For example, wind direction and wind speed will affect the diffusion direction and speed of pollutants, and temperature and humidity may affect the chemical reaction and sedimentation rate of pollutants, so as to comprehensively predict the degree of superimposed influence on air quality parameters and meteorological parameters.

[0081] Step S144, based on the spatial coordinates of the monitoring sub-area where the degree of superposition influence exceeds the preset safety threshold, an environmental abnormality event type including an emergency regulation priority is generated.

[0082] If it is predicted that the air quality parameters in the office area exceed the preset safety threshold due to the diffusion and superposition of pollutants in the workshop, for example, the sulfur dioxide concentration in the office area exceeds the prescribed health standard, then an environmental abnormality event type containing an emergency control priority is generated based on the spatial coordinates of the office area in the industrial park. Since the office area is densely populated and has high requirements for air quality, it is defined as a high-urgency environmental abnormality event, and the event type is the spread of workshop pollution to the office area. If the green area is affected by a certain degree of pollution, but it is below the safety threshold, it may be defined as a low-urgency environmental abnormality event.

[0083] Step S145, determining the impact range boundary and duration prediction value of the environmental anomaly event according to the propagation speed and attenuation coefficient of the pollution diffusion model.

[0084] According to the propagation speed and attenuation coefficient of the pollution diffusion model calculated previously, the diffusion range boundary of pollutants in the industrial park is determined. For example, starting from the workshop, as time goes by, the farthest area to which pollutants can diffuse in the future is calculated according to the propagation speed and attenuation coefficient. This area is the impact range boundary of the environmental abnormality event. At the same time, based on factors such as the pollution diffusion model and meteorological parameters, the duration of the entire environmental abnormality event is predicted. If the workshop equipment failure continues to produce pollutants, the possible duration of this pollution event is predicted by combining factors such as the diffusion speed of pollutants, the attenuation coefficient, and the possible time for repairing the workshop equipment.

[0085] In a possible implementation, step S140 further includes: Step S146, selecting the type of target execution equipment and the range of control parameters from the equipment control strategy library according to the type of pollution source corresponding to the type of environmental abnormal event.

[0086] When it is determined that the type of environmental abnormal event is the spread of workshop pollution to the office area, and the corresponding pollution source type is workshop equipment failure, the target execution device is screened from the equipment control strategy library. Since the pollution source is the excessive emission of pollutants caused by the failure of workshop equipment, the target execution device first considers the exhaust gas treatment equipment in the workshop. For the exhaust gas treatment equipment, its control parameter range needs to be determined according to the specific situation of the workshop equipment failure and the type and concentration of the pollutants. For example, if the sulfur dioxide emission exceeds the standard, the control parameter range may involve the amount of desulfurizer in the exhaust gas treatment equipment, which may be adjusted from the current X kilograms per hour to Y kilograms per hour. At the same time, the equipment's processing air volume may also need to be adjusted from M cubic meters per hour to N cubic meters per hour. In addition, it may also be necessary to consider the control of the workshop ventilation equipment. The wind speed control parameter range of the ventilation equipment may be adjusted from the current A meters per second to B meters per second to control the diffusion speed of pollutants in the workshop, so as to better cooperate with the exhaust gas treatment equipment for pollution control.

[0087] Step S147, based on the spatial coordinates of the boundary of the impact range and the emergency control priority, assign a startup time sequence and a parameter adjustment gradient to each target execution device.

[0088] For example, in the scenario where the pollution in the workshop spreads to the office area, the spatial coordinates of the boundary of the impact range determine the areas that need to be focused on, including the office area and the area on the diffusion path between the workshop and the office area. Since the office area is densely populated and the air quality requirements are high, the emergency control of the exhaust gas treatment equipment and ventilation equipment has a higher priority. For the exhaust gas treatment equipment, according to the spatial coordinates of the boundary of the impact range, the startup time sequence is determined to be immediate startup. In terms of parameter adjustment gradient, within the first 10 minutes after startup, the amount of desulfurizer added is increased from X kilograms per hour to (X+ΔX) kilograms per hour, and the treated air volume is increased from M cubic meters per hour to (M+ΔM) cubic meters per hour; in the next 20 minutes, the amount of desulfurizer added is further increased to (X+2ΔX) kilograms per hour, and the treated air volume is increased to (M+2ΔM) cubic meters per hour. For the ventilation equipment, the startup time sequence is also immediate startup. Within the first 5 minutes after startup, the wind speed is increased from A meters per second to (A+ΔA) meters per second, and then to (A+2ΔA) meters per second in the next 15 minutes. As for the air purification equipment (if any) in the area between the workshop and the office area, since its emergency control priority is relatively low, the startup time sequence can be set to start 10 minutes after the exhaust gas treatment equipment is started, and the parameter adjustment gradient can be adjusted accordingly for a relatively gentler time.

[0089] Step S148, combining the duration prediction value and the wind direction and speed data in the meteorological parameters, calculating the expected coverage rate of the parameter adjustment gradient of the target execution device within the influence range.

[0090] For example, it is known that the duration of the abnormal environmental event of the workshop pollution spreading to the office area is predicted to be T hours, the wind direction in the meteorological parameters is southeast wind, and the wind speed is V meters per second. Based on this information and the parameter adjustment gradient of the target execution equipment (exhaust gas treatment equipment, ventilation equipment, etc.), the expected coverage rate is calculated. Taking the exhaust gas treatment equipment as an example, according to the adjustment gradient of its processing air volume, wind direction and speed, and the predicted duration value, the amount of pollutants that can be treated within T hours is calculated as the proportion of the total amount of pollutants in the entire impact range. Assuming that the total amount of pollutants in the impact range is Q kilograms, and the amount of pollutants treated by the exhaust gas treatment equipment according to the adjustment gradient within T hours is q kilograms, then the expected coverage rate is q / Q. For ventilation equipment, according to the control effect of its wind speed adjustment gradient on the diffusion of pollutants, the proportion of the area within the impact range that can effectively prevent the diffusion of pollutants to the entire impact area is calculated, which is also part of its expected coverage rate.

[0091] Step S149: when the expected coverage is lower than a preset efficiency threshold, adding a location deployment suggestion and a collaborative working mode of an auxiliary control device to the target execution device.

[0092] For example, if after calculation, the expected coverage of the exhaust gas treatment equipment and ventilation equipment is lower than the preset efficiency threshold, it means that it is difficult to effectively control the spread of pollution within the specified time by relying on these two equipment alone. At this time, add the location deployment suggestions and collaborative working mode of the auxiliary control equipment to the target execution device. For example, it is recommended to add air purification equipment around the office area, and its location is deployed in the upwind direction of the office area and near the area with dense personnel activities. In terms of collaborative working mode, the air purification equipment works in coordination with the exhaust gas treatment equipment and ventilation equipment. The exhaust gas treatment equipment is responsible for treating the pollutants generated in the workshop, the ventilation equipment controls the diffusion direction and speed of the pollutants in the workshop, and the air purification equipment purifies the pollutants that have spread to the surrounding of the office area. The operating parameters of the air purification equipment are adjusted according to the operating status of the exhaust gas treatment equipment and the ventilation equipment, such as adjusting the purification efficiency of the air purification equipment according to the treatment air volume of the exhaust gas treatment equipment and the wind speed of the ventilation equipment to achieve the best synergistic effect.

[0093] Step S1410: encapsulate the control parameter ranges, startup time sequences and auxiliary control device deployment suggestions of all target execution devices into an executable instruction set to generate the environment optimization strategy set.

[0094] For example, the control parameter range of the exhaust gas treatment equipment (the adjustment range of the desulfurization agent dosage, the treatment air volume, etc.), the startup time sequence (immediate startup and subsequent adjustment time points), the control parameter range of the ventilation equipment (wind speed adjustment range), the startup time sequence, and the deployment recommendations of the additional air purification equipment (location and collaborative working mode) are integrated. Then, this information is packaged in a specific format to form an executable instruction set, which constitutes an environmental optimization strategy set. This environmental optimization strategy set covers all the equipment control information required to deal with the environmental abnormality event of the workshop pollution spreading to the office area.

[0095] In a possible implementation, step S150 includes: Step S151: converting the executable instruction set into an instruction format compatible with a device control protocol according to the type of the target execution device.

[0096] For example, for waste gas treatment equipment, its equipment control protocol may be a specific industrial control protocol. Convert the range of control parameters (such as desulfurizer dosage, treatment air volume), startup time sequence and other information about the waste gas treatment equipment in the environmental optimization strategy set into an instruction format that the device can recognize. For example, convert the adjustment instruction of the desulfurizer dosage into a specific data format specified in the equipment control protocol, such as hexadecimal encoding, and arrange them in the order of instructions specified by the equipment. For ventilation equipment and possible additional air purification equipment, the relevant control information in the executable instruction set is also converted into the corresponding instruction format according to their respective equipment control protocols.

[0097] Step S152: sending an initialization instruction to the target execution device according to the startup time sequence, and receiving a status confirmation signal returned by the target execution device.

[0098] For example, according to the start-up time sequence previously set for the exhaust gas treatment equipment, when the time point of immediate start-up is reached, an initialization instruction can be sent to the exhaust gas treatment equipment. The initialization instruction contains the basic parameters required for the equipment to start, such as equipment self-test, initialization working mode and other information. After receiving the initialization instruction, the exhaust gas treatment equipment performs the corresponding operation and returns a status confirmation signal after the operation is completed. If the equipment self-test is normal and the initialization working mode is set successfully, the returned status confirmation signal indicates that the equipment is ready. Similarly, according to the start-up time sequence of the ventilation equipment and air purification equipment (if deployed), send initialization instructions to them respectively, and receive the status confirmation signals returned by them.

[0099] Step S153, when the status confirmation signal indicates that the target execution device is ready, control instructions are sent in stages according to the parameter adjustment gradient, and the actual parameter value fed back by the target execution device is monitored in real time.

[0100] For example, when the status confirmation signal returned by the exhaust gas treatment equipment indicates that it is ready, the control instructions are sent in stages according to the previously set parameter adjustment gradient. Within the first 10 minutes after startup, control instructions for increasing the amount of desulfurizer added and the amount of air treated are gradually sent in stages every 5 minutes. After sending instructions at each stage, the actual parameter values ​​fed back by the exhaust gas treatment equipment are monitored in real time, such as the actual amount of desulfurizer added, the actual amount of air treated, etc. For ventilation equipment, after receiving its ready status confirmation signal, the wind speed control instructions are sent in stages according to the set parameter adjustment gradient, and the actual wind speed value fed back by the ventilation equipment is monitored in real time. For air purification equipment (if deployed), control instructions are also sent in stages according to its parameter adjustment gradient, and its actual purification efficiency and other parameter values ​​are monitored.

[0101] Step S154, calculating the deviation between the actual parameter value and the parameter target value corresponding to the expected coverage rate, and triggering a coordination instruction of the auxiliary control device if the deviation exceeds the fault tolerance range.

[0102] For example, for waste gas treatment equipment, the actual desulfurization agent dosage and actual treatment air volume monitored in real time are compared with the parameter target value corresponding to the expected coverage rate. For example, the expected desulfurization agent dosage at a certain point in time should be (X+ΔX) kg, but the actual monitored dosage is (X+ΔX-δ) kg (δ is the deviation value), and the calculated deviation ratio is δ / (X+ΔX). If this deviation ratio exceeds the fault tolerance range, it means that the operating effect of the waste gas treatment equipment has not met expectations, and the coordination instruction of the air purification equipment (auxiliary control equipment) is triggered. The air purification equipment adjusts its own operating parameters according to the coordination instruction, such as improving the purification efficiency, to compensate for the impact of the deviation of the operation of the waste gas treatment equipment. Similar deviation calculations and coordination instruction triggering operations are also performed between ventilation equipment and air purification equipment.

[0103] Step S155, recording the execution timestamp of the control instruction, the device response delay and the number of parameter calibrations, and generating a device execution performance report.

[0104] For example, during the execution of the entire control instruction, the execution timestamp of each control instruction is recorded. For example, when does the exhaust gas treatment equipment receive the initialization instruction and when does it receive the control instruction for each stage? The same is true for ventilation equipment and air purification equipment. At the same time, the response delay of the equipment is recorded, that is, the time interval from receiving the instruction to starting to execute the instruction. For the exhaust gas treatment equipment, if it takes a period of time to adjust the amount of desulfurizer after receiving the control instruction to adjust the amount of desulfurizer, this time interval is the equipment response delay. In addition, the number of parameter calibrations is recorded. When the actual parameter value of the exhaust gas treatment equipment deviates greatly from the target value, parameter calibration may be required, and the situation of each calibration is recorded. Finally, an equipment execution performance report is generated based on the recorded information. This report can reflect the actual execution effect of the target execution equipment in the process of executing the environmental optimization strategy, and provide a basis for subsequent equipment optimization and strategy adjustment.

[0105] In a possible implementation, step S150 further includes: Step S156, extracting the correlation between the number of parameter calibration times and the expected coverage from the device execution performance report, and constructing a strategy execution effect evaluation matrix.

[0106] For example, in the previous equipment execution process, equipment execution performance reports were generated for exhaust gas treatment equipment, ventilation equipment, and possible air purification equipment. Taking the exhaust gas treatment equipment as an example, the number of parameter calibrations is obtained from the equipment execution performance report, such as n parameter calibrations during the entire control process, and the corresponding expected coverage rate is clarified. The expected coverage rate reflects the theoretical coverage range of pollutant treatment by the equipment according to the set control strategy. If the number of parameter calibrations is large and the expected coverage rate is low, this indicates that the operation of the equipment deviates greatly from expectations. The relationship between the number of parameter calibrations and the expected coverage rate of these different equipment (exhaust gas treatment equipment, ventilation equipment, air purification equipment, etc.) is integrated to construct a strategy execution effect evaluation matrix. The rows of the matrix can represent different equipment, and the columns represent related indicators such as the number of parameter calibrations and expected coverage rate. The elements in the matrix are the corresponding numerical relationships, which fully reflect the correlation between the equipment execution status and the expected effect during the entire strategy execution process.

[0107] Step S157, calculating the time prediction error coefficient of the dynamic anomaly analysis model according to the difference between the actual duration of the environmental anomaly event and the predicted value.

[0108] For example, in the environmental anomaly event where the pollution in the workshop spreads to the office area, the duration of the event was predicted before, assuming that the predicted value is T 1 In the actual development of the incident, through comprehensive monitoring of the operation of various equipment, the diffusion of pollutants, and the recovery of air quality in the office area, the actual duration was determined to be T 2 Hours. Calculate the difference between the actual duration and the predicted value, that is, (T 2 -T 1 ) / T 1 (Assuming it is expressed in the form of relative error), the value of this difference is used as the time prediction error coefficient of the dynamic anomaly analysis model, which reflects the accuracy of the model in predicting the duration of environmental anomaly events.

[0109] Step S158, input the strategy execution effect evaluation matrix and the time prediction error coefficient into the model weight optimizer to generate a parameter weight adjustment gradient.

[0110] For example, the constructed policy execution effect evaluation matrix (including various correlations related to device execution performance) and the calculated time prediction error coefficient can be input into a dedicated model weight optimizer. The model weight optimizer processes the input information based on the internal algorithm mechanism. For example, it is possible to analyze which devices in the policy execution effect evaluation matrix have a greater impact on the overall effect, and combine the time prediction error coefficient to determine the need for weight adjustment due to the deviation of the model in time prediction. Based on these comprehensive analysis results, a weight adjustment gradient is generated for each parameter in the dynamic anomaly analysis model. The adjustment gradient clarifies how the weight of each parameter should be adjusted, whether to increase or decrease, and the magnitude of the adjustment.

[0111] Step S159, using a back propagation algorithm to iteratively update the weights of the fully connected layer of the dynamic anomaly analysis model until the time prediction error coefficient drops below a convergence threshold.

[0112] For example, in the dynamic anomaly analysis model, the back propagation algorithm is used to update the weights of the fully connected layer according to the generated parameter weight adjustment gradient. The back propagation algorithm starts from the output layer of the dynamic anomaly analysis model, and according to the difference between the predicted result and the actual result (here reflected as the time prediction error coefficient), the error information is back propagated to gradually adjust the weights of the fully connected layer. After each iterative update, the time prediction error coefficient is recalculated to determine whether it has dropped below the convergence threshold. If it has not yet been reached, iterative updates are continued, and the weights are continuously adjusted until the time prediction error coefficient meets the convergence conditions. This process ensures that the dynamic anomaly analysis model can more accurately predict the relevant parameters of environmental abnormal events, such as duration, after adjustment.

[0113] Step S1510, synchronizing the updated model parameter weights to the dynamic anomaly analysis model copies corresponding to all monitoring sub-areas.

[0114] For example, there are multiple monitoring sub-areas in a large industrial park, and each monitoring sub-area has a corresponding copy of the dynamic anomaly analysis model, which is used to independently analyze the environmental data of this area. After the parameter weights of the main dynamic anomaly analysis model are updated, the updated weights are synchronized to the model copies of each monitoring sub-area. For example, the dynamic anomaly analysis model copies of the high-pollution monitoring sub-area around the workshop, the low-pollution monitoring sub-area in the office area, and other sub-areas will all receive updated model parameter weights, thereby ensuring that environmental monitoring and analysis in the entire industrial park are based on unified and optimized model parameters, improving the accuracy and consistency of prediction and response to environmental anomaly events.

[0115] For example, in a possible implementation, the method further includes: Step S310: When it is detected that the target execution device fails to respond to the control instruction for N consecutive times, the device fault diagnosis process is started: Take the exhaust gas treatment equipment as an example. If it fails to respond to the control command for N consecutive times, first obtain its last online time. This time point is the key information to determine when the equipment lost contact. At the same time, query its historical maintenance records to understand the past maintenance of the equipment, including the time of maintenance, the reason for maintenance (such as component replacement, software upgrade, etc.), and the operation status after maintenance. In addition, obtain the operation status log of adjacent equipment (such as ventilation equipment related to the exhaust gas treatment equipment in the workshop, other production equipment, etc.). The operation status log contains the changes of various operating parameters of the equipment over time, such as the change curves of parameters such as current, voltage, and temperature.

[0116] Step S320, obtaining the last online time, historical maintenance records and operating status logs of adjacent devices of the target execution device.

[0117] For example, in the operating status log of the adjacent equipment of the exhaust gas treatment equipment, analyze the current fluctuation characteristics. If it is found that the current suddenly increases or decreases at a certain moment and remains abnormal, this may indicate a problem with the electrical connection with the exhaust gas treatment equipment or a circuit failure of the exhaust gas treatment equipment itself. At the same time, observe the temperature change curve. If the temperature of the equipment rises abnormally, it may be a hardware problem caused by a failure of the heat dissipation components inside the equipment or an overload operation. For the communication link, if the communication-related parameters of the adjacent equipment (such as signal strength, communication frequency, etc.) fluctuate abnormally during the period when the exhaust gas treatment equipment does not respond to the control instructions, combined with the communication protocol and network topology, the approximate location of the communication link interruption can be inferred, such as a communication node in the workshop or a link between the equipment and the control center.

[0118] Step S330: Identify the device hardware failure type or the communication link interruption location based on the current fluctuation characteristics and temperature change curve in the operation status log.

[0119] Step S340: If it is a hardware failure, matching available devices of the same type from the spare device library and recalculating the impact of the deployment position of the available devices on the expected coverage.

[0120] For example, when it is determined that the exhaust gas treatment equipment has a hardware failure, the same type of available equipment is searched from the spare equipment library. Assume that there is an exhaust gas treatment equipment model X available in the spare equipment library. When deploying the spare equipment to the workshop, it is necessary to recalculate the impact of its deployment location on the expected coverage. Due to factors such as the location of the pollution source, ventilation conditions, and pollutant diffusion paths in the workshop, different deployment locations will affect the equipment's treatment effect on pollutants. Based on the layout of the workshop, the design of the ventilation system, and the previous analysis of the pollutant diffusion model, the proportion of pollutants that the spare equipment can treat at different potential deployment locations to the total amount of pollutants generated in the entire workshop is calculated, that is, the expected coverage. By comparing the expected coverage of different locations, the optimal deployment location is selected.

[0121] Step S350: If the communication is interrupted, switch to the backup communication frequency band to send an encrypted test signal, and locate the link interruption node according to the signal delay time.

[0122] If it is determined that the communication is interrupted, the communication of the exhaust gas treatment equipment is switched to the backup communication frequency band. An encrypted test signal is sent on the backup communication frequency band. The encrypted test signal contains the identification information of the equipment and some special codes for detecting the link status. The link interruption node is located based on the propagation of the encrypted test signal in the link, especially the signal delay time. If the time from sending the test signal to receiving the feedback signal is too long, which exceeds the normal signal propagation delay range, combined with the topological structure of the communication network in the workshop and the location information of each communication node, it can be determined which node or which link has been interrupted.

[0123] Step S360: Push the equipment failure type, standby equipment deployment plan and communication link repair suggestion to the maintenance terminal.

[0124] In detail, whether it is determined to be a hardware failure or a communication interruption, the relevant information can be pushed to the maintenance terminal. If it is a hardware failure, the information pushed includes the failure type of the exhaust gas treatment equipment (such as circuit failure, heat dissipation component failure, etc.), the selected backup equipment deployment plan (the model of the backup equipment, the deployment location, etc.). If it is a communication interruption, the location information of the communication link interruption and the corresponding repair suggestions (such as repairing the interrupted node, adjusting the communication frequency band parameters, etc.) are pushed. Maintenance personnel can repair or adjust the equipment in time based on this information to ensure the normal operation of the environmental monitoring and pollution control equipment in the industrial park.

[0125] For example, in a possible implementation, the method further includes: Step S410: Count the distribution of environmental abnormal event types and strategy execution success rates of all monitored sub-areas within a preset period and generate a regional security assessment report: Step S420, extract the occurrence frequency, average impact range and evolution trend of the associated pollution source type of each environmental abnormal event, calculate the difference in execution success rate of the same type of environmental optimization strategy in different time periods, and identify the periodic characteristics of external environmental interference factors.

[0126] For example, in each monitoring sub-area of ​​a large industrial park, different types of environmental anomalies are statistically analyzed. Taking the environmental anomaly event of workshop pollution spreading to the office area as an example, its frequency of occurrence within a preset period (such as one month) is counted. The impact range of each event is calculated, and then the average impact range is calculated. At the same time, the evolution trend of the associated pollution source type (such as pollution caused by workshop equipment failure) is analyzed to observe whether the frequency of a certain equipment failure has increased or decreased over time, or whether the type of failure has changed. For the same type of environmental optimization strategy (such as the equipment control strategy for the spread of workshop pollution to the office area), the difference in its execution success rate in different time periods (such as each week) is calculated. For example, the execution success rate in the first week is 80%, and in the second week it is 75%. By analyzing the fluctuations of these success rates, combined with factors such as meteorological conditions outside the industrial park (such as periodic changes in wind direction and wind speed), surrounding traffic flow (such as the difference between working days and non-working days), etc., the periodic characteristics of external environmental interference factors can be identified.

[0127] Step S430, according to the evolution trend and periodic characteristics, optimize the parameter similarity judgment threshold in the dynamic grid segmentation scheme, and adjust the noise filtering threshold in the data feature extraction rule so that the noise filtering threshold is dynamically associated with the optimized parameter similarity judgment threshold.

[0128] Based on the evolution trend of the pollution source types associated with environmental anomalies and the periodic characteristics of external environmental interference factors obtained in the previous analysis, the parameter similarity judgment threshold in the dynamic grid segmentation scheme is optimized. For example, if it is found that the pollution diffusion pattern in the area around the workshop has changed over time, it may be necessary to adjust the parameter similarity judgment threshold to more accurately divide the monitoring sub-areas. At the same time, adjust the noise filtering threshold in the data feature extraction rule. If the periodic changes in external environmental interference factors (such as traffic flow) lead to periodic fluctuations in the noise level, adjust the noise filtering threshold accordingly to better filter the noise data. And ensure that the noise filtering threshold is dynamically associated with the optimized parameter similarity judgment threshold, for example, according to the adjustment range of the parameter similarity judgment threshold, adjust the noise filtering threshold according to a certain proportional relationship, so that the two can work together in the environmental monitoring and analysis process to improve the accuracy of environmental data processing.

[0129] Step S440, writing the updated dynamic grid segmentation scheme and data feature extraction rules into the version control queue of the preset rule library.

[0130] For example, the optimized dynamic grid segmentation scheme and the adjusted data feature extraction rules can be integrated and then written into the version control queue of the preset rule base. In the version control queue, these rules are managed in chronological order or according to the logic of version update. This makes it easy to trace the historical versions of the rules and ensure that the latest and optimized dynamic grid segmentation scheme and data feature extraction rules are used in the environmental monitoring system of the industrial park, thereby improving the effectiveness and adaptability of the entire environmental monitoring and management system.

[0131] For example, in a possible implementation, the method further includes: Step S510: When it is detected that a new sensor node is connected to the target monitoring area, a node data fusion verification process is executed: Step S520: Acquire the device identifier, sampling frequency and initial sequence of environmental data collected by the newly added sensor node.

[0132] In this embodiment, the device identifier is the unique identifier of the newly added sensor node in the entire monitoring system, for example, it may be a specific code composed of letters and numbers, which is used to accurately identify the node among many devices. The sampling frequency indicates the time interval regularity of the sensor node collecting environmental data, such as collecting environmental data such as air quality parameters, noise intensity parameters and meteorological parameters every 10 minutes to form an initial sequence of environmental data. The initial sequence of environmental data includes a series of data collected at the sampling frequency since the sensor node started working.

[0133] Step S530: Time-align the initial sequence with the historical environmental data of adjacent sensor nodes, and calculate the parameter correlation coefficient within the same time window.

[0134] In the industrial park, there are existing adjacent sensor nodes around the newly added sensor nodes, and these adjacent sensor nodes have their own historical environmental data records. In order to perform data fusion verification, it is necessary to align the initial sequence collected by the newly added sensor node with the historical environmental data of the adjacent sensor nodes in time. For example, if the newly added sensor node starts collecting data at 10 am, and the historical data records of the adjacent sensor nodes cover the entire morning period, then the data after 10 am are aligned so that they are in the same time frame. Then, for each environmental parameter (such as sulfur dioxide concentration, noise decibel value, temperature, etc.), the parameter correlation coefficient is calculated in the same time window (for example, 1 hour as a time window). This coefficient reflects the degree of correlation between the data collected by the newly added sensor node and the historical data of the adjacent sensor node in the corresponding environmental parameter. If the sulfur dioxide concentration data collected by the newly added sensor node and the sulfur dioxide concentration data in the historical records of the adjacent sensor node show similar trends in a certain time window, then the parameter correlation coefficient between them will be high; conversely, if the data of the two are unrelated, the parameter correlation coefficient will be low.

[0135] Step S540: if the parameter correlation coefficient is lower than the node consistency threshold, start the node calibration mode: send a standard test signal to the newly added sensor node and analyze the error distribution of its response data.

[0136] Assume that the node consistency threshold is set to 0.8. If the calculated parameter correlation coefficient is lower than this threshold, it means that the data collected by the newly added sensor node is less consistent with the data of the adjacent sensor nodes and needs to be calibrated. At this time, a standard test signal is sent to the newly added sensor node. This standard test signal is a known signal with specific environmental parameter values, such as a simulated, fixed-concentration sulfur dioxide standard test signal, or a fixed-intensity noise standard test signal. After receiving the standard test signal, the newly added sensor node will generate response data, and then analyze the error distribution of these response data and the standard test signal. For the sulfur dioxide concentration response data, if there is a deviation from the concentration value in the standard test signal, analyze the distribution of this deviation at different time points and in different concentration ranges, whether the deviation is large at high concentrations, large at low concentrations, or random deviations, etc.

[0137] Step S550: adjusting the sampling frequency compensation parameter of the newly added sensor node according to the error distribution, and recalculating the parameter correlation coefficient until the node consistency threshold is reached.

[0138] For example, the sampling frequency compensation parameters of the newly added sensor nodes can be adjusted based on the error distribution obtained from the previous analysis. If the error distribution shows that the response data deviation is large under the high concentration of sulfur dioxide test signal, the sampling frequency compensation parameters may need to be reduced to make the sensor collect data more accurately in a high concentration environment; if the deviation is large at low concentrations, the sampling frequency compensation parameters may need to be increased. After the adjustment, the data collected by the newly added sensor node is time-aligned with the historical environmental data of the adjacent sensor nodes again, and the parameter correlation coefficient within the same time window is recalculated. Repeat this process, continuously adjust the sampling frequency compensation parameters and recalculate the coefficients until the parameter correlation coefficient reaches the node consistency threshold of 0.8.

[0139] Step S560: adding the newly added sensor nodes that have passed the verification to the spatial distribution matrix, and triggering the incremental update calculation of the dynamic grid segmentation solution.

[0140] For example, when the parameter correlation coefficient of the newly added sensor node reaches the node consistency threshold, it means that the data collected by the node has good consistency with the surrounding environment data, and it can be added to the spatial distribution matrix. The spatial distribution matrix records the location coordinates and parameter change gradients of each sensor node in the industrial park. After the new sensor node is added, the spatial distribution matrix is ​​updated. Since the sensor node layout has changed, it is necessary to trigger the incremental update calculation of the dynamic grid segmentation scheme. In the original dynamic grid segmentation scheme, it is generated based on the previous sensor node layout. The addition of new sensor nodes may affect the parameter mutation boundary, the division of the steady-state area, and the layout of the grid unit. Through the incremental update calculation, the spatial distribution characteristics of the environmental data in the industrial park are re-evaluated, the division of the grid unit is adjusted, and the definition of the parameter mutation boundary and the steady-state area is updated to adapt to the new sensor node layout, ensuring that the entire environmental monitoring system can accurately monitor and analyze the environmental conditions of the industrial park.

[0141] Figure 2 The schematic diagram shows exemplary hardware and software components of the server 100 that can implement the concept of the present application provided by some embodiments of the present application. For example, the processor 120 can be used on the server 100 and used to execute the functions in the present application.

[0142] The server 100 may be a general-purpose server or a special-purpose server, both of which may be used to implement the data processing method based on AI environment monitoring of the present application. Although only one server is shown in the present application, for convenience, the functions described in the present application may be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0143] For example, the server 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and storage media 140 in different forms, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the server 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application may be implemented according to these program instructions. The server 100 also includes an I / O interface 150 between a computer and other input and output devices.

[0144] For ease of explanation, only one processor is described in the server 100. However, it should be noted that the server 100 in the present application may also include multiple processors, so the steps performed by one processor described in the present application may also be performed jointly or individually by multiple processors. For example, if the processor of the server 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually in one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor perform steps A and B together.

[0145] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer executable instructions are preset. When the processor executes the computer executable instructions, the above-mentioned data processing method based on AI environment monitoring is implemented.

[0146] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, various features are sometimes combined into one embodiment, drawing or description thereof.

Claims

1. A data processing method based on AI environmental monitoring, characterized in that: The method comprises: Acquire a set of environmental data collected in real time by multiple sensor nodes in a target monitoring area, wherein the set of environmental data includes air quality parameters, noise intensity parameters and meteorological parameters in a time series; According to the spatial distribution characteristics of the environmental data set, the target monitoring area is dynamically divided into a plurality of monitoring sub-areas, and a corresponding data feature extraction rule is assigned to each monitoring sub-area; Based on the data feature extraction rules, abnormal fluctuation indicators and regional correlation features are extracted from the environmental data of each monitoring sub-region to generate a sub-region feature set; Using the pre-trained dynamic anomaly analysis model, a multi-dimensional correlation analysis is performed on the abnormal fluctuation indicators in the sub-region feature set to determine the type and impact range of environmental abnormal events in the target monitoring area, and generate an environmental optimization strategy set; At least one environment optimization strategy in the environment optimization strategy set is executed, a control instruction is sent to a target execution device, and a parameter weight of the dynamic anomaly analysis model is updated based on feedback data of the control instruction.

2. The data processing method based on AI environment monitoring according to claim 1 is characterized in that: The target monitoring area is dynamically divided into a plurality of monitoring sub-areas according to the spatial distribution characteristics of the environmental data set, and a corresponding data feature extraction rule is assigned to each monitoring sub-area, including: Extract the position coordinates and parameter change gradient of each sensor node from the environmental data set to construct a spatial distribution matrix; Performing density cluster analysis on the spatial distribution matrix to identify parameter mutation boundaries and steady-state areas within the target monitoring area; Generating a dynamic grid segmentation scheme covering the target monitoring area according to the geometric shape of the parameter mutation boundary and the parameter similarity of the steady-state area; Merge continuous grid cells whose parameter similarity in the dynamic grid segmentation scheme is higher than a preset threshold into the same monitoring sub-area, and annotate each monitoring sub-area with a spatial attribute label; According to the environmental parameter type corresponding to the spatial attribute label, the noise filtering threshold and the feature fusion weight in the data feature extraction rule are matched from a preset rule library.

3. The data processing method based on AI environment monitoring according to claim 2 is characterized in that: The method of generating a dynamic grid segmentation scheme covering the target monitoring area according to the geometric shape of the parameter mutation boundary and the parameter similarity of the steady-state area includes: Extracting a spatial coordinate sequence of boundary points from the parameter mutation boundary, and identifying a geometric shape type of the boundary line according to the continuity of the spatial coordinate sequence; According to the curvature change characteristics of the geometric shape type, the boundary line is divided into a plurality of geometric feature segments, and a curvature change direction is marked for each geometric feature segment; Based on the curvature change direction, initial grid cells are generated on both sides of the boundary line, and the side lengths of the initial grid cells are matched with the parameter similarity of the steady-state region to determine the side length adjustment ratio of the initial grid cells; According to the side length adjustment ratio, the initial grid unit is expanded or contracted to generate a boundary grid unit adapted to the geometric shape of the boundary line; Extracting a sensor node group with the highest parameter similarity from the steady-state region, and determining a grid unit division density of the steady-state region according to the spatial distribution density of the sensor node group; Based on the grid unit division density, uniformly distributed steady-state grid units are generated in the steady-state region, and the side lengths of the steady-state grid units are aligned with the side lengths of the boundary grid units to generate transitional grid units; According to the changing trend of the side length of the transitional grid unit, the boundary grid unit is seamlessly connected with the steady-state grid unit to generate a dynamic grid segmentation scheme covering the target monitoring area; Mapping the spatial coordinate sequences of the boundary grid cells, transition grid cells and steady-state grid cells in the dynamic grid segmentation scheme with the parameter similarities of the parameter mutation boundary and the steady-state region to generate grid cell attribute labels; According to the grid cell attribute labels, the grid cells in the dynamic grid segmentation scheme are clustered according to parameter similarity to generate multiple grid cell groups, and a spatial attribute label is marked for each grid cell group; Based on the spatial attribute labels, the grid unit groups in the dynamic grid segmentation scheme are matched with the spatial distribution characteristics of the target monitoring area to generate a dynamic grid segmentation scheme covering the target monitoring area.

4. The data processing method based on AI environment monitoring according to claim 1 is characterized in that: The method of extracting abnormal fluctuation indicators and regional correlation features from the environmental data of each monitoring sub-area based on the data feature extraction rule includes: Performing sliding window normalization processing on the time series environmental data of all sensor nodes in the monitoring sub-area to generate a normalized data sequence; According to the noise filtering threshold in the data feature extraction rule, outlier points are eliminated and data interpolation is performed on the normalized data sequence to generate a smoothed data sequence; Calculating the parameter change rate of adjacent time windows in the smoothed data sequence, and marking the time window where the parameter change rate exceeds the upper limit of the historical fluctuation range as an abnormal fluctuation interval; Extracting the parameter peak value, duration and spatial propagation direction within the abnormal fluctuation interval to generate the abnormal fluctuation index; The propagation directions of all abnormal fluctuation indicators in the same monitoring sub-region are matched with the abnormal fluctuation indicators of adjacent sub-regions to generate cross-regional association paths, and the topological structure of the cross-regional association paths is used as the regional association feature.

5. The data processing method based on AI environment monitoring according to claim 1 is characterized in that: The pre-trained dynamic anomaly analysis model is used to perform multi-dimensional correlation analysis on the abnormal fluctuation indicators in the sub-region feature set to determine the type and impact range of environmental abnormal events in the target monitoring area, including: Using a pre-trained dynamic anomaly analysis model, the parameter peaks in the abnormal fluctuation index are pattern matched with the historical environmental event database to identify potential pollution source types and pollution diffusion models; Calculate the propagation speed and attenuation coefficient of the pollution diffusion model in the target monitoring area according to the cross-regional association path in the regional association feature; In combination with the environmental parameter type corresponding to the spatial attribute label of the monitoring sub-area, predict the degree of superposition of the pollution diffusion model on the air quality parameters and meteorological parameters; Based on the spatial coordinates of the monitoring sub-areas where the degree of superimposed impact exceeds a preset safety threshold, an environmental abnormality event type including an emergency regulation priority is generated; According to the propagation speed and attenuation coefficient of the pollution diffusion model, the impact range boundary and duration prediction value of the environmental abnormal event are determined.

6. The data processing method based on AI environment monitoring according to claim 5 is characterized in that: The generation environment optimization strategy set includes: According to the pollution source type corresponding to the environmental abnormal event type, the type of target execution equipment and the range of control parameters are selected from the equipment control strategy library; Based on the spatial coordinates of the impact range boundary and the emergency control priority, assigning a startup time sequence and a parameter adjustment gradient to each target execution device; Calculate the expected coverage rate of the parameter adjustment gradient of the target execution device within the influence range by combining the duration prediction value and the wind direction and speed data in the meteorological parameters; When the expected coverage rate is lower than a preset efficiency threshold, adding a location deployment suggestion and a collaborative working mode of an auxiliary control device to the target execution device; The control parameter ranges, startup time sequences and auxiliary control device deployment suggestions of all target execution devices are encapsulated into an executable instruction set to generate the environment optimization strategy set.

7. The data processing method based on AI environment monitoring according to claim 6 is characterized in that: The executing at least one environment optimization strategy in the environment optimization strategy set and sending a control instruction to a target execution device includes: According to the type of the target execution device, converting the executable instruction set into an instruction format compatible with the device control protocol; Sending an initialization instruction to the target execution device according to the startup time sequence, and receiving a status confirmation signal returned by the target execution device; When the status confirmation signal indicates that the target execution device is ready, control instructions are sent in stages according to the parameter adjustment gradient, and actual parameter values ​​fed back by the target execution device are monitored in real time; Calculate the deviation between the actual parameter value and the parameter target value corresponding to the expected coverage rate, and trigger a coordination instruction of the auxiliary control device if the deviation exceeds the fault tolerance range; The execution timestamp of the control instruction, the device response delay and the number of parameter calibrations are recorded to generate a device execution performance report.

8. The data processing method based on AI environment monitoring according to claim 7 is characterized in that: The updating of the parameter weights of the dynamic abnormality analysis model based on the feedback data of the control instruction includes: Extracting the correlation between the number of parameter calibrations and the expected coverage from the device execution performance report, and constructing a strategy execution effect evaluation matrix; Calculating the time prediction error coefficient of the dynamic anomaly analysis model according to the difference between the actual duration of the environmental anomaly event and the predicted value; Input the strategy execution effect evaluation matrix and the time prediction error coefficient into the model weight optimizer to generate a parameter weight adjustment gradient; Iteratively updating the weights of the fully connected layer of the dynamic anomaly analysis model using a back propagation algorithm until the time prediction error coefficient drops below a convergence threshold; The updated model parameter weights are synchronized to the corresponding dynamic anomaly analysis model copies of all monitoring sub-areas.

9. The data processing method based on AI environment monitoring according to claim 1 is characterized in that: The method further comprises: When it is detected that the target execution device fails to respond to the control command for N consecutive times, the device fault diagnosis process is started: Obtain the last online time, historical maintenance records, and operating status logs of adjacent devices of the target execution device; Identify the type of device hardware failure or the location of communication link interruption based on the current fluctuation characteristics and temperature change curve in the operation status log; If it is a hardware failure, match available devices of the same type from the backup device library, and recalculate the impact of the deployment position of the available devices on the expected coverage; If the communication is interrupted, switch to the backup communication frequency band to send the encrypted test signal, and locate the link interruption node according to the signal delay time; Push equipment failure types, backup equipment deployment plans, and communication link repair suggestions to the maintenance terminal.

10. A server, characterized in that: The server includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the data processing method based on AI environmental monitoring as described in any one of claims 1 to 9 above.

Images