Method and system for automatically predicting various environmental parameters based on GIS map

Abstract

The invention relates to the technical field of environment detection, and discloses a GIS map-based multi-class environment parameter automatic prediction method, which comprises the following steps: collecting noise intensity, vibration spectrum, sewage turbidity, illumination intensity and PM2.5 concentration in real time through a distributed sensor network, combining social behavior data, carrying out space-time alignment, and generating a multi-dimensional space-time matrix; calculating an energy overlapping degree by adopting a space-time diagram attention network, marking a high-risk collaborative pollution area, and predicting a pollution diffusion path; constructing a self-adaptive prediction model containing physical and behavior driving channels, and dynamically adjusting the weight to optimize the prediction precision; based on the optimization model, reversely deducing a pollution source of an overproof area, matching equipment characteristics and generating a control instruction; and opening an AR (Augmented Reality) interface to verify the treatment effect, and when the virtual-real data deviation exceeds a data deviation threshold, triggering federal learning to update the global model, and generating an environmental protection compliance report. According to the invention, the accurate decision-making efficiency of environmental governance can be improved.

Description

Technical Field

[0001] This application relates to the technical field of environmental detection, and in particular, to a multi-class environmental parameter automatic prediction method and system based on a GIS map. Background Art

[0002] Currently, there are common problems of insufficient integration of multi-source data and weak dynamic prediction ability in the field of environmental monitoring. Traditional environmental detection systems mostly rely on single sensors to collect basic parameters such as noise and particulate matter, and it is difficult to effectively integrate dynamic behavior data such as traffic flow and social media public opinion, resulting in a significant lag in the spatio-temporal correlation analysis of pollution events.

[0003] Although existing GIS platforms can achieve visualization of basic data, they generally adopt static grid division and linear prediction models, and cannot capture the non-linear interaction characteristics between complex environmental parameters. Especially in sudden pollution events, the existing systems have insufficient ability to dynamically correct the pollution diffusion path, resulting in a large deviation between the prediction results and the measured data, seriously restricting the precision decision-making efficiency of environmental governance.

[0004] As can be seen from the above, how to improve the precision decision-making efficiency of environmental governance remains to be solved. Summary of the Invention

[0005] In order to improve the precision decision-making efficiency of environmental governance, this application provides a multi-class environmental parameter automatic prediction method and system based on a GIS map.

[0006] In a first aspect, this application provides a multi-class environmental parameter automatic prediction method based on a GIS map, adopting the following technical solution: A multi-class environmental parameter automatic prediction method based on a GIS map, comprising: Real-time collect data on noise intensity, vibration spectrum, sewage turbidity, light intensity, and PM2.5 concentration in the target area, transmit it to the GIS platform through a distributed sensor network, and synchronously obtain real-time changes in traffic flow, spatio-temporal distribution of environmental keywords in social media, and short-term fluctuation data of meteorological parameters in the target area. Based on the GIS spatial grid technology, align the environmental parameters and dynamic behavior data in a preset format in space and time to generate a multi-dimensional spatio-temporal matrix integrating physical parameters and social behaviors, where the physical parameters include data on noise intensity, vibration spectrum, sewage turbidity, light intensity, and PM2.5 concentration, and the social behaviors include real-time changes in traffic flow, spatio-temporal distribution of environmental keywords in social media, and short-term fluctuation data of meteorological parameters; Extract the energy distribution characteristics of the main noise frequency band and the spectral energy distribution characteristics of the vibration sensor based on the multi-dimensional spatio-temporal matrix, calculate the energy overlap degree between the energy distribution characteristics and the spectral energy distribution characteristics, and mark it as a high-risk collaborative pollution area when the energy overlap degree exceeds the overlap degree threshold; obtain the sewage flow, real-time wind speed, precipitation data, and collaborative pollution characteristics corresponding to the high-risk collaborative pollution area, predict the pollution diffusion path and the change of concentration gradient by combining the sewage flow, real-time wind speed, precipitation data, and collaborative pollution characteristics, and use the spatio-temporal graph attention network to dynamically learn the non-linear association between parameters and output the collaborative influence weight; Construct an adaptive prediction model including a physical-driven channel and a behavior-driven channel according to the collaborative influence weight. The physical channel generates a benchmark prediction result based on the noise attenuation equation and CFD fluid simulation. The behavior channel analyzes the temporal correlation between traffic flow and social data through the LSTM network to generate a prediction correction factor. When the deviation between the physical prediction and the measured data exceeds the data deviation threshold for multiple consecutive periods, reduce the weight of the physical channel and increase the weight of the behavior channel. At the same time, start the digital twin sandbox in GIS to simulate the pollution source migration scenario to generate virtual training data to optimize the adaptive prediction model; Based on the optimized adaptive prediction model, reverse the pollution diffusion path of the exceeded standard area and locate the target device in combination with the sensor data; retrieve the device feature matching model, perform wavelet packet decomposition on the noise and vibration signals, and match the vibration characteristics in the device feature fingerprint library. When the matching is successful, generate a control instruction and send it to the IoT terminal device to execute the operation of shutting down the exceeded standard pump or starting the noise reduction barrier; According to the execution result of the IoT terminal and the real-time monitoring data, open the AR augmented reality verification interface, retrieve the prediction data and the measured data of the GIS platform, generate a deviation heat map based on the prediction data and the measured data, and obtain the corresponding virtual-real data deviation. When the virtual-real data deviation exceeds 20%, trigger the federated learning framework to aggregate the local gradients of the edge nodes to update the global prediction model, and retain the historical version to construct a model evolution tree for anomaly backtracking analysis; generate the corresponding environmental compliance report, and the environmental compliance report includes details of exceeded standard parameters, regulatory correlation analysis, control effect evaluation, executable suggestions, and blockchain evidence data.

[0007] By adopting the above technical solutions, through the spatio-temporal fusion and intelligent modeling of multi-source heterogeneous data, a dynamic decision-making closed-loop for environmental governance has been constructed, significantly improving the efficiency of precise decision-making; the system integrates physical environment parameters and social behavior data in real time, and through an adaptive prediction model with dynamic weight adjustment and digital twin simulation, realizes the rapid positioning of collaborative pollution risks and the precise prediction of diffusion paths; combined with equipment feature matching and AR verification feedback, targeted control instructions can be quickly generated and governance strategies can be dynamically optimized. At the same time, the prediction accuracy is continuously iterated through federated learning and model evolution tree mechanisms, and finally an automated closed-loop of "monitoring - prediction - control - verification - optimization" is formed, greatly shortening the decision response time, reducing human intervention errors, and turning environmental governance from passive response to active, precise, and dynamic intelligent regulation.

[0008] Optionally, the method further includes: Extract the energy distribution characteristics of the main frequency band of the noise through short-time Fourier transform. The energy distribution characteristics include the peak value of the main frequency energy, the proportion of the energy in the frequency band, and the slope of the energy distribution curve; Extract the spectral energy distribution characteristics of the vibration sensor through vibration spectrum analysis. The spectral energy distribution characteristics include the peak frequency of the vibration energy, the fluctuation amplitude of the energy in the frequency band, and the spectral energy density; Calculate the dynamic overlap degree between the energy distribution of the main frequency band of the noise and the spectral energy distribution of the vibration, including: standardize the frequency-domain energy distribution curves of the energy distribution characteristics and the spectral energy distribution characteristics, and use the weighted cosine similarity calculation formula, where the weight of the high-frequency band is higher than that of the low-frequency band; When the dynamic overlap degree exceeds the overlap degree threshold and the difference in the proportion of the high-frequency band energy is less than the high-frequency band energy proportion difference threshold, it is marked as a high-risk collaborative pollution area.

[0009] By adopting the above technical solutions, through multi-dimensional feature analysis and dynamic similarity evaluation, the precise identification and positioning of collaborative environmental pollution risks have been achieved, significantly improving the decision-making efficiency of environmental governance: through weighted analysis of the high-frequency band and the low-frequency band, combined with the dynamic matching of energy distribution characteristics, the pollution source area under the combined action of noise and vibration can be quickly locked, avoiding misjudgment caused by single parameters or static thresholds in traditional methods, enabling governance resources to be directed to high-risk pollution sources, reducing the redundancy of large-area inspections, and at the same time ensuring the accuracy and timeliness of decisions through real-time dynamic evaluation and threshold constraints, ultimately achieving targeted, intelligent, and rapid response to pollution prevention and control.

[0010] Optionally, during the process of dynamically learning the non-linear correlation between parameters using the spatio-temporal graph attention network, the method further includes: Construct the node feature matrix of the spatio-temporal graph. The node feature matrix includes environmental parameters, dynamic behavior parameters, and spatial correlation parameters. Among them, the environmental parameters include noise intensity, peak value of vibration spectrum energy, sewage turbidity, PM2.5 concentration, and light intensity. The dynamic behavior parameters include traffic flow density, frequency of keyword mentions in social media, wind speed, and precipitation intensity. The spatial correlation parameters include predicted values of pollution diffusion paths and concentration gradient change values of adjacent grids; Introduce a dual-channel attention mechanism in the graph attention network. Among them, the spatial channel calculates spatial attention weights based on grid adjacency relationships, and the temporal channel calculates temporal attention weights based on the autocorrelation of time series; When the noise intensity is lower than the noise lower threshold, dynamically reduce the environmental parameter weight of the corresponding grid to 30% of the original weight.

[0011] By adopting the above technical solutions, through the dual-channel mechanism (spatial and temporal dimensions) and dynamic weight adjustment of the spatio-temporal graph attention network, the accuracy and robustness of complex environmental parameter correlation modeling are significantly improved. Especially in low-noise scenarios, interference factors are effectively suppressed, enabling the model to more accurately capture the spatio-temporal laws of pollution diffusion, thus providing more reliable decision-making support for environmental governance.

[0012] Optionally, during the process of starting the digital twin sandbox to simulate the pollution source migration scenario, the method further includes: Construct a multi-physical field coupling simulation model. The multi-physical field coupling simulation model includes an air pollution diffusion module and a ground vibration diffusion module. The air pollution diffusion module simulates the air propagation paths of PM2.5 and noise based on CFD fluid simulation, and the ground vibration diffusion module simulates the vibration energy attenuation in high-risk areas based on the vibration wave equation; Introduce spatio-temporal attenuation factors in the GIS grid. The spatio-temporal attenuation factors are calculated based on the exponential decay function of distance and time; When the deviation between the simulated path and the measured pollution diffusion path exceeds the path deviation threshold, trigger the sandbox to adaptively adjust the migration probability distribution parameters, specifically including: increasing the contribution weight of wind speed to pollutant diffusion by 20%, and recalculating the non-linear relationship between precipitation intensity and pollutant settlement rate according to historical data.

[0013] By adopting the above technical solutions, through multi-physical field coupling simulation and adaptive parameter calibration, the accuracy and dynamic adaptability of pollution diffusion path simulation are significantly improved. It can quickly adjust model parameters (such as wind speed weight, precipitation influence) when measurement deviations occur, ensuring that the simulation results are highly consistent with the real scenario, and providing a reliable basis for accurate pollution source positioning and formulation of governance strategies.

[0014] Optionally, during the implementation process of the device feature matching model, the method further includes: Perform three-layer wavelet packet decomposition on noise and vibration signals to extract vibration characteristics in high frequency band, medium frequency band and low frequency band; Perform weighted voting matching on the decomposed vibration features and the device fingerprint library; When the similarities of the three frequency bands all exceed the similarity threshold and the weighted total similarity ≥ the weighted total similarity threshold, the match is considered successful; when the matching failure rate continuously exceeds the matching failure rate threshold, the fingerprint library incremental update mechanism is triggered, the vibration data of the preset similar time period is automatically retrieved, and the fingerprint library feature vector is updated using the online learning algorithm.

[0015] By adopting the above technical solution, through weighted matching of multi-band vibration characteristics and dynamic fingerprint library update mechanism, the accurate identification capability of pollution source equipment is significantly improved, ensuring the rapid positioning of pollution sources in complex environments. At the same time, the matching model is continuously optimized through adaptive learning, reducing misjudgment and improving the speed and accuracy of governance response.

[0016] Optionally, in the process of inferring the pollution diffusion path of the exceeding-standard area based on the optimized adaptive prediction model, the method further includes: Construct a multi-scale spatiotemporal inversion model, dividing the pollution diffusion path into micro-scale, meso-scale and macro-scale. The micro-scale locates the precise coordinates of the pollution source equipment based on the data of the sensor grid, the meso-scale combines traffic flow and wind speed data to infer the propagation path of pollutants from the pollution source to the excessive area, and the macro-scale analyzes the spatiotemporal distribution of regional pollution sources through historical pollution event data; When locating the target device, a dynamic confidence evaluation mechanism is used to calculate the device position based on the confidence weight of the sensor data, where the confidence is the matching degree between the noise intensity and the vibration energy. When the confidence is lower than 80%, the collaborative inversion of adjacent grids is triggered, and the positioning accuracy is improved through multi-grid data fusion. When generating control instructions, a graded response strategy is introduced. When the exceeding parameter is noise and its duration exceeds the preset duration threshold, the noise reduction barrier is activated first. When the exceeding parameter is sewage turbidity and its concentration exceeds the sewage concentration trigger value, the pump is triggered to shut down and the surrounding sewage treatment facilities are linked to start the purification process.

[0017] By adopting the above technical solutions, the accuracy and efficiency of pollution tracing and governance have been significantly improved through multi-scale inversion and intelligent hierarchical response mechanism: the multi-scale model combines micro-positioning, meso-path analysis and macro-space-time laws to achieve high-precision dynamic tracking of pollution sources; dynamic confidence assessment and collaborative inversion ensure positioning reliability, while the hierarchical response strategy automatically triggers targeted control measures according to the type and severity of pollution, greatly shortening the decision-making time and improving the governance effect.

[0018] In a second aspect, the present application provides a multi-class environmental parameter automatic prediction system based on a GIS map, adopting the following technical solutions: A multi-class environmental parameter automatic prediction system based on a GIS map, comprising: A multi-dimensional spatio-temporal matrix generation module that collects data on noise intensity, vibration spectrum, sewage turbidity, light intensity, and PM2.5 concentration in a target area in real time, transmits the data to a GIS platform through a distributed sensor network, synchronously obtains data on the real-time changes in traffic flow, the spatio-temporal distribution of keywords in the social media environment, and the short-term fluctuations in meteorological parameters in the target area, and aligns the environmental parameters and dynamic behavior data in space and time in a preset format based on the GIS spatial grid technology for generating a multi-dimensional spatio-temporal matrix that combines physical parameters and social behaviors, where the physical parameters include data on noise intensity, vibration spectrum, sewage turbidity, light intensity, and PM2.5 concentration, and the social behaviors include data on the real-time changes in traffic flow, the spatio-temporal distribution of keywords in the social media environment, and the short-term fluctuations in meteorological parameters; A collaborative impact weight output module that extracts the energy distribution characteristics of the main frequency band of noise and the spectral energy distribution characteristics of a vibration sensor based on the multi-dimensional spatio-temporal matrix, calculates the energy overlap degree of the energy distribution characteristics and the spectral energy distribution characteristics, and marks it as a high-risk collaborative pollution area when the energy overlap degree exceeds the overlap degree threshold; obtains the sewage flow, real-time wind speed, precipitation data, and collaborative pollution characteristics corresponding to the high-risk collaborative pollution area, predicts the pollution diffusion path and the change in concentration gradient in combination with the sewage flow, real-time wind speed, precipitation data, and collaborative pollution characteristics, and uses a spatio-temporal graph attention network to dynamically learn the non-linear association between parameters and outputs the collaborative impact weight; An adaptive prediction model optimization module that constructs an adaptive prediction model including a physical-driven channel and a behavior-driven channel according to the collaborative impact weight, where the physical channel generates a benchmark prediction result based on a noise attenuation equation and CFD fluid simulation, and the behavior channel generates a prediction correction factor by analyzing the temporal sequence correlation of traffic flow and social data through an LSTM network. When the deviation between the physical prediction and the measured data exceeds the data deviation threshold for multiple consecutive periods, the weight of the physical channel is reduced and the weight of the behavior channel is increased. At the same time, a digital twin sandbox is started in the GIS to simulate the pollution source migration scenario to generate virtual training data for optimizing the adaptive prediction model; A matching module that, based on the optimized adaptive prediction model, reversely infers the pollution diffusion path of the exceeded standard area and locates the target device in combination with sensor data; retrieves the device feature matching model, performs wavelet packet decomposition on the noise and vibration signals, and uses it to match the vibration characteristics in the device feature fingerprint library. When the matching is successful, a control instruction is generated and sent to the IoT terminal device to perform operations such as shutting down the exceeded standard pump or starting a noise reduction barrier; An environmental protection compliance report generation module, according to the execution results of the IoT terminals and real-time monitoring data, opens an AR augmented reality verification interface, retrieves the predicted data and measured data of the GIS platform, generates a deviation heat map based on the predicted data and the measured data, and obtains the corresponding virtual-real data deviation. When the virtual-real data deviation exceeds 20%, it triggers the federated learning framework to aggregate the local gradients of edge nodes to update the global prediction model, and retains historical versions to construct a model evolution tree for anomaly backtracking analysis; it is used to generate corresponding environmental protection compliance reports, and the environmental protection compliance reports include details of exceeded parameters, regulatory correlation analysis, control effect evaluation, executable suggestions and blockchain-certified data.

[0019] In a third aspect, the present application provides a method for automatically predicting multiple types of environmental parameters based on a GIS map, adopting the following technical solution: A method for automatically predicting multiple types of environmental parameters based on a GIS map, including a processor, and a program of the method for automatically predicting multiple types of environmental parameters based on a GIS map as described in any one of the above is run in the processor.

[0020] In a fourth aspect, the present application provides a storage medium, adopting the following technical solution: A storage medium stores a program of the method for automatically predicting multiple types of environmental parameters based on a GIS map as described in any one of the above.

[0021] In summary, the present application includes at least one of the following beneficial technical effects: First of all, through the real-time fusion and intelligent modeling of multi-source heterogeneous data, a dynamic decision-making closed loop for environmental governance is constructed, significantly improving the accuracy of pollution co-control risk identification and positioning; through high-frequency band and low-frequency band weighted analysis, multi-scale inversion models and dynamic weight adjustment mechanisms, the pollution source can be quickly locked and the diffusion path can be accurately predicted, avoiding misjudgments caused by traditional methods relying on single parameters or static thresholds, enabling targeted allocation of governance resources, reducing the redundancy of large-area inspections, and greatly shortening the pollution positioning time.

[0022] Secondly, through intelligent hierarchical response and adaptive model optimization, the transformation of environmental governance from passive response to active regulation is realized. The multi-scale inversion combined with dynamic confidence evaluation ensures the reliability of pollution source positioning. The hierarchical control strategy automatically triggers targeted measures (such as noise reduction barriers, pump shutdowns) according to the pollution type (such as noise, sewage turbidity) and severity, significantly shortening the decision-making response time; at the same time, the model continuously iterates through digital twin simulation and federated learning to ensure prediction accuracy and timeliness, reducing human intervention errors.

[0023] Finally, through a closed-loop verification and continuous optimization mechanism, the long-term effectiveness of governance strategies is ensured. AR augmented reality feedback and deviation heat map analysis can correct model deviations in real time, while federated learning and the model evolution tree mechanism enable dynamic updates of the prediction model to adapt to environmental changes; this fully automated closed-loop process significantly improves the intelligent level of environmental governance, making decision-making more accurate and execution more efficient, ultimately achieving targeted, dynamic, and sustainable pollution prevention and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flowchart of a method for automatically predicting multiple types of environmental parameters based on a GIS map shown according to an exemplary embodiment.

[0025] Figure 2 is a structural block diagram of a system for automatically predicting multiple types of environmental parameters based on a GIS map shown according to an exemplary embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following describes in detail the embodiments of the present application, and the examples of the embodiments are shown in the drawings.

[0027] In the description of this specification, the description with reference to the terms "certain embodiments", "one embodiment", "some embodiments", "schematic embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiments or examples. Moreover, the specific features, structures, materials, or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0028] An embodiment of the present application discloses a method for automatically predicting multiple types of environmental parameters based on a GIS map. Refer to Figure 1 , including: S100, collect data on noise intensity, vibration spectrum, sewage turbidity, light intensity, and PM2.5 concentration in the target area in real time, transmit it to the GIS platform through a distributed sensor network, synchronously obtain data on real-time changes in traffic flow, spatio-temporal distribution of environmental keywords in social media, and short-term fluctuations in meteorological parameters in the target area, and align the environmental parameters and dynamic behavior data in space and time in a preset format based on the GIS space grid technology to generate a multi-dimensional spatio-temporal matrix integrating physical parameters and social behaviors.

[0029] Among them, the physical parameters include data on noise intensity, vibration spectrum, sewage turbidity, light intensity, and PM2.5 concentration, and the social behaviors include data on real-time changes in traffic flow, spatio-temporal distribution of environmental keywords in social media, and short-term fluctuations in meteorological parameters.

[0030] It should be noted here that in the embodiments of the present application, the corresponding parameters are obtained through sensor network deployment or other means, as follows: Noise sensor: Use a sound level meter (such as a MEMS microphone array) to collect the noise intensity in real time, and a spectrum analyzer to extract the vibration spectrum; Water quality sensor: A conductivity sensor monitors the turbidity of sewage, and a pH sensor assists in judging water quality anomalies; Environmental parameter sensor: PM2.5 sensor (such as laser scattering type), light intensity sensor (photoresistor or photodiode); Traffic flow: The traffic flow is counted in real time through a geomagnetic sensor or a camera (combined with image recognition); Social media data: The spatio-temporal distribution of environmental keywords (such as "noise" and "pollution") in the area is captured through an API interface; Meteorological data: Integrate the API of the meteorological bureau to obtain short-term fluctuation data such as wind speed and precipitation intensity.

[0031] For spatio-temporal alignment and gridification, first perform the corresponding GIS grid division, that is, divide the target area into square grids of 50m×50m, and each grid corresponds to a spatial unit; then design the time slice, which can be a time slice of every 5 minutes to ensure the real-time nature of data collection and analysis. That is to say, the preset format is a square grid of 50m×50m and a time slice of every 5 minutes.

[0032] In the data alignment method in the embodiments of the present application, for data with different sampling frequencies (such as sensors collecting data per second and social media updating per minute), the time resolution is unified through interpolation or the sliding window averaging method; in addition, the spatial alignment method is achieved by merging sensor data into the grid to which they belong, and mapping social media data to the nearest grid according to the user's GPS location.

[0033] Finally, for the matrix dimension, the matrix dimension includes the spatial dimension (grid ID), the time dimension (time slice), and the parameter dimension (physical + social behavior parameters); standardize the parameters with different dimensions (such as noise dB, PM2.5 μg / m³) (such as Z-score standardization).

[0034] By unifying the modeling of the physical environment and social behavior data, for example, the correlation of abnormal increase in PM2.5 concentration during peak traffic hours is discovered; while the 50m×50m grid can accurately locate the pollution source (such as the noise source of a certain construction site), and the 5-minute time slice can capture short-term pollution events (such as sudden sewage leakage).

[0035] S200. Based on the energy distribution characteristics of the main frequency bands of noise extracted from the multi-dimensional spatio-temporal matrix and the spectral energy distribution characteristics of the vibration sensor, calculate the energy overlap degree between the energy distribution characteristics and the spectral energy distribution characteristics. When the energy overlap degree exceeds the overlap degree threshold, mark it as a high-risk collaborative pollution area; obtain the sewage flow rate, real-time wind speed, precipitation data, and collaborative pollution characteristics corresponding to the high-risk collaborative pollution area, and predict the pollution diffusion path and concentration gradient change by combining the sewage flow rate, real-time wind speed, precipitation data, and collaborative pollution characteristics. Use the spatio-temporal graph attention network to dynamically learn the non-linear association between parameters and output the collaborative influence weight. Among them, in the embodiment of the present application, the overlap degree threshold is 70%.

[0036] It should be noted here that the method further includes: S211. Extract the energy distribution characteristics of the main frequency bands of noise through short-time Fourier transform. The energy distribution characteristics include the main frequency energy peak value, the energy proportion of the frequency band, and the slope of the energy distribution curve.

[0037] Among them, in the embodiment of the present application, the noise sensor collects the noise intensity in real time through a sound level meter (such as a MEMS microphone array), and uses a spectrum analyzer to extract the vibration spectrum. In order to accurately capture high-frequency components (such as mechanical equipment vibration noise), short-time Fourier transform (STFT) is used for analysis: select a 1-second analysis window, smooth the edges through a Hamming window to reduce spectrum leakage, and set the window overlap rate to 50% to balance time and spectrum resolution.

[0038] It should be noted here that the key features include: the main frequency energy peak value, locate the frequency with the maximum energy in the STFT spectrum (such as 160 Hz), which reflects the dominant frequency of the noise source; the energy proportion of the frequency band, the proportion of the energy of the high-frequency band (150 - 200 Hz) and the low-frequency band (50 - 100 Hz) in the total energy, which is used to distinguish equipment noise from environmental noise; the slope of the energy distribution curve, perform linear fitting on the spectral curves of the high-frequency and low-frequency bands, and calculate the slope difference to identify abnormal pollution sources (such as high-speed rotating equipment).

[0039] S212. Extract the spectral energy distribution characteristics of the vibration sensor through vibration spectrum analysis. The spectral energy distribution characteristics include the vibration energy peak frequency, the energy fluctuation amplitude of the frequency band, and the spectral energy density.

[0040] Among them, in the embodiment of the present application, the time-domain signal collected by the vibration sensor (such as an accelerometer) is converted into a frequency-domain energy distribution through fast Fourier transform (FFT). In order to meet the Nyquist sampling theorem, the sampling rate is set to 2 times the highest target frequency (such as a 200 Hz target frequency corresponding to a 400 Hz sampling rate).

[0041] It should be noted here that the key features include: the peak frequency of vibration energy, locating the maximum energy value in the FFT spectrum (such as 180 Hz), which reflects the inherent vibration frequency of the device; the amplitude of energy fluctuation in the frequency band, the standard deviation of the energy in the high-frequency band (150 - 200 Hz), which is used to evaluate the operation stability of the device; the spectral energy density, the energy value within the unit frequency band, which is used for the correlation analysis with the noise spectrum.

[0042] S213, calculate the dynamic overlap degree between the energy distribution of the main frequency band of noise and the energy distribution of the vibration spectrum, including: standardize the frequency-domain energy distribution curve of the energy distribution characteristics and the frequency-domain energy distribution curve of the spectrum energy distribution characteristics, and use the weighted cosine similarity calculation formula, where the weight of the high-frequency band is higher than that of the low-frequency band. Among them, 150 - 200 Hz is the high-frequency band with a weight of 0.6, and 50 - 100 Hz is the high-frequency band with a weight of 0.4; when the dynamic overlap degree exceeds the overlap degree threshold and the difference in the proportion of high-frequency band energy is less than the difference threshold of the proportion of high-frequency band energy, it is marked as a high-risk co-pollution area.

[0043] Among them, in the embodiment of the present application, the frequency-domain energy distribution curves of noise and vibration are standardized (such as normalized to the range of [0, 1]) to eliminate sensor differences, and the weighted cosine similarity can be used to calculate the similarity between the high-frequency band (weight 60%) and the low-frequency band (weight 40%), and the total overlap degree is the weighted sum of the two.

[0044] It should be noted here that the difference threshold of the proportion of high-frequency band energy is 15%. That is to say, the threshold constraints include: the dynamic overlap degree > 70%, indicating that the noise and vibration energy distributions are highly similar and may originate from the same pollution source (such as device abnormality); the difference in the proportion of high-frequency band energy < 15%, ensuring the consistency of the high-frequency band and avoiding sensor noise or environmental interference. For example, if the proportions of high-frequency band energy of noise and vibration in a certain area are 40% and 45% respectively, and the overlap degree reaches 75%, it is marked as a "high-risk co-pollution area".

[0045] Precisely identify high-risk pollution areas through multi-dimensional feature extraction and dynamic correlation analysis: First, use the short-time Fourier transform (STFT) and Hamming window to capture the high-frequency components of the noise signal, extract the peak main frequency energy, the proportion of energy in frequency bands, and the slope of the energy distribution curve to quantify the noise spectrum characteristics. Second, extract the peak vibration energy frequency, the amplitude of energy fluctuation in frequency bands, and the spectral energy density through FFT analysis of vibration sensors to locate the equipment vibration source and evaluate its stability. Finally, through standardization processing and weighted cosine similarity calculation (weight of 60% for high-frequency bands and 40% for low-frequency bands), dynamically evaluate the similarity of the noise and vibration energy distributions. Combining the dual-threshold constraints of a dynamic overlap degree > 70% and a difference in high-frequency band energy proportion < 15%, exclude environmental noise interference, precisely mark high-risk collaborative pollution areas (such as abnormal equipment operation), achieve rapid positioning of pollution sources and suppression of misjudgments, and improve the accuracy and timeliness of environmental governance.

[0046] In addition, in the process of dynamically learning the non-linear correlation between parameters using the spatio-temporal graph attention network, the method further includes: S214, construct the node feature matrix of the spatio-temporal graph. The node feature matrix includes environmental parameters, dynamic behavior parameters, and spatial correlation parameters. Among them, the environmental parameters include noise intensity, peak vibration spectral energy, sewage turbidity, PM2.5 concentration, and light intensity. The dynamic behavior parameters include traffic flow density, frequency of keyword mentions on social media, wind speed, and precipitation intensity. The spatial correlation parameters include predicted values of pollution diffusion paths and concentration gradient change values of adjacent grids.

[0047] Among them, in the embodiments of the present application, multi-source data is integrated through the node feature matrix, specifically including: environmental parameters, noise intensity, peak vibration spectral energy, sewage turbidity (through conductivity sensors), PM2.5 concentration (laser scattering sensors), light intensity; dynamic behavior parameters, traffic flow (counted by geomagnetic sensors or cameras), frequency of keyword mentions of "pollution" on social media (captured by APIs), wind speed and precipitation intensity (meteorological bureau APIs); spatial correlation parameters, predicted values of pollution diffusion paths and concentration gradient change values of adjacent grids.

[0048] S215, introduce a dual-channel attention mechanism in the graph attention network. Among them, the spatial channel calculates the spatial attention weight based on the grid adjacency relationship, and the time channel calculates the time attention weight based on the autocorrelation of the time series; when the noise intensity is lower than the noise lower threshold, the weight of the environmental parameters of the corresponding grid is dynamically reduced to 30% of the original weight.

[0049] It should be noted here that for the dual-channel attention mechanism: in the spatial channel, weights are calculated based on grid adjacency relationships (such as 4-neighborhood), for example, higher weights are assigned when the PM2.5 concentration suddenly increases in adjacent grids; in the temporal channel, weights are calculated based on the autocorrelation of historical data, for example, higher weights are assigned when the traffic flow surges in the past 1 hour. The noise threshold is 50 dB. In terms of dynamic weight adjustment, when the grid noise intensity is lower than 50 dB, the weights of environmental parameters are dynamically reduced to 30% of the original value to reduce the interference of minor vibrations or environmental noise.

[0050] Then, for high-risk area marking and pollution diffusion path prediction, it is necessary to combine the dynamic overlap degree and the high-frequency band ratio difference threshold in the third step to mark high-risk areas. It should be noted here that the spatio-temporal graph attention network outputs the contribution degrees of various parameters to pollution diffusion (such as the weight of noise to PM2.5 diffusion), and integrates parameters such as sewage flow and wind speed, and predicts the pollution diffusion path and concentration change through the spatio-temporal graph convolutional network (ST-GCN). For example, if the wind speed is towards the residential area, the model predicts that the PM2.5 concentration will increase in this area, guiding the precise allocation of governance resources.

[0051] In addition, multi-dimensional data is obtained through the deployment of sensor networks: physical environment data, such as noise sensors, water quality sensors, PM2.5 sensors, etc.; social behavior data, such as traffic flow, social media keywords, and meteorological data.

[0052] Among them, in the embodiment of this application, the GIS grid division divides the area into 50m×50m square grids, and each grid corresponds to a spatial unit; the time slice is designed to be updated every 5 minutes to ensure real-time performance. It should be noted here that in terms of data alignment: for time alignment, for the sensor data per second and the social media data per minute, interpolation or sliding window averaging is used to unify the time resolution; for spatial alignment, the sensor data is merged into the grid to which it belongs, and the social media data is mapped to the nearest grid according to the user's GPS. The dimensions of the multi-dimensional spatio-temporal matrix include space (grid ID), time (5 minutes), and parameters (normalized physical + social behavior parameters), for example, the Z-score normalization is used to process the dimensional difference between noise dB and PM2.5 μg / m³.

[0053] It should be noted here that in the spatio-temporal graph attention network, the non-linear correlation between parameters is dynamically captured through the dual-channel attention mechanism (spatial channel and temporal channel), and the collaborative influence weights of various parameters on pollution diffusion are output. The specific implementation is as follows: First is the spatial channel attention mechanism, which calculates spatial weights based on grid adjacency relationships (such as 4-neighborhood or 8-neighborhood). For example, by analyzing the similarity of node features in adjacent grids (such as the similarity of parameters like noise intensity, PM2.5 concentration, etc.), and combining with the adjacency relationship matrix (defining the physical distance between grids or the pollution diffusion path), the influence of neighboring grids on the current grid is weighted. If the PM2.5 concentration in an adjacent grid suddenly increases, the model will assign it a higher spatial weight to identify the pollution diffusion direction. The spatial weight is dynamically calculated through the Graph Attention Layer, specifically: performing softmax normalization on the product of the node feature similarity and the adjacency relationship matrix to obtain the spatial correlation strength between adjacent grids.

[0054] Then is the temporal channel attention mechanism, which calculates temporal weights based on the autocorrelation of time series. For example, by analyzing the correlation between historical data (such as past 1-hour traffic flow, wind speed changes, etc.) and current pollution parameters, and combining with the temporal decay coefficient (such as an exponential decay factor to ensure that recent data has a greater impact on the prediction), the dynamic association of parameters over time is captured. If a sudden increase in traffic flow in a certain area leads to an increase in PM2.5 concentration, the model will assign a higher temporal weight to the historical traffic data. The temporal weight is dynamically calculated through the Temporal Attention Layer, specifically: performing softmax normalization on the product of the historical data correlation and the temporal decay coefficient to determine the contribution degree of historical data to the current prediction.

[0055] In a low-noise scenario (such as when the grid noise intensity < 50 dB), the weights of environmental parameters (such as the peak vibration energy, sewage turbidity) are dynamically reduced to 30% of their original values to suppress the interference of small environmental noises or sensor errors on the model. For example, if the noise intensity of a certain grid is 45 dB, the weight of its vibration energy parameter will be reduced from 0.5 to 0.15, reducing the noise interference of irrelevant parameters on the model.

[0056] In addition, the network fuses the spatial and temporal attention weights through multi-layer graph convolution (Graph Convolution), and finally outputs the collaborative influence weights of each parameter on pollution diffusion. For example, the model may output "the contribution degree of noise to PM2.5 diffusion is 40%" or "the weight of wind speed to sewage leakage diffusion is 60%". These weights are used to quantify the non-linear association between parameters: when the high-frequency energy overlap degree between noise and vibration is relatively high, the model will enhance their collaborative weights for pollution source localization; while for parameters with low correlation (such as light intensity), the weights are significantly suppressed.

[0057] Furthermore, by combining parameters such as the collaborative influence weight, sewage flow rate, and wind speed, the spatio-temporal graph convolutional network (ST-GCN) is used to predict the pollution diffusion path and the change of concentration gradient. For example, if the wind direction points to the residential area and the collaborative weight shows a high correlation between PM2.5 and traffic flow, the model will predict an increase in the PM2.5 concentration in this area and generate a diffusion path map to guide the precise allocation of treatment resources.

[0058] Through the spatial and temporal attention mechanisms, the implicit relationships between complex environmental parameters (such as the non-linear dependence between wind speed and PM2.5 diffusion) can be identified, rather than relying on predefined rules; irrelevant parameter interference can be suppressed in low-noise scenarios, improving the model's sensitivity to real pollution sources (such as accurately distinguishing equipment vibration from environmental noise); the output weights provide an interpretable basis for pollution control, for example, "70% of the pollution in a certain area is caused by abnormal equipment vibration", guiding precise control; moreover, with a high resolution of 5-minute time slices and 50m×50m grids, combined with the spatio-temporal modeling ability of ST-GCN, the diffusion path of short-term pollution events (such as sudden leaks) can be quickly predicted, shortening the treatment response time.

[0059] S300, construct an adaptive prediction model that includes a physically driven channel and a behaviorally driven channel based on the collaborative influence weight. The physical channel generates a benchmark prediction result based on the noise attenuation equation and CFD fluid simulation. The behavioral channel analyzes the temporal correlation between traffic flow and social data through an LSTM network to generate a prediction correction factor. When the deviation between the physical prediction and the measured data exceeds the data deviation threshold for multiple consecutive periods, the weight of the physical channel is reduced and the weight of the behavioral channel is increased; at the same time, a digital twin sandbox is launched in GIS to simulate the pollution source migration scenario to generate virtual training data to optimize the adaptive prediction model.

[0060] First, the noise attenuation equation for constructing the physically driven channel is based on acoustic principles and describes the propagation and attenuation process of noise in the air through a mathematical model. Among them, the considerations include distance, obstacles, reflection surfaces, etc. Then there is the CFD fluid simulation. Computational Fluid Dynamics (CFD) is used to simulate the diffusion path of PM2.5 particles in the air, and factors such as wind speed and temperature gradient need to be considered for their influence on pollutant diffusion.

[0061] By providing a benchmark prediction result for pollution diffusion, it can provide a reference basis for subsequent corrections, and the accuracy based on physical laws helps to understand how environmental variables affect pollutant diffusion.

[0062] Next is to construct a behavior-driven channel. The long short-term memory network (LSTM) is a special type of recurrent neural network (RNN) that is good at processing time series data. In the embodiments of this application, it is used here to analyze the changing trends and their correlations of traffic flow and the frequency of social media keyword mentions over time. The generated prediction correction factor is a time series pattern derived from historical data analysis, which adjusts the results of the physical prediction model to reflect the actual impact of human activities on pollution diffusion.

[0063] Considering the impact of social behavior on pollution diffusion, such as the increase in PM2.5 concentration during traffic peak hours, the prediction accuracy of the model is improved by introducing social behavior data, especially in such application scenarios in urban environments.

[0064] When the deviation between the physical prediction and the measured data exceeds the data deviation threshold for multiple consecutive periods. In the embodiments of this application, the data deviation threshold is 15%, which can also be modified. Here, multiple periods can be 3 periods or more. The weight of the physical channel is reduced and the weight of the behavior channel is increased. For example, the system automatically reduces the weight of the physical channel by 0.3 and correspondingly increases the weight of the behavior channel to 0.7. This adjustment mechanism aims to optimize the prediction accuracy of the model and reduce the errors caused by the incomplete conformity of the physical model assumptions with the actual situation. Thus, the adaptability and flexibility of the model can be improved, especially in complex and changeable urban environments, and it can also ensure that the prediction results are closer to the actual situation and enhance the effectiveness of decision support.

[0065] Among them, when starting the digital twin sandbox simulation, the digital twin sandbox simulation is started in the GIS, that is, a virtual environment is created using the Geographic Information System (GIS) platform to simulate the migration of pollution sources in the real world. Then, by simulating the pollution diffusion process under different scenarios, a large amount of virtual data is collected to further optimize the adaptive prediction model. Thus, a safe and controllable environment is provided to test and improve the model.

[0066] It should be noted here that during the process of starting the digital twin sandbox simulation of the pollution source migration scenario, the method further includes: S310, constructing a multi-physical field coupling simulation model. The multi-physical field coupling simulation model includes an air pollution diffusion module and a ground vibration diffusion module. The air pollution diffusion module simulates the air propagation paths of PM2.5 and noise based on CFD fluid simulation, and the ground vibration diffusion module simulates the vibration energy attenuation in high-risk areas based on the vibration wave equation.

[0067] Among them, for the air pollution diffusion module, computational fluid dynamics (CFD) technology is needed to accurately simulate the propagation paths of PM2.5 particles and noise in the atmosphere. The air pollution diffusion module takes into account the influence of meteorological factors such as wind speed, temperature gradient, humidity, etc. on pollutant diffusion; for complex environment modeling, it includes topographic features such as buildings and terrain undulations, as well as the influence of factors such as vegetation cover on the airflow and pollutant transmission paths. Through fine grid division and boundary condition setting, ensure that the simulation results are as close to the actual situation as possible. It should be noted here that not only the diffusion of a single pollutant is simulated, but also the interaction between different pollutants, such as chemical reactions or synergistic effects, is analyzed to provide a more comprehensive environmental impact assessment.

[0068] For the ground vibration diffusion module, it is based on the vibration wave equation in elastodynamics to simulate how the vibration energy in the high-risk area decays over time and space, and the influence of factors such as geological structure and soil type on vibration propagation is also taken into account. In order to distinguish high-frequency vibrations (such as vibrations generated by mechanical equipment) and low-frequency vibrations (such as seismic waves), separate modeling and analysis are carried out to accurately capture the energy distribution and propagation characteristics of different types of vibrations.

[0069] That is to say, by combining noise and vibration data, evaluate the potential environmental pollution sources in the area and their impact on the surrounding environment, providing a scientific basis for accurately locating the pollution source; thus providing a simulation platform that integrates multiple physical phenomena, which can consider the impact of air pollution and ground vibration simultaneously, is crucial for evaluating the comprehensive impact of pollution sources in a specific area, helps to formulate more effective environmental protection strategies, thereby enhancing the understanding of the pollutant diffusion mechanism under complex environmental conditions and improving the accuracy of prediction.

[0070] S320, introduce a spatio-temporal attenuation factor into the GIS grid, and the spatio-temporal attenuation factor is calculated based on the exponential decay function of distance and time.

[0071] Among them, the spatio-temporal attenuation factor is calculated according to the exponential decay function of distance and time. As the distance increases or time passes, the pollutant concentration gradually decreases; specifically, for each GIS grid cell, its pollutant concentration will be adjusted according to the distance from the pollution source and the length of time since the pollution occurred. According to historical monitoring data and experimental research results, determine appropriate attenuation coefficients; these coefficients reflect the diffusion rate and attenuation law of different pollutants under different environmental conditions.

[0072] Applying the spatiotemporal attenuation factor to the GIS grid system enables the pollutant concentration value of each grid unit to reflect the actual diffusion situation; it not only takes into account the direct diffusion path, but also includes the impact of obstacles, terrain and other factors on the pollutant transmission path; thus, it can more accurately simulate the change pattern of pollutants over time and space, especially the long-distance transmission effect, which helps to identify potential pollution hotspots and provide scientific decision-making support for environmental management departments.

[0073] S330: When the deviation between the simulated path and the measured pollution diffusion path exceeds the path deviation threshold, the sandbox is triggered to adaptively adjust the migration probability distribution parameters, including: increasing the contribution weight of wind speed to pollutant diffusion by 20%, and recalculating the nonlinear relationship between precipitation intensity and pollutant deposition rate based on historical data.

[0074] Among them, in the embodiment of the present application, the path deviation threshold is 10%. When the deviation between the simulated path and the measured pollution diffusion path exceeds 10%, the system automatically starts the adaptive adjustment mechanism; first, the system will identify the main cause of the deviation, which may be that certain key environmental variables (such as wind speed, precipitation intensity) fail to accurately reflect the actual situation.

[0075] For example, if wind speed is found to be one of the main factors causing the deviation, the system will automatically increase the weight of wind speed's contribution to pollutant diffusion by 20% to better match the actual observation data.

[0076] Based on the latest measured data, the nonlinear relationship between precipitation intensity and pollutant deposition rate is recalculated; this process may involve the application of machine learning algorithms to mine complex patterns hidden in large amounts of data. After each adjustment, the system records the performance indicators before and after the adjustment and conducts comparative analysis. In this way, the model parameters are continuously optimized to better adapt to changing environmental conditions.

[0077] S400, based on the optimized adaptive prediction model, reverses the pollution diffusion path of the exceeding area and locates the target equipment in combination with sensor data; calls the equipment feature matching model, performs wavelet packet decomposition on the noise and vibration signals, and matches the vibration features in the equipment feature fingerprint library. When the match is successful, a control instruction is generated and sent to the IoT terminal device to execute the operation of shutting down the exceeding pump or starting the noise reduction barrier.

[0078] Among them, first, through the optimized adaptive prediction model, combined with real-time monitoring data in the GIS grid (such as PM2.5 concentration, noise intensity, etc.), the areas with excessive pollutant concentrations are quickly identified; then, for the areas with excessive concentrations, the pollution diffusion path is inversely deduced. The physical driving channels of pollution diffusion (such as the noise attenuation equation and CFD fluid simulation) and behavioral driving channels (such as the influence of traffic flow and social media data) can be used to inversely deduce the diffusion path of pollutants from the source to the area with excessive concentrations. Specifically: through CFD simulation, the influence of factors such as wind speed and terrain on pollutant transmission is analyzed inversely; with the help of the spatio-temporal graph attention network, the correlation between the pollution source and the area with excessive concentrations in historical data is analyzed to further verify the accuracy of the inversely deduced path.

[0079] It should be noted here that since pollution diffusion may be affected by various factors (such as wind direction change, precipitation deposition, etc.), the system will generate multiple possible diffusion paths and perform probability ranking based on sensor data and model prediction results to finally determine the most likely pollution source.

[0080] Through the above steps, the potential pollution sources in the areas with excessive pollution can be quickly locked, providing a direction for subsequent precise treatment. Moreover, the process of inversely deducing the path not only relies on physical laws but also combines social behavior data, improving the accuracy and reliability of tracing. In addition, the multi-path possibility evaluation mechanism ensures that the optimal solution can be found even in a complex environment, avoiding errors that may be brought by single-path inference.

[0081] Then, the data of noise sensors, vibration sensors, and other environmental sensors (such as PM2.5 sensors) in the areas with excessive concentrations are fused and processed. Through dynamic overlap calculation (such as high-frequency band similarity > 70% and energy ratio difference < 15%), high-risk collaborative pollution areas are screened out. In the high-risk areas, combined with the noise and vibration spectrum characteristics collected by the sensors, the scope is further narrowed down to locate specific equipment or facilities (such as pumps, generators, etc.). For example: the vibration sensor detects a specific frequency vibration signal generated when a certain equipment is operating; the noise sensor captures the high-frequency noise emitted when the equipment is operating. By analyzing the vibration intensity and noise spectrum distribution of the equipment, it is judged whether it is in an abnormal operating state (such as too high speed or bearing wear), so as to confirm whether it is the main source of pollution.

[0082] Furthermore, in the implementation process of the equipment feature matching model, the method also includes: S410, perform three-layer wavelet packet decomposition on the noise and vibration signals to extract the vibration characteristics in the high-frequency band, middle-frequency band, and low-frequency band.

[0083] Among them, first, the noise and vibration signals of the target device are decomposed by three - layer wavelet packet decomposition. Each layer of decomposition further divides the signal into finer frequency bands. After three - layer decomposition, the original signal is divided into eight sub - frequency bands, corresponding to high - frequency band, medium - frequency band and low - frequency band respectively.

[0084] High - frequency band: Usually contains sharp noise and vibration signals generated during device failure or abnormal operation, such as vibrations caused by bearing wear or imbalance; Medium - frequency band: Reflects the main vibration frequencies under normal operating conditions of the device, such as vibrations related to motor speed; Low - frequency band: Captures environmental noise and background vibrations, and may also include some low - frequency mechanical fault signals.

[0085] Wavelet packet decomposition can provide information in both time domain and frequency domain simultaneously, and is especially suitable for the analysis of non - stationary signals, which helps to extract subtle features of the device operating state; Through three - layer decomposition, more detailed spectral information can be obtained, providing a rich data basis for subsequent feature matching.

[0086] Within each sub - frequency band, characteristic parameters such as the energy distribution and main frequency peak of the vibration signal are calculated. For example: Energy distribution, calculate the total energy within each frequency band and its proportion in the total energy of the entire signal; Main frequency peak, identify the main vibration frequency and its corresponding amplitude within each frequency band; Other features, features such as spectral shape and spectral slope can also be extracted to comprehensively describe the characteristics of the vibration signal.

[0087] By extracting characteristic parameters in different frequency bands, it helps to distinguish different operating states of the device, especially abnormal operating states; These characteristic parameters will be used as key inputs in the subsequent matching process to ensure the accuracy and reliability of the matching results.

[0088] S420, perform weighted voting matching on the decomposed vibration characteristics and the device fingerprint library.

[0089] Among them, calculate the similarity between the extracted vibration characteristics and the feature vectors in the device fingerprint library. Use cosine similarity or other distance measurement methods to calculate the similarity scores for the high - frequency band, medium - frequency band and low - frequency band respectively; According to the importance of each frequency band, set different weights. Specifically, the similarity weight for the high - frequency band is 0.5, the medium - frequency band is 0.3, and the low - frequency band is 0.2. Multiply the similarity scores of each frequency band by their corresponding weights, and then sum to obtain the total similarity score.

[0090] The weighted voting mechanism takes into account the importance differences of different frequency bands in device fault diagnosis, improving the accuracy of matching; And by setting reasonable weights, it can better balance the sensitivity of the high - frequency band and the stability of the low - frequency band, ensuring the robustness of the matching results.

[0091] S430. When the similarity of all three frequency bands exceeds the similarity threshold and the weighted total similarity ≥ the weighted total similarity threshold, it is determined that the match is successful; when the matching failure rate continuously exceeds the matching failure rate threshold, the fingerprint database incremental update mechanism is triggered, and the vibration data of the preset similar time period is automatically retrieved, and the fingerprint database feature vector is updated using an online learning algorithm.

[0092] It should be noted here that in the embodiment of the present application, the similarity threshold is 80%, the weighted total similarity threshold is 85%, the matching failure rate threshold is 20%, and the preset similar time period is the last 30 days.

[0093] That is to say, when the similarity of all three frequency bands exceeds 80% and the weighted total similarity ≥ 85%, it is determined that the match is successful. Specifically: the high-frequency band similarity > 80%, the middle-frequency band similarity > 80%, the low-frequency band similarity > 80%, and the weighted total similarity ≥ 85% Strict matching conditions are set to ensure that a match is only determined to be successful when all frequency bands are highly similar, thus avoiding misjudgment; by comprehensively considering the similarity of each frequency band, the credibility of the matching result is improved, and the occurrence probability of false positives or false negatives is reduced.

[0094] The system continuously monitors the success rate of device feature matching. If the matching failure rate continuously exceeds 20%, the fingerprint database incremental update mechanism is triggered. At the same time, the vibration data of the preset similar time period is automatically retrieved, and these data contain the actual vibration conditions of the device under different operating states.

[0095] By using an online learning algorithm (such as adaptive boosting, incremental PCA, etc.) to process the newly collected data, new feature vectors are generated and added to the existing device fingerprint database; by continuously updating the feature vectors in the fingerprint database, the model can adapt to changes in the device operating state and improve the accuracy of long-term prediction and matching.

[0096] The fingerprint database incremental update mechanism can dynamically adjust the model to make it adapt to changes in the device operating state and extend the validity period of the model; in addition, the application of the online learning algorithm enables the fingerprint database to be gradually optimized without interrupting the system operation, improving the flexibility and adaptability of the system.

[0097] S500, based on the execution results of the IoT terminal and the real-time monitoring data, opens the AR augmented reality verification interface, retrieves the predicted data and measured data of the GIS platform, generates a deviation heat map based on the predicted data and the measured data, and obtains the corresponding virtual-real data deviation. When the virtual-real data deviation exceeds 20%, it triggers the federated learning framework to aggregate the local gradients of the edge nodes to update the global prediction model, and retains the historical versions to construct a model evolution tree for anomaly backtracking analysis; generates the corresponding environmental compliance report, which includes details of exceeded parameters, regulatory correlation analysis, control effect evaluation, executable suggestions, and blockchain-certified data.

[0098] Among them, based on the Internet of Things (IoT) platform, an augmented reality (AR) verification interface is developed and opened, which allows users to view and interact with the pollution dispersion prediction results and actual measurement data through mobile devices or dedicated hardware; integrates the execution results from the IoT terminal (such as the change in noise level after shutting down the over-standard pump) and the real-time monitoring data provided by the GIS platform into the AR environment, enabling users to intuitively see the comparison between the prediction and the actual situation.

[0099] Extract the pollution dispersion path prediction data (including pollutant concentration distribution, wind direction influence, etc.) from the GIS platform, and compare it with the data from real-time monitoring sensors (such as PM2.5 concentration, noise intensity, etc.); then calculate the difference between the predicted value and the measured value to generate a deviation heat map. This step not only considers the numerical difference but also includes the consistency analysis of the spatial distribution. The deviation heat map helps to identify areas with larger prediction errors for targeted optimization.

[0100] Among them, using GIS tools, the deviation between the predicted data and the measured data is displayed in the form of a heat map. Different colors represent different degrees of deviation, with red indicating high deviation areas and green indicating low deviation areas; in addition to the visual heat map, these deviations also need to be quantified, that is, obtain the specific percentage of virtual-real data deviation.

[0101] If the deviation heat map shows that the virtual-real data deviation in some areas exceeds 20%, the federated learning framework is triggered. Each edge node (such as a sensor network distributed in different geographical locations) trains the model based on its local data and uploads the trained gradients to the central server, which is responsible for aggregating the gradient information of all edge nodes and updating the global prediction model. Among them, it is necessary to retain the historical versions before each update and construct a model evolution tree for future anomaly backtracking analysis.

[0102] Using federated learning technology, the model accuracy is improved without sharing the original data, protecting privacy while enhancing prediction accuracy, and the model evolution tree records the historical trajectory of the model development, which helps to understand and analyze the changing trend of the model performance.

[0103] Generate a detailed environmental compliance report based on the latest prediction results, measured data, and deviation analysis. The report content includes: details of exceeded parameters, listing the specific types of pollutants that exceed the standard and their concentrations; regulatory correlation analysis, comparing with current environmental protection laws and regulations to point out which regulations are violated; control effect evaluation, evaluating the actual effect of the implemented control measures on improving the pollution situation; actionable suggestions, putting forward specific improvement suggestions for the current situation; blockchain-certified data, using blockchain technology to store key data in the report to ensure data immutability and enhance the credibility of the report.

[0104] In the embodiment of the present application, during the process of reverse inferring the pollution diffusion path in the exceeded standard area based on the optimized adaptive prediction model, the method further includes: First, construct a multi-scale spatio-temporal inversion model, and divide the pollution diffusion path into micro-scale, meso-scale, and macro-scale.

[0105] Among them, use the sensor grid data of 50m×50m to accurately locate the specific coordinates of the pollution source equipment. This step relies on a high-resolution sensor network that can capture subtle changes in a local area; the meso-scale combines environmental data such as traffic flow and wind speed to reverse infer the specific path of the pollutant from the pollution source to the exceeded standard area. This scale takes into account environmental factors in a larger range, such as the influence of wind direction, terrain, etc. on pollutant diffusion; the macro-scale studies the spatio-temporal distribution law of pollution sources in the region by analyzing historical pollution event data. This long-term data analysis helps to identify potential pollution hotspots and trends.

[0106] Then, when positioning the target equipment, adopt a dynamic confidence evaluation mechanism, and calculate the equipment position by weighted confidence of the sensor data, where the confidence is the matching degree of noise intensity and vibration energy; when the confidence is lower than 80%, trigger the collaborative inversion of adjacent grids to improve the positioning accuracy through multi-grid data fusion.

[0107] Among them, assign a confidence score to each sensor grid according to the quality of the sensor data (such as the matching degree of noise intensity and vibration energy). A higher confidence means more reliable data and can be used for more accurate positioning. When the confidence of a certain grid is lower than 80%, the system will trigger the data fusion of adjacent grids to improve the accuracy of the target equipment position. For example, combine the data of multiple adjacent grids to correct the initial estimate value. The dynamic confidence evaluation mechanism ensures that only high-quality data is used for the final positioning decision, reducing the possibility of misjudgment. Through collaborative inversion, the problem of insufficient single-grid data can be compensated for, and the positioning accuracy can be improved.

[0108] Finally, when generating control instructions, a hierarchical response strategy is introduced. When the excessive parameter is noise and the duration exceeds the preset duration threshold, the noise reduction barrier is preferentially activated. When the excessive parameter is sewage turbidity and the concentration exceeds the sewage concentration trigger value, the pump is triggered to shut down and the surrounding sewage treatment facilities are linked to start the purification process.

[0109] Among them, the duration threshold is 1 hour, and the sewage concentration trigger value is twice the set threshold. When it is detected that the noise exceeds the standard and the duration exceeds 1 hour, the noise reduction barrier is preferentially activated. This method can effectively reduce the noise level without interrupting the operation of the equipment and reduce the impact on the surrounding environment. If it is monitored that the sewage turbidity exceeds the standard and the concentration exceeds twice the set threshold, the relevant pump is immediately shut down, and the surrounding sewage treatment facilities are linked to start the purification process. This step aims to quickly prevent the spread of pollution and take measures to purify the affected area.

[0110] The hierarchical response strategy takes targeted measures according to different types of pollution situations, improving the response efficiency and effect. It not only ensures the ability to respond quickly in emergency situations but also takes into account the characteristics of different pollution types, achieving precise governance.

[0111] Based on the above steps, the entire solution can not only quickly and accurately identify the pollution source and locate the target equipment but also take the most appropriate control measures according to the specific situation, thus effectively managing and controlling environmental pollution problems.

[0112] An embodiment of this application discloses a multi-class environmental parameter automatic prediction system based on a GIS map. Refer to Figure 2 , including: A multi-dimensional spatio-temporal matrix generation module 001, which collects data on noise intensity, vibration spectrum, sewage turbidity, light intensity, and PM2.5 concentration in the target area in real time, transmits them to the GIS platform through a distributed sensor network, and synchronously obtains data on the real-time changes in traffic flow, the spatio-temporal distribution of environmental keywords in social media, and the short-term fluctuations in meteorological parameters in the target area. Based on the GIS space grid technology, the environmental parameters and dynamic behavior data are aligned in space and time in a preset format for generating a multi-dimensional spatio-temporal matrix that integrates physical parameters and social behaviors. Among them, the physical parameters include data on noise intensity, vibration spectrum, sewage turbidity, light intensity, and PM2.5 concentration, and the social behaviors include data on the real-time changes in traffic flow, the spatio-temporal distribution of environmental keywords in social media, and the short-term fluctuations in meteorological parameters. The collaborative impact weight output module 002 extracts the energy distribution characteristics of the main noise frequency band and the spectral energy distribution characteristics of the vibration sensor based on the multi-dimensional spatio-temporal matrix, calculates the energy overlap degree between the energy distribution characteristics and the spectral energy distribution characteristics, and marks it as a high-risk collaborative pollution area when the energy overlap degree exceeds the overlap degree threshold; obtains the sewage flow, real-time wind speed, precipitation data, and collaborative pollution characteristics corresponding to the high-risk collaborative pollution area, predicts the pollution diffusion path and concentration gradient change in combination with the sewage flow, real-time wind speed, precipitation data, and collaborative pollution characteristics, and uses the spatio-temporal graph attention network to dynamically learn the non-linear association between parameters and output the collaborative impact weight; The adaptive prediction model optimization module 003 constructs an adaptive prediction model including a physical-driven channel and a behavior-driven channel according to the collaborative impact weight. The physical channel generates a benchmark prediction result based on the noise attenuation equation and CFD fluid simulation. The behavior channel analyzes the time-series correlation between traffic flow and social data through the LSTM network to generate a prediction correction factor. When the deviation between the physical prediction and the measured data exceeds the data deviation threshold for multiple consecutive periods, the weight of the physical channel is reduced and the weight of the behavior channel is increased. At the same time, the digital twin sandbox simulation of the pollution source migration scenario is started in the GIS to generate virtual training data for optimizing the adaptive prediction model; The matching module 004 reversely infers the pollution diffusion path for the exceeded standard area based on the optimized adaptive prediction model, and locates the target device in combination with the sensor data; retrieves the device feature matching model, performs wavelet packet decomposition on the noise and vibration signals, and uses it to match the vibration characteristics in the device feature fingerprint library. When the matching is successful, a control instruction is generated and sent to the IoT terminal device to execute the operation of shutting down the exceeded standard pump or starting the noise reduction barrier; The environmental protection compliance report generation module 005 opens the AR augmented reality verification interface according to the execution result of the IoT terminal and the real-time monitoring data, retrieves the predicted data and measured data of the GIS platform, generates a deviation heat map based on the predicted data and measured data, and obtains the corresponding virtual-real data deviation. When the virtual-real data deviation exceeds 20%, it triggers the federated learning framework to aggregate the local gradients of the edge nodes to update the global prediction model, and retains the historical version to construct a model evolution tree for anomaly backtracking analysis; used to generate the corresponding environmental protection compliance report, and the environmental protection compliance report includes details of exceeded standard parameters, regulatory correlation analysis, control effect evaluation, executable suggestions, and blockchain deposit data.

[0113] The embodiment of the present application also discloses a multi-class environmental parameter automatic prediction system based on a GIS map, including a processor, and a program of the multi-class environmental parameter automatic prediction method based on the GIS map as described in any one of the above is run in the processor.

[0114] The embodiment of the present application also discloses a storage medium, which stores a program of the multi-class environmental parameter automatic prediction method based on the GIS map as described in any one of the above.

[0115] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. An automatic prediction method for multiple types of environmental parameters based on a GIS map, characterized in that, Including: Collecting in real time the noise intensity, vibration spectrum, sewage turbidity, light intensity and PM2.5 concentration data of the target area, transmitting them to the GIS platform through a distributed sensor network, synchronously obtaining the real-time changes in traffic flow, the spatio-temporal distribution of keywords in the social media environment and the short-term fluctuation data of meteorological parameters in the target area, and aligning the environmental parameters and dynamic behavior data in space and time in a preset format based on the GIS space grid technology to generate a multi-dimensional spatio-temporal matrix integrating physical parameters and social behaviors, where the physical parameters include noise intensity, vibration spectrum, sewage turbidity, light intensity and PM2.5 concentration data, and the social behaviors include the real-time changes in traffic flow, the spatio-temporal distribution of keywords in the social media environment and the short-term fluctuation data of meteorological parameters; Extracting the energy distribution characteristics of the main frequency band of noise and the spectral energy distribution characteristics of the vibration sensor based on the multi-dimensional spatio-temporal matrix, calculating the energy overlap degree of the energy distribution characteristics and the spectral energy distribution characteristics, and marking it as a high-risk co-pollution area when the energy overlap degree exceeds the overlap degree threshold; obtaining the sewage flow, real-time wind speed, precipitation data and co-pollution characteristics corresponding to the high-risk co-pollution area, predicting the pollution diffusion path and the change of concentration gradient in combination with the sewage flow, real-time wind speed, precipitation data and co-pollution characteristics, and dynamically learning the non-linear correlation between parameters and outputting the co-influence weight by using a spatio-temporal graph attention network; Constructing an adaptive prediction model including a physical-driven channel and a behavior-driven channel according to the co-influence weight, generating a benchmark prediction result by the physical channel based on the noise attenuation equation and CFD fluid simulation, generating a prediction correction factor by the behavior channel through analyzing the temporal and sequential correlation of traffic flow and social data by the LSTM network, when the deviation between the physical prediction and the measured data exceeds the data deviation threshold for multiple consecutive periods, reducing the weight of the physical channel and increasing the weight of the behavior channel, and at the same time starting a digital twin sandbox in the GIS to simulate the pollution source migration scenario to generate virtual training data to optimize the adaptive prediction model; Based on the optimized adaptive prediction model, inversely inferring the pollution diffusion path of the over-standard area and locating the target device in combination with the sensor data; retrieving the device feature matching model, performing wavelet packet decomposition on the noise and vibration signals, and matching the vibration characteristics in the device feature fingerprint library, and when the matching is successful, generating a control instruction and sending it to the IoT terminal device to execute the operation of shutting down the over-standard pump or starting the noise reduction barrier; According to the execution result of the IoT terminal and the real-time monitoring data, opening an AR augmented reality verification interface, retrieving the prediction data and the measured data of the GIS platform, generating a deviation heat map based on the prediction data and the measured data, and obtaining the corresponding virtual-real data deviation. When the virtual-real data deviation exceeds 20%, triggering the federated learning framework to aggregate the local gradients of the edge nodes to update the global prediction model, and retaining the historical versions to construct a model evolution tree for anomaly backtracking analysis; generating a corresponding environmental compliance report, which includes details of over-standard parameters, regulatory correlation analysis, control effect evaluation, executable suggestions and blockchain evidence data.

2. The automatic prediction method for multiple types of environmental parameters based on a GIS map according to claim 1, wherein, The method further includes: Extract the energy distribution characteristics of the main frequency band of noise through short-time Fourier transform. The energy distribution characteristics include the peak value of the main frequency energy, the proportion of the energy in the frequency band, and the slope of the energy distribution curve. Extract the spectral energy distribution characteristics of the vibration sensor through vibration spectrum analysis. The spectral energy distribution characteristics include the peak frequency of the vibration energy, the amplitude of the energy fluctuation in the frequency band, and the spectral energy density. Calculate the dynamic overlap degree between the energy distribution of the main frequency band of noise and the spectral energy distribution of vibration, including: standardize the frequency-domain energy distribution curves of the energy distribution characteristics and the spectral energy distribution characteristics, and use the weighted cosine similarity calculation formula, where the weight of the high-frequency band is higher than that of the low-frequency band. When the dynamic overlap degree exceeds the overlap degree threshold and the difference in the proportion of high-frequency band energy is less than the high-frequency band energy proportion difference threshold, it is marked as a high-risk collaborative pollution area.

3. The automatic prediction method for multiple types of environmental parameters based on a GIS map according to claim 2, characterized in that, In the process of using the spatio-temporal graph attention network to dynamically learn the non-linear association between parameters, the method further includes: Construct a node feature matrix of the spatio-temporal graph. The node feature matrix includes environmental parameters, dynamic behavior parameters, and spatial association parameters. Among them, the environmental parameters include noise intensity, peak value of vibration spectral energy, sewage turbidity, PM2.5 concentration, and light intensity. The dynamic behavior parameters include traffic flow density, frequency of keyword mentions in social media, wind speed, and precipitation intensity. The spatial association parameters include the predicted value of the pollution diffusion path and the concentration gradient change value of adjacent grids. Introduce a dual-channel attention mechanism in the graph attention network, where the spatial channel calculates the spatial attention weight based on the grid adjacency relationship, and the time channel calculates the time attention weight based on the autocorrelation of the time series. When the noise intensity is lower than the noise lower threshold, dynamically reduce the weight of the environmental parameters of the corresponding grid to 30% of the original weight.

4. The automatic prediction method for multiple types of environmental parameters based on a GIS map according to claim 3, characterized in that, In the process of starting the digital twin sandbox to simulate the pollution source migration scenario, the method further includes: Construct a multi-physical field coupling simulation model. The multi-physical field coupling simulation model includes an air pollution diffusion module and a ground vibration diffusion module. The air pollution diffusion module simulates the air propagation path of PM2.5 and noise based on CFD fluid simulation. The ground vibration diffusion module simulates the vibration energy attenuation in the high-risk area based on the vibration wave equation. Introduce a spatio-temporal attenuation factor in the GIS grid. The spatio-temporal attenuation factor is calculated based on the exponential decay function of distance and time. When the deviation between the simulated path and the measured pollution diffusion path exceeds the path deviation threshold, trigger the sandbox to adaptively adjust the migration probability distribution parameters, specifically including: increasing the contribution weight of the wind speed to pollutant diffusion by 20%, and recalculating the non-linear relationship between precipitation intensity and pollutant sedimentation rate according to historical data.

5. The automatic prediction method for multiple types of environmental parameters based on a GIS map according to claim 4, wherein, In the process of implementing the device feature matching model, the method further includes: Perform three-layer wavelet packet decomposition on the noise and vibration signals to extract the vibration characteristics in the high-frequency band, middle-frequency band, and low-frequency band. Perform weighted voting matching on the decomposed vibration characteristics and the device fingerprint library. It is determined that the matching is successful when the similarities of all three frequency bands exceed the similarity threshold and the weighted total similarity ≥ the weighted total similarity threshold; when the matching failure rate continuously exceeds the matching failure rate threshold, the fingerprint database incremental update mechanism is triggered, and the vibration data of the preset similar time period is automatically retrieved, and the fingerprint database feature vector is updated using the online learning algorithm.

6. The automatic prediction method for multiple types of environmental parameters based on a GIS map according to claim 5, characterized in that, During the process of inversely inferring the pollution diffusion path of the exceeding-standard area based on the optimized adaptive prediction model, the method further includes: Construct a multi-scale spatio-temporal inversion model, and divide the pollution diffusion path into micro-scale, meso-scale and macro-scale. Among them, the micro-scale locates the precise coordinates of the pollution source equipment based on the data of the sensor grid, the meso-scale inversely infers the propagation path of pollutants from the pollution source to the exceeding-standard area by combining traffic flow and wind speed data, and the macro-scale analyzes the spatio-temporal distribution law of regional pollution sources through historical pollution event data; When positioning the target device, a dynamic confidence evaluation mechanism is adopted, and the device position is calculated by weighted calculation according to the confidence of the sensor data, where the confidence is the matching degree of the noise intensity and the vibration energy; when the confidence is lower than 80%, the collaborative inversion of adjacent grids is triggered, and the positioning accuracy is improved through multi-grid data fusion; When generating the control instruction, a hierarchical response strategy is introduced. When the exceeding-standard parameter is noise and the duration exceeds the preset duration threshold, the noise reduction barrier is preferentially started. When the exceeding-standard parameter is the sewage turbidity and the concentration exceeds the sewage concentration trigger value, the pump is triggered to close and the surrounding sewage treatment facilities are linked to start the purification process.

7. A multi-class environmental parameter automatic prediction system based on a GIS map, characterized in that, Including: A multi-dimensional spatio-temporal matrix generation module, which real-time collects the noise intensity, vibration spectrum, sewage turbidity, light intensity and PM2.5 concentration data of the target area, transmits them to the GIS platform through a distributed sensor network, and synchronously obtains the real-time changes of the traffic flow in the target area, the spatio-temporal distribution of keywords in the social media environment and the short-term fluctuation data of meteorological parameters. Based on the GIS space grid technology, the environmental parameters and dynamic behavior data are aligned in space and time in a preset format, and are used to generate a multi-dimensional spatio-temporal matrix that integrates physical parameters and social behaviors, where the physical parameters include noise intensity, vibration spectrum, sewage turbidity, light intensity and PM2.5 concentration data, and the social behaviors include the real-time changes of traffic flow, the spatio-temporal distribution of keywords in the social media environment and the short-term fluctuation data of meteorological parameters; A collaborative influence weight output module, which extracts the energy distribution characteristics of the main frequency band of noise and the spectral energy distribution characteristics of the vibration sensor based on the multi-dimensional spatio-temporal matrix, calculates the energy overlap degree of the energy distribution characteristics and the spectral energy distribution characteristics, and marks it as a high-risk collaborative pollution area when the energy overlap degree exceeds the overlap degree threshold; obtains the sewage flow, real-time wind speed, precipitation data and collaborative pollution characteristics corresponding to the high-risk collaborative pollution area, predicts the pollution diffusion path and the change of concentration gradient by combining the sewage flow, real-time wind speed, precipitation data and collaborative pollution characteristics, and uses the spatio-temporal graph attention network to dynamically learn the non-linear correlation between parameters and outputs the collaborative influence weight; Adaptive prediction model optimization module, which constructs an adaptive prediction model including a physical-driven channel and a behavior-driven channel according to the collaborative influence weight. The physical channel generates a benchmark prediction result based on the noise attenuation equation and CFD fluid simulation. The behavior channel analyzes the temporal correlation between traffic flow and social data through an LSTM network to generate a prediction correction factor. When the deviation between the physical prediction and the measured data exceeds the data deviation threshold for multiple consecutive periods, the weight of the physical channel is reduced and the weight of the behavior channel is increased. At the same time, a digital twin sandbox simulation of the pollution source migration scenario is started in the GIS to generate virtual training data for optimizing the adaptive prediction model; Matching module, based on the optimized adaptive prediction model, inversely deduces the pollution diffusion path of the over-standard area, and locates the target device in combination with sensor data; retrieves the device feature matching model, performs wavelet packet decomposition on the noise and vibration signals, and is used to match the vibration features in the device feature fingerprint library. When the matching is successful, a control instruction is generated and sent to the IoT terminal device to execute the operation of shutting down the over-standard pump or starting the noise reduction barrier; Environmental compliance report generation module, according to the execution results of the IoT terminal and the real-time monitoring data, opens an AR augmented reality verification interface, retrieves the prediction data and the measured data of the GIS platform, generates a deviation heat map based on the prediction data and the measured data, and obtains the corresponding virtual-real data deviation. When the virtual-real data deviation exceeds 20%, it triggers the federated learning framework to aggregate the local gradients of the edge nodes to update the global prediction model, and retains the historical versions to construct a model evolution tree for anomaly backtracking analysis; used to generate the corresponding environmental compliance report, and the environmental compliance report includes details of over-standard parameters, regulatory correlation analysis, control effect evaluation, executable suggestions and blockchain deposit data.

8. An automatic prediction system for multiple types of environmental parameters based on a GIS map, characterized in that, It includes a processor, and a program of the multi-class environmental parameter automatic prediction method based on the GIS map as described in any one of claims 1-6 runs in the processor.

9. A storage medium, characterized in that, Stores a program of the multi-class environmental parameter automatic prediction method based on the GIS map as described in any one of claims 1-6.

Images