Regional pollution process real-time monitoring regulation and control system and method based on artificial intelligence

Abstract

The invention relates to a regional pollution process real-time monitoring, regulation and control system and method based on artificial intelligence. The system is composed of a data acquisition module, a communication module, an artificial intelligence analysis module, an intelligent regulation and control module, an execution terminal and a user interaction module. The data acquisition module acquires multi-source data such as pollutant concentration and meteorological parameters by means of a multi-modal sensor, the data are processed and transmitted by the data communication module, and the artificial intelligence analysis module realizes pollution source tracing and accurate prediction by using technologies such as a space-time diagram convolutional network and an LSTM-Transform hybrid neural network. The intelligent regulation and control module generates a regulation and control strategy based on an NSGA-II algorithm, and the execution terminal is responsible for implementation. The user interaction module provides a visual interface and a manual intervention channel. The method comprises the steps of data acquisition and processing, model construction and prediction, regulation and control strategy formulation and execution and feedback optimization closed-loop operation. According to the invention, comprehensive real-time monitoring and accurate regulation and control of regional pollution are realized, the prediction accuracy is improved, the environmental, economic and social benefits are balanced, and an efficient technical means is provided for regional pollution treatment.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial pollution control, specifically to a system and method for real-time monitoring and control of regional pollution processes based on artificial intelligence. Background Art

[0002] With the rapid development of industrialization and urbanization, regional environmental pollution has become increasingly serious, posing a serious threat to the ecological environment and human health. Traditional regional pollution monitoring and control methods have many limitations.

[0003] In terms of monitoring, previous monitoring equipment was often sparsely distributed and had a single function, making it impossible to obtain comprehensive, real-time, multi-dimensional pollution data. For example, it could only monitor the concentrations of a few pollutants, making it difficult to cover emerging pollutants such as volatile organic compounds (VOCs). Data collection closely related to pollution, such as meteorological parameters and geospatial information, was also incomplete, resulting in a one-sided understanding of the pollution situation. Furthermore, traditional monitoring data transmission methods are inefficient, susceptible to interference, and prone to frequent data delays and loss, making it impossible to provide timely and accurate data support for real-time control.

[0004] In the past, regulatory strategies relied heavily on experience, lacking scientific precision. Traditional regulatory decisions failed to fully consider the balance between environmental capacity, economic costs, and social benefits. When implementing measures like industrial production restrictions and traffic controls, the lack of quantitative analysis often led to poor regulatory results, either failing to effectively improve environmental quality or causing significant economic impacts. Furthermore, the implementation of regulatory measures lacked effective oversight and feedback mechanisms, making it difficult to adjust and optimize them based on actual conditions, making it difficult to achieve the desired results in regional pollution control. Summary of the Invention

[0005] The purpose of the present invention is to provide a real-time monitoring and control system and method for regional pollution processes based on artificial intelligence to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a real-time monitoring and control system for regional pollution processes based on artificial intelligence, comprising:

[0007] Data acquisition module: It is composed of a multimodal sensor array distributed in the monitoring area. The array integrates lidar, differential absorption spectrometer, and micro sensor node equipment to collect PM2.5 data in real time. 2.5 , VOCs, SO2, NOx pollutant concentration data, as well as temperature, humidity, wind speed, wind direction meteorological parameters, while obtaining geographic spatial information and pollution source emission intensity data;

[0008] Data communication module: Adopting a hybrid 5G / 6G and LoRa networking architecture, the high speed and low latency of 5G / 6G are used to quickly upload monitoring data. The long distance and low power consumption of LoRa are used to dynamically schedule edge computing nodes. Before data transmission, the collected data is preliminarily encoded to ensure the accuracy and stability of data transmission.

[0009] Artificial intelligence analysis module: including,

[0010] Pollution source tracing submodule: This module builds a pollutant diffusion model based on a spatiotemporal graph convolutional network. It uses geospatial information, meteorological parameters, and pollutant concentration data as inputs. Through the convolutional layers and graph convolution operations in the model, it achieves three-dimensional precise positioning of pollution sources.

[0011] Dynamic prediction submodule: This module uses an LSTM-Transformer hybrid neural network architecture. LSTM is responsible for long-term and short-term memory learning of time series data, while the Transformer's multi-head attention mechanism is used to integrate the features of multi-source monitoring data to generate pollution evolution forecasts for the next 12-72 hours.

[0012] Intelligent control module: including,

[0013] Multi-objective optimization unit: Based on the NSGA-II algorithm, a pollutant control decision-making model under environmental capacity constraints is constructed. This model comprehensively considers pollution prediction results, environmental capacity constraints, economic costs, and social benefits. Through non-dominated sorting and congestion calculation, a set of Pareto optimal control solutions is generated.

[0014] Dynamic response unit: Based on real-time prediction results, it selects appropriate solutions from the set of solutions generated by multi-objective optimization and generates a set of control instructions, including industrial production restriction plans, traffic control strategies, and environmental governance facility operating parameter adjustment instructions;

[0015] Execution terminal module: Consists of an intelligent valve controller, a swarm of drones, and a variable information release system. The intelligent valve controller automatically adjusts the valve opening at the pollution source according to control instructions. The drone swarm, equipped with monitoring and treatment equipment, conducts aerial monitoring and local treatment of polluted areas. The variable information release system provides real-time information on pollution status and control measures to the public.

[0016] User Interaction Module: Provides a WebGL-based 3D pollution situation visualization interface, supporting multi-dimensional data comparison and analysis; at the same time, it sets up a manual intervention channel, allowing managers to manually adjust the control strategy according to actual conditions.

[0017] Furthermore, the multimodal sensor array in the data acquisition module has a self-calibration function; before each data acquisition cycle begins, the sensor automatically performs an internal calibration procedure, generates a known physical quantity signal through a built-in standard signal source, and compares it with the sensor's measurement result; if there is a deviation, the system corrects the measurement data according to a pre-set calibration algorithm; for PM 2.5 The sensor's measurement accuracy is regularly calibrated using standard particulate matter samples, and the sensor's sensitivity parameters are adjusted based on the calibration results. This self-calibration function ensures that the sensor maintains high measurement accuracy during long-term operation, reduces measurement errors caused by sensor aging and changes in environmental factors, and provides a reliable data basis for subsequent data processing and analysis, thereby improving the accuracy of the entire system in monitoring and controlling regional pollution processes.

[0018] Furthermore, the data communication module has a data caching and retransmission mechanism; when the network is congested or the signal is interrupted, the data communication module will temporarily store the collected data in a local cache device. The cache device uses high-speed flash memory technology and has the characteristics of large-capacity storage and fast reading and writing; at the same time, the system starts a retransmission timer. Once the network returns to normal, the older data in the cache will be retransmitted in priority according to the time sequence of data storage to ensure the integrity and continuity of the data; in the case of unstable 5G / 6G network signals, the monitoring data can be temporarily stored in the cache and retransmitted in order after the signal is restored; this mechanism effectively avoids data loss and ensures that the pollution monitoring data can be transmitted to the subsequent processing module in a timely and complete manner, providing comprehensive data support for pollution prediction and control strategy formulation.

[0019] Furthermore, the pollution source tracing submodule in the artificial intelligence analysis module introduces a particle filter algorithm for auxiliary positioning when performing three-dimensional positioning of pollution sources; after the spatiotemporal graph convolutional network outputs the preliminary pollution source positioning results, the particle filter algorithm optimizes the positioning results according to the dynamic changes of the monitoring data; the particle filter algorithm represents the possible location of the pollution source by randomly sampling a large number of particles in the state space, updates the weights of the particles according to the similarity between the measured data and the predicted state of the particles, and then eliminates particles with lower weights through the resampling process, retains and copies particles with higher weights, and gradually approaches the actual location of the pollution source; in a complex urban environment, when there are multiple possible pollution sources and the spread of pollutants is interfered with by multiple factors, the particle filter algorithm can more accurately lock the pollution source, improve the accuracy of pollution source tracing, and provide strong support for precise pollution control;

[0020] Furthermore, the dynamic prediction submodule takes seasonal and cyclical factors into account when generating pollution evolution forecasts; by analyzing historical data, it extracts the patterns of seasonal changes in the concentrations of different pollutants and the characteristics of weekly and monthly cyclical fluctuations; these seasonal and cyclical characteristics are used as additional input features and input into the LSTM-Transformer hybrid neural network together with multi-source monitoring data; when predicting pollution conditions during the winter heating period, the accuracy of the pollution evolution trend forecast is improved by combining the historical pattern of increased pollutant emissions during this period and the unfavorable meteorological conditions for pollutant diffusion; this helps to formulate more targeted pollution control strategies in advance, rationally arrange environmental governance resources, and effectively respond to pollution problems in different seasons and cycles.

[0021] Furthermore, the multi-objective optimization unit in the intelligent control module takes public health risk assessment into consideration when constructing the control decision-making model; by establishing a quantitative relationship model between pollutant concentration and public health impact, it evaluates the health risks that the public may face under different control schemes; the health risk assessment results are used as an independent objective function to participate in the multi-objective optimization process together with the environmental capacity, economic cost and social benefit objective functions; when formulating industrial production restriction plans, not only the improvement of environmental quality and the control of economic costs are considered, but also the degree of reduction in public health risks is fully evaluated; this makes control decisions pay more attention to the public interest, protect the health of residents, and enhance the comprehensive benefits of regional pollution control.

[0022] Furthermore, the drone swarm in the execution terminal module has the ability to autonomously avoid obstacles and work collaboratively; the drones are equipped with advanced lidar and visual sensors to perceive obstacle information in the surrounding environment in real time; when an obstacle is detected, the drone automatically plans a new flight path through the built-in obstacle avoidance algorithm to ensure flight safety; at the same time, the drone swarm exchanges real-time information through wireless communication technology and works collaboratively according to task requirements; when monitoring large-scale polluted areas, multiple drones can fly in a preset formation mode to achieve synchronous monitoring of different areas and improve monitoring efficiency; when performing pollution control tasks, the drone swarm can work collaboratively, some are responsible for spraying control agents, and some are responsible for monitoring the control effects, thereby improving the overall effect of pollution control.

[0023] Furthermore, the three-dimensional pollution situation visualization interface of the user interaction module supports virtual reality (VR) and augmented reality (AR) interaction modes; after wearing VR or AR devices, users can immersively view the regional pollution status, freely shuttle in the virtual environment, and observe the distribution of pollutant concentrations from different angles; through gesture recognition and voice control technology, users can naturally interact with virtual scenes, zoom in or out of specific areas, query data from specific monitoring points, and switch the display of pollution data at different times; in AR mode, users can combine virtual pollution information with real scenes to intuitively understand the pollution situation of the surrounding environment; this interaction method provides managers and scientific researchers with more intuitive and convenient data analysis and decision support means, which helps to more deeply understand pollution problems and formulate more effective regulation strategies.

[0024] A method for real-time monitoring and control of regional pollution processes based on artificial intelligence, comprising the following steps:

[0025] Monitoring network construction and data collection: Deploy a data collection module consisting of a multimodal sensor array. Each sensor collects real-time pollutant concentration data, meteorological parameters, geospatial information, and pollution source emission intensity data within the target area at preset time intervals.

[0026] Data processing and feature extraction: The data communication module uses a federated learning framework to extract privacy-protected features from distributed monitoring data. It also extracts features from local data at each edge computing node, reducing data transmission volume and privacy leakage risks. Furthermore, it uses an attention mechanism to clean up abnormal data from collected data and highlight key data features.

[0027] Model construction and dynamic optimization: The AI analysis module establishes a spatiotemporal coupled deep reinforcement learning model, modeling the pollution diffusion process as a Markov decision process. Based on the output of the pollution diffusion model and actual monitoring data, the sampling frequency of monitoring nodes is dynamically optimized to improve monitoring efficiency and data quality.

[0028] Pollution prediction and control strategy generation: The pollution source tracing submodule in the artificial intelligence analysis module uses a pollutant diffusion model constructed based on a spatiotemporal graph convolutional network to trace pollution sources and achieve three-dimensional positioning of pollution sources. The dynamic prediction submodule uses an LSTM-Transformer hybrid neural network architecture to predict pollution evolution trends in the next 12-72 hours. The intelligent control module uses the NSGA-II algorithm to construct a multi-objective optimization model under environmental capacity constraints, combines it with the fuzzy hierarchical analysis method to determine the optimal control combination scheme and generate a control instruction set.

[0029] Control execution and effect feedback: The intelligent valve controller, drone swarm, and variable information release system in the execution terminal module execute corresponding operations according to the control instruction set; the control execution effect is monitored in real time, and feedback data after execution is collected, including changes in pollutant concentrations and improvements in environmental quality;

[0030] Model updating and closed-loop control: Utilize online transfer learning algorithms to continuously update the parameters of the prediction model in the artificial intelligence analysis module based on feedback data; evaluate the prediction accuracy and control effect, adjust the model and optimize the control strategy based on the evaluation results, and form a closed-loop control loop of "monitoring-prediction-control-verification".

[0031] Furthermore, an adaptive control strategy adjustment mechanism is introduced into the closed-loop control loop of "monitoring-prediction-control-verification"; when the system finds that the control measures fail to achieve the expected goals based on the feedback of the execution effect, the system automatically analyzes the reasons, predicts model errors, and the control measures are not implemented in place; the prediction model error leads to the system to conduct more in-depth optimization of the prediction model based on the feedback data, adjust the model parameters or structure, and solve the problem of control measure execution. The system re-evaluates the feasibility and effectiveness of the control strategy, and adjusts the intensity and execution time of the control instructions according to the actual situation; when it is found that the pollution concentration in a certain area is not declining at the expected rate, the system analyzes that the industrial production restrictions may be insufficient, and then increases the production restriction ratio or extends the production restriction time; this adaptive mechanism enables the system to respond to complex and changeable pollution situations more flexibly, continuously optimize the control strategy, and improve the effect and efficiency of regional pollution control.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] Through distributed multimodal sensor arrays and advanced data communication technology, it is possible to comprehensively collect various pollution data and related environmental information and achieve low-latency transmission. PM 2.5 The concentration data of various pollutants such as VOCs and meteorological parameters provide a rich and accurate data basis for timely grasp of regional pollution dynamics, greatly improving the comprehensiveness and timeliness of monitoring.

[0034] Leveraging artificial intelligence technologies such as spatiotemporal graph convolutional networks and LSTM-Transformer hybrid neural networks, we achieve precise three-dimensional location of pollution sources and accurate prediction of pollution evolution over the next 12-72 hours. Compared to traditional methods, this method can more accurately pinpoint pollution sources and predict pollution trends in advance, providing a scientific basis for developing targeted prevention and control measures and improving the accuracy of pollution control.

[0035] A multi-objective optimization model based on the NSGA-II algorithm comprehensively considers factors such as environmental capacity, economic costs, and social benefits to generate scientific and reasonable control strategies. Furthermore, control instructions are dynamically adjusted based on real-time prediction results, such as accurately formulating industrial production restrictions and traffic control strategies. This effectively controls pollution while minimizing negative economic and social impacts, achieving intelligent and efficient control of regional pollution.

[0036] The execution terminal module automates the execution of control commands through intelligent valve controllers, drone swarms, and other devices, and monitors control results in real time. Integrating feedback data, it utilizes an online transfer learning algorithm to continuously optimize model parameters, forming a closed-loop control system of "monitoring-prediction-control-verification." This approach continuously improves system performance, ensuring timely adjustment of control strategies based on actual conditions and consistently enhancing regional pollution control effectiveness.

[0037] The user interaction module provides a WebGL-based 3D pollution situation visualization interface, supporting multi-dimensional data comparison and analysis, VR and AR interaction, and a manual intervention channel. This provides managers and researchers with an intuitive and convenient interactive experience, enabling them to gain a deeper understanding of pollution conditions, combine their professional knowledge with decision-making, and improve the scientific and rationality of their decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Schematic diagram of the method flow of the present invention;

[0039] Figure 2 Schematic diagram of the system flow of the present invention. DETAILED DESCRIPTION

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0041] See also Figure 1 —2. The present invention provides a system and method for real-time monitoring and control of regional pollution processes based on artificial intelligence. The implementation methods are as follows: Hardware of the system for real-time monitoring and control of regional pollution processes based on artificial intelligence

[0042] Data acquisition module

[0043] In target monitoring areas, multimodal sensor arrays are strategically deployed based on factors such as topography, population distribution, and industrial layout. In industrially concentrated areas, sensors monitoring pollutants such as sulfur dioxide, nitrogen oxides, and volatile organic compounds are prioritized. In densely populated urban areas, sensors for fine particulate matter and inhalable particulate matter are prioritized. Furthermore, based on meteorological monitoring needs, meteorological sensors for temperature, humidity, wind speed, and direction are strategically distributed. For example, meteorological monitoring points are established both upwind and downwind of a city to more accurately assess the impact of meteorological conditions on pollutant dispersion.

[0044] Mount the sensor on a stable support or structure to ensure it is protected from external interference and can collect data normally. For outdoor sensors, equip them with waterproof, dustproof, and sun-proof protective casings, and install lightning protection devices to ensure stable operation in adverse weather conditions. For example, in areas with high humidity and salt spray, such as the seaside, choose a corrosion-resistant sensor casing and regularly clean and maintain the sensor.

[0045] According to the changing characteristics of different pollutants and monitoring requirements, set a reasonable data collection frequency. For pollutants that change rapidly, such as PM during peak traffic hours, 2.5 Data is collected every five minutes for nitrogen oxides and nitrogen oxides; for pollutants that change relatively slowly, such as some persistent organic pollutants (POPs), data is collected hourly. Meteorological data is collected every 10-15 minutes, depending on its impact on pollution dispersion.

[0046] Data communication module

[0047] Build a hybrid 5G / 6G and LoRa network architecture. Within the monitoring area, select locations with good signal coverage to install 5G / 6G base stations to ensure that sensor data can be quickly transmitted to the data processing center via the 5G / 6G network. For monitoring points with weak signals or in remote areas, use LoRa gateways for data aggregation and transmission. By configuring LoRa node parameters such as spreading factor and bandwidth, optimize data transmission performance and ensure stable data transmission and low power consumption.

[0048] Use efficient data transmission protocols, such as MQTT. MQTT is lightweight, low-power, and supports a publish / subscribe messaging model, making it suitable for transmitting sensor data. Before transmission, the collected data is compressed to reduce data volume and improve transmission efficiency. Furthermore, setting message quality of service (QoS) ensures reliable transmission of important data.

[0049] Edge computing nodes should be strategically deployed within the monitoring area, such as setting up edge servers near monitoring stations with large data volumes. Edge computing nodes are responsible for preliminary processing of collected data, such as data cleaning, denoising, and feature extraction. This reduces the computational burden on the data processing center while enabling real-time data analysis and rapid response. For example, edge computing nodes could be deployed at monitoring stations in industrial parks to conduct real-time analysis of pollution data within the park and issue timely warnings if abnormal data is detected.

[0050] Execute terminal module

[0051] Install an intelligent valve controller on the pollution source's emission pipeline, ensuring it is tightly connected to the pipeline and accurately controlling the valve opening. After installation, debug the intelligent valve controller and set its control parameters, such as maximum opening, minimum opening, and control accuracy. By communicating with the data processing center, it receives control instructions and achieves precise control of pollution source emissions.

[0052] Select drones with high stability, long flight time, and the ability to carry a variety of monitoring and control equipment. Equip drones with appropriate equipment, such as air quality monitors and pesticide sprayers, based on the needs of the monitoring and control mission. Conduct formation training for drone swarms to ensure they can coordinate operations according to pre-set routes and missions. For example, when monitoring a large polluted area, drone swarms can fly in a triangular formation to achieve simultaneous monitoring of different areas.

[0053] Variable information dissemination systems, such as electronic display screens, are installed in key roads and public spaces within the monitoring area. Connected to a data processing center, these systems acquire real-time information on pollution conditions and control measures, displaying them on the screens for public dissemination. Furthermore, these systems can receive feedback from the public, enabling interaction with the public.

[0054] System software

[0055] Data processing and communication module

[0056] The collected data are cleaned using statistical methods. For outliers, the mean and standard deviation of the data are calculated, and a reasonable threshold range is set. Data exceeding the threshold is considered as outliers and is removed or corrected. For example, for PM 2.5 For concentration data, if a data point deviates from the mean by more than three standard deviations, the data is considered an outlier. At the same time, a sliding average filter algorithm is used to denoise the data, removing noise interference and improving data quality.

[0057] A federated learning framework is deployed on both edge computing nodes and the data processing center. Edge computing nodes extract features from collected data locally, encrypt the extracted feature data, and upload it to the data processing center. The data processing center uses a federated learning algorithm to aggregate and train the feature data uploaded by multiple edge computing nodes without directly accessing the original data, thereby building a global model. The federated learning framework protects data privacy and security while fully leveraging distributed data for model training, improving model accuracy and generalization.

[0058] During data processing, an attention mechanism is introduced to clean up abnormal data. By calculating the importance of data, the attention mechanism highlights key data features and suppresses the impact of abnormal data. For example, when processing multi-source monitoring data, the attention mechanism can automatically assign weights based on the importance of different data types to pollution prediction, allowing the model to focus more on important data features, thereby improving the accuracy and reliability of data processing.

[0059] Artificial intelligence analysis module

[0060] A pollutant diffusion model was constructed based on a spatiotemporal graph convolutional network (ST-GCN). During model training, a large amount of historical pollution data, meteorological data, and geographic information were collected as training samples. The graph convolution operation within the ST-GCN modeled the spatial propagation of pollutants, while temporal convolution was used to capture the temporal trends in pollutant concentrations. After multiple iterations of training, the model was able to accurately locate pollution sources in three dimensions. Furthermore, a particle filter algorithm was used to optimize pollution source tracing results, improving their accuracy.

[0061] Pollution evolution prediction is achieved using a hybrid LSTM-Transformer neural network architecture. The LSTM network is responsible for learning long-term and short-term memory (LSTM) on time series data, capturing long-term dependencies within the data. The Transformer network utilizes a multi-head attention mechanism to fuse and extract features from multi-source monitoring data. During model training, historical pollution data, meteorological data, and pollution source emission data are used as input. A backpropagation algorithm is used to adjust model parameters, enabling the model to accurately predict pollution evolution trends over the next 12-72 hours. Furthermore, considering the impact of seasonal and cyclical factors on pollution, seasonal and cyclical characteristics are used as additional input features to improve prediction accuracy.

[0062] Regularly evaluate the models in the AI analysis module, using metrics such as accuracy, recall, and mean squared error (MSE) to quantitatively assess model performance. Based on the evaluation results, optimize the model. For example, if the model's prediction accuracy is low, the amount of training data can be increased, the model structure can be adjusted, or its parameters can be optimized to improve performance. Furthermore, online transfer learning algorithms are used to continuously update model parameters based on newly collected data, enabling the model to adapt to dynamic changes in pollution conditions.

[0063] Intelligent control module

[0064] A pollutant control decision-making model under environmental capacity constraints was constructed using the NSGA-II algorithm. During the model construction process, multiple objective functions were determined, including environmental capacity, economic costs, and social benefits. The environmental capacity objective function was calculated based on regional environmental quality standards and pollutant dispersion models; the economic cost objective function considered costs related to industrial production restrictions and pollution control facility operation; and the social benefit objective function comprehensively considered factors such as public health and employment. The multiple objective functions were optimized using the NSGA-II algorithm to generate a set of Pareto-optimal control plans.

[0065] Based on real-time prediction results, an appropriate control plan is selected from the set of plans generated by multi-objective optimization, and a set of control instructions is generated. The dynamic response unit further refines and adjusts the control instructions based on actual conditions. For example, when formulating industrial production restrictions, factors such as the company's production plan and equipment operating conditions are considered to rationally arrange the duration and extent of the restrictions. When formulating traffic control strategies, information such as traffic flow and road conditions is combined to determine the sections and timing of the restrictions. Furthermore, based on feedback on the control results, control instructions are promptly adjusted to ensure the effectiveness of the control measures.

[0066] Before implementing a control strategy, digital twin technology is used to create a virtual environment mirror image to rehearse the environmental impacts of different control strategies. By comparing the rehearsal results, the optimal control strategy is selected for implementation. During implementation, the control effects are monitored in real time, and the control strategy is optimized and adjusted based on the feedback data. For example, if the pollution concentration in a certain area is not decreasing as expected, the cause is promptly analyzed and the intensity or implementation time of the control measures are adjusted.

[0067] User interaction module

[0068] A three-dimensional pollution situation visualization interface was developed using WebGL technology. During the interface design process, the geographic information of the monitored area was presented as a three-dimensional map, and pollutant concentration data and meteorological data were displayed using different colors, icons, and animations. Users can freely zoom, rotate, and pan the map using the mouse and keyboard to view pollution conditions in different areas. Multi-dimensional data comparison and analysis capabilities are also provided, allowing users to compare data by time period, pollutant type, and monitoring point to gain a deeper understanding of pollution trends.

[0069] A manual intervention channel is set up in the user interaction module, allowing managers to manually adjust control strategies based on actual conditions. Managers can enter new control instructions through the interface or modify the control plan generated by the system. Manual intervention operations are recorded and compared with the control strategy automatically generated by the system to provide reference for subsequent control decisions. The manual intervention channel also allows for public feedback, incorporating public needs into pollution control decisions.

[0070] A strict data security and permissions management mechanism is established to ensure data security in user interaction modules. User identity authentication and authorization management are implemented, with different operational permissions assigned based on user roles and responsibilities. For example, managers have the ability to view and modify control policies, while ordinary users can only view pollution status information. Furthermore, data encryption technology is used to encrypt transmitted and stored data to prevent data leakage and tampering.

[0071] Before starting the system, perform an initialization check on hardware devices, including the data acquisition module, data communication module, and execution terminal module. Check that sensors are functioning properly, network connections are stable, and that the intelligent valve controller and drone swarm are in standby mode. If any hardware failures are detected, repair or replace them promptly.

[0072] Start the data processing and communication module, artificial intelligence analysis module, intelligent control module, and user interaction module. Initialize the parameters in the software system, such as the initial parameters of the model and data processing thresholds. At the same time, load historical data and pre-trained models to provide data and model support for the normal operation of the system.

[0073] Establish a hardware device status monitoring system to monitor the operating status of hardware devices such as sensors, network equipment, and execution terminals in real time. By monitoring parameters such as voltage, current, and temperature, the system determines whether the device is functioning properly. If any abnormality is detected, an alarm is immediately issued, notifying maintenance personnel for action. For example, if a sensor's temperature is too high, indicating poor heat dissipation or a device malfunction, the system will automatically issue an alarm, prompting maintenance personnel to conduct inspection and repairs.

[0074] Monitor the performance of the software system in real time, including metrics such as data processing speed, model prediction accuracy, and control strategy execution efficiency. By monitoring these metrics, problems in the software system can be promptly identified and optimized. For example, if data processing speed slows, it may be due to excessive data volume or inefficient algorithms. In this case, optimize the data processing algorithm or increase computing resources to increase data processing speed.

[0075] Regularly maintain hardware equipment, including sensor calibration and cleaning, network equipment inspection and maintenance, and terminal repair and maintenance. For example, calibrate sensors every six months to ensure the accuracy of their measurement data. Regularly inspect and maintain the drone fleet, replacing vulnerable parts to ensure normal flight. Furthermore, timely upgrades should be made to hardware based on its lifespan and technological developments to improve system performance and reliability.

[0076] Software systems are regularly upgraded based on technological developments and actual application needs. These upgrades include algorithm optimization, functional expansion, and interface improvements. For example, with the continuous advancement of AI technology, new algorithms can be applied to AI analysis modules to improve the model's predictive accuracy. Based on user feedback, the user interface of user interaction modules can be improved to enhance the user experience. During software system upgrades, data security and integrity must be ensured to prevent data loss or corruption.

[0077] Select industrial cities as application cases.

[0078] System Deployment: Based on the city's geographical characteristics and pollution distribution, hardware equipment such as data acquisition modules, data communication modules, and execution terminal modules will be rationally deployed within the city. Simultaneously, a software system platform will be built, and software systems such as artificial intelligence analysis modules, intelligent control modules, and user interaction modules will be deployed in the data processing center.

[0079] Data collection and analysis: After the system is running, the data collection module collects various pollution data and meteorological data in real time and transmits them to the data processing center through the data communication module. The artificial intelligence analysis module analyzes and processes the collected data to achieve pollution source tracing and pollution evolution trend prediction. For example, through the pollution tracing submodule, it was determined that a chemical company in a certain industrial park was the cause of a recent PM 2.5 The main source of pollution incidents.

[0080] Control Strategy Development and Implementation: Based on the predictions from the AI analysis module, the intelligent control module formulates appropriate control strategies. Production restrictions were implemented for the aforementioned chemical companies, and they were required to strengthen the operation and management of pollution control facilities. Traffic control measures were also implemented along congested urban roads to reduce vehicle exhaust emissions. The execution terminal module automatically adjusted the opening of the intelligent valve controllers based on control instructions to control pollutant emissions from the chemical companies. A swarm of drones conducted aerial monitoring and control of polluted areas. A variable information release system disseminated information on pollution status and control measures to the public.

[0081] Effect Evaluation: After a period of operation, the system's effectiveness was evaluated. Results showed significant improvements in the city's air quality, with pollutant concentrations decreasing to varying degrees. Furthermore, by optimizing and adjusting control strategies, the impact on industrial production and urban transportation was minimized while ensuring environmental quality, achieving a balance between environmental, economic, and social benefits.

Claims

1. A real-time monitoring and control system for regional pollution processes based on artificial intelligence, characterized by: include: Data acquisition module: It is composed of a multimodal sensor array distributed in the monitoring area. The array integrates lidar, differential absorption spectrometer, and micro sensor node equipment to collect PM2.5 data in real time. 2.5 , VOCs, SO2, NOx pollutant concentration data, as well as temperature, humidity, wind speed, wind direction meteorological parameters, while obtaining geographic spatial information and pollution source emission intensity data; Data communication module: Adopting a hybrid 5G / 6G and LoRa networking architecture, the high speed and low latency of 5G / 6G are used to quickly upload monitoring data. The long distance and low power consumption of LoRa are used to dynamically schedule edge computing nodes. Before data transmission, the collected data is preliminarily encoded to ensure the accuracy and stability of data transmission. Artificial intelligence analysis module: include, Pollution source tracing submodule: This module builds a pollutant diffusion model based on a spatiotemporal graph convolutional network. It uses geospatial information, meteorological parameters, and pollutant concentration data as inputs. Through the convolutional layers and graph convolution operations in the model, it achieves three-dimensional precise positioning of pollution sources. Dynamic prediction submodule: This module uses an LSTM-Transformer hybrid neural network architecture. LSTM is responsible for long-term and short-term memory learning of time series data, while the Transformer's multi-head attention mechanism is used to integrate the features of multi-source monitoring data to generate pollution evolution forecasts for the next 12-72 hours. Intelligent control module: include, Multi-objective optimization unit: Based on the NSGA-II algorithm, a pollutant control decision-making model under environmental capacity constraints is constructed. This model comprehensively considers pollution prediction results, environmental capacity constraints, economic costs, and social benefits. Through non-dominated sorting and congestion calculation, a set of Pareto optimal control solutions is generated. Dynamic response unit: Based on real-time prediction results, it selects appropriate solutions from the set of solutions generated by multi-objective optimization and generates a set of control instructions, including industrial production restriction plans, traffic control strategies, and environmental governance facility operating parameter adjustment instructions; Execution terminal module: Consists of an intelligent valve controller, a swarm of drones, and a variable information release system. The intelligent valve controller automatically adjusts the valve opening at the pollution source according to control instructions. The drone swarm, equipped with monitoring and treatment equipment, conducts aerial monitoring and local treatment of polluted areas. The variable information release system provides real-time information on pollution status and control measures to the public. User Interaction Module: Provides a WebGL-based 3D pollution situation visualization interface, supporting multi-dimensional data comparison and analysis; at the same time, it sets up a manual intervention channel, allowing managers to manually adjust the control strategy according to actual conditions.

2. A method for real-time monitoring and control of regional pollution processes based on artificial intelligence, characterized by: Utilizing the system of claim 1, comprising the steps of: Monitoring network construction and data collection: Deploy a data collection module consisting of a multimodal sensor array. Each sensor collects real-time pollutant concentration data, meteorological parameters, geospatial information, and pollution source emission intensity data within the target area at preset time intervals. Data processing and feature extraction: The data communication module uses a federated learning framework to extract privacy-protected features from distributed monitoring data. It also extracts features from local data at each edge computing node, reducing data transmission volume and privacy leakage risks. Furthermore, it uses an attention mechanism to clean up abnormal data from collected data and highlight key data features. Model construction and dynamic optimization: The AI analysis module establishes a spatiotemporal coupled deep reinforcement learning model, modeling the pollution diffusion process as a Markov decision process. Based on the output of the pollution diffusion model and actual monitoring data, the sampling frequency of monitoring nodes is dynamically optimized to improve monitoring efficiency and data quality. Pollution prediction and control strategy generation: The pollution source tracing submodule in the artificial intelligence analysis module uses a pollutant diffusion model constructed based on a spatiotemporal graph convolutional network to trace pollution sources and achieve three-dimensional positioning of pollution sources. The dynamic prediction submodule uses an LSTM-Transformer hybrid neural network architecture to predict pollution evolution trends in the next 12-72 hours. The intelligent control module uses the NSGA-II algorithm to construct a multi-objective optimization model under environmental capacity constraints, combines it with the fuzzy hierarchical analysis method to determine the optimal control combination scheme and generate a control instruction set. Control execution and effect feedback: The intelligent valve controller, drone swarm, and variable information release system in the execution terminal module execute corresponding operations according to the control instruction set; the control execution effect is monitored in real time, and feedback data after execution is collected, including changes in pollutant concentrations and improvements in environmental quality; Model updating and closed-loop control: Utilize online transfer learning algorithms to continuously update the parameters of the prediction model in the artificial intelligence analysis module based on feedback data; evaluate the prediction accuracy and control effect, adjust the model and optimize the control strategy based on the evaluation results, and form a closed-loop control loop of "monitoring-prediction-control-verification".

3. The artificial intelligence-based real-time monitoring and control system for regional pollution processes according to claim 1 is characterized by: The multimodal sensor array in the data acquisition module has a self-calibration function. Before each data acquisition cycle, the sensor automatically performs an internal calibration procedure to generate a known physical quantity signal through a built-in standard signal source and compare it with the sensor's measurement results. If there is a deviation, the system corrects the measurement data according to a pre-set calibration algorithm. 2.5 The sensor's measurement accuracy is regularly calibrated using standard particulate matter samples, and the sensor's sensitivity parameters are adjusted based on the calibration results. This self-calibration function ensures that the sensor maintains high measurement accuracy during long-term operation, reduces measurement errors caused by sensor aging and changes in environmental factors, and provides a reliable data basis for subsequent data processing and analysis, thereby improving the accuracy of the entire system in monitoring and controlling regional pollution processes.

4. The artificial intelligence-based real-time monitoring and control system for regional pollution processes according to claim 1 is characterized by: The data communication module is equipped with a data cache and retransmission mechanism. When network congestion or signal interruption occurs, the data communication module temporarily stores the collected data in a local cache device. The cache device uses high-speed flash memory technology and has large-capacity storage and fast read and write capabilities. At the same time, the system starts a retransmission timer. Once the network returns to normal, the older data in the cache is retransmitted first in the order of data storage to ensure data integrity and continuity. In the case of unstable 5G / 6G network signals, the monitoring data can be temporarily stored in the cache and retransmitted in an orderly manner after the signal is restored. This mechanism effectively avoids data loss and ensures that pollution monitoring data can be transmitted to subsequent processing modules in a timely and complete manner, providing comprehensive data support for pollution prediction and control strategy formulation.

5. The artificial intelligence-based real-time monitoring and control system for regional pollution processes according to claim 1 is characterized by: The pollution source tracing submodule in the artificial intelligence analysis module introduces a particle filter algorithm to assist in positioning when performing three-dimensional positioning of pollution sources; after the spatiotemporal graph convolutional network outputs preliminary pollution source positioning results, the particle filter algorithm optimizes the positioning results based on the dynamic changes of monitoring data; the particle filter algorithm represents the possible location of the pollution source by randomly sampling a large number of particles in the state space, updates the weights of the particles based on the similarity between the measured data and the predicted state of the particles, and then eliminates particles with lower weights through the resampling process, retains and copies particles with higher weights, and gradually approaches the actual location of the pollution source; in complex urban environments, when there are multiple possible pollution sources and the spread of pollutants is interfered with by multiple factors, the particle filter algorithm can more accurately lock the pollution source, improve the accuracy of pollution source tracing, and provide strong support for precise pollution control.

6. The artificial intelligence-based real-time monitoring and control system for regional pollution processes according to claim 1 is characterized by: The dynamic prediction submodule takes seasonal and cyclical factors into account when generating pollution evolution forecasts. By analyzing historical data, it extracts the patterns of seasonal changes in the concentrations of different pollutants, as well as the characteristics of weekly and monthly cyclical fluctuations. These seasonal and cyclical characteristics are used as additional input features and input into the LSTM-Transformer hybrid neural network along with multi-source monitoring data. When predicting pollution conditions during the winter heating period, the accuracy of pollution evolution trend predictions is improved by combining the historical patterns of increased pollutant emissions during this period and meteorological conditions that are unfavorable for pollutant diffusion. This helps to formulate more targeted pollution control strategies in advance, rationally allocate environmental governance resources, and effectively respond to pollution problems in different seasons and cycles.

7. The artificial intelligence-based real-time monitoring and control system for regional pollution processes according to claim 1 is characterized by: The multi-objective optimization unit in the intelligent control module takes public health risk assessment into consideration when constructing the control decision-making model; by establishing a quantitative relationship model between pollutant concentration and public health impact, it evaluates the health risks that the public may face under different control schemes; the health risk assessment results are used as an independent objective function to participate in the multi-objective optimization process together with the environmental capacity, economic cost and social benefit objective functions; when formulating industrial production restriction plans, not only the improvement of environmental quality and the control of economic costs are considered, but also the degree of reduction in public health risks is fully evaluated; this makes control decisions more focused on the public interest, protects the physical health of residents, and improves the comprehensive benefits of regional pollution control.

8. The artificial intelligence-based real-time monitoring and control system for regional pollution processes according to claim 1 is characterized by: The drone swarm in the execution terminal module has the ability to autonomously avoid obstacles and work collaboratively; the drones are equipped with advanced lidar and visual sensors to perceive obstacle information in the surrounding environment in real time; when an obstacle is detected, the drone automatically plans a new flight path through the built-in obstacle avoidance algorithm to ensure flight safety; at the same time, the drone swarm exchanges real-time information through wireless communication technology and works collaboratively according to task requirements; when monitoring large-scale polluted areas, multiple drones can fly in a preset formation mode to achieve synchronous monitoring of different areas and improve monitoring efficiency; when performing pollution control tasks, the drone swarm can work collaboratively, some are responsible for spraying control agents, and some are responsible for monitoring the control effects, thereby improving the overall effect of pollution control.

9. The artificial intelligence-based real-time monitoring and control system for regional pollution processes according to claim 1 is characterized by: The three-dimensional pollution situation visualization interface of the user interaction module supports virtual reality (VR) and augmented reality (AR) interaction modes. After wearing VR or AR devices, users can immersively view the regional pollution status, freely shuttle in the virtual environment, and observe the distribution of pollutant concentrations from different angles. Through gesture recognition and voice control technology, users can naturally interact with virtual scenes, zoom in or out of specific areas, query data from specific monitoring points, and switch the display of pollution data at different times. In AR mode, users can combine virtual pollution information with real scenes to intuitively understand the pollution situation in the surrounding environment. This interaction method provides managers and scientific researchers with more intuitive and convenient data analysis and decision support means, which helps to gain a deeper understanding of pollution problems and formulate more effective control strategies.

10. The method for real-time monitoring and control of regional pollution processes based on artificial intelligence according to claim 2, characterized in that: An adaptive control strategy adjustment mechanism is introduced into the closed-loop control loop of "monitoring-prediction-control-verification". When the system finds that the control measures fail to achieve the expected goals based on feedback from the execution effect, the system automatically analyzes the reasons, predicts model errors, and fails to implement the control measures in place. If the prediction model error causes the system to conduct more in-depth optimization of the prediction model based on the feedback data, adjust the model parameters or structure, and solve the problem of implementing the control measures, the system will re-evaluate the feasibility and effectiveness of the control strategy and adjust the intensity and execution time of the control instructions according to the actual situation. When it is found that the pollution concentration in a certain area has not decreased as fast as expected, the system analyzes that the industrial production restrictions may be insufficient, and then increases the production restriction ratio or extends the production restriction time. This adaptive mechanism enables the system to respond more flexibly to complex and changing pollution situations, continuously optimize the control strategy, and improve the effectiveness and efficiency of regional pollution control.

Images