Intelligent environment pollution detection control method and system based on internet of things

Abstract

The application discloses an intelligent environmental pollution detection control method and system based on Internet of Things, belongs to the technical field of environmental monitoring, collects pollution data from multiple sources such as water, soil and atmosphere through a wireless sensor network, generates a standardized data set by format conversion using an embedded system, extracts key indicators in combination with a DSP chip and performs trend prediction using a prediction model, generates a control instruction when the predicted pollution level exceeds a threshold value and remotely sends the control instruction to an execution terminal, and simultaneously compares feedback response data with the predicted value, dynamically adjusts instruction parameters to optimize control effect. The application realizes full-process automation and precision from data collection to closed-loop control, effectively reduces control deviation, improves the real-time performance and accuracy of environmental pollution control, and has remarkable overall technical effect, thereby providing efficient and intelligent support for environmental monitoring and control.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring technology, and in particular discloses an intelligent environmental pollution detection and control method and system based on the Internet of Things. Background Technology

[0002] Environmental monitoring, as a crucial means of ensuring ecological security, plays an irreplaceable role in pollution control and early warning. However, current environmental monitoring methods have significant shortcomings in practical applications.

[0003] Many monitoring methods are limited to detecting specific types of pollution, making it difficult to address the complex interactions of multiple environmental factors and resulting in incomplete monitoring results. More critically, existing technologies often exhibit delays or fragmentation in the post-collection processing and response stages, failing to form a complete closed loop from problem detection to resolution. This deficiency is particularly pronounced in the face of sudden pollution events, often missing the optimal intervention window. Against this backdrop, the environmental monitoring field faces a core technical challenge: how to achieve real-time integration and rapid response to multi-dimensional pollution data. In particular, the timeliness of data transmission has become a key factor limiting the overall system effectiveness. If data cannot be transmitted from monitoring points to the control center in a short time, pollution spread may not be contained. Furthermore, data transmission delays not only affect the real-time nature of monitoring but also weaken the targeting of subsequent analysis and control measures, creating a series of chain reactions. For example, if pollutant exceedances are detected upstream of a river, if the data cannot be quickly transmitted back and trigger downstream early warning and interception mechanisms, the pollutants may have already spread to a wider area, causing greater ecological damage.

[0004] Therefore, ensuring efficient connection of pollution data from collection to transmission to control in multiple scenarios has become a key issue in environmental monitoring and governance. Summary of the Invention

[0005] This invention provides an intelligent environmental pollution detection and control method and system based on the Internet of Things, aiming to solve the technical problems of difficulty in integrating multi-dimensional data, low prediction accuracy and large control deviation in environmental pollution data collection, processing and control.

[0006] One aspect of the present invention relates to an intelligent environmental pollution detection and control method based on the Internet of Things, comprising the following steps: S100: Collect multi-dimensional pollution data from water, soil and air sources through a wireless sensor network, and use a detection program in the embedded Linux kernel to perform preliminary format conversion on the collected pollution data to obtain a standardized pollution dataset. S200: For a standardized pollution dataset, a DSP chip is used to execute a data grouping algorithm to determine the distribution characteristics of different pollution types, and key indicator sequences are extracted from the distribution characteristics. S300: The predictive analysis model is used to predict the trend of the key indicator sequence to obtain the predicted pollution level. If the predicted pollution level exceeds the preset threshold, a control command sequence is generated based on the predicted pollution level and sent to the execution terminal through the remote transmission module. S400: Obtain the response data fed back by the execution terminal, compare the response data with the predicted pollution level, and determine the control deviation of the closed-loop control. S500: To address control deviations, the parameters of the control command sequence are adjusted to obtain an optimized control command sequence, which is then retransmitted via the remote transmission module to achieve closed-loop control of environmental pollution.

[0007] Further, step S100 includes: S110, a wireless sensor network node, collects multi-dimensional raw pollution signals of water, soil and atmosphere. The analog voltage signal is converted into a digital quantization sequence through an analog-to-digital converter circuit. The digital quantization sequence is then time-aligned using a preset sampling frequency to obtain a time-aligned pollution data stream. S120: The embedded system reads the timing-aligned contaminated data stream and performs dimensional mapping on the digital quantization sequence according to the preset physical quantity conversion function to obtain the physical quantity contaminated dataset. S130. The detection program removes outliers from the physical quantity contamination dataset and smooths the fluctuations of the physical quantity contamination dataset using a sliding window mean filtering algorithm to obtain a smoothed contamination dataset. S140. The detection program encapsulates the multi-dimensional pollution values ​​in the smoothed pollution dataset into fixed-length binary data frames according to the timestamp order based on the preset communication protocol structure, thereby obtaining a standardized pollution dataset and realizing the standardized collection and format conversion of multi-dimensional pollution data.

[0008] Further, step S200 includes: S210. For the standardized pollution dataset, the spatial dimension of the multi-dimensional pollution values ​​is divided by a clustering analysis algorithm to determine the spatial distribution matrix. S220. Based on the spatial distribution matrix, the DSP chip is used to execute the time-series correlation analysis algorithm to detect the synchronicity of numerical changes of different pollution sources and determine the spatiotemporal coupling correlation map. S230. For the spatiotemporal coupling correlation map, the principal component analysis algorithm is used to reduce the dimensionality of multi-dimensional pollution values ​​and determine the pollution type feature vector. S240. Based on the pollution type feature vector, use the threshold segmentation algorithm to divide the pollution type feature vector into numerical intervals and determine the key indicator sequence for different pollution types.

[0009] Further, step S300 includes: S310. Construct a long short-term memory network model based on the key indicator sequence, and determine the predicted pollution level within the future time window through iterative calculation of historical pollution values. S320. If the predicted pollution level exceeds the preset threshold, the pollution source response function is invoked to determine the operating power parameters of the pollution control equipment based on the difference between the predicted pollution level and the preset threshold. S330. Match the predefined control instruction set according to the operating power parameters, map the operating power parameters into a binary control instruction sequence, and determine the instruction data packet to be sent; S340. Establish an encrypted communication link with the execution terminal through the remote transmission module, and encapsulate the instruction data packet into a transmission protocol frame; S350: The execution terminal receives transmission protocol frames and parses instruction data packets, drives pollution control equipment according to the parsed operating power parameters, determines the real-time operating status of pollution control equipment, and realizes automatic control of pollution levels.

[0010] Further, step S400 includes: S410. Obtain the response data fed back by the execution terminal. The response data includes real-time running status values. S420: Align the real-time operating status values ​​with the pre-stored predicted pollution levels to determine the control deviation of the closed-loop control between the two.

[0011] Further, step S500 includes: S510. Compare the control deviation with the preset deviation threshold. If the control deviation exceeds the preset deviation threshold, extract the correlation mapping relationship between the weight of each instruction and the corresponding response delay in the current control instruction sequence, and determine the correction step size of each instruction weight. S520. Dynamically reconstruct the weights of each instruction in the control instruction sequence according to the correction step size to obtain the corrected instruction weight set, and determine the execution frequency of each instruction weight by combining the matching degree between the current operating load and the environmental baseline. S530. The modified instruction weight set is rearranged in time according to the execution frequency to obtain the time-optimized instruction sequence, and the encoding compression ratio of each instruction is determined in combination with the real-time occupancy of the current transmission bandwidth. S540. The timing optimization instruction sequence is encapsulated according to the encoding compression ratio to obtain the encapsulated control instruction sequence, and the encapsulated control instruction sequence is sent to the execution terminal through the remote transmission module to realize closed-loop control of environmental pollution.

[0012] Another aspect of the present invention relates to an Internet of Things (IoT)-based intelligent environmental pollution detection and control system for implementing the above-described IoT-based intelligent environmental pollution detection and control method, comprising: The standardized pollution dataset acquisition module is used to collect multi-dimensional pollution data from water, soil and atmosphere sources through a wireless sensor network. The detection program in the embedded Linux kernel performs preliminary format conversion on the collected pollution data to obtain a standardized pollution dataset. The key indicator sequence extraction module is used to determine the distribution characteristics of different pollution types by using a data grouping algorithm executed by a DSP chip on a standardized pollution dataset, and then extract key indicator sequences from the distribution characteristics. The remote transmission control module is used to predict the trend of key indicator sequences using a predictive analysis model to obtain the predicted pollution level. If the predicted pollution level exceeds a preset threshold, a control command sequence is generated based on the predicted pollution level and sent to the execution terminal through the remote transmission module. The control deviation judgment module is used to acquire the response data fed back by the execution terminal, compare the response data with the predicted pollution level, and judge the control deviation of the closed-loop control. The environmental pollution closed-loop control module is used to adjust the parameters of the control command sequence in response to control deviations, obtain an optimized control command sequence, and resend it through the remote transmission module to achieve closed-loop control of environmental pollution.

[0013] The beneficial effects achieved by this invention are as follows: The IoT-based intelligent environmental pollution detection and control method and system provided by this invention address the challenges of multi-dimensional data integration, low prediction accuracy, and large control deviations in environmental pollution data collection, processing, and control. It proposes an integrated solution by combining multi-source data collection, standardized processing, trend prediction, and dynamic adjustment of control commands through a logically interconnected process. This invention collects pollution data from multiple sources such as water, soil, and air using a wireless sensor network. It then uses an embedded system to convert the data into a standardized dataset, extracts key indicators using a DSP chip, and employs a predictive model for trend prediction. When the predicted pollution level exceeds a threshold, a control command is generated and remotely sent to the execution terminal. Simultaneously, by comparing the feedback response data with the predicted value, the command parameters are dynamically adjusted to optimize the control effect. This invention achieves full-process automation and precision from data collection to closed-loop control, effectively reducing control deviations and improving the real-time performance and accuracy of environmental pollution control. The overall technical effect is significant, providing efficient and intelligent support for environmental monitoring and control. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an embodiment of the Internet of Things-based intelligent environmental pollution detection and control method of the present invention.

[0015] Figure 2 This is a functional block diagram of an embodiment of the Internet of Things-based intelligent environmental pollution detection and control system of the present invention.

[0016] The following are the diagram labels: 10, Standardized Pollution Dataset Acquisition Module; 20, Key Indicator Sequence Extraction Module; 30, Remote Transmission Control Module; 40, Control Deviation Judgment Module; 50, Environmental Pollution Closed-Loop Control Module. Detailed Implementation

[0017] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0018] like Figure 1 As shown, the first embodiment of this invention proposes an intelligent environmental pollution detection and control method based on the Internet of Things (IoT). The core of this method is to achieve real-time collection, standardized processing, trend prediction, and closed-loop control of multi-dimensional pollution data from the three major environmental sources: water, soil, and air. This solves the problems of traditional environmental pollution detection being lagging, control being untimely, accuracy being low, and lacking closed-loop optimization. It achieves intelligent, precise, and real-time management and control of environmental pollution, and is applicable to various environmental pollution detection and control scenarios such as industrial parks, urban environments, and farmland ecosystems. The method includes the following steps: Step S100: Collect multi-dimensional pollution data from water, soil and atmosphere sources through a wireless sensor network, and use a detection program in the embedded Linux kernel to perform preliminary format conversion on the collected pollution data to obtain a standardized pollution dataset.

[0019] Based on an IoT architecture, a wireless sensor network is deployed, with corresponding sensing devices deployed at key monitoring points of the three major pollution sources: water, soil, and air. Multi-dimensional pollution data is collected in real time through the wireless sensor network, covering the core pollution indicators of water, soil, and air. The pre-set detection program in the embedded Linux kernel is called to perform preliminary processing on the collected raw pollution data, such as noise reduction, format unification, and normalization, to eliminate interference errors and format differences during the data collection process. Finally, a standardized pollution dataset is obtained, providing unified and high-quality basic data support for subsequent data grouping and trend prediction. In this step, data acquisition and standardization processing must meet the following parameter requirements: The number of sensor nodes in the wireless sensor network should be 20-100, preferably 30-60. The criteria for this selection are: 20 or more nodes can cover key monitoring points of the three major pollution sources, ensuring comprehensive data acquisition; fewer than 100 nodes can control equipment deployment costs and network transmission pressure; 30-60 nodes balance coverage and economy, adapting to small- to large-scale pollution monitoring scenarios. The data acquisition frequency should be 0.1-1Hz, preferably 0.2-0.5Hz. The criteria for this selection are: 0.1Hz can meet the basic pollution data monitoring needs; 1Hz can accurately capture the dynamic fluctuations of pollution indicators; and 0.2-0.5Hz balances monitoring real-time performance and... The equipment's energy consumption and network bandwidth usage meet the regular frequency requirements for environmental monitoring; the processing latency of the embedded Linux kernel detection program is ≤500ms, determined to ensure rapid completion of data format conversion and preliminary processing, avoid data accumulation, and adapt to real-time monitoring needs; the update cycle of the standardized pollution dataset is 1~5min, preferably 2~3min, determined to ensure timely updates of pollution data, reflecting the real-time pollution status of the environment, while 2~3min balances real-time performance and data processing efficiency; the data standardization error is ≤3%, determined to ensure that the standardized data can accurately reflect the original pollution status, avoid processing errors leading to subsequent analysis deviations, and meet the accuracy requirements of environmental monitoring data.

[0020] The Internet of Things (IoT) refers to an integrated network architecture that uses technologies such as wireless sensor networks and remote transmission modules to achieve interconnection and interoperability of sensing devices, execution terminals, and data processing modules, thereby completing the collection, transmission, analysis, and control of pollution data. The network transmission latency is ≤1000ms, preferably ≤500ms. The criteria for this value are: ensuring timely data transmission and avoiding detection and control delays caused by transmission delays; network stability ≥99.5% to ensure long-term continuous monitoring without interruption; and meeting the technical requirements of IoT environmental monitoring.

[0021] Wireless sensor networks refer to wireless communication networks composed of multiple sensor nodes and gateways. They are used to collect pollution data of water, soil, and air and transmit it to a data processing module. They adopt low-power, long-distance communication protocols such as LoRa and NB-IoT, with a communication distance of 1 to 5 km, preferably 2 to 3 km. The criteria for this value are: 1 km can meet the needs of small monitoring areas, 5 km can cover large industrial parks or urban areas, and 2 to 3 km balances communication distance and device power consumption. The network packet loss rate is ≤1%, which is determined to ensure the integrity of pollution data transmission and avoid data loss that could lead to monitoring omissions.

[0022] Water, soil, and air sources refer to the main sources of environmental pollution. Water sources include rivers, lakes, groundwater, and industrial wastewater discharge outlets. Soil sources include farmland soil, industrial park soil, and contaminated site soil. Air sources include industrial exhaust outlets, urban dust, and traffic exhaust emission points. The number of sensor nodes deployed for each source should be ≥3. The basis for this is to ensure that the pollution data collected from each source is representative and to avoid monitoring distortion caused by single-point data bias.

[0023] Multidimensional pollution data refers to various pollution-related data collected from water, soil, and air sources, covering core pollution indicators of these three environmental media. Specifically, it includes: water quality indicators (pH, COD, BOD, ammonia nitrogen, total phosphorus, heavy metal concentration), soil indicators (heavy metal concentration, organic matter content, pH), and air quality indicators (PM2.5, PM10, SO2, NO). 2 The specific value ranges and environmental standards for the following substances are as follows: pH 6~9 (water / soil, national standard limit), COD ≤ 50 mg / L (surface water, GB 3838-2002), ammonia nitrogen ≤ 1.0 mg / L (surface water), PM2.5 ≤ 35 μg / m³ (ambient air quality first-class standard, GB3095-2012), PM10 ≤ 50 μg / m³ (first-class standard), heavy metals (lead ≤ 0.1 mg / L (water), ≤ 80 mg / kg (soil), GB 15618-2018). The values ​​are determined in strict accordance with national environmental standards to ensure the compliance and reference value of the collected data.

[0024] An embedded Linux kernel refers to the Linux operating system kernel embedded in a data acquisition terminal. It features lightweight, low power consumption, and high stability, supports the efficient operation of the detection program, and has a memory footprint of ≤128MB. The value is determined based on the following criteria: adapting to the hardware resources of small data acquisition terminals, avoiding excessive memory usage that could cause device lag, and ensuring kernel stability of ≥99.9% to guarantee continuous operation of the detection program.

[0025] The detection program is a program embedded in the embedded Linux kernel used to perform preliminary processing on raw contaminated data. Its core functions include data denoising, format conversion, and normalization. The program's power consumption is ≤50mW, and the value is determined based on the following criteria: adapting to the power supply requirements of low-power sensor nodes, extending the device's battery life (≥7 days on a single charge), and ensuring the reliability of data processing with a processing accuracy of ≥98%.

[0026] A standardized pollution dataset refers to a collection of pollution data that has been processed by the detection procedure, resulting in a unified format, standardized indicators, and elimination of interference. The data format adopts the JSON standard format to facilitate subsequent data grouping and analysis. The data integrity is ≥99.8%, and the values ​​are determined based on ensuring that the dataset has no missing or invalid data, avoiding data omissions that could lead to biases in subsequent analysis, and meeting the requirements for environmental monitoring data management.

[0027] In step S100, multi-dimensional pollution data is collected from water, soil, and air sources via a wireless sensor network. A detection program in the embedded Linux kernel performs preliminary format conversion on the collected pollution data to obtain a standardized pollution dataset. Based on an IoT architecture, a wireless sensor network is deployed at key monitoring points of the three major pollution sources (water, soil, and air). Multi-dimensional pollution data is collected in real time via the wireless sensor network, covering core pollution indicators for water, soil, and air. A pre-defined detection program in the embedded Linux kernel performs preliminary processing on the collected raw pollution data, including noise reduction, format unification, and normalization, eliminating interference errors and format differences during data collection. Finally, a standardized pollution dataset is obtained, providing unified and high-quality basic data support for subsequent data grouping and trend prediction.

[0028] Step S200: For the standardized pollution dataset, use the DSP chip to execute the data grouping algorithm to determine the distribution characteristics of different pollution types, and extract the key indicator sequence from the distribution characteristics.

[0029] Using the standardized pollution dataset obtained in step S100 as input, the high-speed data processing capability of the DSP chip is utilized to execute a preset data grouping algorithm. The standardized pollution data is grouped according to pollution type (water pollution, soil pollution, air pollution) and specific pollution indicators (such as COD, PM2.5, heavy metals). The distribution characteristics of different pollution types (such as concentration distribution, spatiotemporal distribution, and fluctuation patterns) are analyzed. From the distribution characteristics of various pollution types, core parameters that can characterize the severity and changing trend of the pollution type are extracted to form a key indicator sequence, providing core input for subsequent trend prediction and highlighting the monitoring and control priority of key pollution indicators. In this step, data grouping and indicator extraction must meet the following parameter requirements: DSP chip processing speed ≥ 100 MIPS, based on the following criteria: ensuring rapid completion of data grouping and feature analysis to meet real-time data processing needs; chip power consumption ≤ 100 mW to meet the low power consumption requirements of IoT devices; data grouping algorithm grouping accuracy ≥ 99%, based on the following criteria: ensuring accurate grouping of contaminated data to avoid confusion between different types of contaminated data and affecting subsequent analysis; distribution feature analysis time ≤ 10s, preferably ≤ 5s, based on the following criteria: ensuring rapid extraction of distribution features to provide timely support for subsequent indicator extraction; key indicator sequence dimension 5~20, preferably 8~15, based on the following criteria: 5 dimensions or more can comprehensively cover core contaminated indicators, less than 20 dimensions can avoid feature redundancy and reduce the processing complexity of subsequent prediction models, and 8~15 dimensions balance comprehensiveness and efficiency; key indicator sequence extraction accuracy ≥ 98.5%, based on the following criteria: ensuring that the extracted core indicators can truly reflect the distribution characteristics of contamination and provide accurate support for trend prediction.

[0030] DSP chips refer to digital signal processing chips, which have the ability to perform high-speed data processing and real-time signal processing. They are used to execute data grouping algorithms, distribution feature analysis, and subsequent data processing tasks. The sampling rate of the chip is ≥100kHz, which is selected to ensure rapid acquisition and processing of pollution data and to adapt to the parallel processing requirements of multi-dimensional data. The operating temperature range of the DSP chip is -20℃ to 60℃, which is selected to adapt to complex outdoor environments (high temperature and low temperature) and to ensure stable operation of the chip in various monitoring scenarios.

[0031] Data grouping algorithms refer to algorithms used to group standardized pollution data according to pollution type and specific indicators. In this embodiment, K-means clustering algorithm and decision tree algorithm are preferred, which can achieve accurate grouping of different pollution types. The clustering error of the algorithm is ≤1%, and the value is determined by ensuring the accuracy of the grouping results and avoiding confusion of data of different pollution types. The processing delay of the algorithm is ≤1s, and the value is determined by ensuring fast completion of grouping and adapting to the needs of real-time data processing.

[0032] The distribution characteristics of pollution types refer to the temporal and spatial distribution patterns and concentration variation characteristics of pollution data of different pollution types (water, soil, and air). Specifically, these include average concentration, fluctuation range, peak concentration, and distribution density. The quantification range of fluctuation range is 0-50%, and the value is based on the normal fluctuation range of various pollution indicators. Exceeding this range indicates pollution anomalies. The analysis accuracy of distribution characteristics is ≤2% to ensure the accuracy of the analysis results.

[0033] The key indicator sequence refers to an ordered set of core pollution indicators extracted from the distribution characteristics of pollution types, which can characterize the severity and trend of pollution. The core indicators include water quality (COD, ammonia nitrogen), soil (heavy metal concentration), and air (PM2.5, SO2). The update cycle of the sequence is consistent with that of the standardized pollution dataset (1~5 min). The selection criteria are: to ensure that the key indicator sequence can reflect pollution changes in real time and provide timely and accurate input for subsequent trend prediction, and the integrity of the sequence is ≥99.5%.

[0034] In step S200, for the standardized pollution dataset, a data grouping algorithm is executed using a DSP chip to determine the distribution characteristics of different pollution types, and key indicator sequences are extracted from the distribution characteristics. Using the standardized pollution dataset obtained in step S100 as input, the high-speed data processing capability of the DSP chip is utilized to execute a preset data grouping algorithm, grouping the standardized pollution data according to pollution type and specific indicators, and analyzing the distribution characteristics of different pollution types. From the distribution characteristics of each pollution type, core parameters that can characterize the severity and changing trend of the pollution type are extracted to form key indicator sequences. In this embodiment, the prediction model is an LSTM (Long Short-Term Memory) neural network time-series prediction model, with the following specific structure and implementation: The model type is an LSTM time-series prediction network; the network structure is: 12-dimensional input layer → 64 neurons in the LSTM layer → 32 neurons in the LSTM layer → 16 neurons in the fully connected layer → 1-dimensional output layer; the activation function is ReLU in the hidden layer and Sigmoid in the output layer; the loss function is the mean squared error (MSE); the optimizer is Adam with a learning rate of 0.001; the training data uses 180 days of historical pollution time-series data, totaling 25,920 samples; the input is a sequence of key indicators, including water quality, soil, air, temperature, humidity, wind speed, air pressure, and temporal location; the output is the predicted pollution level for the next 1–6 hours; the convergence condition is a loss < 0.005 and a test set R² ≥ 0.96. Those skilled in the art can directly reproduce this model using the above structure and parameters.

[0035] Step S300: Use a predictive analysis model to predict the trend of key indicator sequences to obtain the predicted pollution level. If the predicted pollution level exceeds a preset threshold, generate a control command sequence based on the predicted pollution level and send it to the execution terminal through a remote transmission module.

[0036] Using the key indicator sequence obtained in step S200 as input, a preset predictive analysis model is adopted, combined with historical pollution data and environmental influencing factors (such as meteorology, hydrology, and industrial production load), to quantitatively predict the changing trend of pollution indicators over a future period, thereby obtaining the predicted pollution level. Preset pollution thresholds (based on national environmental protection standards) for different pollution types and key indicators are used. The predicted pollution level is compared with the corresponding preset threshold. If the predicted pollution level exceeds the preset threshold, it is determined that there is a risk of pollution exceeding the standard. Based on the degree of pollution exceeding the standard, a corresponding control instruction sequence is generated, specifying the control actions, control parameters, and execution timing of the execution terminal. Through a remote transmission module, the control instruction sequence is sent to the corresponding execution terminal in real time to initiate pollution control actions, achieving early intervention for pollution exceeding the standard. In this step, trend prediction and command transmission must meet the following parameter requirements: The preferred predictive analysis model is LSTM or ARIMA, with a prediction accuracy ≥95%. This is based on ensuring accurate prediction of pollution trends, avoiding missed detections of exceeding standards or misjudging normal conditions, and meeting the stringent requirements of environmental early warning. The prediction duration is 1-24 hours, preferably 6-12 hours. This is based on the fact that 1 hour allows for short-term early warning, 24 hours provides sufficient preparation time for control, and 6-12 hours balances timely warning and control feasibility. The preset threshold strictly adheres to national environmental standards; specific values ​​are defined in the new terminology definition. The threshold misjudgment rate is ≤0.5%, ensuring accurate threshold determination and avoiding over-control or control lag. The generation time of the control command sequence is ≤15 seconds, preferably ≤10 seconds, ensuring rapid command generation and timely initiation of control actions to prevent further pollution spread. The transmission rate of the remote transmission module is ≥1Mbps, ensuring rapid command transmission, a transmission delay ≤500ms, and a bit error rate ≤10%. -6 The value is determined based on the following: ensuring accurate and distortion-free instruction transmission, and avoiding abnormal terminal actions caused by transmission errors.

[0037] Predictive analysis models refer to artificial intelligence models used to predict the trend of key indicator sequences and obtain predicted pollution levels. They can combine historical pollution data and environmental influencing factors to achieve accurate prediction of pollution trends. The training iterations of the model are 1,000 to 5,000 times, with 2,000 to 3,000 times being preferred. The convergence threshold is a loss function value ≤ 0.05, which is determined based on ensuring that the model is fully trained and the prediction accuracy meets the standard. The inference delay of the model is ≤ 1 second, which is determined based on ensuring rapid output of prediction results and adapting to real-time early warning requirements.

[0038] Predicted pollution levels refer to the numerical values ​​obtained by quantitatively predicting the concentration levels of key pollution indicators within the next 1 to 24 hours through predictive analysis models. The predicted values ​​are consistent with the dimensions of the key indicator sequence, and the range of values ​​is consistent with the actual concentration range of the corresponding pollution indicators (e.g., the predicted range of PM2.5 is 0 to 500 μg / m³). The prediction error is ≤10%, and the basis for the values ​​is to ensure that the prediction results can truly reflect the future pollution status and provide a reliable basis for the generation of control instructions.

[0039] The preset thresholds refer to the pollution exceedance thresholds set based on national environmental protection standards for different pollution types and key indicators. Specific values ​​are as follows (strictly adhering to national standards): Water quality (COD ≤ 50 mg / L (surface water), ≤ 100 mg / L (industrial wastewater discharge), GB 8978-1996; Ammonia nitrogen ≤ 1.0 mg / L (surface water), ≤ 5.0 mg / L (industrial wastewater)); Soil (Lead ≤ 80 mg / kg (farmland), ≤ 300 mg / kg (industrial park), GB 15618-2018); Air (PM2.5 ≤ 35 μg / m³ (Level 1), ≤ 75 μg / m³ (Level 2); SO2 ≤ 50 μg / m³ (Level 1), ≤ 150 μg / m³ (Level 2), GB 3095-2012). The basis for these values ​​is strict adherence to national environmental monitoring standards and pollutant emission standards to ensure the compliance of pollution exceedance determinations and avoid control deviations caused by artificially set thresholds.

[0040] A control instruction sequence refers to a set of instructions generated to control the execution terminal for pollution control in response to predicted pollution levels exceeding standards. It includes the terminal's equipment number, control actions (such as spraying, adsorption, and filtration), control parameters (such as spray flow rate and adsorption time), and execution duration. The instruction sequence format is adapted to the execution terminal's control protocol, and the execution rate is ≥99%. The values ​​are determined based on ensuring the instructions can be accurately identified and executed by the execution terminal, avoiding incompatibility between instructions and equipment, and ensuring the accuracy of control parameters (0.1 L / min for flow rate parameters and 1 min for time parameters) to guarantee the precision of the control actions.

[0041] The remote transmission module refers to a module used to remotely transmit information such as control command sequences and response data. It adopts communication methods such as 4G / 5G and LoRa, with a communication distance of 1~10km, preferably 3~5km. The selection criteria are: 1km can meet the needs of small monitoring areas, 10km can cover large areas, and 3~5km balances communication distance and transmission stability. The power consumption of the module is ≤80mW, which is suitable for the low power consumption requirements of IoT devices.

[0042] An execution terminal refers to equipment used to perform pollution control actions, specifically including water purification equipment, soil remediation equipment, and air purification equipment (such as spray towers, activated carbon adsorption devices, and dust collectors). The response time of the execution terminal is ≤30s, and the value is determined based on the following criteria: ensuring rapid response to control commands and timely initiation of pollution control actions; the operational stability of the execution terminal is ≥99%; and ensuring continuous and uninterrupted pollution control process.

[0043] In step S300, a predictive analysis model is used to predict the trend of key indicator sequences to obtain the predicted pollution level. If the predicted pollution level exceeds a preset threshold, a control command sequence is generated based on the predicted pollution level and sent to the execution terminal via a remote transmission module. Using the key indicator sequence obtained in step S200 as input, a preset predictive analysis model is used, combined with historical pollution data and environmental impact factors, to quantitatively predict the trend of pollution indicator changes over a future period, obtaining the predicted pollution level. Preset pollution thresholds for different pollution types and key indicators are used, and the predicted pollution level is compared with the corresponding preset threshold. If the predicted pollution level exceeds the preset threshold, it is determined that there is a risk of pollution exceeding the standard. Based on the degree of exceedance of the predicted pollution level, a corresponding control command sequence is generated, specifying the control actions, control parameters, and execution timing of the execution terminal. The control command sequence is sent to the corresponding execution terminal in real time via the remote transmission module to initiate pollution control actions, achieving early intervention for pollution exceeding the standard. In this embodiment, the classification model is an RBF kernel support vector machine (SVM) classifier with the following specific parameters: the kernel function is the radial basis function (RBF); the penalty coefficient C = 1.2; the kernel parameter gamma = 0.08; the input features include pollution concentration, rate of change, fluctuation amplitude, temporal slope, and spatial distribution coefficient; the output is seasonal fluctuation and sudden fluctuation; the training set consists of 8000 labeled samples; the classification threshold is a posterior probability ≥ 0.6 for determining sudden fluctuation; the sudden fluctuation proportion coefficient = sudden period length / total window length, threshold = 0.65. Those skilled in the art can directly reproduce this implementation.

[0044] Step S400: Obtain the response data fed back from the execution terminal, compare the response data with the predicted pollution level, and determine the control deviation of the closed-loop control.

[0045] After the control command sequence is sent to the execution terminal, the response data fed back by the execution terminal is obtained in real time. The response data includes core information such as the operating status of the execution terminal, actual control parameters, and pollution control effect (real-time pollution concentration). The actual pollution control effect (real-time pollution concentration) in the response data is compared with the predicted pollution level obtained in step S300, the difference between the two is calculated, the control deviation in the closed-loop control process is quantified, and it is determined whether the control deviation is within the allowable range. This provides a basis for subsequent control command optimization and ensures that the pollution control effect reaches the expected level. In this step, the data comparison and deviation judgment must meet the following parameter requirements: the acquisition frequency of response data is 0.1~1Hz, preferably 0.2~0.5Hz, based on the following criteria: ensuring real-time acquisition of the operating status and treatment effect of the execution terminal, and timely capture of control deviations; the feedback delay of response data is ≤1s, based on the following criteria: ensuring rapid acquisition of feedback information and avoiding delays that lead to lagging deviation judgment; the calculation error of control deviation is ≤2%, based on the following criteria: ensuring accurate deviation calculation and providing reliable support for subsequent instruction optimization; the allowable range of control deviation is ±5% (relative to the predicted pollution level), based on the following criteria: considering the control accuracy of the environmental protection treatment equipment, a deviation of ±5% is acceptable; if it exceeds this range, the control instructions need to be optimized to avoid poor control effect.

[0046] Response data refers to the real-time data fed back to the data processing module after the execution terminal receives the control command sequence. The core includes the execution terminal's operating status (normal, fault), actual control parameters (such as actual spray flow rate and adsorption time), and real-time pollution concentration (concentration of pollution indicators after treatment). The data format is consistent with the standardized pollution dataset, and the data integrity is ≥99.8%. The basis for the values ​​is to ensure that the response data is free of missing or errors, so as to provide an accurate basis for deviation judgment.

[0047] Closed-loop control refers to a closed-loop management process of "acquisition-processing-prediction-control-feedback-optimization". By executing the response data feedback from the terminal, the control commands are dynamically optimized to ensure that the pollution control effect reaches the expected level. The cycle of closed-loop control is 1 to 5 minutes, preferably 2 to 3 minutes. The basis for the selection is to ensure timely closed-loop optimization, adapt to the dynamic characteristics of pollution changes, and ensure the stability of closed-loop control ≥99.5% to ensure long-term continuous and fault-free management.

[0048] Control deviation refers to the difference between the actual pollution concentration (after treatment) fed back by the execution terminal and the pollution level predicted in step S300. The quantitative formula is: Control deviation = |actual pollution concentration - predicted pollution level| / predicted pollution level × 100%, with a value range of 0~20% and an allowable deviation range of ±5%. The basis for the value is: considering the control accuracy of the environmental protection equipment and the fluctuation characteristics of the pollution index, a deviation of ±5% is acceptable. If it exceeds this range, it indicates that the control effect is not good and the control command needs to be optimized.

[0049] In step S400, response data fed back from the execution terminal is acquired, and the response data is compared with the predicted pollution level to determine the control deviation of the closed-loop control. After the control command sequence is sent to the execution terminal, the response data fed back from the execution terminal is acquired in real time. The response data includes core information such as the operating status of the execution terminal, actual control parameters, and pollution control effect. The actual pollution control effect in the response data is compared with the predicted pollution level obtained in step S300, and the difference between the two is calculated to quantify the control deviation in the closed-loop control process. It is determined whether the control deviation is within the allowable range, providing a basis for subsequent control command optimization and ensuring that the pollution control effect achieves the expected results.

[0050] Step S500: Adjust the parameters of the control command sequence to address the control deviation, obtain an optimized control command sequence, and resend it through the remote transmission module to achieve closed-loop control of environmental pollution.

[0051] For the control deviation determined in step S400, if the control deviation exceeds the allowable range (±5%), the cause of the deviation is analyzed by combining the response data and the predicted pollution level (such as deviation of the execution terminal control parameters or pollution diffusion speed exceeding expectations). The control parameters in the control command sequence are adjusted (such as increasing the spray flow rate or extending the adsorption time), and the control actions and execution timing are optimized to obtain an optimized control command sequence. The optimized control command sequence is resent to the execution terminal through the remote transmission module to execute new pollution control actions. The process of steps S400 to S500 is repeated to continuously optimize the control commands until the control deviation is reduced to the allowable range, thereby achieving closed-loop control of environmental pollution and ensuring that the pollution level remains stable below the preset threshold. In this step, instruction optimization and closed-loop control must meet the following parameter requirements: The parameter adjustment range of the control instruction sequence is 5%~20% of the original parameters. The basis for this value is to avoid excessive parameter adjustment leading to abnormal operation of the execution terminal. An adjustment range of 5%~20% can achieve precise optimization and ensure that the control effect is gradually improved. The generation time of the optimized control instruction sequence is ≤10s, preferably ≤8s. The basis for this value is to ensure that instruction optimization is completed quickly, control actions are adjusted in a timely manner, and pollution is prevented from spreading further. The transmission success rate of instruction retransmission is ≥99.5%. The basis for this value is to ensure that the optimized instructions can be successfully transmitted to the execution terminal and to avoid optimization delays due to transmission failures. The maximum number of optimizations for closed-loop control is 5. The basis for this value is to issue an alarm signal if the control deviation still does not drop to the allowable range after 5 optimizations, prompting staff to intervene manually and preventing infinite loops that could lead to pollution exceeding the standard. After the control deviation drops to the allowable range (±5%), the continuous monitoring time is ≥30min. The basis for this value is to ensure that the pollution level is stable within a safe range and to avoid short-term fluctuations that could lead to control failure, meeting the stability requirements of environmental protection management.

[0052] Parameter adjustment of the control command sequence refers to the process of adjusting the control parameters (such as spray flow rate, adsorption time, and purification power) in the control command sequence in response to control deviations. The adjustment needs to be combined with the magnitude of the deviation and the response data. The larger the deviation, the larger the adjustment range (not exceeding 20%). The adjustment accuracy is 0.1L / min (flow rate) and 1min (time). The basis for the value selection is to ensure that the parameter adjustment is accurate and to avoid the control effect deterioration caused by the adjustment deviation. The adjustment response time is ≤5s to ensure that the parameter adjustment is completed quickly.

[0053] The optimized control instruction sequence refers to the set of control instructions that, after parameter adjustment, can reduce control deviation and improve pollution control effectiveness. It has the same format as the original control instruction sequence, adds deviation adjustment instructions, clarifies the reasons for adjustment and the adjustment range, and the execution rate of the instructions is ≥99.5%. The criteria for the values ​​are: ensuring that the optimized instructions can be accurately identified and executed by the execution terminal, improving the control effect by ≥10%, and ensuring that the deviation can be quickly reduced to the allowable range.

[0054] Closed-loop environmental pollution control refers to a management and control mode that continuously adjusts control commands through a cycle of "prediction-control-feedback-optimization" to keep pollution levels stable below a preset threshold. The response time of closed-loop control is ≤30s, and the basis for this value is to ensure rapid response to changes in pollution and timely optimization of control strategies. The control accuracy of closed-loop control is ≥98%, and the basis for this value is to ensure that pollution levels can be stably controlled below the preset threshold, which meets the core objectives of environmental protection management and control.

[0055] In step S500, the parameters of the control command sequence are adjusted to obtain an optimized control command sequence for the control deviation, and then retransmitted through the remote transmission module to achieve closed-loop control of environmental pollution. For the control deviation determined in step S400, if the control deviation exceeds the allowable range, the cause of the deviation is analyzed by combining response data and predicted pollution levels. The control parameters in the control command sequence are adjusted, and the control actions and execution timing are optimized to obtain an optimized control command sequence. The optimized control command sequence is then retransmitted to the execution terminal through the remote transmission module to execute new pollution control actions. The process of steps S400 to S500 is repeated to continuously optimize the control commands until the control deviation is reduced to the allowable range, achieving closed-loop control of environmental pollution and ensuring that the pollution level remains stable below the preset threshold.

[0056] Furthermore, in the IoT-based intelligent environmental pollution detection and control method provided in this embodiment, step S100 includes:

[0057] Step S110: The wireless sensor network node collects multi-dimensional raw pollution signals of water, soil and atmosphere, converts the analog voltage signal into a digital quantization sequence through an analog-to-digital converter circuit, and performs time-series alignment of the digital quantization sequence using a preset sampling frequency to obtain a time-series aligned pollution data stream.

[0058] The time-aligned contaminated data stream is derived using the following formula: (1); (2); In formulas (1)~(2), For the first The data collected by the first sensor node is the first... The first of the pollution indicators These are digital quantized values ​​without physical units; they originate from analog-to-digital conversion calculation results; their value range is... Defined as a standardized digital quantity obtained after analog-to-digital conversion of the original analog pollution signal, it serves as the basic input for all subsequent data processing. For the first The data collected by the first sensor node is the first... The first of the pollution indicators A simulated voltage signal, in volts (V); sourced from raw signals collected in real time by wireless sensor network nodes; value range: Defined as the analog voltage output after the original pollution signal is converted by the sensing node, it directly reflects the original intensity of the pollution index. This represents the minimum input voltage of the analog-to-digital converter (ADC), measured in volts (V); it originates from the preset parameters of the ADC; and its value range is... Defined as the lowest analog voltage that an analog-to-digital converter circuit can effectively identify, used for normalization processing of the original analog signal. This represents the maximum input voltage of the analog-to-digital converter circuit, measured in volts (V); it originates from the preset parameters of the analog-to-digital converter circuit; its value range is... ; defined as the highest analog voltage that an analog-to-digital converter circuit can effectively identify, and The normalization process was completed in conjunction with this. This refers to the quantization bits in the analog-to-digital conversion (ADC), which has no physical unit; it originates from the hardware parameters of the ADC circuit; its value range is... Defined as the number of bits in a digital quantization sequence, it directly determines the accuracy of analog-to-digital conversion; the higher the number of bits, the higher the conversion accuracy. It is a floor function with no physical unit; it is defined as taking the nearest integer down from the normalized value of the input, ensuring that the digital quantization value is an integer, which meets the requirements of digital signal processing. After timing alignment, the first The first sensing node Pollution indicators in time The pollution data stream has no physical units; it originates from the results of time-series alignment processing; its value is a continuous time-series sequence of digital quantized values; it is defined as a standardized pollution data stream after time synchronization calibration to ensure that data from different nodes and different types of pollution indicators can be compared and fused on the same time axis. This is a unified timeline after time alignment, with units in seconds (s); the source is a system-preset time base; the value range is... Defined as a unified time reference for all pollution data, used to eliminate sampling time deviations between different sensor nodes. The sampling period is a preset period in seconds (s); the source is a preset parameter of the wireless sensor network; the value range is... Defined as the sampling time interval of the sensing node, it directly determines the temporal resolution of pollution data acquisition. It is an impulse function with no physical unit; it is defined as outputting 1 when the input parameter is 0 and outputting 0 when the input parameter is not 0. Its core function is to map discrete digital quantized values ​​to a continuous time axis to achieve time alignment of multi-node data. This is a sensor node index, with no physical unit; the value range is... ( (Total number of sensor nodes); defined as a unique identifier to distinguish different wireless sensor nodes, used to locate the data source. This is an index for pollution indicator types; it has no physical units; the value range is... ( (This refers to the number of pollution indicator types, such as water quality, soil, and air quality indicators); it is defined as a unique identifier to distinguish different pollution types and is used for classifying and processing multi-dimensional pollution data. This is the sampling point index, with no physical unit; the value range is... ; is defined as the sampling point number continuously collected by the sensor node, used to distinguish different sampling data of the same node and the same index. The control logic of formulas (1) to (2) is to first use formula (1) to convert the original analog voltage signal collected by the wireless sensor network node into a signal. Based on the preset input range of the analog-to-digital converter circuit , and quantization bits Complete the conversion of analog signals into digital quantization sequences. The conversion; then, using formula (2), the sampling period corresponding to the preset sampling frequency is used. Based on the impulse function By mapping discrete digital quantized sequences to a unified time axis, time-series alignment of multi-node, multi-indicator data is achieved, ultimately yielding a time-series aligned contamination data stream. This effectively eliminates the data inconsistency problem caused by the sampling time deviation of different sensor nodes, laying the foundation for subsequent multi-source data fusion. Formulas (1) and (2) break through the technical limitations of traditional analog-to-digital conversion, which only realizes the simple conversion of "analog signal to digital signal". They introduce a voltage range normalization processing mechanism to effectively avoid the quantization error caused by the voltage reference difference of different sensor nodes. At the same time, they combine the impulse function to complete the timing alignment, and specifically solve the problem of time asynchrony in the process of multi-dimensional and multi-node contamination data acquisition. This significantly improves the accuracy and consistency of data acquisition and provides core data support for the accuracy of subsequent contamination data analysis.

[0059] Wireless sensor network nodes collect water pollution signals, such as pH value and dissolved oxygen, through sensors deployed along the river. These analog voltage signals are converted into digital quantization sequences by analog-to-digital converter circuits. The preset sampling frequency is 10 times per second to ensure time alignment and obtain a continuous pollution data stream.

[0060] Step S120: The embedded system reads the timing-aligned contaminated data stream, performs dimensional mapping on the digital quantization sequence according to the preset physical quantity conversion function, and obtains the physical quantity contaminated dataset.

[0061] The physical pollution values ​​are obtained using the following formula: (3); In formula (3), After timing alignment, the first The first sensing node Pollution indicators in time The actual physical quantity pollution value, the unit of which depends on the type of pollution indicator (e.g., mg / L, μg / m³, mg / kg); the source is the result of physical quantity conversion; the value range depends on the type of pollution indicator (e.g., the value range for COD is...). ); defined as pollution values ​​with practical environmental monitoring significance, which are the core basic data for subsequent pollution analysis, prediction and control. For the first The physical quantity conversion coefficient for pollution indicators is expressed in units of "actual physical quantity units / digital quantification units". It is derived from the physical quantity conversion parameters preset based on the sensor node calibration results. The value range depends on the type of pollution indicator (e.g., the conversion coefficient for PM2.5 is 0.1 μg / m³ / quantification unit). It is defined as the proportional relationship parameter between the digital quantification value and the actual physical quantity, used to realize the proportional conversion from digital quantity to physical quantity. For the first The physical quantity conversion offset of pollution indicators is in units consistent with the actual physical quantity units of the corresponding pollution indicator; it originates from the physical quantity conversion parameters preset based on the sensor node calibration results; the value range depends on the type of pollution indicator (e.g., the offset value range for some water quality indicators is...). ); is defined as the offset parameter to correct the system error of the sensing node, ensuring that the converted physical quantity data meets the national environmental monitoring standards. The control logic of formula (3) is to use a preset linear physical quantity conversion function to convert the time-aligned digital quantization sequence. Multiply by the corresponding number Specific conversion coefficients for pollution indicators and add offset This process maps digital quantities to actual physical quantities, yielding pollution values ​​with clear environmental significance. Ultimately, a physical quantity pollution dataset is formed, realizing the transformation of digital quantities into actual pollution physical quantities, making the data interpretable and practically applicable. Formula (3) abandons the traditional extensive processing method of using a fixed ratio in physical quantity conversion, and sets exclusive conversion coefficients for different types of pollution indicators (such as water COD, soil heavy metal content, atmospheric PM2.5, etc.). and offset It effectively corrects the systematic errors of different sensing nodes and the dimensional differences of different pollution indicators, and greatly improves the accuracy of physical quantity conversion. At the same time, it solves the technical problem of inconsistent dimensions of multiple types of pollution indicators and the inability to directly compare and analyze them, providing directly usable physical quantity data for subsequent pollution level assessment and control.

[0062] After reading the time-aligned contamination data stream, the embedded system performs dimensional mapping according to a preset physical quantity conversion function. For example, it maps the digital quantization sequence from the range of 0-1023 to the actual pH value of 0-14. Through a linear function such as physical quantity = (digital value / 1023)*14 + offset, the physical quantity contamination dataset is obtained. This involves a calibration process, in which known standard sample data is first collected to fit the function parameters, thereby achieving an accurate conversion from quantized values ​​to real physical quantities. This mapping not only preserves the original information of the signal but also facilitates subsequent analysis and avoids data distortion.

[0063] Step S130: The detection program removes outliers from the physical quantity contamination dataset and smooths the fluctuations of the physical quantity contamination dataset using a sliding window mean filtering algorithm to obtain a smoothed contamination dataset.

[0064] The contamination value of physical quantities after smoothing is obtained by the following formula: (4); (5); In formulas (4)~(5), For the first The first sensing node The average physical quantity of a pollution index is defined as the unit of the actual physical quantity of the corresponding pollution index. The source is the statistical average of the physical quantity pollution data before outliers are removed. The range of values ​​depends on the type of pollution index. It is defined as the normal fluctuation benchmark value of this type of pollution index and is used to determine outliers. For the first The first sensing node The standard deviation of the physical quantity of a pollution index is given, with the unit consistent with the actual physical quantity unit of the corresponding pollution index. It is derived from the statistical standard deviation of the physical quantity pollution data before outliers are removed. The range of values ​​depends on the type of pollution index. It is defined as the normal fluctuation range of the pollution index and is used in conjunction with the mean to complete the outlier determination. After smoothing, the first The first sensing node Pollution indicators in time The physical quantity pollution value is consistent with the actual physical quantity unit of the corresponding pollution index; the source is the sliding window mean filtering result; the value range depends on the pollution index type; it is defined as stable pollution data after eliminating random fluctuations, and is used for subsequent standardized packaging processing. The length of the sliding window has no physical unit; it is derived from preset filter parameters; its value range is... ; is defined as the number of sampling points contained in the sliding window. The larger the window length, the more significant the filtering and smoothing effect, and the stronger the data stability (it is necessary to balance the smoothing effect with the preservation of time series trends). This is the offset index of the sampling point within the sliding window, with no physical unit; its value range is... ; is defined as the time offset of different sampling points within the sliding window relative to the current time t, used to determine the range of sampling points participating in the mean calculation. The control logic of formulas (4) to (5) is to first calculate the first using formula (4). The first sensing node Mean value of physical quantity data of pollution indicators and standard deviation If the pollution value of a physical quantity at a certain moment If the value exceeds the reasonable range of [mean ± 3 standard deviations], it is considered an outlier and removed; subsequently, using formula (5), a length of... A sliding window is used to calculate the moving average of the physical quantity contamination data after outlier removal, resulting in smoothed contamination data. This effectively eliminates data fluctuations caused by sensor noise and environmental interference, improving data stability and reliability. Formulas (4) to (5) use the "3σ criterion" to accurately identify and remove outliers. Compared with the traditional fixed threshold removal method, it can adapt to the fluctuation characteristics of different pollution indicators and effectively avoid the erroneous removal of normal data. At the same time, combined with the sliding window mean filtering technology, the time series data is smoothed through a dynamic window. While preserving the time series change trend of pollution data, it effectively suppresses random interference, solves the technical problem of poor stability and large interference of the original pollution data, and provides high-quality preprocessed data for subsequent data grouping and trend prediction.

[0065] The detection program removes outliers from the physical quantity contamination dataset. Specifically, it uses statistical methods such as calculating the mean and standard deviation to identify and remove values ​​exceeding three times the standard deviation. Then, it applies a sliding window mean filter with a window size of five data points to calculate the average value point by point for the digital quantization sequence to smooth out fluctuations. For example, for the digital quantization sequence [7.2, 7.5, 10.0, 7.3, 7.4], removing 10.0 results in a more stable sequence, which helps reduce noise interference and improve data reliability.

[0066] Step S140: The detection program encapsulates the multi-dimensional pollution values ​​in the smoothed pollution dataset into fixed-length binary data frames according to the timestamp order based on the preset communication protocol structure, thereby obtaining a standardized pollution dataset and realizing the standardized collection and format conversion of multi-dimensional pollution data.

[0067] Binary data frames are derived using the following formula: (6); In formula (6), For the first The first sensing node Pollution indicators in time A binary data frame with no physical unit; its source is the result of data encapsulation; its value is a fixed-length binary sequence (such as 64-bit or 128-bit, preset according to the communication protocol); it is defined as a standardized contaminated data transmission unit for remote transmission in the Internet of Things and subsequent data parsing and processing. This is the header of the data frame, without physical units; the source is a preset IoT communication protocol; the value is a fixed-length binary sequence (e.g., 8 bits); it is defined as the identification field of the data frame, used to identify the type, source, and communication protocol version of the data frame, ensuring that the data frame can be correctly parsed. This is a timestamp field, with the unit being seconds (s); the source is the current system time; the value is the binary sequence corresponding to the time; it is defined as the time identifier of the data frame, used to associate the data collection time, ensuring the consistency and traceability of time-series data. This is a checksum field with no physical unit; its source is the data frame header, timestamp, and verification calculations of corrupted data (such as CRC checksum algorithm); its value is a fixed-length binary sequence (such as 8 bits); it is defined as the integrity check field of the data frame, used to detect whether there are any abnormalities such as loss or tampering during data transmission. XOR concatenation is an operation without physical units; it is defined as concatenating multiple binary fields into a complete binary sequence according to a preset order, ensuring the data frame has a standardized structure and format, and is suitable for remote transmission requirements. The control logic of formula (6) is to concatenate the data frame header according to the preset IoT communication protocol structure. Current timestamp Smoothed contaminated data and checksum It is encapsulated into a fixed-length binary data frame through XOR concatenation operation. After integrating the binary data frames of all sensor nodes and all pollution indicators, a standardized pollution dataset is obtained, realizing the standardized encapsulation of multi-dimensional pollution data and ensuring the standardization, integrity and reliability of remote data transmission. Formula (6) breaks through the limitation of traditional data encapsulation that only contains the core data itself, and innovatively introduces timestamp and checksum fields. The timestamp ensures the temporal correlation of the data, which is convenient for subsequent temporal analysis and data traceability; the checksum realizes the integrity verification in the data transmission process, effectively avoiding data loss, tampering and other problems; at the same time, it adopts fixed-length encapsulation and XOR splicing method to accurately adapt to the bandwidth requirements of IoT remote transmission, solves the technical problem of non-standardized and error-prone transmission of multi-dimensional and multi-node pollution data, and significantly improves the reliability and efficiency of data transmission.

[0068] The detection program encapsulates data according to a preset communication protocol structure. For example, it places multi-dimensional pollution values ​​such as pH and dissolved oxygen into fixed-length binary frames in timestamp order. The frame header includes synchronization bytes, timestamps, and data length, followed by pollution value bytes, resulting in a standardized dataset that achieves efficient transmission and compatibility. This standardized acquisition and format conversion improves the real-time performance and accuracy of data processing in environmental monitoring.

[0069] Preferably, the IoT-based intelligent environmental pollution detection and control method provided in this embodiment includes step S200 as follows:

[0070] Step S210: For the standardized pollution dataset, the spatial dimensions of the multi-dimensional pollution values ​​are divided using a clustering analysis algorithm to determine the spatial distribution matrix.

[0071] The spatial distribution matrix is ​​determined using the following formula: (7); (8); In formulas (7)~(8), This is the clustering objective function value, without physical units; it is derived from the sum of squared Euclidean distances from all samples to their corresponding cluster centers during the clustering process; its value range is... Defined as an evaluation metric for clustering performance. The smaller the value, the higher the similarity of samples within the cluster, and the better the clustering effect. The number of clusters is preset and has no physical units; it is derived from preset clustering parameters (set according to the complexity of the contamination distribution); the value range is... Defined as the total number of cluster categories in a polluted area, ensuring that areas with different levels of pollution and different spatial distributions can be effectively distinguished. For the first There are 10 clusters, with no physical units; the source is the output of the K-means clustering algorithm; the value is (sensor node index). Pollution Indicator Type Index A sample set is defined as a set of samples with similar pollution values, corresponding to a specific spatial distribution area of ​​pollution. For the first The central value of each cluster is in units consistent with the actual physical quantity of the pollution index; it is derived from the mean of all samples within the cluster; the range of values ​​depends on the type of pollution index; it is defined as the typical pollution concentration value corresponding to the cluster, which can represent the overall pollution level of the polluted area. The square of the Euclidean distance has no physical unit; it is defined as the square of the difference between two numerical values ​​and is used to quantify the similarity between a sample and a cluster center. The smaller the distance, the higher the similarity between the sample and the cluster center. This is a spatial distribution matrix with no physical units; it originates from the integration of core parameters in the clustering results; its values ​​are... A real matrix is ​​defined as a structured matrix that characterizes the spatial distribution features of multi-dimensional pollution data, clearly presenting the core feature parameters of each pollution area. For the first The number of samples in each cluster, without physical units; sourced from the statistical results of samples within that cluster; value range is... ( This represents the total number of sensor nodes. (Number of pollution index types); defined as the number of pollution data samples contained in the cluster, indirectly reflecting the coverage of the polluted area. For the first The standard deviation of each cluster is in units consistent with the actual physical quantity units of the pollution index; it is derived from the standard deviation calculation of all samples within the cluster; the range of values ​​depends on the type of pollution index; it is defined as the degree of fluctuation of pollution data within the cluster, reflecting the stability of the pollution level in the polluted area. The control logic of formula (7) is to use the K-means clustering algorithm to smooth the polluted data. For clustering samples, the goal is to minimize all samples to their respective cluster centers. Sum of squared Euclidean distances "Using '" as the objective function, the multi-dimensional pollution values ​​are divided according to their spatial distribution characteristics into... Clusters Then, using formula (8), the core parameters (cluster center values) of each cluster are calculated. Sample size Standard deviation Integrate into a spatial distribution matrix This enables the spatial division of multi-dimensional pollution data, clearly defining the distribution characteristics and pollution levels of different pollution areas. Formulas (7) and (8) innovatively apply the K-means clustering algorithm to the spatial division of multi-dimensional pollution data, breaking through the limitations of traditional division of pollution areas based on geographical location. By clustering based on the similarity of pollution values, it can accurately identify areas with similar pollution concentrations, effectively avoiding the misclassification problem of "similar geographical locations but large differences in pollution levels". At the same time, by integrating key clustering parameters through the spatial distribution matrix, it provides structured and standardized data support for subsequent spatiotemporal coupling analysis, solving the technical problem of unclear spatial distribution characteristics of multi-source pollution data and difficulty in accurate analysis, and improving the scientificity and accuracy of pollution area division.

[0072] The standardized pollution dataset contains concentration data of various pollutants collected from industrial areas, such as heavy metal content and organic compound levels. This data is acquired through wireless sensor networks and preliminarily processed to form a multi-dimensional numerical sequence. For example, for spatial dimension partitioning, clustering algorithms such as the K-means method are applied to these values. First, the Euclidean distance of each data point is calculated, and similar pollution values ​​are grouped. For instance, heavy metal data points from the upper reaches of a river are clustered into one cluster, and organic matter data points from the lower reaches into another. The cluster center positions are iteratively optimized until convergence, thus determining the spatial distribution matrix. This spatial distribution matrix uses rows to represent different regions and columns to represent pollution dimensions. For example, the matrix element value is the average heavy metal concentration in a specific region. This organizes the scattered data into a structured spatial representation, facilitating subsequent analysis of differences in pollution patterns across different regions.

[0073] Step S220: Based on the spatial distribution matrix, use the DSP chip to execute the time-series correlation analysis algorithm to detect the synchronicity of numerical changes of different pollution sources and determine the spatiotemporal coupling correlation map.

[0074] The spatiotemporal coupling correlation map is determined using the following formula: (9); (10); In formulas (9)~(10), For the first The cluster and the first The temporal correlation coefficients of each cluster, without physical units; sourced from temporal correlation analysis results; value range is [value range missing]. Defined as the degree of synchronicity in the temporal changes of contamination data between two clusters. The closer to 1, the stronger the synchronicity of pollution changes between the two clusters; the closer to -1, the stronger the antisynchronicity; and the closer to 0, the less obvious the temporal correlation. Covariance is a function whose unit is the square of the actual physical quantity of the corresponding pollution index; it is defined as a quantitative index of the degree of linear correlation between two time series, and the larger the absolute value of the covariance, the stronger the linear correlation. It is a variance function, with units equal to the square of the actual physical quantity of the corresponding pollution index; it is defined as a quantitative index of the degree of fluctuation of a single time series itself. The larger the variance, the more drastic the fluctuation of the series. For the first Clusters in time The average pollution value, with units consistent with the actual physical quantities of the pollution indicators; sourced from all samples within this cluster over time. The mean value is calculated; the range of values ​​depends on the type of pollution indicator; it is defined as the time-series change sequence of the pollution level of the cluster, and is used for time-series correlation analysis. This is a spatiotemporal coupling correlation matrix, without physical units; it originates from the aggregation of correlation coefficients between all clusters; its value is... A real matrix is ​​defined as a matrix that characterizes the spatiotemporal correlation between different clusters (contaminated areas), i.e., a spatiotemporal coupling correlation map. The matrix elements are the temporal correlation coefficients of the corresponding clusters, which intuitively reflect the spatiotemporal coupling characteristics of pollution. This is a cluster index with no physical units; its value range is... Defined as an auxiliary index to distinguish different clusters, used to calculate the temporal correlation coefficient between any two clusters. The control logic of formula (9) utilizes the high-speed computing power of the DSP chip to calculate the temporal correlation coefficient between any two clusters. and Pollution data time series , Pearson correlation coefficient The Pearson correlation coefficient is used to quantify the temporal synchronicity of pollution levels in two clusters (different pollution areas); subsequently, the temporal correlation coefficients between all clusters are integrated using formula (10). Association graph matrix In other words, the spatiotemporal coupling correlation map clearly shows the synchronicity and correlation of pollution value changes between different pollution sources and different pollution areas, and realizes the quantitative characterization of the spatiotemporal coupling relationship of pollution. Formulas (9) to (10) combine the high-speed computing advantages of DSP chips to realize the rapid and efficient calculation of the temporal correlation coefficient of multiple clusters, and specifically solve the technical problem of large computational load and low efficiency in the time series analysis of multi-source pollution data; at the same time, the degree of pollution correlation in different regions is quantified through the correlation map matrix, breaking through the limitations of traditional qualitative analysis, realizing the quantitative and visual characterization of the spatiotemporal coupling relationship of pollution, providing accurate correlation basis for subsequent pollution type identification and trend prediction, and significantly improving the accuracy and efficiency of pollution analysis.

[0075] Based on this spatial distribution matrix, DSP chips such as the TMS320 series processors execute time-series correlation analysis algorithms to detect the synchronicity of numerical changes in different pollution sources, such as factory emissions and agricultural runoff. Specifically, the time-series correlation analysis algorithm is used to calculate the Pearson correlation coefficient and compare the similarity between time series. For example, for the heavy metal series from factory emissions and the organic matter series from agricultural runoff, if the coefficient is greater than 0.8, they are considered to be synchronous changes. By constructing a graph with nodes representing pollution sources and edges representing correlation strength, a spatiotemporal coupled correlation graph is obtained. This spatiotemporal coupled correlation graph visualizes the spatial and temporal linkage of pollution, such as how upstream emissions affect downstream water quality.

[0076] Step S230: For the spatiotemporal coupling correlation map, the principal component analysis algorithm is used to reduce the dimensionality of the multi-dimensional pollution values ​​and determine the pollution type feature vector.

[0077] The pollution type feature vector is determined using the following formula: (11); (12); (13); In formulas (11)~(13), This is the covariance matrix of multi-dimensional contamination data, without physical units; it originates from covariance calculation results; and its values ​​range from [value missing]. real matrix ( (where is the number of pollution index types); defined as a matrix representing the linear correlation between different pollution indices, and is the core input for eigenvalue decomposition. This represents the total number of contaminated data samples, without physical units; the source is the statistical results of samples from a standardized contaminated dataset; the value range is... Defined as the number of valid contaminated data samples participating in the dimensionality reduction process, ensuring the reliability of the dimensionality reduction results. For the first A multi-dimensional vector of each contaminated data sample, without physical units; values ​​range from... 3D real vector ( (This represents the number of pollution index types); it is defined as a vector containing the values ​​of all pollution indexes for a single sample, and is the basic sample for calculating the covariance matrix. This is the mean vector of multi-dimensional contamination data, without physical units; its value ranges from... A dimensional real vector is defined as a vector formed by integrating the means of all contaminated data samples and is used for calculating the covariance matrix. The first covariance matrix is ​​the first... Each eigenvalue has no physical unit; it originates from the eigenvalue decomposition result; its value range is [value range missing]. It is defined as a quantitative indicator of the importance of the corresponding feature vector. The larger the feature value, the higher the contribution of the corresponding feature vector and the richer the contamination information it contains. The first covariance matrix is ​​the first... Each eigenvector has no physical unit; it originates from the eigenvalue decomposition result; and its value ranges from [value]. Dimensional real vector; defined as a vector representing a core pollution feature, used for dimensionality reduction processing of multi-dimensional data. This is a feature vector representing the pollution type, without physical units; its source is the integration of key feature vectors; its value is... real matrix ( The feature dimensions after dimensionality reduction. ); is defined as a structured vector containing all core pollution features, used for subsequent pollution type classification and key indicator extraction. This is an eigenvalue / eigenvector index, with no physical unit; its value range is... Defined as a unique index that distinguishes different eigenvalues ​​and eigenvectors. The feature dimension after dimensionality reduction has no physical units; it is derived from the cumulative contribution rate of feature values; the value range is... ; is defined as the number of core features retained after dimensionality reduction, ensuring a cumulative contribution rate ≥ 85%, balancing dimensionality reduction efficiency and information retention rate. The control logic of formulas (11) to (13) is: to calculate the multi-dimensional contamination data matrix through formula (11). covariance matrix The linear correlation between different pollution indicators is quantified; the covariance matrix is ​​decomposed using formula (12) to obtain the corresponding eigenvalues. and eigenvectors Select the top-ranked feature values. The feature vectors (cumulative contribution rate ≥ 85%) are integrated into a pollution type feature vector using formula (13). This achieves dimensionality reduction of multi-dimensional pollution data, significantly reducing data dimensionality while retaining core pollution features, thus improving subsequent processing efficiency. Formulas (11) to (13) use principal component analysis (PCA) to achieve dimensionality reduction of multi-dimensional pollution data, specifically addressing the technical challenges of data redundancy and high computational complexity of multiple pollution indicators; by using the screening criterion of "cumulative contribution rate ≥ 85%", redundant information is effectively eliminated while retaining core pollution information to the maximum extent, and multicollinearity interference between different pollution indicators is avoided; the generated pollution type feature vector This provides concise and efficient feature data for subsequent pollution type classification and key indicator extraction, significantly improving the efficiency and accuracy of subsequent processing.

[0078] For this spatiotemporal coupling correlation map, the principal component analysis algorithm calculates the covariance matrix and extracts the first few principal components to reduce the dimensionality of multi-dimensional values. For example, it reduces 10 pollution dimensions to 3 principal components. These components capture more than 80% of the variance, thereby determining the pollution type feature vector. For example, the first component in the pollution type feature vector represents the pollution type dominated by heavy metals.

[0079] Step S240: Based on the pollution type feature vector, use the threshold segmentation algorithm to divide the pollution type feature vector into numerical intervals and determine the key indicator sequence for different pollution types.

[0080] The key indicator sequence is determined using the following formula: (14); (15); In formulas (14)~(15), For the first The cluster in the th order of ... A binary identifier for the feature dimension, without physical units; sourced from threshold segmentation results; value is 0 or 1; defined as indicating whether the cluster contains the feature dimension. Each feature dimension corresponds to a binary parameter of the pollution type, where 1 indicates inclusion and 0 indicates exclusion. For the first The cluster in the th order of ... The feature values ​​of each feature dimension have no physical units; their source is the pollution type feature vector. The range of values ​​is Defined as the core feature quantification value of the cluster in this feature dimension, it is the core basis for threshold segmentation. For the first The minimum threshold for each feature dimension, without physical units; sourced from preset pollution type feature threshold parameters; value range is... The threshold value is defined as the minimum feature value threshold corresponding to the pollution type of the feature dimension. If the feature value is lower than the threshold value, it is determined that the pollution type is not included. For the first The maximum threshold for each feature dimension has no physical units; the source is a preset pollution type feature threshold parameter; the value range is... The threshold value is defined as the highest feature value threshold corresponding to the pollution type of the feature dimension. If the feature value is higher than the threshold value, it is determined that the pollution type is not included. For the first The key indicator sequence corresponding to each feature dimension has no physical unit; it is derived from the binary identifier integration result of formula (15); and its value is... 3D binary vector ( (This refers to the number of clusters); it is defined as a structured sequence that characterizes the distribution of the pollution type corresponding to this feature dimension in each cluster, i.e., the key indicator sequence of this pollution type. This is a feature dimension index, without physical units; the value range is... ( (The feature dimension after dimensionality reduction); defined as a unique index to distinguish different feature dimensions, each feature dimension corresponding to a key feature of a pollution type. The control logic of formula (14) is for the pollution type feature vector. Each feature dimension Preset the corresponding numerical range threshold , For each cluster eigenvalues ​​in this feature dimension Threshold segmentation is performed to obtain binary identifiers. (1 indicates that the cluster contains the pollution type corresponding to this feature dimension, and 0 indicates that it does not contain it); through formula (15), each feature dimension is... The corresponding binary identifiers of all clusters are integrated into a key indicator sequence. This method clarifies the key indicators corresponding to different pollution types, providing core input for subsequent pollution trend prediction and control instruction generation. Formulas (14) to (15) adopt an adaptive threshold segmentation algorithm, setting exclusive threshold intervals for the feature dimensions of different pollution types. This breaks through the limitations of traditional fixed threshold segmentation, accurately distinguishing the feature differences of different pollution types and avoiding misjudgment. At the same time, through binary identifiers and key indicator sequences, the dimensionality-reduced feature data is transformed into a structured sequence that can be directly used for trend prediction, solving the technical problem of unclear pollution type features and difficulty in extracting key indicators. This provides accurate and efficient core basis for subsequent prediction analysis and control instruction generation.

[0081] Based on the feature vector of the pollution type, threshold segmentation algorithms such as the Otsu method are used to divide the numerical range. For example, the range of vector values ​​from 0 to 1 is divided into three intervals: low, medium, and high. This determines the key indicator sequences for different pollution types, such as the high interval sequence for heavy metals, thereby achieving accurate extraction of the distribution characteristics of pollution types. This approach supports targeted governance strategies in environmental monitoring.

[0082] Furthermore, in the IoT-based intelligent environmental pollution detection and control method provided in this embodiment, step S300 includes:

[0083] Step S310: Construct a long short-term memory network model based on the key indicator sequence, and determine the predicted pollution level within the future time window by iteratively calculating historical pollution values.

[0084] The predicted pollution level is determined using the following formula: (16); (17); (18); In formulas (16)~(18), For LSTM models in time The cell state has no physical unit; it is derived from the cell state update result of formula (16); its value is a real vector; it is defined as the core state of the LSTM model to memorize the features of historical pollution data, and is used to store key information of long-term pollution data. For LSTM models in time The forget gate output has no physical units; its value range is... Defined as controlling the historical cell state. The closer the retention ratio parameter is to 1, the more historical data information is retained. It is an element-wise multiplication (Hadamard product) with no physical unit; it is defined as the element-wise multiplication of two vectors, used to achieve precise regulation of cell state and hidden state by gating signals. For LSTM models in time The cell state has no physical unit; its value is a real vector; it is defined as the model cell state at the previous time step and is used for updating and iterating the cell state at the current time step. For LSTM models in time The input gate output has no physical unit; its value range is... Defined as controlling the current input information The closer the value of the percentage of information entering the cell state is to 1, the higher the retention rate of the current input information. It is a hyperbolic tangent activation function with no physical unit; its value range is... Defined as a nonlinear transformation of input parameters, used to adjust the numerical range of cell states and hidden states, thereby improving the model's nonlinear fitting ability. This is the weight matrix for updating cell states, with no physical units; it originates from the training parameters of the LSTM model; its values ​​are real matrices; it is defined as a weight parameter that controls the degree of influence of input information on cell states, used to adjust the contribution of input information. For LSTM models in time The hidden state has no physical unit; its value is a real vector; it is defined as the model output state of the previous time step and is used for the fusion processing of the input information at the current time step. For time The key indicator sequence has no physical units; its value is... A binary vector; defined as the input feature of the LSTM model, containing the core contamination feature information at the current time step. This is the bias term for cell state updates, with no physical units; it originates from the training parameters of the LSTM model; its value is a real vector; it is defined as a parameter to correct the bias in cell state updates and improve the model's prediction accuracy. For LSTM models in time The hidden state has no physical unit; it is derived from the hidden state update result of formula (17); its value is a real vector; it is defined as the output state of the model at the current moment, containing the core feature information of the current and historical pollution data. For LSTM models in time The output gate outputs a value that has no physical unit; its value range is... Defined as controlling cell state Output to hidden state The closer the value of the proportional parameter is to 1, the higher the output ratio of the cell state. For the future The predicted pollution level is in units consistent with the actual physical quantity units of the pollution index; it is derived from the LSTM model prediction results of formula (18); the range of values ​​depends on the type of pollution index; it is defined as the predicted pollution concentration at a certain future moment, used to determine whether the pollution control process needs to be initiated. This is the output layer weight matrix, without physical units; it originates from the training parameters of the LSTM model; its values ​​are real matrices; it is defined to control the hidden states. The mapping weights to the predicted values ​​are used to adjust the contribution of the hidden states to the prediction results. This is the output layer bias term, with no physical unit; it originates from the training parameters of the LSTM model; its value is a real number; it is defined as a parameter to correct the bias of the prediction value, further improving the prediction accuracy of the model. The forecast time window is in hours (h); the source is preset forecast parameters; the value range is... ; is defined as the predicted future time interval, used to determine the time node of the predicted pollution level, adapting to the pollution control needs of different scenarios. The control logic of formulas (16) to (18) is to construct a Long Short-Term Memory (LSTM) prediction model using formulas (16) and (17), with key indicator sequences Input to the model, passing through the forget gate Input gate and output gate Synergistic effect, dynamically updating the cell state of the model and hidden state This enables iterative memorization of the long-term and short-term characteristics of historical pollution data; through formula (18), the hidden state is... Output layer weights and bias Mapping to obtain the future time window Predicted pollution levels within This enables accurate prediction of pollution trends. Formulas (16) to (18) introduce the LSTM model to solve the technical problem of difficulty in capturing long-term and short-term dependencies in traditional time series forecasting. It can effectively remember the long-term change trend and short-term fluctuation characteristics of historical pollution data. Compared with traditional prediction models such as ARIMA and BP neural networks, the prediction accuracy is significantly improved. At the same time, by using the key indicator sequence as the model input, the interference of redundant data is effectively avoided, and the prediction efficiency is greatly improved. It solves the technical problems of low accuracy in environmental pollution trend prediction and difficulty in capturing long-term and short-term change patterns, and provides an early prediction basis for subsequent pollution control, realizing the foresight of pollution control.

[0085] For air pollution monitoring in industrial parks, key indicator sequences include historical concentration data of PM2.5 and NO2, collected from sensors to form a time-series input. Long Short-Term Memory (LSTM) network models process these key indicator sequences by constructing a multi-layered neural network structure. For example, the LTM network model includes input gates, forget gates, and output gates. The input gate determines the process of adding new information to the memory unit, the forget gate controls the retention ratio of old information, and the output gate generates the current state output. Specifically, during the training phase, the LTM network model uses the backpropagation algorithm to adjust weights and iteratively processes historical values. For example, inputting the PM2.5 key indicator sequence from the past 24 hours into the LTM network model, the predicted value for the next 6 hours is calculated through updating the hidden state. This process involves time-step vector operations to ensure the capture of long-term dependencies in the key indicator sequence, thereby obtaining accurate pollution level predictions, such as predicting that the PM2.5 concentration will rise to 150 micrograms per cubic meter.

[0086] Step S320: If the predicted pollution level exceeds the preset threshold, the pollution source response function is invoked to determine the operating power parameters of the pollution control equipment based on the difference between the predicted pollution level and the preset threshold.

[0087] The operating power of pollution control equipment is calculated using the following formula: (19); In formula (19), The operating power of the pollution control equipment is expressed in kilowatts (kW); the source is the calculation result of the response function of formula (19); the value range is... Defined as the core operating parameter of pollution control equipment, it directly determines the treatment effect and energy consumption level. This represents the difference between predicted and standard pollution levels, with units consistent with the actual physical quantities of the pollution indicators; the source is the calculation of the difference between predicted and standard pollution levels. The range of values ​​is Defined as the degree of deviation between the predicted pollution level and the standard value, it is used to determine whether enhanced treatment needs to be initiated and to determine the appropriate power. The minimum operating power of the pollution control equipment, expressed in kilowatts (kW); sourced from the equipment's preset hardware parameters; value range is... Defined as the minimum power required for equipment to maintain basic operation, used for basic monitoring and light treatment when pollution levels are not exceeded. The maximum operating power of the pollution control equipment is expressed in kilowatts (kW); it is derived from the equipment's preset hardware parameters; the value range is... Defined as the maximum power at which the equipment achieves the best treatment effect, used for enhanced treatment when pollution levels are severely exceeded. This is the maximum permissible exceedance difference, with units consistent with the actual physical quantity units of the pollution indicator; the source is the maximum permissible exceedance limit preset by the national environmental standards; the value range depends on the type of pollution indicator (e.g., 50 mg / L for COD); it is defined as the critical value of the degree of pollution exceedance, used for normalization calculation of operating power, and to avoid power regulation imbalance. The pollution value is the national environmental standard value, and the unit is consistent with the actual physical quantity unit of the pollution indicator; the source is the national environmental monitoring standard (such as GB 3838-2002); the range of values ​​depends on the type of pollution indicator; it is defined as the benchmark value for pollution control, used to calculate the pollution prediction difference and determine the degree of pollution exceeding the standard. The control logic of formula (19) is to call the preset pollution source response function, and first calculate the pollution level prediction value. pollution values ​​in accordance with national environmental standards Difference ;like This indicates that the predicted pollution level did not exceed the standard, and the minimum operating power was taken. Maintain basic monitoring; if Then, based on the difference between the difference and the maximum allowable exceedance difference... The ratio of adaptive calculation of the operating power of pollution control equipment This ensures that the operating power is precisely matched with the degree of pollution exceeding the standard, achieving a balance between energy saving and treatment effect, and providing core power parameters for the generation of subsequent control commands. Formula (19) breaks through the extensive mode of "fixed power operation" of traditional pollution control equipment, constructs an adaptive response function based on the pollution prediction difference, and realizes dynamic adjustment of operating power; by introducing the maximum allowable difference of exceedance. This avoids excessive or insufficient power adjustment, ensuring that pollution control results meet standards while minimizing equipment energy consumption. It solves the technical problems of high energy consumption and poor adaptability of traditional treatment equipment. At the same time, it connects with LSTM prediction results to achieve forward-looking adjustment of treatment strategies, improving the accuracy and economy of pollution control.

[0088] If the predicted pollution level exceeds a preset threshold, such as 100 micrograms per cubic meter, the pollution source response function is invoked. This function is based on the difference. For example, if the difference is 50, the operating power parameter is determined to be 80% through a linear mapping formula. This operating power parameter corresponds to the adjustment of the fan speed of the purification equipment to match the response requirements.

[0089] Step S330: Match the predefined control instruction set according to the operating power parameters, map the operating power parameters into a binary control instruction sequence, and determine the instruction data packet to be sent.

[0090] The standardized control commands for the corresponding treatment equipment are derived using the following formula: (20); (twenty one); In formulas (20)~(21), For the first Each cluster corresponds to the control command intensity of the governance equipment, which has no physical unit; it is derived from the quantitative calculation results of formula (20); the value range is... Defined as the execution intensity of control commands, the higher the value, the higher the operating priority and execution intensity of the governance equipment. This is a power weighting coefficient, with no physical unit; it originates from preset control parameters; its value range is... Defined as the weight of the influence of operating power on the intensity of control commands, ensuring that power parameters are the core basis for control. These are correlation weighting coefficients, without physical units; they originate from preset control parameters; their value range is... Defined as the weight of the influence of spatiotemporal correlation characteristics on the strength of control commands, to ensure the synergy of governance in different regions. . For the first Each cluster corresponds to the operating power of the treatment equipment, in kilowatts (kW); the source is the calculation result of formula (19); the value range is the same as... Consistency; defined as the power parameters of the associated regional governance equipment, used to achieve coordinated control. For the first Each cluster corresponds to a standardized control instruction for the governance device, without physical units; the source is the encoding result of formula (21); the value is a binary instruction sequence of fixed length; it is defined as a control instruction that the governance device can directly parse and execute, containing key information such as execution intensity, device address, and execution time. This is an instruction encoding function with no physical unit; it is defined as a function that converts information such as control instruction strength, device address, and timestamp into standardized binary instructions, ensuring the standardization and compatibility of the instructions. For the first Each cluster corresponds to a unique address of the governance device, without a physical unit; it originates from the hardware address preset by the IoT device; its value is a fixed-length string (such as a MAC address); it is defined as a unique identifier to distinguish different governance devices, ensuring that control commands are accurately delivered to the target device. Through formula (20), combined with the operating power determined in step S320... Temporal correlation coefficients in spatiotemporal coupling correlation maps Introducing weighting coefficients (Power weight) and (Association weight), quantification calculation of the first The control command strength of the governance equipment corresponding to each cluster This achieves the coupling of power parameters with spatiotemporal correlation characteristics; subsequently, through formula (21), the control command strength and device address are... and predicted timestamp Encode and generate standardized control instructions. This ensures the uniqueness, standardization, and executability of the instructions, providing a clear basis for equipment execution. Formulas (20) and (21) break through the limitations of traditional control instructions based solely on a single power parameter, innovatively incorporating spatiotemporal coupling characteristics (time-series correlation coefficient). This approach considers the interconnected impacts of different pollution areas to avoid pollution spread caused by the treatment of a single area; it adjusts the contribution of power and correlation features through weighting coefficients to adapt to different pollution scenarios; and it adopts standardized coding methods to ensure the compatibility of instructions with IoT governance devices, solving the technical problems of poor targeting of control instructions and easy occurrence of governance blind spots, and improving the systematicness and synergy of pollution control.

[0091] The control instruction set is matched according to the operating power parameter. For example, 80% power is mapped to a binary sequence such as 10100000, and then an instruction data packet is formed.

[0092] Step S340: Establish an encrypted communication link with the execution terminal through the remote transmission module, and encapsulate the instruction data packet into a transmission protocol frame.

[0093] The reliability of control command transmission is derived from the following formula: (twenty two); In formula (22), To control the reliability of command transmission, the unit is percentage (%); the source is the calculation result of formula (22); the value range is... Defined as the pass / fail level of instruction transmission, used to determine whether the transmission is valid, with a preset pass / fail threshold of 99%. The number of successfully received control commands, without physical units; sourced from statistical results of IoT transmission modules; value range is [missing value]. Defined as the number of instructions successfully delivered to the target device and parsed. This represents the number of control commands sent, without physical units; it originates from statistical results from the system's sending module; its value range is... Defined as the total number of control commands sent by the system to the target device.

[0094] The execution feedback information is generated using the following formula: (twenty three); in, For the first The execution feedback information of each governance device has no physical unit; the source is the execution feedback result of formula (23); the value is structured data containing execution status (success / failure), actual operating power and execution time; it is defined as the status feedback after the device executes the instruction, and is used for subsequent adjustment of the control strategy of the system. This represents the instruction execution time, in minutes (min); the source is the actual time taken by the device to execute the instruction; the value range is... Defined as the time from when the governance device receives an instruction to when it completes execution, used for time calibration of feedback callbacks. This is a device execution function with no physical unit; it is defined as a function that parses control commands, executes pollution control operations at preset power, and generates feedback information. It is the core logic of command execution. For the first The current operating status of each treatment device has no physical unit; the value is "normal / fault / standby"; it is defined as the initial state before the device executes the command, used to determine whether the device has the conditions to execute, and to avoid invalid execution by faulty devices. The formula control logic calculates the transmission reliability of the control command through formula (22). That is, the number of successfully received instructions. With the number of sent instructions The ratio, if If the transmission fails, a retransmission mechanism is triggered to ensure that the instructions are accurately delivered to the target treatment equipment. Then, the treatment equipment executes the received standardized control instructions through formula (23). Combined with preset operating power and the current operating status of the equipment Complete the pollution control operation and generate execution feedback information. Feedback is sent to the system terminal to achieve closed-loop control of instruction transmission and execution. Formulas (22) to (23) introduce a transmission reliability verification mechanism. By quantifying the transmission success rate, the technical problem of instruction loss and mistransmission during IoT transmission is solved, ensuring the effective delivery of control instructions. At the same time, a device execution feedback function is constructed to collect the execution status in real time, realizing a closed-loop process of "instruction sending - transmission verification - device execution - feedback callback". This breaks through the limitation of "only sending and not checking" in traditional control, making it easier for the system to detect execution anomalies in a timely manner and adjust strategies, thereby improving the reliability and traceability of pollution control.

[0095] The remote transmission module uses AES encryption to establish a link and encapsulates data packets into TCP / IP protocol frames to send them to the execution terminal.

[0096] Step S350: The terminal receives the transmission protocol frame and parses the instruction data packet, drives the pollution control equipment according to the parsed operating power parameters, determines the real-time operating status of the pollution control equipment, and realizes automatic control of the pollution level.

[0097] Operating power deviation is calculated using the following formula: (twenty four); In formula (24), For the first The operating power deviation of each treatment device is expressed in kilowatts (kW); the source is the deviation calculation result of formula (24); the value range is... Defined as the difference between the actual power of the device and the preset power, it is used to quantify the execution deviation and provide a basis for parameter optimization. For the first The actual operating power of each treatment device is expressed in kilowatts (kW); the source is the device's execution feedback information. ; range of values ​​and Consistency; defined as the power of the equipment when it actually performs the treatment operation, used to compare with the preset power to calculate the deviation. For the first The preset operating power of each treatment device is in kilowatts (kW); the source is the calculation result of formula (19); the value range is the same as... Consistency; defined as the operating power preset by the system for the device, which is the benchmark parameter for the device to perform.

[0098] The optimized preset operating power is obtained using the following formula: (25); In formula (25), For the next forecast period, the first The optimized preset operating power of each treatment device is expressed in kilowatts (kW); the source is the parameter optimization result of formula (25); the value range is consistent with... Consistent; defined as the optimized reference power parameter, used to improve the control accuracy of the next cycle. This is a parameter adjustment coefficient, without physical units; it originates from preset optimization parameters; its value range is... Defined as the constraint coefficient of power deviation on the optimization step size, it is used to limit the parameter adjustment range, ensure that the optimization process is stable and efficient, and avoid equipment damage and energy waste caused by excessive power adjustment. This is a sign-based function for determining exceedances, with no physical units; the user-defined rule is: when the input is the pollution prediction difference... The function outputs 1 when the input is... The time function outputs 0; this is used to determine whether the current pollution is in an excessive state, and to determine the execution logic of parameter optimization: when the pollution exceeds the standard, power reduction optimization is initiated; when the pollution does not exceed the standard, the baseline power is maintained and no additional adjustments are made. For the current cycle number The baseline preset power of the pollution control equipment is used for time-series recursive optimization to improve the pollution control accuracy in the next cycle. The formula control logic is to calculate the actual operating power of the pollution control equipment using formula (24). With preset operating power deviation The degree of execution deviation is quantified; then, an adjustment coefficient is introduced through formula (25). Combined with feedback deviation The sign of the difference between the pollution forecast and the actual pollution level (to determine whether pollution still exceeds the standard) is used for the next forecast period. The preset operating power is optimized and adjusted; at the same time, the feedback deviation and the optimized parameters are fed back to the LSTM model in step S310 to realize the iterative update of the model parameters, forming a closed-loop iterative mechanism of "prediction-control-feedback-optimization" to continuously improve the pollution control accuracy. Formulas (24) to (25) break through the limitation of "one-time command execution" in traditional control by constructing a closed-loop iterative optimization mechanism, and optimize the preset power parameters in real time through feedback deviation, thus solving the problem of poor treatment effect caused by equipment execution deviation; at the same time, the feedback information is integrated into the LSTM prediction model to realize the synergistic optimization of the prediction model and the control strategy, thereby improving the subsequent prediction accuracy and control adaptability; adjustment coefficient The introduction of this technology ensures the stability of parameter optimization, avoids equipment wear and energy waste caused by excessive adjustments, and further improves the accuracy, economy and stability of pollution control.

[0099] After the terminal parses the protocol frame, it drives the pollution control equipment, such as adjusting the filter power, to achieve real-time status monitoring and thus automatically control the pollution level.

[0100] Preferably, in the IoT-based intelligent environmental pollution detection and control method provided in this embodiment, step S400 includes:

[0101] Step S410: Obtain the response data fed back by the execution terminal. The response data includes real-time running status values.

[0102] For air quality monitoring scenarios within industrial parks, execution terminals such as air purification devices provide feedback data. This data includes real-time operating status values, such as the current SO2 concentration and device wind speed. To acquire this data, data packets are first received from the execution terminal via a wireless communication link. These packets contain a sequence of timestamped values, such as a real-time SO2 concentration reading of 80 micrograms per cubic meter, along with device operating parameters like a wind speed of 5 meters per second. This acquisition process involves a data verification mechanism to ensure the received values ​​are error-free. For example, checksum calculations are used to compare the original and received data for consistency. If the checksum matches, the data is confirmed as valid, providing a reliable foundation for subsequent processing. Further, when processing the response data, the numerical sequence is parsed into a structured format. For instance, the SO2 concentration value is extracted and stored in a local database to support real-time analysis. This method allows intelligent environmental pollution detection and control systems to dynamically respond to environmental changes. For example, when the device detects a sudden pollution event, the response data is updated instantly, including peak concentration and device response time, thus forming a complete feedback loop.

[0103] Step S420: Align the real-time operating status values ​​with the pre-stored predicted pollution levels to determine the control deviation of the closed-loop control between the two.

[0104] The control deviation is determined using the following formula: (26); In formula (26), To control for deviations, the same physical unit as the pollution concentration is used, which is the final output of the formula. The magnitude of the value represents the degree of deviation from the pollution level. For the first The actual pollution value of the area is in the same physical unit as the pollution concentration. It comes from real-time feedback from the Internet of Things monitoring terminal and is taken as the true level of environmental pollution. For the first The actual pollution value fed back from the terminal after full time delay compensation in the region. Pre-store the time-series predicted contamination levels for the system. The original sampling reference time, For IoT data transmission latency, To regulate the execution delay of the control command. The control logic of formula (26) is to calculate the absolute difference between the predicted value and the actual value after aligning the time axis of the entire link delay, and accurately quantify the deviation of the pollution level; through delay compensation and numerical alignment, the timing misalignment error caused by the transmission of IoT and the execution delay of the device is eliminated, and the closed-loop control deviation is accurately quantified, providing a reliable data basis for subsequent adaptive parameter optimization and control command iteration.

[0105] When aligning real-time operating status values ​​with pre-stored predicted pollution levels, the process involves first retrieving predicted values ​​from a database, such as a predicted SO2 concentration of 100 micrograms per cubic meter. Then, a time-axis alignment method is used to match the real-time and predicted values, for example, using linear interpolation to place them at the same time point and calculate the difference. The specific process includes step-by-step comparison: assuming a real-time value of 80 and a predicted value of 100, the deviation is calculated as an absolute difference of 20. The closed-loop control deviation is further quantified, for example, by using a percentage formula to determine a deviation rate of 20%, where the deviation rate is defined as (real-time value - predicted value) / predicted value * 100%. This alignment mechanism considers time delay factors; for example, if the real-time data lags by 1 minute, the system adjusts the predicted value's time window to match, thus accurately determining the deviation. Furthermore, after the deviation is determined, a threshold judgment can be introduced; for example, a deviation exceeding 10% is marked as a high deviation state. This approach ensures the closed-loop characteristics of the intelligent environmental pollution detection and control system. In one embodiment, the above deviation determination helps optimize subsequent control strategies.

[0106] Furthermore, in the IoT-based intelligent environmental pollution detection and control method provided in this embodiment, step S500 includes:

[0107] Step S510: Compare the control deviation with the preset deviation threshold. If the control deviation exceeds the preset deviation threshold, extract the correlation mapping relationship between the weight of each instruction and the corresponding response delay in the current control instruction sequence, and determine the correction step size of each instruction weight.

[0108] The correction step size is determined using the following formula: (27); In formula (27), The step size is a fixed unit with no physical unit. It is the final output of the formula and is used to limit the single adjustment range of the instruction weight. This is a correction factor, without physical units, derived from system preset optimization parameters, and its value range is [range missing]. This is used to constrain the global correction strength and ensure the stability of weight iteration. The preset deviation threshold is a physical unit that is the same as the pollution concentration and is derived from the system's preset upper limit for allowable emission deviation. The delay mapping factor has no physical unit and is derived from the statistical fitting results of historical operating data, representing the correlation between command response delay and command weight. The control logic of formula (27) dynamically calculates the weight correction step size based on the normalized ratio of the control deviation relative to the preset threshold. The greater the control deviation exceeds the threshold, the stronger the weight correction magnitude. At the same time, it combines the command delay mapping factor to realize differentiated weight updates and constrains the overall adjustment boundary through the correction coefficient. This formula realizes deviation-driven adaptive dynamic weight correction, fundamentally avoiding the two technical defects of over-correction and insufficient correction caused by traditional fixed step size updates. It optimizes the command weight allocation by combining command delay characteristics, further improving the command adaptability, operational stability and pollution control accuracy of the entire IoT closed-loop control system.

[0109] In air quality optimization scenarios within industrial parks, when the intelligent environmental pollution detection and control system detects a control deviation exceeding a preset threshold, it initiates an instruction adjustment process. For example, when handling excessive deviations, the system first extracts the correlation mapping between the weight of each instruction in the current control instruction sequence and its corresponding response delay. This correlation mapping is presented in a table format formed from accumulated historical data, where each instruction weight corresponds to a delay value; for example, a purification instruction weight of 0.6 corresponds to a delay of 2 seconds. By analyzing these correlations, the correction step size can be determined; for example, a step size of 0.1 can be calculated based on a linear function of the delay, thus providing a basis for weight adjustment. This extraction process involves database queries and relational modeling to ensure the accuracy of the mapping relationship, supporting subsequent dynamic reconstruction.

[0110] Step S520: Dynamically reconstruct the weights of each instruction in the control instruction sequence according to the correction step size to obtain the corrected instruction weight set, and determine the execution frequency of each instruction weight by combining the matching degree between the current operating load and the environmental baseline.

[0111] The revised instruction weights are derived using the following formula: (28); In formula (28), This represents the corrected instruction weight, which has no physical unit and is the final output of the formula, used to update the instruction issuance priority. This is the original instruction weight, which has no physical unit and comes from the system's initial preset, representing the initial priority of the instruction.

[0112] The execution frequency is determined using the following formula: (29); In formula (29), The execution frequency is expressed in Hz, and the final output of the formula represents the frequency of instruction execution per unit time. The preset frequency for the instruction reference is in Hz, which comes from the system's preset basic parameters and is the original value of the instruction frequency reference. The data used to manage the real-time operating load of the equipment has no physical unit; it originates from real-time data collected by IoT terminals, and the value range is [missing information]. , representing the current load status of the equipment. The control logic of formulas (28) to (29) is based on the modified instruction weight linkage benchmark frequency, combined with the real-time operating load of the equipment to make adaptive inverse proportional adjustment; the higher the operating load of the equipment, the lower the instruction execution frequency is accordingly. While ensuring the pollution control effect, it avoids the equipment overload operation loss under high load, realizes the coordinated adaptation of instruction priority, execution frequency and equipment load, and improves the whole closed-loop control system.

[0113] The weights of each instruction in the control instruction sequence are dynamically reconstructed based on a determined correction step size, resulting in a corrected set of instruction weights. For example, the original weight of 0.6 is adjusted to 0.7, while the sum of all weights in the instruction weight set remains at 1 to maintain instruction balance. Furthermore, the execution frequency of each instruction is determined by combining the matching degree between the current operating load and the environmental baseline. For instance, the frequency increases to once per minute when the load is high. This matching degree is assessed using a similarity algorithm to evaluate how close the environmental parameters are to the baseline, thereby optimizing frequency allocation.

[0114] Step S530: Reorder the modified instruction weight set according to the execution frequency to obtain the timing-optimized instruction sequence, and determine the encoding compression ratio of each instruction in combination with the real-time occupancy of the current transmission bandwidth.

[0115] The encoding compression ratio is determined using the following formula: (30); In formula (30), The compression ratio has no physical unit and is the output of the formula. Its range represents the degree of compression of communication data. This is the baseline compression ratio, which has no physical unit and is derived from the protocol preset. This represents the remaining bandwidth, measured in Mbps (megabits per second), derived from real-time monitoring of the IoT network, and indicates the currently available bandwidth. The total bandwidth is expressed in Mbps (megabits per second), derived from the transmission hardware parameters, and represents the maximum bandwidth limit of the device. The control logic of formula (30) calculates the remaining bandwidth by using the total bandwidth and the real-time occupancy. Then, the ratio of the remaining bandwidth to the baseline bandwidth is used as the core adjustment factor, and the compression degree of a single instruction is dynamically calculated in combination with the baseline compression ratio. When the bandwidth is sufficient, the compression ratio is reduced to improve the data transmission fidelity. When the bandwidth is tight, the compression ratio is increased to ensure timely transmission of instructions, thereby achieving a dynamic balance between the efficiency of pollution control instruction issuance and the quality of data transmission.

[0116] The corrected weight set is then time-reordered to obtain a time-optimized instruction sequence, for example, by placing high-frequency instructions at the beginning of the sequence. Next, the encoding compression ratio is determined based on real-time transmission bandwidth utilization; for example, if bandwidth utilization is 80%, the compression ratio is set to 2:1.

[0117] Step S540: The timing optimization instruction sequence is encapsulated according to the encoding compression ratio to obtain the encapsulated control instruction sequence, and the encapsulated control instruction sequence is sent to the execution terminal through the remote transmission module to realize closed-loop control of environmental pollution.

[0118] The timing-optimized instruction sequence is encapsulated based on the encoding compression ratio to obtain the encapsulated control instruction sequence, which is then sent to the execution terminal via a remote transmission module. This closed-loop control mechanism helps improve the response efficiency of intelligent environmental pollution detection and control systems.

[0119] Please see Figure 2This invention provides an IoT-based intelligent environmental pollution detection and control system for implementing the aforementioned IoT-based intelligent environmental pollution detection and control method. It includes a standardized pollution dataset acquisition module 10, a key indicator sequence extraction module 20, a remote transmission control module 30, a control deviation judgment module 40, and an environmental pollution closed-loop control module 50. The standardized pollution dataset acquisition module 10 collects multi-dimensional pollution data from water, soil, and air sources via a wireless sensor network, and uses a detection program in an embedded Linux kernel to perform preliminary format conversion on the collected pollution data to obtain a standardized pollution dataset. The key indicator sequence extraction module 20 uses a DSP chip to perform data grouping on the standardized pollution dataset. The algorithm determines the distribution characteristics of different pollution types and extracts key indicator sequences from these characteristics. The remote transmission control module 30 uses a predictive analysis model to predict the trend of the key indicator sequences, obtaining the predicted pollution level. If the predicted pollution level exceeds a preset threshold, a control command sequence is generated based on the predicted pollution level and sent to the execution terminal via the remote transmission module. The control deviation judgment module 40 acquires the response data fed back from the execution terminal, compares the response data with the predicted pollution level, and determines the control deviation of the closed-loop control. The environmental pollution closed-loop control module 50 adjusts the parameters of the control command sequence to obtain an optimized control command sequence, which is then retransmitted via the remote transmission module to achieve closed-loop environmental pollution control.

[0120] Comparison data of implementation results:

[0121] To verify the technical effectiveness of this invention, a comparative test was conducted using an online environmental pollution monitoring and control system for an industrial park. The results are shown in Table 1 below: .

[0122] The IoT-based intelligent environmental pollution detection and control method and system provided in this embodiment, compared with the prior art, significantly improves the accuracy of pollution prediction, the speed of treatment response and the compliance rate through multi-source data fusion, LSTM accurate prediction, SVM fluctuation classification, PSO collaborative optimization and closed-loop adaptive adjustment, effectively reducing energy consumption, intensity of manual intervention and pollution risk. The technical effects are clear and reproducible.

[0123] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if these modifications and modifications of the invention fall within the scope of the claims and their equivalents, the invention is also intended to include these modifications and modifications.

Claims

1. A smart environmental pollution detection and control method based on the Internet of Things, characterized in that, Includes the following steps: S100: Collect multi-dimensional pollution data from water, soil and air sources through a wireless sensor network, and use a detection program in the embedded Linux kernel to perform preliminary format conversion on the collected pollution data to obtain a standardized pollution dataset. S200. For the standardized pollution dataset, a data grouping algorithm is executed using a DSP chip to determine the distribution characteristics of different pollution types, and key indicator sequences are extracted from the distribution characteristics. S300. A predictive analysis model is used to predict the trend of the key indicator sequence to obtain the predicted pollution level. If the predicted pollution level exceeds a preset threshold, a control instruction sequence is generated based on the predicted pollution level and sent to the execution terminal through a remote transmission module. S400: Obtain the response data fed back by the execution terminal, compare the response data with the predicted pollution level, and determine the control deviation of the closed-loop control; S500. For the control deviation, adjust the parameters of the control command sequence to obtain an optimized control command sequence, and resend it through the remote transmission module to realize closed-loop control of environmental pollution.

2. The intelligent environmental pollution detection and control method based on the Internet of Things according to claim 1, characterized in that, Step S100 includes: S110: The wireless sensor network node collects multi-dimensional raw pollution signals of water, soil and atmosphere, converts the analog voltage signal into a digital quantization sequence through an analog-to-digital converter circuit, and performs time-series alignment of the digital quantization sequence using a preset sampling frequency to obtain a time-series aligned pollution data stream. S120. The embedded system reads the time-aligned contaminated data stream and performs dimensional mapping on the digital quantization sequence according to a preset physical quantity conversion function to obtain a physical quantity contaminated dataset. S130. The detection program removes outliers from the physical quantity contamination dataset and smooths the fluctuations of the physical quantity contamination dataset using a sliding window mean filtering algorithm to obtain a smoothed contamination dataset. S140. The detection program encapsulates the multi-dimensional pollution values ​​in the smoothed pollution dataset into fixed-length binary data frames according to the timestamp order based on the preset communication protocol structure, thereby obtaining a standardized pollution dataset and realizing the standardized collection and format conversion of multi-dimensional pollution data.

3. The intelligent environmental pollution detection and control method based on the Internet of Things according to claim 1, characterized in that, Step S200 includes: S210. For the standardized pollution dataset, the spatial dimension of the multi-dimensional pollution values ​​is divided by a clustering analysis algorithm to determine the spatial distribution matrix. S220. Based on the spatial distribution matrix, the DSP chip is used to execute a time-series correlation analysis algorithm to detect the synchronicity of numerical changes of different pollution sources and determine the spatiotemporal coupling correlation map. S230. For the spatiotemporal coupling correlation map, the principal component analysis algorithm is used to reduce the dimensionality of the multi-dimensional pollution values ​​to determine the pollution type feature vector. S240. Based on the pollution type feature vector, the pollution type feature vector is divided into numerical intervals using a threshold segmentation algorithm to determine the key indicator sequence for different pollution types.

4. The intelligent environmental pollution detection and control method based on the Internet of Things according to claim 3, characterized in that, Step S300 includes: S310. Construct a long short-term memory network model based on the key indicator sequence, and determine the predicted pollution level within the future time window through iterative calculation of historical pollution values. S320. If the predicted pollution level exceeds a preset threshold, the pollution source response function is invoked to determine the operating power parameters of the pollution control equipment based on the difference between the predicted pollution level and the preset threshold. S330. Match the predefined control instruction set according to the operating power parameters, map the operating power parameters into a binary control instruction sequence, and determine the instruction data packet to be sent; S340. Establish an encrypted communication link with the execution terminal through the remote transmission module, and encapsulate the instruction data packet into a transmission protocol frame; S350. The execution terminal receives the transmission protocol frame and parses the instruction data packet, drives the pollution control equipment according to the parsed operating power parameters, determines the real-time operating status of the pollution control equipment, and realizes automatic control of the pollution level.

5. The intelligent environmental pollution detection and control method based on the Internet of Things according to claim 1, characterized in that, Step S400 includes: S410. Obtain response data fed back by the execution terminal, wherein the response data includes real-time running status values; S420. Align the real-time operating status value with the pre-stored predicted pollution level to determine the control deviation of the closed-loop control between the two.

6. The intelligent environmental pollution detection and control method based on the Internet of Things according to claim 5, characterized in that, Step S500 includes: S510. Compare the control deviation with the preset deviation threshold. If the control deviation exceeds the preset deviation threshold, extract the correlation mapping relationship between the weight of each instruction and the corresponding response delay in the current control instruction sequence, and determine the correction step size of each instruction weight. S520. Dynamically reconstruct the weights of each instruction in the control instruction sequence according to the correction step size to obtain the corrected set of instruction weights, and determine the execution frequency of each instruction weight by combining the matching degree between the current operating load and the environmental baseline. S530. The modified instruction weight set is rearranged in time according to the execution frequency to obtain a time-optimized instruction sequence, and the encoding compression ratio of each instruction is determined in combination with the real-time occupancy of the current transmission bandwidth. S540. The timing optimization instruction sequence is encapsulated according to the encoding compression ratio to obtain the encapsulated control instruction sequence, and the encapsulated control instruction sequence is sent to the execution terminal through the remote transmission module to realize closed-loop control of environmental pollution.

7. The intelligent environmental pollution detection and control method based on the Internet of Things according to claim 6, characterized in that, In step S510, the correction step size is determined using the following formula: ； in, To correct the step size, For correction factor, To control deviation, To set a preset deviation threshold, This is the time delay mapping factor.

8. The intelligent environmental pollution detection and control method based on the Internet of Things according to claim 7, characterized in that, In step S520, the corrected instruction weights are obtained using the following formula: ； in, The corrected instruction weights, The original instruction weight; The execution frequency is determined using the following formula: ； in, For execution frequency, Preset frequency for instruction reference. To monitor the real-time operating load of the treatment equipment.

9. The intelligent environmental pollution detection and control method based on the Internet of Things according to claim 8, characterized in that, In step S530, the encoding compression ratio is determined using the following formula: ； in, The compression ratio is... As the baseline compression ratio, For the remaining bandwidth, This represents the total bandwidth.

10. An Internet of Things (IoT)-based intelligent environmental pollution detection and control system, used to implement the IoT-based intelligent environmental pollution detection and control method as described in any one of claims 1 to 9, characterized in that, include: The standardized pollution dataset acquisition module is used to collect multi-dimensional pollution data from water, soil and atmosphere sources through a wireless sensor network. The detection program in the embedded Linux kernel performs preliminary format conversion on the collected pollution data to obtain a standardized pollution dataset. The key indicator sequence extraction module is used to determine the distribution characteristics of different pollution types by using a data grouping algorithm executed by a DSP chip for the standardized pollution dataset, and to extract key indicator sequences from the distribution characteristics. The remote transmission control module is used to perform trend prediction on the key indicator sequence using a predictive analysis model to obtain the predicted pollution level. If the predicted pollution level exceeds a preset threshold, a control command sequence is generated based on the predicted pollution level and sent to the execution terminal through the remote transmission module. The control deviation judgment module is used to acquire the response data fed back by the execution terminal, compare the response data with the predicted pollution level, and judge the control deviation of the closed-loop control. The environmental pollution closed-loop control module is used to adjust the parameters of the control command sequence in response to the control deviation, obtain an optimized control command sequence, and resend it through the remote transmission module to realize environmental pollution closed-loop control.

Images