Purification system based on multiple pollutants in waste gas and intelligent regulation and control method

Abstract

The invention relates to the technical field of waste gas purification, and discloses a multi-pollutant purification system based on waste gas and an intelligent regulation and control method. The method comprises a waste gas monitoring module, a pollution feature extraction module, a purification decision generation module, a trend deduction module, a decision compensation module and a purification execution module. The waste gas monitoring module obtains waste gas components and working condition parameter data flow in real time; the pollution feature extraction module analyzes the multi-pollution features and generates a map; the purification decision generation module outputs a primary scheme based on decision space and efficiency constraint in combination with a characteristic spectrum and working condition parameters; the trend deduction module predicts dynamic evolution of pollution characteristics and working condition parameters; the decision compensation module compensates the primary scheme according to the prediction result to generate an optimization scheme; and the purification execution module drives the purification device to operate according to the optimization scheme. According to the system, intelligent and precise treatment of waste gas purification is achieved, the complex and changeable waste gas purification requirements can be met, the purification effect is improved, and the operation cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of waste gas purification, and in particular to a waste gas multi-pollutant purification system and an intelligent control method. Background Art

[0002] Industrial production, energy conversion, and other processes generate waste gas containing a variety of pollutants. Directly discharging these gases without effective treatment can adversely impact the ecological environment and human health. Currently, a variety of technologies are available for waste gas purification, including adsorption, catalytic oxidation, and absorption. However, these methods have numerous limitations in practical application. Traditional exhaust gas purification systems often use fixed processing processes and parameter settings, making them difficult to adapt to dynamic changes in exhaust gas composition. For example, when the types of pollutants in the exhaust gas suddenly increase or the concentration rises sharply, the fixed purification strategy cannot be adjusted in time, resulting in a significant decrease in purification efficiency and difficulty meeting environmental emission standards. Most existing systems lack the ability to monitor and comprehensively analyze exhaust gas composition and operating parameters in real time. During the purification process, they are unable to accurately grasp the changing characteristics of pollutants and fluctuations in equipment operating conditions. This results in a lack of scientific basis for purification decisions and can easily lead to over-purification or under-purification. Over-purification wastes energy and consumables, increasing operating costs; while under-purification fails to achieve the desired purification effect, posing environmental risks. Traditional systems typically lack the ability to predict pollution trends and make compensatory decisions. They can only passively purify waste gas based on current conditions, unable to predict pollutant evolution trends and adjust purification strategies in time to address future pollution conditions. This results in insufficient adaptability and foresight, making it difficult to meet complex and ever-changing waste gas purification needs. Summary of the Invention

[0003] The purpose of the present invention is to provide a waste gas multi-pollutant purification system and an intelligent control method to solve the problems raised in the above background technology.

[0004] To achieve the above objectives, the present invention provides a system for purifying multiple pollutants from waste gas, the method comprising: An exhaust gas monitoring module, which is used to obtain exhaust gas composition data stream and operating condition parameter data stream in real time; A pollution feature extraction module, which performs multi-pollution feature analysis based on preset pollutant threshold constraints and the exhaust gas component data stream to generate a pollution feature map; A purification decision generation module, which outputs a primary purification decision plan based on an embedded purification decision space and purification efficiency constraints, combined with the pollution characteristic map and the operating condition parameter data stream; A trend deduction module, which performs dynamic evolution prediction on the pollution characteristic map and the operating condition parameter data stream to obtain a predicted pollution characteristic map and a predicted operating condition parameter set; a decision compensation module, which performs strategic compensation on the primary purification decision plan based on the predicted pollution characteristic map and the predicted operating condition parameter set to generate an optimized purification decision plan; A purification execution module drives the purification device to operate based on the optimized purification decision plan.

[0005] Preferably, the pollution feature extraction module includes: Get the basic attribute set of exhaust emission sources; constructing a pollution diffusion model based on the basic attribute set; Inputting the exhaust gas composition data stream into the pollution diffusion model to generate a pollutant concentration distribution cloud map; The pollution characteristic map is analyzed based on the preset pollutant threshold constraint and the pollutant concentration distribution cloud map.

[0006] Preferably, the steps of constructing the pollution feature extraction module include: Set up clusters of similar exhaust emission sources; Loading the historical pollution feature record library corresponding to the same type of exhaust emission source cluster; Load the real-time pollution characteristic record library of the target exhaust emission source; Constructing a pollution characteristic benchmark model based on the historical pollution characteristic record library; The pollution feature extraction module is generated by performing feature analysis sensitivity optimization on the pollution feature benchmark model based on the real-time pollution feature record library.

[0007] Preferably, the performing feature analysis sensitivity optimization on the pollution feature benchmark model based on the real-time pollution feature record library includes: Using the real-time pollution feature record library to test the pollution feature benchmark model, and obtain a feature analysis sensitivity coefficient; When the feature analysis sensitivity coefficient is lower than a preset sensitivity threshold, incremental training is performed on the pollution feature benchmark model based on the real-time pollution feature record library.

[0008] Preferably, the purification decision generation module includes: calibrating the purification decision space according to the pollution characteristic map and the operating condition parameter data stream to obtain a first calibrated decision space; identifying a decision feature triggering interval in the first calibration decision space to generate a first decision feature triggering domain; generating an initial purification decision based on the first decision feature triggering domain; Calculating the purification efficiency matching degree of the initial purification decision; When the purification efficiency matching degree satisfies the purification efficiency constraint condition, the initial purification decision is added to the primary purification decision scheme.

[0009] Preferably, calibrating the purification decision space according to the pollution characteristic map and the operating condition parameter data stream includes: Traversing the decision records in the purified decision space; Analyze the similarity depth between the current pollution feature map and the sample pollution features in the decision record; Analyze the similarity depth between the current working condition parameter data stream and the sample working condition parameters in the decision record; Weighted calculation of the correlation strength coefficient between the similar depth of the pollution characteristics and the similar depth of the operating condition parameters; When the correlation strength coefficient exceeds a preset correlation threshold, the corresponding decision record is included in the first calibration decision space.

[0010] Preferably, the decision compensation module includes: calibrating the purification decision space based on the predicted pollution characteristic map and the predicted operating condition parameter set to obtain a second calibrated decision space; identifying a decision feature triggering interval in the second calibration decision space to generate a second decision feature triggering domain; Iteratively generating a compensation purification decision under the purification efficiency constraint condition; The compensatory purification decision is strategically integrated with the primary purification decision scheme.

[0011] Preferably, the purification execution module includes: parsing a purification parameter instruction set in the optimized purification decision plan; Converting the purification parameter instruction set into a control signal sequence for a purification device; The purification medium dosage and reaction conditions are adjusted according to the control signal sequence.

[0012] Preferably, the system further includes a purification efficiency feedback module: Collect data on the composition of the purified exhaust gas; Compare the exhaust gas composition data after purification with the preset purification target value; A purification efficiency deviation index is generated and fed back to the pollution feature extraction module.

[0013] Preferably, the present invention also includes an intelligent control method based on an exhaust gas multi-pollutant purification system, the method comprising: Obtain exhaust gas composition data stream and operating parameter data stream in real time; Analyzing the exhaust gas composition data stream based on preset pollutant threshold constraints to generate a pollution characteristic map; Outputting a primary purification decision plan based on the embedded purification decision space and purification efficiency constraints, combining the pollution characteristic map and the operating condition parameter data stream; Deducing the dynamic change trend of the pollution characteristic map and the operating condition parameter data stream to obtain a predicted pollution characteristic map and a predicted operating condition parameter set; Performing strategic compensation on the primary purification decision plan based on the predicted pollution characteristic map and the predicted operating condition parameter set; Drive the purification device to execute the optimized purification decision plan; Collect the data of exhaust gas composition after purification and generate the purification efficiency deviation index; The pollution characteristic analysis rules are updated according to the purification efficiency deviation index.

[0014] Compared with the prior art, the present invention has the following beneficial effects: By setting up the exhaust gas monitoring module, the exhaust gas composition data stream and the operating parameter data stream can be obtained in real time, allowing the system to have a comprehensive and timely understanding of the current exhaust gas status and equipment operation status, providing basic information for subsequent pollution characteristic analysis and purification decision-making. The pollution feature extraction module is based on preset pollutant threshold constraints and combines the exhaust gas composition data stream to perform multi-pollution feature analysis and generate a pollution feature map. It can accurately identify the characteristics of various pollutants in the exhaust gas, including the type, relative content and distribution of pollutants, making the system's understanding of the pollution situation clearer and more specific, and avoiding vague judgments on pollution characteristics. The purification decision generation module is based on the embedded purification decision space and purification efficiency constraints, and combines the pollution characteristic map and operating parameter data stream to output the primary purification decision plan. The decision-making not only takes into account the current pollution characteristics, but also takes into account the operating conditions of the equipment, ensuring that the primary plan is theoretically feasible and targeted and can initially meet the purification needs. The trend deduction module dynamically predicts the evolution of pollution characteristic maps and operating parameter data streams to obtain predicted pollution characteristic maps and predicted operating parameter sets, enabling the system to transcend the limitations of the current state and grasp the changing trends of pollutants and possible fluctuations in equipment operating conditions in advance, providing a forward-looking basis for decision-making optimization and avoiding the lag that may be caused by making decisions based solely on the current state. The decision compensation module performs strategic compensation on the primary purification decision plan based on the predicted pollution characteristic map and the predicted operating condition parameter set, and generates an optimized purification decision plan. It can make up for the shortcomings of the primary plan in responding to future changes, making the final decision plan more perfect. It is not only suitable for the current pollution situation, but also can adapt to possible pollution trends in the future, enhancing the adaptability and flexibility of decision-making. The purification execution module drives the operation of the purification device based on the optimized purification decision-making plan, ensuring that the purification device can work according to the optimal strategy, making the purification process more accurate and efficient. It can not only effectively remove various pollutants in the exhaust gas, but also reasonably adjust the operating parameters according to actual conditions, reduce unnecessary energy and consumables consumption, and achieve good purification effects while helping to reduce operating costs and improve the overall performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a timing diagram of the exhaust gas multi-pollutant purification system according to the present invention; Figure 2 Flowchart of the pollution feature extraction module; Figure 3 Flowchart constructed for the pollution feature extraction module; Figure 4 Flowchart of the module for purification decision generation; Figure 5 Flowchart of the decision compensation module. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] See also Figure 1 The present invention provides a system for purifying multiple pollutants from waste gas, the method comprising: The exhaust gas monitoring module, pollution signature extraction module, purification decision generation module, trend deduction module, decision compensation module, and purification execution module operate in coordination. The exhaust gas monitoring module collects real-time exhaust gas composition data streams and operating parameter data streams via a sensor array. The pollution signature extraction module receives the exhaust gas composition data stream and, combined with preset pollutant threshold constraints, generates a pollution signature map by analyzing pollutant concentration distribution. The map includes pollutant types, concentration gradients, and spatial distribution characteristics. The purification decision generation module utilizes an embedded purification decision space (which stores a library of historical purification strategies) to match the pollution signature map with the operating parameter data stream, identifying the optimal decision and generating a preliminary purification decision plan. The plan covers purification process selection, reaction condition parameters, and resource allocation strategy. The trend deduction module uses a time series prediction algorithm to dynamically evolve the pollution signature map and operating parameter data stream, outputting a predicted pollution signature map and a set of predicted operating condition parameters. The decision compensation module compares the deviation between the predicted data and the preliminary plan, searches for compensation strategies in the purification decision space, and generates an optimized purification decision plan. The purification execution module converts the optimized plan into control instructions, driving the purification device to adjust the purification medium dosage and reaction conditions.

[0018] Example 1: See Figure 2 The operation process of the pollution feature extraction module begins with the acquisition of the basic attribute set of the exhaust emission source. The basic attribute set covers physical structure parameters, emission characteristic parameters, and environmental correlation parameters. Physical structure parameters include chimney height, exhaust pipe inner diameter, and pipe inclination; emission characteristic parameters involve the initial flow velocity of exhaust gas, emission temperature fluctuation range, and exhaust continuity coefficient; environmental correlation parameters refer to coupling parameters with the surrounding topography, such as building obstruction factor and surface roughness index. These parameters are retrieved in real time through the equipment archive and stored in the feature extraction buffer in the form of structured data tables.

[0019] The pollution diffusion model is constructed using multi-source data fusion technology. Based on a basic attribute set, the model's core utilizes the Gaussian plume diffusion equation framework, which includes submodules for the dynamic calculation of lateral and vertical diffusion coefficients. The lateral diffusion coefficient automatically matches the Pasquier stability level based on the atmospheric stability classification table, while the vertical diffusion coefficient is linked to the temperature stratification curve and inversion layer height data. During the model initialization phase, a geographic information system base map is loaded, and the emission source location coordinates are mapped to a three-dimensional grid coordinate system. The grid cell resolution is set to 10 meters x 10 meters x 5 meters (length x width x height) based on monitoring accuracy requirements. When the exhaust gas composition data stream is input into the model, independent transmission channels are created for each pollutant type, and each channel is embedded with species-specific diffusion correction factors, such as the gas-phase conversion attenuation factor for sulfur dioxide and the photochemical reaction correction value for nitrogen oxides.

[0020] The generation process of the pollutant concentration distribution cloud map includes spatial interpolation and dynamic rendering. The raw diffusion data output by the model is first subjected to Kriging spatial interpolation calculations to fill in the concentration values ​​in areas not covered by the sensors. The interpolation results are stored in a voxel matrix, with the matrix axes corresponding to the three-dimensional coordinates of longitude, latitude, and altitude. The cloud map rendering engine extracts the matrix data and maps the concentration values ​​to the RGBA color space: the lowest concentration range is rendered as transparent blue, the threshold boundary range transitions to yellow, and the concentration range exceeding the standard is marked in red. The rendering process overlays the real-time meteorological flow field map, using dynamic particle trajectories to demonstrate the impact of the dominant wind direction on the pollutant transmission path. The cloud map update frequency is synchronized with the monitoring data stream, and the default setting is to generate a frame of holographic projection view every minute.

[0021] The analysis operation of the pollution characteristic map is associated with a three-layer data architecture. The first-layer pollutant classification index table sorts the excessive substances according to the toxicity level and marks the chemical characteristic codes, such as marking PM2.5 as inhalable particulate matter and benzene series as carcinogens. The second-layer spatial distribution heat map is stored in a tile pyramid structure. The top layer shows the overall pollution outline of the region. Drilling down step by step can obtain the specific concentration value in the grid unit, and record the migration trajectory of the center of the pollution cluster. The third-layer time dimension cumulative curve records the concentration change trend of each pollutant, calculates the hourly concentration extreme value and the daily average fluctuation amplitude in a sliding time window manner, and automatically marks the abnormal fluctuation timestamp. When there are continuous exceeding areas in the cloud map, the map automatically activates the pollution source tracing function to reversely trace the main contributing sources of pollutants in the area.

[0022] The real-time correction mechanism is implemented through data assimilation technology. The micro-meteorological stations deployed in the emission area upload the measured values ​​of temperature, humidity, wind direction and speed every 30 seconds. These measured data are synchronously input into the deviation corrector with the model prediction values, and the Kalman filter algorithm is used to calculate the prediction error covariance matrix. The corrector outputs parameter adjustment instructions to the diffusion model, focusing on correcting the matching degree between the vertical turbulent diffusion coefficient and the actual atmospheric boundary layer height. Under calm wind conditions, the system automatically switches to the smoke puff integration mode and recalculates the residence time of pollutants in the near-ground layer. All correction operations are completed within the data stream processing cycle to ensure that the spatial matching deviation rate between the cloud map and the measured data in the next calculation cycle is lower than the set threshold.

[0023] The feature extraction module is equipped with a threshold adaptive adjustment mechanism. Preset pollutant threshold constraints are loaded in the form of a configuration file, which contains two sets of standard libraries: national standard limits and local special emission limits. When new pollutants are detected in the exhaust gas composition, the system searches the material safety database to obtain temporary threshold recommended values. In the case of continuous exceeding of the standard, the atlas generation system automatically increases the spatial resolution to a 1-meter grid accuracy to increase the sensitivity of identifying the edges of pollution clusters. The historical pollution feature library regularly performs morphological analysis on the atlas, identifies high-frequency pollution patterns and generates feature templates. When real-time data matches the template features, the early warning and accelerated response mechanism is triggered.

[0024] The output interface of the feature extraction module uses a multi-level buffering design. The raw cloud map data is compressed and stored in a distributed file system, allowing the trend deduction module to access the basic data set. The standard pollution feature map is converted into a lightweight JSON format and pushed to the decision-making center system in real time. It contains a list of pollutant types, a set of coordinates of the boundary of the exceeded area, and pollution intensity classification labels. A metadata description file is also generated, recording traceability information such as the map generation timestamp, correction operation records, and spatial accuracy level. During sudden changes in exhaust emission conditions, the system automatically activates the fast channel, bypassing the regular data verification process and directly transmitting key pollution indicators.

[0025] Example 2: See Figure 3 The construction process of the pollution feature extraction module is based on the dual-source data collaboration mechanism of the historical pollution feature record library and the real-time pollution feature record library. The historical pollution feature record library stores the long-term monitoring data of the same type of exhaust gas emission source clusters. The data collection period is not less than three years, covering the exhaust gas composition changes in different seasons and different operating conditions. Each record contains four core fields: timestamp, exhaust gas composition spectrum, operating parameter vector, and purification effect evaluation value. The exhaust gas composition spectrum is organized in the form of a multi-dimensional array. The array dimension corresponds to the type of pollutant, and the value of each dimension is the normalized concentration value. The operating parameter vector integrates continuous variables such as temperature, pressure, flow, humidity, and discrete variables such as equipment operating status and maintenance records. The purification effect evaluation value is quantitatively expressed in percentage, which is converted from the ratio of the emission concentration after purification to the standard limit.

[0026] The real-time pollution feature record library adopts a circular buffer structure design, and the buffer capacity is set to the most recent 72 hours of continuous monitoring data. The data writing process implements sliding window management, with the window size defaulting to 1 hour, and each sliding triggers a feature extraction operation. The buffer is equipped with an outlier filter layer. When the concentration of a pollutant exceeds three standard deviations of the historical maximum value, the data review process is automatically triggered. The review method includes sensor cross-validation and manual review and marking. The data that passes the review is attached with a confidence label and stored in the record library. The confidence labels are divided into three levels: high, medium, and low, which affect the sample weight distribution in subsequent model training.

[0027] The pollution characteristic benchmark model is constructed using an integrated learning framework. The model input layer receives the exhaust gas composition spectrum and operating condition parameter vectors from the historical record library and generates high-order feature combinations through feature engineering. The feature combination includes time-lagged variables (concentration values ​​in the previous hour), spatial correlation variables (concentration differences between adjacent monitoring points), and operating condition-derived variables (temperature-pressure interaction terms). The improved random forest algorithm is applied during the model training phase, and the node splitting criteria for each decision tree use the following custom indicators: : ; in: Indicates the total number of candidate features for the current node, For the The number of samples corresponding to each feature, is the total number of node samples, is the information gain increment after the feature split, Indicates the average detection delay of this feature in the historical records (unit: minute). is the latency penalty coefficient (default value 0.01). This formula ensures that the model prioritizes features with high information gain and timely data acquisition for splitting. The baseline model output layer generates a probability distribution for pollution feature classification, with class labels including eight preset types: regular pollution pattern, sudden pollution pattern, and complex pollution pattern.

[0028] The implementation process of feature parsing sensitivity optimization is divided into two stages: offline testing and online incremental training. In the offline testing stage, the real-time record library data is divided into training set, validation set and test set in a ratio of 7:2:1. The sensitivity coefficient is calculated during the testing process. , which is defined as the ratio of the model’s correct recognition rate of newly added pollution features to the historical benchmark recognition rate. , the system automatically starts the incremental training process. Incremental training uses a small batch gradient boosting method, extracting 500 latest records from the real-time record library as incremental samples each time, and adjusting the model parameters through the following strategies: Feature weight redistribution: Dynamically adjust the sampling probability of each feature in the random forest based on changes in feature importance in incremental samples. For newly emerging pollutant feature combinations, the initial sampling probability is set to 1.5 times the historical mean.

[0029] Tree pruning: Remove subtrees whose performance on incremental samples continues to deteriorate and grow new subtrees in their place. Classification boundary adjustment: Use fuzzy clustering algorithms to recalculate the center vectors of various pollution patterns and expand the classification boundaries to accommodate new pollution features.

[0030] Optimized models are managed through a version control mechanism. Each update retains snapshots of the previous three versions, enabling rapid rollback to a stable state. The model performance monitoring panel displays key metrics such as the sensitivity coefficient change curve, feature recognition confusion matrix, and incremental training time in real time. If the sensitivity coefficient improves by less than 1% after five consecutive incremental training sessions, the system automatically switches to full training mode, reloading all data from the historical record library for model reconstruction.

[0031] The interactive interface between real-time data and the model utilizes an event-driven architecture. When the exhaust gas composition data stream is input, a feature extraction request event is triggered. This event carries payload information such as the data time window, pollutant concentration matrix, and operating parameter snapshots. After monitoring this event, the feature extraction service performs the following sequence of operations: First, data integrity is verified; requests with more than 15% missing values ​​will return an error code. Complete data enters the preprocessing pipeline for normalization, scaling, time alignment, and outlier smoothing. The processed data is then fed into the baseline model to obtain preliminary classification results. When the model returns a low-confidence prediction (probability value <0.6), the incremental model is automatically called for a secondary prediction. Finally, a structured message is output that integrates the pollution feature label, confidence score, and feature importance ranking.

[0032] The historical data retrospective analysis module regularly performs pattern mining tasks. This module loads the entire historical record repository and applies a time series clustering algorithm to identify recurring patterns in the evolution of pollution signatures. Each pattern is abstracted as a state transition diagram, where nodes represent typical pollution states and edge weights represent the transition probabilities between states. When new patterns are discovered, feature extraction rule recommendations are automatically generated and, after manual review, embedded into the decision rule set of the baseline model. The retrospective analysis results are also used to optimize the data storage strategy of the real-time record repository. Raw data corresponding to high-frequency access patterns is cached in memory, while data from low-frequency patterns is transferred to cold storage.

[0033] The system maintenance subsystem is responsible for ensuring the continuous operation of the feature extraction module. A dual-machine hot standby architecture is deployed at the hardware level, and service switching is completed within 10 seconds in the event of a master node failure. At the software level, resource isolation containers are set up to limit the maximum memory usage of a single feature extraction task to no more than 4GB. The data pipeline implements flow control. When the input data rate exceeds the processing capacity, a degraded processing mode is activated, and only core pollutant feature extraction is performed. The log system records the detailed trajectory of each feature extraction operation, including audit information such as the input data hash value, model version number, and computing resource consumption. The log file retention period is six months.

[0034] Example 3: See Figure 4The purification decision generation module processes the pollution signature map and operating parameter data streams through a two-layer calibration mechanism. The first layer of calibration traverses the decision records in the purification decision space. Each record contains the sample pollution signature, the sample operating parameters, and the corresponding purification strategy. The similarity depth between the current pollution signature map and the sample is calculated using a dynamic time warping algorithm to match the temporal alignment of the concentration curves. The similarity depth of the operating parameter is quantified using Euclidean distance, and the weight coefficient is dynamically adjusted based on the parameter importance (for example, temperature weight is set to 0.6 and pressure weight is set to 0.4). The correlation strength coefficient is generated by normalizing the weighted similarity depth, with a threshold set to 0.7. The second layer of calibration determines the trigger intervals for the decision features in the first calibration decision space. These intervals are divided based on the peak pollutant concentration and the fluctuation range of the operating parameters. Initial purification decisions are generated using a genetic algorithm. Individuals in the population represent different purification process combinations. The fitness function is the purification efficiency matching degree, calculated as (actual purification rate / target purification rate) × 100%. When the matching degree is ≥90%, the strategy is included in the primary purification decision plan.

[0035] The operational process of the purification decision generation module is based on a two-tier calibration mechanism. The first tier of calibration focuses on the dynamic screening process of the purification decision space. The purification decision space is constructed as a graph database, storing a set of decision records formed by historical valid decision cases. Each decision record contains five dimensions of data: the sample contamination characteristic fingerprint (MD5 hash value of the contamination characteristic graph), the sample operating condition parameter vector (normalized parameter array), the purification strategy instruction set, the expected purification efficiency value, and the actual execution performance rating. When calibration is initiated, the system traverses the current decision record set and processes the current input data using a multi-dimensional similarity parallel computing architecture.

[0036] The dynamic time warping algorithm is introduced to calculate the similarity depth between the current pollution feature map and the sample pollution feature. The algorithm first converts the map data into a time series form: a snapshot of the pollutant concentration distribution at a fixed interval (the default is 5 minutes) is extracted along the time axis, and each snapshot is reduced to a 128-dimensional feature vector. The similarity depth calculation is implemented in three steps: the first step is to standardize the sequence length, and adjust the sequences of different lengths to the same length through cubic spline interpolation; the second step is to calculate the cosine similarity of each pair of feature vectors to form a similarity matrix; the third step is to find the optimal curved path on the matrix, and the cumulative similarity of the path is the final similarity depth value. The calculation formula is as follows: ; in: Indicates the similar depth of pollution characteristics, is the optimal bending path length, and Represent the length of the current sequence and the sample sequence respectively, is the path indicator function (1 when the path passes through point (i, j) and 0 otherwise), and Represent the feature vectors of the i-th frame of the current sequence and the j-th frame of the sample sequence respectively. This calculation method can overcome the comparison bias caused by the difference in the acquisition frequency of monitoring data.

[0037] The similarity depth between the current operating condition parameter data stream and the sample operating condition parameters uses a weighted Euclidean distance framework. Parameter weight assignment follows the parameter sensitivity rule: the temperature parameter weight is set to 0.6, the pressure parameter weight is set to 0.3, and the flow parameter weight is 0.1. The weight coefficient is dynamically adjusted based on the parameter's impact on purification efficiency, and the weight table is updated every 24 hours through the parameter importance assessment module. Before similarity depth calculation, continuous parameters must be Z-score normalized, and discrete parameters must be converted to one-hot encoding. The correlation strength coefficient is generated through dual similarity depth fusion: ; in: represents the correlation strength coefficient, and are the fusion weights of pollution characteristics and operating parameters respectively (default α=0.7, β=0.3), is the depth of similar pollution characteristics, Indicates the similarity depth of working condition parameters, is the distance scaling factor (default 0.5). When , the corresponding decision record is included in the first calibration decision space. This space is stored in a minimum heap structure and only the top 100 highly correlated records are retained in descending order of R value.

[0038] The second level of calibration establishes the identification rules for the decision feature trigger interval. The interval division is based on the distribution pattern of the pollution cluster exceeding the standard and the fluctuation characteristics of the operating parameters in the pollution feature map. The pollution cluster feature extractor scans the continuous exceeding standard area in the map and calculates three key indicators: the area of ​​the area, the extreme value of the core concentration, and the azimuth angle with the emission source. The operating parameter fluctuation analysis uses the window variance detection method to calculate the coefficient of variation of the parameter within a 30-minute time window. The trigger interval judgment condition is to meet the following conditions at the same time: the pollution cluster area is greater than 1.5 times the baseline value and the coefficient of variation of the operating parameter exceeds 0.25. The spatiotemporal domain that meets the conditions is marked as the first decision feature trigger domain, and the trigger domain is mapped to the decision space to form a three-dimensional decision coordinate (pollutant type axis, concentration range axis, operating condition fluctuation level axis).

[0039] The generation of initial purification decisions adopts a multi-objective genetic algorithm framework. The highly correlated decision records in the decision space are converted into the initial population, and each individual is represented by the purification process gene code. The code adopts a binary string structure, with the first 8 bits representing the catalytic oxidation process intensity (0000000011111111 corresponding to 0100% power), the middle 6 bits controlling the adsorbent injection rate (000000111111 corresponding to 0300kg / h), and the last 4 bits managing the pH value of the washing liquid (00001111 corresponding to 5.09.0). The population size is set to 200 individuals, and the upper limit of the evolutionary generation is 50 generations. The fitness function is defined as the purification efficiency matching degree: ; Where: represents the fitness of individual θ, For the expected purification efficiency, is the target purification rate, is the predicted energy consumption, is the maximum energy consumption allowed by the system, represents the equipment loss coefficient, Is the theoretical minimum loss value. Weight coefficient 、 、 These correspond to the priority levels of purification effectiveness, energy efficiency, and equipment lifespan (default values ​​are 0.8, 0.15, and 0.05, respectively). Genetic operations include two-point crossover (with a probability of 0.7) and site mutation (with a probability of 0.01), and the selection mechanism uses a tournament selection strategy.

[0040] The rule for determining whether the fitness evaluation reaches the standard is: when the optimal individual and When , the purification strategy corresponding to that individual is adopted. The decision generator decodes the binary gene into executable instructions: catalytic oxidation power = first 8 bits of the gene × 100 / 255 (%), adsorbent dosage = middle 6 bits × 300 / 63 (kg / h), washing liquid pH = last 4 bits × 4 / 15 + 5.0. A constraint checking module is also added to verify whether the instruction parameters exceed the safe operating boundaries of the equipment. The resulting primary purification decision plan contains four core data: strategy code, expected purification efficiency, resource consumption forecast, and equipment loss estimate, and is stored in the decision plan cache pool.

[0041] The decision-making solution cache pool uses a versioning management mechanism. Each solution is accompanied by a solution fingerprint, which is generated by combining the input feature hash and the decision parameters. When the similarity between the new input data and the historical decision scenario exceeds a threshold, the system directly calls the cached solution and executes a rapid correction process: using the current operating parameters to fine-tune the adsorbent dosage (±5% adjustment range) and calibrate the pH value of the washing liquid (±0.3 adjustment range). The cache pool implements an LRU replacement strategy, retaining the last 100 efficient decision solutions. The solution failure detector continuously monitors the actual execution performance of each solution. If the purification efficiency match is lower than 0.85 after three consecutive executions, the solution is automatically removed from the cache pool.

[0042] The module's output interface is designed as a bidirectional data channel. The primary channel transmits a structured, primary purification decision plan, including an executable set of process control parameters. The secondary channel outputs a decision analysis report, documenting an index list of associated decision records, the three-dimensional coordinates of the decision feature trigger domain, and a map of the genetic algorithm evolution trajectory. An exception handling mechanism is built into the solution generation terminal: if 50 generations of evolution fail to produce a qualified individual, the system switches to a conservative strategy mode, initiating pre-set purification plans based on contamination type, and simultaneously sending a decision failure alert to the operations and maintenance system. All output data is accompanied by a timestamp and decision sequence number, establishing a complete solution traceability chain.

[0043] Example 4: See Figure 5 The implementation process of the decision-making compensation module is based on a dynamic calibration mechanism between the predicted pollution characteristic map and the predicted operating parameter set. Taking the waste gas treatment system of a chemical plant as an example, the system detected an abnormal fluctuation in sulfur dioxide concentration at 8:00 a.m., and the trend deduction module predicted that a pollution peak would occur in the next two hours. At this time, the primary purification decision plan has been generated, including the operating parameters of the activated carbon adsorption device with a dosage of 200 kg / h and a catalytic oxidation temperature set at 350°C. After the decision-making compensation module is activated, it first loads the prediction dataset, which contains the predicted pollutant concentration values ​​and the operating parameter change curve in time series format.

[0044] The spatial grid marking method is used to analyze the predicted pollution characteristic map. The emission area is divided into 10m×10m grid cells, and each cell records the predicted concentration value of the future time slice (one slice every 15 minutes). The system identifies that the grid cells numbered G-17 to G-23 will form a continuous exceeding standard area at 9:30, and the core concentration will reach 2.3 times the standard limit. The predicted operating condition parameter set shows that the exhaust gas temperature will rise from 280℃ to 320℃ during the same period, and the fan speed needs to be increased by 15% to maintain the system pressure balance. When these data are input into the second calibration decision space, the spatial reconstruction process is triggered: the historical decision records with a similarity of more than 70% with the predicted concentration distribution pattern and a matching degree of more than 65% in the operating condition change trend are screened out to form a temporary decision library containing 32 decision records.

[0045] The generation of the second decision-making feature trigger domain relies on a coupled analysis of pollution diffusion dynamics and equipment response characteristics. The system creates a trigger domain determination table and updates the association between pollution characteristics and equipment parameters in each area in real time (see Table 1).

[0046] Table 1: Trigger domain determination table.

[0047] ;

[0048] Trigger priority is determined based on a weighted score combining the number of exceedances and the rate of parameter change. Areas with a score exceeding 80 are marked as "emergency." For area G-18, the system searched the temporary decision database and discovered three relevant historical decisions: Case A increased the activated carbon dosage to 240 kg / h and supplemented it with alkali spray; Case B maintained the original adsorption capacity but raised the catalyst temperature to 380°C; and Case C combined these two measures but reduced the fan speed by 10%. The decision compensation module initiated a multi-dimensional assessment, calculating the adaptability index of each case under the current forecast conditions.

[0049] The generation process of compensatory purification decisions implements phased iterative optimization. In the first phase, the activated carbon dosing parameters are adjusted: based on the predicted concentration gradient curve, the system makes tentative adjustments in the range of 200-250kg / h with a step size of 5kg / h. After each adjustment, the change in pollutant adsorption efficiency is simulated and calculated. When the dosage reaches 230kg / h, the simulation shows that the predicted residual concentration at 9:30 in the G-18 area can be reduced to 1.2 times the standard limit. In the second phase, the catalytic oxidation conditions are optimized: based on the predicted exhaust gas temperature rise curve, the optimal reaction temperature window is recalculated, and the original solution of 350℃ is adjusted to 365℃ to compensate for the impact of the temperature increase on the catalytic efficiency. In the third phase, the system parameters are balanced: it is detected that the increase in fan speed may cause the pressure drop of the washing tower to increase, and the bypass valve opening is adjusted to 45% to maintain the system pressure difference stable.

[0050] The strategy fusion phase uses a decision tree conflict resolution mechanism. When the base adsorption capacity (200kg / h) in the primary solution conflicts with the incremental demand (230kg / h) in the compensation solution, the system initiates a compromise algorithm: a transitional solution of 215kg / h is executed for the first 30 minutes, switching to 230kg / h when the predicted concentration reaches its peak; a linear temperature increase strategy is adopted for the catalytic temperature, increasing by 5°C every 10 minutes from the initial 350°C to the target temperature. The fused optimization solution generates an execution schedule, with control instruction switching accurate to the minute: 09:00-09:20: Activated carbon 215kg / h + catalyst 355℃ + fan speed 105%; 09:20-09:40: Activated carbon 230kg / h + catalyst 365℃ + fan speed 110%; after 09:40: return to benchmark parameters based on actual monitoring data.

[0051] The real-time adaptive monitoring system continuously collected key indicators during plan execution. Portable monitors deployed in area G-18 uploaded actual concentration data every five minutes, analyzing deviations from predicted values. When the actual concentration exceeded the predicted value by 15% at 9:28 a.m., the system immediately triggered dynamic compensation: the peak response plan was implemented two minutes in advance and a backup adsorption tower was activated. Simultaneously, temperature sensor feedback indicated that the actual heating rate of the catalytic bed was lower than expected, so the system automatically added 5kW of auxiliary power to the electric heater to maintain the temperature climbing curve.

[0052] The update mechanism of the historical decision database is activated after the compensation process is completed. After the actual execution effect data of this optimization plan is verified, a new decision record is generated and stored in the database. The record contains fields such as the predicted feature map fingerprint, the actual operating parameter matrix, the compensation strategy instruction set, and the final purification efficiency. The system specially marks the key operation nodes in this compensation: the inhibitory effect of the 230kg / h dosage on the peak concentration, the by-product control performance at a catalytic temperature of 365℃, and the coordinated adjustment parameters of the fan speed and the bypass valve. The difference analysis results between the new record and the original case are used to adjust the similarity calculation weight in the decision space and enhance the sensitivity to the recognition of sudden pollution characteristics.

[0053] The exception handling subsystem provides safety assurance for the compensation process. When predicted data indicates a potential excess of the equipment's design capacity (e.g., predicted concentration exceeding three times the limit), the system automatically enters emergency mode: an alert is immediately sent to the central control room and the pre-set, highest-level purification plan (maximum adsorption capacity plus full activation of backup equipment) is implemented. All compensation operations are subject to safety constraints, such as activated carbon dosage not exceeding 120% of the equipment's nameplate value and catalytic temperature not exceeding the material's tolerance limit. Before each parameter adjustment, the system cross-checks the available margin of the actuator to ensure that the operation is executed within safety boundaries.

[0054] Example 5: The operation of the purification execution module starts with the instruction parsing process of optimizing the purification decision plan. The system loads the plan file, which adopts a hierarchical coding structure: the top layer is the global execution strategy code, which identifies the priority ranking of the purification targets for this operation; the middle layer contains the equipment group control parameter package, each parameter package is associated with the identifier of a specific purification device; the bottom layer is the timing control instruction block, which defines the time nodes and transition curves for parameter adjustment. The parsing engine disassembles the file content layer by layer and maps the parameter instruction classification to the device controller address space. Taking the parameter package parsing of the activated carbon adsorption device as an example, the target dosage is 215kg / h, the dosage acceleration change gradient is 5kg / min, and the maximum allowable instantaneous error is ±3%. The parameter package of the catalytic oxidation unit contains a temperature setting value of 365℃, a heating rate limit of 15℃ / min, and a constant temperature maintenance period identifier.

[0055] The generation of control signal sequences adheres to industrial protocol conversion specifications. Each parameter instruction is first converted into a standard control message, the message header of which contains the device address code, instruction type identifier, and payload length. The dosage control signal uses a pulse frequency modulation mechanism: the target value of 215 kg / h is converted to a pulse frequency reference value of 1.2 kHz, which is related to the speed control motor drive current. The temperature control signal uses analog output mode, converting the 365°C target value into a 4-20 mA current loop signal, corresponding to the PLC output module channel address 0x3F5A. The timing instruction block drives the signal sequence generator, arranging control events according to the timeline set by the plan: a frequency ramp instruction is issued at 09:00:00 to increase the motor from 800 rpm to 950 rpm, and 09:20:00 triggers the second frequency jump to 1050 rpm. Each event is marked with a 5-second execution tolerance window.

[0056] The purification unit drive subsystem implements a distributed control architecture. The field controller network is divided into three levels: the first-level controller is the central command distributor, deployed in the purification system's main control cabinet; the second-level controller is the equipment cluster coordinator, with each purification unit equipped with an independent coordination node; and the third-level controller is the execution terminal, directly connected to the driver and sensor. The central command distributor splits the overall control signal sequence into device subsequences and sends them to the coordination node via industrial Ethernet. After receiving the command, the adsorption unit coordinator generates a speed control closed loop locally: the Hall sensor collects the motor speed in real time, and the data is refreshed every 200ms; the PID regulator compares the actual speed with the target value and outputs a PWM duty cycle correction signal to the drive circuit. The coordinator of the catalytic oxidation unit performs temperature interlock control: when the thermocouple detects that the actual bed temperature is 2°C lower than the set value, it automatically increases the thyristor conduction angle and activates the auxiliary heater.

[0057] The purification medium adjustment mechanism adopts a composite execution mechanism. The discharge mechanism of the activated carbon storage silo is equipped with a two-stage control: the coarse adjustment mode is performed by the variable frequency screw conveyor to control the speed, and the fine adjustment mode is adjusted by the pneumatic gate to adjust the discharge cross section. When the control signal is triggered, the coarse adjustment mode is first enabled to quickly approach the target dosage. When the actual dosage enters the range of ±5% of the target value, the fine adjustment mode is switched. The alkali liquid supply system of the spray tower implements dual pressure-flow regulation: the plunger pump receives the basic flow signal, and the differential pressure sensor monitors the atomization status of the nozzle; when the pipeline pressure exceeds 0.6MPa, the reflux valve is automatically activated to release the pressure to maintain the injection pressure in a stable range.

[0058] The purification efficiency feedback module deploys a multi-channel monitoring network. The system sets up three parallel sampling points in the exhaust main pipe of the purification device and configures different types of sensor arrays. Sensor array 1 is equipped with a non-dispersive infrared analyzer to continuously measure the concentration of SO2 and NOx; array 2 is equipped with a hydrogen flame ionization detector to monitor the total amount of VOCs; array 3 is equipped with a laser scattering particle counter to capture suspended particles in the PM2.5-PM10 range. All sensors are sampled synchronously through the data collector. The sampling period defaults to 60 seconds, and a 10-second high-speed sampling mode is started at the pollution peak. The monitoring data is verified in real time: when the readings of the three groups of sensors for the same pollutant differ by more than 15%, the calibration gas injection program is triggered to recalibrate the sensor.

[0059] The purification efficiency evaluation system establishes a standardized comparison process. The preset purification target value database stores the control standards for different pollutants, including instantaneous limit values ​​(such as SO2 peak concentration ≤ 150mg / m³) and moving average limit values ​​(such as 24-hour average PM10 ≤ 100μg / m³). After each sampling window is closed, the data processing unit automatically associates the timestamp of the current sample with the purification plan execution timeline to generate a performance evaluation table that matches the time period. The comparison algorithm uses direct threshold judgment for instantaneous values ​​and calculates the forward cumulative curve value for the moving average. The system generates a deviation index vector, and each dimension represents the extent to which a specific pollutant exceeds the standard within a specified time period: for example, the SO2 index is calculated as the difference between the measured maximum value and the limit value, and the particulate matter index uses the ratio of the 24-hour average value to the limit value.

[0060] Feedback closed-loop control enables dynamic parameter updates. The deviation indicator vector is pushed to the threshold constraint library of the pollution feature extraction module through the data bus. The update logic includes two modes: in normal mode, the threshold fine-tuning algorithm is executed. When the same pollutant exceeds the standard for three consecutive sampling periods, the corresponding threshold is automatically reduced by 5%; in emergency mode, threshold reconstruction is triggered. When the sudden increase in pollutant concentration exceeds 120% of the historical maximum value, a temporary threshold rule is created and the emergency analysis level is activated. The change record of the threshold constraint library generates an update notification file, which contains four metadata items: original threshold, new threshold, effective time, and change reason code. The file is synchronously distributed to the purification decision space management system to trigger the re-evaluation process of historical decision records.

[0061] The equipment status monitoring network performs a holographic record of the purification process. The system has a built-in equipment operation log memory, which records four types of operation events according to the timeline: the time and parameter details of the control instruction, the response time and action confirmation signal of the actuator, the monitoring value curve of the key sensor element, and the triggering record of the safety interlock device. The log storage implements a hierarchical compression strategy: the original time series record of routine operation data is retained for 30 days, and the abnormal event data is marked with a permanent storage flag. The maintenance diagnostic interface regularly scans the equipment status code. When the activated carbon saturation status flag is detected to be set, the adsorbent replacement recommendation is attached to the performance evaluation report; the continuous increase in the pressure loss of the catalytic bed triggers the catalyst activity detection process, and the result data is automatically associated with the equipment loss estimation model correction parameters.

[0062] A communication fault-tolerant mechanism ensures control link reliability. The main control signal channel uses a dual-network redundant architecture, automatically switching to the PROFIBUS backup channel when the Ethernet signal delay exceeds 500ms. A triple-check mechanism is implemented for data transmission: CRC cyclic redundancy check at the message level, data hash check at the application layer, and execution confirmation feedback at the transaction level. The purification execution module sets an instruction timeout monitor. Any control signal that does not receive a response confirmation from the device will be resent after 2 seconds. After three consecutive failures, the device is considered offline and the backup device takeover procedure is initiated. The secure communication protocol encrypts and signs key control instructions, and the key is rotated and updated every 24 hours to prevent the injection of illegal instructions.

[0063] The system sets up multiple layers of emergency control safety boundaries. The hardware layer is equipped with independent limit protection devices: when the temperature of the adsorption tower exceeds the safety threshold, the physical cooling system is directly triggered; when the pressure difference of the bag dust collector is too high, the compressed air pulse backblowing is automatically activated. The software layer sets up a parameter permission interval verification module to perform parameter range verification before the control signal is issued: when the activated carbon dosage instruction value exceeds the equipment rating by 90%-110%, it is automatically truncated to the allowable boundary value; when the catalytic temperature setting value is higher than the material tolerance temperature, the safety cooling coefficient is automatically injected into the calculation process. The execution effect emergency stop criterion is implanted at the end of the control chain. When the concentration of purified emissions exceeds the emergency limit for 15 minutes, the system is forced to switch to the preset highest intensity purification mode.

[0064] This implementation method builds a complete decision-making execution and feedback control chain. The purification execution module ensures the accurate transmission of control intent through fine-grained instruction parsing and protocol conversion; the distributed control system realizes the coordinated operation of the equipment cluster; the performance feedback system establishes a scientific evaluation index system; and the closed-loop update mechanism enables the continuous optimization of the pollution feature analysis rules. The design of all technical links focuses on traceability: every control instruction, every state change, and every threshold adjustment retains a complete operation trajectory and decision-making basis record. The equipment status monitoring and safety protection mechanism provides technical support for long-term continuous operation, and the communication fault-tolerant design enhances the applicability of industrial sites. There is no subjective evaluation content in the entire implementation process, and all descriptions are based on verifiable technical features and working principles.

[0065] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0066] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A system for purifying multiple pollutants from waste gas, characterized in that: The system comprises: An exhaust gas monitoring module, which is used to obtain exhaust gas composition data stream and operating condition parameter data stream in real time; A pollution feature extraction module, which performs multi-pollution feature analysis based on preset pollutant threshold constraints and the exhaust gas component data stream to generate a pollution feature map; A purification decision generation module, which outputs a primary purification decision plan based on an embedded purification decision space and purification efficiency constraints, combined with the pollution characteristic map and the operating condition parameter data stream; A trend deduction module, which performs dynamic evolution prediction on the pollution characteristic map and the operating condition parameter data stream to obtain a predicted pollution characteristic map and a predicted operating condition parameter set; a decision compensation module, which performs strategic compensation on the primary purification decision plan based on the predicted pollution characteristic map and the predicted operating condition parameter set to generate an optimized purification decision plan; A purification execution module drives the purification device to operate based on the optimized purification decision plan.

2. The system according to claim 1, wherein The pollution feature extraction module includes: Get the basic attribute set of exhaust emission sources; constructing a pollution diffusion model based on the basic attribute set; Inputting the exhaust gas composition data stream into the pollution diffusion model to generate a pollutant concentration distribution cloud map; The pollution characteristic map is analyzed based on the preset pollutant threshold constraint and the pollutant concentration distribution cloud map.

3. The system according to claim 1, wherein The steps of constructing the pollution feature extraction module include: Set up clusters of similar exhaust emission sources; Loading the historical pollution feature record library corresponding to the same type of exhaust emission source cluster; Load the real-time pollution characteristic record library of the target exhaust emission source; Constructing a pollution characteristic benchmark model based on the historical pollution characteristic record library; The pollution feature extraction module is generated by performing feature analysis sensitivity optimization on the pollution feature benchmark model based on the real-time pollution feature record library.

4. The system according to claim 3, wherein: The performing feature analysis sensitivity optimization on the pollution feature benchmark model based on the real-time pollution feature record library includes: Using the real-time pollution feature record library to test the pollution feature benchmark model, and obtain a feature analysis sensitivity coefficient; When the feature analysis sensitivity coefficient is lower than a preset sensitivity threshold, incremental training is performed on the pollution feature benchmark model based on the real-time pollution feature record library.

5. The system according to claim 1, wherein: The purification decision generation module includes: calibrating the purification decision space according to the pollution characteristic map and the operating condition parameter data stream to obtain a first calibrated decision space; identifying a decision feature triggering interval in the first calibration decision space to generate a first decision feature triggering domain; generating an initial purification decision based on the first decision feature triggering domain; Calculating the purification efficiency matching degree of the initial purification decision; When the purification efficiency matching degree satisfies the purification efficiency constraint condition, the initial purification decision is added to the primary purification decision scheme.

6. The system according to claim 5, wherein: The calibrating the purification decision space according to the pollution characteristic map and the operating condition parameter data stream includes: Traversing the decision records in the purified decision space; Analyze the similarity depth between the current pollution feature map and the sample pollution features in the decision record; Analyze the similarity depth between the current working condition parameter data stream and the sample working condition parameters in the decision record; Weighted calculation of the correlation strength coefficient between the similar depth of the pollution characteristics and the similar depth of the operating condition parameters; When the correlation strength coefficient exceeds a preset correlation threshold, the corresponding decision record is included in the first calibration decision space.

7. The system according to claim 1, wherein: The decision compensation module includes: calibrating the purification decision space based on the predicted pollution characteristic map and the predicted operating condition parameter set to obtain a second calibrated decision space; identifying a decision feature triggering interval in the second calibration decision space to generate a second decision feature triggering domain; iteratively generating a compensation purification decision under the purification efficiency constraint condition; The compensatory purification decision is strategically integrated with the primary purification decision scheme.

8. The system according to claim 1, wherein: The purification execution module includes: parsing a purification parameter instruction set in the optimized purification decision plan; Converting the purification parameter instruction set into a control signal sequence for a purification device; The purification medium dosage and reaction conditions are adjusted according to the control signal sequence.

9. The system according to claim 1, wherein: The system also includes a purification efficiency feedback module: Collect data on the composition of the purified exhaust gas; Compare the exhaust gas composition data after purification with the preset purification target value; A purification efficiency deviation index is generated and fed back to the pollution feature extraction module.

10. An intelligent control method based on an exhaust gas multi-pollutant purification system, characterized in that: The method comprises: Obtain exhaust gas composition data stream and operating parameter data stream in real time; Analyzing the exhaust gas composition data stream based on preset pollutant threshold constraints to generate a pollution characteristic map; Outputting a primary purification decision plan based on the embedded purification decision space and purification efficiency constraints, combining the pollution characteristic map and the operating condition parameter data stream; Deducing the dynamic change trend of the pollution characteristic map and the operating condition parameter data stream to obtain a predicted pollution characteristic map and a predicted operating condition parameter set; Performing strategic compensation on the primary purification decision plan based on the predicted pollution characteristic map and the predicted operating condition parameter set; Drive the purification device to execute the optimized purification decision plan; Collect the data of exhaust gas composition after purification and generate the purification efficiency deviation index; The pollution characteristic analysis rules are updated according to the purification efficiency deviation index.

Images