Dynamic cooperative regulation and control method and system for carbonaceous adsorption medium and malodorous gas neutralizer

Abstract

The invention relates to the technical field of waste gas purification, and discloses a dynamic cooperative regulation and control method and system for a carbonaceous adsorption medium and a malodorous gas neutralizer. The method comprises the following steps: collecting system inlet multi-gas time sequence concentration data and environmental parameters, and constructing a pollutant response phase lattice in combination with an activated carbon adsorption rate model and a chemical deodorant reaction kinetic equation; real-time working conditions are mapped to phase lattices to construct a directed state transition diagram, a dynamic cooperative regulation strategy tree is generated through community discovery and key path extraction, and an execution mechanism is driven to regulate the wind resistance of an activated carbon channel and the release rate of a deodorant. The problems of cooperative purification mismatch and peculiar smell rebound caused by inconsistent response dynamics of the activated carbon and the deodorant under the dynamic disturbance working condition are solved, the low ebb of purification efficiency is avoided, the method is suitable for high-cleanliness places such as medical treatment and laboratories, and the user experience is improved.

Description

Technical Field

[0001] This invention relates to the field of waste gas purification technology, specifically to a dynamic synergistic control method and system for carbonaceous adsorption media and odor neutralizer. Background Technology

[0002] In air purification systems, a combination of activated carbon and chemical deodorizers is commonly used to treat air pollutants. Activated carbon primarily removes volatile organic compounds through physical adsorption, while chemical deodorizers rely on specific chemical reaction mechanisms to decompose or neutralize odorous gases such as hydrogen sulfide and ammonia. This combination approach expands the purification spectrum and enhances overall removal capacity; the ratio is typically set empirically or adjusted based on sensor feedback under steady-state operating conditions.

[0003] However, in actual operation, when the types and concentrations of pollutants in the environment change rapidly, the system may experience a brief but clearly perceptible odor rebound. A typical scenario is a sudden increase in kitchen fumes followed by rapid dissipation. At this time, activated carbon temporarily occupies its microporous structure due to preferential adsorption of large organic molecules, resulting in a significant decrease in the adsorption rate of small odorous molecules. Simultaneously, chemical deodorizers, having not reached the critical humidity or pollutant concentration threshold required for their effective reaction, maintain a low reaction rate and cannot compensate for the instantaneous decrease in the adsorption capacity of activated carbon. This phenomenon is not caused by the exhaustion of filter material lifespan or sensor malfunction, but rather stems from a phased mismatch in the synergistic purification capabilities caused by the inconsistent response kinetics of the two purification materials under dynamic disturbance conditions. Existing control strategies are mostly based on feedback regulation using static ratios or single gas sensor signals (such as TVOC), without establishing a dynamic coupling model of the competitive relationship between adsorption and reaction pathways during transient processes. Therefore, it is impossible to predict and avoid this phased trough in purification efficiency. Although such short-term odor leaks are limited in duration, the human nose has an extremely low threshold for perceiving certain odor substances. Even a brief exceedance can cause significant discomfort, leading users to often misinterpret it as equipment malfunction or filter failure. In engineering practice, traditional designs focus on steady-state removal efficiency, filter material lifespan, or energy consumption, generally neglecting the high sensitivity of subjective odor perception to instantaneous concentration peaks. This has resulted in the problem not being recognized as a technical defect requiring proactive control for a long time, and it may even pose a risk of cross-contamination in medical, laboratory, or high-cleanliness environments. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, the present invention provides a method for dynamic synergistic regulation of carbonaceous adsorption media and odor gas neutralizers, comprising:

[0005] Step S1: Collect time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system, and construct a pollutant response phase lattice by combining the activated carbon adsorption rate model and the chemical deodorizing agent reaction kinetic equation.

[0006] Step S2: Map the real-time operating data to the pollutant response grid, connect the grid cells traversed in chronological order, and construct a directed state transition diagram;

[0007] Step S3: Perform community detection and critical path extraction on the directed state transition graph to generate a dynamic collaborative regulation strategy tree;

[0008] Step S4: Based on the ratio instructions output by the dynamic collaborative control strategy tree, drive the actuator to jointly adjust the air resistance of the activated carbon channel and the release rate of the deodorizing agent.

[0009] Furthermore, the steps for constructing a pollutant response lattice include:

[0010] Step S11: Collect time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system to obtain synchronized time-series concentration data and environmental parameters from multiple gas sensors.

[0011] Step S12: Based on the time-series concentration data and environmental parameters collected from multiple gas sensors, combine the instantaneous data collected synchronously into a set of instantaneous operating condition data, and then map each set of instantaneous operating condition data into a multi-dimensional state point.

[0012] Step S13: Based on the adsorption hysteresis characteristics of activated carbon for organic matter of different molecular weights and the nonlinear reaction threshold of chemical deodorizers for malodorous gases, the multidimensional state points are clustered by time sliding window to obtain multiple clusters.

[0013] Step S14: Divide the space into grid cells in a four-dimensional space consisting of pollutant type, concentration gradient, humidity range, and reactivity. Record the theoretical response delay difference, synergistic efficiency index, and local kinetic mismatch risk level for each grid cell. The boundaries of the grid cells are adjusted according to the cluster distribution.

[0014] Step S15: Organize all grid cells into a structured pollutant response phase grid according to the coordinate relationship in four-dimensional space, and store the pollutant response phase grid in the form of a four-dimensional tensor.

[0015] Furthermore, the step of mapping each set of instantaneous operating condition data to a multi-dimensional state point includes:

[0016] Step S121: Determine the dimensional composition of the multidimensional state point, which consists of the target pollutant concentration dimension, temperature and humidity dimension, airflow velocity dimension, and reactivity dimension.

[0017] Step S122: Calculate the reaction rate using the chemical deodorizer reaction kinetic equation, normalize the reaction rate, and obtain the value of the reaction activity dimension.

[0018] Step S123: Extract the values ​​of target pollutant concentration, temperature and humidity, and airflow velocity from each set of instantaneous operating data, and combine them with the value of reactivity to integrate the values ​​of the four dimensions into a multi-dimensional state point.

[0019] Furthermore, the steps for clustering multidimensional state points using a time-sliding window include:

[0020] Step S131: Calculate the working condition similarity between each multidimensional state point and other state points within the current time sliding window. The working condition similarity is calculated by weighted Euclidean distance.

[0021] Step S132: Set a dynamic density threshold. The dynamic density threshold is set as an adjustment coefficient multiple of the median of the distances from all multidimensional state points in the current window to their preset nearest neighbors. The adjustment coefficient is between 0 and 1.

[0022] Step S133: Divide the state points that meet the similarity threshold and density threshold into the same cluster, and each cluster represents a set of instantaneous working conditions with similar characteristics.

[0023] Furthermore, the steps for constructing a directed state transition graph include:

[0024] Step S21: Collect real-time operating data to obtain real-time multi-gas sensor time-series concentration data and environmental parameters;

[0025] Step S22: Map the real-time operating data to the pollutant response phase grid and determine the mapping unit corresponding to the real-time operating condition;

[0026] Step S23: Based on the chronological order, establish directed edges between consecutively hit mapping units to form a directed state transition graph;

[0027] Step S24: When the trajectory corresponding to the real-time operating condition enters a high-risk area, the outgoing edge weight of the relevant nodes is enhanced to highlight potential odor rebound paths; the potential odor rebound path refers to the operating condition evolution path that starts from the current high-risk node and will lead to subsequent odor rebound, and its outgoing edge weight is enhanced to be identified first.

[0028] Step S25: Maintain and update the directed state transition graph.

[0029] Furthermore, the steps for mapping real-time operating data to pollutant response grids include:

[0030] Step S221: Normalize the time-series concentration data and environmental parameters collected in real time from multiple gas sensors to ensure that their value range is consistent with the division range of each dimension of the four-dimensional space of the pollutant response phase.

[0031] Step S222: Calculate the spatial distance between the multi-dimensional state point corresponding to the real-time operating condition and the center of all grid cells in the pollutant response grid. The spatial distance is calculated using Manhattan distance.

[0032] Step S223: Determine the grid cell with the smallest distance as the mapping cell of the real-time operating data in the pollutant response phase grid; if there are multiple grid cells with equal distance, select the topologically adjacent grid cell as the final mapping cell according to the position of the mapping cell at the previous moment of the real-time operating condition.

[0033] Furthermore, the steps for generating a dynamic collaborative regulation strategy tree include:

[0034] Step S31: Using a modularity optimization algorithm suitable for directed graphs, the directed state transition graph is divided into communities to obtain multiple communities;

[0035] Step S32: Within each community, the K-shortest path algorithm that preserves edge directionality is used to extract critical transfer paths from low-risk starting nodes to high-risk ending nodes, thus obtaining a set of critical transfer paths within each community.

[0036] Step S33: Bind a set of pre-validated material ratio adjustment rules to each critical transfer path. The material ratio adjustment rules are a set of operation instructions formulated to maximize the synergistic effect of the two purification materials based on the working condition evolution characteristics corresponding to the critical transfer path. Each critical transfer path is an element in the critical transfer path set.

[0037] Step S34: Organize the material ratio adjustment rules into a dynamic collaborative control strategy tree according to the community-path hierarchy.

[0038] Furthermore, the steps for community partitioning of a directed state transition graph using a modularity optimization algorithm suitable for directed graphs include:

[0039] Step S311: Define a directed graph modularity index. The directed graph modularity index is calculated based on the directed transition probability between nodes. The modularity is the sum of the actual transition probabilities of directed edges within the community minus the sum of the expected transition probabilities under random distribution.

[0040] Step S312: Initialize community partitioning, treat each node as an independent community, and calculate the initial modularity value;

[0041] Step S313: Iteratively merge communities. Each time, select the two communities with the greatest improvement in modularity after merging and merge them. During the merging process, keep the direction information of the directed edges intact and record the directed transition relationship between the merged communities.

[0042] Step S314: Set termination conditions. When the increase in modularity during the iteration process is less than a preset threshold, the merging will stop. Each community corresponds to a typical disturbance condition mode.

[0043] Furthermore, the step of jointly adjusting the air resistance of the activated carbon channel and the release rate of the deodorizing agent by driving the actuator includes:

[0044] Step S41: Query the dynamic collaborative control strategy tree generated in step S3 in real time, and match the community to which the current working condition trajectory belongs and the nearest critical path;

[0045] Step S42: Based on the matched nearest critical path, obtain the corresponding control instructions from the leaf nodes of the dynamic collaborative control strategy tree. The control instructions are a set of combined instructions including activated carbon channel wind resistance adjustment parameters and deodorant release rate adjustment parameters.

[0046] Step S43: Drive the actuator to synchronously execute the control command, and the adjustment process follows the pre-set coordination constraints in the strategy tree;

[0047] Step S44: After execution, record the closed-loop feedback data to provide a basis for online optimization of the dynamic collaborative control strategy tree;

[0048] Step S45: Optimize the dynamic collaborative control strategy tree online based on closed-loop feedback data to ensure that the strategy adapts to changes in system operation.

[0049] A dynamic synergistic regulation system for carbonaceous adsorption media and odor neutralizing agents is provided to achieve the aforementioned dynamic synergistic regulation method for carbonaceous adsorption media and odor neutralizing agents. The system comprises:

[0050] Pollutant response phase grid construction module: used to collect time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system, and combine them with the activated carbon adsorption rate model and the reaction kinetic equation of chemical deodorizer to construct the pollutant response phase grid;

[0051] Directed state transition diagram generation module: used to map real-time operating data to pollutant response phase grids, connect the traversed grid cells in chronological order, and construct a directed state transition diagram;

[0052] Dynamic Cooperative Regulation Strategy Tree Generation Module: Used to perform community detection and critical path extraction on the directed state transition graph, and generate a dynamic cooperative regulation strategy tree;

[0053] The execution adjustment and strategy optimization module is used to drive the actuator to jointly adjust the air resistance of the activated carbon channel and the release rate of the deodorizing agent according to the ratio instructions output by the dynamic collaborative control strategy tree, and optimize the strategy tree based on closed-loop feedback data.

[0054] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs a pollutant response phase lattice by combining the activated carbon adsorption rate model with the chemical deodorizing agent reaction kinetic equation, accurately characterizing the difference in response characteristics of the two purification materials under dynamic disturbance conditions. It solves the technical pain point that traditional static ratios cannot capture the transient adsorption-reaction coupling relationship, and breaks through the limitation of using a single steady-state index to characterize dynamic conditions. It can completely preserve the evolution trajectory of operating conditions under sudden changes in pollutant concentration and fluctuations in environmental parameters, providing accurate state basis for coordinated regulation. Based on community discovery and critical path extraction using directed state transition graphs, this invention achieves the classification and identification of typical disturbance patterns such as "sudden increase in kitchen fumes" and "reagent leakage," as well as the precise location of risk evolution trajectories. This avoids the problem of insufficient adaptability of traditional single regulation strategies to different disturbance scenarios. The dynamic collaborative control strategy tree establishes a precise mapping of "operating mode - evolution path - control command," which can drive the actuator to synchronously adjust the air resistance of the activated carbon channel and the release rate of the deodorizing agent. It actively compensates for the decrease in adsorption capacity caused by the temporary occupancy of activated carbon micropores, while ensuring that the deodorizing agent reaction reaches the critical condition. This fundamentally avoids the mismatch in collaborative purification caused by the inconsistency of the response kinetics of the two materials, completely solving the problem of short-term odor rebound and preventing users from misjudging equipment failure. In addition, when applied in high-cleanliness places such as medical and laboratory settings, this invention can effectively block the risk of cross-contamination caused by instantaneous odor leakage, taking into account both purification efficiency and the user's subjective odor experience, filling the technical gap of traditional designs that neglect the sensitivity of instantaneous concentration peak perception. This invention achieves precise prediction and proactive intervention of purification efficiency mismatch under dynamic disturbance conditions by dynamically and synergistically regulating the reaction rate of activated carbon adsorption and chemical deodorizing agents during gas separation and purification. This effectively avoids the risk of odor rebound and improves the stability and reliability of the waste gas treatment system. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This is a flowchart of a dynamic synergistic regulation method for carbonaceous adsorption media and odor gas neutralizer in this invention;

[0057] Figure 2 This is a schematic diagram of community division according to the present invention;

[0058] Figure 3 This is a schematic diagram of the dynamic collaborative control strategy tree of the present invention;

[0059] Figure 4This is a functional block diagram of a dynamic synergistic regulation system for carbonaceous adsorption medium and odor gas neutralizer in this invention. Detailed Implementation

[0060] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] Example 1:

[0062] Please see Figure 1 As shown, this embodiment provides a method for dynamic synergistic regulation of carbonaceous adsorption media and odor gas neutralizers, including:

[0063] Step S1: Collect time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system, and construct a pollutant response phase lattice by combining the activated carbon adsorption rate model and the chemical deodorizing agent reaction kinetic equation.

[0064] This step focuses on the dynamic operating condition characterization requirements of the gas purification system. First, time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system are collected. Then, combined with the activated carbon adsorption rate model and the reaction kinetic equation of the chemical deodorizer, pollutant response phase grids are constructed through data processing, clustering, and grid partitioning, providing a state characterization basis for subsequent operating condition mapping and control strategy formulation.

[0065] The time-series concentration data from the multi-gas sensors represents the continuous time-series concentration changes of various target pollutants at the inlet of the gas purification system. This data is synchronously acquired through a high-response-rate gas sensor array deployed at the inlet. Target pollutants include volatile organic compounds (VOCs), hydrogen sulfide, and ammonia. VOC concentration data is acquired using sensors based on photoionization detection principles, while hydrogen sulfide and ammonia concentration data are acquired using sensors based on electrochemical principles. The time interval for time-series acquisition is determined by the average response period of historical disturbance conditions, with values ​​ranging in the millisecond range to ensure the capture of instantaneous changes in pollutant concentrations.

[0066] The environmental parameters are key external conditions affecting the effectiveness of the two purification materials, including temperature, humidity and airflow velocity. Temperature and humidity data are collected by an integrated temperature and humidity sensor, and airflow velocity data is collected by a thermal anemometer built into the pipe. The collection frequency is consistent with that of the multi-gas sensor to achieve data synchronization.

[0067] The activated carbon adsorption rate model is a mathematical model describing the variation of the adsorption rate of activated carbon for organic compounds of different molecular weights under different operating conditions. The core input variables include the molecular weight and concentration of the pollutants, the airflow velocity, and the temperature. The output variable is the amount of adsorption per unit time, i.e., the adsorption rate. This model adopts the well-established Langmuir adsorption rate equation, a classic model widely used in adsorption kinetics and proven to be suitable for characterizing the adsorption rate of single-component or multi-component organic compounds on activated carbon. The core expression of the model is: r = k a ·C·(1-θ)-k_d·θ, where each character has the following meaning: r is the adsorption rate (unit: mg / (g·s)), k a The adsorption rate constant (unit: L / (mg·s)) is determined by the characteristics of the activated carbon material and the type of target pollutant, and is obtained by fitting static adsorption experimental data; C is the pollutant concentration (unit: mg / L); θ is the coverage of adsorption sites on the activated carbon surface (dimensionless), θ=q / q m q represents the actual adsorption amount (unit: mg / g). m The saturated adsorption capacity of activated carbon (unit: mg / g) is determined by the pore structure of the activated carbon material itself; k_d is the desorption rate constant (unit: s). -1 The adsorption capacity (k) is related to the interaction forces between pollutant molecules and the activated carbon surface, and was determined through adsorption-desorption cycle experiments. For the adsorption differences of organic compounds with different molecular weights, k was experimentally calibrated. a The values ​​of k_d represent the characteristics of macromolecular organic compounds. a The value was determined experimentally to be less than that of small molecule organic compounds. a The values ​​naturally reflect the differences in adsorption rates. This method of characterizing differences in material properties through experimental calibration parameters is a common approach in this field for applying classical models.

[0068] The chemical deodorizer reaction kinetic equation is a quantitative equation characterizing the reaction rate between the chemical deodorizer and the odorous gas. Input variables include pollutant concentration, humidity, temperature, and the content of the effective ingredient in the deodorizer; the output variable is the reaction rate. This equation employs a mature bimolecular reaction kinetic equation, a classic kinetic model describing the reaction rate between two reactants. It has been widely used in chemical deodorization and gas purification, possessing a mature theoretical foundation and engineering application cases. The core expression of the model is: r'=k×Cg×Cd, where each character has the following meaning: r' is the reaction rate (unit: mol / (L·s)); k is the reaction rate constant (unit: L / (mol·s)), determined by the type of deodorizer, the type of odorous gas, temperature, and humidity, and calibrated experimentally under different humidity, concentration, and temperature conditions; Cg is the odorous gas concentration (unit: mol / L); and Cd is the content of the effective ingredient in the deodorizer (unit: mol / L). For the critical humidity or pollutant concentration conditions required for an effective reaction, the effective range of k is determined experimentally to characterize the reaction. That is, when the pollutant concentration or humidity is below the critical value, the experimentally calibrated value of k approaches 0, resulting in a low reaction rate. This method of reflecting the critical conditions of the reaction by limiting the effective range of parameters is a conventional approach in this field for applying classical kinetic models and is an engineering application of existing technology.

[0069] Specifically, the steps for constructing the pollutant response lattice are as follows:

[0070] Step S11: Collect time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system to obtain synchronized time-series concentration data and environmental parameters from multiple gas sensors.

[0071] Specifically, a high-response-rate gas sensor array deployed at the inlet collects time-series concentration data from multiple gas sensors, while an integrated temperature and humidity sensor and a thermal anemometer built into the pipeline collect environmental parameters, ensuring that the two types of data are collected at the same frequency and achieving synchronization.

[0072] Step S12: Based on the time-series concentration data and environmental parameters collected from multiple gas sensors, combine the instantaneous data collected synchronously into a set of instantaneous operating condition data, and then map each set of instantaneous operating condition data into a multi-dimensional state point.

[0073] Specifically, to achieve a multi-dimensional representation of the operating conditions, the process for constructing multi-dimensional state points is as follows:

[0074] Step S121: Determine the dimensional composition of the multidimensional state point, which consists of the target pollutant concentration dimension, temperature and humidity dimension, airflow velocity dimension, and reactivity dimension.

[0075] Step S122: Calculate the reaction rate using the chemical deodorizer reaction kinetic equation, normalize the reaction rate, and obtain the value of the reaction activity dimension.

[0076] Step S123: Extract the values ​​of target pollutant concentration, temperature and humidity, and airflow velocity from each set of instantaneous operating data. Combine the values ​​of the reactivity dimension with the values ​​of the four dimensions to form a multi-dimensional state point. The value range of each dimension corresponds to the effective range of the sensor measurement range or the calculation result.

[0077] Step S13: Based on the adsorption hysteresis characteristics of activated carbon for organic matter of different molecular weights and the nonlinear reaction threshold of chemical deodorizers for malodorous gases, the multidimensional state points are clustered by time sliding window to obtain multiple clusters.

[0078] The length of the time sliding window is determined based on the average duration of typical disturbance conditions in historical data, ensuring that the window includes the complete process of condition abrupt change and recovery. Specifically, the dynamic condition density clustering method proceeds as follows:

[0079] Step S131: Calculate the operating condition similarity between each multidimensional state point and other state points within the current time sliding window. The operating condition similarity is calculated by weighted Euclidean distance, and the weight coefficient is determined according to the degree of influence of each dimension on the purification efficiency. For example, the pollutant concentration dimension has a higher weight than the temperature dimension.

[0080] Step S132: Set a dynamic density threshold. This threshold is set as d times the median distance from all multidimensional state points in the current window to their k-th nearest neighbor, where k is the preset number of nearest neighbors and d∈(0,1) is the adjustment coefficient. The optimal value range is determined by verifying the historical clustering effect. This threshold is adaptively adjusted according to the distribution density of state points in the window. The denser the distribution, the lower the threshold, and vice versa.

[0081] Step S133: Divide the state points that meet the similarity threshold and density threshold into the same cluster, and each cluster represents a set of instantaneous working conditions with similar characteristics.

[0082] Step S14: Divide the space into grid cells in a four-dimensional space consisting of pollutant type, concentration gradient, humidity range and reactivity, and record the theoretical response delay difference, synergistic efficiency index and local dynamic mismatch risk level for each grid cell; the boundary of the grid cell is adjusted according to the cluster distribution.

[0083] Specifically, the process of mesh generation and parameter recording is as follows:

[0084] Step S141: Divide the four-dimensional space into intervals for each dimension. The pollutant type dimension is divided into several sub-intervals based on the chemical properties of the target pollutants, such as volatile organic compounds, sulfur-containing compounds, and nitrogen-containing compounds, and sorted by chemical property similarity (e.g., in ascending order of molecular polarity), considered as an ordered discrete dimension. The concentration gradient dimension is divided into several intervals based on multiples of the human nasal perception threshold. The humidity interval dimension is divided into three intervals: low humidity, medium humidity, and high humidity, based on the effective humidity range of the chemical deodorizer reaction. The boundary value of each interval is determined by the critical humidity parameter of the reaction kinetic equation. The reactivity dimension is divided into three intervals: low activity, medium activity, and high activity, based on the reaction rate level of the chemical deodorizer. The level classification is determined based on the ratio of the reaction rate to the maximum reaction rate. The boundaries of the grid cells are adjusted according to the cluster distribution to ensure that the state points within the same cluster fall within the same or adjacent grid cells, improving the grid cell's fit to the operating conditions.

[0085] Step S142: Calculate the theoretical response delay difference and synergistic efficiency index for each grid cell using the activated carbon adsorption rate model and the chemical deodorizer reaction kinetic equation. The theoretical response delay difference is the time difference between the time it takes for the activated carbon adsorption rate to reach a stable value and the time it takes for the chemical deodorizer reaction rate to reach a stable value. The synergistic efficiency index is the ratio of the purification efficiency when the two materials work together to (the sum of purification efficiencies when they work alone + ε) multiplied by the reaction activity dimension correction coefficient, where ε is the minimum regularization term (e.g., 1e). -6 The reaction activity dimension correction coefficient is positively correlated with the reaction activity value corresponding to the grid cell. The higher the reaction activity, the larger the correction coefficient, so as to reflect the enhancing effect of reaction activity on synergistic effect.

[0086] Step S143: The local dynamic mismatch risk level of each grid cell is obtained by weighted calculation of theoretical response delay difference and synergistic efficiency index, and divided into three levels: low risk, medium risk and high risk. The weights are determined according to the degree of influence of the two indicators on odor rebound.

[0087] Step S15: Organize all grid cells into a structured pollutant response phase grid according to the coordinate relationship in four-dimensional space, and store the pollutant response phase grid in the form of a four-dimensional tensor.

[0088] Each dimension of the four-dimensional tensor corresponds to an attribute in the four-dimensional space. The dimension corresponding to the pollutant type is organized in blocks, with each block corresponding to one type of pollutant. The block is divided into grids according to the other three dimensions. The tensor element is all the information contained in the grid cell. The four-dimensional tensor preserves the topological adjacency relationship of the state evolution path, which is defined as follows: if two tensor elements are adjacent in only one of the four dimensions and the other dimensions are completely consistent, they are determined to be topologically adjacent.

[0089] Traditional methods rely solely on a single steady-state index, failing to characterize the spatiotemporal coupling features of asynchronous responses between two types of materials during transient processes. This step constructs a pollutant response phase lattice through processes S11 to S15, transforming the dynamic mismatch problem into a problem of identifying local geometric structures in phase space. This provides a topologically faithful state representation foundation for subsequent precise intervention, which is unattainable with ordinary time series analysis or scalar thresholding. Under dynamic disturbance conditions, the pollutant response phase lattice can completely retain the state information of the entire process from steady state to disturbance and then to recovery, avoiding the omission of transient features by traditional methods. This allows subsequent control strategies to be based on the complete evolution trajectory of the operating conditions.

[0090] For example, when a disturbance occurs due to a sudden increase in kitchen fumes, step S11 collects the instantaneous peak value of volatile organic compound concentration and small fluctuations in airflow velocity; S12 maps the synchronous data into multi-dimensional state points; S13 assigns the state point to the corresponding cluster through dynamic operating condition density clustering; S14 determines the corresponding grid cell in four-dimensional space, and records that the theoretical response delay difference is large, the synergistic efficiency index is less than 1, and the local dynamic mismatch risk level is high; S15 incorporates the grid cell into the pollutant response phase grid in the form of a four-dimensional tensor, which maintains topological adjacency with the low-risk grid cell corresponding to the steady-state operating condition, clearly presenting the operating condition transition relationship.

[0091] Step S2: Map the real-time operating data to the pollutant response phase grid, connect the grid cells traversed in chronological order, and construct a directed state transition diagram.

[0092] This step focuses on the dynamic representation requirements of the operating condition evolution trajectory. First, real-time operating condition data is collected and mapped to the pollutant response phase grid generated in step S1. Then, by establishing directed edges, enhancing risk path weights, and maintaining the graph structure, a directed state transition graph is constructed, providing a foundation for subsequent community discovery and critical path extraction.

[0093] The real-time operating condition data and the time-series concentration data and environmental parameters collected by the multi-gas sensors in step S1 use the same acquisition method and index system to ensure data consistency and comparability. The core difference is that the real-time operating condition data focuses on dynamic acquisition and real-time processing at the current moment, while the data in step S1 is used to build the basic data support for pollutant response grids.

[0094] Specifically, the steps for constructing the directed state transition graph are as follows:

[0095] Step S21: Collect real-time operating data to obtain time-series concentration data and environmental parameters from real-time multi-gas sensors.

[0096] Using the same sensor array and acquisition method as in step S1, real-time target pollutant concentration data, temperature and humidity data, and airflow velocity data are collected synchronously to ensure that the acquisition frequency is consistent with that in step S1 and to maintain data synchronization.

[0097] Step S22: Map the real-time operating data to the pollutant response grid generated in step S1 to determine the mapping unit corresponding to the real-time operating conditions.

[0098] Specifically, to achieve real-time and accurate mapping, the mapping process is as follows:

[0099] Step S221: Normalize the time-series concentration data and environmental parameters collected in real time from multiple gas sensors to ensure that their value range is consistent with the division range of each dimension of the four-dimensional space of the pollutant response phase.

[0100] Step S222: Calculate the spatial distance between the multidimensional state point corresponding to the real-time operating condition and the center of all grid cells in the pollutant response grid. The spatial distance is calculated using Manhattan distance. Under the premise that the scales of each dimension are similar after normalization, the weighted Euclidean distance is approximated to reduce the computational complexity.

[0101] Step S223: Determine the grid cell with the smallest distance as the mapping cell for real-time operating data in the pollutant response phase grid; if there are multiple grid cells with equal distance, select the topologically adjacent grid cell as the final mapping cell based on the position of the mapping cell at the previous moment of the real-time operating condition to ensure the continuity of the mapping process.

[0102] Step S23: Based on the chronological order, establish directed edges between consecutively hit mapping units to form a directed state transition graph.

[0103] When establishing directed edges, topologically adjacent grid cells are prioritized to ensure that the directed state transition diagram accurately reflects the continuous evolution characteristics of the operating conditions and avoids discontinuous, skip-like edge connections. The structure of the directed state transition diagram includes two core parts: nodes and directed edges. Each node is a grid cell in the pollutant response phase grid, storing core information such as the theoretical response delay difference, synergistic efficiency index, and local kinetic mismatch risk level corresponding to that grid cell. The directed edges carry information on the transition time interval and the state change rate. The transition time interval is calculated by the difference in timestamps between two adjacent mappings, and the state change rate is the ratio of the change amplitude of the core indicators corresponding to two adjacent grid cells to the transition time interval. Core indicators include pollutant concentration, reaction rate, adsorption rate, etc.

[0104] Step S24: When the trajectory corresponding to the real-time operating condition enters a high-risk area, the outgoing edge weight of the relevant nodes is increased to highlight potential odor rebound paths; the potential odor rebound path refers to the operating condition evolution path that starts from the current high-risk node and may lead to subsequent odor rebound. After its outgoing edge weight is increased, it can be preferentially identified in subsequent path analysis.

[0105] Specifically, the outgoing edge weight enhancement method is as follows: the weight adjustment ratio is determined based on the local dynamic mismatch risk level of the nodes. The outgoing edge weight coefficient of high-risk nodes is increased by a preset ratio, the outgoing edge weight coefficient of medium-risk nodes remains unchanged, and the outgoing edge weight coefficient of low-risk nodes is appropriately reduced. The initial value of the weight coefficient is determined by the probability of different paths causing odor rebound in historical operating data. The path with a higher probability of causing odor rebound has a larger initial weight coefficient. The enhanced potential odor rebound path will serve as an important input for community discovery and critical path extraction in step S3. When calculating the weighted path length in step S32, because the outgoing edge weight coefficient of high-risk nodes is increased, the corresponding potential odor rebound path will be assigned a higher weight, making it easier to identify as a critical transfer path and ensuring that subsequent control strategies can specifically cover the risk evolution trajectory.

[0106] Step S25: Maintain and update the directed state transition graph to ensure that the graph structure reflects the latest state of the system operation.

[0107] Specifically, the maintenance and update process is as follows:

[0108] Step S251: Node information update. When the theoretical response delay difference, synergistic efficiency index, etc. corresponding to the grid cell are deviated due to filter material aging or long-term environmental changes, the node information is corrected by periodically collected calibration data.

[0109] Step S252: Update edge information. Based on the newly generated working condition transition data, adjust the statistical values ​​of the transition time interval and the statistical values ​​of the state change rate of the directed edges.

[0110] Existing control strategies lack the ability to model state transition processes, easily misjudging dynamic disturbances as isolated anomalies. This step, however, constructs a directed state transition graph, fully preserving the system's topology from steady state to a disturbed state and back. This allows subsequent analysis to be based on paths rather than single points, capturing dynamic behavioral characteristics that cannot be obtained using only state snapshots or statistical means. For example, when a system experiences brief fluctuations in pollutant concentration, isolated node analysis might classify it as an abnormal disturbance and ignore it. However, path analysis using the directed state transition graph reveals that the fluctuation originates from a specific steady-state path, and the state change rate of the directed edges matches the characteristics of a typical disturbance. This accurately identifies the dynamic condition requiring attention, providing a precise basis for subsequent control strategies.

[0111] For example, when a chemical reagent in the laboratory suddenly leaks, causing a sharp increase in the concentration of malodorous gas, step S21 collects real-time data on the sudden increase in pollutant concentration and synchronous environmental parameters; S22 normalizes the data and maps it to the pollutant response grid, initially mapping it to a low-risk node, and then transitioning to a medium-risk node and a high-risk node as the concentration increases; S23 establishes directed edges between continuous mapping units and records the time interval and concentration change rate of the transition process; S24 automatically enhances the outgoing edge weights of high-risk nodes; S25 continuously maintains the graph structure, and the final directed state transition graph clearly presents the complete risk evolution path from the steady-state path to the high-risk area.

[0112] Step S3: Perform community detection and critical path extraction on the directed state transition graph to generate a dynamic collaborative regulation strategy tree.

[0113] This step focuses on the need for precise formulation of dynamic control strategies. Taking the directed state transition graph generated in step S2 as the core input, it identifies typical disturbance operating conditions through community discovery, extracts key risk transfer paths, binds adjustment rules, and organizes them into a dynamic collaborative control strategy tree, providing targeted solutions for subsequent adjustments by the implementing agency.

[0114] Specifically, the steps for generating the dynamic collaborative regulation strategy tree are as follows:

[0115] Step S31: Using a modularity optimization algorithm suitable for directed graphs, the directed state transition graph generated in step S2 is divided into communities to obtain multiple communities, i.e., subgraph clusters.

[0116] The modularity optimization algorithm for directed graphs fully considers the directional characteristics and transition frequency of directed edges, avoiding the loss of key temporal information by traditional undirected graph community detection algorithms. Its specific process is as follows:

[0117] Step S311: Define the modularity index of the directed graph. This index is calculated based on the directed transition probability between nodes. The formula is: Modularity = (Sum of actual transition probabilities of directed edges within a community - Sum of expected transition probabilities under random distribution). The actual transition probability is calculated by the ratio of the number of transitions of directed edges to the total number of transitions, and the expected transition probability is derived from the out-degree and in-degree statistics of nodes. This index is the core quantitative standard for measuring the quality of community partitioning. It is used to judge whether "directed transitions between nodes within a community are denser and transitions between communities are sparser". The higher the index value, the more reasonable the community structure after partitioning is, and the better it reflects the clustering of similar disturbance conditions.

[0118] Step S312: Initialize community partitioning, treat each node as an independent community, and calculate the initial modularity value.

[0119] Step S313: Iteratively merge communities. Each time, select the two communities with the largest increase in module degree after merging and merge them. During the merging process, keep the direction information of the directed edges intact and record the directed transition relationship between the merged communities.

[0120] Step S314: Set termination conditions. When the increase in modularity during the iteration process is less than a preset threshold, the merging stops. Each community corresponds to a typical disturbance condition mode, such as kitchen oil fume disturbance mode, chemical reagent leakage disturbance mode, etc.

[0121] Step S32: Within each community, using the K-shortest path algorithm that preserves edge directionality, extract the critical transition paths from low-risk starting nodes to high-risk ending nodes, obtaining the set of critical transition paths within each community. Please refer to the community division diagram. Figure 2 .

[0122] The K-shortest path algorithm, which preserves edge directionality, strictly follows the direction constraints of directed edges to ensure that the extracted paths conform to the temporal order and causal relationship of the working conditions. Its specific process is as follows:

[0123] Step S321: Determine the set of low-risk starting nodes and the set of high-risk ending nodes in each community. The set of low-risk starting nodes consists of all nodes in the community with a low risk level of local dynamic mismatch, and the set of high-risk ending nodes consists of all nodes in the community with a high risk level of local dynamic mismatch.

[0124] Step S322: Taking each low-risk starting node as the starting point and the high-risk ending node as the ending point, calculate the weighted path length of all possible directed paths. The weighted path length is obtained by summing the weights of each directed edge on the path. The edge weight is defined as w(e) = -(α·ΔR + β·v_change), where α and β are weight coefficients (determined by the analytic hierarchy process, α > β), ΔR is the difference in local dynamic mismatch risk level between adjacent nodes, and v_change is the rate of state change. At the same time, the negative cycle problem caused by cyclic perturbation is avoided by limiting the maximum number of nodes on the path.

[0125] Step S323: Sort the paths corresponding to each start-end point pair in ascending order of weighted path length, and select the top K paths as candidate critical paths. The value of K is determined based on the total number of paths in the community and the actual control needs.

[0126] Step S324: Deduplicate and filter the candidate critical paths, remove redundant paths with path similarity higher than a preset threshold, and retain the core paths with significant differences to obtain the set of critical transfer paths.

[0127] Step S33: Bind a set of pre-validated material ratio adjustment rules to each critical transfer path. The material ratio adjustment rules are a set of specific operation instructions formulated to maximize the synergistic effect of the two purification materials in response to the working condition evolution characteristics corresponding to the path.

[0128] The pre-verification process of the regulation rules was completed through a combination of simulation experiments and actual working condition tests: First, simulations were performed based on the activated carbon adsorption rate model and the reaction kinetic equation of the chemical deodorizer to obtain preliminary regulation rules; then, the corresponding critical transfer path working conditions were simulated on the experimental platform to iteratively optimize the regulation rules; the core contents of the regulation rules include the adjustment range and timing of the activated carbon channel wind resistance, the adjustment ratio of the chemical deodorizer release rate and response delay, etc. These parameters are dynamically set according to the theoretical response delay difference and synergistic efficiency index of each node on the path.

[0129] Step S34: Organize the material ratio adjustment rules into a dynamic collaborative control strategy tree according to the community-path hierarchy.

[0130] The dynamic collaborative control strategy tree is essentially a hierarchical multi-branch decision tree in terms of data structure. Internal nodes represent operating condition categories, including typical disturbance operating condition modes corresponding to the top-level community and specific operating condition evolution trajectories corresponding to the critical transfer paths at the lower levels. Leaf nodes store specific wind resistance and deodorizing agent release combination instructions, with each leaf node uniquely corresponding to a critical transfer path. The strategy tree structure is scalable; when a new typical disturbance operating condition mode is added to the system, rules can be expanded by adding community nodes, corresponding path nodes, and leaf nodes without reconstructing the entire tree structure. Please refer to the schematic diagram of the dynamic collaborative control strategy tree. Figure 3 .

[0131] Conventional control methods struggle to differentiate response requirements for different types of disturbances, leading to lags or over-adjustments. This step, however, uses holistic community and path analysis of the directed state transition graph to structure complex dynamic mismatch scenarios into an indexable policy tree. This ensures precise matching between control actions and disturbance evolution paths. Furthermore, the generation of the policy tree relies entirely on the global topological properties of the graph, making it impossible to achieve the same effect by processing nodes individually. Community partitioning distinguishes different types of disturbance conditions, avoiding insufficient targeting due to a uniform control strategy. Critical path extraction accurately pinpoints the core trajectory of risk evolution, allowing control commands to act on key nodes in advance, effectively avoiding periods of low purification efficiency. The hierarchical structure of the dynamic collaborative control policy tree ensures rapid response to control decisions, directly indexing the corresponding control commands based on the community and path of the real-time condition, eliminating the need for complex real-time calculations.

[0132] For example, for communities experiencing kitchen fume disturbances, the critical path extracted in step S32 exhibits a rapid transition from "low-risk node → medium-risk node → high-risk node," with a high rate of state change. Step S33 assigns a regulation rule to this path: at the medium-risk node, the release rate of the chemical deodorizer is increased, and the air resistance of the activated carbon channel is appropriately increased to prolong the contact time. Step S34 organizes this rule, along with the community and path information, into a dynamic collaborative control strategy tree, forming corresponding leaf nodes. For communities experiencing chemical reagent leakage disturbances, the critical path exhibits a slow transition. The assigned regulation rule is to slightly increase the release rate of the deodorizer at the later stages of the low-risk node, and then adjust the air resistance at the medium-risk node to avoid energy waste caused by premature adjustment.

[0133] Step S4: Based on the ratio instructions output by the dynamic collaborative control strategy tree, drive the actuator to jointly adjust the air resistance of the activated carbon channel and the release rate of the chemical deodorizer.

[0134] This step focuses on the active collaborative control requirements of the gas purification system. It takes the dynamic collaborative control strategy tree generated in step S3 and the current operating trajectory data in the directed state transition diagram updated in real time in step S2 as the core inputs. Through path matching, instruction acquisition, execution adjustment, feedback recording and strategy optimization, it realizes the dynamic collaborative control of the two purification materials and avoids the trough of purification efficiency.

[0135] Specifically, the steps for joint regulation and strategy optimization are as follows:

[0136] Step S41: Query the dynamic collaborative control strategy tree generated in step S3 in real time, and match the current working condition trajectory to the community and the nearest critical path.

[0137] Specifically, the matching process is as follows:

[0138] Step S411: Community matching. Calculate the proportion of each community node in the current working condition trajectory, and determine the typical disturbance working condition mode corresponding to the current working condition, i.e., the matching community, according to the principle of maximum membership.

[0139] Step S412: Path matching. Within the matching community, calculate the similarity between the current working condition trajectory and each critical transfer path within the community. The path similarity is calculated by accumulating the differences in theoretical response delay, collaborative efficiency index, and state change rate of corresponding nodes on the trajectory. The critical transfer path with the highest similarity is the nearest critical path.

[0140] Step S413: If the similarity between the current trajectory and all key transition paths is lower than the preset threshold, it is determined to be a new working condition evolution path. The system temporarily stores the trajectory and triggers the online update process of the dynamic collaborative control strategy tree.

[0141] Step S42: Based on the matched nearest critical path, obtain the corresponding control instructions from the leaf nodes of the dynamic collaborative control strategy tree. The control instructions are a set of combined instructions including activated carbon channel wind resistance adjustment parameters and chemical deodorizer release rate adjustment parameters.

[0142] The activated carbon channel resistance adjustment parameters include a target resistance value and an adjustment rate. The target resistance value is determined based on the adsorption rate requirements of each node on the critical path, and the adjustment rate is set based on the rate of state change. The faster the state change, the higher the adjustment rate. The chemical deodorizer release rate adjustment parameters include a target release rate and a start-up delay time. The target release rate is calculated based on the reaction rate requirements on the critical path, and the start-up delay time is determined based on the theoretical response delay difference to ensure that the deodorizer's reaction timing is synchronized with the adsorption changes of the activated carbon.

[0143] Step S43: Drive the actuator to synchronously execute the control instructions, and the adjustment process strictly follows the pre-set coordination constraints in the strategy tree.

[0144] Specifically, the execution and constraint process is as follows:

[0145] Step S431: Activated carbon channel air resistance adjustment. The variable frequency fan changes its speed according to the air resistance adjustment parameters, and adjusts the airflow and velocity through the activated carbon layer to achieve precise control of the air resistance value.

[0146] Step S432: Adjustment of chemical deodorizing agent release rate. The micro metering pump changes the pump's operating frequency according to the release rate adjustment parameter to control the output of the deodorizing agent. At the same time, the deodorizing agent is atomized into tiny droplets through the atomizing nozzle to ensure full contact and reaction with the gas.

[0147] Step S433: Follow the synergistic constraints. The synergistic constraints are conditions set to avoid mutual interference between the two adjustment operations and to ensure maximum synergistic efficiency. For example, the increase in the deodorizer release rate must not exceed the upper limit of the compensation requirement corresponding to the current activated carbon adsorption rate; the time difference between wind resistance adjustment and deodorizer release rate adjustment must not exceed a preset threshold to ensure that the action conditions of the two materials remain matched.

[0148] Step S44: After execution, record the closed-loop feedback data to provide a basis for online optimization of the dynamic collaborative control strategy tree.

[0149] The closed-loop feedback data includes adjusted pollutant concentration data, actual adsorption rate, actual reaction rate, actual synergistic efficiency, user feedback information, and constraint satisfaction indicators: pollutant concentration data is collected by a gas sensor at the outlet; the actual adsorption rate is calculated based on the concentration difference of volatile organic compounds at the inlet and outlet and the airflow rate; the actual reaction rate is calculated based on the concentration difference of odorous gases at the inlet and outlet and the airflow rate; the actual synergistic efficiency is calculated as the ratio of the actual purification efficiency when the two materials work together to (the sum of the actual purification efficiencies when they work alone + ε), where ε is 1e -6 Level regularization terms; user feedback information is collected through the device's interactive interface, including subjective evaluations such as whether an odor is perceived; the constraint satisfaction index is calculated by judging whether the collaborative constraint conditions are met during the adjustment process, and the value is 0 or 1, where 1 indicates satisfaction and 0 indicates non-satisfaction.

[0150] Step S45: Optimize the dynamic collaborative control strategy tree online based on closed-loop feedback data to ensure that the strategy adapts to changes in system operation.

[0151] Specifically, the optimized process is as follows:

[0152] Step S451: Deviation calculation. The system calculates the deviation between each indicator in the feedback data and the target value stored in the strategy tree.

[0153] Step S452: Parameter adjustment. If the deviation exceeds the allowable range, adjust the control parameters of the corresponding leaf node according to the direction and magnitude of the deviation. If the user reports a perceived odor, increase the adjustment range of the control parameters of the corresponding leaf node by δ times (δ is an adjustment coefficient greater than 1, determined by calibration through historical optimization data). At the same time, adjust the collaborative constraint parameters according to the constraint satisfaction index. If the constraint satisfaction is consistently 0, appropriately relax the corresponding constraint threshold. If the constraint satisfaction is consistently 1, appropriately tighten the constraint to improve collaborative efficiency.

[0154] Step S453: Path and community optimization. If the deviation persists, re-optimize the corresponding critical path extraction results, and even adjust the community division boundaries.

[0155] Step S454: New path processing. For the new operating condition evolution path temporarily stored in step S413, when the accumulated sample size reaches a preset threshold and the path is verified to be repeatable and typical, it is included in the corresponding community, generating a new critical path and binding regulation rules, and extending it to the dynamic collaborative control strategy tree to achieve adaptive expansion of the strategy tree.

[0156] Traditional systems respond passively only after odor rebound occurs, failing to prevent instantaneous leakage below the human nasal perception threshold. This step, however, achieves proactive collaborative control based on dynamic path prediction by executing joint adjustment commands derived from the complex preceding structure. This effectively blocks the trough in purification efficiency caused by differences in material response kinetics. Through precise matching of the dynamic collaborative control strategy tree, the control commands can act in advance at key nodes in the evolution of operating conditions, preventing odor rebound. Synchronous adjustment of the actuators and strict adherence to collaborative constraints ensure synchronized responses from both purification materials, maximizing collaborative efficiency. Closed-loop feedback and online optimization enable the control strategy to continuously adapt to changes in system state, maintaining excellent control performance over the long term. In medical and laboratory settings with extremely high cleanliness requirements, this proactive collaborative control effectively avoids the risk of cross-contamination caused by instantaneous odor leakage, improving system reliability and applicability.

[0157] For example, after the system matches the critical path corresponding to the kitchen fume disturbance through the current trajectory, step S42 obtains the control instructions from the strategy tree: increase the activated carbon channel resistance by 20% from the current value, with an adjustment rate of 5% / ms; increase the chemical deodorizer release rate by 30%, with a start-up delay of 0ms. In step S43, the variable frequency fan immediately increases its speed at a rate of 5% / ms, and the micro-metering pump synchronously increases its operating frequency, strictly adhering to the cooperative constraints during the adjustment process; after the adjustment is completed, the sensor at the outlet shows that the pollutant concentration drops rapidly to below the sensing threshold, and the closed-loop feedback data shows that the actual value of the cooperative efficiency reaches 1.2, which meets the preset target; step S44 records the feedback data, and step S45 fine-tunes the parameters of the corresponding leaf nodes of the strategy tree based on the data to complete the optimization.

[0158] Example 2:

[0159] This embodiment, based on Embodiment 1, provides a dynamic synergistic control system for carbonaceous adsorption media and odor neutralizing agents, such as... Figure 4 As shown, it includes:

[0160] A dynamic synergistic regulation system for carbonaceous adsorption media and odor neutralizing agents is provided to achieve the aforementioned dynamic synergistic regulation method for carbonaceous adsorption media and odor neutralizing agents. The system comprises:

[0161] Pollutant response phase grid construction module: used to collect time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system, and combine them with the activated carbon adsorption rate model and the reaction kinetic equation of chemical deodorizer to construct the pollutant response phase grid;

[0162] Directed state transition diagram generation module: used to map real-time operating data to pollutant response phase grids, connect the traversed grid cells in chronological order, and construct a directed state transition diagram;

[0163] Dynamic Cooperative Regulation Strategy Tree Generation Module: Used to perform community detection and critical path extraction on the directed state transition graph, and generate a dynamic cooperative regulation strategy tree;

[0164] The execution adjustment and strategy optimization module is used to drive the actuator to jointly adjust the air resistance of the activated carbon channel and the release rate of the deodorizing agent according to the ratio instructions output by the dynamic collaborative control strategy tree, and optimize the strategy tree based on closed-loop feedback data.

Claims

1. A method for dynamic synergistic regulation of carbonaceous adsorption medium and odor neutralizing agent, characterized in that, The method includes: S1: Collect time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system, and construct a pollutant response phase grid by combining the activated carbon adsorption rate model and the chemical deodorizing agent reaction kinetic equation; S2: Map real-time operating data to pollutant response grids, connect the traversed grid cells in chronological order, and construct a directed state transition diagram; S3: Perform community detection and critical path extraction on the directed state transition graph to generate a dynamic collaborative regulation strategy tree; S4: Based on the ratio instructions output by the dynamic collaborative control strategy tree, drive the actuator to jointly adjust the air resistance of the activated carbon channel and the release rate of the chemical deodorizer.

2. The method for dynamic synergistic regulation of carbonaceous adsorption medium and odor neutralizer according to claim 1, characterized in that, The steps for constructing the pollutant response phase grid include: S11: Collect time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system to obtain synchronized time-series concentration data and environmental parameters from multiple gas sensors; S12: Based on the time-series concentration data and environmental parameters collected from multiple gas sensors, combine the instantaneous data collected synchronously into a set of instantaneous operating condition data, and then map each set of instantaneous operating condition data into a multi-dimensional state point. S13: Based on the adsorption hysteresis characteristics of activated carbon for organic matter of different molecular weights and the nonlinear reaction threshold of chemical deodorizers for malodorous gases, multidimensional state points are clustered by time sliding window to obtain multiple clusters. S14: Divide the space into grid cells in a four-dimensional space consisting of pollutant type, concentration gradient, humidity range, and reactivity. Record the theoretical response delay difference, synergistic efficiency index, and local kinetic mismatch risk level for each grid cell. The boundaries of the grid cells are adjusted according to the cluster distribution. S15: Organize all grid cells into a structured pollutant response phase grid according to the coordinate relationship in four-dimensional space, and store the pollutant response phase grid in the form of a four-dimensional tensor.

3. The method for dynamic synergistic regulation of carbonaceous adsorption medium and odor neutralizer according to claim 2, characterized in that, The step of mapping each set of instantaneous operating condition data to a multi-dimensional state point includes: S121: Determine the dimensional composition of the multidimensional state point, which consists of the target pollutant concentration dimension, temperature and humidity dimension, airflow velocity dimension, and reactivity dimension; S122: The reaction rate is calculated using the reaction kinetic equation of the chemical deodorizer, and the reaction rate is normalized to obtain the value of the reaction activity dimension. S123: Extract the values ​​of target pollutant concentration, temperature and humidity, and airflow velocity from each set of instantaneous operating data, and combine them with the value of reactivity to form a multi-dimensional state point.

4. The method for dynamic synergistic regulation of carbonaceous adsorption medium and odor neutralizer according to claim 2, characterized in that, The step of clustering multidimensional state points by time sliding window includes: S131: Calculate the working condition similarity between each multidimensional state point and other state points within the current time sliding window, wherein the working condition similarity is calculated by weighted Euclidean distance; S132: Set a dynamic density threshold, which is set as an adjustment factor multiple of the median of the distances from all multidimensional state points in the current window to their preset nearest neighbors. The adjustment factor is between 0 and 1. S133: State points that meet the similarity threshold and density threshold are divided into the same cluster, and each cluster represents a set of instantaneous working conditions with similar characteristics.

5. The method for dynamic synergistic regulation of carbonaceous adsorption medium and odor neutralizer according to claim 1, characterized in that, The steps for constructing the directed state transition graph include: S21: Collect real-time operating data to obtain real-time time-series concentration data and environmental parameters from multiple gas sensors; S22: Map the real-time operating data to the pollutant response phase grid to determine the mapping unit corresponding to the real-time operating condition; S23: Based on the temporal sequence, establish directed edges between consecutively hit mapping units to form a directed state transition graph; S24: When the trajectory corresponding to the real-time operating condition enters a high-risk area, the outgoing edge weight of the relevant nodes is enhanced to highlight potential odor rebound paths; the potential odor rebound path refers to the operating condition evolution path that starts from the current high-risk node and leads to subsequent odor rebound, and its outgoing edge weight is enhanced to be prioritized for identification. S25: Maintain and update the directed state transition graph.

6. The method for dynamic synergistic regulation of carbonaceous adsorption medium and odor neutralizer according to claim 5, characterized in that, The step of mapping real-time operating data to pollutant response grids includes: S221: Normalize the time-series concentration data and environmental parameters acquired in real time from multiple gas sensors to ensure that their value range is consistent with the division range of each dimension of the pollutant response phase four-dimensional space. S222: Calculate the spatial distance between the multi-dimensional state point corresponding to the real-time operating condition and the center of all grid cells in the pollutant response grid. The spatial distance is calculated using Manhattan distance. S223: The grid cell with the smallest distance is determined as the mapping cell of the real-time operating data in the pollutant response phase grid; if there are multiple grid cells with equal distance, the topologically adjacent grid cell is selected as the final mapping cell according to the position of the mapping cell at the previous moment of the real-time operating condition.

7. The method for dynamic synergistic regulation of carbonaceous adsorption medium and odor neutralizer according to claim 1, characterized in that, The steps for generating the dynamic collaborative regulation strategy tree include: S31: Using a modularity optimization algorithm suitable for directed graphs, the directed state transition graph is divided into communities to obtain multiple communities; S32: Within each community, using the K-shortest path algorithm that preserves edge directionality, extract the critical transfer path from the low-risk starting node to the high-risk ending node, and obtain the set of critical transfer paths within each community. S33: Bind a set of pre-validated material ratio adjustment rules to each critical transfer path. The material ratio adjustment rules are a set of operational instructions formulated to maximize the synergistic effect of the two purification materials based on the working condition evolution characteristics corresponding to the critical transfer path. Each critical transfer path is an element in the critical transfer path set. S34: Organize the material ratio adjustment rules into a dynamic collaborative control strategy tree according to the community-path hierarchy.

8. The method for dynamic synergistic regulation of carbonaceous adsorption medium and odor neutralizer according to claim 7, characterized in that, The steps of performing community partitioning on a directed state transition graph using a modularity optimization algorithm suitable for directed graphs include: S311: Define a directed graph modularity index, which is calculated based on the directed transition probability between nodes. The modularity is the sum of the actual transition probabilities of directed edges within the community minus the sum of the expected transition probabilities under random distribution. S312: Initialize community partitioning, treat each node as an independent community, and calculate the initial modularity value; S313: Iteratively merge communities. Each time, select the two communities with the greatest improvement in modularity after merging and merge them. During the merging process, keep the direction information of the directed edges intact and record the directed transition relationships between the merged communities. S314: Set termination conditions. When the increase in modularity during the iteration process is less than a preset threshold, the merging will stop. Each community corresponds to a typical disturbance condition mode.

9. The method for dynamic synergistic regulation of carbonaceous adsorption medium and odor neutralizer according to claim 1, characterized in that, The steps of the drive actuator in jointly adjusting the air resistance of the activated carbon channel and the release rate of the chemical deodorizer include: S41: Real-time query of the dynamic collaborative control strategy tree generated in the process, matching the current working condition trajectory to its community and the nearest critical path; S42: Based on the matched nearest critical path, obtain the corresponding control instructions from the leaf nodes of the dynamic collaborative control strategy tree. The control instructions are a set of combined instructions including activated carbon channel wind resistance adjustment parameters and chemical deodorizer release rate adjustment parameters. S43: Drive the actuator to execute control commands synchronously, and the adjustment process follows the pre-set coordination constraints in the strategy tree; S44: Record closed-loop feedback data after execution; S45: Optimize the dynamic collaborative control strategy tree online based on closed-loop feedback data to ensure that the strategy adapts to changes in system operation.

10. A dynamic synergistic control system for carbonaceous adsorption media and odor neutralizing agents, used to implement the dynamic synergistic control method for carbonaceous adsorption media and odor neutralizing agents as described in any one of claims 1-9, characterized in that, The system includes: Pollutant response phase grid construction module: used to collect time-series concentration data and environmental parameters from multiple gas sensors at the inlet of the gas purification system, and combine them with the activated carbon adsorption rate model and the reaction kinetic equation of chemical deodorizer to construct the pollutant response phase grid; Directed state transition diagram generation module: used to map real-time operating data to pollutant response phase grids, connect the traversed grid cells in chronological order, and construct a directed state transition diagram; Dynamic Cooperative Regulation Strategy Tree Generation Module: Used to perform community detection and critical path extraction on the directed state transition graph, and generate a dynamic cooperative regulation strategy tree; The execution adjustment and strategy optimization module is used to drive the actuator to jointly adjust the air resistance of the activated carbon channel and the release rate of the chemical deodorizer according to the ratio instructions output by the dynamic collaborative control strategy tree, and optimize the strategy tree based on closed-loop feedback data.

Images