Air purification system for atomic fluorescence

Abstract

The invention provides an air purification system for atomic fluorescence, which relates to the field of air purification, and is characterized in that an air quality data packet is acquired and matched with a current optimal excitation wavelength combination, and then the air quality data packet is input into a light source driving module to generate a customized multi-wavelength mixed light field into a light reaction cavity. Meanwhile, original air flow is input into the light reaction cavity to obtain processed air flow, and candidate strategies closest to or most relevant to the current air quality characteristics are quickly found in a strategy semantic set in a purification strategy semantic dynamic query mode. And finally, a purification strategy, namely an optimal excitation wavelength combination, which most meets the current requirement is finally determined in a dynamic evaluation mode, so that intelligent control over a light source is achieved, and it is ensured that the system can provide the optimal purification performance in various complex air environments.

Description

Technical Field

[0001] This application relates to the field of air purification, and more particularly, in the embodiments of this application, it relates to an air purification system for atomic fluorescence. Background Art

[0002] Air pollution has become a severe challenge faced by modern society. Toxic and harmful gases and particulate pollutants emitted by various industrial, transportation, and life activities pose a serious threat to the environment and human health. Traditional air purification technologies, such as filtration, adsorption, catalytic oxidation, etc., often face problems such as low efficiency, filter material saturation, secondary pollution, or high energy consumption when dealing with complex mixed pollutants or specific refractory substances. There is an urgent need to develop an innovative technology that can decompose or remove air pollutants more efficiently and precisely. Drawing on the highly specific interaction between light and matter in spectroscopic principles such as atomic fluorescence, by applying light energy of a specific wavelength to excite or decompose target pollutant molecules, a potential new approach to air purification is provided, promising precise targeting and efficient conversion of pollutants.

[0003] However, current existing air purification solutions, especially those technologies that attempt to decompose pollutants using the principle of light irradiation (such as some photocatalytic or UV oxidation technologies), often adopt fixed or limited spectral combinations and lack adaptability to the dynamic changes in the actual air composition. Systems that specifically use principles similar to atomic fluorescence to efficiently and directionally decompose pollutants are even less, which results in their inability to always maintain the optimal purification effect when dealing with different types and concentrations of pollutant mixtures. There may be problems such as energy waste, incomplete treatment, or the generation of adverse by-products, making it difficult to meet the requirements of complex and changing application scenarios.

[0004] Therefore, an optimized air purification system for atomic fluorescence is desired. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. The embodiments of this application provide an air purification system for atomic fluorescence. First, it obtains an air quality data packet and matches the current optimal excitation wavelength combination, and then inputs it into a light source driving module to generate a customized multi-wavelength mixed light field into a photoreaction chamber. At the same time, the original air flow is input into the photoreaction chamber to obtain a treated air flow, and the purification strategy semantic dynamic query method is used to quickly find the candidate strategy closest or most relevant to the current air quality characteristics in the strategy semantic set. Finally, the purification strategy that best meets the current requirements, that is, the optimal excitation wavelength combination, is finally determined through a dynamic evaluation method, so as to realize the intelligent control of the light source and ensure that the system can provide the optimal purification performance in various complex air environments.

[0006] According to one aspect of the present application, an air purification system for atomic fluorescence is provided, which includes:

[0007] An air quality data generation module for inputting the raw air flow into an integrated multi-sensor module to obtain an air quality data packet;

[0008] An excitation wavelength matching module for matching the current optimal excitation wavelength combination from a pollutant-wavelength-efficiency database based on the air quality data packet to obtain a light source control instruction;

[0009] A customized multi-wavelength light source generation module for inputting the light source control instruction into a light source driving module to generate a customized multi-wavelength mixed light field to an optical reaction chamber through the light source driving module;

[0010] An air flow processing module for inputting the raw air flow into the optical reaction chamber, wherein the raw air flow is decomposed under the action of the customized multi-wavelength mixed light field to obtain a processed air flow;

[0011] A purified air quality data packet generation module for inputting the processed air flow into an outlet sensor module to obtain a purified air quality data packet;

[0012] A purification skill report generation module for generating a purification skill report based on the comparison between the air quality data packet and the purified air quality data packet.

[0013] Compared with the prior art, an air purification system for atomic fluorescence provided by the present application first obtains an air quality data packet and matches the current optimal excitation wavelength combination, and then inputs it into a light source driving module to generate a customized multi-wavelength mixed light field to an optical reaction chamber. At the same time, the raw air flow is input into the optical reaction chamber to obtain a processed air flow, so as to quickly find the candidate strategy closest or most relevant to the current air quality characteristics in the strategy semantic set through the purification strategy semantic dynamic query method. Finally, the purification strategy that best meets the current requirements, that is, the optimal excitation wavelength combination, is finally determined through the dynamic evaluation method, so as to realize the intelligent control of the light source and ensure that the system can provide the optimal purification performance in various complex air environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0015] Figure 1It is a system block diagram of an air purification system for atomic fluorescence according to an embodiment of the present application.

[0016] Figure 2 It is a block diagram of an excitation wavelength matching module in an air purification system for atomic fluorescence according to an embodiment of the present application.

[0017] Figure 3 It is a schematic diagram of data flow of an excitation wavelength matching module in an air purification system for atomic fluorescence according to an embodiment of the present application.

[0018] Figure 4 It is a block diagram of a policy query response encoding unit in an air purification system for atomic fluorescence according to an embodiment of the present application. Detailed implementation manners

[0019] Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the drawings denote elements having the same or similar functions. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0020] The special term "exemplary" here means "serving as an example, embodiment or illustration". Any embodiment described as "exemplary" here is not necessarily to be construed as superior or better than other embodiments.

[0021] In addition, for a better illustration of the present application, numerous specific details are given in the following detailed implementation manners. Those skilled in the art should understand that the present application can also be implemented without some specific details. In some instances, methods, means, elements and circuits well known to those skilled in the art are not described in detail so as to highlight the gist of the present application.

[0022] Furthermore, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, "a plurality" means two or more unless otherwise specifically defined.

[0023] Air pollution has become a severe challenge faced by modern society. Harmful gases and particulate matters emitted from industries, transportation, and daily activities pose a serious threat to the environment and health. Traditional purification technologies such as filtration, adsorption, and catalytic oxidation often encounter problems such as low efficiency, filter media saturation, secondary pollution, and high energy consumption when dealing with complex or refractory pollutants. There is an urgent need to develop more efficient new purification technologies. Drawing on the principles of spectroscopy such as atomic fluorescence, using light energy of specific wavelengths to excite or decompose pollutant molecules provides a new idea for precisely and efficiently purifying air. However, most existing light-based purification technologies adopt fixed spectral combinations, making it difficult to adapt to the dynamic changes in the types and concentrations of pollutants in the air, lacking pertinence and adaptability, resulting in problems such as unstable purification effects, energy waste, and by-products, and it is difficult to meet the requirements of complex scenarios.

[0024] In view of the above technical problems, the present application proposes an air purification system for atomic fluorescence. Figure 1 FIG. is a system block diagram of an air purification system for atomic fluorescence according to an embodiment of the present application. As Figure 1 shown, an air purification system 100 for atomic fluorescence according to an embodiment of the present application includes: an air quality data generation module 110 for inputting an original air stream into an integrated multi-sensor module to obtain an air quality data packet; an excitation wavelength matching module 120 for, based on the air quality data packet, matching the current optimal excitation wavelength combination from a pollutant-wavelength-efficiency database to obtain a light source control instruction; a customized multi-wavelength light source generation module 130 for inputting the light source control instruction into a light source driving module to generate a customized multi-wavelength mixed light field through the light source driving module into a photoreaction chamber; an air stream processing module 140 for inputting the original air stream into the photoreaction chamber, wherein the original air stream is decomposed under the action of the customized multi-wavelength mixed light field to obtain a processed air stream; a purified air quality data packet generation module 150 for inputting the processed air stream into an outlet sensor module to obtain a purified air quality data packet; and a purification skill report generation module 160 for generating a purification skill report based on the comparison between the air quality data packet and the purified air quality data packet.

[0025] In the above air purification system 100 for atomic fluorescence, the air quality data generation module 110 is configured to input the raw air stream into the integrated multi-sensor module to obtain an air quality data packet. It should be understood that inputting the raw air stream into the integrated multi-sensor module provides basic data support for subsequent purification strategy selection and light source control. Considering the variety of pollutants in the air and the erratic concentration changes, an integrated multi-sensor module capable of comprehensively and accurately detecting this information is required. This module typically includes multiple different types of sensors, each sensor being specifically designed to monitor one or a class of specific air pollutants or air quality indicators, such as particulate matter (PM2.5, PM10), volatile organic compounds (VOCs), carbon monoxide (CO), carbon dioxide (CO2), sulfur dioxide (SO2), nitrogen oxides (NOx), etc. Through such a design, it is possible to ensure that as detailed pollution component information as possible is obtained from the raw air stream. When the raw air stream enters the integrated multi-sensor module, it undergoes a series of preprocessing steps to meet the working requirements of each sensor. For example, for some sensors sensitive to humidity and temperature changes, it may be necessary to first adjust the temperature and humidity of the air stream; while for those sensors measuring particulate matter concentration, special sampling methods may be required to ensure that particulate matter can be accurately captured in the detection area. At the same time, to ensure the accuracy and reliability of the data, a reasonable air flow channel and flow control mechanism need to be set inside the module, so that each incoming air stream can evenly contact all sensors and complete the detection process within the specified response time. During actual operation, each sensor reacts to the corresponding components in the air stream according to its working principle and generates an electrical signal. These electrical signals reflect the measured concentrations of various pollutants or air quality parameter values. However, due to the significant differences in the technical characteristics of different sensors, the output signal forms may also vary significantly. Therefore, further data processing steps are required to unify the format. This step includes, but is not limited to, signal amplification, filtering, calibration, and conversion into digital signals, etc., with the aim of eliminating noise interference, compensating for the influence of environmental factors, and converting the original physical quantity into a standardized data representation for subsequent analysis.

[0026] In particular, in the above air purification system for atomic fluorescence, since the types and concentrations of air pollutants are constantly changing, and different pollutants have different absorption and reaction characteristics for light of different wavelengths. Therefore, for the specific pollutant composition in the current air, the most suitable light wavelength combination is intelligently selected from the preset pollutant-wavelength-efficiency relationship database to maximize the purification efficiency, reduce energy consumption, and possibly reduce the generation of undesirable by-products, achieving customized and precise purification. Specifically, during the process of selecting the light wavelength combination, the semantic dynamic query method of the purification strategy is used to quickly find the candidate strategy that is closest or most relevant to the current air quality characteristics in the strategy semantic set, and through a dynamic evaluation method, the purification strategy that best meets the current requirements is finally determined, that is, the optimal excitation wavelength combination, so as to realize the intelligent control of the light source and ensure that the system can provide the optimal purification performance in various complex air environments.

[0027] Figure 2 It is a block diagram of an excitation wavelength matching module in an air purification system for atomic fluorescence according to an embodiment of the present application. Figure 3 It is a schematic diagram of data flow of an excitation wavelength matching module in an air purification system for atomic fluorescence according to an embodiment of the present application. As Figure 2 and Figure 3 shown, in the embodiment of the present application, the excitation wavelength matching module 120 includes: an air quality data packet structured encoding unit 121 for performing structured embedding encoding on the air quality data packet to obtain an air quality data packet structured embedding encoding vector; a preset purification strategy structured encoding unit 122 for performing structured embedding encoding on each preset purification strategy in the pollutant-wavelength-efficiency database to obtain a set of preset purification strategy structured embedding encoding vectors; a strategy query response encoding unit 123 for inputting the air quality data packet structured embedding encoding vector and the set of preset purification strategy structured embedding encoding vectors into a purification strategy semantic dynamic query network to obtain an air quality-purification strategy query response encoding vector; and an optimal excitation wavelength combination generation unit 124 for performing feature decoding on the air quality-purification strategy query response encoding vector to obtain the current optimal excitation wavelength combination.

[0028] Specifically, in the air quality data packet structured encoding unit 121 and the preset purification strategy structured encoding unit 122. It should be understood that since the original air quality data packet contains various information from multiple sensors, it may have diverse formats, high dimensions, and a complex internal structure; while the preset purification strategies in the pollutant-wavelength-efficiency database are also a structured information set. Directly processing and comparing these original structured data is neither efficient nor easy to accurately capture their deep semantic associations. In order to enable efficient and semantic matching and querying, these structured information must first be uniformly converted into standard, low-dimensional numerical vectors that can represent their features and semantics. Based on this, in the technical solution of this application, the air quality data packet is subjected to structured embedding encoding to obtain an air quality data packet structured embedding encoding vector, and each preset purification strategy in the pollutant-wavelength-efficiency database is subjected to structured embedding encoding to obtain a set of preset purification strategy structured embedding encoding vectors. In this way, the encoded structured embedding encoding vectors can capture the key features, interrelationships, and the "semantics" contained in the original data, that is, what kind of pollution situation the air quality data packet represents, or what kind of pollution a certain purification strategy can handle and what kind of effect it can bring.

[0029] Figure 4 The block diagram of the strategy query response encoding unit in an air purification system for atomic fluorescence according to an embodiment of the present application. As Figure 4As shown, in the embodiment of the present application, the policy query response encoding unit 123 includes: a preset purification policy structured embedding concentration subunit 1231, configured to perform information concentration on each preset purification policy structured embedding encoding vector in the set of preset purification policy structured embedding encoding vectors to obtain a set of preset purification policy structured embedding concentration encoding vectors; a semantic association degree construction subunit 1232, configured to construct a sparse association topology matrix between preset purification policy feature nodes based on the semantic association degree between any two preset purification policy structured embedding concentration encoding vectors in the set of preset purification policy structured embedding concentration encoding vectors; a preset purification policy feature graph spectrum encoding subunit 1233, configured to input the set of preset purification policy structured embedding concentration encoding vectors and the sparse association topology matrix between preset purification policy feature nodes into a graph spectrum construction engine based on a graph convolutional network model to obtain a preset purification policy feature graph spectrum encoding matrix; an information query subunit 1234, configured to perform information query in the preset purification policy feature graph spectrum encoding matrix based on the air quality data packet structured embedding encoding vector to obtain the air quality - purification policy query response encoding vector. It should be understood that although the step of structured embedding encoding has transformed air quality and each purification policy into vector representations respectively, these independent vectors have not fully captured the possible high - order, non - linear associations between air quality and different purification policies, as well as the internal structure and mutual influence of the preset policy set itself. The original vector embedding may only reflect point - to - point similarity, and to find the optimal policy, it is necessary to understand the "position" of air quality in the complex semantic space composed of all potential policies and its deep semantic association patterns with these policies. Based on this, in the technical solution of the present application, the air quality data packet structured embedding encoding vector and the set of preset purification policy structured embedding encoding vectors are further input into a purification policy semantic dynamic query network to obtain an air quality - purification policy query response encoding vector. In particular, through the purification policy semantic dynamic query network, the set of preset purification policy structured embedding encoding vectors can be regarded as a data set with a potential structure (manifold). By constructing and optimizing the graph structure (graph spectrum encoding), the semantic associations and global context information between policies are explicitly modeled and integrated into a structured and semantically rich representation. Then, taking the current air quality data packet structured embedding encoding vector as a query, efficient and semantic search and reasoning are performed in this encoded policy manifold space through the purification policy semantic dynamic query method.The core lies in dynamically evaluating the potential match, feasibility, or expected effect between the current air quality semantics and different purification strategy semantics, thereby generating a comprehensive air quality - purification strategy query response encoding vector. This vector precisely encodes the semantic relationship and potential response between the current air quality state and the optimal (or most relevant) purification strategy, providing an extremely optimized input basis for the subsequent feature decoding step. It enables the system to go beyond simple feature - dimension - based matching and truly achieve intelligent strategy selection based on semantics and potential associations, improving the ability to accurately and quickly find the current optimal excitation wavelength combination from complex and variable air quality, thus ensuring that the entire purification system always maintains efficient, accurate, and robust performance in a dynamic environment.

[0030] Specifically, the preset purification strategy structured embedding and concentration subunit 1231 is used to perform information concentration on each preset purification strategy structured embedding encoding vector in the set of preset purification strategy structured embedding encoding vectors to obtain a set of preset purification strategy structured embedding concentration encoding vectors, which is represented by the information concentration formula as follows:

[0031] S = {s1, s2,..., s i ,..., s n}

[0032] H = {h1, h2,..., h i ,..., h n}

[0033]

[0034] Among them, S is the set of preset purification strategy structured embedding encoding vectors, s1, s2, s i , s n are respectively the 1st, 2nd, i - th, and n - th preset purification strategy structured embedding encoding vectors in the set of preset purification strategy structured embedding encoding vectors, W c and b c are respectively the trainable weight matrix and the trainable bias vector, ReLu is the ReLu activation function, ||s i || is the L1 - norm of s i , H is the set of preset purification strategy structured embedding concentration encoding vectors, h1, h2, h i , h nThey are the 1st, 2nd, i-th, and n-th preset purification strategy structured embedded concentrated coding vectors in the set of preset purification strategy structured embedded concentrated coding vectors respectively. It should be understood that the core motivation for information concentration of the set of preset purification strategy structured embedded coding vectors lies in the significant high-dimensional sparsity and semantic noise interference of the original coding data. Since the preset strategy representations in the pollutant-wavelength-efficiency database usually cover multi-dimensional characteristic parameters, directly using the original high-dimensional vectors for semantic association analysis will lead to an exponential increase in computational complexity and is prone to falling into the curse of dimensionality. In addition, the original data collected by the sensor inevitably introduces redundant noise during the digitization process, such as signal offset caused by sensor temperature drift, multi-source data synchronization error, etc. These noises will obscure the essential associations between strategy features. By non-linearly projecting the high-dimensional embedded vectors into a low-dimensional semantic space, redundant dimensions that have no significant association with the purification efficiency can be filtered out through a feature selection mechanism, such as the saturation noise signals of some sensors in a specific pollutant concentration range. And, by using manifold learning techniques to reconstruct the internal association topology between strategy features, local semantic clusters scattered in the original space are condensed into core semantic units with high cohesion and low coupling. This processing method significantly improves the semantic modeling efficiency of the subsequent graph convolutional network, avoids the problem of over-smoothing of the graph structure caused by noise propagation, and enhances the generalization ability of the model to complex air pollution scenarios by retaining key discriminative features.

[0035] In an embodiment of the present application, the semantic association degree construction subunit 1232 includes: calculating the semantic association degree between any two preset purification strategy structured embedded concentrated coding vectors in the set of preset purification strategy structured embedded concentrated coding vectors to obtain an association topology matrix between preset purification strategy feature nodes; inputting the association topology matrix between preset purification strategy feature nodes into a gated mask network to obtain a sparse association topology matrix between preset purification strategy feature nodes.

[0036] Specifically, calculating the semantic association degree between any two preset purification strategy structured embedded concentrated coding vectors in the set of preset purification strategy structured embedded concentrated coding vectors to obtain an association topology matrix between preset purification strategy feature nodes, which is expressed by the semantic association degree calculation formula as:

[0037]

[0038] where, [·;·] is vector concatenation, W r and b r are the semantic association weight matrix and the semantic association bias vector respectively, r i,j is h i and h jThe preset purification strategy structured embedded concentrated semantic association vector, where L is the length of the preset purification strategy structured embedded concentrated semantic association vector, z is the eigenvalue position of the preset purification strategy structured embedded concentrated semantic association vector, and A i,j is the semantic association degree between h i and h j in the association topology matrix of preset purification strategy feature nodes. It should be understood that since the preset strategies in the pollutant-wavelength-efficiency database may still retain non-linear potential associations after structured embedding, directly using the original high-dimensional vectors for strategy retrieval will lead to distorted semantic distance calculation. By constructing the association topology matrix, the system can explicitly characterize the fine-grained semantic association patterns between strategy features. This association modeling essentially maps the discrete strategy vectors to a continuous graph manifold space, using the global connection characteristics of the graph structure to make up for the limitations of the simple vector space metric. The choice of the metric function needs to take into account the physical meaning and mathematical separability of the strategy features. For example, the cosine similarity is used to capture the directional association or the Mahalanobis distance is used to incorporate the feature covariance information. This design enables the edge weights of the graph spectrum to reflect both the instantaneous matching degree between strategies and the coupling strength of the long-term purification effect. The finally formed association topology matrix provides an input rich in semantic gradients for the subsequent graph convolutional network, enabling the model to dynamically aggregate neighborhood strategy features through the message passing mechanism, effectively improving the retrieval recall rate and decision-making robustness of the optimal excitation wavelength combination.

[0039] Specifically, input the association topology matrix between the preset purification strategy feature nodes into the gated mask network to obtain the sparse association topology matrix between the preset purification strategy feature nodes, which is represented by the gated mask formula:

[0040]

[0041] where W g is the mask weight vector, b g is the mask bias weight parameter, [h i ||h j is the association topology matrix between the preset purification strategy feature nodes, Sigmoid is the Sigmoid function, and M i,j is the semantic association degree between h i and h jThe mask semantic correlation degree between them. It should be understood that when the initial correlation topology matrix quantifies the semantic correlation between policy features, although all node pair correlation information is completely retained, a large number of weakly correlated or noisy edges among them will lead to redundant feature propagation paths during the graph convolution process, exacerbating the over-smoothing phenomenon and increasing the computational load. The gated mask network introduces a parameterized dynamic mask mechanism to non-linearly recalibrate the initial correlation degree in a data-driven manner. Its essence is to construct a learnable edge importance evaluation function. This network fuses the node feature vectors and the original correlation strength through a multi-layer perception structure, adaptively generates a mask probability between 0 and 1, making the redundant edge weights approach zero and the key edge weights significantly enhanced. This sparsification process is not a simple threshold truncation, but through joint training with the overall optimization goal of the system, enabling the mask mechanism to identify the key topological structures of the policy semantic manifold under different pollution scenarios. The finally formed sparse correlation topology matrix not only retains the integrity of the core semantic correlation but also reduces the computational complexity of graph convolution by reducing the propagation of invalid edges, enabling the model to significantly improve the real-time inference efficiency while maintaining a high recall rate, providing a structural guarantee for fast policy matching in a dynamic air environment.

[0042] Specifically, the preset purification policy feature spectrum encoding subunit 1233 is configured to input the set of preset purification policy structured embedded concentrated encoding vectors and the sparse correlation topology matrix between the preset purification policy feature nodes into a spectrum construction engine based on a graph convolution network model to obtain a preset purification policy feature spectrum encoding matrix, which is represented by the preset purification policy feature spectrum encoding formula as follows:

[0043] A i,j ’ = M i,j ·A i,j

[0044] Where, A i,j ’ is the encoding matrix of the preset purification policy feature spectrum regarding h i and h jThe semantic correlation degree of preset purification strategy features therebetween. It should be understood that the graph spectrum construction engine based on the graph convolutional network model is to realize high-order feature fusion through topological relationship modeling of the graph structure while retaining the core semantics of the strategy features. The graph convolutional network enables the representation of each strategy node to not only contain its own features but also dynamically fuse multi-level association information of adjacent nodes through an iterative neighborhood information aggregation mechanism. The sparse association topology matrix plays a role of structural guidance in this process. The key edges screened by the gated mask network define the information propagation path to ensure that the graph convolution operation is only performed between semantically related nodes. This design enables the final representation of the strategy node to capture its multi-hop neighborhood structure relationship in the feature manifold, such as the synergistic effect or temporal dependence characteristics of a certain wavelength combination on the degradation of composite pollutants. The graph spectrum encoding matrix integrates global context information to elevate a single strategy vector to a comprehensive representation containing spatial association and semantic hierarchy, thereby providing a strategy feature space with both local sensitivity and global consistency for subsequent retrieval of the optimal excitation wavelength combination, significantly improving the strategy matching accuracy and decision robustness in complex pollution scenarios.

[0045] In an embodiment of the present application, the information query sub-unit 1234 includes: inputting the structured embedding encoding vector of the air quality data packet and the preset purification strategy feature graph spectrum encoding matrix into a feature query response engine to obtain the air quality - purification strategy query response encoding vector, which is represented by the air quality - purification strategy query response encoding formula as:

[0046] αi = softmax(uTgi)

[0047] v r = Σ i α i (u ⊙ g i )

[0048] Wherein, u is the structured embedding encoding vector of the air quality data packet, g i is each row vector in the preset purification strategy feature graph spectrum encoding matrix, softmax is the softmax function, α i is the feature query factor, ⊙ is element-wise multiplication by position, and v rThe air quality-purification strategy query response encoding vector. It should be understood that due to the heterogeneity between the real-time features of the air quality data packet and the multi-dimensional semantics in the strategy map, relying solely on static encoding matching cannot adapt to the rapid changes in complex pollution scenarios. The feature query response engine realizes two-way dynamic interaction through the attention mechanism: on the one hand, the real-time feature vector of the air quality data packet is used as a query condition to locate key nodes in the global semantic space of the strategy map; on the other hand, the graph-like encoding matrix of the strategy map provides structured semantic routing for the query through the sparse association topology optimized by the gated mask network. This interaction mechanism enables the system to adaptively adjust the focus of the strategy map according to the current pollutant component characteristics, such as strengthening the weight of the photocatalyst-related wavelength strategy in the formaldehyde-exceeding scenario, and focusing on the association strength of the ultraviolet photolysis strategy when particulate matter is aggregated. The final generated air quality-purification strategy query response encoding vector not only integrates the dynamic characteristics of real-time monitoring data, but also integrates the global semantic information in the strategy map that has been propagated through multi-hop neighborhoods, forming a deep semantic representation of the current pollution scene, and providing the light source driving module with a wavelength combination decision-making basis that is both timely and accurate.

[0049] Preferably, the graph neural network has completed the extraction of high-order relationships and global structural information at the graph structure level for each preset purification strategy structured embedding encoding vector. However, the sparse processing of the graph topology by the gated mask network may cause each row vector g in the preset purification strategy feature imitation graph encoding matrix to be i At the macro level, the higher-order distribution outside the graph still needs to follow a unified statistical law, that is, to realize that each row vector g i The higher-order global correlation between them.

[0050] First, use the semicircular rate framework to calculate each row vector g i Rigid neighborhood projection of :

[0051]

[0052] g u For all row vectors g i The mean vector is obtained by taking the mean of the corresponding positions of .

[0053] Rigid neighborhood projection vector g' i It aims to explore the high-order statistical dependency characteristics between strategy features, break through the limitations of traditional covariance analysis, and capture the nonlinear and asymmetric deep correlation patterns between variables.

[0054] Then, calculate the rigid neighborhood projection vector g' i Dynamic aggregation measure of:

[0055]

[0056] where g' ij is the eigenvalue at the j-th position of the vector g' i .

[0057] Finally, using the dynamic aggregation measure r i to perform weighted optimization on the row vector g i to obtain g″ i = r i g i .

[0058] That is, in view of the fact that the high-order cumulant dominates the macroscopic distribution law, it is necessary to constrain the global symmetry through the dynamic cumulant measure under local rigid constraints to ensure the consistency of different row vectors g i at the high-order structure, thereby promoting the uniform propagation of statistics and avoiding irrelevant responses such as cascading failures in the graph structure.

[0059] After that, the preset purification strategy eigen-spectrum coding optimization matrix composed of the optimized row vectors is combined with the structured embedded coding vector input feature query response engine of the air quality data packet to improve the processing accuracy, and the air quality - purification strategy query response coding vector is obtained.

[0060] In an embodiment of the present application, the optimal excitation wavelength combination generation unit 124 is configured to: pass the air quality - purification strategy query response encoding vector through an optimal excitation wavelength combination analyzer based on a classifier to obtain an analysis result, and the analysis result is used to represent the label of the current optimal excitation wavelength combination. It should be understood that although the air quality - purification strategy query response encoding vector generated in the previous stage efficiently encodes the complex semantic relationship between the current air quality state and potential purification strategies and indicates the location of the optimal strategy in the vector space, it is itself an abstract numerical vector and cannot be directly used as an instruction to control a physical light source. In order to convert the result of this intelligent matching into an actual executable purification action, this high - dimensional, abstract vector representation must be "translated" or "decoded" back into specific parameters in the physical world - that is, the excitation wavelength combination of the light source (including specific wavelength values, intensities, or mixing ratios) or the label of the current optimal excitation wavelength combination. Therefore, in the technical solution of the present application, the air quality - purification strategy query response encoding vector is further decoded for features to obtain the current optimal excitation wavelength combination. That is to say, by receiving the air quality - purification strategy query response encoding vector output from the previous step as input and learning a mapping from this semantic vector space to the specific excitation wavelength parameter space. In particular, in a specific example of the present application, the air quality - purification strategy query response encoding vector is passed through an optimal excitation wavelength combination analyzer based on a classifier to obtain an analysis result, and the analysis result is used to represent the label of the current optimal excitation wavelength combination. In another specific example of the present application, the output of this decoding network directly corresponds to the specific parameters of the current optimal excitation wavelength combination, such as a vector or data structure containing multiple wavelength values and their corresponding intensities. This decoding module is optimized together with the encoding and query parts during the system training process, so as to learn how to most effectively convert the semanticized strategy encoding into actual light source control instructions, ensuring the accuracy and effectiveness of decoding.

[0061] In the above-mentioned air purification system 100 for atomic fluorescence, the customized multi-wavelength light source generation module 130 is configured to input the light source control instruction into the light source driving module to generate a customized multi-wavelength mixed light field into the photoreaction chamber through the light source driving module. It should be understood that inputting the light source control instruction into the light source driving module is the core step to ensure that the air purification system can dynamically adjust the purification strategy according to the current air quality characteristics. First of all, considering the wide variety of air pollutants and their different absorption and reaction characteristics to light of different wavelengths, in order to achieve efficient and precise decomposition or removal of target pollutant molecules, a flexible and adjustable light source scheme must be adopted. This light source scheme is no longer limited to traditional fixed wavelengths or limited wavelength combinations, but is the optimal excitation wavelength combination obtained through intelligent analysis based on real-time acquired air quality data packets. Therefore, the light source control instruction is essentially a data set containing information about which specific wavelengths and their intensities should be used, which provides the necessary parameter configuration basis for the light source driving module. After the light source control instruction is generated, the next step is to accurately transmit it to the light source driving module. This step requires the system to have a highly reliable communication mechanism to ensure that the instruction will not be lost or in error during transmission. Usually, this communication link needs to support high-rate data transmission and have a certain anti-interference ability to adapt to electromagnetic interference and other factors in complex industrial environments or household application scenarios. In addition, since the light source control instruction may involve relatively complex data structures (such as multiple wavelength values and their corresponding intensities), corresponding encoding and decoding protocols also need to be designed to facilitate accurate information exchange between the sending end and the receiving end. Once the light source control instruction successfully reaches the light source driving module, the next task is to regulate the actual light source device according to these instructions. The light source driving module, as the bridge connecting the control system and the physical light source, plays a crucial role. It not only needs to have powerful signal processing capabilities to parse the control instructions from the upper-level system and convert them into specific hardware operation commands; at the same time, it also needs to have sufficient power output capabilities to drive the light source components that can emit light within the required wavelength range. Specifically, after receiving the light source control instruction, the light source driving module will first parse the instruction content to identify the specified wavelength list and the energy level corresponding to each wavelength. Then, according to these parameters, the circuit design inside the driving module will make corresponding adjustments, such as adjusting the current magnitude, changing the voltage level or activating specific light-emitting elements, so that the entire light source system can generate a mixed light field composed of multiple wavelengths as expected. It should be noted that in order to achieve the ideal purification effect, this customized multi-wavelength mixed light field not only needs to meet the basic requirements regarding wavelength and intensity, but also needs to consider issues such as the spatial distribution uniformity and temporal stability of the light field.This means that in actual operation, the light source driving module also needs to combine optical principles and technical means, such as using special lens groups or mirror arrays to optimize the light propagation path, ensuring that every air molecule entering the photoreaction chamber can be fully irradiated. At the same time, to address the performance drift phenomenon that may occur during long-term operation, a self-monitoring and correction mechanism should be integrated within the light source driving module to regularly check the working state of the light source and make appropriate adjustments according to the actual situation, ensuring that the light field characteristics always meet the preset standards. Further, considering that different types of pollutants may exhibit different response characteristics to the same wavelength, and there may even be requirements for the optimal action time and cumulative dose at the same wavelength, when constructing a customized multi-wavelength mixed light field, the influence of these factors also needs to be comprehensively considered. This means that in addition to directly setting the initial wavelength and intensity according to the light source control instructions, the light source driving module may also need to introduce some advanced algorithms or models to simulate and predict the purification efficiency under different light field configurations and make fine-tuning accordingly to obtain the best overall purification efficiency. For example, by introducing a feedback loop mechanism, the purification process in the photoreaction chamber can be monitored in real time, and various parameters of the light source can be dynamically adjusted accordingly to achieve more refined control. Finally, after the above series of precise control measures, a customized multi-wavelength mixed light field is successfully generated and introduced into the photoreaction chamber. Here, this light field will interact with the original air flow passing through it to complete the purification process by exciting or decomposing pollutant molecules in the air.

[0062] In the above-mentioned air purification system 100 for atomic fluorescence, the air flow processing module 140 is used to input the raw air flow into the photoreaction chamber. Among them, the raw air flow is decomposed under the action of the customized multi-wavelength mixed light field to obtain the processed air flow. It should be understood that different types of air pollutants have different absorption and reaction characteristics for specific wavelengths of light. By precisely controlling the light source to generate a customized multi-wavelength mixed light field, these pollutant molecules can be targeted for excitation or decomposition. Before the raw air flow enters the photoreaction chamber, it must pass through a carefully designed air flow channel and flow control mechanism to ensure that the air can be evenly distributed throughout the reaction space, thereby maximizing the contact area and time with the light field. This means that in actual operation, not only how to effectively guide the air flow path needs to be considered, but also the flow rate and pressure conditions need to be optimized to maintain the stable operation of the system while ensuring the purification efficiency. Specifically, when the raw air flow is introduced into the photoreaction chamber, it will first encounter a pretreatment area where some preliminary adjustments may be made, such as temperature and humidity regulation, to adapt to the optimal environmental conditions for subsequent optical reactions. This pretreatment is crucial for improving the overall purification effect because many chemical reaction rates and photolysis processes highly depend on the environmental temperature and humidity. Subsequently, the air flow will be guided to the core area of the photoreaction chamber where the customized multi-wavelength mixed light field is already ready. To ensure that all incoming air can be fully exposed to the light field, special designs are usually adopted inside the reaction chamber to promote the uniform dispersion of air. For example, a series of flow deflectors, turbulators can be set up, or the tiny turbulence generated by a fan can be used to increase the interaction opportunities between air molecules and light. Once the raw air flow starts to contact the customized multi-wavelength mixed light field, a series of complex physical and chemical changes will immediately occur. Depending on the air composition, photons of certain specific wavelengths can excite the electronic transitions within the target pollutant molecules, causing the molecular bonds to break, and then triggering a series of chain reactions, ultimately decomposing the originally harmful substances into relatively safe small molecule compounds or even completely mineralizing them into harmless products such as carbon dioxide and water. In this process, the key lies in accurately matching the optimal excitation wavelength combination corresponding to each pollutant, which not only requires an in-depth understanding of the photochemical properties of various pollutants, but also needs to use advanced sensing technologies and data analysis methods to monitor the air quality in real time and dynamically adjust the light source parameters accordingly to achieve the best purification effect. In addition, it is worth noting that during the entire photolysis process, in addition to the main purification effect, there may also be a risk of side reactions. To avoid generating undesirable by-products, appropriate wavelengths and energy densities need to be carefully selected, and at the same time, reaction conditions such as residence time and air flow rate and other parameters need to be optimized. For example, too high an energy density may lead to unnecessary high-temperature thermal effects, thereby causing secondary pollution; on the contrary, if the energy is insufficient, the pollutants may not be completely degraded. Therefore, precisely controlling these variables is particularly crucial for ensuring the purification quality.For this reason, the photoreaction chamber is often equipped with a precise control system to monitor and adjust various parameters, ensuring that each link operates under optimal conditions. Further, to improve the purification efficiency and reduce energy consumption, the design of the photoreaction chamber also needs to consider maximizing the utilization rate of light energy. On the one hand, the design of optical elements (such as mirrors and lenses) can be improved to enhance the concentration and coverage of the light field; on the other hand, the application of new materials can also be explored, such as using coatings with high reflectivity to reduce energy loss. At the same time, it is also important to reasonably arrange the position and layout of the light source. An ideal configuration should enable the light to penetrate the air layer as much as possible, thereby increasing the effective irradiation area. This not only improves the purification rate per pass but also helps to shorten the total processing time, which is particularly beneficial for dealing with sudden severe pollution events. Finally, after the original air flow undergoes the above series of complex and delicate optical and chemical reactions, it is transformed into a processed clean air flow and flows out of the photoreaction chamber. During this process, although most pollutants have been effectively removed, in order to ensure that the air quality finally discharged into the external environment meets the standards, additional filtering devices or detection sensors sometimes need to be set at the outlet for final quality control. Only when it is confirmed that all pollutants have been reduced to below the safe level is this part of the air considered to have been properly treated.

[0063] In the above-mentioned air purification system 100 for atomic fluorescence, the purified air quality data packet generation module 150 is configured to input the processed air flow into the outlet sensor module to obtain a purified air quality data packet. It should be understood that inputting the processed air flow into the outlet sensor module not only verifies the effectiveness of the purification strategy but also provides a scientific basis for subsequent adjustments. First of all, considering that the ultimate goal of the entire air purification system is to improve air quality and ensure that the output air meets health and safety standards, it is particularly important to perform accurate quality detection on the processed air before it exits the system. This step is achieved through the outlet sensor module, which consists of a series of highly sensitive and specific sensors that can accurately measure different types of pollutants, including but not limited to particulate matter (such as PM2.5, PM10), volatile organic compounds (VOCs), carbon monoxide (CO), carbon dioxide (CO2), sulfur dioxide (SO2), and nitrogen oxides (NOx), etc. When the processed air flow enters the outlet sensor module, it undergoes a carefully designed path that aims to maximize the effective contact area with each sensor while ensuring the uniformity and stability of air flow. To achieve this, the outlet sensor module is usually equipped with an optimized airway structure that can guide the air flow smoothly through each sensor area, ensuring that each sensor receives a sufficient and consistent sample volume. In addition, there may be temperature and humidity adjustment devices inside the module to maintain ideal detection environmental conditions because some sensors are very sensitive to temperature and humidity changes, and any slight change may cause reading deviations or extended response times. During the actual operation process, the processed air flow first enters the pre-filtering area, where the main task is to preliminarily screen out the remaining large particulate matter in the air to prevent them from interfering with or clogging the subsequent high-precision sensors. The pre-filtered air is then distributed to different sensor units, each of which is responsible for monitoring a specific type of pollutant. For example, an optical particle counter can be used to measure the particulate matter concentration; an electrochemical sensor is suitable for detecting toxic gas components; and an infrared absorption sensor is good at quantifying the greenhouse gas content. Each sensor generates corresponding electrical signals according to its working principle, and these signals directly reflect the specific concentration levels of the measured pollutants. However, simply obtaining the original electrical signals is not sufficient to form a complete purified air quality data packet. The next crucial step is to perform detailed data processing and calibration on these signals. Since various sensors may be affected by multiple factors during actual use, such as decreased sensitivity due to long-term operation and baseline drift caused by environmental changes, advanced algorithms and technologies must be introduced to correct these problems. This includes but is not limited to signal amplification, noise filtering, temperature compensation, and regular calibration procedures, aiming to eliminate external interference and restore the optimal performance state of the sensors.On this basis, it is also necessary to convert all processed signals into digital information in a unified format for subsequent analysis and storage. Further, in order to ensure the authenticity and reliability of the obtained data, the outlet sensor module often needs to have a self-diagnosis function. This means that it can monitor its own operating status in real time and automatically identify potential problems or abnormal situations. For example, if a certain sensor shows an obvious trend of deviation in readings, the system should be able to issue an alarm in a timely manner and recommend appropriate maintenance measures, such as cleaning the sensor surface, replacing aging components or recalibrating the equipment. This self-protection mechanism helps to maintain the consistency and continuity of the data and avoid affecting the overall evaluation results due to individual fault points. In addition, considering the spatio-temporal variability characteristics of air pollution conditions, the outlet sensor module should also pay attention to flexibility and adaptability in design. On the one hand, it needs to be able to quickly respond to changes in external conditions and instantaneously update measurement parameters to reflect the latest air quality status; on the other hand, with the progress of technology and the upgrading of standards, the module should also reserve sufficient expansion space to facilitate the integration of new sensors or the improvement of existing components in the future. The advantage of doing so is that it can not only extend the service life of the equipment, but also continuously improve the detection ability and accuracy of the system. Finally, after all necessary data collection and processing are completed, a purified air quality data packet is generated. This data packet not only contains a series of numerical information about the concentrations of various pollutants, but more importantly, it also covers rich background information, such as sampling timestamps, geographical location coordinates, and meteorological parameters at that time. These additional information is crucial for deeply understanding the laws of air quality changes and the reasons behind them, and also provides strong support for formulating more effective environmental protection policies.

[0064] In the above air purification system 100 for atomic fluorescence, the purification skill report generation module 160 is used to generate a purification skill report based on the comparison between the air quality data packet and the post-purification air quality data packet. It should be understood that considering the complex and variable composition of air pollution and the different impacts of different pollutants on health, accurately quantifying the change in air quality before and after purification is crucial for verifying the system performance. The air quality data packet mentioned here contains detailed information such as the concentrations of various pollutants collected from the raw air, the particle size distribution, and the content of volatile organic compounds; while the post-purification air quality data packet records the changes in these parameters in the processed air. By carefully comparing these two data sets, the actual effect of the system in removing specific pollutants can be comprehensively understood, thus providing a scientific basis for technical improvement. When making the comparison, the first thing to solve is how to standardize the information in the two data packets for effective comparative analysis. Since the raw and post-purification air quality data may come from different sensors or measurement methods and may vary in data format, range, and even accuracy, a series of data preprocessing steps are needed to unify the standards. This includes but is not limited to operations such as unit conversion, error correction, and outlier removal to ensure the comparability of the two sets of data. In addition, considering the extremely low concentration of some pollutants or the detection limit problem, advanced mathematical models and statistical methods also need to be applied to fill in the missing values or smooth the curve, thereby improving the accuracy of the analysis results. Next, enter the core comparison stage. This step aims to deeply explore the specific change trends of the air quality indicators before and after purification and the reasons behind them. On the one hand, the purification efficiency can be intuitively reflected by directly calculating the reduction ratio of the concentration of each pollutant; on the other hand, more complex analysis methods, such as multiple regression analysis and time series prediction models, can also be used to explore the interaction relationships between various factors. For example, studying the correlation between the degradation rate of a certain pollutant under a specific wavelength combination and other environmental variables (such as temperature, humidity) helps to reveal the potential mechanism in the purification process and provides guidance for optimizing the light source configuration. At the same time, this in-depth comparison can also help identify the deficiencies in the system, such as whether there are some stubborn pollutants that are difficult to remove, or the phenomenon of decreased purification efficiency under specific conditions, etc. After completing the above basic comparison, the next step is to convert the results of these quantitative analyses into an easy-to-understand visual form. Graphical display can not only enable non-professionals to quickly grasp the main conclusions but also highlight some key trends and patterns. Commonly used chart types include bar charts, line charts, pie charts, and heat maps, etc., which are respectively suitable for showing different types of comparison relationships. For example, bar charts can be used to intuitively display the change range of the concentrations of various pollutants; line charts are suitable for tracking the development trend of a specific indicator over time; while heat maps can clearly present the complex interaction effects between multiple variables.In addition, to enhance the professionalism and persuasiveness of the report, some advanced visualization tools and technologies can be introduced, such as three-dimensional stereograms, dynamic interactive interfaces, etc., enabling readers to comprehensively examine the data analysis results from multiple dimensions. Further, in addition to simple numerical comparisons, a certain professional knowledge background and industry standards need to be incorporated as a reference framework during the process of generating the purification skills report. This is because it is sometimes difficult to accurately judge whether the purification effect has reached the expected goal based solely on absolute values, and a fair evaluation can only be made by placing it in a broader context. For example, relevant air quality standards promulgated by a country or region can be referred to determine whether the various indicators after purification meet the specified requirements; or the performance levels of other advanced devices in the same field can be used for reference to evaluate the relative advantages and disadvantages of this system. Doing so can not only provide users with more objective and real evaluation results, but also help to identify the positioning and future development direction of their own products in the global market. In addition, when compiling the purification skills report, special attention should be paid to how to effectively convey the deep-seated information hidden behind the data. This means not only reporting the facts seen on the surface, but also trying to explain the root causes of these phenomena. For example, if it is found that a certain wavelength combination performs excellently in removing specific pollutants, then the underlying physical and chemical principles should be explored, and whether it is possible to promote its application in other similar scenarios; conversely, if there is an obvious decline in efficiency in a certain link, then it is necessary to deeply explore whether there are design defects, material aging or other unforeseen factors behind it. Through in-depth analysis of these issues, valuable improvement suggestions can be provided for the technology R & D team, promoting the continuous optimization and upgrading of the entire system. Finally, after sorting out and summarizing all relevant information, the final purification skills report is formed. This report is not just a simple list of data, but a comprehensive document integrating detailed technical analysis, professional interpretation suggestions and forward-looking strategic thinking. It is both a summary of the current air purification work achievements and a guide for future improvement and development.

[0065] In summary, an air purification system 100 for atomic fluorescence based on the embodiments of the present application is elucidated. It obtains an air quality data packet and matches the current optimal excitation wavelength combination, and then inputs it into the light source driving module to generate a customized multi-wavelength mixed light field into the photoreaction cavity. At the same time, the raw air flow is input into the photoreaction cavity to obtain a processed air flow, and the candidate strategy closest or most relevant to the current air quality characteristics is quickly found in the strategy semantic set through the purification strategy semantic dynamic query method. Finally, the purification strategy that best meets the current needs, that is, the optimal excitation wavelength combination, is finally determined through a dynamic evaluation method, so as to realize the intelligent control of the light source and ensure that the system can provide the optimal purification performance in various complex air environments.

[0066] As described above, the air purification system 100 for atomic fluorescence according to the embodiments of the present application can be implemented in various terminal devices. In one example, the air purification system 100 for atomic fluorescence can be integrated into the terminal device as a software module and / or a hardware module. For example, the air purification system 100 for atomic fluorescence can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the air purification system 100 for atomic fluorescence can also be one of the many hardware modules of the terminal device.

[0067] Alternatively, in another example, the air purification system 100 for atomic fluorescence and the terminal device can also be separate devices, and the air purification system 100 for atomic fluorescence can be connected to the terminal device through a wired and / or wireless network and transmit interaction information in accordance with a predefined data format.

[0068] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there can be other division methods in actual implementation.

[0069] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0070] In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware or in the form of a hardware plus a software functional module.

[0071] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention.

[0072] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

[0073] In addition, it is obvious that the word "including" does not exclude other units or steps, and the singular form does not exclude the plural form. A plurality of units or devices stated in the system claims can also be implemented by one unit or device through software or hardware. The terms such as "second" are used to denote names and do not denote any particular order.

Claims

1. An air purification system for atomic fluorescence, characterized in that, Including: An air quality data generation module, configured to input an original air stream into an integrated multi-sensor module to obtain an air quality data packet; An excitation wavelength matching module, configured to match a current optimal excitation wavelength combination from a pollutant-wavelength-efficiency database based on the air quality data packet to obtain a light source control instruction; A customized multi-wavelength light source generation module, configured to input the light source control instruction into a light source driving module to generate a customized multi-wavelength mixed light field into a photoreaction chamber through the light source driving module; An air stream processing module, configured to input the original air stream into the photoreaction chamber, wherein the original air stream is decomposed under the action of the customized multi-wavelength mixed light field to obtain a processed air stream; A purified air quality data packet generation module, configured to input the processed air stream into an outlet sensor module to obtain a purified air quality data packet; A purification skill report generation module, configured to generate a purification skill report based on a comparison between the air quality data packet and the purified air quality data packet.

2. The air purification system for atomic fluorescence according to claim 1, wherein The excitation wavelength matching module includes: An air quality data packet structured encoding unit, configured to perform structured embedding encoding on the air quality data packet to obtain an air quality data packet structured embedding encoding vector; A preset purification strategy structured encoding unit, configured to perform structured embedding encoding on each preset purification strategy in the pollutant-wavelength-efficiency database to obtain a set of preset purification strategy structured embedding encoding vectors; A strategy query response encoding unit, configured to input the air quality data packet structured embedding encoding vector and the set of preset purification strategy structured embedding encoding vectors into a purification strategy semantic dynamic query network to obtain an air quality-purification strategy query response encoding vector; An optimal excitation wavelength combination generation unit, configured to perform feature decoding on the air quality-purification strategy query response encoding vector to obtain the current optimal excitation wavelength combination.

3. The air purification system for atomic fluorescence according to claim 2, wherein, The strategy query response encoding unit includes: A preset purification strategy structured embedding concentration sub-unit, configured to perform information concentration on each preset purification strategy structured embedding encoding vector in the set of preset purification strategy structured embedding encoding vectors to obtain a set of preset purification strategy structured embedding concentration encoding vectors; A semantic association degree construction sub-unit, configured to construct a sparse association topology matrix between preset purification strategy feature nodes based on the semantic association degree between any two preset purification strategy structured embedding concentration encoding vectors in the set of preset purification strategy structured embedding concentration encoding vectors; A preset purification strategy feature graph spectrum encoding sub-unit, configured to input the set of preset purification strategy structured embedding concentration encoding vectors and the sparse association topology matrix between preset purification strategy feature nodes into a graph spectrum construction engine based on a graph convolutional network model to obtain a preset purification strategy feature graph spectrum encoding matrix; An information query sub-unit, configured to perform information query in the preset purification strategy feature graph spectrum encoding matrix based on the air quality data packet structured embedding encoding vector to obtain the air quality-purification strategy query response encoding vector.

4. An air purification system for atomic fluorescence according to claim 3, characterized in that, The semantic association degree construction subunit includes: Calculating the semantic association degree between any two preset purification strategy structured embedding concentrated coding vectors in the set of the preset purification strategy structured embedding concentrated coding vectors to obtain a correlation topology matrix between preset purification strategy feature nodes; Inputting the correlation topology matrix between the preset purification strategy feature nodes into a gated mask network to obtain a sparse correlation topology matrix between the preset purification strategy feature nodes.

5. An air purification system for atomic fluorescence according to claim 4, characterized in that The information query subunit includes: Inputting the air quality data packet structured embedding coding vector and the preset purification strategy feature spectrum coding matrix into a feature query response engine to obtain the air quality - purification strategy query response coding vector.

6. The air purification system for atomic fluorescence according to claim 5, characterized in that The optimal excitation wavelength combination generation unit is configured to: Pass the air quality - purification strategy query response coding vector through an optimal excitation wavelength combination analyzer based on a classifier to obtain an analysis result, and the analysis result is used to represent the label of the current optimal excitation wavelength combination.

Images