Method, system and storage medium for pollution monitoring of river water

Abstract

The present application relates to the technical field of pollution monitoring, and more particularly to a pollution monitoring method and system for river water and a storage medium, comprising: obtaining river sediment spectral data through multi-point synchronous acquisition technology, and extracting characteristic vectors using principal component analysis to determine the types of pollutants; combining hydrodynamic and molecular polarization effects, using back propagation neural networks to predict the diffusion direction of pollutants and the interface polarization intensity, and generating the distribution pattern of pollutants in sediments; obtaining the potential gradient and chemical bond amplitude characteristics of the spatial distribution of pollution sources; matching the signal attenuation rate of the spectral signal with the chemical bond strength model to determine the emission frequency of pollutants and the interface scattering effect; combining the persistence of pollution sources and the chemical bond amplitude to trace the spatial distribution, emission frequency and potential gradient of pollution sources. Through these steps, the present application realizes accurate tracing and dynamic monitoring of pollution sources, effectively improving the accuracy of pollutant detection and the scientificity of water quality management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pollution monitoring, in particular to a pollution monitoring method and system for river water and a storage medium. BACKGROUND

[0002] River sediment pollutant analysis and tracing is a complex technical problem. Traditional methods are difficult to accurately identify the types, sources and dynamic variation rules of multiple pollutants. The diffusion process of pollutants in the river is affected by hydrodynamic conditions, and also involves polarization effects at the molecular level, which increases the difficulty of analysis. The spatial distribution of pollution sources presents the characteristics of potential gradient and chemical bond amplitude, but how to extract effective information from these characteristics and establish a correlation with the pollution emission frequency has not been solved. In addition, the dynamic variation trend of pollutants on the time scale and the persistence characteristics of pollution sources also need to be further studied. Most importantly, how to organically combine these complex physical and chemical processes with spectral data analysis methods to establish a systematic river sediment pollutant analysis and tracing technology system. This not only needs to solve the problem of identifying a single pollutant, but also needs to consider the interaction between multiple pollutants and their migration and transformation rules in the river ecosystem. At the same time, how to ensure the accuracy and reliability of the analysis results so that they can provide strong support for water environment governance decision-making is also a problem to be solved. SUMMARY

[0003] The present application provides a pollution monitoring method for river water, mainly comprising:

[0004] Step S1: obtaining river sediment spectral data, extracting a first feature vector according to principal component analysis, and determining the type of pollutant according to the cosine similarity of the first feature vector and a preset pollutant spectral template;

[0005] Step S2: determining the diffusion direction of the pollutant and the interface polarization intensity according to the type of the pollutant, the hydrodynamic influence and the molecular polarization effect through a back propagation neural network, and obtaining the distribution mode of the pollutant in the sediment;

[0006] Step S3: obtaining the potential gradient distribution of the spatial distribution of the pollution source and the spatial characteristics of the chemical bond amplitude through cluster analysis according to the distribution mode and the wavelength scattering intensity, and matching the signal attenuation rate of the spectral signal with a preset chemical bond strength model to determine the pollution emission frequency and the interface scattering effect;

[0007] Step S4: obtaining the dynamic trend of the pollutant pulse duration and the deposition ratio of the pollutant in the sediment by analyzing the time series of the pollution emission frequency and the interface scattering effect;

[0008] Step S5: judging the persistence of the pollution source according to the deviation of the dynamic trend from the preset pollutant diffusion model, and obtaining the potential gradient distribution of the spatial distribution of the pollution source and the tracing result of the pollutant emission frequency through weighted average.

[0009] As a preferred technical solution of the present application, in step S1, the river sediment spectrum data is acquired, including:

[0010] The spectrum data is acquired from the surface sediment and deep sediment of the river through multi-point synchronous acquisition, the original spectrum data set containing spectrum phase shift, time delay distribution and wavelength absorption coefficient is obtained, the characteristic parameters of the spectrum signal are extracted from the original spectrum data set, the characteristic parameters include spectrum phase shift, time delay distribution and wavelength absorption coefficient, the initial feature matrix of the spectrum data set is constructed according to the characteristic parameters, and the optimized spectrum data set is obtained by denoising the initial feature matrix.

[0011] As a preferred technical solution of the present application, the first feature vector is extracted by principal component analysis, including:

[0012] The original spectrum data set is standardized to obtain a standardized spectrum matrix, the standardized spectrum matrix is decomposed by principal component analysis, the first feature vector containing spectrum phase shift, surface charge density and time delay distribution is extracted, the variance contribution rate of the first feature vector is calculated, the principal components are screened according to the variance contribution rate to obtain a principal component feature set, and the optimized first feature vector is generated by orthogonalizing the principal component feature set.

[0013] As a preferred technical solution of the present application, in step S1, the pollutant species is determined according to the cosine similarity between the first feature vector and the preset pollutant spectrum template, including:

[0014] The preset pollutant spectrum template is acquired, the cosine similarity between the first feature vector and the preset pollutant spectrum template is calculated, the cosine similarity is compared with a preset threshold, if the cosine similarity is greater than the preset threshold, the pollutant species is determined, the spectrum characteristic parameters of the determined pollutant species are extracted to generate a pollutant feature vector, and the accuracy of the pollutant species is verified according to the matching result of the pollutant feature vector and the preset pollutant spectrum template.

[0015] As a preferred technical solution of the present application, in step S2, the diffusion direction of the pollutant and the interface polarization intensity are judged by the back propagation neural network, including:

[0016] The input feature set is constructed according to the first feature vector corresponding to each pollution category, training data set is generated by combining the hydrodynamic effect and the molecular polarization effect, the pollution diffusion prediction model is obtained by training the training data set through the back propagation neural network, the pollution diffusion direction and the interface polarization intensity are calculated through the pollution diffusion prediction model, the distribution mode of the pollution in the sediment is generated, and the spatial characteristics of the pollution distribution are obtained by verifying the distribution mode.

[0017] As a preferred technical solution of the present application, in step S3, the potential gradient distribution and the spatial characteristics of the chemical bond amplitude of the pollution source spatial distribution are obtained through clustering analysis, including:

[0018] According to the spatial characteristic parameters of the pollution distribution in the sediment, the feature data set is generated by combining the time delay distribution and the wavelength scattering intensity, the potential gradient distribution of the pollution source spatial distribution is obtained by classifying the feature data set through clustering analysis, the spatial characteristics of the chemical bond amplitude are calculated, and the feature matrix of the pollution source spatial distribution is generated by standardizing the potential gradient distribution and the chemical bond amplitude.

[0019] As a preferred technical solution of the present application, in step S3, the signal attenuation rate of the spectrum signal is matched with the preset chemical bond strength model, including:

[0020] The signal attenuation rate of the spectrum signal is obtained, the preset chemical bond strength model is extracted, the matching degree of the signal attenuation rate and the preset chemical bond strength model is calculated, the matching degree is compared with the preset threshold, if the matching degree is greater than the preset threshold, the pollution emission frequency and the interface scattering effect are determined, the spectrum analysis is performed on the pollution emission frequency, the feature parameters of the interface scattering effect are generated, and the accuracy of the pollution emission frequency is verified.

[0021] As a preferred technical solution of the present application, the potential gradient distribution of the pollution source spatial distribution and the tracing result of the pollution emission frequency are obtained by weighted average, including:

[0022] According to the persistence of the pollution source and the chemical bond amplitude, the feature parameters are extracted, the weighted feature set is generated by combining the potential gradient distribution and the wavelength modulation characteristic, the weighted feature set is processed by weighted average, the potential gradient distribution of the pollution source spatial distribution is obtained, the tracing result of the pollution emission frequency is calculated, the tracing result is verified, and the final feature matrix of the pollution source spatial distribution is generated.

[0023] In a second aspect, the present application also provides a pollution monitoring system for river water, which is used to realize the above-mentioned method, and the system comprises:

[0024] The determining unit is configured to acquire river sediment spectrum data, extract a first feature vector according to principal component analysis, and determine a pollutant type according to a cosine similarity between the first feature vector and a preset pollutant spectrum template;

[0025] The predicting unit is configured to determine a pollutant diffusion direction and an interface polarization intensity by back propagation neural network according to the pollutant type, a water dynamic effect and a molecular polarization effect, and obtain a distribution mode of the pollutant in the sediment;

[0026] The matching unit is configured to obtain a potential gradient distribution of a pollution source spatial distribution and a spatial feature of a chemical bond amplitude by cluster analysis according to the distribution mode and a wavelength scattering intensity, and determine a pollutant emission frequency and an interface scattering effect by matching a signal attenuation rate of the spectrum signal with a preset chemical bond strength model.

[0027] The analyzing unit is configured to obtain a pollutant pulse duration and a dynamic trend of a pollutant deposition ratio in the sediment by analyzing a time sequence of the pollutant emission frequency and the interface scattering effect.

[0028] The tracing unit is configured to determine a persistence of the pollution source according to a deviation of the dynamic trend from a preset pollutant diffusion model, and obtain a tracing result of the potential gradient distribution of the pollution source spatial distribution and the pollutant emission frequency by weighted average.

[0029] In a third aspect, the present application also provides a computer readable storage medium, which has instructions stored thereon, and the instructions are executed by a processor to implement the method described above.

[0030] The technical scheme provided by the embodiments of the present application can have the following beneficial effects:

[0031] The present application acquires river sediment spectrum data by multi-point synchronous acquisition technology, extracts a feature vector by principal component analysis, and determines a pollutant type according to a similarity with a preset template. In combination with a water dynamic effect and a molecular polarization effect, a back propagation neural network is used to determine a pollutant diffusion direction and an interface polarization intensity, and a pollutant distribution mode is obtained. By cluster analysis, a potential gradient of a pollution source spatial distribution and a chemical bond amplitude feature are obtained, and a pollutant emission frequency is determined by matching a spectrum signal attenuation rate with a chemical bond strength model. By time sequence analysis, a pollutant dynamic trend is obtained, a persistence of a pollution source is determined, and finally a weighted average algorithm is used to trace a pollution source distribution and an emission frequency. The present application realizes accurate analysis and tracing of pollutants in river sediment, and provides effective technical support for water environment management. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 A flowchart of a pollution monitoring method for river water in the embodiments of the present application;

[0033] Figure 2 The structure diagram of the pollution monitoring system for river water in the embodiment of the present application. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present application will be clearly and detailedly described with reference to the drawings in the embodiments of the present application. The described embodiments are only some embodiments of the present application.

[0035] As Figure 1 The pollution monitoring method for river water in the embodiment can specifically include the following steps.

[0036] Step S1: acquiring river sediment spectrum data, extracting a first feature vector according to principal component analysis, and determining a pollutant type according to a cosine similarity of the first feature vector and a preset pollutant spectrum template;

[0037] The acquiring of the river sediment spectrum data includes the following steps.

[0038] The spectrum data of the river surface sediment and deep sediment are acquired through multi-point synchronous acquisition to obtain an original spectrum data set containing spectrum phase shift, time delay distribution and wavelength absorption coefficient, the original spectrum data set is preprocessed, the feature parameters of the spectrum signal are extracted, the feature parameters include the spectrum phase shift, the time delay distribution and the wavelength absorption coefficient, an initial feature matrix of the spectrum data set is constructed according to the feature parameters, and the initial feature matrix is denoised to obtain an optimized spectrum data set.

[0039] Specifically, the spectrum data of the river surface sediment and deep sediment are acquired through multi-point synchronous acquisition, which includes spectrum phase shift, time delay distribution and wavelength absorption coefficient and other features. These original spectrum data can reflect the distribution characteristics of pollutants in different sediment layers. The collected spectrum data is preprocessed to extract feature parameters containing key information, such as spectrum phase shift, time delay distribution and wavelength absorption coefficient. These parameters can reflect the physical and chemical properties of pollutants. Based on the above feature parameters, an initial feature matrix is constructed to organize these spectrum feature parameters into a matrix format suitable for further analysis, so as to process the data. In order to remove the noise in the acquisition process and eliminate the interference of the noise on the analysis results, the initial feature matrix is preprocessed and denoised to obtain the optimized spectrum data set. The above technical solution can effectively extract the feature information of pollutants and eliminate the influence of environmental noise on data, providing high-quality input data for subsequent pollutant type determination, diffusion model analysis and other processes, and more accurately reflecting the real distribution of pollutants in the sediment, thereby providing a reliable basis for pollution source tracing and water quality management decision-making.

[0040] Further, the first feature vector is extracted through principal component analysis, including the following steps.

[0041] The original spectrum data set is standardized to obtain a standardized spectrum matrix, the standardized spectrum matrix is decomposed by principal component analysis, the first characteristic vector containing spectral phase shift, surface charge density and time delay distribution is extracted, the variance contribution rate of the first characteristic vector is calculated, the principal components are screened according to the variance contribution rate, the principal component characteristic set is obtained, and the optimized first characteristic vector is generated by orthogonalizing the principal component characteristic set.

[0042] Specifically, by standardizing the original spectrum data set, the absorption coefficient data of different wavelengths is adjusted to a unified scale, so that the difference between different data dimensions will not affect the subsequent analysis; the spectrum matrix after standardization is decomposed by the PCA algorithm, PCA projects the original data to a new feature space through linear transformation, extracts several principal components reflecting the main variation of the data, and selects key features such as spectral phase shift, surface charge density and time delay distribution; the variance contribution rate of each principal component reflects its importance in the data, and the principal components selected according to the variance contribution rate can effectively retain most of the effective information in the data and remove noise; the selected principal component characteristic set is orthogonalized to eliminate the correlation between the principal components and ensure their linear independence, thereby improving the stability and accuracy of subsequent analysis, obtaining the optimized first characteristic vector, and using it as the basis input for pollutant detection, which can help accurately identify and quantitatively analyze different types of pollutants. The above technical solution combines standardization and PCA to effectively suppress noise in spectral data and extract key features of pollutant distribution, and then through subsequent pollution source tracing and diffusion prediction, the monitoring accuracy and efficiency are improved.

[0043] Further, the pollutant type is determined according to the cosine similarity between the first characteristic vector and the preset pollutant spectrum template, comprising:

[0044] The preset pollutant spectrum template is obtained, the cosine similarity between the first characteristic vector and the preset pollutant spectrum template is calculated, the cosine similarity is compared with the preset threshold, if the cosine similarity is greater than the preset threshold, the pollutant type is determined, the spectral feature parameters of the determined pollutant type are extracted to generate a pollutant characteristic vector, and the accuracy of the pollutant type is verified according to the matching result of the pollutant characteristic vector and the preset pollutant spectrum template.

[0045] Specifically, by performing cosine similarity calculation on the first feature vector and the spectral templates, the cosine similarity is a method for measuring the similarity of two vectors in the direction, and the closer the value is to 1, the higher the similarity of the two vectors. If the calculated cosine similarity is greater than a preset threshold, it can be determined that the pollutant species is the pollutant represented by the template; after determining the pollutant species, the spectral feature parameters related to the pollutant species are extracted to generate a pollutant feature vector. The feature vector integrates the absorption characteristics, phase changes and other spectral information of the pollutant at different wavelengths, which is used for subsequent pollutant analysis; by matching the pollutant feature vector with the preset pollutant spectral template again, the accuracy of the pollutant species is verified, and the reliability of the identification result is further ensured; the above technical solution combines feature extraction of spectral data and cosine similarity algorithm, so that in complex river water pollution monitoring, the pollutant species can be accurately identified, and the difficulty of traditional methods in multiple pollutant identification and pollution source tracing is solved.

[0046] Step S2: determining the diffusion direction and interface polarization intensity of the pollutant in the sediment according to the pollutant species, the hydrodynamic effect and the molecular polarization effect by the back propagation neural network, to obtain the distribution mode of the pollutant in the sediment;

[0047] The back propagation neural network for determining the diffusion direction and interface polarization intensity of the pollutant includes:

[0048] According to the first feature vector corresponding to each pollutant species, an input feature set is constructed, a training data set is generated in combination with the hydrodynamic effect and the molecular polarization effect, the training data set is trained by the back propagation neural network, a pollutant diffusion prediction model is obtained, the diffusion direction and interface polarization intensity of the pollutant are calculated by the pollutant diffusion prediction model, the distribution mode of the pollutant in the sediment is generated, and the spatial characteristics of the pollutant distribution are obtained by verifying the distribution mode.

[0049] Specifically, the core of step S2 is to use the back propagation neural network (BP neural network) to determine the diffusion direction and interface polarization intensity of the pollutant, and generate the distribution mode of the pollutant in the sediment through these information. This process first involves comprehensive analysis of the pollutant species, the hydrodynamic effect and the molecular polarization effect, and uses the neural network model to process and predict these factors, and then deduces the spatial distribution characteristics of the pollutant.

[0050] Specifically, by obtaining spectral data from river sediment, the first feature vector corresponding to each pollutant species is extracted, which is generated by principal component analysis method and is the key data for inputting neural network. Based on the above first feature vector, combined with hydrodynamic influence and molecular polarization effect, these feature vectors and hydrological environment information such as flow velocity, turbulence intensity and charge distribution of pollutant molecules are integrated into a complete input feature set. By simulating the diffusion behavior of different pollutant species under various hydrodynamic and molecular polarization effects, the data set is labeled. In the training process, the neural network adjusts its network weights according to these input feature data, gradually learns the relationship between the diffusion mode of pollutants and the interface polarization intensity, and constantly optimizes its prediction error through back propagation, so that the network can accurately predict the diffusion direction of pollutants under different hydrological conditions. After training, the back propagation neural network will generate a pollutant diffusion prediction model, which can use known input features to predict the diffusion direction of pollutants in sediment and interface polarization intensity. Among them, the diffusion direction and interface polarization intensity are two key factors to describe how pollutants distribute in river sediment, which reflect the migration path of pollutants and the interaction between pollutants and sediment interface. Through the prediction model, the distribution pattern of pollutants in sediment can be generated, which shows the distribution of pollutant concentration in different spatial positions. In this process, the back propagation neural network can not only predict the migration route of pollutants according to the difference of hydrodynamic and molecular polarization effect, but also accurately analyze the deposition of pollutants in sediment. This distribution pattern can be further verified by comparing with the field sampling data to verify the accuracy of the model prediction results, ensuring that the spatial characteristics of pollutant distribution are consistent with the actual situation.

[0051] The technical solution solves several core problems existing in the traditional pollutant monitoring method. First, the traditional method often cannot simultaneously consider the comprehensive influence of multiple factors such as pollutant types, river water dynamics and molecular polarization effect, and through the introduction of the back propagation neural network, these complex multivariate relationships are effectively modeled, which can more accurately predict the diffusion and deposition behavior of pollutants in the river; second, the method can realize real-time dynamic monitoring, and through analyzing the diffusion mode of the pollutants, the direction and intensity change of the pollutant migration are obtained in time, which provides strong support for pollution source tracing. More importantly, the neural network-based model can be continuously optimized with more data accumulation, improving the prediction accuracy and having strong adaptability; for example, in actual application, if the river is in strong water flow conditions, the diffusion direction of the pollutants may be significantly affected by the water dynamics, and under static water conditions, the molecular polarization effect may dominate the migration path of the pollutants, and the back propagation neural network can combine these two factors to predict the dynamic behavior of the pollutants through the trained model, providing accurate data support for pollution source tracing and water environment management.

[0052] Step S3: obtaining the potential gradient distribution of the spatial distribution of the pollution source and the spatial characteristics of the chemical bond amplitude by cluster analysis according to the distribution mode and the wavelength scattering intensity, matching the signal attenuation rate of the spectral signal with the preset chemical bond strength model to determine the pollutant emission frequency and the interface scattering effect;

[0053] The potential gradient distribution of the spatial distribution of the pollution source and the spatial characteristics of the chemical bond amplitude obtained by cluster analysis include:

[0054] According to the distribution mode of the pollutants in the sediment, the spatial feature parameters are extracted, the feature data set is generated in combination with the time delay distribution and the wavelength scattering intensity, the feature data set is classified through cluster analysis, the potential gradient distribution of the spatial distribution of the pollution source is obtained, the spatial characteristics of the chemical bond amplitude are calculated, the potential gradient distribution and the chemical bond amplitude are standardized, and the feature matrix of the spatial distribution of the pollution source is generated.

[0055] Specifically, spatial feature parameters are extracted according to the distribution pattern of pollutants in river sediments, including the concentration, distribution density of pollutants, and the influence of water flow on pollutant diffusion, etc. These data reflect the spatial distribution of pollutants in sediments and can provide a basis for subsequent analysis. On this basis, a feature dataset is generated by combining time delay distribution and wavelength scattering intensity. Time delay distribution refers to the propagation delay of optical signals during the diffusion process of pollutants, and wavelength scattering intensity reflects the interaction intensity of pollutants and optical signals. These data can provide information on the changes of pollutants in different sediment layers. After the dataset is constructed, clustering analysis is used to classify the feature data to further reveal the spatial distribution of pollution sources. Clustering analysis can identify the concentration area and diffusion direction of pollutants by grouping data according to similarity, thereby forming the spatial features of potential gradient distribution and chemical bond amplitude. Potential gradient distribution can reflect the diffusion trend of pollutants in sediments, and chemical bond amplitude provides information at the molecular level of pollutants, showing the chemical bonding strength between pollutants and sediment particles. These two features together represent the behavior and state of pollutants in the river ecosystem. By standardizing the potential gradient distribution and chemical bond amplitude, these two features can be compared and analyzed at the same scale. Standardization helps to eliminate the dimensional differences in the data and ensures the comparability between different features, thereby generating a feature matrix of pollution source spatial distribution. This matrix contains spatial information of pollution sources, pollutant diffusion intensity, and interaction intensity with sediments, providing accurate spatial positioning for pollution source tracing and water body management.

[0056] The above technical solution can achieve efficient and accurate pollution source identification and distribution analysis during pollutant monitoring, and can reflect the diffusion trend of pollutants and the persistence of pollution sources in real time. It solves the problem that traditional monitoring methods cannot capture the change pattern of pollutants. By combining spectral data and physical and chemical characteristics, this method lays the foundation for pollution source tracing, diffusion direction prediction, and optimization of pollution control measures. For example, in the pollution monitoring of a certain river, clustering analysis technology can be used to identify the specific location of the pollution source and understand the diffusion trend of pollutants. When this technology is applied to water pollution management, managers can take targeted management measures based on the spatial distribution characteristics of pollution sources, such as adjusting water flow scheduling or setting up pollution source isolation belts, thereby optimizing the management effect.

[0057] Further, the signal attenuation rate of the matching spectral signal is matched with a preset chemical bond strength model, including:

[0058] The signal attenuation rate of the spectral signal is obtained, a preset chemical bond strength model is extracted, the matching degree of the signal attenuation rate and the preset chemical bond strength model is calculated, the matching degree is compared with a preset threshold, if the matching degree is greater than the preset threshold, the pollutant emission frequency and the interface scattering effect are determined, spectral analysis is performed on the pollutant emission frequency, the characteristic parameters of the interface scattering effect are generated, and the accuracy of the pollutant emission frequency is verified.

[0059] Specifically, the spectral signal in the river water is obtained by a spectral acquisition technology. After a series of processing, the signal attenuation rate of the spectral signal is calculated, which reflects the degree of attenuation of the optical signal in the river sediment. The signal attenuation rate is usually closely related to the concentration, distribution and nature of the pollutant. The attenuation rate is obtained by measuring the intensity change of the light wave when propagating in the water body, combined with the wavelength absorption coefficient and other optical parameters; a preset chemical bond strength model is also used, which is established in advance according to the type of pollutant and the characteristics of the sediment, and can describe the strength of the chemical bond between the pollutant molecules and the sediment and its change over time. The model reflects the emission characteristics of the pollutant and its behavior in the sediment, including the diffusion, adsorption and interface reaction of the pollutant; the model provides a theoretical basis for analyzing the behavior of the pollutant, and can be matched with the actually measured spectral signal attenuation rate; by calculating the matching degree between the spectral signal attenuation rate and the chemical bond strength model, the distribution pattern of the pollutant in the river sediment can be evaluated, and when the matching degree exceeds the preset threshold, the emission frequency of the pollutant and the characteristics of the interface scattering effect can be determined. This process provides accurate data support for monitoring and tracing of pollution sources, especially in the determination of the emission rules and diffusion direction of the pollutant, which can effectively improve the accuracy of the detection results.

[0060] The pollutant emission frequency refers to the emission intensity of the pollution source within a certain time, which directly affects the concentration change of the pollutant in the water body, and the interface scattering effect reflects the interaction between the pollutant and the interface of the water body and the sediment, which is usually related to the chemical properties, physical state and diffusion characteristics of the pollutant. By analyzing the emission frequency of the pollutant through spectral analysis, the time information about the emission of the pollutant can be extracted from the spectral data, which provides a reference for the dynamic change of the pollution source. This spectral analysis can effectively reveal the short-term and long-term emission patterns of the pollutant, supporting targeted pollution control; after determining the emission frequency, the accuracy of the frequency is further verified by combining the actual river water environment data and historical pollution data. The verification step can eliminate errors and false positive results to ensure the reliability of the pollution monitoring data. By comparing the pollutant emission data of different time periods, the correlation between the data and factors such as environmental changes and flow changes is analyzed, so as to accurately predict the dynamic change trend of the pollutant.

[0061] The technical scheme solves the problem that the emission frequency, diffusion direction and deposition distribution of pollutants cannot be monitored in real time in the prior art, provides a more accurate pollutant monitoring method by matching the spectral attenuation rate and the chemical bond strength model, and can reflect the diffusion and change trend of the pollutants in real time, thereby improving the accuracy of the pollutant detection and making the tracing and treatment of the pollution source more efficient.

[0062] Step S4: obtaining the dynamic trend of the pollutant pulse duration and the deposition ratio of the pollutants in the sediment by analyzing the time series of the pollutant emission frequency and the interface scattering effect;

[0063] Specifically, since the pollutant emission frequency reflects the emission intensity of the pollutants, and the interface scattering effect indicates the scattering and adsorption degree of the pollutants in the sediment, the combination of the two can reveal the diffusion and accumulation process of the pollutants in the river water, and therefore, by time series analysis of the data, the duration of the pollutant emission can be determined, and the change trend of the deposition ratio of the pollutants in the river sediment can be inferred, which reflects the activity cycle, diffusion speed and deposition capacity of the pollution source, thereby providing a basis for the tracing and treatment of the pollution source.

[0064] By analyzing the time series of the pollutant emission frequency and the interface scattering effect, a dynamic model of pollutant diffusion is generated according to the data input of the frequency of pollutant emission and the interface scattering effect, and the duration of the pollutant emission is further derived. The dynamic model is a model trained by the historical emission frequency of the pollutants, the corresponding interface scattering effect and the corresponding historical duration. The above analysis process closely combines the diffusion behavior of the pollutants and their deposition characteristics, can track the dynamic change of the pollutants in real time, quantifies the migration mode, diffusion law and deposition rate of the pollutants, enhances the timeliness and accuracy of the pollution monitoring, can effectively judge whether the pollution source is a continuous source, and predict the potential risk and distribution mode of the pollutants, and especially provides data support for water pollution monitoring and river pollution treatment. For example, in a certain river section, it can be determined by the technical scheme that when the emission frequency of a certain pollutant is high, the accumulation ratio of the pollutant in the sediment will increase, and vice versa. The short-term change of the pollutant pulse indicates that the pollution source may be an intermittent emission source, thereby providing data basis for the environmental management department to take corresponding treatment measures.

[0065] Step S5: judging the continuity of the pollution source according to the deviation of the dynamic trend from a preset pollutant diffusion model, and obtaining the potential gradient distribution of the spatial distribution of the pollution source and the tracing result of the pollutant emission frequency by weighted averaging.

[0066] The potential gradient distribution of the spatial distribution of the pollution source and the tracing result of the pollutant emission frequency obtained by weighted averaging include:

[0067] According to the persistence of the pollution source and the amplitude of the chemical bond, a weighted feature set is generated combining the potential gradient distribution and the wavelength modulation characteristics, the weighted feature set is processed through weighted average, the potential gradient distribution of the spatial distribution of the pollution source is obtained, the tracing result of the pollutant emission frequency is calculated, the tracing result is verified, and the final feature matrix of the spatial distribution of the pollution source is generated.

[0068] Specifically, the technical solution aims to utilize the pollutant diffusion model and weighted average to determine the persistence of the pollution source, and obtain the potential gradient distribution of the spatial distribution of the pollution source and the tracing result of the pollutant emission frequency through data processing. The implementation principle is based on analyzing the dynamic trend and diffusion characteristics of the river pollutant, combining the weighted average technology, optimizing and integrating different data characteristics, and generating high-credibility pollution source spatial distribution information.

[0069] In the process of pollutant diffusion, pollutants will diffuse with water flow, sediment distribution, and molecular polarization effect, etc. The diffusion model has the characteristics of time and space. The training data of the diffusion model includes historical pollutant spectral data, hydrodynamic influence, molecular polarization effect, interface polarization intensity, diffusion direction, and corresponding pollutant distribution trend. When the detected dynamic trend deviates greatly from the predicted dynamic trend obtained by inputting the current pollutant spectral data, hydrodynamic influence, molecular polarization effect, diffusion direction, and interface polarization intensity into the pollutant diffusion model, it can be inferred that the pollution source may have changed, or its pollution emission is more persistent. This deviation analysis will provide basic data for the weighted average algorithm, helping to further optimize the spatial distribution and emission characteristics of the pollution source; the weighted feature set is constructed by chemical bond amplitude, potential gradient distribution, and wavelength modulation characteristics, etc. These feature parameters provide quantitative support for the distribution, migration, and chemical reaction process of pollutants in sediments; for example, the chemical bond amplitude can reflect the binding strength of pollutants and sediments, the potential gradient distribution reveals the diffusion trend of pollutants under the action of water flow, and the wavelength modulation characteristics further describe the interaction between pollutants and spectral signals. Integrating these feature information helps to improve the description accuracy of the spatial distribution of pollution sources and accurately reflects the diffusion range and concentration gradient of pollutants in the river; the influence of different features is fused according to the preset weight by weighted average, so as to obtain the potential gradient distribution of the spatial distribution of pollution sources and the emission frequency of pollutants, which not only optimizes the processing effect of feature data, but also reduces the error that may be caused by a single data source. The weighted average method can find a suitable balance point between different features by reasonably configuring the weight; for example, the potential gradient has a large weight in evaluating the spatial expansion of pollution sources, while the chemical bond amplitude can effectively distinguish the type of pollution source. In this way, the result of weighted processing will more accurately reflect the real distribution and emission of pollution sources. By tracing the spatial distribution and emission frequency of pollution sources, the final feature matrix of pollution sources can be generated, which contains the potential gradient and emission frequency of pollution sources at different time points and spatial positions, and can provide accurate decision basis for pollutant control. By verifying the tracing result, the model can be further optimized, making the tracing result of pollution sources more reliable, and providing scientific guidance for subsequent pollution source control and water quality improvement work.For example, in the pollution monitoring of urban river channels, the location of the pollution source can be accurately tracked, and the emission frequency of the pollutants can be evaluated. If a pollution source is located upstream of the river channel and frequently emits heavy metal pollutants, the algorithm can generate a clear pollution source distribution map by combining the potential gradient and the chemical bond amplitude characteristics, and determine the persistence of the pollution source through dynamic trend analysis. At this time, the weighted average method can be used to combine different characteristics, such as water flow and chemical properties of pollutants, for accurate analysis, and ultimately form the spatial distribution and emission frequency tracking results of the pollution source, thereby achieving accurate monitoring and management of the pollution source.

[0070] The above technical solution solves the problems of inaccurate pollution source tracking and difficult pollution source persistence judgment in the pollution monitoring process by combining the pollutant diffusion model, dynamic trend analysis and weighted average, and can achieve accurate positioning and dynamic monitoring of the pollution source.

[0071] The present application also provides a pollution monitoring system for river water, which is used to implement the above method, as shown in Figure 2 The system comprises:

[0072] A determination unit is configured to obtain river sediment spectrum data, extract a first feature vector according to principal component analysis, and determine the type of pollutants according to the cosine similarity between the first feature vector and a preset pollutant spectrum template.

[0073] A prediction unit is configured to determine the diffusion direction of the pollutants and the interface polarization intensity by backpropagation neural network according to the type of pollutants, hydrodynamic influence and molecular polarization effect, and obtain the distribution mode of the pollutants in the sediment.

[0074] A matching unit is configured to obtain the potential gradient distribution of the spatial distribution of the pollution source and the spatial characteristics of the chemical bond amplitude through cluster analysis according to the distribution mode and the wavelength scattering intensity, and match the signal attenuation rate of the spectrum signal with a preset chemical bond strength model to determine the emission frequency of the pollutants and the interface scattering effect.

[0075] An analysis unit is configured to obtain the dynamic trend of the pulse duration of the pollutants and the deposition ratio of the pollutants in the sediment by analyzing the time series of the emission frequency of the pollutants and the interface scattering effect.

[0076] A tracking unit is configured to determine the persistence of the pollution source according to the deviation of the dynamic trend from a preset pollutant diffusion model, and obtain the tracking results of the potential gradient distribution of the spatial distribution of the pollution source and the emission frequency of the pollutants through weighted average.

[0077] The present application also provides a computer readable storage medium, which stores instructions that are executed by a processor to implement the above method.

[0078] To sum up, the present application uses multi-point synchronous acquisition technology to obtain the spectral data of river sediments, and combines principal component analysis to reduce the dimension of the data and extract the key features of pollutants. These features provide a high-quality data basis for subsequent pollutant species identification and distribution analysis. Through cosine similarity matching with the preset pollutant spectral template, the pollutant species can be accurately determined, which provides a basis for further analysis of pollution source distribution and diffusion. By using back propagation neural network combined with hydrodynamic influence and molecular polarization effect, the diffusion direction and interface polarization intensity of pollutants are predicted, so as to obtain the distribution pattern of pollutants in sediments. This step improves the accuracy of diffusion prediction by considering the interaction between water flow and pollutants and simulating the migration process of pollutants. In the analysis of spatial distribution of pollution sources, clustering analysis method is used combined with the distribution pattern of pollutants, potential gradient and chemical bond amplitude characteristics to obtain the spatial distribution map of pollution sources. This step provides detailed spatial information for pollution source positioning and helps to identify high-risk areas. Finally, the weighted average algorithm integrates various characteristics, combined with the persistence and dynamic trend of pollutants, to accurately calculate the potential gradient distribution and discharge frequency of pollution sources, and finally generate the traceability result of pollution sources. The above technical scheme can realize real-time monitoring and evaluation of pollution sources, and judge the change of pollution sources through continuous dynamic trend analysis. Through the cooperation between each step, a complete pollution source analysis chain is formed, from data acquisition to pollution source positioning, to dynamic monitoring of diffusion trend, realizing high-precision tracing and analysis of pollution sources. The technical scheme solves the problem that the pollution source is difficult to accurately identify and dynamically monitor in the traditional method, and provides strong data support for water environment governance.

[0079] The above merely describes the preferred embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can make equivalent replacements or changes to the technical solutions and inventive concepts of the present application within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for monitoring pollution of river water, characterized by, The method comprises: Step S1: acquiring river sediment spectrum data, extracting a first feature vector according to principal component analysis, and determining a pollutant type according to a cosine similarity of the first feature vector and a preset pollutant spectrum template; Step S2: determining a pollutant diffusion direction and an interface polarization intensity according to the pollutant type, a hydrodynamic effect and a molecular polarization effect through a back propagation neural network to obtain a distribution mode of the pollutant in the sediment; Step S3: obtaining a potential gradient distribution of a pollutant source spatial distribution and a spatial feature of a chemical bond amplitude through clustering analysis according to the distribution mode and a wavelength scattering intensity, matching a signal attenuation rate of a spectrum signal with a preset chemical bond strength model to determine a pollutant emission frequency and an interface scattering effect; Step S4: obtaining a pollutant pulse duration and a dynamic trend of a pollutant deposition ratio in the sediment through analysis of a time sequence of the pollutant emission frequency and the interface scattering effect; Step S5: determining a persistence of the pollutant source according to a deviation of the dynamic trend from a preset pollutant diffusion model, obtaining a backtracking result of the potential gradient distribution of the pollutant source spatial distribution and the pollutant emission frequency through weighted averaging, including: extracting a characteristic parameter according to the persistence of the pollutant source and the chemical bond amplitude, combining the potential gradient distribution and a wavelength modulation characteristic to generate a weighted feature set, processing the weighted feature set through weighted averaging to obtain the potential gradient distribution of the pollutant source spatial distribution, calculating the backtracking result of the pollutant emission frequency, verifying the backtracking result, and generating a final feature matrix of the pollutant source spatial distribution; The first feature vector is extracted through principal component analysis, including: performing standardization processing on an original spectrum data set to obtain a standardized spectrum matrix, decomposing the standardized spectrum matrix through principal component analysis to extract a first feature vector containing a spectrum phase shift, a surface charge density and a time delay distribution, calculating a variance contribution rate of the first feature vector, screening principal components according to the variance contribution rate to obtain a principal component feature set, and performing orthogonalization processing on the principal component feature set to generate an optimized first feature vector for subsequent pollutant type determination; The back propagation neural network is used to determine the pollutant diffusion direction and the interface polarization intensity, including: constructing an input feature set according to the first feature vector, combining the hydrodynamic effect and the molecular polarization effect to generate a training data set, training the training data set through the back propagation neural network to obtain a pollutant diffusion prediction model, calculating the pollutant diffusion direction and the interface polarization intensity through the pollutant diffusion prediction model to generate the distribution mode of the pollutant in the sediment, and verifying the distribution mode to obtain a spatial feature of the pollutant distribution.

2. The method of claim 1, wherein, The river sediment spectrum data is acquired, including: The spectrum data set is obtained by multi-point synchronous acquisition from the surface sediment and deep sediment of the river channel, and the original spectrum data set containing spectrum phase shift, time delay distribution and wavelength absorption coefficient is obtained, the spectrum signal feature parameters are extracted from the original spectrum data set, the initial feature matrix of the spectrum data set is constructed according to the feature parameters, and the optimized spectrum data set is obtained by optimizing the initial feature matrix through the noise reduction algorithm.

3. The method of claim 1, wherein, The pollutant species is determined according to the cosine similarity of the first feature vector and the preset pollutant spectrum template, including: The preset pollutant spectrum template is obtained, the cosine similarity of the first feature vector and the preset pollutant spectrum template is calculated, the cosine similarity is compared with the preset threshold, if the cosine similarity is greater than the preset threshold, the pollutant species is determined, the spectrum feature parameters of the determined pollutant species are extracted, the pollutant feature vector is generated, and the accuracy of the pollutant species is verified according to the matching result of the pollutant feature vector and the preset pollutant spectrum template.

4. The method of claim 1, wherein, The potential gradient distribution of the spatial distribution of the pollution source and the spatial characteristics of the chemical bond amplitude are obtained through clustering analysis, including: The spatial feature parameters are extracted according to the distribution mode of the pollutant in the sediment, the feature data set is generated by combining the time delay distribution and the wavelength scattering intensity, the potential gradient distribution of the spatial distribution of the pollution source is obtained by classifying the feature data set through clustering analysis, the spatial characteristics of the chemical bond amplitude are calculated, and the feature matrix of the spatial distribution of the pollution source is generated by standardizing the potential gradient distribution and the chemical bond amplitude.

5. The method of claim 1, wherein, The signal attenuation rate of the spectrum signal is matched with the preset chemical bond strength model, including: The signal attenuation rate of the spectrum signal is obtained, the preset chemical bond strength model is extracted, the matching degree of the signal attenuation rate and the preset chemical bond strength model is calculated, the matching degree is compared with the preset threshold, if the matching degree is greater than the preset threshold, the pollutant emission frequency and the interface scattering effect are determined, the frequency spectrum analysis is carried out on the pollutant emission frequency, the feature parameters of the interface scattering effect are generated, and the accuracy of the pollutant emission frequency is verified.

6. The method of claim 1, wherein, The potential gradient distribution of the spatial distribution of the pollution source and the tracing result of the pollutant emission frequency are obtained by weighted average, including: The feature parameters are extracted according to the persistence of the pollution source and the chemical bond amplitude, the weighted feature set is generated by combining the potential gradient distribution and the wavelength modulation characteristics, the potential gradient distribution of the spatial distribution of the pollution source is obtained by processing the weighted feature set through weighted average, the tracing result of the pollutant emission frequency is calculated, the tracing result is verified, and the final feature matrix of the spatial distribution of the pollution source is generated.

7. A pollution monitoring system for river water for implementing the method according to any one of claims 1-6, characterized in that, The system comprises: A determination unit is configured to obtain river sediment spectrum data, extract a first feature vector according to principal component analysis, and determine a pollutant species according to the cosine similarity of the first feature vector and a preset pollutant spectrum template; A prediction unit is configured to determine a pollutant diffusion direction and an interface polarization intensity according to the pollutant species, hydrodynamic influence and molecular polarization effect through a back propagation neural network, and obtain a distribution mode of the pollutant in the sediment. The matching unit is configured to obtain the potential gradient distribution and the spatial characteristics of the chemical bond amplitude of the spatial distribution of the pollution source by cluster analysis according to the distribution mode and the wavelength scattering intensity, match the signal attenuation rate of the spectral signal with the preset chemical bond strength model to determine the pollution emission frequency and the interface scattering effect. The analysis unit is configured to obtain the dynamic trend of the pollution pulse duration and the pollution deposition ratio in the sediment by analyzing the time sequence of the pollution emission frequency and the interface scattering effect. The tracing unit is configured to judge the persistence of the pollution source according to the deviation of the dynamic trend from the preset pollution diffusion model, and obtain the tracing result of the potential gradient distribution and the pollution emission frequency of the spatial distribution of the pollution source by weighted average.

8. A computer-readable storage medium having stored thereon instructions, the computer-readable storage medium comprising: The instructions are executed by the processor to implement the method in any one of claims 1-6.

Images