Pollution source identification method and system based on data fusion
Through data fusion technology, combined with conventional water quality, ultraviolet-visible absorption spectrum and three-dimensional fluorescence spectrum data, a pollution source identification model is established, which solves the problems of low traceability efficiency and high misjudgment rate in the existing technology, and realizes the intelligence and accuracy of pollution sources. and efficient identification improve the accuracy and timeliness of traceability.
Patent Information
- Application Number
- CN202110246412.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-03-05
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-03-05
AI Technical Summary
The existing technology relies on manual inspections in the traceability of pollution sources, and lacks automatic comparison methods, resulting in insufficient traceability accuracy and timeliness, and a high misjudgment rate in complex situations.
Using the pollution source identification method based on data fusion, the pollution source samples are collected, pretreated, and conventional water quality, ultraviolet-visible absorption spectrum and three-dimensional fluorescence spectrum data tests are carried out, and characteristic data is extracted and identification models are established in combination with classification algorithms to achieve intelligent and accurate pollution source and efficient discrimination.
It improves the accuracy and timeliness of pollution source traceability, reduces dependence on technical experts, reduces costs, enhances the identification ability in complex emission scenarios, and realizes real-time comparison.
Smart Images

Figure CN113011478B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of environmental supervision, and in particular to a pollution source identification method and system based on data fusion. Background Art
[0002] At present, the source of illegal discharge and leakage by pollutant discharge units mainly relies on manual investigation. Manual investigation usually refers to the investigation of the water quality of the discharge of pollutants from the accident site to the upstream to trace the source of illegal discharge. However, this method is time-consuming and labor-intensive, and it is easy to lose timeliness and has low efficiency.
[0003] In recent years, manual investigation methods assisted by characteristic pollutant databases have emerged. Comparing pollutant data of polluted water, soil or atmosphere with the characteristic pollutant database of pollution sources can narrow the scope of investigation, reduce the workload of investigation, and improve the efficiency of investigation. Previously, Wan Pingyu and others from Beijing University of Chemical Technology proposed a chemical watermark information database of pollution sources including anion types, organic types, metal element types, and fluorescence information for water pollution source tracing. A new type of water pollution emission source database with strong operability is proposed in the related technology. The database contains three sub-databases: a basic information database of pollution sources, a conventional water quality database, and a water quality fingerprint database. These databases are relatively complex and costly, making it difficult to achieve online early warning and tracing.
[0004] In addition, this method still lacks an automatic comparison method and complete system for pollution sources. In actual work, the source of pollution is often determined by manual comparison. This comparison method requires the staff to have strong professional knowledge and experience, the judgment results are highly subjective, and lack scientific and quantitative data support. In complex situations, the misjudgment rate of pollution sources is high and there is still a large lag in the identification of pollution sources. Therefore, it is necessary to develop an automatic comparison method and system for pollution sources to improve the accuracy and real-time nature of pollution traceability.
[0005] Based on data-driven, the use of chemometric analysis methods to establish a discrimination model can achieve intelligent, accurate and efficient discrimination of pollution sources. The pollutant composition of pollution sources is usually complex, and a single indicator has limitations in characterizing pollution sources. For similar pollution sources (such as different companies in the same industry), the accuracy of pollution source identification based on a single indicator is often not high. Data fusion strategies can integrate different pollution information together and can more comprehensively reflect the characteristics of pollution sources. Therefore, identification models based on data fusion often have better recognition performance than models based on single data sources, and can improve the accuracy of pollution source identification. Summary of the invention
[0006] The present invention aims to solve one of the technical problems in the related art at least to a certain extent.
[0007] To this end, an object of the present invention is to propose a pollution source identification method based on data fusion, which is accurate, intelligent, low-cost, highly operational, conducive to large-scale promotion, and of great significance for pollution source tracing.
[0008] Another object of the present invention is to provide a pollution source identification system based on data fusion.
[0009] To achieve the above object, an embodiment of the present invention provides a pollution source identification method based on data fusion, comprising the following steps:
[0010] Collect samples from pollution sources;
[0011] Pre-treating the pollution source sample;
[0012] The pollution index test is carried out on the pre-treated pollution source samples to obtain conventional water quality data, UV-visible absorption spectrum data and three-dimensional fluorescence spectrum data;
[0013] Preprocessing the conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data;
[0014] Performing feature extraction on the pre-processed conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data;
[0015] The characteristic data extracted from the conventional water quality data, the ultraviolet-visible absorption spectrum data and the three-dimensional fluorescence spectrum data are spliced to construct fusion data;
[0016] Establishing a pollution source identification model and training it according to the fused data and classification algorithm;
[0017] Pollution source identification is performed through the trained pollution source identification model.
[0018] In addition, the pollution source identification method based on data fusion according to the above embodiment of the present invention may also have the following additional technical features:
[0019] Furthermore, in one embodiment of the present invention, the pollution source samples include: water samples, soil samples and atmospheric samples, and the pretreatment of the pollution source samples includes: filtering the water samples with a 0.2-10.0 μm filter membrane, dissolving the soil samples with ultrapure water first, and then filtering the soil extract with a 0.2-10.0 μm filter membrane, and dissolving the gas samples in ultrapure water and then filtering them with a 0.2-10.0 μm filter membrane.
[0020] Further, in one embodiment of the present invention, the pollution index test includes conventional water quality, ultraviolet-visible absorption spectrum and three-dimensional fluorescence spectrum;
[0021] The conventional water quality includes but is not limited to pH value, conductivity, chemical oxygen demand, total nitrogen, ammonia nitrogen and total phosphorus;
[0022] The ultraviolet-visible absorption spectrum scanning range is 200-800nm, and the scanning interval is 0.1-10nm;
[0023] The excitation wavelength scanning range of the three-dimensional fluorescence spectrum is 200-600nm, the emission wavelength scanning range is 230-700nm, and the scanning interval is 1-10nm.
[0024] Furthermore, in one embodiment of the present invention, the conventional water quality data, the ultraviolet-visible absorption spectrum data and the three-dimensional fluorescence spectrum data are preprocessed, including:
[0025] Preprocessing of conventional water quality data: using the Laida criterion to remove abnormal data;
[0026] Preprocessing of UV-visible absorption spectra: removing invalid data from the spectral data, and then performing standard normal transformation on the spectral data;
[0027] Preprocessing of the three-dimensional fluorescence spectrum: The fluorescence intensity of the original fluorescence fingerprint is converted into Raman units (RU) by integrating the Raman scattering intensity of ultrapure water with an excitation wavelength of 350 nm and an emission wavelength of 370 to 430 nm.
[0028] Further, in one embodiment of the present invention, feature extraction is performed on the preprocessed test data, including feature extraction of conventional water quality data, ultraviolet-visible absorption spectrum data, and three-dimensional fluorescence spectrum data;
[0029] The feature extraction methods of the conventional water quality data and the ultraviolet-visible absorption spectrum data include principal component analysis, non-negative matrix decomposition, and independent component analysis;
[0030] The feature extraction of the three-dimensional fluorescence spectrum data is to extract the main fluorescence components of the three-dimensional fluorescence spectrum using parallel factor analysis.
[0031] Furthermore, in one embodiment of the present invention, the classification algorithm includes but is not limited to partial least squares resolution analysis, support vector machine, and K nearest neighbor node algorithm.
[0032] Furthermore, in one embodiment of the present invention, a pollution source identification model is established and trained based on the fusion data and the classification algorithm, further comprising:
[0033] Model initialization: 75% to 95% of the sample data are selected as the training set, the pollution source identification model is established by the cross-validation method, and the optimal number of latent variables is selected according to the principle of minimizing the cross-validation error;
[0034] Model training: setting the number of variables to the optimal number of potential variables to fit the pollution source identification model again;
[0035] Model prediction: The remaining 5% to 25% of the sample data are predicted through the fitted pollution source identification model, and the model performance is evaluated based on the prediction results, wherein the evaluation parameters are the sensitivity, specificity, precision and accuracy of the pollution source identification model.
[0036] To achieve the above object, another embodiment of the present invention proposes a pollution source identification method system based on data fusion, comprising:
[0037] Sampling module, sample pretreatment module, injection module, analysis and testing module, data transmission module, display module and system control module;
[0038] The sampling module is used to collect pollution source samples;
[0039] The sample pretreatment module is used to pretreatment the pollution source sample;
[0040] The sample injection module is used to transport the pre-treated sample to the analysis and testing module;
[0041] The analysis and testing module is used to perform pollution index testing on the pre-treated pollution source samples to obtain conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data;
[0042] The data transmission module is used for data transmission between modules;
[0043] The display module is used to display data;
[0044] The system control module is used to embed the pollution source identification model and perform control, obtain conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data through the data transmission module, and use the configuration software to display the results in real time on the display module, and transmit the conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data collected in the configuration software to the pollution source identification model. The pollution source identification model distinguishes the sample category and then returns the distinction result to the display module to display the distinction result.
[0045] Furthermore, in one embodiment of the present invention, the analysis and testing module includes but is not limited to an online pH meter, an online conductivity meter, an online COD analyzer, an online ammonia nitrogen analyzer, an online total nitrogen meter and an online total phosphorus analyzer, a UV-visible absorption spectrophotometer and a fluorescence spectrophotometer.
[0046] The pollution source identification method and system based on data fusion according to the embodiment of the present invention have the following advantages:
[0047] (1) Data fusion technology was applied to pollution source identification and a pollution source identification model was established, which broke the reliance on technical experts and improved the efficiency of source tracing; the accuracy of source tracing was improved, especially in complex pollution emission scenarios;
[0048] (2) The pollution index analysis and testing methods selected by data fusion are mature and reliable, easy to operate, low-cost, and rich in information, which is conducive to large-scale promotion;
[0049] (3) Manual comparison usually takes several hours. The pollution source identification system constructed by the present invention can perform real-time comparison, which enhances the timeliness of pollution tracing and is of great significance to pollution tracing.
[0050] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:
[0052] Figure 1 A flow chart of a pollution source identification method based on data fusion according to an embodiment of the present invention;
[0053] Figure 2 Establishing a flow chart for a pollution source identification model according to an embodiment of the present invention;
[0054] Figure 3 A schematic diagram showing the variation of the cross-validation error rate with the number of potential variables according to an embodiment of the present invention;
[0055] Figure 4 is the calculated response value of the model for category 1 (ZQ) in the training and prediction phase according to one embodiment of the present invention;
[0056] Figure 5 The figure is a schematic diagram of the structure of a pollution source identification system based on data fusion according to an embodiment of the present invention. DETAILED DESCRIPTION
[0057] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.
[0058] The following describes a pollution source identification method and system based on data fusion according to an embodiment of the present invention with reference to the accompanying drawings.
[0059] First, a pollution source identification method based on data fusion proposed according to an embodiment of the present invention will be described with reference to the accompanying drawings.
[0060] Figure 1 The figure is a flow chart of a pollution source identification method based on data fusion according to an embodiment of the present invention.
[0061] Figure 2 A flow chart is established for a pollution source identification model according to an embodiment of the present invention.
[0062] like Figure 1 and Figure 2 As shown, the pollution source identification method based on data fusion includes the following steps:
[0063] Step S1, collecting pollution source samples.
[0064] Step S2, pre-treating the pollution source sample.
[0065] Pollution source samples include: water samples, soil samples and air samples. The pretreatment of pollution source samples includes: filtering water samples with a 0.2-10.0μm filter membrane, dissolving soil samples with ultrapure water first, and then filtering the soil extract with a 0.2-10.0μm filter membrane, and dissolving gas samples in ultrapure water and then filtering them with a 0.2-10.0μm filter membrane.
[0066] Step S3, performing pollution index testing on the pretreated pollution source samples to obtain conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data.
[0067] Furthermore, pollution index analysis tests include, but are not limited to, conventional water quality, UV-visible absorption spectroscopy, and three-dimensional fluorescence spectroscopy.
[0068] Conventional water quality includes but is not limited to pH value, conductivity, chemical oxygen demand, total nitrogen, ammonia nitrogen and total phosphorus; the ultraviolet-visible absorption spectrum scanning range is 200-800nm, and the scanning interval is 0.1-10nm; the excitation wavelength scanning range of the three-dimensional fluorescence spectrum is 200-600nm, the emission wavelength scanning range is 230-700nm, and the scanning interval is 1-10nm.
[0069] Step S4, preprocessing the conventional water quality data, the ultraviolet-visible absorption spectrum data and the three-dimensional fluorescence spectrum data.
[0070] Data preprocessing includes the preprocessing of conventional water quality, UV-visible absorption spectrum and three-dimensional fluorescence spectrum data.
[0071] Conventional water quality data preprocessing refers to the use of the Laida criterion (3σ criterion) to eliminate abnormal data.
[0072] UV-visible absorption spectrum preprocessing refers to removing the part of the spectrum data without valid data (that is, basically all solvent background absorption), and then performing a standard normal transformation on the spectrum to reduce the influence of scattering. The calculation formula is:
[0073]
[0074] Among them, S SNV is the transformed data, S k is the original data, is the mean of all wavelength points of the original spectrum, and m is the number of wavelength points of the spectrum.
[0075] Three-dimensional fluorescence spectrum preprocessing refers to converting the fluorescence intensity of the original fluorescence fingerprint into Raman units (RU) by integrating the Raman scattering intensity of ultrapure water with an excitation wavelength of 350 nm and an emission wavelength of 370 to 430 nm.
[0076] Step S5, extracting features from the pre-processed conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data.
[0077] Data feature extraction includes feature extraction of conventional water quality data, UV-visible absorption spectrum and three-dimensional fluorescence spectrum data.
[0078] The feature extraction methods of conventional water quality data and UV-visible absorption spectra include but are not limited to principal component analysis, non-negative matrix decomposition, and independent component analysis. As a preferred solution, the present invention adopts principal component analysis, and its basic principle is:
[0079]
[0080] Where X represents conventional water quality data or UV-visible absorption spectrum data, F represents the number of principal components, and t f represents the score vector of the fth principal component, p f represents the load vector of the fth principal element. The score vector t f is the coordinate of the sample on the fth principal element, that is, the coordinate of the new variable. Load vector p f It represents the correlation coefficient between the original variable and the fth principal component. The larger the load, the more fully the principal component explains the variable. F score vectors constitute the score matrix T, and F load vectors constitute the load matrix P. Multiplying T and P gives the part simulated by the F principal components. E is the model residual.
[0081] Feature extraction of three-dimensional fluorescence spectrum data refers to the use of parallel factor analysis to obtain the main fluorescent components of the three-dimensional fluorescence spectrum. Its basic principle is:
[0082]
[0083] Among them, x ijk is the fluorescence intensity of the i-th sample at emission wavelength j and excitation wavelength k; F represents the factor number; a if , b jf 、c kf Represent the elements in the load matrices A, B, and C respectively; ε ijk is the model residual, which represents the part that cannot be explained by the model.
[0084] Step S6, combining the characteristic data extracted from the conventional water quality data, the ultraviolet-visible absorption spectrum data and the three-dimensional fluorescence spectrum data to construct fused data.
[0085] Fusion data refers to the splicing of characteristic data matrices extracted from conventional water quality data, UV-visible absorption spectrum data, and three-dimensional fluorescence spectrum data into a new matrix. The formula is:
[0086] F=[A,B,C]
[0087] Where F is the fused data matrix, A, B, and C are the characteristic data matrices extracted from conventional water quality data, UV-visible absorption spectrum data, and three-dimensional fluorescence spectrum data, respectively.
[0088] Step S7, establishing a pollution source identification model based on the fused data and classification algorithm and performing training.
[0089] Step S8: identifying pollution sources using the trained pollution source identification model.
[0090] Establishing a pollution source identification model refers to establishing a classification model based on feature fusion data using fusion data and classification algorithms. Classification algorithms include but are not limited to partial least squares resolution analysis (PLS-DA), support vector machine (SVM), and K nearest neighbor (KNN) node algorithm. As a preferred solution, the present invention uses partial least squares resolution analysis as the classification algorithm of the identification model. The process of establishing the identification model includes model initialization, training, prediction, and performance evaluation.
[0091] Model initialization refers to selecting 75% to 95% of the sample data as the training set, using the cross-validation method to build the model, and selecting the optimal number of latent variables based on the principle of minimizing the cross-validation error.
[0092] Model training means setting the number of latent variables to the optimal number of latent variables and fitting the model again.
[0093] Model prediction refers to using the trained model to predict the remaining 5% to 25% of the sample set and evaluating the model performance based on the prediction results.
[0094] Model performance evaluation parameters include sensitivity (sn), specificity (sp), precision (pr) and accuracy (Acc). sn, sp and pr are single-category performance parameters.
[0095] Sensitivity refers to the ability of a classifier to correctly identify a certain category. Suppose there are two categories, category 1 is positive and category 2 is negative. Taking category 1 as an example, the sensitivity of category 1 describes the proportion of samples whose true value is positive that the model predicts correctly. The calculation formula is:
[0096] sn=TPN / (TPN+FNN)
[0097] TPN represents the number of samples whose true value is positive and whose classification result is also positive. FNN represents the number of samples whose true value is positive but whose classification result is negative, that is, the number of false negative samples.
[0098] Specificity indicates the ability of the classifier to reject samples of another category (category 2), that is, the ability of the classifier to correctly identify negative samples. It describes the proportion of correct predictions made by the model among all samples whose true values are negative. The calculation formula is:
[0099] sp=TNN / (FPN+TNN)
[0100] Among them, TNN represents the number of samples whose true value is negative and whose classification result is also negative. FPN represents the number of samples whose true value is negative but whose classification result is positive, that is, the number of false positive samples.
[0101] Precision indicates the ability of a classifier to avoid misclassifying samples of a certain category. Taking category 1 as an example, precision describes the proportion of samples with positive prediction values that the model predicts correctly. The calculation formula is:
[0102] pr=TPN / (TPN+FPN)
[0103] The accuracy rate describes the ratio of correctly classified samples to the total number of classified samples. It does not consider any classification performance information about a single class. For the above categories 1 and 2, the accuracy rate calculation formula is:
[0104] Acc=(TPN+TNN) / (TPN+FNN+TNN+FPN)
[0105] The present invention is further described below in conjunction with specific embodiments and the accompanying drawings.
[0106] 1) Pollution source sample collection
[0107] A and B are two pollution sources, and 30 water samples are collected from each of them.
[0108] 2) Sample pretreatment
[0109] The samples were filtered through a 0.45 μm filter membrane.
[0110] 3) Sample analysis and testing
[0111] The samples were tested for pH, conductivity, chemical oxygen demand, ammonia nitrogen, total nitrogen and total phosphorus. The results showed that there was no significant difference in the conventional water quality between A and B: pH was between 6 and 8, conductivity was between 900 and 1800 μS / cm, and total phosphorus was between 1 and 11 mg / L. The chemical oxygen demand of A was between 150 and 1300 mg / L, and that of B was between 70 and 600 mg / L. The ammonia nitrogen of A was between 50 and 150 mg / L, and that of B was between 30 and 130 mg / L. The total nitrogen of A was between 60 and 130 mg / L, and that of B was between 20 and 110 mg / L.
[0112] The UV-visible absorption spectrum and three-dimensional fluorescence spectrum of the test samples were measured. The scanning range of the UV-visible absorption spectrum was 200-800nm, and the scanning interval was 0.2nm. The results showed that the UV-visible absorption spectra of A and B basically showed a single exponential downward trend, and there was a weak absorption peak near 260-280nm, and the difference was not obvious.
[0113] The three-dimensional fluorescence spectrum of the test sample has an excitation wavelength of 220-600nm, an emission wavelength of 230-650nm, and a scanning interval of 5nm. The typical three-dimensional fluorescence spectra of A and B are both at the excitation / emission wavelength (denoted as E x / E m )=225 / 340nm and E x / E m = There are two fluorescence peaks near 275 / 340nm, and the difference is not significant.
[0114] 4) Data preprocessing
[0115] No abnormal samples were found when analyzing the conventional water quality data using the Laida criterion, so the conventional water quality data of all samples can be used for modeling.
[0116] The original UV-visible absorption spectrum has almost zero absorbance after 500nm, so this part of the spectrum is removed. Then the remaining spectrum is subjected to standard normal transformation.
[0117] The three-dimensional fluorescence spectrum uses the integration of the Raman scattering intensity between the ultrapure water excitation wavelength of 350nm and the emission wavelength of 370-430nm to convert the fluorescence intensity of the original three-dimensional spectrum into Raman units (RU).
[0118] 5) Feature extraction
[0119] The principal component analysis of conventional water quality data shows that the cumulative variance contribution rate of the first three principal components has reached 95.53%, which contains most of the information of the original data. Therefore, three principal components PC1 to PC3 are extracted from the six conventional water quality data.
[0120] The principal component analysis of the preprocessed UV-visible absorption spectrum shows that the cumulative variance contribution rate of the first six principal components reaches 98.78%, which contains most of the information of the original data. Therefore, six principal components, UVPC1 to UVPC6, are also extracted from the UV-visible absorption spectrum data.
[0121] The preprocessed three-dimensional fluorescence spectrum data were subjected to parallel factor decomposition to obtain six main fluorescence components F1-F6.
[0122] 6) Data Fusion
[0123] PC1~PC3, UVPC1~UVPC6 and F1~F6 are spliced together to construct the fused data FD=[PC1, PC2, PC3, UVPC1, UVPC2, UVPC3, UVPC4, UVPC5, UVPC6, F1, F2, F3, F4, F5, F6].
[0124] 7) Establish a pollution source identification model
[0125] a) Consider samples of A as category 1 and samples of B as category 2. Take 23 samples of A and 23 samples of B as training samples, and the rest as prediction samples;
[0126] b) Import training samples and perform data normalization;
[0127] c) Preliminarily establish the PLS-DA classification model and select the optimal number of variables based on the principle of minimizing the cross-validation error rate. Figure 3 As shown in the figure, when the number of latent variables is 1 to 5, the cross-validation error rate is the minimum value of 0. Theoretically, the number of latent variables can be 1 to 5. However, a low number of latent variables may contain less information, which may reduce the model performance during prediction; a large number of latent variables may have more redundant information, which increases noise during prediction. As a preferred option, the optimal number of latent variables in this case is selected as the intermediate value 3;
[0128] d) Set the number of latent variables to 3, refit the model and save the model;
[0129] e) Import prediction samples and perform data normalization;
[0130] f) Using the established model to classify and identify the predicted samples;
[0131] g) Check the model prediction performance. The results show that the established recognition model has perfect prediction performance for A and B. No sample was misclassified during the training and prediction stages. Figure 4 The single-category performance parameters of sensitivity, specificity, and precision are all 1, and the overall performance parameter accuracy is also 1.
[0132] The pollution source identification method based on data fusion proposed in the embodiment of the present invention includes sample collection, sample analysis and testing, data preprocessing, feature extraction, data fusion, identification model establishment and pollution source identification system construction. Conventional water quality, ultraviolet-visible absorption spectrum and three-dimensional fluorescence spectrum are fused and applied to pollution source identification. Compared with traditional pollution source identification methods, this method is accurate, intelligent, low-cost, and highly operational, which is conducive to large-scale promotion and has important significance for pollution source tracing.
[0133] Next, a pollution source identification system based on data fusion proposed according to an embodiment of the present invention will be described with reference to the accompanying drawings.
[0134] Figure 5 The figure is a schematic diagram of the structure of a pollution source identification system based on data fusion according to an embodiment of the present invention.
[0135] like Figure 5 As shown, the pollution source identification system based on data fusion includes: a sampling module, a sample pretreatment module, a sampling module, an analysis and testing module, a data transmission module, a display module and a system control module;
[0136] The sampling module is used to collect samples from pollution sources;
[0137] The sample pretreatment module is used to pretreat the pollution source samples;
[0138] The sampling module is used to transport the pre-treated sample to the analysis and testing module;
[0139] The analysis and testing module is used to test the pollution index of the pre-treated pollution source samples to obtain conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data;
[0140] The data transmission module is used for data transmission between modules;
[0141] The display module is used to display data;
[0142] The system control module is used to embed the pollution source identification model and perform control. It obtains conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data through the data transmission module, and uses the configuration software to display the results in real time on the display module. The conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data collected in the configuration software are transmitted to the pollution source identification model. The pollution source identification model distinguishes the sample category of the pollution source and then returns the distinction result to the display module to display the distinction result.
[0143] The analysis and testing modules include, but are not limited to, an online pH meter, an online conductivity meter, an online COD analyzer, an online ammonia nitrogen analyzer, an online total nitrogen analyzer, an online total phosphorus analyzer, an ultraviolet-visible absorption spectrophotometer, and a fluorescence spectrophotometer.
[0144] Building a pollution source identification system includes the following steps:
[0145] 1) connecting a sampling module, a sample pretreatment module, a sample injection module, an analysis and testing module, a data transmission module, a display module and a system control module together, wherein the analysis and testing module comprises an online pH meter, an online conductivity meter, an online COD analyzer, an online ammonia nitrogen analyzer, an online total nitrogen meter and an online total phosphorus analyzer, an ultraviolet-visible absorption spectrophotometer and a fluorescence spectrophotometer;
[0146] 2) Embed the pollution source identification model into the system control module;
[0147] 3) The system control module starts the sample collection module, sends the sample to the analysis and testing module through the sampling module, and then transmits the measured data to the system control model through the data transmission module, and uses the configuration software MCGS to display the results in real time on the display module;
[0148] 4) Use OPC technology to exchange data between the pollution source identification model and the MCGS industrial computer configuration software. The water quality data collected in the configuration software is transmitted to the pollution source identification model, which analyzes the sample category and then returns the discrimination result to the display module to display the discrimination result;
[0149] 5) Repeat steps 3) and 4) continuously to achieve real-time comparison of pollution sources.
[0150] It should be noted that the above explanation of the method embodiment is also applicable to the system of this embodiment and will not be repeated here.
[0151] The pollution source identification system based on data fusion proposed in the embodiment of the present invention includes sample collection, sample analysis and testing, data preprocessing, feature extraction, data fusion, identification model establishment and pollution source identification system construction. Conventional water quality, ultraviolet-visible absorption spectrum and three-dimensional fluorescence spectrum are fused and applied to pollution source identification. Compared with traditional pollution source identification methods, this method is accurate, intelligent, low-cost, and highly operational, which is conducive to large-scale promotion and has important significance for pollution source tracing.
[0152] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0153] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.
[0154] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.
Claims
1. A pollution source identification method based on data fusion, characterized in that: The following steps are involved: Collect samples from pollution sources; Pre-treating the pollution source sample; The pollution index test is carried out on the pre-treated pollution source samples to obtain conventional water quality data, UV-visible absorption spectrum data and three-dimensional fluorescence spectrum data; Preprocessing the conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data; Preprocessing the conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data includes: Preprocessing of conventional water quality data: using the Laida criterion to remove abnormal data; Preprocessing of UV-visible absorption spectrum: remove the part without valid data in the spectrum data, and then perform standard normal transformation on the spectrum to reduce the influence of scattering. The calculation formula is: Among them, S SNV is the transformed data, S k is the original data, is the mean of all wavelength points of the original spectrum, and m is the number of wavelength points of the spectrum; Preprocessing of the three-dimensional fluorescence spectrum: The fluorescence intensity of the original fluorescence fingerprint was converted into Raman units (RU) using the integral of the Raman scattering intensity of ultrapure water with an excitation wavelength of 350 nm and an emission wavelength of 370 to 430 nm; Performing feature extraction on the pre-processed conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data; The feature extraction method of the conventional water quality data and the ultraviolet-visible absorption spectrum data adopts principal component analysis; The basic principle of principal component analysis is: Where X represents conventional water quality data or UV-visible absorption spectrum data, F represents the number of principal components, and t f represents the score vector of the fth principal component, p f represents the load vector of the fth principal element, and the score vector t f is the coordinate of the sample on the fth principal element, the load vector p f It represents the correlation coefficient between the original variable and the fth principal component. The larger the load, the more fully the principal component explains the variable. The F score vectors constitute the score matrix T, and the F load vectors constitute the load matrix P. The multiplication of T and P gives the part simulated by the F principal components. E is the model residual; Feature extraction of three-dimensional fluorescence spectrum data refers to the use of parallel factor analysis to obtain the main fluorescent components of the three-dimensional fluorescence spectrum. Its basic principle is: Among them, x ijk is the fluorescence intensity of the i-th sample at emission wavelength j and excitation wavelength k, F represents the factor number, a if 、b jf 、c kf Represent the elements in the load matrices A, B, and C respectively, ε ijk is the model residual, which represents the part that cannot be explained by the model; The characteristic data extracted from the conventional water quality data, the ultraviolet-visible absorption spectrum data and the three-dimensional fluorescence spectrum data are spliced to construct fusion data; Establishing a pollution source identification model and training it according to the fused data and classification algorithm; Identify pollution sources through the trained pollution source identification model; The fusion data refers to splicing the characteristic data matrix extracted from the conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data into a new matrix, and the formula is: F = [A, B, C]; Wherein, F is a fused data matrix, and A, B, and C are characteristic data matrices extracted from the conventional water quality data, ultraviolet-visible absorption spectrum data, and three-dimensional fluorescence spectrum data, respectively.
2. The method according to claim 1, characterized in that The pollution source samples include: water samples, soil samples and air samples. The pretreatment of the pollution source samples includes: filtering the water samples with a 0.2-10.0 μm filter membrane, dissolving the soil samples with ultrapure water first, and then filtering the soil leachate with a 0.2-10.0 μm filter membrane, and dissolving the gas samples in ultrapure water and then filtering them with a 0.2-10.0 μm filter membrane.
3. The method according to claim 1, characterized in that The pollution index test includes conventional water quality, ultraviolet-visible absorption spectrum and three-dimensional fluorescence spectrum; The conventional water quality includes pH value, conductivity, chemical oxygen demand, total nitrogen, ammonia nitrogen and total phosphorus; The ultraviolet-visible absorption spectrum scanning range is 200-800nm, and the scanning interval is 0.1-10nm; The excitation wavelength scanning range of the three-dimensional fluorescence spectrum is 200-600nm, the emission wavelength scanning range is 230-700nm, and the scanning interval is 1-10nm.
4. The method according to claim 1, characterized in that: The classification algorithms include partial least squares resolution analysis, support vector machine, and K nearest neighbor node algorithm.
5. The method according to claim 1, characterized in that Establishing a pollution source identification model and training it according to the fusion data and classification algorithm further includes: Model initialization: 75% to 95% of the sample data are selected as the training set, the pollution source identification model is established by the cross-validation method, and the optimal number of latent variables is selected according to the principle of minimizing the cross-validation error; Model training: setting the number of variables to the optimal number of potential variables to fit the pollution source identification model again; Model prediction: The remaining 5% to 25% of the sample data are predicted through the fitted pollution source identification model, and the model performance is evaluated based on the prediction results, wherein the evaluation parameters are the sensitivity, specificity, precision and accuracy of the pollution source identification model.
6. A pollution source identification system based on data fusion, the system is used to execute the method according to any one of claims 1 to 5, characterized in that: include: Sampling module, sample pretreatment module, injection module, analysis and testing module, data transmission module, display module and system control module; The sampling module is used to collect pollution source samples; The sample pretreatment module is used to pretreatment the pollution source sample; The sample injection module is used to transport the pre-treated sample to the analysis and testing module; The analysis and testing module is used to perform pollution index testing on the pre-treated pollution source samples to obtain conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data; The data transmission module is used for data transmission between modules; The display module is used to display data; The system control module is used to embed the pollution source identification model and perform control, obtain conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data through the data transmission module, and use the configuration software to display the results in real time on the display module, and transmit the conventional water quality data, ultraviolet-visible absorption spectrum data and three-dimensional fluorescence spectrum data collected in the configuration software to the pollution source identification model. The pollution source identification model distinguishes the sample category and then returns the distinction result to the display module to display the distinction result.
7. The system according to claim 6, characterized in that The analysis and testing module comprises an online pH meter, an online conductivity meter, an online COD analyzer, an online ammonia nitrogen analyzer, an online total nitrogen analyzer, an online total phosphorus analyzer, an ultraviolet-visible absorption spectrophotometer and a fluorescence spectrophotometer.
Citation Information
Patent Citations
High-risk pollution source classification forecasting method based on principal component analysis and random forest
CN107480839A
Method for tracing pollutants by combining water quality parameters with three-dimensional fluorescence spectrum
CN109470667A
A pollution source abnormal data identification method based on a deep learning algorithm
CN109711547A
Water pollution discharge source database and establishment method thereof
CN110083585A
KLXS multi-model fusion method and system based on HRRP target recognition
CN111222575A
