Industrial park VOCs tracing method based on fingerprint spectrum and data driving

Abstract

The invention discloses an industrial park VOCs traceability method based on fingerprint spectrum and data driving. The method comprises the steps of establishment of a park enterprise emission VOCs source spectrum library, establishment of an emission scene library, establishment of a meteorological scene library, establishment of a meteorological-emission pollution traceability library, establishment of a meteorological-emission pollution traceability model and pollution traceability. A traditional fingerprint spectrum comparison method and a data-driven traceability model are combined, a suspicious exhaust port can be quickly and accurately found out when a pollutant standard exceeding event occurs, accurate and quick traceability of small-scale VOCs in a park is realized, the requirements of high accuracy and high timeliness of a VOCs traceability result in an industrial park can be met, and the traceability of the VOCs in the park is improved. And the pollution control capability of the industrial park is greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of atmospheric environmental protection technology, and in particular to a VOCs tracing method for an industrial park based on fingerprint and data-driven. Background Art

[0002] Chemical parks are modern comprehensive industrial parks that integrate chemical production, scientific research, and life, and are an important part of the chemical industry. Due to the high density of enterprises in chemical parks, the integration of a large number of chemical production facilities and equipment, and the complex temporal and spatial distribution characteristics of pollution emission sources and pollutant components, it is difficult to trace the source after pollution occurs, making the air pollution control and risk management of chemical parks face challenges. At the same time, when pollutants exceed the alarm or there are complaints about odors, there is a phenomenon of mutual buck-passing between enterprises, which makes it difficult for management departments to carry out law enforcement inspections and control.

[0003] At present, for the smaller-scale atmospheric pollution source tracing of parks or enterprises, cruise monitoring, receptor model method and source model method are mainly used. Cruise monitoring is mainly used to obtain high-value points, but due to the certain distance between the cruise points and the production area of ​​the enterprise, it is difficult to accurately locate the pollution source. The receptor model method includes the chemical mass balance model method (CMB), the positive definite matrix factorization method (PMF) and the principal component analysis method (PCA). Although this method can better distinguish the emission source contribution rate of known pollutants, it cannot determine the spatial location of the pollution source. At the same time, it is difficult to obtain the actual pollutant concentration distribution in the region with limited receptor points. In addition, the integrity of the pollution source components has a significant impact on the accuracy of the receptor model method, which greatly limits its application in small-scale source tracing. The source model methods mainly include FLEXPART, AERMOD, CALPUFF, etc. These models can predict the diffusion trend of pollutants and the emission contribution of pollution sources. However, the source model method has the problem of long model running time, resulting in low timeliness of the source tracing results and inability to quickly locate the pollution source. As the control requirements of industrial parks continue to increase, the requirements for accuracy and speed of pollution source tracing in industrial parks have also increased. However, the current model results are lagging and cannot meet the needs of rapid pollution source tracing in industrial parks. Therefore, this application proposes a VOCs source tracing method for industrial parks based on fingerprints and data-driven to solve this problem. Summary of the invention

[0004] The purpose of this application is to provide a VOCs source tracing method for industrial parks based on fingerprint maps and data-driven, which is used to solve the problems of low accuracy and slow speed in existing pollution source tracing in industrial parks.

[0005] The present invention establishes grouped emission scenarios based on the VOCs emission inventory of enterprises in the park and the VOCs source spectrum of the enterprises, and establishes a meteorological scenario library through equal interval division based on the meteorological data of the park over the years. The diffusion model is used to simulate the diffusion trend and emission impact of enterprise pollutants under different scenarios, and the meteorological scenarios and emission impacts are matched one by one by grouping to obtain a meteorological-emission pollution source tracing library. The data in the pollution source tracing library is used as a training sample, and the ANN artificial neural network is used for training to obtain a meteorological-emission pollution source tracing model. Based on the source tracing model, potential polluting enterprises can be obtained. The comparison of the VOCs spectra of polluting enterprises and receptor points can accurately locate the organized outlets and unorganized surface sources that cause excessive emissions, thereby improving the timeliness and accuracy of VOCs source tracing in industrial parks.

[0006] A VOCs source tracing method for industrial parks based on fingerprint and data-driven, specifically including the following steps:

[0007] S1, Establishment of VOCs source spectrum library of park enterprises: Through the collection and composition analysis of VOCs emitted from the organized / unorganized outlets of enterprises, the test results are normalized and analyzed to form the VOCs source spectrum of park enterprises;

[0008] S2, establishment of emission scenario library: collect VOCs emission inventories of enterprises in the park, distribute the emission inventories in time and space based on the location of enterprises in the park and the production situation throughout the year, and build an emission scenario library of different OVOCs components;

[0009] The VOCs emission inventory must be accurate to the enterprise, and the emission data must be accurate to the component; the source spectrum includes 115 VOCs components, which are divided into 29 alkanes, 11 alkenes, 17 aromatic hydrocarbons, 21 OVOCs, 35 halogenated hydrocarbons, carbon disulfide, and acetylene.

[0010] The emission inventory should include organized and unorganized emissions. Organized emissions from enterprises shall be treated as point sources, and unorganized emissions shall be treated as surface sources.

[0011] S3, establishment of meteorological scenario library: according to the existing meteorological station monitoring data and WRF simulated meteorological field data of the park, the range of meteorological parameters of the park is determined. The meteorological parameters include wind speed, wind direction, temperature, cloud cover, precipitation, and air pressure. The wind direction is set to 0-360°, where 0° and 360° are both due north. The cloud cover is in a ten-point system, ranging from 0 to 10. The intervals between each parameter are divided according to the range of meteorological parameters of the park over the years. The divided meteorological parameters are arranged and combined to obtain different meteorological scenarios.

[0012] S4, establishment of a meteorological-emission pollution source tracing library: combine emission scenarios and meteorological scenarios to form multiple meteorological-emission scenarios, simulate the contribution of park enterprise emissions to the concentration of each component of VOCs at the receptor point under different meteorological-emission scenarios through diffusion models (existing technology), summarize the simulation results under each meteorological-emission scenario, and correspond the meteorological scenarios with the model simulation results to obtain meteorological-emission source tracing samples, which are compiled to form a meteorological-emission source tracing library;

[0013] The meteorological-emission scenario refers to multiple component emission scenarios corresponding to multiple meteorological scenarios, so each meteorological scenario also corresponds to multiple model simulation results. The models include but are not limited to AERMOD and CALPUFF models. When the model is applied, localization processing of model parameters such as terrain and land use type is required.

[0014] S5, Construction of meteorological-emission pollution source tracing model: The data set in the meteorological-emission source tracing library is used as training samples, meteorological conditions and pollution source names are used as input, and the pollution source contribution ratio is used as output. The meteorological-emission source tracing model is obtained through machine learning training. The learning method is artificial neural network ANN. The accuracy of the meteorological-emission source tracing model is improved by learning and adjusting parameters;

[0015] The name of the pollution source is the name of the enterprise.

[0016] The ANN neural network structure consists of three parts: input layer, hidden layer and output layer. Each layer is composed of multiple neurons, each neuron is interconnected with all neurons in the previous layer, and each connection corresponds to a weight. In the three layers of the neural network, the input layer is used to receive and input data, the hidden layer processes the data, and the output layer generates and outputs the generated results.

[0017] When learning and adjusting the parameters, the parameters are adjusted by minimizing the mean square error RMSE.

[0018]

[0019] Among them, y i is the true value of the i-th sample point, y' i is the predicted value.

[0020] S6, Pollution source tracing: When the VOCs concentration at the receptor point exceeds the standard, the real-time meteorological data and the characteristic components of VOCs at the time of exceeding the standard are input into the meteorological-emission source tracing model, which can obtain the concentration contribution values ​​of various source emissions to the characteristic components of the receptor point and quickly find potential polluting enterprises. The VOCs source spectrum of the potential polluting enterprise is retrieved and the cosine similarity analysis is performed with the VOCs spectrum of the receptor point at the time of exceeding the standard, and finally the discharge outlet of the suspicious enterprise is locked, realizing the rapid and precise source tracing of VOCs in the park;

[0021] The real-time meteorological data includes wind direction, wind speed, and temperature, and the characteristic components are components that account for more than 10% of the VOCs at the receptor point.

[0022] The advantages of the present invention are: combining the traditional fingerprint comparison method with the data-driven tracing model, it is possible to quickly and accurately find suspicious outlets when pollutants exceed the standard, and to achieve precise and rapid tracing of small and medium-scale VOCs in the industrial park. It can meet the requirements of industrial parks for high accuracy and timeliness of VOCs tracing results, and greatly improve the pollution control capabilities of industrial parks. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 It is a flow chart of the technical solution of the present invention.

[0025] Figure 2 This is a schematic diagram of the enterprise source spectrum of the present invention.

[0026] Figure 3 This is a schematic diagram of the emission scenario library of the present invention.

[0027] Figure 4 This is a schematic diagram of the meteorological-emission traceability library of the present invention.

[0028] Figure 5 Schematic diagram of the traceability result of this embodiment. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] Example: Taking an industrial park as an example, a VOCs tracing method for an industrial park based on fingerprint and data-driven, such as Figure 1 As shown, the following steps are included:

[0031] S1, Establishment of VOCs source spectrum library of park enterprises: Investigate the production links of park enterprises, collect organized / unorganized emissions in the production links, and detect 115 VOCs components, including 29 alkanes, 11 alkenes, 17 aromatic hydrocarbons, 21 OVOCs, 35 halogenated hydrocarbons, carbon disulfide, and acetylene. Use Suma tank sampling and use gas chromatography-mass spectrometry (GC-MS) for sample analysis. The detected VOCs component data are normalized to obtain the percentage of each component, and the VOCs emission source spectrum of different outlets of the enterprise is obtained.

[0032] S2, establishment of emission scenario library: according to the location of the enterprise, emission inventory, and the monthly production situation of the enterprise, the inventory is distributed in time and space, and multiple emission scenarios are created by grouping. Each emission scenario includes the location of the enterprise, the name of the component, and the emission source strength of the component.

[0033] Furthermore, the number of enterprises in step S2 is 218.

[0034] Furthermore, the emission inventory in step S2 is specific to VOCs components. The enterprise VOCs emission inventory in the embodiment includes a total of 47 components.

[0035] Furthermore, the temporal variation of the source intensity of each component in step S2 is related to the monthly production situation of the enterprise, and the basis for time allocation includes but is not limited to the monthly production situation of the enterprise.

[0036] Furthermore, the emission scenarios in step S2 are constructed on a monthly basis, and the number of emission scenarios is 47 × 12, for a total of 564.

[0037] S3, establishment of meteorological scenario database: collect historical meteorological data of the park, including wind direction, wind speed, cloud cover, temperature, count the range of meteorological parameters and divide them into equal intervals, among which cloud cover is divided into total cloud cover and low cloud cover.

[0038] Further, in step S3, wind direction, wind speed, total cloud cover, low cloud cover, and temperature are defined as W, V, TC, LC, and T, respectively, with ranges of 0-365°, 0-12 m / s, 0-10, 0-10, and -8-40°C, and intervals of 45°, 2 m / s, 2, 2, and 4°C, respectively.

[0039] Furthermore, in step S3, different meteorological parameters are arranged and combined to obtain 8×6×5×5×12 meteorological scenarios, for a total of 14,400.

[0040] S4, establishment of meteorological-emission pollution source tracing library: the 564 emission scenarios contained in the emission scenario library established in step S2 and the 14,400 meteorological scenarios contained in the meteorological scenario library established in step S3 are arranged and combined to obtain meteorological-emission scenarios. There are 564×14,400 types in total. Different meteorological-emission scenarios are input into the diffusion model to obtain the contribution of each enterprise's emissions to the pollutant concentration at the receiving point under each meteorological-emission scenario.

[0041] Furthermore, the diffusion models in step S4 include AERMOD and CALPUFF, but are not limited to these two diffusion models. CALPUFF is mainly used to simulate the diffusion process of pollutants under quiet and low wind conditions.

[0042] Furthermore, the diffusion model in step S4 needs to be localized. For AERMOD, the terrain data and land use data need to be localized. For the CALPUFF model, the terrain data, land use data, and ozone background concentration need to be localized.

[0043] Furthermore, the receptor points in step S4 are environmentally sensitive points, which may be located at the factory boundary or near residential areas.

[0044] Furthermore, a set of model results obtained in step S4 is the contribution of 218 enterprises to the mass concentration of a certain component of VOCs at the receptor site.

[0045] Furthermore, in step S4, there are 564×14400 groups of model results.

[0046] Furthermore, the model results in step S4 need to correspond to the meteorological scenario and the emission scenario.

[0047] S5, construction of meteorological-emission pollution source tracing model: the meteorological-emission source tracing results obtained in step S4 are used as training samples, and the training samples are transferred to the artificial neural network ANN. The meteorological conditions and emission scenarios are used as the input layer of the training, and the contribution value of enterprise emissions to the mass concentration of pollutants at the receiving point is used as the output layer of the training. The ratio of training set, validation set, and test set is 8:1:1 respectively.

[0048] Furthermore, the training set in step S5 is used to train the model and fit the data distribution law, that is, to determine the parameters such as the weight and bias of the model. The validation set is used to verify the data set of the model performance. During the model training process, the validation set is used to adjust the model parameters to optimize the model performance. The test set is used to evaluate the data set of the model performance. After the model training is completed, the test set is used to evaluate the generalization ability of the model.

[0049] Furthermore, the ANN artificial neural network in step S5 has two hidden layers.

[0050] S6, pollution source tracing: When the VOCs concentration at the receptor point exceeds the standard, the VCOs characteristic component information and real-time meteorological parameters are input into the meteorological-emission pollution source tracing model. The model is executed to obtain the contribution ratio of 218 enterprises' emissions to the VOCs characteristic component concentration at the receptor point, and the list of potential polluting enterprises is obtained. The VOCs source spectrum of organized / unorganized emissions of potential polluting enterprises in the enterprise source spectrum library is retrieved, and the cosine similarity analysis is performed with the VOCs spectrum at the receptor point when the concentration exceeds the standard. Finally, the name of the pollution source that caused the VOCs concentration at the receptor point to exceed the standard is locked.

[0051] Furthermore, in step S6, the real-time weather conditions in the embodiment are: wind direction is 45°, wind speed is 5 m / s, temperature is 30°C, and the constructed weather scenario is (45, 5, 30).

[0052] Furthermore, in step S6, the receptor point VOCs spectrum is sampled for detection and analysis only when the VOCs concentration exceeds the standard.

[0053] Furthermore, the determination of the number of receptor points in step S6 needs to be based on the traceability area and surrounding building facilities. The contribution value of enterprise emissions to the pollutant concentration at the receptor point in the result includes and only includes the contribution of enterprise emissions to the pollutant concentration at the receptor point where VOCs exceed the standard.

[0054] Furthermore, the VCOs characteristic component in step S6 refers to the component that accounts for more than 10% in the VOCs spectrum of the receptor point.

[0055] Furthermore, the name of the pollution source that causes the VOCs concentration at the receptor point to exceed the standard in step S6 can be an organized emission outlet or an unorganized emission surface source.

[0056] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A VOCs source tracing method for industrial parks based on fingerprint and data-driven, characterized by: The specific steps include: S1, Establishment of VOCs source spectrum library of park enterprises: Through the collection and composition analysis of VOCs emitted from the organized / unorganized outlets of enterprises, the test results are normalized and analyzed to form the VOCs source spectrum of park enterprises; S2, establishment of emission scenario library: collect VOCs emission inventories of enterprises in the park, distribute the emission inventories in time and space based on the location of enterprises in the park and the production situation throughout the year, and build an emission scenario library of different VOCs components; S3, establishment of meteorological scenario library: according to the existing meteorological station monitoring data and WRF simulated meteorological field data of the park, the range of meteorological parameters of the park is determined. The intervals of each parameter are divided according to the range of meteorological parameters of the park over the years. After the divided meteorological parameters are arranged and combined, different meteorological scenarios are obtained; S4, establishment of meteorological-emission pollution source tracing library: combine emission scenarios and meteorological scenarios to form multiple meteorological-emission scenarios, simulate the contribution of park enterprise emissions to the concentration of each component of VOCs at the receptor point under different meteorological-emission scenarios through diffusion models, summarize the simulation results under each meteorological-emission scenario, and correspond the meteorological scenarios with the model simulation results to obtain meteorological-emission source tracing samples, which are compiled to form a meteorological-emission source tracing library; S5, Construction of meteorological-emission pollution source tracing model: The data set in the meteorological-emission source tracing library is used as training samples, meteorological conditions and pollution source names are used as input, and the pollution source contribution ratio is used as output. The meteorological-emission source tracing model is obtained through machine learning training. The learning method is artificial neural network ANN. The accuracy of the meteorological-emission source tracing model is improved by learning and adjusting parameters; S6, Pollution source tracing: When the VOCs concentration at the receptor point exceeds the standard, the real-time meteorological data and the characteristic components of VOCs at the time of exceeding the standard are input into the meteorological-emission source tracing model, which can obtain the concentration contribution value of each enterprise's emissions to the characteristic components of the receptor point, quickly find potential polluting enterprises, retrieve the VOCs source spectrum of the potential polluting enterprises, and perform cosine similarity analysis with the VOCs spectrum of the receptor point at the time of exceeding the standard, and finally lock the discharge outlet of the suspicious enterprise to achieve rapid and precise tracing of VOCs in the park.

2. The method for tracing the source of VOCs in industrial parks based on fingerprint and data-driven according to claim 1, characterized in that: The emission inventory includes organized and unorganized emissions. Organized emissions from enterprises are treated as point sources, and unorganized emissions are treated as surface sources.

3. The VOCs source tracing method for industrial parks based on fingerprint and data-driven according to claim 1, characterized in that: The meteorological-emission scenario refers to multiple component emission scenarios corresponding to multiple meteorological scenarios, so each meteorological scenario also corresponds to multiple model simulation results. The models include AERMOD and CALPUFF models. When the model is applied, localization processing of terrain and land use type model parameters is required.

4. The method for tracing the source of VOCs in industrial parks based on fingerprint and data-driven according to claim 1, characterized in that: The name of the pollution source is the name of the enterprise.

5. The method for tracing the source of VOCs in industrial parks based on fingerprint and data-driven according to claim 1, characterized in that: The ANN neural network structure consists of three parts: input layer, hidden layer and output layer. Each layer is composed of multiple neurons, each neuron is interconnected with all neurons in the previous layer, and each connection corresponds to a weight. In the three layers of the neural network, the input layer is used to receive and input data, the hidden layer processes the data, and the output layer generates and outputs the generated results.

6. The method for tracing the source of VOCs in industrial parks based on fingerprint and data-driven according to claim 1, characterized in that: When learning and adjusting parameters, the parameters are adjusted by minimizing the root mean square error RMSE. Among them, y i is the true value of the i-th sample point, y' i is the predicted value.

7. The method for tracing the source of VOCs in industrial parks based on fingerprint and data-driven according to claim 1, characterized in that: The real-time meteorological data includes wind direction, wind speed, and temperature, and the characteristic components are components that account for more than 10% of the VOCs at the receptor point.

8. The method for tracing the source of VOCs in industrial parks based on fingerprint and data-driven according to claim 1, characterized in that: The meteorological parameters include wind speed, wind direction, temperature, cloud cover, precipitation, and air pressure. The wind direction is set to 0-360°, where 0° and 360° are both due north. The cloud cover is in a ten-point scale, ranging from 0 to 10.

Images