Pollution treatment facility discrimination method and system based on multi-source data, and electronic equipment

Abstract

The invention provides a pollution control facility discrimination method and system based on multi-source data, and electronic equipment. The method comprises the following steps: collecting enterprise power consumption data, working condition parameters and CEMS online monitoring data; processing the data, including data cleaning, conversion and feature extraction; training and testing are carried out according to data of historical operation states, a CatBoost model is optimized by utilizing grid search and cross validation, and parameter characteristics are acquired by utilizing an optimal hyper-parameter training test model; and finally, calculating a factor contribution value by using an SHAP model, and determining an abnormal factor for early warning. According to the invention, through multi-source data fusion and an advanced model, the discrimination accuracy and interpretability of pollution control facilities are improved, and powerful support is provided for air pollution control.

Description

Technical Field

[0001] The present invention belongs to the technical field of pollution control facility identification, and in particular relates to a pollution control facility identification method, system and electronic equipment based on multi-source data. Background Art

[0003] Efficient operation of pollution control facilities is one of the key means to achieve pollutant emission reduction.

[0004] Traditional methods for identifying pollution control facilities mainly rely on manual monitoring and a single data source, such as regular equipment inspections, emissions testing, etc. This approach has many drawbacks. On the one hand, manual monitoring is costly, inefficient, and easily affected by subjective factors, resulting in inaccurate results. On the other hand, a single data source is difficult to fully reflect the actual operating status of pollution control facilities.

[0005] At present, although some advanced monitoring technologies and equipment are constantly emerging, most of them can only monitor specific pollutants or parameters, and cannot comprehensively consider the impact of multiple factors on pollution control facilities. Moreover, different types of pollution control facilities have different working principles and characteristics, which require targeted identification methods.

[0006] With the rapid development of information technology, the acquisition and processing of multi-source data has become easier. The multi-source data in the present invention mainly include electricity consumption data, working condition monitoring data, enterprise CEMS online data, etc. Electricity consumption data can reflect the energy consumption of pollution control facilities, and by analyzing the electricity consumption pattern, it can be determined whether the facilities are operating normally. Working condition monitoring data can monitor the working status of pollution control facilities in real time, such as temperature, pressure, flow and other parameters, which helps to detect abnormal situations in time. Enterprise CEMS online data can provide information such as pollutant emission concentrations before and after treatment of pollution control facilities, providing an important basis for evaluating the treatment effect of facilities. These multi-source data contain rich information and can provide a more comprehensive and accurate basis for the identification of inefficient and failed pollution control facilities. However, there is currently a lack of an effective system and method that can fully integrate multi-source data to accurately identify inefficient and failed pollution control facilities.

[0007] After searching, there is no patent disclosure on multi-source data to identify the inefficiency and failure of pollution control facilities, but there are patents related to online monitoring systems and methods for pollution control facilities, such as China Patent Application No.: 201810539544.3, the name of the invention patent is: an online monitoring system and method for pollution control facilities based on the Internet of Things, and the embodiment of the invention discloses an online monitoring system and method for pollution control facilities based on the Internet of Things. The system is mainly based on the Internet of Things technology, with the environmental pollution control facility data acquisition instrument, referred to as the "environmental protection black box" as the hardware core, which can obtain the operating time, on-site process parameters, characteristic pollutant concentrations before and after pollution control of small and micro enterprises in real time. Combined with the data processing platform, it provides effective data evidence for environmental law enforcement and provides a strong basis for the calculation and collection of environmental protection taxes. This scheme is a good exploration of the supervision and management of pollution control facilities, but the data collected from environmental protection facilities only includes the vibration of the core motor of the control facility, the flow of pollutants, and the concentration of pollutants, and does not consider the actual electricity consumption of the enterprise and the operating parameters of the enterprise. There may be certain errors in the judgment.

[0008] In summary, the current methods for identifying inefficiency and failure of industrial enterprises still have certain uncertainties. It is difficult to judge the actual operation situation of treatment facilities by considering only a single parameter. Therefore, multi-source data can be fully integrated to accurately identify inefficiency and failure of pollution control facilities. Summary of the invention

[0009] The purpose of the embodiments of the present invention is to provide a method, system and electronic device for identifying pollution control facilities based on multi-source data, aiming to solve the problems raised in the background technology.

[0010] To achieve the above object, the present invention provides the following technical solutions:

[0011] A method, system and electronic device for identifying pollution control facilities based on multi-source data, the method comprising the following steps:

[0012] Collect enterprise electricity consumption data, operating parameters and CEMS online monitoring data.

[0013] The enterprise electricity consumption data collection refers to connecting with the enterprise's power management system to obtain the electricity consumption data of air pollution control facilities. The electricity consumption data includes information such as power consumption, peak power consumption, and valley power consumption during equipment operation. The electricity consumption data can reflect the operating intensity and time pattern of the control facilities. For example, high power consumption may mean that the facilities are operating at full capacity to treat a large amount of pollutants, while abnormal power consumption fluctuations may indicate that the equipment is faulty or operating inefficiently.

[0014] The operating parameter collection requires the installation of sensors to collect the operating parameters of the treatment facility, such as the pressure, temperature, pH value, etc. of the equipment. These parameters directly reflect the working status of the treatment facility, and different combinations of operating parameters can indicate whether the facility is operating within the normal range. For example, a specific pressure and temperature combination may correspond to the best treatment effect, and deviation from this combination may mean that there is a problem with the operation of the facility.

[0015] The CEMS online monitoring data collection is connected to the CEMS system to obtain real-time atmospheric pollutant emission data, including the concentration and flow rate of pollutants such as sulfur dioxide (SO2), nitrogen oxides (NOx), and particulate matter. CEMS data can directly reflect the treatment effect of the treatment facilities on atmospheric pollutants. By comparing the pollutant concentrations at the inlet and outlet, the removal efficiency of the treatment facilities can be calculated.

[0016] Processing the power consumption data, operating parameters and CEMS online monitoring data;

[0017] The data processing includes data cleaning, data conversion and feature extraction;

[0018] The data cleaning is to remove the noise caused by factors such as power grid fluctuations and metering errors from the enterprise electricity consumption data. For example, by calculating the average power consumption over a period of time, remove abnormal data points that deviate significantly from the average value. For the operating parameter data, identify and remove abnormal values ​​caused by sensor failure or interference. For example, according to the normal working range of the equipment, set a reasonable range of pressure and temperature, and data points outside this range are regarded as abnormal values ​​and processed. For CEMS online monitoring data, remove abnormal data caused by instrument failure, shutdown, etc. For example, for continuous extremely high or extremely low pollutant concentration values, further inspection and processing are carried out.

[0019] The data conversion is to convert data from different sources into a unified format and unit for comprehensive analysis. For example, the enterprise electricity consumption data is converted from kilowatt-hours to megawatt-hours, the different units of operating parameters are unified into international standard units, and the units of pollutant concentrations in CEMS data are unified.

[0020] The feature extraction is performed by analyzing the relationship between the operating parameters and the pollutant removal efficiency. For example, the optimal removal efficiency under a specific pressure and temperature combination is determined, and these relationships are extracted as feature parameters. It is also possible to calculate the statistical characteristics of the enterprise's electricity consumption data, operating parameters, and CEMS data, such as mean, variance, standard deviation, etc. These statistical characteristics can reflect the centralization trend and degree of dispersion of the data, and are of reference value for judging the operational stability of the treatment facilities. Perform time series analysis to observe the changing trends of the enterprise's electricity consumption data, operating parameters, and CEMS data over time, as well as the correlation between them. For example, analyze the relationship between peak electricity consumption and peak pollutant emissions in different time periods.

[0021] Determine whether the processed monitoring data is running normally at the corresponding moment, perform correlation calculation, set the correlation threshold, and retain the monitoring factors that are greater than the correlation threshold;

[0022] After processing the power consumption data, operating parameters and CEMS online monitoring data, the parameters of normal operation of the historical treatment facilities are marked as "0", and those of abnormal operation are marked as "1", and a training set and a test set are constructed.

[0023] Using the training set, training the CatBoost model;

[0024] The CatBoost model is optimized using grid search and cross validation to obtain the optimal hyperparameters of the model.

[0025] The CatBoost is a GBDT model based on a symmetric decision tree. It solves the problem of difficult handling of categorical features by adding hyperparameters to the frequency of occurrence of categorical features. It also solves the problem of offset and gradient deviation during model prediction, reduces overfitting, and greatly improves the generalization ability and accuracy of the algorithm. 2 As the performance evaluation index, select R 2 The largest parameter combination is taken as the optimal parameter combination.

[0026] Model parameters include the number of base models (estimators) and the depth of the tree (depth). Calculation index R 2 The calculation of is shown in formula (1):

[0027]

[0028] Where n is the number of samples, y i is the true value of the i-th sample, is the predicted value of the i-th sample, is the true value y i The mean of .

[0029] According to the optimal hyperparameter combination of the CatBoost model, the model is tested after removing the marked "0" and "1" in the test set, and the obtained results are compared with the marked parameters to verify the accuracy of the model.

[0030] After the CatBoost model is trained and tested, it can be used to analyze real-time data and obtain the specific characteristics of each parameter;

[0031] After obtaining the specific characteristics of each parameter, the SHAP model is used to further calculate the contribution value of each factor to obtain the contribution value of different factors under abnormal operating conditions;

[0032] The SHAP model combines the idea of ​​game theory to calculate the marginal contribution of the added features to the model output, and uses a unified method to explain the degree of influence of each driving factor of the machine learning model on the predicted value. As a post-explanation tool, SHAP can still explain the contribution of each feature to the prediction results in the face of the "black box" nature of the machine learning algorithm, and enhance the interpretability of the machine learning model. Through the SHAP package of the python language, you can obtain the model interpreter of TreeExplainer and calculate the influence of each feature on the prediction results of the machine learning model.

[0033] The relationship between the SHAP value and the predicted value is shown in formula (2):

[0034]

[0035] Where f(x) represents the predicted value of the model; represents the benchmark value of the model, that is, the mean of all predicted values ​​of the model; p represents the number of features, is the SHAP value of the ith driving factor, indicating the relative impact of the ith driving factor on the predicted value.

[0036] According to the contribution values ​​of different factors, the factors that contribute most to the outliers can be obtained;

[0037] Issue early warnings for the factors with the greatest impact and provide timely guidance to personnel to conduct on-site inspections.

[0038] After the model is established, it is visualized through real-time data access, and the identification results are output in a visual manner, such as displaying the operating status of air pollution control facilities, inefficiency and failure risks, and other information through charts and reports.

[0039] The visualization includes, but is not limited to, drawing a line graph to show the changes in the enterprise's electricity consumption data, operating parameters and CEMS data in different time periods, as well as their relationship with the status of the treatment facilities. For example, displaying the curve of electricity consumption over time, while marking the changes in pollutant concentration and operating parameters in the corresponding time period.

[0040] A method, system and electronic equipment for identifying pollution control facilities based on multi-source data, the system comprising:

[0041] The data acquisition module 31 is responsible for collecting multi-source data such as enterprise power consumption data, operating parameters and CEMS online monitoring data, including connection interfaces with the enterprise power management system, sensor equipment and CEMS system.

[0042] The data processing module 32 cleans, converts and analyzes the collected data to extract characteristic parameters, and specifically includes a data cleaning unit, a data conversion unit and a characteristic extraction unit.

[0043] The model building module 33 builds an air pollution control facility identification model based on the extracted characteristic parameters, including a model selection unit, a model training unit and a model optimization unit.

[0044] The result output module 34 outputs the screening results in a visual way to provide decision support for users, including a visual display unit and a decision support unit.

[0045] A method, system and electronic device for identifying pollution control facilities based on multi-source data, the electronic device comprising: a processor 41, a power supply 43 and a memory 42, the instructions of each device module are input 44 and executed by the processor, and the memory is used to store the instructions and output results 45 of the device modules.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] 1. Improve identification accuracy: Traditional methods for identifying pollution control facilities often rely on a single data source or manual monitoring, and accuracy is difficult to guarantee. The present invention can comprehensively reflect the operating status of pollution control facilities from multiple angles by integrating multi-source data such as electricity consumption data, operating condition monitoring data, and enterprise CEMS online data. Electricity consumption data can reveal the energy consumption of the facility and determine whether the equipment is operating normally; operating condition monitoring data such as temperature, pressure, flow and other parameters can reflect the working status of the facility in real time and detect anomalies in a timely manner; enterprise CEMS online data provides pollutant emission concentration information before and after treatment to evaluate the treatment effect. The comprehensive use of multi-source data has greatly improved the accuracy of identifying inefficient and failed states of pollution control facilities.

[0048] 2. Realize real-time monitoring and early warning: It is difficult to conduct real-time and effective monitoring and early warning of pollution control facilities with existing technologies. The present invention can collect multi-source data in real time, and conduct real-time analysis of the operating status of facilities through efficient data processing and identification models. Once signs of inefficiency or failure of facilities are found, the system can issue an early warning in time, so that relevant personnel can quickly take measures for maintenance and adjustment to avoid further deterioration of pollution problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0050] Figure 1 A schematic diagram of the method of the present invention.

[0051] Figure 2 It is a flow chart of the identification method of the present invention.

[0052] Figure 3 It is a block diagram of the identification device of the present invention.

[0053] Figure 4 It is a block diagram of the electronic equipment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0055] Understandably, traditional methods for identifying pollution control facilities rely mainly on manual monitoring and a single data source, such as regular equipment inspections and emissions testing. This approach has many drawbacks. On the one hand, manual monitoring is costly, inefficient, and easily affected by subjective factors, resulting in inaccurate results. On the other hand, a single data source is difficult to fully reflect the actual operating status of pollution control facilities. At present, although some advanced monitoring technologies and equipment continue to emerge, most of them can only monitor specific pollutants or parameters, and cannot comprehensively consider the impact of multiple factors on pollution control facilities. Moreover, different types of pollution control facilities have different working principles and characteristics, requiring targeted identification methods.

[0056] To solve the above problems, the present invention provides a pollution control facility identification method, system and electronic equipment based on multi-source data. After integrating the electricity consumption monitoring of industrial enterprises, operating parameters and CEMS online data, the data is processed and input into the Catboost model for calculation. The obtained parameter features are further input into the SHAP model to calculate the contribution of each factor under abnormal operation conditions, so as to provide early warning for abnormal factors.

[0057] Figure 1 A flow chart of a method provided by an embodiment of the present invention is shown.

[0058] Specifically, in a preferred embodiment provided by the present invention, a method, system and electronic device for identifying pollution control facilities based on multi-source data, taking the industrial coating industry as an example, the method specifically includes the following steps:

[0059] First, the specific interfaces in the data acquisition module 31 are used to connect with the enterprise power management system, sensor equipment and CEMS system respectively.

[0060] (1) Enterprise electricity consumption data collection: Connect with the power management system of the coating enterprise to obtain the electricity consumption data of the working equipment in the coating workshop and the corresponding air pollution control facilities (such as waste gas treatment equipment). This includes information such as power consumption in different power stages when the equipment is running, the time periods when peak power consumption occurs (for example, during high-load production), and valley time periods. These electricity consumption data can reflect the operating intensity and time patterns of the control facilities during the coating production process. For example, when a workshop is carrying out large-scale coating operations, the increase in power consumption of the control facilities may mean that the facilities are operating at full capacity to treat a large amount of organic waste gas and other pollutants generated by coating; and abnormal power consumption fluctuations, such as a sudden drop or increase in power consumption, may indicate that the equipment is faulty or operating inefficiently.

[0061] (2) Collection of operating parameters: Install sensors on the treatment facilities in the paint shop to collect operating parameters. Such as the pressure of the exhaust gas treatment equipment (reflecting the smoothness of gas flow), temperature (may affect the efficiency of chemical reactions), pH value (if acid-base treatment is involved), etc. Different combinations of operating parameters can indicate whether the facility is operating within the normal range. For example, a specific combination of pressure and temperature may correspond to the best exhaust gas treatment effect. If it deviates from this combination, it may mean that there is a problem with the operation of the facility, such as abnormal increase in pressure due to blockage of the filter material, or abnormal decrease in temperature due to failure of the heating element.

[0062] (3) CEMS online monitoring data collection: Connect to the CEMS system to obtain real-time atmospheric pollutant emission data. For the industrial coating industry, the concentration, flow rate and other information of pollutants such as particulate matter (such as paint mist particles) and volatile organic compounds (VOCs) are mainly monitored. CEMS data can directly reflect the treatment effect of the treatment facilities on the atmospheric pollutants generated by industrial coatings. By comparing the pollutant concentrations at the inlet and outlet of the exhaust gas treatment equipment, the removal efficiency of the treatment facilities can be calculated. For example, the VOCs concentration at the inlet is high, and the outlet concentration is significantly reduced after treatment, so the VOCs removal efficiency can be calculated.

[0063] After the data collection is completed, the data processing module 32 processes the data.

[0064] (1) Data cleaning:

[0065] For the enterprise electricity consumption data, taking into account the possible power fluctuations in industrial coating workshops (such as the impact of large equipment startup on the power grid), measurement errors and other factors, the average power consumption over a period of time (such as a day or a week) is calculated to remove abnormal data points that deviate significantly from the average. For example, if the power consumption at a certain moment is several times higher than the average power consumption during that period, and there is no reasonable explanation for production activities, it is considered abnormal data and removed.

[0066] For the working condition parameter data, identify and eliminate abnormal values ​​caused by sensor failure or interference. For example, according to the working range of the exhaust gas treatment equipment during normal painting operations, set a reasonable range for pressure and temperature. If the pressure is too high beyond the normal range, it may be a sensor failure or pipe blockage; if the temperature is too low beyond the reasonable range, it may be a heating element damage. Data points outside this range are regarded as abnormal values ​​and processed.

[0067] For CEMS online monitoring data, remove abnormal data caused by instrument failure, shutdown, etc. For example, for extremely high or low pollutant concentration values ​​that appear continuously, further inspection and processing are carried out. If it is found that the data abnormality is caused by instrument failure, repair or calibration are carried out in time, and the abnormal data is removed.

[0068] (2) Data conversion:

[0069] Convert enterprise electricity consumption data from kilowatt-hours to megawatt-hours to better unify with other data in terms of magnitude and facilitate comprehensive analysis. For example, after converting the electricity consumption data of multiple equipment in a paint shop, it is possible to more intuitively compare the relationship between the electricity consumption of different equipment and the operation of treatment facilities.

[0070] The different units of operating parameters are unified into international standard units. For example, the temperature is converted from Celsius to Kelvin, making it easier to analyze different parameters.

[0071] Unify the units of pollutant concentrations in CEMS data. For example, convert particulate matter concentrations from mg / m3 to μg / m3 to keep them consistent with other data, facilitating subsequent calculations and analysis.

[0072] (3) Feature extraction:

[0073] Analyze the relationship between operating parameters and pollutant removal efficiency. In the industrial coating industry, determine the optimal paint mist particle removal efficiency under a specific pressure and temperature combination, and extract these relationships as characteristic parameters. For example, through multiple experiments and data analysis, it is found that when the pressure of the exhaust gas treatment equipment is within a certain range and the temperature is maintained at a specific value, the paint mist particle removal efficiency is the highest. Extract these relationships as important features for subsequent model analysis.

[0074] Calculate the statistical characteristics of the enterprise's electricity consumption data, operating parameters and CEMS data, such as mean, variance, standard deviation, etc. These statistical characteristics can reflect the concentration trend and dispersion of the data, and have reference value for judging the operational stability of the treatment facilities. For example, calculate the mean and variance of the electricity consumption data of the paint shop treatment facilities over a period of time. If the variance is large, it means that the electricity consumption is unstable, which may indicate that there is a problem with the equipment.

[0075] Conduct time series analysis to observe the change trends of enterprise electricity consumption data, operating parameters and CEMS data over time, as well as the correlation between them. For example, analyze the relationship between peak electricity consumption and peak pollutant emissions in different time periods. In the industrial coating industry, it may be found that during peak production periods, electricity consumption is high, pollutant emissions also increase, and the pressure and temperature of the treatment facilities will change accordingly. Through this analysis, we can better understand the operation of the treatment facilities under different production conditions.

[0076] The model selection unit in the model building module 33 selects the CatBoost model.

[0077] (1) Determine whether the processed monitoring data is operating normally at the corresponding moment, perform correlation calculations, set correlation thresholds, and retain monitoring factors that are greater than the correlation thresholds. In the industrial coating industry, the correlation thresholds between different data are determined based on the actual conditions and historical data of the coating workshop. For example, if a certain correlation is found between electricity consumption data and paint mist particle concentration, when the correlation is lower than the set threshold, it may mean that there is a problem in a certain link. In this way, factors that have a greater impact on the operating status of the treatment facilities can be screened out for further analysis.

[0078] (2) The parameters of normal operation of historical treatment facilities are marked as "0", and those of abnormal operation are marked as "1", and training and test sets are constructed. The historical data of industrial coating workshops over a period of time are collected, and the data of normal operation of treatment facilities are marked as "0", such as the period of stable equipment operation and high pollutant removal efficiency; the data of abnormal operation are marked as "1", such as equipment failure, pollutant emission exceeding the standard, etc. These marked data are divided into training sets and test sets for subsequent model training and testing.

[0079] (3) Use the training set to train the CatBoost model. By performing grid search on different parameter combinations and combining the cross-validation method, we adjust the number of base models (estimators) and the depth of the tree (depth). 2 As the performance evaluation indicator, R 2 The largest parameter combination is taken as the optimal parameter combination. During the model training process, the parameters are continuously adjusted to make the model better adapt to the data characteristics of the industrial coating industry.

[0080] (4) Based on the optimal hyperparameter combination of the CatBoost model, remove the marked "0" and "1" in the test set, test the model, and compare the results with the marked parameters to verify the accuracy of the model. For example, if the model predicts that the governance facility operates normally in a certain period of time, but it is actually marked as abnormal, it means that the model has deviations and needs further adjustment and optimization. In this way, the accuracy of the model can be continuously improved.

[0081] (5) After the CatBoost model is trained and tested, it can be used to analyze real-time data and obtain the specific characteristics of each parameter. In an industrial paint shop, real-time access to electricity consumption data, operating parameters, and CEMS online monitoring data is used to analyze the trained CatBoost model and obtain the specific characteristics of each parameter at the current moment. For example, the operating status of the treatment facility and the efficiency of paint mist particle removal can be determined in real time.

[0082] (6) After obtaining the specific characteristics of each parameter, the SHAP model is used to further calculate the contribution value of each factor. Through the SHAP package of the Python language, the model interpreter of TreeExplainer can be obtained to calculate the impact of each feature on the prediction results of the machine learning model and obtain the contribution value of different factors under abnormal operating conditions. For example, when the treatment facility has an abnormal operating state, the SHAP model can be used to analyze whether the factors such as changes in electricity data, abnormal operating parameters, or excessive pollutant emissions contribute the most to the abnormal state.

[0083] (7) Based on the contribution values ​​of different factors, the factors that contribute most to the abnormal values ​​can be obtained. In the industrial coating industry, the factors that contribute most to the abnormal operation of the treatment facilities are determined, such as abnormal increase in power consumption in a certain period of time, excessive pressure or low temperature. These factors are focused on and analyzed so that timely measures can be taken to adjust and repair them.

[0084] (8) Issue early warnings for the factors that have the greatest impact, and promptly guide personnel to conduct on-site inspections. When the factors that contribute the most to abnormal values ​​are found, the system issues an early warning signal and notifies relevant personnel to conduct on-site inspections. For example, if the pressure sensor shows abnormally high pressure, it may mean that the filter system of the exhaust gas treatment equipment is clogged and the filter material needs to be cleaned or replaced in time. In this way, problems can be discovered and measures can be taken in a timely manner to avoid environmental pollution and production interruptions caused by failures in treatment facilities.

[0085] The result output module 34 outputs the screening results in a visual manner to provide decision support for users.

[0086] (1) After the model is established, it can be visualized by accessing data in real time. For example, a line graph can be drawn to show the changes in the enterprise's electricity consumption data, operating parameters, and CEMS data in different time periods, as well as their relationship with the status of the treatment facilities. The curve of electricity consumption over time can be displayed, and the changes in pollutant concentrations and operating parameters in the corresponding time period can be marked.

[0087] (2) At the same time, the identification results can be output in a visual way, such as by displaying the operating status of air pollution control facilities, inefficiency and failure risks, and other information in the form of charts and reports. Through visual display, relevant personnel can more intuitively understand the operation of the control facilities, identify problems in a timely manner, and take measures.

[0088] The processor 41 in the electronic device executes corresponding operations after the instructions of each device module are input 44, the memory 42 is used to store the instructions and output results 45 of the device modules, and the power supply 43 provides power guarantee for the entire system.

[0089] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0090] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory.

[0091] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0092] The above-mentioned embodiment only expresses one implementation mode of the present invention, and its description is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

[0093] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying pollution control facilities based on multi-source data, characterized in that: The following steps are involved: S1. Collect enterprise electricity consumption data, operating parameters and CEMS online monitoring data; S2. Processing the power consumption data, operating parameters and CEMS online monitoring data, wherein the data processing includes data cleaning, data conversion and feature extraction; S3. Determine whether the operation of the processed monitoring data at the corresponding moment is normal, perform correlation calculation, set a correlation threshold, and retain monitoring factors greater than the correlation threshold; S4. Mark the parameters of normal operation of historical governance facilities as "0" and abnormal operation as "1", and construct training sets and test sets; S5. Using the training set, training the CatBoost model; S6. Optimizing the CatBoost model by using grid search and cross validation to obtain the optimal hyperparameters of the model; S7. Testing the model according to the optimal hyperparameter combination of the CatBoost model to verify the accuracy of the model; S8. Analyze the real-time data using the trained and tested CatBoost model to obtain the specific characteristics of each parameter; S9. Calculate the contribution value of each factor using the SHAP model to obtain the contribution value of different factors under abnormal operation conditions; S10. According to the contribution values ​​of different factors, determine the factor that contributes most to the abnormal value and issue an early warning; S11. Through real-time access to data, visual display is performed and the identification results are output in a visual manner.

2. The method for identifying pollution control facilities based on multi-source data according to claim 1 is characterized in that: The enterprise electricity consumption data collection in step S1 refers to connecting with the enterprise's power management system to obtain the electricity consumption data of air pollution control facilities; The operating parameter collection in step S1 is to collect the operating parameters of the treatment facility by installing sensors; the CEMS online monitoring data collection is to access the CEMS system to obtain real-time atmospheric pollutant emission data; The data cleaning described in step S2 includes removing noise caused by power grid fluctuations and metering errors from the enterprise's electricity consumption data; identifying and eliminating abnormal values ​​caused by sensor failure or interference from the operating parameter data; and removing abnormal data caused by instrument failure, outage, etc. from the CEMS online monitoring data.

3. The method for identifying pollution control facilities based on multi-source data according to claim 1 is characterized in that: The CatBoost model is constructed in R 2 As the performance evaluation index, select R 2 The largest parameter combination is taken as the optimal parameter combination; The CatBoost model parameters include the number of base models (estimators) and the depth of the tree (depth); the calculation index R 2 The calculation is shown in the formula: Where n is the number of samples, y i is the true value of the i-th sample, is the predicted value of the i-th sample, is the true value y i The mean of According to the optimal hyperparameter combination of the CatBoost model, the model is tested after removing the marked "0" and "1" in the test set, and the obtained results are compared with the marked parameters to verify the accuracy of the model.

4. The method for identifying pollution control facilities based on multi-source data according to claim 1 is characterized in that: The relationship between the SHAP value and the predicted value is shown in the formula: Where f(x) represents the predicted value of the model; represents the benchmark value of the model, that is, the mean of all predicted values ​​of the model; p represents the number of features, is the SHAP value of the i-th driving factor, indicating the relative impact of the i-th driving factor on the predicted value; According to the contribution values ​​of different factors, the factor that contributes most to the outliers can be obtained.

5. The method for identifying pollution control facilities based on multi-source data according to claim 1 is characterized in that: The data conversion is to convert data from different sources into a unified format and unit for comprehensive analysis.

6. The method for identifying pollution control facilities based on multi-source data according to claim 1 is characterized in that: The feature extraction includes analyzing the relationship between the operating parameters and the pollutant removal efficiency, determining the optimal removal efficiency under a specific pressure and temperature combination, and extracting these relationships as feature parameters; Calculate the statistical characteristics of enterprise electricity consumption data, operating parameters and CEMS data; Conduct time series analysis to observe the changing trends of enterprise electricity consumption data, operating parameters and CEMS data over time, as well as the correlation between them.

7. The method for identifying pollution control facilities based on multi-source data according to claim 1 is characterized in that: The CatBoost model is a GBDT model based on a symmetric decision tree. It generates new numerical features by adding hyperparameters to the frequency of occurrence of categorical features, thereby solving the problem of categorical features being difficult to handle. It also solves the problems of offset and gradient deviation during model prediction, reduces overfitting, and improves the generalization ability and accuracy of the algorithm.

8. The method for identifying pollution control facilities based on multi-source data according to claim 1 is characterized in that: The SHAP model combines the idea of ​​game theory to calculate the marginal contribution of the added features to the model output, and uses a unified method to explain the influence of each driving factor of the machine learning model on the predicted value, thereby enhancing the interpretability of the machine learning model.

9. A pollution control facility identification system based on multi-source data, characterized in that: include: A data acquisition module (31), which is responsible for collecting multi-source data such as enterprise electricity consumption data, operating parameters and CEMS online monitoring data according to the pollution control facility identification method according to any one of claims 1 to 8, and includes connection interfaces with the enterprise power management system, sensor equipment and CEMS system; A data processing module (32) cleans, converts and analyzes the collected data to extract characteristic parameters, and specifically includes a data cleaning unit, a data conversion unit and a characteristic extraction unit; A model building module (33), which builds an air pollution control facility identification model based on the extracted characteristic parameters according to the pollution control facility identification method according to any one of claims 1 to 8, and comprises a model selection unit, a model training unit and a model optimization unit; The result output module (34) outputs the screening result in a visual manner to provide decision support for the user, and includes a visual display unit and a decision support unit.

10. A pollution control facility identification electronic device based on multi-source data, characterized in that: include: A processor (41), a power supply (43) and a memory (42), wherein the instructions of each device module are input (44) and executed by the processor, and the memory is used to store the instructions of the device module and output (45) results; the pollution control facility identification electronic device is used to execute the pollution control facility identification method described in any one of claims 1 to 8.

Images