A dilution method flue gas monitoring system

Abstract

The application discloses a dilution method flue gas monitoring system, which is used for continuous monitoring in the emission process of waste gas pollution sources and comprises a particulate matter monitoring unit, a gaseous pollutant monitoring unit, a flue gas parameter monitoring unit and a data acquisition and processing unit. The data acquisition and processing unit comprises a sample collection and transmission device, a pretreatment equipment, an analysis instrument, a data acquisition and transmission equipment and auxiliary equipment. The data acquisition and processing unit further comprises an analysis module and an execution module. The analysis module is used for acquiring a flue gas pressure change value. A BP neural network prediction model is constructed with the flue gas pressure change value as input and a probe dilution ratio correction value as output. The probe dilution ratio correction value is predicted through the model. The execution module is in communication connection with the analysis module and is used for acquiring the predicted probe dilution ratio correction value and inputting the probe dilution ratio correction value into the analysis instrument, so that automatic correction of the probe dilution ratio is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of flue gas monitoring, in particular to a dilution method flue gas monitoring system. BACKGROUND

[0002] The flue gas CEMS online monitoring system mainly monitors the real-time of the environmental protection chimney of the enterprise, and sends the monitoring data in the form of a digital string to the provincial environmental control center, so as to understand and supervise the production and emission of the enterprise. The monitoring system collects a series of related data such as flue gas temperature, humidity, SO2, NOx, content, dynamic and static pressure, and flow from the environmental protection chimney. The maintenance of the detection system by the enterprise and the operator plays a crucial role. Unfamiliarity with the hardware and software of the entire system will lead to errors and deviations in uploading data, and long maintenance period, resulting in no uploading, etc. Directly affect the supervision of the provincial department to the enterprise, and even cause serious consequences such as punishment.

[0003] At present, there are two commonly used measurement methods: one is the direct suction method, that is, the flue gas after pretreatment, removal of water, dust and other impurities is sucked into the analyzer for PPM content measurement; the second is the dilution method, which uses compressed air to dilute the flue gas by a certain proportion through the Venturi tube to form a negative pressure system, and directly enters the analyzer for analysis and measurement. The pretreatment process is relatively simple, and the flue gas concentration is low, which is convenient for daily maintenance.

[0004] When the dilution method is used for measurement, a dilution probe needs to be used. At present, the dilution probe is mostly made of stainless steel, and a critical small hole made of quartz material and a vacuum generator made of stainless steel are built-in. The gas flow of the vacuum generator is 1-10 L / min, and the sample gas flow through the critical small hole is 50-500 mL / min. In order to make the nozzle small hole get the critical flow, the ratio of the absolute pressure of the throat of the Venturi tube to the static pressure of the flue gas should be less than or equal to 0.53. At this time, the flow rate of the gas through the critical small hole is limited within the range of sound speed, so the flow rate of the gas through the critical small hole is also a constant value. According to this principle, the dilution ratio of the sample gas can be calculated, wherein the calculation formula of the dilution ratio is: R=(Q1+Q2) / Q2, wherein Q1 is the dilution gas flow, L / min; Q2 is the sample gas flow, mL / min.

[0005] Because the dilution probe is very sensitive to the change of flue gas pressure. The change of flue gas pressure will affect the flow rate Q of sample gas entering the critical nozzle, thereby affecting the dilution ratio. In order to overcome the change of dilution ratio caused by pressure change, high-precision flow and pressure gauges and alarm systems need to be configured to take necessary correction measures when the pressure changes, such as when the vacuum pressure gauge value changes from the original 75kPa to 72kPa, the probe dilution ratio needs to be corrected. The specific correction method currently used is to pass zero gas into the system, adjust the instrument zero point, calibrate the analyzer with a medium concentration standard gas, and store the display value of the stabilized analyzer into the instrument. The calculation formula of the corrected dilution ratio is: R=a / b, wherein R is the dilution ratio; a is the standard gas concentration; b is the display value data stored in the analyzer. The corrected dilution ratio is input into the computer, for example: the original dilution ratio R is 100, the standard gas concentration a is 150ppm, the original display value data b stored in the analyzer is 1.5, and the display value data stored in the analyzer after correction is 1.3. At this time, the corrected dilution ratio R=150 / 1.3≈115.

[0006] In the prior art, the whole process of correcting the probe dilution ratio still needs manual collection of basic change data, and the display value data in the analyzer needs to be modified, which is low in construction efficiency, high in error rate and low in fault tolerance. Therefore, we propose a dilution method flue gas monitoring system. SUMMARY

[0007] The dilution method flue gas monitoring system disclosed by the application is based on big data technology, uses artificial intelligence machine learning algorithm to mine the relationship between the change amount of vacuum pressure gauge and the correction value, thereby realizing automatic correction of the probe dilution ratio, reducing the labor intensity, improving the construction efficiency, reducing the error rate of system operation, and improving the accuracy of flue gas monitoring, which can effectively solve the problems in the background technology.

[0008] To achieve the above purpose, the technical scheme adopted by the application is:

[0009] A dilution method flue gas monitoring system for continuous monitoring during the emission process of waste gas pollution sources, comprising a particulate matter monitoring unit, a gaseous pollutant monitoring unit, a flue gas parameter monitoring unit and a data acquisition and processing unit, the data acquisition and processing unit comprising a sample collection and transmission device, a pretreatment device, an analyzer, a data acquisition and transmission device and auxiliary equipment, wherein the data acquisition and transmission device is in communication connection with the analyzer, the particulate matter monitoring unit, the gaseous pollutant monitoring unit and the flue gas parameter monitoring unit, respectively, the data acquisition and processing unit further comprising an analysis module and an execution module, the analysis module being used to obtain the flue gas pressure change value, to construct a BP neural network prediction model taking the flue gas pressure change value as input and the probe dilution ratio correction value as output, and to predict the probe dilution ratio correction value through the model;

[0010] The execution module is in communication connection with the analysis module, for obtaining the predicted probe dilution ratio correction value and inputting the probe dilution ratio correction value into the analysis instrument to realize automatic correction of the probe dilution ratio.

[0011] Further, in order to enrich the means of obtaining flue gas pressure parameter data and improve the accuracy of flue gas pressure parameter data, the flue gas pressure parameter data obtaining method comprises direct measurement through a flue gas parameter monitoring unit and mining through a machine learning algorithm.

[0012] Further, the mining of flue gas pressure parameter data through a machine learning algorithm is to predict the development trend by analyzing the law of change of flue gas pressure with time through a big data evolution analysis technology.

[0013] Further, the construction steps of the BP neural network prediction model include:

[0014] Step 1, data preprocessing: normalizing the obtained flue gas pressure change value data to obtain a standardized data set, using the maximum and minimum value normalization method to convert the data to the [0, 1] interval; the formula is:

[0015] Wherein, x is the original data, x' is the normalized data, min and max are the minimum and maximum values in the original data, respectively;

[0016] Step 2, design the structure of BP neural network: the structure of BP neural network includes input layer, hidden layer and output layer; the number of nodes of the input layer is equal to the dimension of the input data, the number of nodes of the output layer is equal to the dimension of the output data, and the number of nodes of the hidden layer needs to be adjusted according to the actual situation; the formula is:

[0017] Wherein, f(x) is a sigmoid function or a ReLU function;

[0018] Step 3, divide the obtained standardized data set into training set and test set, wherein the proportion of the training set is 70% to 80%, and the proportion of the test set is 20% to 30%;

[0019] Step 4, use the training set data to train the BP neural network, set the maximum period of training, target error and learning rate; during the training process, the back propagation algorithm is used for weight update to minimize the error function; the error function is:

[0020]

[0021] Wherein, y_k is the label value of the training set, and \hat{y_k} is the predicted value of the BP neural network.

[0022] Step 5, verification of the model: using the test set to test the trained neural network, calculate the prediction error and accuracy; the prediction error is calculated by mean square error (MSE) or mean absolute error (MAE); the formula is:

[0023]

[0024]

[0025] Wherein, yi is the label value of the test set, is the predicted value of the BP neural network;

[0026] Step 6, model optimization: according to the results in the verification process, adjust the structure, learning rate and weight optimization algorithm of the neural network, retrain and verify until the prediction error and accuracy of the model are less than the set threshold.

[0027] Further, the input data of the training set is the flue gas pressure parameter data directly obtained by the flue gas parameter monitoring unit, and the output data of the training set is the probe dilution ratio correction value corresponding to the directly obtained flue gas pressure parameter data:

[0028] Wherein, the probe dilution ratio correction value corresponding to the directly obtained flue gas pressure parameter data is determined by the following steps:

[0029] ①Set the flue gas pressure change amount safety threshold, and obtain the flue gas pressure standard value under the normal operation state of the system;

[0030] ②Read the real-time value of the flue gas pressure, calculate the difference between the real-time value of the flue gas pressure and the flue gas pressure standard value;

[0031] ③Determine whether the difference between the real-time value of the flue gas pressure and the flue gas pressure standard value is within the flue gas pressure change amount safety threshold;

[0032] ④When the difference between the real-time value of the flue gas pressure and the flue gas pressure standard value is outside the flue gas pressure change amount safety threshold, adjust the instrument zero point by using the correction method of passing zero gas into the system, calibrate the analyzer with medium concentration standard gas, and calculate the probe dilution ratio correction value by formula, the calculation formula is: b = a / R, wherein, R is the dilution ratio; a is the standard gas concentration; b is the probe dilution ratio correction value.

[0033] Further, in order to realize the comprehensive monitoring of the discharged flue gas, the flue gas parameter monitoring unit is also used to collect other flue gas parameter data of the sample gas, including flue gas flow, flue gas temperature, flue gas humidity, flue gas oxygen content.

[0034] Further, in order to improve the stability of system operation, reduce the failure rate and repair frequency of the equipment, the sample collection and transmission device has the functions of filtering particulate matter, heating, heat preservation and back blowing purification, through the above functions, the self-cleaning of the system can be realized, the service life of the equipment is prolonged, and the heating temperature is not less than 120 DEG C, and at least 10 DEG C higher than the dew point temperature of flue gas.

[0035] Further, in order to improve the stability of the device operation, the material of the sample collection device should be selected from high temperature resistant, corrosion resistant, non-absorbing and non-reactive materials with gaseous pollutants, and the material of the sample collection device is preferably stainless steel or titanium alloy.

[0036] Further, in order to facilitate the calibration of the system, the sample collection device has at least two sample transmission pipes, respectively used for sample gas collection and transmission and standard gas system calibration.

[0037] Further, in order to prolong the service life of the analysis instrument and reduce the failure rate during operation, the pretreatment equipment includes sample filtration equipment and dehumidification condensation equipment, and the operation temperature of the dehumidification equipment is 4±2 DEG C.

[0038] Compared with the prior art, the present application has the following beneficial effects:

[0039] 1) The dilution method flue gas monitoring system provided by the present application is based on big data technology, and the relationship between the change amount of the vacuum pressure gauge and the correction value is mined, a BP neural network prediction model is constructed by using artificial intelligence machine learning algorithm, the probe dilution ratio correction value is predicted, and the execution module and the analysis module are communicated and connected, used for obtaining the predicted probe dilution ratio correction value, inputting the probe dilution ratio correction value into the analysis instrument, realizing the automatic correction of the probe dilution ratio, and realizing the automatic correction of the probe dilution ratio, which can reduce the labor intensity, improve the construction efficiency, reduce the error rate of system operation, and improve the accuracy of flue gas monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The overall structure block diagram of the dilution method flue gas monitoring system of the present application. DETAILED DESCRIPTION

[0041] The present application will be further described below in conjunction with specific embodiments, wherein the drawings are only used for exemplary description, and the representation is only a schematic diagram, not a physical diagram, and cannot be understood as a limitation of the present application, in order to better illustrate the specific embodiments of the present application, some components of the drawings will be omitted, enlarged or reduced, and do not represent the size of the actual product.

[0042] The following provides a specific embodiment

[0043] The scheme adopted in this embodiment is: a dilution method flue gas monitoring system for continuous monitoring during waste gas pollution source emission, comprising a particulate matter monitoring unit, a gaseous pollutant monitoring unit, a flue gas parameter monitoring unit and a data acquisition and processing unit, the data acquisition and processing unit comprising a sample collection and transmission device, a pretreatment device, an analysis instrument, a data acquisition and transmission device and auxiliary equipment, wherein the data acquisition and transmission device is in communication connection with the analysis instrument, the particulate matter monitoring unit, the gaseous pollutant monitoring unit and the flue gas parameter monitoring unit respectively, the data acquisition and processing unit further comprises an analysis module and an execution module, the analysis module is used to obtain a flue gas pressure change value, a BP neural network prediction model is constructed by taking the flue gas pressure change value as input and taking a probe dilution ratio correction value as output, and the probe dilution ratio correction value is predicted through the model, the execution module is in communication connection with the analysis module and is used to obtain the predicted probe dilution ratio correction value and input the probe dilution ratio correction value into the analysis instrument, so as to realize automatic correction of the probe dilution ratio.

[0044] In this embodiment, the steps of the analysis module include:

[0045] Step S1, flue gas pressure change value acquisition: using a sensor in the flue gas parameter monitoring unit to collect the flue gas pressure change value and transmitting it to the data acquisition and processing unit;

[0046] Step S2, data preprocessing: performing filtering, denoising and calibration preprocessing operations on the collected flue gas pressure change value;

[0047] Step S3, construction of a BP neural network prediction model: taking the preprocessed flue gas pressure change value as input data of the BP neural network and taking the probe dilution ratio correction value as output data, a stable prediction model is obtained through training; wherein the construction process of the BP neural network model is as follows:

[0048] Assuming that the input layer has m neurons, the hidden layer has n neurons, the output layer has k neurons, the input vector is X, the output vector is Y, the output of the i-th hidden layer neuron is h(i), and the output of the j-th output layer neuron is y(j);

[0049] The calculation formula from the input layer to the hidden layer is:

[0050] h(i) = f(sum(w(j, i) * X(j)) + b(i));

[0051] Wherein, w(j, i) represents the connection weight from the j-th neuron of the input layer to the i-th neuron of the hidden layer, b(i) represents the bias of the i-th neuron of the hidden layer, and f is a sigmoid function or a ReLU function;

[0052] The calculation formula from the hidden layer to the output layer is:

[0053] y(j) = g(sum(w(i,j)*h(i))+b(j));

[0054] wherein w(i,j) represents the connection weight from the i th neuron of the hidden layer to the j th neuron of the output layer, b(j) represents the bias of the j th neuron of the output layer, and g is an activation function;

[0055] The training process of the BP neural network includes two stages of forward propagation and back propagation. In the forward propagation, the input vector X is input into the network to obtain the output vector Y. In the back propagation, the output error is calculated, and the connection weight and bias are updated by using the error back propagation to minimize the output error of the network. The error formula of the BP neural network is:

[0056] E = 1 / 2*sum((y(j)-Y(j))^2);

[0057] wherein Y(j) is the expected output of the j th sample in the training set, and y(j) is the actual output of the network. The weight and bias update formula of the BP neural network is:

[0058] w(i,j)(t+1) = w(i,j)(t)-η*δ(j)*h(i) b(j)(t+1) = b(j)(t)-η*δ(j);

[0059] w(j,i)(t+1) = w(j,i)(t)-η*δ(j)*X(i) b(i)(t+1) = b(i)(t)-η*δ(j);

[0060] wherein η is a learning rate, and δ(j) is an error term of the j th neuron.

[0061] Step S4, model prediction: using the trained BP neural network model to predict the new flue gas pressure change value to obtain the probe dilution ratio correction value;

[0062] Step S5, execution module processing: transmitting the predicted probe dilution ratio correction value to the execution module for corresponding correction operation.

[0063] In this embodiment, the specific implementation steps of the system are:

[0064] 1) setting the flue gas pressure change safety threshold in the normal operation state of the system, and obtaining the flue gas pressure standard value in the normal operation state of the system;

[0065] 2) constructing a BP neural network prediction model through an analysis module, and predicting the probe dilution ratio correction value through the model;

[0066] 3) The real-time value of the flue gas pressure is obtained by the flue gas parameter monitoring unit, and when the difference between the real-time value of the flue gas pressure and the standard value of the flue gas pressure is outside the safe threshold range of the flue gas pressure variation, the analysis module sends the probe dilution ratio correction value to the execution module;

[0067] 4) The execution module inputs the probe dilution ratio correction value into the analysis instrument to realize automatic correction of the probe dilution ratio,

[0068] 5) After the automatic correction of the probe dilution ratio is completed, the flue gas sample is collected by the sample collection and transmission device, and after dilution, it is transmitted to the pretreatment equipment, and after the flue gas sample is treated by the pretreatment equipment, it is transported to the analysis instrument for analysis, and the corrected gaseous pollutant data is obtained;

[0069] 6) The particulate matter concentration and flue gas parameters during the flue gas emission process are monitored by the particulate matter monitoring unit and the flue gas parameter monitoring unit respectively, and the monitoring result data of each unit is obtained;

[0070] 7) The gaseous pollutant data and the monitoring result data of each unit are transmitted to the relevant departments through the data collection and transmission equipment.

[0071] In this embodiment, for the acquisition of the particulate matter concentration parameter, a dust measuring instrument can be used. The particulate matter CEMS can be divided into direct reading type, direct extraction type and dilution extraction type according to the sampling method, and can be divided into optical method, electrical method and radioactive method according to the analysis principle. The optical technology is the most widely used in the particulate matter CEMS concentration monitoring method, and the principle of the particulate matter CEMS in use is mainly based on optical technology. The optical method is divided into transmission method and scattering method, and the scattering method is divided into front scattering method and back scattering method. The transmission type dust measuring instrument is suitable for measuring the smoke dust concentration with high concentration in flue gas, and the scattering type dust measuring instrument is suitable for measuring the smoke dust concentration with low concentration in flue gas. If the flue gas contains small droplets, the transmitted light beam cannot distinguish between droplets and smoke dust particles, which will cause the result to be high.

[0072] For the environment with high humidity or even condensation, direct extraction sampling method is suitable. Extraction sampling method needs to be equipped with a jet pump to directly extract the original flue gas from the flue for measurement without changing the nature of the flue gas. A high-power heater with high-precision temperature control is used to heat the wet flue gas at a constant temperature to avoid measurement errors caused by condensation of wet flue gas.

[0073] In this embodiment, for the acquisition of the flue gas parameters, an integrated thermal resistance temperature transmitter can be used to measure the temperature, a pressure transmitter can be used to measure the pressure, zirconia or electrochemical principle can be used to measure the oxygen content, and a capacitive principle can be used to measure the humidity. Differential pressure method, thermal sensing method and ultrasonic method can be used to measure the flow.

[0074] The commonly used method is to use the pitot tube flowmeter to measure the flue gas flow by differential pressure method. For the condition that the flue gas emission is large and stable, and the flue gas flow rate is 10-15 m / s, the pitot tube flowmeter is suitable. The pitot tube flowmeter is mainly composed of a pitot tube and a differential pressure transmitter. The pitot tube has two pressure measuring holes, one hole is aligned with the direction of gas flow, which measures the total pressure, and the other hole is perpendicular to the flow direction, which measures the static pressure. The differential pressure transmitter measures the difference between them, that is, the dynamic pressure, and the flow rate is proportional to the square root of the dynamic pressure. The measurement accuracy of the upper half of the measurement range of the pitot tube flowmeter is high, but the measurement accuracy of the lower half of the measurement range is low, that is, when the flow is low, the measurement accuracy is poor.

[0075] In this embodiment, when the flow rate is relatively stable and higher than 5 m / s, the measurement accuracy of the pitot tube flowmeter is still good, and when the flow rate is lower than 3 m / s, the accuracy is obviously reduced. Therefore, the pitot tube flowmeter is not suitable for low flow rate occasions below 5 m / s. The measurement range should not be set too large, and under normal circumstances, the flow should be located at 50%-70% of the instrument range. Because the flue gas contains dust, once the dust adheres to the total pressure hole, the wind speed coefficient will change, and the measurement error will increase significantly. In order to solve this problem, the pitot tube type flue gas flow measurement system is equipped with a back blowing pipeline and a solenoid valve, which is back blown at regular intervals to remove pollutants, so as to ensure that the pitot tube flow rate coefficient does not change.

[0076] In this embodiment, in order to enrich the means of obtaining flue gas pressure parameter data and improve the accuracy of flue gas pressure parameter data, the flue gas pressure parameter data acquisition method includes direct acquisition by measuring through a flue gas parameter monitoring unit, and also includes acquisition through a machine learning algorithm.

[0077] In this embodiment, the flue gas pressure parameter data is obtained through a machine learning algorithm by using big data evolution analysis technology to analyze the law of change of flue gas pressure with time to predict its development trend.

[0078] In this embodiment, we provide a method for constructing a BP neural network prediction model, and the construction steps of the method include:

[0079] Step 1, data preprocessing: normalize the obtained flue gas pressure change value data to obtain a standardized data set, and use the maximum and minimum value normalization method to convert the data to the [0, 1] interval; the formula is:

[0080] Where x is the original data, x' is the normalized data, min and max are the minimum and maximum values of the original data, respectively;

[0081] Step 2, design the structure of BP neural network: the structure of BP neural network includes input layer, hidden layer and output layer; the number of nodes of input layer is equal to the dimension of input data, the number of nodes of output layer is equal to the dimension of output data, and the number of nodes of hidden layer needs to be adjusted according to actual situation; the formula is:

[0082] Wherein, f(x) is sigmoid function or ReLU function;

[0083] Step 3, divide the obtained standardized data set into training set and test set, wherein the proportion of training set is 70%-80%, and the proportion of test set is 20%-30%;

[0084] Step 4, train the BP neural network using the training set data, set the maximum training period, target error and learning rate; during the training process, the weight value is updated by using back propagation algorithm to minimize the error function; the error function is:

[0085]

[0086] Wherein, y_k is the label value of training set, and \hat{y_k} is the predicted value of BP neural network;

[0087] Step 5, model verification: test the trained neural network using the test set, calculate the prediction error and accuracy; the prediction error is calculated by using mean square error (MSE) or mean absolute error (MAE); the formula is:

[0088]

[0089]

[0090] Wherein, yi is the label value of test set, is the predicted value of BP neural network;

[0091] Step 6, model optimization: according to the results in the verification process, adjust the structure, learning rate and weight optimization algorithm of neural network, and retrain and verify until the prediction error and accuracy of the model are less than the set threshold.

[0092] In this embodiment, in the optimization process of the model, mainly including adjusting learning rate, adjusting weight optimization algorithm and adjusting the structure of neural network: the specific process is:

[0093] Adjust the learning rate: the size of learning rate has great influence on the convergence speed and accuracy of the model, and needs to be adjusted properly. If the learning rate is too small, the convergence speed of the model will be slow; if the learning rate is too large, the model will not converge. In this embodiment, the optimal learning rate is determined by trial and error and experience; the adjustment formula is:

[0094] l = l * d ^ (ep / de) ; wherein, l is the initial learning rate, d is the learning rate decay factor, ep is the current iteration number, and de is the learning rate decay step number.

[0095] (2) Adjust the weight optimization algorithm: the weight optimization algorithm directly affects the convergence speed and stability of the neural network; in this embodiment, the Adam algorithm is used for optimization.

[0096] (3) Adjust the structure of the neural network: increase or decrease the number of layers and the number of neurons in each layer of the neural network to optimize the model.

[0097] Further, the input data of the training set is the flue gas pressure parameter data directly obtained by the flue gas parameter monitoring unit, and the output data of the training set is the probe dilution ratio correction value corresponding to the directly obtained flue gas pressure parameter data:

[0098] Further, in the implementation of the present patent, a method for obtaining the probe dilution ratio correction value corresponding to the directly obtained flue gas pressure parameter data is provided, and the specific steps of the method include:

[0099] ① Set the flue gas pressure change amount safety threshold, and obtain the flue gas pressure standard value under the normal operating state of the system;

[0100] ② Read the real-time value of the flue gas pressure, and calculate the difference between the real-time value of the flue gas pressure and the flue gas pressure standard value;

[0101] ③ Determine whether the difference between the real-time value of the flue gas pressure and the flue gas pressure standard value is within the flue gas pressure change amount safety threshold;

[0102] ④ When the difference between the real-time value of the flue gas pressure and the flue gas pressure standard value is outside the flue gas pressure change amount safety threshold, adjust the instrument zero point by using the correction method of passing zero gas into the system, calibrate the analyzer with the medium concentration standard gas, and calculate the probe dilution ratio correction value by formula: b = a / R, wherein R is the dilution ratio; a is the standard gas concentration; b is the probe dilution ratio correction value.

[0103] In this embodiment, in order to realize comprehensive monitoring of the discharged flue gas, the flue gas parameter monitoring unit is also used to collect other flue gas parameter data of the sample gas, including flue gas flow, flue gas temperature, flue gas humidity, and flue gas oxygen content.

[0104] In the embodiment, in order to improve the stability of system operation, reduce the failure rate and repair frequency of the equipment, the sample collection and transmission device has the functions of filtering particulate matter, heating, heat preservation and back blowing purification, and through the above functions, the self-cleaning of the system can be realized, the service life of the equipment is prolonged, the heating temperature is not less than 120 DEG C, and is at least 10 DEG C higher than the dew point temperature of the flue gas.

[0105] In the embodiment, in order to improve the stability of the device operation, the material of the sample collection device should be selected from materials resistant to high temperature, corrosion, non-absorption and non-reaction with gaseous pollutants, and the material of the sample collection device is preferably stainless steel or titanium alloy. In order to facilitate the calibration of the system, the sample collection device has at least two sample transmission tubes, which are respectively used for collecting and transmitting sample gas and calibrating the whole system with standard gas.

[0106] In the embodiment, in order to prolong the service life of the analysis instrument and reduce the failure rate during operation, the pretreatment equipment includes sample filtration equipment and dehumidification condensation equipment, and the operation temperature of the dehumidification equipment is 4±2 DEG C.

[0107] The above shows and describes the basic principles and main features of the present application and the advantages of the present application. It should be understood by those skilled in the art that the present application is not limited by the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application, and various changes and improvements can be made without departing from the spirit and scope of the present application, and these changes and improvements all fall within the scope of the present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A dilution-based flue gas monitoring system for continuous monitoring of exhaust gas pollution sources, comprising a particulate matter monitoring unit, a gaseous pollutant monitoring unit, a flue gas parameter monitoring unit, and a data acquisition and processing unit, wherein the data acquisition and processing unit includes a sample acquisition and transmission device, pretreatment equipment, analytical instruments, data acquisition and transmission equipment, and auxiliary equipment, wherein, The data acquisition and transmission equipment is communicatively connected to the analytical instrument, particulate matter monitoring unit, gaseous pollutant monitoring unit, and flue gas parameter monitoring unit, respectively. Its key feature is that the data acquisition and processing unit further includes an analysis module and an execution module. The analysis module acquires flue gas pressure change values, constructs a BP neural network prediction model with flue gas pressure change values ​​as input and probe dilution ratio correction values ​​as output, and predicts probe dilution ratio correction values ​​through the model. The execution module is communicatively connected to the analysis module, acquires the predicted probe dilution ratio correction values, and inputs the probe dilution ratio correction values ​​into the analytical instrument to achieve automatic correction of the probe dilution ratio. The specific steps of the flue gas pressure change analysis module include: Step S1, Flue gas pressure change value acquisition: The flue gas pressure change value is acquired using the sensor in the flue gas parameter monitoring unit and transmitted to the data acquisition and processing unit; Step S2, Data Preprocessing: The collected flue gas pressure change values ​​are filtered, denoised, and calibrated. Step S3, Construction of BP neural network prediction model: The pre-processed flue gas pressure change value is used as the input data of BP neural network, and the probe dilution ratio correction value is used as the output data. A stable prediction model is obtained through training. Step S4, Model Prediction: Use the trained BP neural network model to predict the new flue gas pressure change value and obtain the probe dilution ratio correction value. Step S5, Execution module processing: The predicted probe dilution ratio correction value is transmitted to the execution module for corresponding correction operations; In step S3, the construction process of the BP neural network model is as follows: Suppose that the input layer has m neurons, the hidden layer has n neurons, the output layer has k neurons, the input vector is X, the output vector is Y, the output of the i-th hidden layer neuron is h(i), and the output of the j-th output layer neuron is y(j). The calculation formula from the input layer to the hidden layer is: h(i)=f(sum(w(j,i)*X(j)) + b(i)); Where w(j,i) represents the connection weight from the j-th neuron in the input layer to the i-th neuron in the hidden layer, b(i) represents the bias of the i-th neuron in the hidden layer, and f is the sigmoid function or the ReLU function; The calculation formula from the hidden layer to the output layer is: y(j)=g(sum(w(i,j)*h(i)) + b(j)) Where w(i,j) represents the connection weight from the i-th neuron in the hidden layer to the j-th neuron in the output layer, b(j) represents the bias of the j-th neuron in the output layer, and g is the activation function; The training process of a BP neural network includes two stages: forward propagation and backward propagation. During forward propagation, the input vector X is fed into the network to obtain the output vector Y. During backward propagation, the output error is calculated, and the connection weights and biases are updated using this error to minimize the network's output error. The error formula for a BP neural network is: E = 1 / 2 * sum((y(j) - Y(j))^2); Where Y(j) is the expected output of the j-th sample in the training set, and y(j) is the actual output of the network; the weight and bias update formulas of the BP neural network are: w(i,j)(t+1)=w(i,j)(t)-η*δ(j)*h(i) b(j)(t+1)=b(j)(t)-η*δ(j); w(j,i)(t+1)=w(j,i)(t)-η*δ(j)*X(i) b(i)(t+1)=b(i)(t)-η*δ(j); Where η is the learning rate, and δ(j) is the error term of the j-th neuron; The steps for constructing a BP neural network prediction model include: Step 1, Data Preprocessing: Normalize the acquired flue gas pressure change data to obtain a standardized dataset. Use the maximum-minimum normalization method to transform the data into the [0,1] interval; the formula is: ； Where x is the original data, x' is the normalized data, and min and max are the minimum and maximum values ​​in the original data, respectively; Step 2, design the structure of the BP neural network: The structure of a BP neural network includes an input layer, hidden layers, and an output layer; the number of nodes in the input layer equals the dimension of the input data, the number of nodes in the output layer equals the dimension of the output data, and the number of nodes in the hidden layer needs to be adjusted according to the actual situation; the formula is: Where f(x) is the sigmoid function or the ReLU function; Step 3: Divide the obtained standardized dataset into a training set and a test set, with the training set accounting for 70%–80% and the test set accounting for 20%–30%. Step 4: Train the BP neural network using the training set data, setting the maximum training period, target error, and learning rate. During training, use the backpropagation algorithm to update the weights to minimize the error function. The error function is: ; in, y k The label values ​​of the training set. These are the predicted values ​​from the BP neural network; Step 5, Model Validation: Test the trained neural network using a test set, and calculate the prediction error and accuracy; the prediction error is calculated using mean squared error (MSE) or mean absolute error (MAE); the formula is: ； ； in, y i The label values ​​for the test set. These are the predicted values ​​from the BP neural network; Step 6, Model Optimization: Based on the results of the validation process, adjust the structure, learning rate, and weight optimization algorithm of the neural network, and retrain and validate until the prediction error and accuracy of the model are less than the set threshold.

2. The dilution method flue gas monitoring system according to claim 1, characterized in that: The input data for the training set consists of flue gas pressure parameter data directly obtained through the flue gas parameter monitoring unit. The output data for the training set consists of the probe dilution ratio correction value corresponding to the directly obtained flue gas pressure parameter data. The steps for determining the probe dilution ratio correction value corresponding to the directly obtained flue gas pressure parameter data include: Set a safe threshold for flue gas pressure change and obtain the standard value of flue gas pressure under normal system operation. Read the real-time value of the flue gas pressure and calculate the difference between the real-time value of the flue gas pressure and the standard value of the flue gas pressure. Determine whether the difference between the real-time value of flue gas pressure and the standard value of flue gas pressure is within the safe threshold range of flue gas pressure change; When the difference between the real-time value of flue gas pressure and the standard value of flue gas pressure is outside the safe threshold range of flue gas pressure change, a correction method is adopted by introducing zero gas into the system to adjust the instrument zero point, calibrate the analyzer with medium concentration standard gas, and calculate the probe dilution ratio correction value using the formula: b=a / R, where R is the dilution ratio; a is the standard gas concentration; and b is the probe dilution ratio correction value.

3. The dilution method flue gas monitoring system according to claim 1, characterized in that: The flue gas parameter monitoring unit is also used to collect other flue gas parameter data of the sample gas, including flue gas flow rate, flue gas temperature, flue gas humidity, and flue gas oxygen content.

4. The dilution method flue gas monitoring system according to claim 1, characterized in that: The sample collection and transmission device has the functions of filtering particulate matter, heating, heat preservation and backflushing purification. The heating temperature is not less than 120℃ and is at least 10℃ higher than the flue gas dew point temperature.

5. The dilution method flue gas monitoring system according to claim 1, characterized in that: The sample collection device is made of stainless steel or titanium alloy.

6. The dilution method flue gas monitoring system according to claim 1, characterized in that: The sample acquisition device has at least two sample transfer tubes, one for the acquisition and transfer of sample gas and the other for the full system calibration of standard gas.

7. The dilution method flue gas monitoring system according to claim 1, characterized in that: The pretreatment equipment includes sample filtration equipment and dehumidification and condensation equipment. The operating temperature of the dehumidification equipment is 4±2℃.

Images