Power plant carbon emission online abnormal value real-time processing method, equipment and medium

Abstract

The invention discloses a power plant carbon emission online abnormal value real-time processing method, device and medium, and the method comprises the steps: obtaining monitoring data in CEMS carbon emission, the detection data comprising flue gas flow and carbon dioxide concentration; performing real-time online processing on the obtained monitoring data, and judging whether data exception exists or not; and if abnormity exists, establishing a mapping relation between the DCS data and the CEMS data through big data and a deep learning algorithm, and correcting or filling abnormal values of monitoring data in the CEMS by adopting DCS real-time operation data. According to the method, the mapping model relationship between the DCS data and the CEMS carbon emission online monitoring data is constructed through deep learning, the abnormal value of the monitoring data in the CEMS is corrected or filled by adopting the DCS real-time operation data, and a feasible scheme is provided for real-time processing of the CEMS carbon emission online monitoring abnormal value. The method can be widely applied to the field of environmental protection monitoring.

Description

Technical Field

[0001] The present invention relates to the field of environmental monitoring, and in particular to a method, equipment and medium for real-time processing of online abnormal values ​​of carbon emissions from power plants. Background Art

[0002] Thermal power generation is the main area of ​​carbon emissions. Accurately measuring the carbon emissions of thermal power generation is an important basis for enterprises to formulate energy-saving and carbon reduction strategies, reduce greenhouse gas emissions and participate in the carbon emissions trading market. The continuous emission monitoring system (CEMS) has become the standard for most fixed emission sources due to its real-time, accuracy, unchangeable and traceable characteristics. However, CEMS has been widely adopted in the online monitoring of carbon emissions in developed countries in Europe and the United States, while thermal power generation in my country is still in the pilot stage. CEMS carbon emission online monitoring directly measures the flue gas flow, carbon dioxide concentration, flue gas temperature, flue gas pressure, flue gas humidity and other parameters in carbon emissions by arranging online monitoring equipment at the outlet of fixed pollution sources of enterprises. Although the accuracy and real-time nature of the data are improved, the monitoring data is often prone to abnormalities due to the fact that the sensor equipment is easily affected by various factors such as physical environment changes, production conditions, production processes, equipment failures, etc., resulting in a decrease in the quality of CEMS carbon emission online monitoring data. Among them, the flue gas flow and carbon dioxide concentration parameters are the main variables that need to be solved urgently.

[0003] Patent application number 201710358297.2 discloses a missing value filling method and system based on deep learning, which belongs to post-statistical missing filling, fails to achieve online solution, and does not belong to the method of processing abnormal values ​​in online carbon emission monitoring; Patent application number 202011118064.3 discloses a flue gas online monitoring data preprocessing method, which identifies, screens and marks abnormal data, but does not involve post-identification processing, which is significantly different from real-time online processing. Patent application number 202410551730.4 discloses a power plant carbon emission accounting method, device and server based on big data, which belongs to the power plant carbon emission accounting data processing technology, but does not involve the relevant content of CEMS. After literature patent retrieval, screening and induction and summary, it was found that there is no public information on real-time online processing of abnormal values ​​in CEMS online carbon emission monitoring at home and abroad.

[0004] There is currently a lack of timely methods for the processing of abnormal carbon emission data after identification, especially the lack of real-time processing methods for abnormal values ​​in CEMS online carbon emission monitoring, which leads to poor real-time and accuracy of carbon emission measurement. In recent years, with the rapid development of big data and deep learning technology, its application in the field of carbon emission monitoring has gradually attracted attention. However, the real-time processing system for abnormal values ​​in online carbon emission monitoring based on big data is still in the initial exploration stage, and has not yet been applied to the real-time processing of abnormal values ​​in CEMS online carbon emission monitoring data. The real-time processing of abnormal values ​​in online CEMS carbon emission monitoring and the in-depth analysis of big data still need to be strengthened. Summary of the invention

[0005] In order to solve at least one of the technical problems existing in the prior art to a certain extent, the object of the present invention is to provide a method, device and medium for real-time processing of online abnormal values ​​of carbon emissions from power plants.

[0006] The first technical solution adopted by the present invention is:

[0007] A method for real-time processing of online abnormal values ​​of carbon emissions from power plants, comprising the following steps:

[0008] Obtain monitoring data of CEMS carbon emissions, the detection data mainly including flue gas flow and carbon dioxide concentration;

[0009] Process the acquired monitoring data in real time online and determine whether there are any data anomalies;

[0010] If there are any abnormalities, a mapping relationship between DCS data and CEMS data is constructed through big data and deep learning algorithms, and the DCS real-time operation data is used to correct or fill in the abnormal values ​​of the monitoring data in the CEMS.

[0011] Furthermore, the mapping relationship between DCS data and CEMS data is constructed by using big data and deep learning algorithms, and the abnormal values ​​of the monitoring data in CEMS are corrected or filled using the DCS real-time operation data, including:

[0012] Obtain DCS historical operation data and CEMS carbon emission online monitoring historical data to construct a data set; divide the data set into a training set and a test set;

[0013] Carry out data preprocessing and dimensionality reduction on the data set, calculate the Pearson correlation coefficient, and determine the correlation between the characteristic variable X and the target variable Y; DCS historical operation data is used as the characteristic variable X, and CEMS carbon emission online monitoring historical data is used as the target variable Y;

[0014] Use time series algorithms to build prediction models (RNN, LSTM, GRU and other deep learning algorithms), and use training sets to train prediction models;

[0015] When an outlier is detected in the monitoring data, the time point corresponding to the outlier is obtained; the DCS real-time data at the corresponding time is obtained according to the time point and used as the input of the trained prediction model to correct or fill in the outlier data after prediction.

[0016] Furthermore, the step of considering the delay time is included:

[0017] Define the delay time between the feature variable X and the target variable Y: define the target variable at time k as y(k) and the delay time as d i , then the corresponding characteristic variable X is x(kd i );

[0018] The mutual information method (MI) is used to calculate the degree of dependence between the feature variable X and the target variable Y;

[0019] Determine the delay time using the differential evolution algorithm (DE): The differential evolution algorithm optimizes the problem through iterative search and determines the optimal delay time d i , so that the model prediction effect is the best;

[0020] The optimized delay time is converted into actual time units (seconds) to obtain the actual delay time from feature variable X to target variable Y.

[0021] Furthermore, the acquisition of DCS historical operation data and CEMS carbon emission online monitoring historical data to construct a data set includes:

[0022] Obtain the DCS operation history data set Seq11, the flue gas flow history data set Seq21 and the carbon dioxide concentration history data set Seq22 measured by the CEMS;

[0023] Extract the DCS operation history data in the combustion process and flue gas treatment desulfurization process from the data set Seq11 to form the data set Seq12;

[0024] The data sets Seq12 and Seq21 are combined into a new data set Seq31; the data sets Seq12 and Seq22 are combined into a new data set Seq32.

[0025] Furthermore, the Pearson correlation coefficient calculation formula is as follows:

[0026]

[0027] In the formula, ρ X,Yis the Pearson correlation coefficient, X is the characteristic variable in DCS, Y is the target variable flue gas flow or carbon dioxide concentration, σ X is the standard deviation of the characteristic variable X, σ Y is the standard deviation of the characteristic variable Y; cov represents covariance;

[0028] Based on the Pearson correlation coefficient, the preset models Model1 and Model2 are used to calculate the delay time from each DCS variable to the flue gas flow and carbon dioxide concentration respectively. The data set Seq31 is processed into Seq41, and the data set Seq32 is processed into Seq42.

[0029] Furthermore, the use of a training set to train the prediction model includes:

[0030] In the dataset Seq41, the smoke flow rate is used as a label, and deep learning is used to build and train the first prediction model Modle3;

[0031] In the dataset Seq42, the carbon dioxide concentration is used as a label, and deep learning is used to build and train the second prediction model Model4;

[0032] The trained first prediction model is used to predict the flue gas flow in real time and process abnormal values; the trained second prediction model is used to predict the carbon dioxide concentration value in real time and process abnormal values.

[0033] Furthermore, both the first prediction model and the second prediction model are constructed using long short-term memory networks.

[0034] Furthermore, during the model training process, model score, root mean square error, and mean absolute error were used as measurement indicators.

[0035] The second technical solution adopted by the present invention is:

[0036] An electronic device comprises a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement a real-time processing method for online abnormal values ​​of carbon emissions from a power plant as described above.

[0037] The third technical solution adopted by the present invention is:

[0038] A computer-readable storage medium stores at least one instruction, at least one program, a code set or an instruction set, wherein the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by a processor to implement a method for real-time processing of online abnormal values ​​of carbon emissions from a power plant as described above.

[0039] The fourth technical solution adopted by the present invention is:

[0040] A computer program product or a computer program, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the above-mentioned method for real-time processing of online abnormal values ​​of carbon emissions from a power plant.

[0041] The beneficial effects of the present invention are as follows: the present invention constructs a mapping model relationship between DCS data and CEMS carbon emission online monitoring data through deep learning, and uses DCS real-time operation data to correct or fill in the abnormal values ​​of the monitoring data in CEMS, which provides a feasible solution for real-time processing of CEMS carbon emission online monitoring abnormal values, and can effectively improve the accuracy and real-time performance of CEMS carbon emission online monitoring data. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the embodiments of the present invention or the drawings of related technical solutions in the prior art are introduced below. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 is a schematic diagram of real-time processing of abnormal flue gas flow and carbon dioxide concentration proposed in an embodiment of the present invention;

[0044] Figure 2 is a carbon emission flow chart of a thermal power plant constructed in an embodiment of the present invention;

[0045] Figure 3 yes Figure 2 Schematic diagram of target variable and feature variable delay time;

[0046] Figure 4 is a flow chart of a method for real-time processing of abnormal values ​​of flue gas flow and carbon dioxide concentration proposed in an embodiment of the present invention;

[0047] Figure 5 yes Figure 4 Flowchart of the delay time calculation step in step;

[0048] Figure 6 is a heat map of characteristic variables in an embodiment of the present invention, and the numbers are Pearson correlation coefficients;

[0049] Figure 7It is a model training and testing iterative calculation diagram in an embodiment of the present invention;

[0050] Figure 8 It is a diagram showing the prediction effect of carbon dioxide concentration in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and are not to be construed as limitations of the present invention. For the step numbers in the following embodiments, they are only provided for the convenience of explanation, and the order between the steps is not limited in any way, and the execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0052] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., and orientations or positional relationships indicated are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0053] In the description of the present invention, "several" means one or more, "more" means more than two, "greater than", "less than", "exceed" etc. are understood as not including the number itself, and "above", "below", "within" etc. are understood as including the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0054] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention based on the specific content of the technical solution.

[0055] In view of the existing technical problems, the present invention proposes a real-time processing solution for abnormal values ​​of online monitoring of carbon emissions in power plants based on big data, combined with actual engineering conditions. Starting from the data of the distributed control system (DCS) of the power plant, the abnormal values ​​of flue gas flow and carbon dioxide concentration in CEMS are predicted and corrected or filled through the real-time data of DCS. At present, all power plants are equipped with DCS, which can realize centralized monitoring and decentralized control of the power generation process. Its main functions include data acquisition and processing, historical data recording and trend analysis, etc. After years of development, the system operation data obtained has high frequency and high reliability, which can be used to measure the operating status of the generator set and predict carbon emissions. Furthermore, the real-time operation data of DCS can timely control the operating status of the generator set, and the operating status of the generator set is closely related to carbon emissions. The abnormal values ​​of flue gas flow and carbon dioxide concentration in CEMS can be predicted and corrected or filled through the real-time data of DCS.

[0056] The present invention proposes a real-time processing solution for abnormal values ​​of online monitoring of carbon emissions in power plants based on big data, which mainly includes four parts, namely, a CEMS carbon emission data acquisition module, a real-time online data processing module, a pre-training module and a DCS data system acquisition module. The parameters collected by the CEMS carbon emission data acquisition module include flue gas flow, carbon dioxide concentration, flue gas temperature, flue gas pressure, flue gas humidity and other parameters. Among them, the most important parameters affecting carbon dioxide calculation are flue gas flow and carbon dioxide concentration. Therefore, the present invention mainly considers the real-time processing of abnormal values ​​of flue gas flow and carbon dioxide concentration; the parameters processed by the real-time online data processing module are flue gas flow and carbon dioxide concentration, and the abnormal value judgment method is used to judge whether it is abnormal (the judgment method can use threshold judgment). After being abnormal, the pre-trained model is used to correct or fill it; the pre-trained model uses DCS and CEMS data to train prediction models Model3 and Model4 for flue gas flow and carbon dioxide concentration respectively (the prediction model can use deep learning algorithms such as RNN, LSTM, GRU, etc. to deeply explore the characteristics of DCS and CEMS data). When the real-time online processing module receives an abnormal value request, the time point corresponding to the abnormal value is obtained, the time point corresponding to the DCS data is found, and the corresponding characteristic variables are found, and used as the input value of Model3 or Model4, the target value is predicted in real time, and fed back to the real-time online data processing module. The method and system for real-time processing of abnormal values ​​in online monitoring of carbon emissions from power plants based on big data proposed in the present invention can provide a feasible solution for real-time processing of abnormal values ​​in online monitoring of carbon emissions from CEMS, and can effectively improve the accuracy and real-time performance of online monitoring data of carbon emissions from CEMS.

[0057] Example 1

[0058] CEMS carbon emission online monitoring can obtain carbon emission rate in real time, but because the measuring equipment is easily affected by various factors such as physical environment changes, production conditions, production processes, equipment failures, etc., the measured data often have abnormal values ​​(including missing values), especially flue gas flow and carbon dioxide concentration parameters, which are the two most important parameters affecting the accurate calculation of carbon dioxide. At present, thermal power plants are equipped with distributed control systems (DCS), which can realize centralized monitoring and decentralized control of the power generation process. Its main functions include data acquisition and processing, historical data recording and trend analysis, etc. Through these functions, the DCS system can achieve goals such as automatic control, optimization of production processes, and improvement of production efficiency and quality. After years of development, the system operation data obtained by DCS has high frequency and high reliability, which can be used to measure the operating status of generator sets and predict carbon emissions. Correcting or filling the abnormal values ​​of flue gas flow and carbon dioxide concentration in CEMS through DCS real-time operation data will help improve the accurate calculation of carbon emissions. After investigation, it was found that there is a lack of literature and patents on this technology in the current field, and it is urgent to fill this technical gap.

[0059] In view of the problem that it is difficult to effectively process abnormal values ​​in CEMS carbon emission online monitoring in real time, this embodiment proposes a real-time processing method for abnormal values ​​in power plant carbon emission online monitoring based on big data. Through big data and deep learning algorithms, a mapping relationship between DCS data and CEMS data is constructed, and the flue gas flow and carbon dioxide concentration in CEMS are corrected and filled in real time to improve the real-time and accuracy of carbon emission accounting.

[0060] The present invention is based on engineering practice and uses DCS data to perform real-time correction and filling of the flue gas flow and carbon dioxide concentration in CEMS, including the following steps:

[0061] Step 1: Obtain the historical operation data of the DCS of the thermal power plant. The thermal power plant is a typical process production process in which the processes influence each other. The operating parameters can measure the operating status of the unit, and the operating status of the unit directly affects carbon emissions.

[0062] Step 2: Obtain the historical data of CEMS carbon emission online monitoring, mainly flue gas flow (Y1) and carbon dioxide concentration (Y2) data, and construct a data set with the data obtained in the first step.

[0063] Step 3: Perform data preprocessing and dimensionality reduction on the data set, calculate the Pearson correlation coefficient, and determine the correlation between the feature variable X and the target variable Y. When the Pearson correlation coefficient is in the interval (-1,-m]∪[m,1), select the variable as the auxiliary variable, where m is an empirically selected value.

[0064] Among them, the power production process is a typical process production process. Affected by the process flow, pipelines and measuring equipment, there is a certain delay time from the characteristic variable to the target variable. However, if the delay time is not considered and the prediction model can achieve better results, it can be considered not to calculate the delay time. Otherwise, the mutual information method (MI) and differential evolution (DE) algorithm are used to calculate the delay time from the auxiliary variable (X) to the target variable (Y) and reconstruct the data set. The delay time calculation steps are as follows:

[0065] 1) Define the delay time between the auxiliary variable and the output variable: define the target variable at time k as y(k) and the delay time as d i , then the corresponding auxiliary variable is x(kd i );

[0066] 2) Use mutual information MI to calculate the degree of dependence between feature variable X and target variable Y;

[0067] 3) Use DE algorithm to determine the delay time: DE algorithm optimizes the problem through iterative search and determines the optimal delay time d i , so that the model prediction effect is the best;

[0068] 4) Convert the optimized delay time into actual time units (seconds) to obtain the actual delay time from feature variable X to target variable Y.

[0069] Step 4: Using time series algorithms (RNN, LSTM, GRU and other deep learning algorithms), divide 80% of the data set into a training set and 20% into a test set to train the flue gas flow and carbon dioxide concentration prediction models respectively.

[0070] Step 5: Process flue gas flow and carbon dioxide concentration anomalies online in real time, and judge the abnormal values ​​of the flue gas flow and carbon dioxide concentration obtained in real time. Set the threshold before judging, and define the data exceeding the threshold or missing as abnormal values. Obtain the time point under the data, find the time point corresponding to the characteristic variable, obtain the DCS real-time data at the corresponding time, and use it as the input of the prediction model to correct or fill in the abnormal value data after prediction.

[0071] The following is a detailed case study with accompanying drawings.

[0072] Figure 1The real-time processing system for flue gas flow and carbon dioxide concentration anomalies proposed in this embodiment mainly includes four parts, namely, a CEMS carbon emission data acquisition module, a real-time online data processing module, a pre-training module and a DCS data system acquisition module. The parameters collected by the CEMS carbon emission data acquisition module include flue gas flow, carbon dioxide concentration, flue gas temperature, flue gas pressure, flue gas humidity and other parameters, among which the most important parameters are flue gas flow and carbon dioxide concentration. Therefore, the present invention mainly considers the abnormal values ​​of flue gas flow and carbon dioxide concentration; the parameters processed by the real-time online data processing module are flue gas flow and carbon dioxide concentration, and the abnormal value judgment method is used to judge whether it is abnormal (the judgment method can use threshold judgment). After being abnormal, the pre-training model is used to perform Correction or filling; The pre-trained model uses DCS and CEMS data to train the prediction models Model 3 and Model 4 for flue gas flow and carbon dioxide concentration respectively (the prediction model can use deep learning algorithms such as RNN, LSTM, GRU, etc. to deeply explore the characteristics of DCS and CEMS data). When receiving an abnormal value request from the real-time online processing module, obtain the time point corresponding to the abnormal value, find the time point corresponding to the DCS data, and find the corresponding characteristic variables, and use them as the input value of Model 3 or Model 4 to predict the target value in real time, and feed it back to the real-time online data processing module to make a real-time prediction of the flue gas flow or carbon dioxide concentration. After obtaining complete carbon emission data, use formula (1) to calculate the carbon dioxide emission rate. Figure 2 The figure shows the installation diagram of CEMS carbon emission online monitoring equipment. The carbon emission rate is calculated by formula (1):

[0073]

[0074] Where M is the carbon emission rate, P ct is the flue gas pressure, V S is the flue gas volume flow rate, w is the flue gas humidity, is the carbon dioxide concentration, R is the standard molar gas constant, and T is the flue gas temperature.

[0075] According to literature research, flue gas flow and carbon dioxide concentration are the two most critical parameters affecting carbon emission rate. Since the sensor is easily affected by the environment, outliers (including missing values) are prone to occur. Traditional outlier processing methods cannot solve this problem well and the processing effect is poor. Prediction through DCS operation data has become a new method.

[0076] Depend on Figure 2It can be seen that the production process of thermal power plants is long, and each process has a lot of operating data, so it is difficult to mine the mapping relationship with CEMS carbon emission data from the massive DCS data. Optionally, when the DCS data provided by the power plant is too high in dimension, it can be started from the combustion process in the main process of carbon emission and the desulfurization process in the flue gas treatment process to avoid the influence of redundant variables on the model.

[0077] Figure 3 The delay time diagram of the target variable and the characteristic variable is shown. Due to the influence of the process flow, pipelines and measuring equipment, there is a certain delay time from the characteristic variable to the target variable. If the variable delay time has a significant impact on the result, it is necessary to calculate the target variable (y) and each characteristic variable (x1, x2, ... x i ) Delay time (d1, d2, ...d i ), find out the time correspondence, the calculation process is as follows Figure 5 shown.

[0078] Figure 4 The flowchart of the method for real-time processing of abnormal flue gas flow and carbon dioxide concentration proposed by the present invention is as follows:

[0079] Step S11: obtaining a power plant unit system DCS operation history data set Seq11, a flue gas flow history data set Seq21 and a carbon dioxide concentration history data set Seq22 measured by a CEMS;

[0080] Step S12: Considering the DCS data that the power plant can provide, when the data dimension is too high, optionally, extract the DCS operation history data in the combustion process and the flue gas treatment desulfurization process from the data set Seq11 to form the data set Seq12;

[0081] Step S13: The processed Seq12 and Seq21 form a new data set Seq31, and the processed Seq12 and Seq22 form a new data set Seq32, and data preprocessing and dimensionality reduction are carried out;

[0082] Step S14: Calculate the Pearson correlation coefficient, determine the auxiliary variables, and use Model 1 and Model 2 to calculate the delay time from each DCS variable to the flue gas flow rate and carbon dioxide concentration. Seq31 is processed into Seq41, and Seq32 is processed into Seq42. The Pearson correlation coefficient calculation formula is as follows:

[0083]

[0084] In the formula, ρ X,Y is the Pearson correlation coefficient, X is the characteristic variable in DCS, Y is the target variable flue gas flow or carbon dioxide concentration, σ Xis the standard deviation of the feature variable X, σ Y is the standard deviation of the characteristic variable Y. After calculating the Pearson correlation coefficient, the correlation between X and Y is determined, ρ X,Y (-1,-m]∪[m,1) is selected as the auxiliary variable, where m is an empirically selected value.

[0085] The steps for calculating the delay time of X and Y are as follows Figure 5 As shown, (the delay time calculation model from characteristic variable to flue gas flow is Model1, and the delay time calculation model from characteristic variable to carbon dioxide concentration is Model2) the specific calculation process is as follows:

[0086] Step S21: Define the delay time between the auxiliary variable and the output variable: define the target variable at time k as y(k), and the delay time as d i , then the corresponding auxiliary variable is x(kd i );

[0087] Step S22: Calculate the degree of dependence between the feature variable X and the target variable Y using mutual information (MI):

[0088] Step S23: Using the differential evolution (DE) algorithm to solve the optimal delay time of the feature variable X(k) and the target variable Y: The differential evolution algorithm optimizes the problem through iterative search and determines the optimal delay time d i , so that the model prediction effect is the best;

[0089] Step S24: Convert the optimized delay time into actual time units (seconds) to obtain the actual delay time from each feature variable to the target variable.

[0090] Step S15: In the data set Seq41, the flue gas flow rate is used as a label, and the prediction model Model3 is constructed by using deep learning; in the data set Seq42, the carbon dioxide concentration is used as a label, and the prediction model Model4 is constructed by using deep learning;

[0091] Step S16: Model 3 and Model 4 predict the flue gas flow rate and carbon dioxide concentration in real time and process abnormal values.

[0092] This implementation case uses the carbon dioxide concentration as the target value and uses the actual production data of DCS and CEMS for modeling and prediction. In this case, the relevant characteristic variables are directly found from the DCS raw data provided by a power plant. The actual variable parameters provided by the power plant are shown in Table 1. The Pearson correlation coefficient between the variables is calculated, and 10 variables are finally selected. The calculation results are shown in Figure 6 As shown, after sorting, the data set Seq42 is obtained.

[0093] Table 1 DCS characteristic variables

[0094]

[0095]

[0096] When constructing the carbon dioxide concentration prediction model Model 4, deep learning algorithms are given priority. This implementation case uses the Long Short-Term Memory (LSTM) network to build a prediction model, and then judges the prediction performance of the model. If the prediction result is not ideal, the calculation of the delay time is considered. LSTM is an advanced variant of the Recurrent Neural Network (RNN), which effectively solves the RNN gradient vanishing and gradient explosion problems, and performs well in processing time series data. It controls the flow of information by introducing a gating mechanism (including input gate, forget gate, and output gate), thereby effectively capturing long-distance dependencies. Its model is as follows:

[0097] 1) Input gate:

[0098] i t =σ(W i ×[h t-1 ,X t ]+b i ) (3)

[0099] 2) Forget Gate:

[0100] f t =σ(W f ×[h t-1 ,X t ]+b f ) (4)

[0101] 3) Output gate:

[0102] o t =σ(W o ×[h t-1 ,X t ]+b o ) (5)

[0103] 4) Update unit:

[0104] c t =f t ×c t-1 +i t *tanh(W c ×[h t-1 ,X t ]+b c ) (6)

[0105] 5) Hidden state:

[0106] h t =o t ×tanh(c t ) (7)

[0107] In the formula, σ and tanh are activation functions, W i ,W f ,W o ,W c is the weight parameter, b i ,b f ,b o ,b c is the bias constant.

[0108] In order to measure the effect of modeling, the model score (R 2 ), Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) are used as measurement indicators. Model score R 2 It can measure the effect of fitting, and is generally between 0 and 1. The closer the value is to 1, the better the fitting effect is. The calculation formula is shown in (8). RMSE and MAE can measure the ability of the model calculation results to be close to the true value. The smaller the value, the higher the model calculation accuracy is. The calculation formulas are shown in (9) and (10).

[0109] 1) Model score:

[0110]

[0111] 2) Root mean square error:

[0112]

[0113] 3) Mean absolute error:

[0114]

[0115] The dataset Seq42 is divided into 80% training set and 20% test set, and the training and test loss values ​​change as follows Figure 7 As shown in the figure, it can be seen that when the iteration reaches 50 times, the loss values ​​of the training set and the test set basically no longer change, and the final loss values ​​of the training set and the test set are 0.06625283 and 0.02200199 respectively. Figure 8 The graph of the measured values ​​of carbon dioxide concentration on the test set and the LSTM predicted values ​​shows that the measured values ​​and the predicted values ​​have a good fit. 2The value is 0.96290840589032, the RMSE value is 0.0027490829786373343, and the MAE value is 0.0013463517760870375, which shows that the LSTM model can well predict the carbon dioxide concentration of this unit. Using the Model4 model for real-time prediction will be able to effectively handle abnormal values.

[0116] From the results obtained from the test set of Model 4, it can be seen that using DCS data as input, the carbon dioxide concentration can be effectively predicted in real time. The prediction results can be used to correct or fill in the abnormal values ​​of carbon dioxide concentration and realize the real-time processing of CEMS carbon emission online abnormal values.

[0117] In summary, compared with the prior art, the method of the present invention has at least the following advantages and beneficial effects:

[0118] (1) The present invention is based on the abnormal values ​​existing in the online measurement data of CEMS carbon emissions, and innovatively proposes a real-time processing method for online abnormal values ​​of power plant carbon emissions based on big data. Existing abnormal value processing methods are often offline processing and lack timeliness. The method proposed in the present invention can process abnormal values ​​of flue gas flow and carbon dioxide concentration in real time.

[0119] (2) The present invention uses deep learning to construct a mapping model relationship between DCS data and CEMS carbon emission online monitoring data. Taking carbon dioxide concentration as an implementation case, the loss values ​​of the training set and the test set are 0.06625283 and 0.02200199, respectively. Using this model to predict, correct and fill in outliers can effectively improve the quality of CEMS carbon emission online monitoring data.

[0120] (3) Aiming at the delay problem of DCS data and CEMS data in power plants, the present invention proposes a method for calculating the delay time using the mutual information method (MI) and the differential evolution algorithm (DE).

[0121] (4) The prediction model can update model parameters based on system operation data to avoid degradation of model performance.

[0122] (5) Compared with the traditional prediction method, the implementation case of the present invention shows that the constructed prediction model can effectively adapt to the data of generator sets with frequent operating conditions, and the fitting degree can reach 0.96290840589032;

[0123] Example 2

[0124] An embodiment of the present invention further provides an electronic device, the electronic device comprising a processor and a memory, the memory storing at least one instruction, at least one program, a code set or an instruction set, the at least one instruction, the at least one program, the code set or the instruction set being loaded and executed by the processor to implement the following Figure 4 A real-time processing method for online abnormal values ​​of carbon emissions from power plants is shown.

[0125] It is understood that the memory may include a random access memory (RAM) or a read-only memory (ROM). Optionally, the memory includes a non-transitory computer-readable storage medium. The memory may be used to store instructions, programs, codes, code sets, or instruction sets. The memory may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function, instructions for implementing the above-mentioned various method embodiments, etc.; the data storage area may store data created according to the use of the server, etc.

[0126] The processor may include one or more processing cores. The processor uses various interfaces and lines to connect the various parts of the entire server, and executes various functions of the server and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory, and calling data stored in the memory. Optionally, the processor can be implemented in at least one hardware form of digital signal processing (DSP), field programmable gate array (FPGA), and programmable logic array (PLA). The processor can integrate one or a combination of a central processing unit (CPU) and a modem. Among them, the CPU mainly processes the operating system and application programs; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor, but implemented separately through a chip.

[0127] Since the electronic device is an electronic device corresponding to a method for real-time processing of online abnormal values ​​of carbon emissions from a power plant in an embodiment of the present invention, and the principle of solving the problem by the electronic device is similar to that of the method, the implementation of the electronic device can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0128] Example 3

[0129] The embodiment of the present invention further provides a computer-readable storage medium, wherein the storage medium stores at least one instruction, at least one program, code set or instruction set, and the at least one instruction, the at least one program, the code set or instruction set is loaded and executed by a processor to implement the following Figure 4 A real-time processing method for online abnormal values ​​of carbon emissions from power plants is shown.

[0130] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0131] Since the storage medium is the storage medium corresponding to a method for real-time processing of online abnormal values ​​of carbon emissions from a power plant in an embodiment of the present invention, and the principle of solving the problem by the storage medium is similar to that of the method, the implementation of the storage medium can refer to the implementation process of the above-mentioned method embodiment, and the repeated parts will not be repeated.

[0132] Example 4

[0133] In some possible implementations, various aspects of the method of the embodiment of the present invention can also be implemented in the form of a program product, which includes a program code. When the program product is run on a computer device, the program code is used to enable the computer device to execute the steps of a real-time processing method for online abnormal values ​​of carbon emissions from a power plant according to various exemplary embodiments of the present application described above in this specification. Among them, the executable computer program code or "code" used to execute various embodiments can be written in a high-level programming language such as C, C++, C#, Smalltalk, Java, JavaScript, Visual Basic, structured query language (e.g., Transact-SQL), Perl, or in various other programming languages.

[0134] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0135] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0136] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable ordinary technicians in the field to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent changes or modifications made based on the essence of the content of the present invention should be included in the protection scope of the present invention.

Claims

1. A real-time processing method for online abnormal values ​​of carbon emissions from power plants, characterized in that: The following steps are involved: Acquire monitoring data of CEMS carbon emissions, wherein the monitoring data includes flue gas flow and carbon dioxide concentration; Process the acquired monitoring data in real time online and determine whether there are any data anomalies; If there are any abnormalities, a mapping relationship between DCS data and CEMS data is constructed through big data and deep learning algorithms, and the DCS real-time operation data is used to correct or fill in the abnormal values ​​of the monitoring data in the CEMS.

2. A method for real-time processing of online abnormal values ​​of carbon emissions from power plants according to claim 1, characterized in that: The mapping relationship between DCS data and CEMS data is constructed by using big data and deep learning algorithms, and the abnormal values ​​of the monitoring data in CEMS are corrected or filled using the DCS real-time operation data, including: Obtain DCS historical operation data and CEMS carbon emission online monitoring historical data to build a data set; Carry out data preprocessing and dimensionality reduction on the data set, calculate the Pearson correlation coefficient, and determine the correlation between the characteristic variable X and the target variable Y; DCS historical operation data is used as the characteristic variable X, and CEMS carbon emission online monitoring historical data is used as the target variable Y; Use time series algorithms to build prediction models and use data sets to train prediction models; When an outlier is detected in the monitoring data, the time point corresponding to the outlier is obtained; the DCS real-time data at the corresponding time is obtained according to the time point and used as the input of the trained prediction model to correct or fill in the outlier data after prediction.

3. A method for real-time processing of online abnormal values ​​of carbon emissions from power plants according to claim 2, characterized in that: Also includes steps to account for delay time: Define the delay time between the feature variable X and the target variable Y: define the target variable at time k as y(k) and the delay time as d i , then the corresponding characteristic variable X is x(kd i ); The mutual information method is used to calculate the degree of dependence between the feature variable X and the target variable Y; Use differential evolution algorithm to determine the delay time: Differential evolution algorithm optimizes the problem through iterative search and determines the optimal delay time d i , so that the model prediction effect is the best; The optimized delay time is converted into actual time units to obtain the actual delay time from the feature variable X to the target variable Y.

4. A method for real-time processing of online abnormal values ​​of carbon emissions from power plants according to claim 2, characterized in that: The acquisition of DCS historical operation data and CEMS carbon emission online monitoring historical data to construct a data set includes: Obtain the DCS operation history data set Seq11, the flue gas flow history data set Seq21 and the carbon dioxide concentration history data set Seq22 measured by the CEMS; Extract the DCS operation history data in the combustion process and flue gas treatment desulfurization process from the data set Seq11 to form the data set Seq12; The data sets Seq12 and Seq21 are combined into a new data set Seq31; the data sets Seq12 and Seq22 are combined into a new data set Seq32.

5. A method for real-time processing of online abnormal values ​​of carbon emissions from power plants according to claim 4, characterized in that: The Pearson correlation coefficient calculation formula is as follows: In the formula, ρ X,Y is the Pearson correlation coefficient, X is the characteristic variable in DCS, Y is the target variable flue gas flow or carbon dioxide concentration, σ X is the standard deviation of the feature variable X, σ Y is the standard deviation of the characteristic variable Y; cov represents covariance; Based on the Pearson correlation coefficient, the preset models Model1 and Model2 are used to calculate the delay time from each DCS variable to the flue gas flow and carbon dioxide concentration respectively. The data set Seq31 is processed into Seq41, and the data set Seq32 is processed into Seq42.

6. A method for real-time processing of online abnormal values ​​of carbon emissions from power plants according to claim 5, characterized in that: The method of using a data set to train a prediction model includes: In the dataset Seq41, the flue gas flow rate is used as a label, and the first prediction model Model3 is constructed and trained using deep learning; in the dataset Seq42, the carbon dioxide concentration is used as a label, and the second prediction model Model4 is constructed and trained using deep learning; The trained first prediction model is used to predict the flue gas flow in real time and process abnormal values; the trained second prediction model is used to predict the carbon dioxide concentration value in real time and process abnormal values.

7. A method for real-time processing of online abnormal values ​​of carbon emissions from power plants according to claim 6, characterized in that: The first prediction model and the second prediction model are both constructed using long short-term memory networks.

8. A method for real-time processing of online abnormal values ​​of carbon emissions from power plants according to claim 7, characterized in that: In the process of training the model, model score, root mean square error and mean absolute error are used as measurement indicators.

9. An electronic device, characterized in that: The electronic device includes a processor and a memory, wherein the memory stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: The storage medium stores at least one instruction, at least one program, a code set or an instruction set, and the at least one instruction, the at least one program, the code set or the instruction set is loaded and executed by the processor to implement the method according to any one of claims 1 to 8.

Images