A method and related device for predicting nitrogen oxide emissions in the waste incineration industry

Abstract

The present invention provides a method and related device for predicting nitrogen oxide emissions in the waste incineration industry, belonging to the field of environmental protection technology. The method of the present invention performs feature importance analysis on all preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry; performs empirical mode decomposition on the key features closely related to nitrogen oxide emissions in the waste incineration industry and continuous nitrogen oxide emission data in the waste industry to obtain intrinsic mode function components and residual components; constructs a training set and a validation set by using the obtained intrinsic mode function components and residual components; trains an LSTM model according to the training set to obtain a trained LSTM model; uses the trained LSTM model to predict the data in the validation set and compares it with the continuous nitrogen oxide emission data in the waste industry to obtain the prediction result of nitrogen oxide emissions in the waste incineration industry. The present invention solves the problem of low accuracy in predicting nitrogen oxide emissions in the waste incineration industry.

Description

Technical Field

[0001] The present invention belongs to the technical field of environmental protection, and particularly relates to a method and related device for predicting nitrogen oxide emissions in the waste incineration industry. Background Art

[0002] With the increasingly strict environmental protection regulations and the continuous improvement of the public's environmental awareness, the waste incineration industry is facing increasing emission reduction pressure and regulatory requirements. In order to reduce nitrogen oxide emissions, the waste incineration industry needs to take a series of effective emission reduction measures, such as optimizing the combustion process and installing advanced denitration equipment. However, the implementation of these measures often requires a large amount of capital and resources. Without accurate nitrogen oxide emission prediction as a guide, the waste industry may over-invest or have poor emission reduction effects during the emission reduction process, resulting in increased operating costs and decreased economic benefits.

[0003] Currently, machine learning technology has been widely applied in multiple industries. However, in the field of environmental protection, especially in the waste incineration industry, the application of machine learning is still insufficient. Traditional prediction methods that rely on simple statistical models or fixed parameter models are difficult to accurately capture the dynamic change rules of nitrogen oxide emissions, resulting in large prediction errors. This limitation is particularly prominent in the environmental protection field because the data in the environmental protection industry usually has high complexity and dynamic variability, which poses great challenges to the accuracy and stability of the model.

[0004] For example, in the process of municipal solid waste incineration, the sources of waste are extensive and the composition is complex. Due to the lack of refined waste sorting, there is significant uncertainty in the composition of the waste entering the incinerator. Different waste components will undergo different physical and chemical reactions during incineration, which directly affects the types and quantities of pollutant emissions, further increasing the difficulty of emission prediction. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and related device for predicting nitrogen oxide emissions in the waste incineration industry, which is used to solve the problem of low accuracy in predicting nitrogen oxide emissions in the existing waste incineration industry.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for predicting nitrogen oxide emissions in the waste incineration industry, including the following steps:

[0008] Obtain continuous nitrogen oxide emission data, related flue gas parameter data, and production monitoring point data;

[0009] Preprocess all the obtained data to remove abnormal data;

[0010] Perform feature importance analysis on all preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry;

[0011] Perform empirical mode decomposition on the key features closely related to nitrogen oxide emissions in the waste incineration industry and continuous nitrogen oxide emission data in the waste industry to obtain intrinsic mode function components and residual components;

[0012] Construct a training set and a validation set using the obtained intrinsic mode function components and residual components;

[0013] Train the LSTM model according to the training set to obtain a trained LSTM model;

[0014] Use the trained LSTM model to predict the data in the validation set and compare it with the continuous nitrogen oxide emission data in the waste industry to obtain the nitrogen oxide emission prediction result in the waste incineration industry.

[0015] A further improvement of the present invention lies in that in the step of preprocessing the obtained data to remove abnormal data, all the obtained data are preprocessed. First, abnormal data are removed to obtain data after removing abnormalities, and then the emission dynamic balance algorithm is used to perform data stationarity processing on the data after removing abnormalities to obtain stationary data.

[0016] A further improvement of the present invention lies in that the specific steps of the emission dynamic balance algorithm include:

[0017] Analyze the time series characteristics of all the obtained data in real time to obtain a number of dynamic features;

[0018] Perform normalization processing on the obtained number of dynamic features;

[0019] Combine the normalized number of dynamic features to obtain the overall fluctuation magnitude of all the obtained data;

[0020] Perform data stationarity processing on the overall fluctuation magnitude of all the obtained data to obtain the original stationary data;

[0021] Perform real-time monitoring and dynamic adjustment on the original stationary data until the overall fluctuation magnitude of all the obtained data reaches the expected fluctuation range to obtain the final stationary data.

[0022] A further improvement of the present invention lies in that in the step of combining the normalized number of dynamic features to obtain the overall fluctuation magnitude of all the obtained data, the calculation formula for the overall fluctuation magnitude of all the obtained data is:

[0023]

[0024] Where, The overall fluctuation magnitude of all the acquired data are the weight factors corresponding to each dynamic feature, and the sum of the five weight factors equals 1, where t is the trend change value after normalization at time where t is the periodic fluctuation value after normalization at time where t is the noise level magnitude after normalization at time where t is the real-time data change rate after normalization at time is the operating condition, and its value is 0 or 1, , , , and are the dynamic adjustment parameters, t and the dynamic adjustment parameter at time , , , (t) and are set according to the volatility index;

[0025] In the step of performing data stationarity processing on the overall fluctuation magnitude of all the acquired data, the calculation formula for performing data stationarity processing on the overall fluctuation magnitude of all the acquired data is:

[0026]

[0027] where is the data after data stationarity processing at time t , i is a natural number, is the size of the smoothing window, which is used to control the number of data points participating in the smoothing calculation, is the smoothing coefficient, which is used to adjust the overall fluctuation and its influence degree on the smoothing process, where t is the data before data stationarity processing at time j is an index, indicating the forward shift from the current time by j time periods, is the overall fluctuation magnitude of all the acquired data, where t is the overall fluctuation magnitude of all the acquired data at time

[0028] A further improvement of the present invention lies in that in the step of performing feature importance analysis on all preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry, an emission pattern mining algorithm is specifically used to perform feature importance analysis on all preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry.

[0029] A further improvement of the present invention lies in that the specific steps of the emission pattern mining algorithm include:

[0030] Construct a tree model for emission pattern mining;

[0031] Use all preprocessed data to train the tree model for emission pattern mining to obtain a trained tree model for emission pattern mining;

[0032] Calculate the SHAP value according to the trained tree model for emission pattern mining;

[0033] According to the calculated SHAP value, determine the contribution of all preprocessed data to the prediction result of nitrogen oxide emissions in the waste incineration industry, sort the contribution sizes, and obtain key features closely related to nitrogen oxide emissions in the waste incineration industry.

[0034] A further improvement of the present invention lies in that in the step of constructing a training set and a validation set by using the obtained intrinsic mode function components and residual components, specifically, 80% of the obtained intrinsic mode function components and residual components are used as the training set, and 20% are used as the validation set.

[0035] In a second aspect, the present invention provides a nitrogen oxide emission prediction system for the waste incineration industry, including a data acquisition module, a data preprocessing module, a feature importance analysis module, an empirical mode decomposition module, a data set construction module, a model training module, and a nitrogen oxide emission prediction module for the waste incineration industry;

[0036] The data acquisition module is used to acquire continuous high-frequency nitrogen oxide emission data, relevant flue gas parameter data, and production monitoring point data;

[0037] The data preprocessing module is used to preprocess all acquired data and remove abnormal data;

[0038] The feature importance analysis module is used to perform feature importance analysis on all preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry;

[0039] The empirical mode decomposition module is used to perform empirical mode decomposition on the key features closely related to nitrogen oxide emissions in the waste incineration industry and continuous nitrogen oxide emission data in the waste industry to obtain intrinsic mode function components and residual components;

[0040] The dataset construction module is used to construct a training set and a validation set by using the obtained intrinsic mode function components and residual components;

[0041] The module training module is used to train the LSTM model according to the training set to obtain a trained LSTM model;

[0042] The nitrogen oxide emission prediction for the waste incineration industry is used to predict the data in the validation set by using the trained LSTM model and compare it with the continuous nitrogen oxide emission data of the waste industry to obtain the nitrogen oxide emission prediction result for the waste incineration industry.

[0043] In a third aspect, the present invention provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the nitrogen oxide emission prediction method for the waste incineration industry introduced above are implemented.

[0044] In a fourth aspect, the present invention provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the nitrogen oxide emission prediction method for the waste incineration industry introduced above are implemented.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The present invention belongs to an improved invention. Compared with the existing nitrogen oxide emission prediction methods for the waste incineration industry, on the one hand, when predicting the nitrogen oxide emissions in the waste incineration industry, the present invention first performs a feature importance analysis on all the preprocessed data to obtain key features closely related to the nitrogen oxide emissions in the waste incineration industry. Compared with the traditional feature selection methods based on information gain and mutual information, it can more accurately screen out the features that have a key impact on the nitrogen oxide emission prediction, thereby improving the accuracy of the nitrogen oxide emission prediction and effectively solving the problem of low accuracy of the nitrogen oxide emission prediction in the existing technology for the waste incineration industry.

[0047] Furthermore, the present invention also discloses that in the step of preprocessing all the obtained data to remove abnormal data, first preprocess all the obtained data to remove abnormal data to obtain the data after removing abnormalities, and then use the emission dynamic balance algorithm to perform data stationarity processing on the data after removing abnormalities to obtain stationary data. By performing data stationarity processing on the data after removing abnormalities, the data fluctuations can be reduced, the stationarity of the data can be improved, and further the accuracy of the nitrogen oxide emission prediction in the waste incineration industry can also be improved.

[0048] Furthermore, the present invention also discloses real-time monitoring and dynamic adjustment of the original stability data until the overall fluctuation magnitude of all the obtained data reaches the expected fluctuation range, thereby obtaining the final stability data. Real-time monitoring and dynamic adjustment of the original stability data can ensure that the data remains stable in the complex and changeable waste incineration environment, providing reliable data for subsequent nitrogen oxide emission prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 is a flowchart of the method for predicting nitrogen oxide emissions in the waste incineration industry of the present invention;

[0050] Figure 2 is a schematic diagram of the system for predicting nitrogen oxide emissions in the waste incineration industry of the present invention;

[0051] Figure 3 is a schematic structural diagram of the electronic device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] To further understand the content of the present invention, the following provides a detailed description of the present invention with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely for explaining the present invention rather than limiting it.

[0053] The method for predicting nitrogen oxide emissions in the waste incineration industry proposed by the present invention uses the trained LSTM model to predict the data in the validation set and compares it with the continuous nitrogen oxide emission data in the waste industry, thereby obtaining the prediction result of nitrogen oxide emissions in the waste incineration industry. Compared with the prior art, the present invention effectively solves the problem of low accuracy in predicting nitrogen oxide emissions in the waste incineration industry in the prior art.

[0054] Example 1:

[0055] The flowchart of the method for predicting nitrogen oxide emissions in the waste incineration industry of the present invention is as Figure 1 shown. The method for predicting nitrogen oxide emissions in the waste incineration industry of the present invention includes the following steps:

[0056] S1. Obtain continuous high-frequency nitrogen oxide emission data, related flue gas parameter data, and production monitoring point data.

[0057] S2. Preprocess all the obtained data to remove abnormal data.

[0058] S3. Conduct feature importance analysis on all the preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry.

[0059] S4. Perform empirical mode decomposition on the key features closely related to the nitrogen oxide emissions in the waste incineration industry and the continuous nitrogen oxide emission data of the waste industry to obtain the intrinsic mode function components and the residual components.

[0060] S5. Use the obtained intrinsic mode function components and residual components to construct the training set and the validation set.

[0061] S6. Train the LSTM model according to the training set to obtain the trained LSTM model.

[0062] S7. Use the trained LSTM model to predict the data in the validation set and compare it with the continuous nitrogen oxide emission data of the waste industry to obtain the prediction result of the nitrogen oxide emissions in the waste incineration industry.

[0063] Embodiment 2:

[0064] The schematic diagram of the nitrogen oxide emission prediction system for the waste incineration industry of the present invention is as Figure 2 shown. The nitrogen oxide emission prediction system for the waste incineration industry of the present invention includes a data acquisition module, a data preprocessing module, a feature importance analysis module, an empirical mode decomposition module, a data set construction module, a model training module, and a nitrogen oxide emission prediction module for the waste incineration industry.

[0065] Among them, the data acquisition module is used to acquire continuous high-frequency nitrogen oxide emission data, relevant flue gas parameter data, and production monitoring point data.

[0066] The data preprocessing module is used to preprocess all the acquired data and clear the abnormal data.

[0067] The feature importance analysis module is used to perform feature importance analysis on all the preprocessed data to obtain the key features closely related to the nitrogen oxide emissions in the waste incineration industry.

[0068] The empirical mode decomposition module is used to perform empirical mode decomposition on the key features closely related to the nitrogen oxide emissions in the waste incineration industry and the continuous nitrogen oxide emission data of the waste industry to obtain the intrinsic mode function components and the residual components.

[0069] The data set construction module is used to construct the training set and the validation set by using the obtained intrinsic mode function components and residual components.

[0070] The module training module is used to train the LSTM model according to the training set to obtain the trained LSTM model.

[0071] The nitrogen oxide emission prediction in the waste incineration industry is used to predict the data in the validation set by using the trained LSTM model, and compare it with the continuous nitrogen oxide emission data in the waste industry to obtain the nitrogen oxide emission prediction results in the waste incineration industry.

[0072] Embodiment 3:

[0073] The method for predicting nitrogen oxide emissions in the waste incineration industry of the present invention includes the following steps:

[0074] S1. Obtain continuous high-frequency nitrogen oxide emission data, relevant flue gas parameter data, and production monitoring point data.

[0075] First, obtain continuous high-frequency nitrogen oxide emission data, relevant flue gas parameter data, and production monitoring point data (where continuous means that data is obtained every minute and the obtained data is continuous). Among them, the relevant flue gas parameter data includes flue gas temperature, flue gas pressure, flue gas flow rate, and flue gas oxygen content. The production monitoring point data includes equipment operation status, equipment liquid level data, and equipment power data.

[0076] In this embodiment, CEMS (Continuous Emission Monitoring Systems) is specifically used to obtain continuous high-frequency nitrogen oxide emission data and relevant flue gas parameter data, and DCS (Distributed Control System) is specifically used to obtain production monitoring point data.

[0077] S2. Preprocess all the obtained data to remove abnormal data.

[0078] Then preprocess all the obtained data to remove abnormal data. Among them, abnormal data includes data during shutdown, data of equipment failures, data during equipment maintenance, data during equipment calibration, and maximum and minimum values in high-frequency nitrogen oxide emission data, relevant flue gas parameter data, and production monitoring point data.

[0079] In this step, when preprocessing all the obtained data, first remove abnormal data to obtain data after removing abnormal data, and then use the emission dynamic balance algorithm to perform data stationarity processing on the data after removing abnormal data to obtain stationary data. The specific steps of the emission dynamic balance algorithm include:

[0080] A. Analyze the time series characteristics of all the obtained data in real time to obtain several dynamic features;

[0081] B. Perform normalization processing on the obtained several dynamic features;

[0082] C. Combine several normalized dynamic features to obtain the overall fluctuation magnitude of all the acquired data;

[0083] D. Perform data stationarity processing on the overall fluctuation magnitude of all the acquired data to obtain the original stationary data;

[0084] E. Perform real-time monitoring and dynamic adjustment on the original stationary data until the overall fluctuation magnitude of all the acquired data reaches the expected fluctuation range to obtain the final stationary data.

[0085] Among the several dynamic features in step A, there are trend change values (represented by ), periodic fluctuation values (represented by ), noise level magnitudes (represented by ), and real-time data change rates (represented by ). The relevant descriptions of the trend change values, periodic fluctuation values, noise level magnitudes, and real-time data change rates are as follows:

[0086] Trend change values: By analyzing long-term time series data, determine whether the nitrogen oxide emissions show an upward, downward, or stable trend, providing a basis for predicting the future trend of nitrogen oxide emissions in the waste incineration industry.

[0087] Periodic fluctuation values: Determine whether there are regular periodic changes in the data, such as daily, weekly, or monthly cycles, which helps to discover the periodic activity patterns of the emission sources.

[0088] Noise level magnitudes: Evaluate the content of noise in the data, remove or reduce the interference of noise on data analysis, and improve the reliability and accuracy of the data.

[0089] Real-time data change rates: Calculate the change rate of the data at each time point, promptly discover the rapid change situations of the emission data, and take corresponding measures.

[0090] The calculation formula for normalizing the several obtained dynamic features in step B is:

[0091]

[0092] where is the several normalized dynamic features at time t , is the several dynamic features before normalization at time t , is the mean value of each dynamic feature, and is the standard deviation.

[0093] Step B can normalize a number of obtained dynamic features (dynamic features are also called factors), so that the values of each dynamic feature fall within a similar scale range, thereby eliminating the differences in dimension and numerical range between different factors.

[0094] The formula for calculating the overall fluctuation magnitude of all the data obtained in Step C is:

[0095]

[0096] where, is the overall fluctuation magnitude of all the data obtained is the weight factor corresponding to each dynamic feature, is t the trend change value after normalization at time is t the periodic fluctuation value after normalization at time is t the noise level magnitude after normalization at time is t the real-time data change rate after normalization at time, OF is the operating condition. The operating condition comprehensively considers the influence of various factors such as the equipment operation state, production process, and environmental conditions on the nitrogen oxide emissions in the waste incineration industry, and analyzes the operating condition as an important dynamic feature. 、 、 、 and are t the dynamically adjusted parameters at time.

[0097] In this embodiment has a value range of [0.10, 0.30], has a value range of [0.10, 0.25], has a value range of [0.10, 0.25], has a value range of [0.10, 0.30], has a value range of [0.15, 0.40]. Select 1 value from the value ranges of 、 、 、 and as the weight factors corresponding to each dynamic feature respectively, 、 、 、 and The sum of the five weight factors is equal to 1.

[0098] In this embodiment The preferred values are as follows:

[0099] This embodiment The value is 0.25 The value of is 0.20 The value of is 0.20 The value of is 0.20, The value of is 0.15, It can be adjusted appropriately according to actual needs.

[0100] This embodiment The value is 0 or 1. Under normal operating conditions, The value of is 1. Under other operating conditions (such as when the incinerator fails or stops feeding garbage into the incinerator until the garbage in the furnace is completely burned out, etc.), The value of is 0.

[0101] This step uses the volatility index to t The dynamic adjustment coefficient at the moment For adjustment, the volatility index ( , VI) is calculated as follows:

[0102]

[0103] Among them, Is the standard deviation of all the data obtained, Is the absolute value of the change rate of all the data obtained, Is the volatility index.

[0104] This step is for t The dynamic adjustment coefficient at the moment The adjustment is specifically described as follows:

[0105] According to The value of adjusts t The dynamic adjustment coefficient at the moment , , , And ;

[0106] When Increases (indicating an increase in volatility), increase , decrease . Through this adjustment method, when the volatility is large, the influence of the trend change value, periodic fluctuation value, noise level, and real-time data change rate on the overall fluctuation can be more prominent, and the influence weight of the operating conditions can be relatively reduced, so that the overall fluctuation size of all the data obtained can more reflect the actual fluctuation of the data.

[0107] When Decreases (indicating a decrease in volatility), decrease , , and , increase . When the data fluctuation is small, appropriately reduce the influence of the trend change value, the periodic fluctuation value, the noise level, and the real-time data change rate, and relatively increase the weight of the working condition, so as to ensure the rationality of calculating the overall fluctuation size of all the obtained data.

[0108] Specifically, when is greater than 1.5 (also called relatively large fluctuation), values are all 1.2, value is 0.7.

[0109] When is greater than 0.8 and less than or equal to 1.5 (also called moderate fluctuation), and values are all 1.

[0110] When is less than or equal to 0.8 (also called relatively small fluctuation), values are all 0.7, value is 1.2.

[0111] The calculation formula for data stationarity processing of the overall fluctuation size of all the obtained data in step D is:

[0112]

[0113] where, is the data after data stationarity processing at t moment, i is a natural number, is the size of the smoothing window, used to control the number of data points participating in the smoothing calculation, is the smoothing coefficient, used to adjust the influence degree of the overall fluctuation on the smoothing process, is t the data before data stationarity processing at j moment, j is an index, indicating the forward shift from the current time by is t the overall fluctuation size of all the obtained data at

[0114] The larger the value of

[0115] is, the greater the influence of the overall fluctuation of all the obtained data on the smoothing effect.

[0116] During the real-time monitoring and dynamic adjustment of the original stationary data, the size of the smoothing window is continuously adjusted. and the smoothing coefficient is adjusted , and it is dynamically adjusted according to the changes in real-time data , until the overall fluctuation size of all the obtained data reaches the expected fluctuation range, and the final stationary data is obtained.

[0117] S3. Perform feature importance analysis on all the preprocessed data to obtain the key features closely related to nitrogen oxide emissions in the waste incineration industry.

[0118] Specifically, in this step, the emission pattern mining algorithm is used to perform feature importance analysis on all the preprocessed data to obtain the key features closely related to nitrogen oxide emissions in the waste incineration industry (in this embodiment, 20 key features closely related to nitrogen oxide emissions in the waste incineration industry are obtained). The specific steps of the emission pattern mining algorithm include:

[0119] a. Construct a tree model for emission pattern mining;

[0120] b. Use all the preprocessed data to train the tree model for emission pattern mining to obtain a trained tree model for emission pattern mining;

[0121] c. Calculate the SHAP (SHapley Additive exPlanations) value according to the trained tree model for emission pattern mining;

[0122] d. According to the calculated SHAP value, determine the contribution size of the preprocessed data to the prediction result of nitrogen oxide emissions in the waste incineration industry, sort the contribution sizes, and obtain the key features closely related to nitrogen oxide emissions in the waste incineration industry.

[0123] In step a, the tree model for emission pattern mining is based on the dynamic modeling technology of time series data, which can capture the pattern association between each feature and nitrogen oxide emissions in the waste incineration industry in real time, and then learn the complex non-linear relationship between each feature and nitrogen oxide emissions in the waste incineration industry.

[0124] The calculation formula for the SHAP value in step c is:

[0125]

[0126] where is the feature subset, is the set of all features, and are the predicted values when including and not including the feature respectively, i is a natural number. Indicates which feature Is the tree model function for emission pattern mining.

[0127] The specific process of step d is described as follows:

[0128] According to the calculated SHAP values, determine the contribution of the preprocessed data to the prediction result of nitrogen oxide emissions in the waste incineration industry, and sort the contributions (specifically, in this embodiment, the contributions are sorted from large to small). In this embodiment, the top 20 data in terms of contribution are selected as the key features closely related to nitrogen oxide emissions in the waste incineration industry.

[0129] S4. Perform empirical mode decomposition on the key features closely related to nitrogen oxide emissions in the waste incineration industry and the continuous nitrogen oxide emission data in the waste industry to obtain the intrinsic mode function components and the residual components.

[0130] Perform empirical mode decomposition on the key features closely related to nitrogen oxide emissions in the waste incineration industry and the real nitrogen oxide emission data in the waste industry to obtain the intrinsic mode function components (Intrinsic Mode Function, IMF) and the residual components.

[0131] Among them, each IMF represents the signal characteristics at different time scales. The high-frequency IMF reflects the short-term fluctuations and noise components in the data, and the low-frequency IMF reflects the long-term trends and main change patterns in the data.

[0132] The specific steps of empirical mode decomposition include:

[0133] Preprocessing: Preprocess the key features closely related to nitrogen oxide emissions in the waste incineration industry and the continuous nitrogen oxide emission data to ensure the integrity and consistency of the data, providing a basis for subsequent empirical mode decomposition.

[0134] Extreme point detection: For each key feature closely related to nitrogen oxide emissions in the waste incineration industry, detect its local maximum points and local minimum points. A local maximum point refers to a point whose value is greater than the values of its adjacent points, and a local minimum point is the opposite.

[0135] Envelope construction: Use the cubic spline interpolation method to fit the local maximum points and local minimum points respectively to construct the upper envelope and the lower envelope.

[0136] Mean calculation and detail component extraction: Calculate the mean of the upper envelope and the lower envelope, and subtract this mean from the continuous nitrogen oxide emission data in the waste industry to obtain a detail component.

[0137] Sifting: Check whether the extracted detail component meets the conditions of the intrinsic mode function component, that is, the number of extreme points and zero-crossing points is equal or differs by one, and the mean of the upper envelope and the lower envelope is zero. If not, take this detail component as a new signal and repeat the above steps until the conditions of the intrinsic mode function component are met.

[0138] Iterative decomposition: Strip the extracted intrinsic mode function components from the continuous nitrogen oxide emission data in the waste industry, and repeat the above process for the residual components until the residual components become a monotonic function or a low-frequency signal.

[0139] S5. Construct a training set and a validation set using the obtained intrinsic mode function components and residual components.

[0140] Construct a training set and a validation set using the obtained intrinsic mode function components and residual components. Take 80% of the obtained intrinsic mode function components and residual components as the training set, and 20% as the validation set.

[0141] To improve the generalization ability and robustness of the LSTM (Long Short-Term Memory) model in the later stage, in this embodiment, the training set adopts the method of time series cross-validation. The training set is divided into multiple subsets, and one subset is used as the validation set in turn, and the remaining subsets are used as the training set for multiple LSTM model training and validation.

[0142] S6. Train the LSTM model according to the training set to obtain a trained LSTM model.

[0143] The LSTM model includes an input gate, a forget gate, an output gate, and a memory unit, and can capture long-term and short-term dependencies in time series data.

[0144] Specifically, after multiple iterations and experiments, the input dimension of the LSTM model in this step is 390, the number of neurons in the hidden layer is 128, the number of hidden layers is 2, and the output dimension is 1, corresponding to the predicted value of nitrogen oxide emissions at the 5th minute.

[0145] In this step, the Adam optimization algorithm is used to train the LSTM model. The Adam optimization algorithm is a variant of the Stochastic Gradient Descent (SGD). It combines the concepts of momentum and RMSprop, and the LSTM model can converge faster. In this step, the learning rate of the LSTM model is set to 0.001, and the L2 regularization method is adopted to prevent the LSTM model from overfitting.

[0146] S7. Use the trained LSTM model to predict the data in the validation set and compare it with the continuous nitrogen oxide emission data of the waste industry to obtain the nitrogen oxide emission prediction results of the waste incineration industry.

[0147] In this step, the evaluation index for the nitrogen oxide emission prediction results of the waste incineration industry is the mean absolute percentage error.

[0148] The formula for calculating the mean absolute percentage error is:

[0149]

[0150] where MAPE is the mean absolute percentage error, is the number of samples, is the true value of the i th sample, is the predicted value of the i th sample, is the percentage of the absolute prediction error of the i th sample.

[0151] To verify the effectiveness of the nitrogen oxide emission prediction method for the waste incineration industry proposed in the present invention, this embodiment selects 4 pollutants (NOx, CO, Dust, and HCl) in the data of Line 5 of Zengcheng of a certain South China Institute project for prediction. This embodiment uses the mean absolute percentage error to evaluate the nitrogen oxide emission prediction results of the waste incineration industry, and the evaluation results are shown in Table 1.

[0152] Table 1 Evaluation results of the mean absolute percentage error

[0153]

[0154] The data in Table 1 shows that the mean absolute percentage errors of the 4 pollutants (NOx, CO, Dust, and HCl) are relatively small, and the mean absolute percentage errors of the 4 pollutants (NOx, CO, Dust, and HCl) are all less than 100%, verifying the effectiveness of the prediction method of the present invention.

[0155] Example 4:

[0156] Please refer to Figure 3 As shown, the present invention also provides an electronic device 100 for the nitrogen oxide emission prediction method of the waste incineration industry; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and operable on the at least one processor 102, and at least one communication bus 104.

[0157] The memory 101 can be used to store the computer program 103. The processor 102 realizes the steps of the nitrogen oxide emission prediction method in the waste incineration industry described in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the electronic device 100 (such as audio data, etc.). In addition, the memory 101 can include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0158] The at least one processor 102 can be a Central Processing Unit (CPU), and can also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 can be a microprocessor or the processor 102 can also be any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and connects various parts of the entire electronic device 100 through various interfaces and lines.

[0159] The memory 101 in the electronic device 100 stores a plurality of instructions to implement the nitrogen oxide emission prediction method in the waste incineration industry. The processor 102 can execute the plurality of instructions to thereby implement:

[0160] Obtain continuous high-frequency nitrogen oxide emission data, related flue gas parameter data, and production monitoring point data;

[0161] Preprocess all the obtained data to clear abnormal data;

[0162] Conduct feature importance analysis on all the preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry;

[0163] Perform empirical mode decomposition on the key features closely related to nitrogen oxide emissions in the waste incineration industry and the continuous nitrogen oxide emission data of the waste industry to obtain the intrinsic mode function components and the residual components;

[0164] Construct a training set and a validation set by using the obtained intrinsic mode function components and the residual components;

[0165] Train the LSTM model according to the training set to obtain a trained LSTM model;

[0166] Use the trained LSTM model to predict the data in the validation set and compare it with the continuous nitrogen oxide emission data of the waste industry to obtain the nitrogen oxide emission prediction result of the waste incineration industry.

[0167] Embodiment 5:

[0168] If the modules / units integrated in the electronic device 100 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-mentioned embodiment methods of the present invention, it can also be completed by a computer program instructing the relevant hardware. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory and read-only memory (ROM, Read-Only Memory).

[0169] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0170] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce means for implementing the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0171] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0172] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in the flow Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0173] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: it is still possible to modify the specific implementation manners of the present invention or make equivalent replacements, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A method for predicting nitrogen oxide emissions in the waste incineration industry, characterized in that, It includes the following steps: Obtain continuous nitrogen oxide emission data, related flue gas parameter data, and production monitoring point data; Preprocess all the obtained data. First, clear the abnormal data to obtain the data after clearing the anomalies, and then use the emission dynamic balance algorithm to perform data stationarity processing on the data after clearing the anomalies to obtain stationary data; The specific steps of the emission dynamic balance algorithm include: Analyze the time series characteristics of all the obtained data in real time to obtain several dynamic features; Perform normalization processing on the obtained several dynamic features; Combine the normalized several dynamic features to obtain the overall fluctuation magnitude of all the obtained data; The calculation formula for the overall fluctuation magnitude of all the obtained data is: wherein, is the overall fluctuation magnitude of all the acquired data is the weight factor corresponding to each dynamic feature, the sum of the five weight factors is equal to 1, is t the trend change value after normalization processing at time is t the periodic fluctuation value after normalization processing at time is t the noise level magnitude after normalization processing at time is t the real-time data change rate after normalization processing at time is the working condition, the value of which is 0 or 1, , , , (t) and are t the dynamic adjustment parameters at time t the dynamic adjustment parameter at time , , , (t) and are set according to the volatility index; Perform data stationarity processing on the overall fluctuation magnitude of all the obtained data to obtain the original stationary data; the calculation formula for performing data stationarity processing on the overall fluctuation magnitude of all the obtained data is: Among them, is t the data after the processing of the moment data stationarity, i is a natural number, is the size of the smoothing window, which is used to control the number of data points participating in the smoothing calculation, is the smoothing coefficient, which is used to adjust t the overall fluctuation size of all the data obtained at the moment, is t the data before the processing of the moment data stationarity, j is an index, indicating the forward shift from the current time by j time periods; Perform real-time monitoring and dynamic adjustment on the original stationary data until the overall fluctuation magnitude of all the obtained data reaches the expected fluctuation range to obtain the final stationary data; Perform feature importance analysis on all the preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry; Perform empirical mode decomposition on the key features closely related to nitrogen oxide emissions in the waste incineration industry and the continuous nitrogen oxide emission data of the waste industry to obtain intrinsic mode function components and residual components; Construct a training set and a validation set using the obtained intrinsic mode function components and residual components; Train the LSTM model according to the training set to obtain a trained LSTM model; Use the trained LSTM model to predict the data in the validation set and compare it with the continuous nitrogen oxide emission data of the waste industry to obtain the nitrogen oxide emission prediction result of the waste incineration industry.

2. The method for predicting nitrogen oxide emissions in the waste incineration industry according to claim 1, wherein In the step of performing feature importance analysis on all the preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry, specifically use the emission pattern mining algorithm to perform feature importance analysis on all the preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry.

3. The method for predicting nitrogen oxide emissions in the waste incineration industry according to claim 2, wherein The specific steps of the emission pattern mining algorithm include: Construct a tree model for emission pattern mining; Use all the preprocessed data to train the tree model for emission pattern mining to obtain a trained tree model for emission pattern mining; Calculate the SHAP value according to the trained tree model for emission pattern mining; According to the calculated SHAP value, determine the contribution size of all the preprocessed data to the nitrogen oxide emission prediction result of the waste incineration industry, sort the contribution sizes, and obtain key features closely related to nitrogen oxide emissions in the waste incineration industry.

4. The method for predicting nitrogen oxide emissions in the waste incineration industry according to claim 1, wherein In the step of constructing a training set and a validation set using the obtained intrinsic mode function components and residual components, specifically use 80% of the obtained intrinsic mode function components and residual components as the training set and 20% as the validation set.

5. A nitrogen oxide emission prediction system for the waste incineration industry, characterized in that, It includes a data acquisition module, a data preprocessing module, a feature importance analysis module, an empirical mode decomposition module, a dataset construction module, a model training module, and a nitrogen oxide emission prediction module for the waste incineration industry; The data acquisition module is used to acquire continuous high-frequency nitrogen oxide emission data, relevant flue gas parameter data, and production monitoring point data; The data preprocessing module is used to preprocess all the acquired data. First, abnormal data is cleared to obtain the data after abnormal data clearing, and then the emission dynamic balance algorithm is used to perform data stationarity processing on the data after abnormal data clearing to obtain stationary data; The specific steps of the emission dynamic balance algorithm include: Real-time analyze the time series characteristics of all the acquired data to obtain several dynamic features; Perform normalization processing on the obtained several dynamic features; Combine the normalized several dynamic features to obtain the overall fluctuation magnitude of all the acquired data; The calculation formula for the overall fluctuation magnitude of all the acquired data is: Among them, is the overall fluctuation magnitude of all the acquired data are the weight factors corresponding to the respective dynamic features, and the sum of the five weight factors is equal to 1, is t the trend change value after normalization at time is t the periodic fluctuation value after normalization at time is t the noise level magnitude after normalization at time is t the real-time data change rate after normalization at time is the operating condition, and the value of 、 、 、 (t) and is t the dynamically adjusted parameter at time t the dynamically adjusted parameter at time 、 、 、 (t) and are set according to the volatility index; Perform data stationarity processing on the overall fluctuation magnitude of all the acquired data to obtain the original stationary data; the calculation formula for performing data stationarity processing on the overall fluctuation magnitude of all the acquired data is: Among them, is t the data after the processing of the stationarity of the moment data, i is a natural number, is the size of the smoothing window, which is used to control the number of data points participating in the smoothing calculation, is the smoothing coefficient, which is used to adjust t the overall fluctuation size of all the data obtained at the moment, is t the data before the processing of the stationarity of the moment data, j is an index, indicating the forward shift from the current time by j time periods; Perform real-time monitoring and dynamic adjustment on the original stationary data until the overall fluctuation magnitude of all the acquired data reaches the expected fluctuation range to obtain the final stationary data; The feature importance analysis module is used to perform feature importance analysis on all the preprocessed data to obtain key features closely related to nitrogen oxide emissions in the waste incineration industry; The empirical mode decomposition module is used to perform empirical mode decomposition on the key features closely related to nitrogen oxide emissions in the waste incineration industry and the continuous nitrogen oxide emission data in the waste industry to obtain intrinsic mode function components and residual components; The dataset construction module is used to construct a training set and a validation set by using the obtained intrinsic mode function components and residual components; The module training module is used to train the LSTM model according to the training set to obtain a trained LSTM model; The nitrogen oxide emission prediction module for the waste incineration industry is used to predict the data in the validation set by using the trained LSTM model and compare it with the continuous nitrogen oxide emission data in the waste industry to obtain the nitrogen oxide emission prediction result for the waste incineration industry.

6. An electronic device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the nitrogen oxide emission prediction method for the waste incineration industry according to any one of claims 1 to 4.

7. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the nitrogen oxide emission prediction method for the waste incineration industry according to any one of claims 1 to 4.

Images