Power plant coal-fired boiler-oriented NOx emission prediction method and system

Abstract

The invention relates to the technical field of emission prediction, and provides a power plant coal-fired boiler-oriented NOx emission prediction method and system, and the method comprises the steps: collecting the emission of nitrogen oxides in a chimney of a power plant coal-fired boiler, and building a nitrogen oxide emission matrix of a collection period; setting a mark value, obtaining a fluctuation characteristic value, and a characteristic distance and a characteristic similarity of adjacent fluctuation sliding windows, and calculating a first characteristic value; calculating a second characteristic value according to the modal decomposition result of the nitrogen oxide emission, and calculating a fusion weight in combination with the mark value and the first characteristic value; and according to all the nitrogen oxide emissions collected by the nitrogen-oxygen detection sensor in the collection period and the fusion weight, obtaining a prediction result of the nitrogen oxide emissions of the coal-fired boiler of the power plant. According to the method, the reliability of predicting the nitrogen oxide emission of the coal-fired boiler of the power plant can be improved.

Description

Technical Field

[0001] This invention relates to the field of emission prediction technology, specifically to a method and system for predicting NOx emissions from coal-fired boilers in power plants. Background Technology

[0002] The combustion of pulverized coal in coal-fired boilers in power plants produces a large amount of nitrogen oxides (NOx). Since the units in operation have preset emission limits, predicting the NOx emissions from coal-fired boilers in power plants can help achieve pollution control and environmental protection. The NOx emissions from coal-fired boilers in power plants are not stable; when power demand changes, the load on the coal-fired boilers changes accordingly, and the NOx emissions will also vary.

[0003] However, the flue gas from coal-fired boilers in power plants mixes and stagnates as it rises inside the chimney. When nitrogen oxide emissions fluctuate due to load changes, the actual monitored nitrogen oxide emissions are affected by the mixing and stagnation of the flue gas, causing the monitored actual values ​​to deviate from the true values ​​and affecting the reliability of nitrogen oxide emission prediction. Summary of the Invention

[0004] This invention provides a method and system for predicting NOx emissions from coal-fired power plant boilers. It addresses the problem that the mixing and retention of nitrogen oxides emitted from coal-fired power plant boilers in the chimney causes deviations in NOx emission values, leading to insufficient reliability in NOx emission prediction. The specific technical solution adopted is as follows: In a first aspect, one embodiment of the present invention provides a method for predicting NOx emissions from coal-fired boilers in power plants, the method comprising the following steps: The nitrogen oxide emissions of flue gas at different nitrogen oxide detection sensor locations inside the chimney of a coal-fired boiler in a power plant were collected during different collection periods and at different collection times, and a nitrogen oxide emission matrix for the collection period was established. The marker value of the nitrogen oxide detection sensor in the acquisition period is assigned according to the nitrogen oxide emission amount. According to all nitrogen oxide emission matrices, the fluctuation feature value of the nitrogen oxide emission matrix is ​​obtained and a fluctuation feature sequence is established. The fluctuation sliding window is identified according to the fluctuation feature sequence and the feature distance between adjacent fluctuation sliding windows is obtained. According to the nitrogen oxide emission matrix and the nitrogen oxide emission amount collected by the same nitrogen oxide detection sensor, the feature similarity of the nitrogen oxide detection sensor in the acquisition period is calculated. According to the marker value and feature similarity of the nitrogen oxide detection sensor in the acquisition period, and the feature distance of all adjacent fluctuation sliding windows where the fluctuation feature value of the nitrogen oxide emission matrix in the acquisition period is located, the first feature value of the nitrogen oxide detection sensor in the acquisition period is calculated. Based on the empirical mode decomposition results of all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period, the second characteristic value of the nitrogen oxide detection sensor during the acquisition period is calculated. Combining the label value and the first characteristic value, the fusion weight of the nitrogen oxide detection sensor during the acquisition period is calculated. Based on the total nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period and the fusion weight, the predicted nitrogen oxide emissions of coal-fired boilers in power plants are obtained.

[0005] Furthermore, the nitrogen oxide emission matrix is ​​specifically as follows: The nitrogen oxide emission matrix contains all nitrogen oxide emissions collected by all nitrogen oxide detection sensors within the same collection period, with nitrogen oxide emissions collected by the same nitrogen oxide detection sensor appearing in the same row of the nitrogen oxide emission matrix.

[0006] Furthermore, the specific method for assigning the marker value of the nitrogen and oxygen detection sensor during the acquisition cycle is as follows: The average nitrogen oxide emissions collected by the same nitrogen oxide detection sensor within the same collection period are recorded as the average nitrogen oxide emissions of the same nitrogen oxide detection sensor within the same collection period. Based on the relationship between the average nitrogen oxide emissions of the nitrogen oxide detection sensor during the acquisition period and the preset first judgment threshold, the marker value of the nitrogen oxide detection sensor during the acquisition period is assigned a value of 1, 2, or 0.

[0007] Furthermore, the specific method for determining the fluctuation characteristic values ​​and the fluctuation characteristic sequence of the nitrogen oxide emission matrix is ​​as follows: Singular value decomposition is performed on the nitrogen oxide emission matrix to obtain different singular values. The ratio of the largest singular value to the sum of all singular values ​​is denoted as the fluctuation characteristic value of the nitrogen oxide emission matrix. A fluctuation characteristic sequence is established based on the fluctuation characteristic values ​​of the nitrogen oxide emission matrix for all collection periods.

[0008] Furthermore, the method for determining the characteristic distance between adjacent fluctuating sliding windows is as follows: A sliding window of a first preset length is established and slides with a step size of 1. The normalized value of the Pearson correlation coefficient of nitrogen oxide emissions in adjacent sliding windows on the fluctuation feature sequence is recorded as the fluctuation similarity of adjacent sliding windows. Based on the relationship between the fluctuation similarity and the preset second judgment threshold, the starting sliding window and the ending sliding window are identified. All sliding windows passed from the starting sliding window to the ending sliding window are recorded as fluctuation sliding windows. When the adjacent number The and the first When the absolute value of the difference in the similarity of the fluctuations of the three fluctuating sliding windows is greater than the preset third judgment threshold, the judgment is based on the adjacent three-dimensional sliding windows. The and the first The difference in the fluctuation characteristic values ​​of the first fluctuation sliding window determines the adjacent first fluctuation characteristic value. The and the first The characteristic distance of each fluctuating sliding window, where... It represents odd numbers that are greater than or equal to 1.

[0009] Furthermore, the method for determining the feature similarity of the nitrogen and oxygen detection sensor during the acquisition cycle is as follows: The values ​​of all nitrogen oxide emissions collected by any nitrogen oxide detection sensor in the nitrogen oxide emission matrix of the acquisition period are set to 0, and the comparison concentration matrix of the nitrogen oxide detection sensor in the acquisition period is obtained. The cosine similarity between the comparison concentration matrix of the nitrogen oxide detection sensor in the acquisition period and the nitrogen oxide emission matrix of the acquisition period is denoted as the feature similarity of the nitrogen oxide detection sensor in the acquisition period.

[0010] Furthermore, the specific method for determining the second characteristic value of the nitrogen and oxygen detection sensor during the acquisition cycle is as follows: For all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period, the information entropy of the distribution probability of the energy value is calculated based on the mode decomposition results with the number of intrinsic mode fractions (IMFs) being 3, 5, and 7, respectively. The positive correlation result of the KL divergence between the energy distribution vectors of the IMF components corresponding to all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period (IMFs with 3, 5, and 7 respectively) is recorded as the second characteristic value of the nitrogen oxide detection sensor during the acquisition period.

[0011] Furthermore, the steps for obtaining the fusion weights of the nitrogen and oxygen detection sensor during the acquisition cycle are as follows: When the flag value of the nitrogen and oxygen detection sensor is 1 in the acquisition period, the product of the first feature value and the second feature value of the nitrogen and oxygen detection sensor in the acquisition period is recorded as the fusion weight of the nitrogen and oxygen detection sensor in the acquisition period. When the marker value of the nitrogen and oxygen detection sensor in the acquisition period is 2, the ratio of the first feature value to the second feature value of the nitrogen and oxygen detection sensor in the acquisition period is recorded as the fusion weight of the nitrogen and oxygen detection sensor in the acquisition period. When the flag value of the nitrogen and oxygen detection sensor is 0 during the acquisition period, the fusion weight of the nitrogen and oxygen detection sensor during the acquisition period is assigned to 1 / 1 of the number of nitrogen and oxygen detection sensors.

[0012] Furthermore, the specific method for obtaining the predicted nitrogen oxide emissions from coal-fired boilers in power plants based on all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period and their fusion weights includes: The mean, variance, first eigenvalue, KL divergence of information entropy of 3 and 5, KL divergence of information entropy of 7 and 5, and KL divergence of information entropy of 3 and 7 of all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period are arranged in sequence to obtain the feature vector of the nitrogen oxide detection sensor during the acquisition period. The sequence of weighted feature vectors from a first preset number of consecutive acquisition cycles, along with all nitrogen oxide emissions collected by all nitrogen oxide detection sensors during the first preset number of consecutive acquisition cycles, are input into the decision tree regression model to obtain the predicted values ​​of nitrogen oxide emissions at all acquisition times in the next acquisition cycle following the last acquisition cycle.

[0013] Secondly, embodiments of the present invention also provide a NOx emission prediction system for coal-fired boilers in power plants, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of any of the methods described above.

[0014] The beneficial effects of this invention are: This application collects nitrogen oxide emissions from flue gas at different locations within the chimney of a coal-fired power plant. Firstly, considering the flue gas transmission process during load adjustment, each nitrogen oxide detection sensor from the chimney inlet to the chimney outlet will sequentially monitor changes in nitrogen oxide emissions according to the order of flue gas flow. Based on the changing trends of all nitrogen oxide emissions collected by the nitrogen oxide detection sensors within the collection period, a label value is assigned to the nitrogen oxide detection sensors for that period. Furthermore, based on the characteristics of the changing trends, more attention is paid to nitrogen oxide emissions showing attenuation, obtaining the first characteristic value of the nitrogen oxide detection sensors for that period. This first characteristic value evaluates the degree of attention given to the nitrogen oxide emissions collected by the nitrogen oxide detection sensors during the collection period in the process of predicting nitrogen oxide emissions. When the nitrogen oxide emissions collected by the nitrogen oxide detection sensors within the same collection period simultaneously include emissions that have not changed... When nitrogen oxide emissions change or decrease, the changing nitrogen oxide emissions increase the low-frequency energy of the collected nitrogen oxide emissions in the frequency domain. To more significantly identify the low-frequency energy, it is extracted at different scales to obtain the second feature value of the nitrogen oxide detection sensor during the acquisition period. Combined with the label value and the first feature value, the fusion weight of the nitrogen oxide detection sensor during the acquisition period is calculated. Finally, based on all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period and the fusion weight, the prediction process is more accurate based on the nitrogen oxide emissions showing a changing trend. This provides the prediction results for nitrogen oxide emissions from coal-fired power plant boilers, addressing the problem that the mixing and retention of nitrogen oxides emitted from coal-fired power plant boilers in the chimney causes the value of nitrogen oxide emissions to deviate, resulting in insufficient reliability of nitrogen oxide emission prediction. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a method for predicting NOx emissions from coal-fired boilers in power plants, provided in one embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Please see Figure 1 The diagram illustrates a flowchart of a NOx emission prediction method for coal-fired boilers in power plants, provided by an embodiment of the present invention. The method includes the following steps: Step S001: Collect the nitrogen oxide emissions of flue gas at different nitrogen oxide detection sensor locations inside the chimney of the coal-fired boiler in the power plant during different collection periods and at different collection times, and establish a nitrogen oxide emission matrix for the collection period.

[0019] Inside the chimney of a coal-fired boiler in a power plant, different nitrogen oxide detection sensors are installed at equal intervals from bottom to top. These sensors are used to collect the amount of nitrogen oxides emitted in the flue gas at different collection periods and at different collection times.

[0020] In this embodiment, 20 nitrogen oxide detection sensors are set. The number of nitrogen oxide detection sensors can be set by those skilled in the art based on the predicted nitrogen oxide emissions. In this embodiment, the duration of the collection period is set to 30 minutes, and the time interval between adjacent collection times is set to 10 seconds. The nitrogen oxide emissions are collected for one day. In actual application, as other implementation methods, implementers can decide the sampling frequency and sampling duration according to the actual situation. This application does not impose any special restrictions.

[0021] This embodiment uses mean filtering to denoise nitrogen oxide emissions, avoiding the influence of air interference and sensor vibration during the data collection process. Mean filtering is a well-known technique and will not be elaborated further. As other embodiments, while achieving the goal of data denoising, implementers may employ other methods in the prior art, such as median filtering, for data denoising; this application does not impose any special limitations.

[0022] Arrange all nitrogen oxide emissions collected by the same nitrogen oxide detection sensor within the same acquisition period in chronological order of acquisition time to obtain the nitrogen oxide emission sequence of the same nitrogen oxide detection sensor within the same acquisition period. Then, arrange the nitrogen oxide emission sequences of all nitrogen oxide detection sensors within the same acquisition period in order from bottom to top of the nitrogen oxide detection sensors to obtain the nitrogen oxide emission matrix of the same acquisition period.

[0023] Thus, a matrix of nitrogen oxide emissions for different collection periods was obtained.

[0024] Step S002: Assign a value to the marker value of the nitrogen oxide detection sensor during the acquisition period based on the nitrogen oxide emission amount; obtain the fluctuation feature value of the nitrogen oxide emission matrix and establish a fluctuation feature sequence based on the matrix of all nitrogen oxide emissions; identify the fluctuation sliding window based on the fluctuation feature sequence and obtain the feature distance between adjacent fluctuation sliding windows; calculate the feature similarity of the nitrogen oxide detection sensor during the acquisition period based on the nitrogen oxide emission matrix and the nitrogen oxide emissions collected by the same nitrogen oxide detection sensor; calculate the first feature value of the nitrogen oxide detection sensor during the acquisition period based on the marker value and feature similarity of the nitrogen oxide detection sensor during the acquisition period, and the feature distance between all adjacent fluctuation sliding windows where the fluctuation feature value of the nitrogen oxide emission matrix of the acquisition period is located.

[0025] When the power demand of a power plant changes, the boiler load will be adjusted accordingly. If the boiler load decreases, the amount of coal fed into the boiler will decrease accordingly, and the total amount of nitrogen oxides emitted by the boiler combustion will decrease. However, the change in the nitrogen oxide content per unit volume of flue gas needs to be judged in conjunction with the adjustment of combustion conditions under low load, and it is not necessarily a synchronous decrease.

[0026] During flue gas transfer under load adjustment, the nitrogen oxide (NOx) sensor located at the chimney inlet will be the first to detect changes in NOx emissions. Since the flue gas flows directionally upwards along the chimney, each NOx sensor from the chimney inlet to the chimney outlet will sequentially monitor NOx emissions according to the order of flue gas flow; that is, the NOx emissions from each sensor will decrease sequentially according to the order of flue gas flow. Therefore, during the dynamic process of load adjustment, the relative importance of NOx emissions from different sensors for NOx emission prediction will vary, and this difference will change dynamically. Therefore, the NOx emission data collected by different sensors should be considered to varying degrees when predicting NOx emissions.

[0027] The singular value decomposition algorithm is used to decompose the nitrogen oxide emission matrix into singular values ​​to obtain three singular values. The ratio of the largest singular value to the sum of all singular values ​​is recorded as the fluctuation characteristic value of the nitrogen oxide emission matrix. The fluctuation characteristic values ​​of the nitrogen oxide emission matrix of all collection periods are arranged in chronological order of collection periods to obtain the fluctuation characteristic sequence.

[0028] The larger the fluctuation characteristic value of the nitrogen oxide emission matrix, the more singular the principal component of the nitrogen oxide emission matrix, that is, the higher the consistency of the changing trend of the data collected by each sensor.

[0029] When nitrogen oxide emissions change, the fluctuation characteristic sequence can reflect the trend of nitrogen oxide emissions.

[0030] Nitrogen oxide (NOx) emissions are collected at all sampling times within a day at different sampling periods when the load of a coal-fired boiler in a power plant remains stable. The second quartile of the NOx emissions collected when the load remains stable is recorded as the first judgment threshold. A sliding window of a first preset length is established and slides with a step size of 1. The average normalized value of the Pearson correlation coefficient of NOx emissions in adjacent sliding windows collected when the load remains stable is recorded as the second judgment threshold. In this embodiment, the first preset length is set to 5. The normalized value is calculated using the maximum-minimum normalization method. The calculation of the Pearson correlation coefficient is a well-known technique and will not be described in detail here.

[0031] The average nitrogen oxide emissions collected by the same nitrogen oxide detection sensor within the same acquisition period are recorded as the average nitrogen oxide emissions of the same nitrogen oxide detection sensor within the same acquisition period. When the average nitrogen oxide emissions of the nitrogen oxide detection sensor within the acquisition period are less than the first judgment threshold, the marker value of the nitrogen oxide detection sensor in the acquisition period is assigned to 1. When the average nitrogen oxide emissions of the nitrogen oxide detection sensor within the acquisition period are greater than the first judgment threshold, the marker value of the nitrogen oxide detection sensor in the acquisition period is assigned to 2. The marker values ​​of the remaining unassigned nitrogen oxide detection sensors in the acquisition period are assigned to 0.

[0032] When the marker value of the nitrogen oxide detection sensor in the acquisition period is 1, and the load of the coal-fired boiler in the power plant remains stable, the value of all nitrogen oxide emissions collected by the nitrogen oxide detection sensor in the acquisition period indicates an upward trend in nitrogen oxide emissions; when the marker value of the nitrogen oxide detection sensor in the acquisition period is 2, and the load of the coal-fired boiler in the power plant remains stable, the value of all nitrogen oxide emissions collected by the nitrogen oxide detection sensor in the acquisition period indicates a downward trend in nitrogen oxide emissions.

[0033] A sliding window of a first preset length is established and slides with a step size of 1. The normalized Pearson correlation coefficient of nitrogen oxide emissions within adjacent sliding windows on the fluctuation feature sequence is denoted as the fluctuation similarity between adjacent sliding windows. Following the order of the sliding windows, the windows with fluctuation similarity less than a second judgment threshold are selected. The first sliding window among the adjacent sliding windows is denoted as the starting sliding window. The sliding window whose fluctuation similarity is less than the second judgment threshold is then selected. The last sliding window among the adjacent sliding windows is denoted as the ending sliding window. All sliding windows traversed from the starting sliding window to the ending sliding window are denoted as the undulating sliding windows. When the adjacent... The and the first When the absolute value of the difference in the similarity of the fluctuations of the first fluctuation sliding window is greater than the third judgment threshold, the adjacent first fluctuation sliding window will be... The and the first The Cook distance between the fluctuation eigenvalues ​​of the i-th fluctuation sliding window is denoted as the adjacent i-th... The and the first The characteristic distance of a fluctuating sliding window.

[0034] in, This represents an odd number greater than or equal to 1; in this embodiment, the third judgment threshold is set to arctan10; this embodiment uses the maximum and minimum value normalization method to calculate the normalized value. In actual application, implementers can use other methods of existing technology, such as the sigmoid function, to calculate the normalized value, which is not limited here; the calculation of the Cook distance is a well-known technology and will not be described in detail here.

[0035] Understandably, all adjacent fluctuating sliding windows constitute a complete change in nitrogen oxide emissions.

[0036] The values ​​of all nitrogen oxide emissions collected by any nitrogen oxide detection sensor in the nitrogen oxide emission matrix of the collection period are set to 0. The comparison concentration matrix of the nitrogen oxide detection sensor in the collection period is obtained. The cosine similarity between the comparison concentration matrix of the nitrogen oxide detection sensor in the collection period and the nitrogen oxide emission matrix of the collection period is denoted as the feature similarity of the nitrogen oxide detection sensor in the collection period.

[0037] Based on the marker value and feature similarity of the nitrogen oxide detection sensor in each acquisition cycle, and the feature distance of all adjacent fluctuation sliding windows of the fluctuation feature value of the nitrogen oxide emission matrix in the acquisition cycle, the first feature value of each nitrogen oxide detection sensor in each acquisition cycle is calculated. The formula for calculating the first feature value is: , in, Indicates the first The nitrogen oxide detection sensor in the first The first characteristic value of each acquisition cycle; This indicates the total number of nitrogen and oxygen detection sensors; Indicates the first The total number of nitrogen and oxygen detection sensors with acquisition cycle marker values ​​of 1 and 2; Indicates the first The mean of the characteristic distances of all adjacent fluctuation sliding windows of the fluctuation characteristic value of the nitrogen oxide emission matrix in each collection period; Indicates the first The nitrogen oxide detection sensor in the first Feature similarity across collection cycles; Indicates the first The nitrogen oxide detection sensor in the first The initial weight for each collection period is set as follows in this embodiment: ; This represents a preset adjustment coefficient. The purpose of the adjustment coefficient is to prevent the denominator from being 0. In this embodiment, the value of the adjustment coefficient is 0.01. Indicates the first The nitrogen oxide detection sensor in the first The marker value for each collection cycle; Indicates the first The nitrogen oxide detection sensor in the first The fluctuation characteristic values ​​of the nitrogen oxide emission matrix in each collection cycle have corresponding adjacent fluctuation sliding windows. Indicates the first The nitrogen oxide detection sensor in the first The fluctuation characteristic values ​​of the nitrogen oxide emission matrix for each collection cycle do not have corresponding adjacent fluctuation sliding windows.

[0038] When calculating the first characteristic value, according to the conditions in the first row, the nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period are in a decaying phase. The nitrogen oxide emissions showing decaying fluctuations are located in the lower and middle parts of the chimney. More than half of the nitrogen oxide emissions collected by the nitrogen oxide detection sensors do not show decay. These nitrogen oxide emissions showing decaying fluctuations should be considered more in the process of predicting nitrogen oxide emissions. When calculating according to the conditions in the second row, the nitrogen oxide emissions showing decaying fluctuations are located in the upper and middle parts of the chimney. More than half of the nitrogen oxide emissions collected by the nitrogen oxide detection sensors have decayed. These nitrogen oxide emissions should be considered more in the process of predicting nitrogen oxide emissions. When calculating according to the conditions in the third row, the fluctuations in nitrogen oxide emissions have returned to normal.

[0039] At this point, the first characteristic value of each nitrogen and oxygen detection sensor in each acquisition cycle is obtained.

[0040] Step S003: Based on the empirical mode decomposition results of all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period, calculate the second characteristic value of the nitrogen oxide detection sensor during the acquisition period. Combine the label value and the first characteristic value to calculate the fusion weight of the nitrogen oxide detection sensor during the acquisition period.

[0041] When the nitrogen oxide emission data collected by the nitrogen oxide detection sensor in the same acquisition period includes both unchanged and increasing or decreasing nitrogen oxide emissions, the changing nitrogen oxide emissions will increase the low-frequency energy of the collected nitrogen oxide emissions in the frequency domain. In order to more significantly identify the low-frequency energy, the low-frequency energy is extracted at different scales.

[0042] For all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period, the empirical mode decomposition algorithm is used to set the number of intrinsic mode fractions (IMFs) to 3, 5, and 7 respectively. The energy values ​​are obtained when the number of IMFs is 3, 5, and 7 respectively. The information entropy of the distribution probability of the energy values ​​when the number of IMFs is 3, 5, and 7 respectively is calculated. The reference frequency band number parameter is set to the maximum number of IMFs. The distribution probability of the energy values ​​when the number of IMFs is 3 and 5 respectively is mapped to the space of the distribution probability when the number of IMFs is 7, so as to achieve frequency alignment.

[0043] The positive correlation result of the KL divergence between the energy distribution vectors of the IMF components corresponding to all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period (IMFs with 3, 5, and 7 respectively) is recorded as the second characteristic value of the nitrogen oxide detection sensor during the acquisition period.

[0044] It is understood that a positive correlation is applied to the KL divergence between the energy distribution vectors of IMF components with IMF numbers of 3, 5, and 7, respectively, ensuring a positive correlation between the KL divergence and the second characteristic value of the nitrogen and oxygen detection sensor during the acquisition period. It is understood that the positive correlation in this application refers to the relationship between the independent and dependent variables. The independent variable is the KL divergence of the information entropy with IMF numbers of 3, 5, and 7, and the dependent variable is the second characteristic value of the nitrogen and oxygen detection sensor during the acquisition period. The positive correlation means that the dependent variable increases (decreases) as the independent variable increases (decreases), and can be an additive or multiplicative relationship.

[0045] Preferably, as an embodiment of this application, the product of the KL divergence of information entropy with IMFs of 3 and 5 and the first weighting coefficient is denoted as the first product; the product of the KL divergence of information entropy with IMFs of 7 and 5 and the second weighting coefficient is denoted as the second product; the product of the KL divergence of information entropy with IMFs of 3 and 7 and the third weighting coefficient is denoted as the third product; and the normalized value of the sum of the first product, the second product, and the third product is denoted as the second characteristic value of the nitrogen and oxygen detection sensor during the acquisition period.

[0046] In this embodiment, the sigmoid function is used to calculate the normalized value. The sigmoid function is a well-known technique and will not be described in detail here. As other implementation methods, implementers can use other methods of the prior art, such as the tanh function.

[0047] The sum of the first weight coefficient, the second weight coefficient, and the third weight coefficient is 1. In this embodiment, the values ​​of the first weight coefficient, the second weight coefficient, and the third weight coefficient are 0.25, 0.25, and 0.5, respectively.

[0048] When the flag value of the nitrogen and oxygen detection sensor is 1 during the acquisition period, the product of the first feature value and the second feature value of the nitrogen and oxygen detection sensor during the acquisition period is recorded as the fusion weight of the nitrogen and oxygen detection sensor during the acquisition period.

[0049] When the marker value of the nitrogen and oxygen detection sensor is 2 in the acquisition period, the ratio of the first feature value to the second feature value of the nitrogen and oxygen detection sensor in the acquisition period is recorded as the fusion weight of the nitrogen and oxygen detection sensor in the acquisition period. In the process of calculating the ratio, in order to avoid the denominator being zero, a preset value needs to be added to the denominator. In this example, the preset value is 0.01.

[0050] When the flag value of the nitrogen and oxygen detection sensor is 0 during the acquisition period, the fusion weight of the nitrogen and oxygen detection sensor during the acquisition period is assigned to 1 / 1 of the number of nitrogen and oxygen detection sensors.

[0051] At this point, the fusion weights of the nitrogen and oxygen detection sensors during the acquisition cycle are obtained.

[0052] Step S004: Based on the total nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period and the fusion weight, obtain the predicted nitrogen oxide emissions from the coal-fired boiler of the power plant.

[0053] The mean, variance, first eigenvalue, KL divergence of information entropy 3 and 5, KL divergence of information entropy 7 and 5, and KL divergence of information entropy 3 and 7 of all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period are arranged in sequence to obtain the feature vector of the nitrogen oxide detection sensor during the acquisition period. The fusion weight of the nitrogen oxide detection sensor during the acquisition period is used as the weight, and the feature vectors of all nitrogen oxide detection sensors in the same acquisition period are weighted and summed to obtain the weighted feature vector of the same acquisition period.

[0054] The sequence of weighted feature vectors from a first preset number of consecutive acquisition cycles, along with all nitrogen oxide emissions collected by all nitrogen oxide detection sensors during the first preset number of consecutive acquisition cycles, are input into the decision tree regression model to obtain the predicted values ​​of nitrogen oxide emissions at all acquisition times in the next acquisition cycle following the last acquisition cycle.

[0055] Wherein, the first preset number is greater than or equal to 3 and less than or equal to 10, and in this embodiment, the value of the first preset number is 5; in this embodiment, the maximum depth of the decision tree regression model is set to 8, the minimum number of samples for leaf nodes is set to 3, and the minimum number of samples for node splitting is set to 5; the training process of the decision tree regression model is a well-known technique and will not be described in detail here.

[0056] This enables the prediction of nitrogen oxide emissions from coal-fired boilers in power plants.

[0057] Based on the same inventive concept as the above method, this embodiment of the invention also provides a NOx emission prediction system for coal-fired boilers in power plants, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described methods for predicting NOx emissions from coal-fired boilers in power plants.

[0058] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting NOx emissions from coal-fired boilers in power plants, characterized in that, The method includes the following steps: The nitrogen oxide emissions of flue gas at different nitrogen oxide detection sensor locations inside the chimney of a coal-fired boiler in a power plant were collected during different collection periods and at different collection times, and a nitrogen oxide emission matrix for the collection period was established. The marker value of the nitrogen oxide detection sensor in the acquisition period is assigned according to the nitrogen oxide emission amount. According to all nitrogen oxide emission matrices, the fluctuation feature value of the nitrogen oxide emission matrix is ​​obtained and a fluctuation feature sequence is established. The fluctuation sliding window is identified according to the fluctuation feature sequence and the feature distance between adjacent fluctuation sliding windows is obtained. According to the nitrogen oxide emission matrix and the nitrogen oxide emission amount collected by the same nitrogen oxide detection sensor, the feature similarity of the nitrogen oxide detection sensor in the acquisition period is calculated. According to the marker value and feature similarity of the nitrogen oxide detection sensor in the acquisition period, and the feature distance of all adjacent fluctuation sliding windows where the fluctuation feature value of the nitrogen oxide emission matrix in the acquisition period is located, the first feature value of the nitrogen oxide detection sensor in the acquisition period is calculated. Based on the empirical mode decomposition results of all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period, the second characteristic value of the nitrogen oxide detection sensor during the acquisition period is calculated. Combining the label value and the first characteristic value, the fusion weight of the nitrogen oxide detection sensor during the acquisition period is calculated. Based on the total nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period and the fusion weight, the predicted nitrogen oxide emissions of coal-fired boilers in power plants are obtained.

2. The method for predicting NOx emissions from coal-fired boilers in power plants according to claim 1, characterized in that, The nitrogen oxide emission matrix is ​​as follows: The nitrogen oxide emission matrix contains all nitrogen oxide emissions collected by all nitrogen oxide detection sensors within the same collection period, with nitrogen oxide emissions collected by the same nitrogen oxide detection sensor appearing in the same row of the nitrogen oxide emission matrix.

3. The method for predicting NOx emissions from coal-fired boilers in power plants according to claim 1, characterized in that, The specific method for assigning the marker value of the nitrogen and oxygen detection sensor during the acquisition cycle is as follows: The average nitrogen oxide emissions collected by the same nitrogen oxide detection sensor within the same collection period are recorded as the average nitrogen oxide emissions of the same nitrogen oxide detection sensor within the same collection period. Based on the relationship between the average nitrogen oxide emissions of the nitrogen oxide detection sensor during the acquisition period and the preset first judgment threshold, the marker value of the nitrogen oxide detection sensor during the acquisition period is assigned a value of 1, 2, or 0.

4. The method for predicting NOx emissions from coal-fired boilers in power plants according to claim 1, characterized in that, The specific method for determining the fluctuation characteristic value and the fluctuation characteristic sequence of the nitrogen oxide emission matrix is ​​as follows: Singular value decomposition is performed on the nitrogen oxide emission matrix to obtain different singular values. The ratio of the largest singular value to the sum of all singular values ​​is denoted as the fluctuation characteristic value of the nitrogen oxide emission matrix. A fluctuation characteristic sequence is established based on the fluctuation characteristic values ​​of the nitrogen oxide emission matrix for all collection periods.

5. The method for predicting NOx emissions from coal-fired boilers in power plants according to claim 1, characterized in that, The method for determining the characteristic distance between adjacent fluctuating sliding windows is as follows: A sliding window of a first preset length is established and slides with a step size of 1. The normalized value of the Pearson correlation coefficient of nitrogen oxide emissions in adjacent sliding windows on the fluctuation feature sequence is recorded as the fluctuation similarity of adjacent sliding windows. Based on the relationship between the fluctuation similarity and the preset second judgment threshold, the starting sliding window and the ending sliding window are identified. All sliding windows passed from the starting sliding window to the ending sliding window are recorded as fluctuation sliding windows. When the adjacent number The and the first When the absolute value of the difference in the similarity of the fluctuations of the three fluctuating sliding windows is greater than the preset third judgment threshold, the judgment is based on the adjacent three-dimensional sliding windows. The and the first The difference in the fluctuation characteristic values ​​of the first fluctuation sliding window determines the adjacent first fluctuation characteristic value. The and the first The characteristic distance of each fluctuating sliding window, where... It represents odd numbers that are greater than or equal to 1.

6. The method for predicting NOx emissions from coal-fired boilers in power plants according to claim 1, characterized in that, The method for determining the feature similarity of the nitrogen and oxygen detection sensor during the acquisition cycle is as follows: The values ​​of all nitrogen oxide emissions collected by any nitrogen oxide detection sensor in the nitrogen oxide emission matrix of the acquisition period are set to 0, and the comparison concentration matrix of the nitrogen oxide detection sensor in the acquisition period is obtained. The cosine similarity between the comparison concentration matrix of the nitrogen oxide detection sensor in the acquisition period and the nitrogen oxide emission matrix of the acquisition period is denoted as the feature similarity of the nitrogen oxide detection sensor in the acquisition period.

7. The method for predicting NOx emissions from coal-fired boilers in power plants according to claim 1, characterized in that, The specific method for determining the second characteristic value of the nitrogen and oxygen detection sensor during the acquisition period is as follows: For all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period, the information entropy of the distribution probability of the energy value is calculated based on the mode decomposition results with the number of intrinsic mode fractions (IMFs) being 3, 5, and 7, respectively. The positive correlation result of the KL divergence between the energy distribution vectors of the IMF components corresponding to all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period (IMFs with 3, 5, and 7 respectively) is recorded as the second characteristic value of the nitrogen oxide detection sensor during the acquisition period.

8. The method for predicting NOx emissions from coal-fired boilers in power plants according to claim 3, characterized in that, The steps for obtaining the fusion weight of the nitrogen and oxygen detection sensor during the acquisition period are as follows: When the flag value of the nitrogen and oxygen detection sensor is 1 in the acquisition period, the product of the first feature value and the second feature value of the nitrogen and oxygen detection sensor in the acquisition period is recorded as the fusion weight of the nitrogen and oxygen detection sensor in the acquisition period. When the marker value of the nitrogen and oxygen detection sensor in the acquisition period is 2, the ratio of the first feature value to the second feature value of the nitrogen and oxygen detection sensor in the acquisition period is recorded as the fusion weight of the nitrogen and oxygen detection sensor in the acquisition period. When the flag value of the nitrogen and oxygen detection sensor is 0 during the acquisition period, the fusion weight of the nitrogen and oxygen detection sensor during the acquisition period is assigned to 1 / 1 of the number of nitrogen and oxygen detection sensors.

9. The method for predicting NOx emissions from coal-fired boilers in power plants according to claim 7, characterized in that, The method for obtaining the predicted nitrogen oxide emissions from coal-fired boilers in power plants based on all nitrogen oxide emissions collected by nitrogen oxide detection sensors during the acquisition period and their fusion weights includes the following specific methods: The mean, variance, first eigenvalue, KL divergence of information entropy of 3 and 5, KL divergence of information entropy of 7 and 5, and KL divergence of information entropy of 3 and 7 of all nitrogen oxide emissions collected by the nitrogen oxide detection sensor during the acquisition period are arranged in sequence to obtain the feature vector of the nitrogen oxide detection sensor during the acquisition period. The sequence of weighted feature vectors from a first preset number of consecutive acquisition cycles, along with all nitrogen oxide emissions collected by all nitrogen oxide detection sensors during the first preset number of consecutive acquisition cycles, are input into the decision tree regression model to obtain the predicted values ​​of nitrogen oxide emissions at all acquisition times in the next acquisition cycle following the last acquisition cycle.

10. A NOx emission prediction system for coal-fired boilers in power plants, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as claimed in any one of claims 1-9.

Images