Heating furnace efficiency calculation method

Abstract

The invention relates to a heating furnace efficiency calculation method. The method comprises the following steps: preprocessing heating furnace sample data; preliminarily screening heating furnace efficiency influence factors, establishing a factor index data matrix, and preliminarily screening influence factor indexes; dimensionality reduction processing is carried out on the correlation coefficient matrix, standardization processing and dimensionality reduction processing are carried out on the heating furnace efficiency influence factors after preliminary screening, and feature vectors with the cumulative variance contribution rate exceeding 80% serve as input quantities of support vector machine regression; determining a support vector machine training set and a test set; selecting a support vector machine kernel function, determining to-be-optimized parameters of a support vector machine model, and training a support vector machine regression model by using the optimal parameters; testing the trained support vector machine regression model by using a test set, and determining an optimal support vector machine regression model as a heating furnace efficiency calculation model; and calculating the heating furnace efficiency by using the heating furnace efficiency calculation model. The efficiency of the heating furnace can be accurately and quickly calculated.

Description

Technical field: The invention relates to a heating furnace energy evaluation and prediction method in the technical field of oil and gas storage and transportation, and in particular to a heating furnace exergy efficiency calculation method. Background technology: As the main energy-consuming equipment in the surface engineering of oil and gas fields, heating furnaces are widely used in various production links such as oil and gas gathering, transportation and processing. As the development of oil and gas fields gradually enters the middle and late stages, the water content gradually increases and its operating efficiency gradually decreases. In order to reasonably observe the exergy efficiency of heating furnaces with greater energy-saving potential, a scientific and reasonable exergy efficiency calculation method needs to be proposed. The combustion principle of heating furnaces is complex, and exergy efficiency is difficult to obtain through actual measurement. At present, artificial intelligence algorithms and numerical simulation methods are mostly used for measurement. Compared with numerical simulation algorithms, artificial intelligence algorithms have the advantages of short time consumption, small sample requirements, and high calculation accuracy. The most widely used artificial intelligence algorithm is the neural network algorithm. However, the neural network algorithm is sensitive to sample data volume, sample integrity and error. The algorithm training time is long and there is an overfitting problem, and the generalization ability is poor. In summary, the current calculation method for heating furnace exergy efficiency still has certain limitations, and it is impossible to accurately and quickly calculate the heating furnace exergy efficiency. Summary of the invention: The object of the present invention is to provide a method for calculating the exergy efficiency of a heating furnace, which is used to solve the problem that the method for calculating the exergy efficiency of a heating furnace in the prior art cannot calculate accurately and quickly. The technical solution adopted by the present invention to solve the technical problem is: the method for calculating the exergy efficiency of the heating furnace comprises the following steps: Step 1: Preprocessing of heating furnace sample data: reviewing the heating furnace sample data obtained on site, removing abnormal values ​​of the sample data, and filling in the missing data of the sample; Step 2: Preliminary screening of factors affecting heating furnace exergy efficiency. Taking heating furnace exergy efficiency as the object and combining it with the actual production and operation of heating furnace, factors affecting energy efficiency are considered, including heating furnace inlet and outlet temperature, moisture content, and excess air coefficient. A factor index data matrix is ​​established to preliminarily screen influencing factor indexes. Step 3: Reduce the dimension of the correlation coefficient matrix, standardize the factors affecting the heating furnace exergy efficiency after preliminary screening, reduce the dimension of the calculated standardized matrix, and use the eigenvectors with cumulative variance contribution rate exceeding 80% as the input of support vector machine regression; Step 4: Determine the support vector machine training set and test set, and determine the training samples and test samples according to the 7:3 division; Step 5: Select the support vector machine kernel function, determine the parameters to be optimized for the support vector machine model, optimize the penalty factor c and the Gaussian kernel function parameter g, and finally select the parameter combination that minimizes the training model error, and use the optimal parameters to train the support vector machine regression model; Step 6: Use the test set to test the trained support vector machine regression model, using the determination coefficient R 2 The optimal support vector machine regression model is determined as the heating furnace exergy efficiency calculation model, with the root mean square error RMSE and mean absolute error MSE as the auxiliary model accuracy measurement standard. In the formula is the calculated value, is the mean Step 7: Calculate the heating furnace exergy efficiency using the heating furnace exergy efficiency calculation model. Step 1 in the above scheme is specifically as follows: (1) The Raida criterion is used to eliminate outliers in sample data; Assume that a set of test data is x 1 , x 2 , x 3 , ..., x n , where n is the number of data, the arithmetic mean t and standard deviation δ of the test data are calculated as follows: If |x i -n|≥3δ, then x i Remove abnormal data; (2) Use Lagrange interpolation method to fill in the missing data. The specific steps of step 2 in the above scheme are: (1) Establish factor indicator data matrix M: In the formula, m 11 is the jth influencing factor of the heating furnace at the i-th moment (i=1,2,…,n; j=1,2,…,m); (2) Calculate the Pearson correlation coefficient between each influencing factor: Calculate the Pearson correlation coefficient between each influencing factor index. First, calculate the mean of each factor index. Variance X j and the covariance X j1j2 , the calculation formula is as follows: Finally, according to the calculated covariance, the correlation coefficient R between the influencing factor indicators is calculated. j1j2 , the calculation formula is as follows: In the formula, is the average value of the j-th factor index data; X j is the variance of the j-th factor indicator data; The jth 1 、j 2 The mean of the factor indicators of the columns; The jth row in the i-th row 1 、j 2 Column factor indicator data; For the jth 1 、j 2 covariance of two columns; is the correlation coefficient between the two factor indicators; The value range of the Pearson correlation coefficient is [-1,1]. If the absolute value of the Pearson correlation coefficient is larger, it means that the factor indicator is more closely related to the typical reference indicator, and the data should be retained, otherwise it should be eliminated. Step 3 in the above scheme is specifically as follows: (1) The MIN-MAX method was used to standardize the factors affecting the heating furnace exergy efficiency after preliminary screening to obtain a normalized matrix; (2) The principal component analysis method is used to standardize the matrix for dimensionality reduction, and the eigenvectors with cumulative variance contribution exceeding 80% are used as the input of the support vector machine regression: First, solve the eigenvalues, variance contribution rates, and cumulative contribution rates of the principal components of the factors: According to the equation |λE-N|=0, calculate the normalization matrix N ij The m eigenvalues ​​of λ are sorted in ascending order 1 >λ 2 >…λ m ≥0, the eigenvector corresponding to each eigenvalue is η 1 ,η 2 ,…,η m , and get m principal components: Where P i is the i-th principal component, η ij is the coefficient in the linear combination of each principal component; Extract t principal components from it, then P i Variance contribution rate ρ of (i=1,2,L,m) i for: Cumulative contribution rate δ hl for: Where ρi P i The variance contribution rate of (i=1,2,…,m), λ i is the i-th eigenvalue; δ hl is 1 ,ρ 2 ,L,ρ q The cumulative contribution rate of hl When ≥80%, the first t principal components are extracted, which is the comprehensive index of t analysis items. Step 5 in the above scheme is specifically: (1) Establish support vector machine function: f(x)=ω·x+b Where ω is the weight variable, b is the linear function coefficient; (2) Introducing the relaxation factor ξ i , And the penalty factor c, the calculation model is: The constraints are: Where ε is the insensitive loss function, and p is the loss function; Will The partial derivative of is set to 0, as follows: Convert the quadratic programming problem into a dual problem: The constraints are: Where α i , α j , is the Lagrange multiplier, Q is the loss function of the Lagrange coefficient; (3) Solve the Lagrangian objective function and use the Gaussian kernel function to obtain the support vector machine regression model: Where g is the parameter of the Gaussian kernel function; c is the penalty factor, which represents the tolerance of the regression model to errors, and its size affects the generalization ability of the model; g is the parameter of the Gaussian kernel function, and its size determines the number of support vectors; (4) Use the cross-validation method to optimize the parameters c and g. Beneficial effects: 1. The present invention uses a method combining principal component analysis and support vector machine regression to calculate the heating furnace exergy efficiency. After the principal component analysis method is used, the input parameters of the heating furnace exergy efficiency are reduced, the training time of the support vector machine model is shortened, and the penalty factor c and the kernel function parameter g are optimized by cross-validation method, thereby improving the calculation accuracy of the support vector machine model. This method improves the calculation accuracy of the heating furnace exergy efficiency and the convergence speed of the calculation method, and has important practical significance and application prospects. 2. When calculating the exergy efficiency of a heating furnace, the present invention proposes a method for calculating the exergy efficiency of a heating furnace based on principal component analysis (PCA) and support vector machine regression (SVR), which can accurately and quickly calculate the exergy efficiency of a heating furnace. Description of the drawings: Figure 1 It is a flow chart of the present invention; Figure 2 This is a visualization diagram of the Pearson correlation coefficient between the factors affecting the exergy efficiency of the heating furnace. Figure 3 This is the calculation result of heating furnace exergy efficiency. Specific implementation method: The present invention will be further described below: See also Figure 1 This method for calculating heating furnace exergy efficiency based on principal component analysis-support vector machine includes the following contents: Step 1: Preprocessing of heating furnace sample data. The heating furnace sample data obtained on site was reviewed. After reviewing the data, the Raida (3σ) criterion was used to eliminate abnormal values ​​of the sample data, and the Lagrange interpolation method was used to fill in the missing data of the sample. (1) The Raida criterion is used to eliminate outliers in sample data; Assume that a set of test data is x 1 , x 2 , x 3 , ..., x n , where n is the number of data, the arithmetic mean t and standard deviation δ of the test data are calculated as follows: If |x i -n|≥3δ, then x i It is abnormal data and should be eliminated; (2) Lagrange interpolation method was used to supplement missing data; Find a known n-1 degree polynomial passing through n points, as follows: y=a 0 +a 1 x+a 2 x 2 +L an-1 x n-1 The coordinates of n points (x 1 ,y 1 ), (x 2 ,y 2 )…,(x n ,y n ) into the polynomial function, we get the following formula: The Lagrange interpolation polynomial is solved as follows: Step 2: Preliminary screening of factors affecting heating furnace exergy efficiency. Taking heating furnace exergy efficiency as the research object, combined with the actual production and operation of heating furnaces, considering factors affecting energy efficiency such as heating furnace inlet and outlet temperature, moisture content, excess air coefficient, etc., establish a factor index data matrix, and use the Pearson correlation coefficient method to preliminarily screen the influencing factor index; (1) Establish factor indicator data matrix M: In the formula, m 11 is the jth influencing factor of the heating furnace at the i-th moment (i=1,2,…,n; j=1,2,…,m). (2) Calculate the Pearson correlation coefficient between each influencing factor: Calculate the Pearson correlation coefficient between each influencing factor index. First, calculate the mean of each factor index. Variance X j and covariance The calculation formula is as follows: Finally, according to the calculated covariance, the correlation coefficient between the influencing factor indicators is calculated. The calculation formula is as follows: In the formula, is the average value of the j-th factor index data; X j is the variance of the j-th factor indicator data; The jth 1 、j 2 The mean of the factor indicators of the columns; The jth row in the i-th row 1 、j 2 Column factor indicator data; For the jth 1 、j 2 covariance of two columns; is the correlation coefficient between two factor indicators. The value range of the Pearson correlation coefficient is [-1,1]. If the absolute value of the Pearson correlation coefficient is larger, it means that the factor indicator is more closely related to the typical reference indicator, and the data should be retained, otherwise it should be eliminated.

[0001] Step 3: Reduce the dimension of the correlation coefficient matrix. Use the MIN-MAX method to standardize the factors affecting the heating furnace exergy efficiency after preliminary screening, use the principal component analysis (PCA) method to reduce the dimension of the calculated standardized matrix, and use the eigenvectors with a cumulative variance contribution rate exceeding 80% as the input of the support vector machine regression; (1)MIN-MAX standardization processing: Since the original data have dimension and magnitude differences among various factor indicators, the MIN-MAX method is required to standardize the data. The standardization formula is: Then we get the normalized matrix N ij =(n ij ) n×m for: (2) Calculate the factor index weights using principal component analysis: First, solve the eigenvalues, variance contribution rates, and cumulative contribution rates of the principal components of the factors: According to the equation |λE-N|=0, calculate the normalization matrix N ij The m eigenvalues ​​of λ are sorted in ascending order 1 >λ 2 >…λ m ≥0, the eigenvector corresponding to each eigenvalue is η 1 ,η 2 ,…,η m . Get m principal components, the calculation formula is: Where P i is the i-th principal component, η ij are the coefficients in the linear combination of the principal components. Extract t principal components from it, then P i Variance contribution rate ρ of (i=1,2,L,m) i for: Cumulative contribution rate δ hl for: Where ρ i P iThe variance contribution rate of (i=1,2,…,m), λ i is the i-th eigenvalue; δ hl is 1 ,ρ 2 ,L,ρ q The cumulative contribution rate of hl When ≥80%, the first t principal components are extracted, which is the comprehensive index of t analysis items. Step 4: Determine the support vector machine training set and test set. Due to the difference in the size of the training sample and the test sample, the regression calculation results will be different. If the training sample is too small, the model will be invalid. On the contrary, if the training sample is too large, the model will be overfitted, resulting in unsatisfactory calculation results. Therefore, the training sample and the calculation sample are determined according to the 7:3 division. Step 5: Select the support vector machine kernel function and determine the parameters to be optimized for the support vector machine model. Select the radial basis kernel function and use the cross-validation method to optimize the penalty factor c and the Gaussian kernel function parameter g. Finally, select the parameter combination that minimizes the training model error and use the optimal parameters to train the support vector machine regression model. (1) Establish support vector machine function: f(x)=ω·x+b Where ω is the weight variable and b is the linear function coefficient. (2) Introducing the relaxation factor ξ i , And the penalty factor c, the calculation model is: The constraints are: Where ε is the insensitive loss function and p is the loss function. Let ω,b,ξ i , The partial derivative of is set to 0, as follows: Convert the quadratic programming problem into a dual problem: The constraints are: Where α i , α j , is the Lagrange multiplier, and Q is the loss function of the Lagrange coefficient. (3) Solve the Lagrangian objective function and use the Gaussian kernel function to obtain the support vector machine regression model, the formula is: Where g is the parameter of the Gaussian kernel function. c and g are important parameters that affect the calculation accuracy of the SVR model. c is the penalty factor, which represents the tolerance of the regression model to errors. Its size affects the generalization ability of the model. g is the parameter of the Gaussian kernel function. Its size determines the number of support vectors, thereby affecting the training speed of the model. (4) Use the cross-validation method to optimize the parameters c and g: Step 6: Calculate the exergy efficiency of the heating furnace. Based on the optimal c and g combination, the support vector machine regression model is trained using the optimal parameters, and the determination coefficient (goodness of fit) R is used. 2 The optimal support vector machine regression model is determined as the heating furnace exergy efficiency calculation model. 2 , RMSE, and MSE calculation formulas are as follows: In the formula is the calculated value, is the mean. Step 7: Calculate the heating furnace exergy efficiency using the heating furnace exergy efficiency calculation model to obtain the heating furnace exergy efficiency index data. Embodiment 1: In order to make the above contents of the present invention more clearly understandable, the actual measured data of a heating furnace in a certain oil field is taken as an example, and the specific method steps are as follows: Step 1: Preprocessing of heating furnace sample data. The heating furnace sample data obtained on site was reviewed. After reviewing the data, the Raida (3σ) criterion was used to eliminate abnormal values ​​of the sample data, and the Lagrange interpolation method was used to fill in the missing data of the sample. The original energy efficiency test data of the heating furnace includes 16 items such as fuel consumption, fuel temperature, ambient temperature, and cold air temperature entering the furnace, totaling 500 sets of heating furnace test data, as shown in Table 1. Table 1 Heating furnace energy efficiency test data (partial) The Raida criterion is used to remove abnormal values ​​of sample data. Taking the exhaust gas temperature data of the heating furnace as an example, 5 abnormal data are identified, which are located in the 25th, 57th, 67th, 75th and 495th rows of the sample data. They are removed, and so on. After removing the abnormal data, there are 480 groups of samples in total. According to the measured data analysis of the above heating furnace, there are fewer abnormal values ​​and they are more scattered; while the missing values ​​are more concentrated and there are more of them. Therefore, a more accurate Lagrange interpolation method is used to supplement them. Taking the average temperature of the outer surface of the furnace body and the ambient temperature as examples, the missing value repair results are shown in Table 2: Table 2 Results of repairing missing values ​​of heating furnace data Step 2: Preliminary screening of factors affecting heating furnace exergy efficiency. Taking heating furnace energy efficiency as the research object, and combining with the actual production and operation of heating furnaces, considering factors affecting heating furnace inlet and outlet temperature, moisture content, excess air coefficient, etc., establish a factor index data matrix, and use the Pearson correlation coefficient method to preliminarily screen the influencing factor index; (1) Establish factor indicator data matrix M: Based on the preprocessed data of heating furnace energy efficiency in step 1, it is matrixed to obtain the factor index data matrix M: The value range of the Pearson correlation coefficient is [-1,1]. If the absolute value of the Pearson correlation coefficient is larger, it means that the factor indicator is more closely related to the typical reference indicator, and the data should be retained, otherwise it should be eliminated. (2) Calculate the Pearson correlation coefficient between each influencing factor: Calculate the Pearson correlation coefficient between each influencing factor index. First, calculate the mean of each factor index. Variance X j and covariance Exergy efficiency and smoke exhaust 2 Take content as an example: Efficiency mean Variance X 1 : Smoke exhaust O 2 Content mean Variance X 1 : Then the covariance between the two is X j1j2 : Similarly, the mean of the remaining factor indicators can be obtained Variance X j and the covariance X j1j2 As shown in Table 3. Table 3 Calculation results of mean, variance and covariance of factor indicators Secondly, the Pearson correlation coefficient was calculated based on the exergy efficiency and the exhaust gas O 2 Taking the two factor indicators of content as an example, the calculation process of Pearson correlation coefficient is explained. Similarly, the calculation results of the Pearson correlation coefficients between exergy efficiency and other factor indicators are shown in Table 4. Table 4 Factors affecting heating furnace energy efficiency—Pearson correlation coefficient In order to intuitively reflect the closeness of the relationship between the typical reference sequence index and the other factor indexes, they are plotted as shown in the attached figure. Figure 2 The visualization shown. From Table 4 and Figure 2 It can be seen that the correlation coefficient between the heating furnace exergy efficiency and the six factors of CO at the exhaust, the average temperature of the furnace outer surface, the fuel temperature, the temperature of the cold air entering the furnace, the ambient temperature, and the outlet temperature is small, and the close relationship is small. Therefore, the six factors should be eliminated, which not only ensures the accuracy of the calculation effect of the heating furnace exergy efficiency, but also shortens the model running speed. The influencing factors of the heating furnace exergy efficiency determined by the Pearson correlation coefficient method are RO at the exhaust 2 , medium flow, excess air coefficient, load rate, exhaust point O 2 , fuel consumption, received basic low calorific value, inlet temperature and exhaust gas temperature are used as influencing factors of heating furnace exergy efficiency. Step 3: Reduce the dimension of the correlation coefficient matrix. Use the MIN-MAX method to standardize the factors affecting the heating furnace exergy efficiency after preliminary screening, use the principal component analysis (PCA) method to reduce the dimension of the calculated standardized matrix, and use the eigenvectors with a cumulative variance contribution rate exceeding 80% as the input of the support vector machine regression; (1)MIN-MAX standardization processing: The MIN-MAX standardization method is used to standardize the impact factor index data and construct the screened factor index data matrix N: Based on the factor index standardization data, it is matrixed to obtain the factor index normalized data matrix N ij : (2) Use principal component analysis to reduce data dimensionality The principal component analysis method is used to reduce the dimension of the factors affecting the exergy efficiency of the heating furnace. First, the eigenvalues, variance explanation rate and cumulative contribution rate of the principal components of the factors are solved. According to the equation: The 9 principal component eigenvalues ​​are: 1 =2.624,λ 2 =1.725,λ 3 =1.195,λ 4 =1.132,λ 5 =0.828,λ 6 =0.687,λ 7 =0.482,λ 8 =0.303,λ 9 =0.025. The variance explanation rate and cumulative contribution rate of each principal component are calculated and shown in Table 5. Table 5 Initial eigenvalues ​​of factor principal component analysis The above table shows that the principal component analysis extracted 4 principal components, all of which have characteristic root values ​​greater than 1. The variance explanation rates of these 4 principal components are 29.152%, 19.162%, 13.277% and 16.253%, respectively, and the cumulative contribution rate is 74.168%. Therefore, the first 4 are selected as principal components, and the information extraction and corresponding relationship with the remaining research items are analyzed. Based on the normalized data matrix R ij , according to the equation The linear combination coefficients of the four principal components of each factor index are shown in Table 10. Table 6 Linear combination coefficients of each factor index for the principal component The linear combination coefficients are multiplied by the variance explanation rate and then added up, and then divided by the cumulative explanation rate to determine the comprehensive score coefficients of each factor indicator, which are then normalized to obtain the weight values ​​of each factor indicator. The calculation results are shown in Table 7. Table 7 Comprehensive score coefficients and weights of each factor indicator It can be seen from Table 7 that the weights of the factor indicators determined by the principal component analysis method are excess air coefficient, exhaust RO 2 、Smoke exhaust O 2, fuel consumption, load rate, inlet temperature, received basic low calorific value, exhaust temperature and medium flow rate. Among them, the weights of exhaust temperature and medium flow rate are relatively small and should be further eliminated. Finally, the 7 heating furnace exergy efficiency input factors are determined. Step 4: Determine the support vector machine training set and test set. Due to the difference in the size of training samples and test samples, the regression calculation results will be different. If the training samples are too few, the model will be invalid. On the contrary, if the training samples are too many, the model will be overfitted, resulting in unsatisfactory calculation results. Therefore, the training samples and calculation samples are determined according to the 7:3 division. That is, the training samples are 336 items and the test samples are 144 items. Step 5: Select the support vector machine kernel function and determine the parameters to be optimized for the support vector machine model. The radial basis kernel function is selected, and the penalty factor c and Gaussian kernel function parameter g of different proportions of training samples and test samples are optimized by cross-validation method to determine the calculation accuracy of the support vector machine regression model under the optimal c and g. After calculation, bestc=2; bestg=0.3536. Step 6: Calculate the exergy efficiency of the heating furnace. Based on the optimal c and g combination, the support vector machine regression model is trained using the optimal parameters, and the determination coefficient (goodness of fit) R is used. 2 The calculation data of heating furnace exergy efficiency index is obtained by taking the model accuracy standard as the main one and the root mean square error RMSE and mean absolute error MSE as the auxiliary ones. Bring the optimal penalty factor c and radial basis vector g into the support vector machine calculation model and calculate its determination coefficient R 2 , RMSE, and MSE are shown in Table 8 below. Table 8 Support vector machine regression model R 2 , RMSE, MSE Step 7: Schematic diagram of the calculation results of the heating furnace based on the combination of principal component analysis and support vector machine is as follows Figure 3 . Based on principal component analysis and support vector machine regression model, the exergy efficiency of heating furnace can be calculated accurately and quickly with a small amount of sample data. This method is conducive to improving the quality and efficiency of high-energy-consuming equipment such as heating furnaces.

Claims

1. A method for calculating the exergy efficiency of a heating furnace, characterized in that The steps include: Step 1: Preprocessing of heating furnace sample data: reviewing the heating furnace sample data obtained on site, removing abnormal values ​​of the sample data, and filling in the missing data of the sample; Step 2: Preliminary screening of factors affecting heating furnace exergy efficiency. Taking heating furnace exergy efficiency as the object and combining it with the actual production and operation of heating furnace, factors affecting energy efficiency are considered, including heating furnace inlet and outlet temperature, moisture content, and excess air coefficient. A factor index data matrix is ​​established to preliminarily screen influencing factor indexes. Step 3: Reduce the dimension of the correlation coefficient matrix, standardize the factors affecting the heating furnace exergy efficiency after preliminary screening, reduce the dimension of the calculated standardized matrix, and use the eigenvectors with cumulative variance contribution rate exceeding 80% as the input of support vector machine regression; Step 4: Determine the support vector machine training set and test set, and determine the training samples and test samples according to the 7:3 division; Step 5: Select the support vector machine kernel function, determine the parameters to be optimized for the support vector machine model, optimize the penalty factor c and the Gaussian kernel function parameter g, and finally select the parameter combination that minimizes the training model error, and use the optimal parameters to train the support vector machine regression model; Step 6: Use the test set to test the trained support vector machine regression model, using the determination coefficient R 2 The optimal support vector machine regression model is determined as the heating furnace exergy efficiency calculation model, with the root mean square error RMSE and mean absolute error MSE as the auxiliary model accuracy measurement standard. In the formula is the calculated value, is the mean Step 7: Calculate the heating furnace exergy efficiency using the heating furnace exergy efficiency calculation model.

2. The method for calculating heating furnace exergy efficiency according to claim 1, characterized in that: The step 1 is specifically as follows: (1) The Raida criterion is used to eliminate outliers in sample data; Assume that a set of test data is x1, x2, x3, ..., x n , where n is the number of data, the arithmetic mean t and standard deviation δ of the test data are calculated as follows: If |x i -n|≥3δ, then x i Remove abnormal data; (2) Use Lagrange interpolation method to fill in the missing data.

3. The method for calculating heating furnace exergy efficiency according to claim 2, characterized in that: The step 2 is specifically as follows: (1) Establish factor indicator data matrix M: In the formula, m 11 is the jth influencing factor of the heating furnace at the i-th moment (i=1,2,…,n; j=1,2,…,m); (2) Calculate the Pearson correlation coefficient between each influencing factor: Calculate the Pearson correlation coefficient between each influencing factor index. First, calculate the mean of each factor index. Variance X j and covariance The calculation formula is as follows: Finally, according to the calculated covariance, the correlation coefficient between the influencing factor indicators is calculated. The calculation formula is as follows: In the formula, is the average value of the j-th factor index data; X j is the variance of the j-th factor indicator data; are the average values ​​of the factor indicators in columns j1 and j2, respectively; is the factor indicator data of the i-th row, j1-th column, and j2-th column; is the covariance of the j1th and j2th columns; is the correlation coefficient between the two factor indicators; The value range of the Pearson correlation coefficient is [-1,1]. If the absolute value of the Pearson correlation coefficient is larger, it means that the factor indicator is more closely related to the typical reference indicator, and the data should be retained, otherwise it should be eliminated.

4. The method for calculating heating furnace exergy efficiency according to claim 3, characterized in that: The step three is specifically as follows: (1) The MIN-MAX method was used to standardize the factors affecting the heating furnace exergy efficiency after preliminary screening to obtain a normalized matrix; (2) The principal component analysis method is used to standardize the matrix for dimensionality reduction, and the eigenvectors with cumulative variance contribution exceeding 80% are used as the input of the support vector machine regression: First, solve the eigenvalues, variance contribution rates, and cumulative contribution rates of the principal components of the factors: According to the equation |λE-N|=0, calculate the normalization matrix N ij The m eigenvalues ​​of are sorted in ascending order λ1>λ2>…λ m ≥0, the eigenvalues ​​corresponding to the eigenvectors are η1, η2, …, η m , and get m principal components: Where P i is the i-th principal component, η ij is the coefficient in the linear combination of each principal component; Extract t principal components from it, then P i Variance contribution rate ρ of (i=1,2,…,m) i for: Cumulative contribution rate δ hl for: Where ρ i P i The variance contribution rate of (i=1,2,…,m), λi is the i-th eigenvalue; δ hl is ρ1,ρ2,…,ρ q The cumulative contribution rate of hl When ≥80%, the first t principal components are extracted, which is the comprehensive index of t analysis items.

5. The method for calculating heating furnace exergy efficiency according to claim 4, characterized in that: The step five is specifically as follows: (1) Establish support vector machine function: f(x)=ω·x+b Where ω is the weight variable, b is the linear function coefficient; (2) Introducing the relaxation factor ξ i , And the penalty factor c, the calculation model is: The constraints are: Where ε is the insensitive loss function, and p is the loss function; Let ω,b,ξ i , The partial derivative of is set to 0, as follows: Convert the quadratic programming problem into a dual problem: The constraints are: Where α i , α j , is the Lagrange multiplier, Q is the loss function of the Lagrange coefficient; (3) Solve the Lagrangian objective function and use the Gaussian kernel function to obtain the support vector machine regression model: Where g is the parameter of the Gaussian kernel function; c is the penalty factor, which represents the tolerance of the regression model to errors, and its size affects the generalization ability of the model; g is the parameter of the Gaussian kernel function, and its size determines the number of support vectors; (4) Use the cross-validation method to optimize the parameters c and g.

Images