Method for predicting oxide scale amount of once-through boiler pipeline based on ant colony optimization SVM

Abstract

The invention discloses a once-through boiler pipeline oxide scale amount prediction method based on ant colony optimization SVM, and belongs to the technical field of once-through boiler online monitoring. The method comprises the following steps: performing magnetic detection on a once-through boiler superheater pipeline, taking a magnetic induction intensity signal of the superheater pipeline as input, taking the oxide skin stacking height in the once-through boiler superheater pipeline as output, establishing a once-through boiler database, selecting and preprocessing modeling data, and ensuring the predictive ability and generalization ability of a model; carrying out optimization training on the SVM model through an ant colony optimization algorithm, searching an optimal penalty factor C and a Gaussian kernel function parameter g, and obtaining an optimal oxide skin accumulation amount prediction model; predicting the test sample according to the oxide skin accumulation amount prediction model; an ant colony optimization algorithm is used for optimizing a penalty factor C and a Gaussian kernel function parameter g of an SVM, the accuracy of oxide skin accumulation amount prediction is improved, and the prediction precision and generalization ability of the model are guaranteed by selecting and preprocessing data.

Description

Technical Field

[0001] The invention relates to the technical field of online monitoring of direct current boilers, and in particular to a method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM. Background Art

[0002] In recent years, as my country's newly commissioned thermal power generating units gradually develop towards large capacity and high parameters, pipe burst accidents caused by blockage of oxide scale in high-temperature tubes of supercritical boilers have occurred frequently. Pipe oxidation corrosion causes the oxide scale to fall off on the steam side of the pipeline. The oxide scale particles enter the steam turbine with the steam, causing the turbine blades to be eroded by solid particles, which has brought huge economic losses to power plants and society. For a long time, due to the fact that the oxide scale problem involves many aspects of professional knowledge and the lack of effective means to detect the thickness of the oxide scale, domestic research on this issue has been relatively backward, and there is a lack of sufficient understanding of the formation mechanism and prevention measures of boiler high-temperature oxide scale. Foreign research focuses more on the impact of the oxide scale problem on the safe and economic operation of the steam turbine, and there is relatively little research on the impact of the oxide scale on the operation of the boiler and how to prevent the oxide scale from peeling through operation technology management. By consulting Chinese and foreign literature, it is found that the commonly used detection technologies for oxide scale in supercritical boiler tubes are: X-ray detection technology, ultrasonic detection technology and magnetic field detection technology. The ultrasonic detection method is aimed at the oxide scale that has not yet separated from the tube wall in the early stage of formation, and mainly detects the growth thickness of the oxide scale based on the echo principle of the tube wall interface. The X-ray detection method uses imaging based on the absorption of radiation by oxide particles to achieve the purpose of observing the accumulation of oxide particles. The magnetic detection method uses the magnetized oxide signal inside the paramagnetic austenite tube to measure the presence or accumulation of oxide particles. The above detection methods are all used to detect oxide particles inside the superheater and reheater during boiler shutdown and maintenance. There is no effective means to evaluate oxide particles during boiler operation. American generators use the PAC detection system to realize online monitoring of oxide shedding in the steam-water system of the generator boiler. The system evaluates various parameters of oxide in steam by detecting the vibration impact signal of specific sensors when the oxide moves, but the high temperature and high pressure environment inside the main steam pipeline is a great test for the penetration and service life of the sensor.

[0003] The support vector machine method relies on fewer data samples, has a fast convergence speed, and can make full use of the real-time operating status information of the direct current boiler to predict the amount of oxide scale accumulation. Therefore, the present invention introduces the ant colony optimization support vector machine (ACO-SVM), uses state evaluation to obtain various parameters in the operation process of the direct current boiler, and trains the current operating parameters combined with the statistical data of historical operating parameters to achieve the prediction of the amount of oxide scale accumulation. Then, the ant colony algorithm is used to optimize the two important parameters of the support vector machine (penalty parameter C and Gaussian kernel function parameter g), which can improve the accuracy of the prediction of the amount of oxide scale accumulation. Summary of the invention

[0004] The purpose of the present invention is to provide a method for predicting the amount of oxidized iron scale in direct current boiler pipes based on ant colony optimization SVM to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for predicting the amount of iron oxide scale in a direct current boiler pipeline based on ant colony optimization SVM, comprising: obtaining magnetic induction intensity data in a direct current boiler superheater pipeline, obtaining oxide scale accumulation height data in the direct current boiler superheater pipeline, and establishing a direct current boiler database; selecting and preprocessing the data in the direct current boiler database to obtain normalized sample data; optimizing and training the SVM model through an ant colony optimization algorithm to obtain an optimal oxide scale accumulation prediction model; and predicting the test sample according to the optimal oxide scale accumulation prediction model.

[0006] As a preferred solution of the method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM described in the present invention, wherein: magnetic detection is performed on the direct current boiler superheater pipeline, and establishing a direct current boiler database includes the following steps: magnetic detection is performed on the direct current boiler superheater pipeline by an instrument, the magnetic induction intensity at various locations of the superheater pipeline is collected, and the accumulation height of the oxidized iron scale in the direct current boiler superheater pipeline is established.

[0007] As a preferred solution of the method for predicting the amount of iron scale on pipes of direct current boilers based on ant colony optimization SVM of the present invention, the following criteria are also included when selecting the historical data of the operating parameters and the amount of iron scale accumulation collected during the operation of the direct current boiler: ① sufficient dispersion, that is, the Euclidean distance between the points formed by the coordinates of each group of input and output is greater than n, where n is a constant; ② sufficient quantity, for modeling the amount of iron scale accumulation, the number of training samples is not less than 30 groups to ensure the prediction ability of the model;

[0008] As a preferred solution of the method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM described in the present invention, the preprocessing of the historical data of various operating parameters and the amount of oxidized iron scale accumulation collected during the operation of the direct current boiler also includes: clearing duplicate data in the direct current boiler database; performing node sequence numbering on the data in the direct current boiler database after the clearing operation, and smoothing the missing data in the direct current boiler database after the clearing operation through the Hermite interpolation algorithm; and normalizing the data in the direct current boiler database after the clearing and smoothing operations to generate sample data.

[0009] The Hermite interpolation formula is:

[0010]

[0011] Where: y i =y(x i ),y i '=y'(x i )

[0012]

[0013] Among them, H(x) is the interpolation polynomial, h i and a i are the first interpolation intermediate term and the second interpolation intermediate term of the i-th node, respectively, and the first interpolation intermediate term is located between the i-th node and the i-1-th node, and the second interpolation intermediate term is located between the i-th node and the i+1-th node, n is the total number of nodes, and the node refers to the data point in the interpolation. The node data is used as the horizontal coordinate, and the corresponding value of the node data is used as the vertical coordinate. x is the horizontal coordinate, y is the vertical coordinate, and x is the horizontal coordinate. i 、x j Respectively represent the i-th and j-th node data, y i is the corresponding value of the i-th node data, the value range of i and j is [1, n], y' and y i 'y and y respectively i The derivative of .

[0014] The data in the direct current boiler database after the clearing and smoothing operations are normalized to generate sample data. The specific method of normalizing the data is as follows:

[0015]

[0016] Among them, X represents the original data that needs to be normalized, X max and X min They represent the maximum and minimum values ​​in the array that need to be normalized, Y represents the normalized data, and Y max and Y minThey represent the range parameters that need to be mapped respectively. According to the normalized data, the actual output vector z is calculated. i .

[0017] As a preferred solution of the method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM described in the present invention, the detailed steps of: optimizing and training the SVM model through the ant colony optimization algorithm to obtain the optimal oxide scale accumulation prediction model are: initializing the support vector machine parameter penalty factor C and kernel function parameter g; training the support vector machine using sample data; setting the maximum number of iterations, and using the ant colony algorithm to perform global optimization on the penalty factor C and kernel function parameter g; judging whether the maximum number of iterations has been reached, and if so, terminating the optimization to obtain the globally optimal prediction model; if not, continuing the optimization to finally obtain the optimal penalty factor C, kernel function parameter g and optimal value of the objective function.

[0018] The method of training the support vector machine using sample data is characterized in that:

[0019] Construct the classification decision function f(m):

[0020]

[0021] Among them, m i ∈R n ,n i ∈{-1,1},i=1,2,...,n,m i is the i-th eigenvector, n i is m i The corresponding class tag, 0≤d i ≤c,d i is the Lagrange multiplier, e is the intercept; if n i =1, it means that for m i The classification is accurate, n i = -1 means that for m i Misclassification of

[0022] K is the kernel function for constructing the classification decision function; Gaussian radial basis kernel function (RBF) is selected, and the specific expression of kernel function K is:

[0023]

[0024] Where m c is the center of the kernel function;

[0025] The normalized sample data is input into the constructed support vector machine to optimize the classification boundary and obtain the predicted output vector z i * .

[0026] As a preferred solution of the method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM of the present invention, wherein: the relevant parameters of the ant colony algorithm are initialized; the parameters of the ant colony algorithm include the initialization ant colony scale M, the maximum number of iterations C max , information evaporation coefficient ρ, heuristic factor α and expected heuristic factor β; binary encode the penalty factor C and kernel function parameter g to form a binary string; construct a binary directed graph G = (V, S); initialize the pheromone concentration; calculate the probability P ij (t) and according to the probability P ij (t) Determine the value of the next node until all ants complete the search process; calculate the penalty factor C of the new iteration * and kernel function parameter g * The value of; respectively calculate the predicted output vector z obtained by the original support vector machine i * And the predicted output vector z obtained by the support vector machine after optimizing the parameters i ** With the actual output vector z i The minimum value of the total error is obtained, and the global optimization is performed with the minimum total error as the objective function; whether minF(t)≤minF(t-1) holds, if so, the pheromone on the road section corresponding to the optimal solution is updated, if not, it is not updated;

[0027] The prediction output vector z of the support vector machine is calculated i * The actual output vector z obtained from the normalized sample data i The calculation formula for the minimum value of the total error minF is:

[0028]

[0029] The pheromone concentration τ of the t+1th search node after update i,j (t+1) is:

[0030] τ i,j (t+1)=(1-λ)τ i,j (t)+Δτ i,j

[0031] Where λ is the pheromone volatility coefficient, the preferred value of λ is 0<λ<1, Δτ i,j is the change in pheromone from node i to node j for all ants in this cycle.

[0032] Preferably, in the binary directed graph G=(V, S), the node set V is:

[0033]

[0034] The expression of the directed arc set S is:

[0035]

[0036] Among them, v s is the starting node, node and Respectively represent b j The state with values ​​of 0 and 1;

[0037] Preferably, the probability P ij (t), the probability P ij The calculation formula of (t) is:

[0038]

[0039] Among them, τ i,j is the pheromone concentration from node i to node j, η i,j is the inverse of the distance from node i to node j, k is the number of candidate solutions x corresponding to node i in the t-th search cycle s The number of bits in the binary code.

[0040] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: firstly, the penalty factor C and the Gaussian kernel function parameter g of the SVM are automatically optimized by using the ant colony algorithm to find a better parameter combination, thereby improving the performance of the SVM and the accuracy of the prediction of the oxide scale accumulation amount; secondly, the data in the database are selected and preprocessed to ensure the prediction accuracy of the model; finally, the theoretical basis of the ant colony algorithm determines that it ultimately obtains the global optimal value, rather than the local minimum value. This ensures the good generalization ability of the SVM for unknown samples and avoids overfitting. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0042] in:

[0043] Figure 1 It is a flow chart of a method for predicting the amount of oxidized iron scale in direct current boiler pipes based on ant colony optimization SVM according to an embodiment of the present invention.

[0044] Figure 2 It is a flow chart of establishing a globally optimal support vector machine prediction model in the present invention. DETAILED DESCRIPTION

[0045] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] See also Figure 1-Figure 2 The present invention provides a method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM, comprising:

[0047] S1: Conduct magnetic detection on the DC boiler superheater pipe, establish a DC boiler database, collect magnetic induction intensity signals at various locations on the superheater pipe as input, and the oxide scale accumulation height in the DC boiler superheater pipe as output.

[0048] S2: Select and preprocess the data in the database to obtain normalized sample data.

[0049] Furthermore, the following criteria are also included when selecting the historical data of the operating parameters and scale accumulation during the operation of the once-through boiler:

[0050] S2.1: Sufficient dispersion, that is, the Euclidean distance between the points formed by the coordinates of each set of input and output is greater than n, where n is 3;

[0051] S2.2: Sufficient number of training samples for modeling scale accumulation, with the number of training samples not less than 30 groups, to ensure the predictive ability of the model;

[0052] Specifically, preprocessing the collected historical data of various operating parameters and oxide scale accumulation during the operation of the once-through boiler also includes the following process:

[0053] The duplicate data in the DC boiler database are cleared; the nodes of the data in the DC boiler database after the clearing operation are sequentially numbered, and the missing data in the DC boiler database after the clearing operation is smoothed by the Hermite interpolation algorithm; the data in the DC boiler database after the clearing and smoothing operations are normalized to generate sample data.

[0054] Furthermore, the Hermite interpolation formula is:

[0055]

[0056] Where: y i =y(x i ),y i'=y'(x i )

[0057]

[0058] Among them, H(x) is the interpolation polynomial, h i and a i are the first interpolation intermediate term and the second interpolation intermediate term of the i-th node, respectively, and the first interpolation intermediate term is located between the i-th node and the i-1-th node, and the second interpolation intermediate term is located between the i-th node and the i+1-th node, n is the total number of nodes, and the node refers to the data point in the interpolation. The node data is used as the horizontal coordinate, and the corresponding value of the node data is used as the vertical coordinate. x is the horizontal coordinate, y is the vertical coordinate, and x is the horizontal coordinate. i 、x j Respectively represent the i-th and j-th node data, y i is the corresponding value of the i-th node data, the value range of i and j is [1, n], y' and y i ' are the derivatives of y and yi respectively.

[0059] Furthermore, the data in the direct current boiler database after the clearing and smoothing operations are normalized to generate sample data. The specific method of normalizing the data is as follows:

[0060]

[0061] Among them, X represents the original data that needs to be normalized, X max and X min They represent the maximum and minimum values ​​in the array that need to be normalized, Y represents the normalized data, and Y max and Y min They represent the range parameters that need to be mapped. They can be selected as [0,1] or [-1,1] according to the actual situation. Based on the normalized data, the actual output vector z is calculated. i .

[0062] S3: The SVM model is optimized and trained by the ant colony optimization algorithm to obtain the optimal oxide scale accumulation prediction model.

[0063] Further, in S3, the SVM model is optimized and trained by using an ant colony optimization algorithm to obtain an optimal oxide scale accumulation prediction model, including the following steps:

[0064] S3.1: Initialize the support vector machine parameters penalty factor C and kernel function parameter g.

[0065] S3.2: Use sample data to train the support vector machine.

[0066] Specifically, the construction of the classification decision function f(m) is provided:

[0067]

[0068] Among them, m i ∈R n ,n i ∈{-1,1},i=1,2,...,n,m i is the i-th eigenvector, n i is m i The corresponding class tag, 0≤d i ≤c,d i is the Lagrange multiplier, e is the intercept; if n i =1, it means that for m i The classification is accurate, n i = -1 means that for m i Misclassification of

[0069] Specifically, a kernel function K for constructing a classification decision function is provided; a Gaussian radial basis kernel function (RBF) is selected, and the expression of the kernel function K is:

[0070]

[0071] Where m c is the center of the kernel function;

[0072] Specifically, the normalized sample data is input into the constructed support vector machine, the classification boundary is optimized, and the prediction output vector z is obtained. i * .

[0073] S3.3: Set the maximum number of iterations and use the ant colony algorithm to globally optimize the penalty factor C and kernel function parameter g.

[0074] Furthermore, in S3.3, the penalty factor C and the kernel function parameter g are globally optimized using the ant colony algorithm, including the following steps:

[0075] S3.3.1: Initialize the relevant parameters of the ant colony algorithm; the parameters of the ant colony algorithm include the initial ant colony size M, the maximum number of iterations C max , information evaporation coefficient ρ, heuristic factor α and expected heuristic factor β;

[0076] S3.3.2: Binary encode the penalty factor C and the kernel function parameter g to form a binary string;

[0077] S3.3.3: Construct a binary directed graph G = (V, S).

[0078] S3.3.4: Initialize pheromone concentration.

[0079] S3.3.5: Calculate the probability P ij (t) and according to the probability P ij (t) Determine the value of the next node until all ants complete the search process.

[0080] S3.3.6: Calculate the penalty factor C for the new iteration * and kernel function parameter g * The value of .

[0081] S3.3.7: Calculate the predicted output vector z obtained by the original support vector machine i * And the predicted output vector z obtained by the support vector machine after optimizing the parameters i ** With the actual output vector z i The minimum value of the total error is obtained, and the global optimization is performed with the minimum total error as the objective function.

[0082] Specifically, the prediction output vector z of the support vector machine is i * The actual output vector z obtained from the normalized sample data i The calculation formula for the minimum value of the total error minF is:

[0083]

[0084] S3.3.8: Determine whether minF(t)≤minF(t-1) holds. If so, update the pheromone on the road section corresponding to the optimal solution. If not, do not update.

[0085] Specifically, the pheromone concentration τ of the t+1th search node after update i,j (t+1) is:

[0086] τ i,j (t+1)=(1-λ)τ i,j (t)+Δτ i,j

[0087] Where λ is the pheromone volatility coefficient, the preferred value of λ is 0<λ<1, Δτ i,j is the change in pheromone from node i to node j for all ants in this cycle.

[0088] Specifically, in the binary directed graph G=(V,S), the node set V is:

[0089]

[0090] The expression of the directed arc set S is:

[0091]

[0092] Among them, v s is the starting node, node and Respectively represent b j The state with values ​​of 0 and 1;

[0093] Specifically, the probability P ij (t), the probability P ij The calculation formula of (t) is:

[0094]

[0095] Among them, τ i,j is the pheromone concentration from node i to node j, η i,j is the inverse of the distance from node i to node j, k is the number of candidate solutions x corresponding to node i in the t-th search cycle s The number of bits in the binary code.

[0096] S3.4: Determine whether the maximum number of iterations has been reached. If so, terminate the optimization search and obtain the globally optimal prediction model.

[0097] S3.5: If the maximum number of iterations is not reached, continue to search for the best solution and eventually obtain the optimal penalty factor C, kernel function parameter g and optimal value of the objective function.

[0098] S4: Predict the test sample based on the oxide scale accumulation prediction model.

[0099] Furthermore, the magnetic induction intensity signal at each location of the direct current boiler superheater pipe is used as input, and the output is the accumulation amount of oxide scale, thereby realizing the prediction of the accumulation amount of oxide scale.

[0100] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0101] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM, the method comprising: Step 1: Obtain the magnetic induction intensity data in the superheater pipe of the direct current boiler, obtain the accumulation height data of the oxide scale in the superheater pipe of the direct current boiler, and establish a direct current boiler database; Step 2: Select and preprocess the data in the once-through boiler database to obtain normalized sample data; It is characterized by further comprising: Step 3: Optimize and train the SVM model through the ant colony optimization algorithm to obtain the optimal oxide scale accumulation prediction model; Step 4: predicting the test sample according to the optimal oxide scale accumulation prediction model; Wherein, the implementation method of step 2 includes: S21: Clearing duplicate data in the once-through boiler database; S22: performing node sequence numbering on the data in the DC boiler database after the clearing operation, and smoothing the missing data in the DC boiler database after the clearing operation by using the Hermite interpolation algorithm; S23: Normalize the data in the DC boiler database after the clearing and smoothing operations to generate sample data.

2. The method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM according to claim 1 is characterized in that: The Hermite interpolation formula is: Where: y i = y(x i ), y i ' = y'(x i ) Among them, H(x) is the interpolation polynomial, h i and a i are the first interpolation intermediate term and the second interpolation intermediate term of the i-th node, respectively, and the first interpolation intermediate term is located between the i-th node and the i-1-th node, and the second interpolation intermediate term is located between the i-th node and the i+1-th node, n is the total number of nodes, and the node refers to the data point in the interpolation. The node data is used as the horizontal coordinate, and the corresponding value of the node data is used as the vertical coordinate. x is the horizontal coordinate, y is the vertical coordinate, and x is the horizontal coordinate. i 、x j Respectively represent the i-th and j-th node data, y i is the corresponding value of the i-th node data, the value range of i and j is [1, n], y' and y i 'y and y respectively i The derivative of .

3. The method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM according to claim 1 is characterized in that: The formula for the normalization process is: Among them, X is the original data that needs to be normalized, X max and X min is the maximum and minimum value in the array to be normalized, Y is the normalized data, and Y max and Y min is the range parameter that needs to be mapped. According to the normalized data, the actual output vector z is calculated. i .

4. The method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM according to claim 1 is characterized in that: The optimization training of the SVM model by the ant colony optimization algorithm to obtain the optimal oxide scale accumulation prediction model comprises the following steps: Initialize the support vector machine parameter penalty factor C and kernel function parameter g; Using the sample data to train a support vector machine; Set the maximum number of iterations and use the ant colony algorithm to globally optimize the penalty factor C and kernel function parameter g; Determine whether the maximum number of iterations has been reached. If so, end the optimization search and obtain the global optimal prediction model. If the maximum number of iterations is not reached, the algorithm is reinitialized and the optimization is continued to finally obtain the optimal penalty factor C, kernel function parameter g and optimal value of the objective function.

5. The method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM according to claim 4 is characterized in that: The use of sample data to train the support vector machine includes: Construct the classification decision function f(m): Among them, m i ∈R n ,n i ∈{-1,1},i=1,2,...,n,m i is the i-th eigenvector, n i is m i The corresponding class tag, 0≤d i ≤c,d i is the Lagrange multiplier, e is the intercept; K is the kernel function for constructing the classification decision function, and its expression is: Where m c is the center of the kernel function; The normalized sample data is input into the constructed support vector machine to optimize the classification boundary and obtain the predicted output vector z i * .

6. The method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM according to claim 4 is characterized in that: The setting of the maximum number of iterations and the use of the ant colony algorithm to perform global optimization on the penalty factor C and the kernel function parameter g include the following steps: Initialize the ant colony algorithm parameters; Binary encoding the penalty factor C and the kernel function parameter g to form a binary string; Construct a binary directed graph G = (V, S); Initialize pheromone concentration; Calculate the probability P ij (t), and according to the probability P ij (t) Decide the value of the next node until all ants complete the search process; Calculate the penalty factor C for the new iteration * and kernel function parameter g * The value of Calculate the predicted output vector z obtained by the original support vector machine respectively i * And the predicted output vector z obtained by the support vector machine after optimizing the parameters i ** The actual output vector z obtained in step 2 i The minimum value of the total error is obtained, and the global optimization is performed with the minimum total error as the objective function; Determine whether min F(t)≤min F(t-1) holds. If so, update the pheromone on the road section corresponding to the optimal solution. If not, do not update.

7. The method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM according to claim 6 is characterized in that: The calculation formula of the minimum value minF is: The pheromone concentration τ of the t+1th search node after update i,j (t+1) is: t i,j (t+1)=(1-λ)τ i,j (t)+Δτ i,j Among them, λ is the pheromone volatility coefficient, Δτ i,j is the change in pheromone from node i to node j for all ants in this cycle.

8. The method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM according to claim 4 is characterized in that: In the binary directed graph G=(V,S), the node set V is: The expression of the directed arc set S is: Among them, v s is the starting node, node and Respectively represent b j The state with values ​​of 0 and 1; The probability P ij The calculation formula of (t) is: Among them, τ i,j is the pheromone concentration from node i to node j, η i,j is the inverse of the distance from node i to node j, k is the number of candidate solutions x corresponding to node i in the t-th search cycle s The number of bits in the binary code.

9. The method for predicting the amount of iron oxide scale in direct current boiler pipelines based on ant colony optimization SVM according to claim 1 is characterized in that: The magnetic detection is performed on the superheater pipe of the direct current boiler by means of an instrument, the magnetic induction intensity at various locations of the superheater pipe and the accumulation height of oxide scale in the superheater pipe of the direct current boiler are collected, and a direct current boiler database is established.

10. The method for predicting the amount of oxidized iron scale in direct current boiler pipelines based on ant colony optimization SVM according to claim 1 is characterized in that: The sample data for modeling in the direct current boiler database is selected in accordance with the following principles: sufficient dispersion, that is, the Euclidean distance between the points composed of coordinates of each group of input and output quantities is greater than n, where n is a constant; sufficient quantity, for modeling of scale accumulation, the number of training samples is not less than 30 groups.

Images