Boiler temperature field distribution optimization control method based on multi-source data driving

Abstract

The invention discloses a boiler temperature field distribution optimization control method based on multi-source data driving, belongs to the technical field of coal-fired unit regulation and control, and aims to solve the problem that the temperature of each point of a whole hearth cannot be covered in the process of optimizing temperature field distribution in the boiler hearth. Comprising the following steps: performing numerical simulation on a combustion state in a boiler to obtain hearth temperature field numerical simulation data under typical operation conditions; carrying out multi-source data fusion on the operation process parameter set and the operation process parameter set; classifying the fused data set according to the boiler operation load; a KNN algorithm is adopted to construct a hearth temperature distribution soft measurement model, and boiler hearth temperature distribution conditions under different working conditions are predicted; selecting a controllable variable, and optimizing the controllable variable; the ideal temperature field distribution serves as an optimization target, a cultural gene algorithm MA integrating a genetic algorithm GA and a hill-climbing search HC is adopted, a secondary air volume optimal solution set meeting the optimization target is obtained through calculation, and combustion of the boiler is regulated and controlled. The method is used for detecting the internal combustion state of the coal-fired boiler.

Description

Technical Field

[0001] The invention relates to a boiler temperature field distribution optimization control method, belonging to the technical field of coal-fired unit regulation. Background Art

[0002] Today, with the vigorous development of new energy, new energy sources with wind and solar power as the main power sources have the characteristics of intermittent nature, unstable power generation, and poor controllability. Thermal power generation, which is mainly coal-fired, has the advantages of fast response speed, flexible site selection, relatively small impact from geographical and climatic environment, and stable output. It can achieve rapid peak shaving and valley filling of the power grid, effectively make up for the shortage of new energy power generation, and plays an irreplaceable role in ensuring the stability and security of the power grid system.

[0003] In the process of thermal power generation, the quality of the internal combustion state of the coal-fired boiler directly affects the safety, efficiency and environmental protection of the boiler system. The temperature field distribution in the furnace is the core parameter reflecting the combustion state. Real-time analysis and acquisition of the temperature field distribution state in the furnace and regulation of the combustion process accordingly are the current research hotspots. The existing methods for detecting the temperature field distribution state in the furnace have the following problems:

[0004] 1. It is difficult to obtain accurate furnace temperature field through traditional contact temperature measurement method due to quantity and installation difficulty limitations;

[0005] 2. As non-contact measurement, acoustic and optical measurement methods are easily affected by factors such as the environment and equipment accuracy, resulting in low measurement accuracy and difficulty;

[0006] 3. The three-dimensional distribution of the temperature field can be accurately obtained through numerical simulation methods, but due to its long calculation cycle, it is difficult to apply it to obtain the three-dimensional distribution of the boiler furnace temperature field in real time.

[0007] At present, there are few studies on boiler furnace temperature optimization, and few can cover the temperature of every point in the entire furnace. Therefore, it is of great significance to achieve real-time measurement of the temperature field distribution inside the boiler and adjust the combustion state by adjusting the controllable variables of the combustion process to maintain the stable and economical operation of coal-fired boilers. Summary of the invention

[0008] The purpose of the present invention is to solve the problem that the prior art cannot cover the temperature of each point in the entire furnace during the optimization of the temperature field distribution in the boiler furnace, and provides a boiler temperature field distribution optimization control method driven by multi-source data.

[0009] The present invention provides a boiler temperature field distribution optimization control method based on multi-source data drive, which comprises:

[0010] S1. Collect and obtain the operating process parameter set of m typical operating conditions of the boiler;

[0011] S2. According to the operating process parameter set, the combustion state in the boiler is numerically simulated m times to obtain the numerical simulation data of the furnace temperature field under typical operating conditions;

[0012] S3, multi-source data fusion is performed on the numerical simulation data of furnace temperature field under typical boiler operating conditions and the operating process parameter set to obtain a multi-source data fusion data set;

[0013] S4. Classify the data set after fusion of multi-source data according to boiler operation load to obtain a modeling data set;

[0014] S5. Based on the modeling data set, the KNN algorithm is used to build a soft measurement model for furnace temperature distribution to predict the furnace temperature distribution of boilers under different working conditions;

[0015] S6. Select controllable variables from the operating process parameters, optimize the controllable variables, and take the minimum deviation between the furnace temperature distribution and the ideal temperature field distribution as the optimization objective function;

[0016] S7. Taking the ideal temperature field distribution as the optimization target, the genetic algorithm MA that integrates the genetic algorithm GA and the hill climbing search HC is used to calculate the optimal solution set of the secondary air volume that meets the optimization target, and the boiler is regulated and burned according to the optimal solution set of the secondary air volume.

[0017] Preferably, the operating process parameter set D in S1 includes boiler operating load Load, total air volume T A , total coal quantity T C , primary air volume P AV , primary air temperature P AT , Secondary air volume S AV , secondary air temperature S AT and burnout air volume SOFA AV ;

[0018] The operating process parameter set D under the typical operating condition i of the boiler i Specifically:

[0019]

[0020] Wherein, i = 1, 2, ..., m represents the number of typical boiler operating conditions, and m represents the total number of typical boiler operating conditions;

[0021] [Load i T Ai T Ci P AVi P ATi S AVi S ATi SOFA AVi] characterizes the operating parameters of the typical boiler operating condition i, where Load i represents the boiler operating load of the typical boiler operating condition i, T Ai represents the total air volume of the boiler in typical operating condition i, T Ci represents the total coal quantity of the boiler in typical operating condition i, P AVi represents the primary air volume of the boiler under typical operating condition i, P ATi represents the primary air temperature of the boiler under typical operating condition i, S AVi represents the secondary air volume of the boiler under typical operating condition i, S ATi represents the secondary air temperature of the boiler under typical operating condition i, SOFA AVi It represents the burnout air volume of the boiler under typical operating condition i.

[0022] Preferably, the specific method of performing m numerical simulations on the combustion state in the boiler in S2 to obtain numerical simulation data of the furnace temperature field under typical operating conditions includes:

[0023] The numerical simulation data C of the furnace temperature field under typical operating conditions includes the X coordinate, Y coordinate, Z coordinate representing the internal position of the furnace and the furnace temperature data T corresponding to the coordinate point;

[0024] Numerical simulation data of furnace temperature field under typical operating condition i C i Specifically:

[0025]

[0026] Where j = 1, 2, ..., n represents the total number of data points obtained by numerical simulation under typical boiler operating condition i, X ij The X coordinate of the typical boiler operating condition i is represented by the furnace width, and Y ij The Y coordinate of the typical boiler operating condition i is represented by the furnace height, Z ij The Z coordinate of the typical boiler operating condition i is represented by the furnace depth, T ij The furnace temperature data corresponding to the coordinate point i of the typical operating condition of the boiler are shown.

[0027] Preferably, the specific method of obtaining the data set after fusion of multi-source data in S3 includes:

[0028] The numerical simulation data of furnace temperature field under typical boiler operating conditions C i With the running process parameter set D i Perform multi-source data fusion to obtain the multi-source data fusion dataset H i :

[0029]

[0030] Preferably, the specific method for obtaining the modeling data set in S4 includes:

[0031] S4-1. Based on the boiler operating load Load as the classification basis, the operating conditions include full load, 75% load and 50% load. i Classify data with similar loads into a class of data G i ;

[0032] S4-2. Randomly select one working condition data from each classified data as the model test data set, perform stratified sampling on the data of the remaining working conditions, and merge the stratified sampled data according to the parameter category to obtain the modeling data set;

[0033] S4-3, use the max-min normalization method to preprocess the data, specifically:

[0034]

[0035] Among them, x i represents the sample value of the parameter, x * represents the data value after normalization, x min Indicates the minimum value of the parameter sample point, x max Indicates the maximum value of the parameter sample points.

[0036] Preferably, the specific method of using the KNN algorithm to construct a furnace temperature distribution soft measurement model in S5 to predict the furnace temperature distribution of the boiler under different working conditions includes:

[0037] S5-1, construct characteristic variables, use the data set R as the model input characteristic variable, and use the furnace temperature data T as the model output value;

[0038] R={Load,T A ,T C ,P AV ,T AT ,S AV ,S AT ,SOFA AV ,X,Y,Z};

[0039] S5-2, use the cross-validation method to obtain the optimal k value to achieve the best sample regression accuracy, where k represents the number of nearby points to be searched;

[0040] S5-3, using Euclidean distance as the similarity judgment index, select samples close to the target to be predicted, and set R(T i ) is the characteristic variable of the point to be predicted, R(T J ) is the Jth historical feature variable, and the similarity between them is d(R(T i),R(T J )) is calculated as follows:

[0041]

[0042] Among them, J∈[1,N], N represents the number of historical data points, R f (T i ) represents R(T i ) The fth feature attribute in the feature variable set, R f (T J ) represents R(T J ) The fth characteristic attribute in the characteristic variable set, f = 1, 2, ..., F, F represents the total number of characteristic variables;

[0043] S5-4. For the neighboring samples of the prediction point, the weighted and average calculation method is used to obtain the temperature prediction result:

[0044]

[0045] Among them, f KNN represents the predicted value of the KNN model, P(R s ) represents the neighboring points of the sample to be predicted, s = 1, 2, ..., k;

[0046] S5-5. According to the temperature prediction result obtained in S5-4, the boiler furnace temperature distribution under different operating conditions is obtained.

[0047] Preferably, the specific method of selecting controllable variables from the operating process parameters in S6, optimizing the controllable variables, and taking the minimum deviation between the furnace temperature distribution and the ideal temperature field distribution as the optimization objective function specifically includes:

[0048] S6-1, the secondary air volume S AV As a controllable variable, the secondary air volume S AV Optimize and adjust the boiler combustion state to reduce the deviation between the furnace temperature distribution and the ideal temperature field distribution, so that the furnace reaches the ideal temperature distribution state. Uncontrollable variables remain unchanged as fixed values ​​during the optimization process.

[0049] S6-2. The minimum deviation between the furnace temperature distribution and the ideal temperature field distribution is taken as the optimization objective function. The optimization objective function is specifically:

[0050]

[0051] Among them, F1(u) represents the deviation of furnace temperature distribution from the ideal temperature field distribution, T p =[t p1 ,t p2 ,…,t pq,…] represents the predicted value of the temperature field distribution of the entire furnace at time p, TI p =[ti p1 ,ti p2 ,…,ti pq ,…] represents the ideal temperature field distribution at time p, q=1,…,2776652 represents the data volume of the temperature field distribution of the entire furnace, Y pre (p) represents the real-time prediction model of furnace temperature field distribution at time p, S AV (p) represents the controllable variable at time p, the lower limit of the value is 380.286, the upper limit of the value is 896.188, x, y, z, T A (p)…,SOFA AV (p) represents the state quantities at time p, S AV (p) and x, y, z, T A (p)…,SOFA AV (p) Input samples that make up the model.

[0052] Preferably, the specific method of taking the ideal temperature field distribution as the optimization target and using the genetic algorithm MA integrating the genetic algorithm GA and the hill climbing search HC to calculate and obtain the optimal solution set of the secondary air volume that meets the optimization target includes:

[0053] S7-1, randomly generate Z individuals as the initial population, and the individuals in each population are composed of the secondary air volume at time p;

[0054] S7-2, read the historical operation data at time p, the total air volume is composed of primary air volume and secondary air volume, update the total air volume when introducing the randomly generated individuals, construct the updated total air volume, other operation data, full furnace coordinate data and randomly generated secondary air volume into a new data set, input the new data set into the furnace temperature field distribution prediction model, and obtain the boiler furnace temperature field distribution at time p;

[0055] S7-3, mapping the constructed objective function to a fitness function, and calculating the fitness value;

[0056] S7-4, select excellent individuals from the population as parents, and obtain new offspring populations through selection, crossover and mutation;

[0057] S7-5. In each round of iteration, the global search strategy of the genetic algorithm GA is adopted to select the individual with the highest fitness in the new generation population as the initial climbing point, retain the part of the individual with the highest fitness, adjust the remaining parameters, and generate new individuals;

[0058] S7-6, determine whether the number of climbing times reaches the set number, if not, return to execute S7-5, if yes, execute S7-7;

[0059] S7-7, determine whether the individual reaches the maximum value of the fitness setting, otherwise return to execute S7-5, if yes, the ideal furnace combustion temperature distribution is achieved, output the current optimal solution, and obtain the target secondary air volume.

[0060] Preferably, m is 54;

[0061] Said n is 2776652;

[0062] The k is 29.

[0063] Preferably, the value range of the boiler load Load is [175,350],

[0064] Total air volume T A The value range is [603,1304], the total coal quantity T C The value range is [90,211],

[0065] Primary air volume P AV The value range is [228,451], the primary air temperature P AT The value range is [315,425],

[0066] Secondary air volume S AV The value range is [332,896], the secondary air temperature S AT The value range is [320,366],

[0067] Burnout air volume SOFA AV The value range is [352,915];

[0068] The value range of the X coordinate is [-7.62, 7.62], the value range of the Y coordinate is [6.5, 64.9], and the value range of the Z coordinate is [-25.6, 7.62];

[0069] The value range of the furnace temperature data T corresponding to the coordinate point is [320,2372].

[0070] Advantages of the present invention: The boiler temperature field distribution optimization control method based on multi-source data driven proposed in the present invention can optimize the combustion temperature field distribution in the boiler furnace, thereby extending the service life of the furnace and preventing certain areas of the furnace from overheating or overcooling, leading to slag scaling, explosion and other dangers. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a flowchart of the boiler temperature field distribution optimization control method based on multi-source data drive according to the present invention;

[0072] Figure 2 This is a schematic diagram of the boiler temperature distribution prediction results based on the KNN prediction model;

[0073] Figure 3 It is the convergence curve solved by MA algorithm under full load (100%);

[0074] Figure 4 The convergence curve is solved by MA algorithm under 75% load;

[0075] Figure 5 It is the convergence curve solved by MA algorithm under 50% load. DETAILED DESCRIPTION

[0076] 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.

[0077] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0078] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0079] Embodiment 1:

[0080] Combine the following Figure 1 This embodiment describes a boiler temperature field distribution optimization control method based on multi-source data drive, which includes:

[0081] S1. Collect and obtain the operating process parameter set of m typical operating conditions of the boiler;

[0082] S2. According to the operating process parameter set, the combustion state in the boiler is numerically simulated m times to obtain the numerical simulation data of the furnace temperature field under typical operating conditions;

[0083] S3, multi-source data fusion is performed on the numerical simulation data of furnace temperature field under typical boiler operating conditions and the operating process parameter set to obtain a multi-source data fusion data set;

[0084] S4. Classify the data set after fusion of multi-source data according to boiler operation load to obtain a modeling data set;

[0085] S5. Based on the modeling data set, the KNN algorithm is used to build a soft measurement model for furnace temperature distribution to predict the furnace temperature distribution of boilers under different working conditions;

[0086] S6. Select controllable variables from the operating process parameters, optimize the controllable variables, and take the minimum deviation between the furnace temperature distribution and the ideal temperature field distribution as the optimization objective function;

[0087] S7. Taking the ideal temperature field distribution as the optimization target, the genetic algorithm MA that integrates the genetic algorithm GA and the hill climbing search HC is used to calculate the optimal solution set of the secondary air volume that meets the optimization target, and the boiler operation is controlled according to the optimal solution set of the secondary air volume.

[0088] Furthermore, the operating process parameter set D in S1 includes boiler operating load Load, total air volume T A , total coal quantity T C , primary air volume P AV , primary air temperature P AT , Secondary air volume S AV , secondary air temperature S AT and burnout air volume SOFA AV ;

[0089] The operating process parameter set D under the typical operating condition i of the boiler i Specifically:

[0090]

[0091] Wherein, i = 1, 2, ..., m represents the number of typical boiler operating conditions, and m represents the total number of typical boiler operating conditions;

[0092] [Load i T Ai T Ci P AVi P ATi S AVi S ATi SOFA AVi ] characterizes the operating parameters of the typical boiler operating condition i, where Load i represents the boiler operating load of the typical boiler operating condition i, T Ai represents the total air volume of the boiler in typical operating condition i, T Ci represents the total coal quantity of the boiler in typical operating condition i, P AVi represents the primary air volume of the boiler under typical operating condition i, P ATi represents the primary air temperature of the boiler under typical operating condition i, S AVi represents the secondary air volume of the boiler under typical operating condition i, S ATi represents the secondary air temperature of the boiler under typical operating condition i, SOFA AVi It represents the burnout air volume of the boiler under typical operating condition i.

[0093] Furthermore, the specific method of performing m numerical simulations on the combustion state in the boiler described in S2 to obtain numerical simulation data of the furnace temperature field under typical operating conditions includes:

[0094] The numerical simulation data C of the furnace temperature field under typical operating conditions includes the X coordinate, Y coordinate, Z coordinate representing the internal position of the furnace and the furnace temperature data T corresponding to the coordinate point;

[0095] Numerical simulation data of furnace temperature field under typical operating condition i C i Specifically:

[0096]

[0097] Where j = 1, 2, ..., n represents the total number of data points obtained by numerical simulation under typical boiler operating condition i, X ij The X coordinate of the typical boiler operating condition i is represented by the furnace width, and Y ij The Y coordinate of the typical boiler operating condition i is represented by the furnace height, Z ij The Z coordinate of the typical boiler operating condition i is represented by the furnace depth, T ij The furnace temperature data corresponding to the coordinate point i of the typical operating condition of the boiler are shown.

[0098] Furthermore, the specific method of obtaining the data set after fusion of multi-source data in S3 includes:

[0099] The numerical simulation data of furnace temperature field under typical boiler operating conditions C i With the running process parameter set D i Perform multi-source data fusion to obtain the multi-source data fusion dataset H i :

[0100]

[0101] Furthermore, the specific method of obtaining the modeling data set in S4 includes:

[0102] S4-1. Based on the boiler operating load Load as the classification basis, the operating conditions include full load, 75% load and 50% load. i Classify data with similar loads into a class of data G i ;

[0103] S4-2. Randomly select one working condition data from each classified data as the model test data set, perform stratified sampling on the data of the remaining working conditions, and merge the stratified sampled data according to the parameter category to obtain the modeling data set;

[0104] S4-3, use the max-min normalization method to preprocess the data, specifically:

[0105]

[0106] Among them, x i represents the sample value of the parameter, x * represents the data value after normalization, x min Indicates the minimum value of the parameter sample point, x max Indicates the maximum value of the parameter sample points.

[0107] Furthermore, the specific method of using the KNN algorithm to construct a furnace temperature distribution soft measurement model in S5 to predict the furnace temperature distribution of boilers under different working conditions includes:

[0108] S5-1, construct characteristic variables, use the data set R as the model input characteristic variable, and use the furnace temperature data T as the model output value;

[0109] R={Load,T A ,T C ,P AV ,T AT ,S AV ,S AT ,SOFA AV ,X,Y,Z};

[0110] S5-2, use the cross-validation method to obtain the optimal k value to achieve the best sample regression accuracy, where k represents the number of nearby points to be searched;

[0111] S5-3, using Euclidean distance as the similarity judgment index, select samples close to the target to be predicted, and set R(T i ) is the characteristic variable of the point to be predicted, R(T J ) is the Jth historical feature variable, and the similarity between them is d(R(T i ),R(T J )) is calculated as follows:

[0112]

[0113] Among them, J∈[1,N], N represents the number of historical data points, R f (T i ) represents R(T i ) The fth feature attribute in the feature variable set, R f (T J ) represents R(T J ) The fth characteristic attribute in the characteristic variable set, f = 1, 2, ..., F, F represents the total number of characteristic variables;

[0114] S5-4. For the neighboring samples of the prediction point, the weighted and average calculation method is used to obtain the temperature prediction result:

[0115]

[0116] Among them, f KNN represents the predicted value of the KNN model, P(R s ) represents the neighboring points of the sample to be predicted, s = 1, 2, ..., k;

[0117] S5-5. According to the temperature prediction result obtained in S5-4, the boiler furnace temperature distribution under different operating conditions is obtained.

[0118] Furthermore, the specific method of selecting controllable variables from the operating process parameters in S6, optimizing the controllable variables, and taking the minimum deviation between the furnace temperature distribution and the ideal temperature field distribution as the optimization objective function specifically includes:

[0119] S6-1, the secondary air volume S AV As a controllable variable, the secondary air volume S AV Optimize and adjust the boiler combustion state to reduce the deviation between the furnace temperature distribution and the ideal temperature field distribution, so that the furnace reaches the ideal temperature distribution state. Uncontrollable variables remain unchanged as fixed values ​​during the optimization process.

[0120] S6-2. The minimum deviation between the furnace temperature distribution and the ideal temperature field distribution is taken as the optimization objective function. The optimization objective function is specifically:

[0121]

[0122] Among them, F1(u) represents the deviation of furnace temperature distribution from the ideal temperature field distribution, T p =[t p1 ,t p2 ,…,t pq ,…] represents the predicted value of the temperature field distribution of the entire furnace at time p, TI p =[ti p1 ,ti p2 ,…,ti pq ,…] represents the ideal temperature field distribution at time p, q=1,…,2776652 represents the data volume of the temperature field distribution of the entire furnace, Y pre (p) represents the real-time prediction model of furnace temperature field distribution at time p, S AV (p) represents the controllable variable at time p, the lower limit of the value is 380.286, the upper limit of the value is 896.188, x, y, z, T A (p)…,SOFA AV(p) represents the state quantities at time p, S AV (p) and x, y, z, T A (p)…,SOFA AV (p) Input samples that make up the model.

[0123] Furthermore, the specific method of taking the ideal temperature field distribution as the optimization target and adopting the genetic algorithm MA integrating the genetic algorithm GA and the hill climbing search HC to calculate and obtain the optimal solution set of the secondary air volume that meets the optimization target includes:

[0124] S7-1, randomly generate Z individuals as the initial population, and the individuals in each population are composed of the secondary air volume at time p;

[0125] S7-2, read the historical operation data at time p, the total air volume is composed of primary air volume and secondary air volume, update the total air volume when introducing the randomly generated individuals, construct the updated total air volume, other operation data, full furnace coordinate data and randomly generated secondary air volume into a new data set, input the new data set into the furnace temperature field distribution prediction model, and obtain the boiler furnace temperature field distribution at time p;

[0126] S7-3, mapping the constructed objective function to a fitness function, and calculating the fitness value;

[0127] S7-4, select excellent individuals from the population as parents, and obtain new offspring populations through selection, crossover and mutation;

[0128] S7-5. In each round of iteration, the global search strategy of the genetic algorithm GA is adopted to select the individual with the highest fitness in the new generation population as the initial climbing point, retain the part of the individual with the highest fitness, adjust the remaining parameters, and generate new individuals;

[0129] S7-6, determine whether the number of climbing times reaches the set number, if not, return to execute S7-5, if yes, execute S7-7;

[0130] S7-7, determine whether the individual reaches the maximum value of the fitness setting, otherwise return to execute S7-5, if yes, the ideal furnace combustion temperature distribution is achieved, output the current optimal solution, and obtain the target secondary air volume.

[0131] Furthermore, the m is 54;

[0132] Said n is 2776652;

[0133] The k is 29.

[0134] Furthermore, the value range of the boiler load Load is [175,350],

[0135] Total air volume TA The value range is [603,1304], the total coal quantity T C The value range is [90,211],

[0136] Primary air volume P AV The value range is [228,451], the primary air temperature P AT The value range is [315,425],

[0137] Secondary air volume S AV The value range is [332,896], the secondary air temperature S AT The value range is [320,366],

[0138] Burnout air volume SOFA AV The value range is [352,915];

[0139] The value range of the X coordinate is [-7.62, 7.62], the value range of the Y coordinate is [6.5, 64.9], and the value range of the Z coordinate is [-25.6, 7.62];

[0140] The value range of the furnace temperature data T corresponding to the coordinate point is [320,2372].

[0141] In the present invention, the method for optimizing the control of boiler temperature field distribution based on multi-source data drive includes:

[0142] Step 1: Obtain an operating parameter set D that characterizes the typical operating conditions of the boiler. The operating process parameter set D includes boiler load Load, total air volume T A , total coal quantity T C , primary air volume P AV , Secondary air temperature T AT , Secondary air volume S AV , secondary air temperature S AT 、Burning air volume SOFA AV , the typical boiler operating conditions refer to some standard or common modes of boiler operation, which are usually the most frequent or most representative operating conditions of the boiler in actual operation;

[0143]

[0144] Among them, [Load i T Ai ...SOFA AVi ] represents the operating parameters of the typical operating condition i of the boiler, m is the total number of typical operating conditions of the boiler, the total number of typical operating conditions of the boiler m=54, and the operating parameter data information table of the typical operating condition is shown in Table 1:

[0145] Table 1

[0146] Serial number Run parameters symbol unit scope 1 Unit load Load MW [175,350] 2 Total air volume <![CDATA[T A ]]> t / h [603,1304] 3 Total coal <![CDATA[T C ]]> t / h [90,211] 4 Primary air volume <![CDATA[P AV ]]> t / h [228,451] 5 Primary air temperature <![CDATA[P AT ]]> t / h [315,425] 5 Secondary air volume <![CDATA[S AV ]]> t / h [332,896] 6 Secondary air temperature <![CDATA[S AT ]]> ℃ [320,366] 7 Burnout air volume <![CDATA[SOFA AV ]]> t / h [352,915]

[0147] Step 2: According to different operating parameters under typical working conditions of the boiler, the combustion state in the furnace is numerically simulated m times to obtain the numerical simulation data C of the furnace temperature field under typical working condition i, wherein the numerical simulation data C includes the X coordinate (furnace width), Y coordinate (furnace height), and Z coordinate (furnace depth) data representing the internal position of the furnace, as well as the furnace temperature data T corresponding to different coordinate points;

[0148]

[0149] in, is the X-coordinate value of the numerical simulation data obtained by simulating the typical working condition i, is the Y coordinate value of the numerical simulation data obtained by simulating the typical working condition i, is the Y coordinate value of the numerical simulation data obtained by simulating the typical working condition i, is the furnace temperature value T corresponding to different coordinate points obtained by simulating the typical working condition i, n is the total number of data points obtained by numerical simulation under the typical working condition i, and the total number of data points obtained by numerical simulation under a typical working condition in this embodiment is n=2776652. The numerical simulation data information table under the typical working condition is shown in Table 2:

[0150] Table 2

[0151] Serial number Numerical simulation data symbol unit scope 1 X coordinate X m [-7.62,7.62] 2 Y coordinate Y m [6.5,64.9] 3 Z coordinate Z m [-25.6,7.62] 4 temperature T K [320,2373]

[0152] Step 3: The numerical simulation data C under the typical working condition i i The operating parameter set D of the typical operating condition i i Perform multi-source data fusion and construct a data set H after multi-source data fusion i =(D i |C i );

[0153]

[0154] Step 4: Reconstruct the multi-source data according to the boiler operating load Load to obtain the data set H i Perform classification and construct a modeling data set, further including steps 4.1-4.3:

[0155] Step 4.1: Based on the different boiler operating loads, the operating conditions include full load (350MW), 75% load (260MW), and 50% load (175MW). 18 data sets H with similar operating loads are collected. i Divide into a class of data G i, a total of 3 categories;

[0156] Step 4.2: Randomly select one working condition from the 18 typical working conditions in each category as the model test data set; in order to reduce the computational cost and improve the model training speed, the remaining 17 typical working conditions are respectively subjected to stratified sampling. The data of each typical working condition are divided into 10 layers based on the Y coordinate of the furnace height. 5,000 groups of data are extracted from each layer, for a total of 50,000 groups. The data obtained after stratified sampling of the 17 typical working conditions are merged according to the parameter category to construct the modeling data set;

[0157] Step 4.3: Due to the dimensional differences of different parameters, the data needs to be preprocessed using the max-min normalization method. The specific formula is:

[0158]

[0159] Among them, x i is the sample value of a parameter, is the normalized value, x min 、x max are the minimum and maximum values ​​of the parameter sample points respectively;

[0160] Step 5: Based on the K nearest neighbor (KNN) algorithm, a furnace temperature distribution soft measurement model is constructed, and the parameters of the KNN model are trained using the training set. The trained model is used to predict the furnace temperature distribution of the boiler under different working conditions, further including steps 5.1 to 5.4:

[0161] Step 5.1: Construct feature variables and prepare a dataset containing input features and corresponding input values ​​to reconstruct the preprocessed multi-source dataset R = {Load, T A ,T C ,P AV ,T AT ,S AV ,S AT ,SOFA AV ,X,Y,Z} as the model input characteristic variables, and the furnace three-dimensional distribution temperature data T as the output value;

[0162] Step 5.2: Determine the optimal k value, and select the optimal k=29 value through the cross-validation method to achieve the optimal sample regression accuracy;

[0163] Step 5.3: Similarity measurement, using Euclidean distance as the similarity judgment index, select samples close to the target to be predicted, set R(T i ) is the characteristic variable of the point to be predicted, R(T j) is the jth historical feature variable, and the specific formula for calculating the similarity between the two is:

[0164]

[0165] Among them, R f (T i ), R f (T j ) are R(T) and R(T j ) The fth characteristic attribute in the characteristic variable set, F is the number of total characteristic variables, and n is the number of historical data points;

[0166] Step 5.4: Calculate the predicted value. Use the sum and average calculation to obtain the temperature prediction result for the 29 neighboring samples of the prediction point:

[0167]

[0168] Among them, f KNN is the predicted value of the KNN model, P(R i ) is the neighboring point of the sample to be predicted, and k is the number of nearby points to be searched;

[0169] Based on the training data set, a coal-fired boiler temperature field distribution prediction model driven by multi-source data is constructed. The temperature distribution prediction results are compared with the target values. Figure 2 As shown. In each category, according to the training data set, a rapid boiler furnace three-dimensional temperature distribution prediction model is established, and a total of 3 prediction models are obtained; for the actual data of typical working conditions collected in each category, the trained prediction model is used to predict the boiler furnace temperature distribution value in this category; the prediction performance of the boiler furnace temperature distribution prediction model in the three categories is shown in Table 3:

[0170] Table 3

[0171] Working condition category MAPE(%) MAE <![CDATA[R 2 ]]> 1 13.846 126.640 0.793 2 13.159 122.455 0.811 3 15.329 135.874 0.771

[0172] Step 6: The several characteristics that have the greatest impact on the boiler temperature described in step 5 can be divided into controllable variables and uncontrollable variables, and the optimization objective function is determined, further including steps 6.1-6.2:

[0173] Step 6.1: Select the secondary air volume S AV As a controllable variable in the optimization process, the secondary air volume is optimized to adjust the boiler combustion state and reduce the deviation between the temperature distribution and the ideal temperature distribution, so that the furnace can reach the ideal temperature distribution state. The uncontrollable variables are kept as fixed values ​​in the optimization process.

[0174] Step 6.2: Determine the optimization objective function to minimize the deviation between the furnace temperature distribution and the ideal temperature field distribution. The optimization objective function formula is as follows:

[0175]

[0176] st

[0177] T m =Y pre (S AV (m),x,y,z,T A (m)…,SOFA AV (m))

[0178] 380.286<S AV (m)<896.188

[0179] Among them, M=5 is set to optimize the number of time windows, T m =[t m1 ,t m2 ,...,t mn ] is the predicted value of the temperature field distribution of the entire furnace at time m, TI m =[ti m1 ,ti m2 ,...,ti mn ] is the ideal distribution of furnace temperature field, n = 2776652 is the data volume of the whole furnace temperature field distribution, Y pre (m) is the real-time prediction model of furnace temperature field distribution at time m, S AV (m) is the controllable variable (secondary air volume) at time m, x, y, z, T A (m)…,SOFA AV (m) is the state variable required for modeling, S AV (m) and x, y, z, T A (m)…,SOFA AV (m) Input samples that make up the model, S AV _down(m)=380.286、S AV_up (m) = 896.188 are the lower and upper limits of the controllable variable u(m);

[0180] Step 7: Taking the ideal temperature field distribution of the boiler as the optimization target, combined with the furnace temperature field distribution prediction model, the memetic algorithm (MA) integrating genetic algorithms (GA) and hill climbing (HC) is used to optimize the controllable variables input by the boiler to obtain the optimal set of controllable variables, further including steps 7.1 to 7.7:

[0181] Step 7.1: Randomly generate Z individuals as the initial population, where each individual consists of 5 moments of secondary air volume;

[0182] Step 7.2: Read the historical operation data of 5 moments in the DCS system. Since the total air volume is composed of primary air volume and secondary air volume, the total air volume is updated when the randomly generated individuals are introduced. The updated total air volume and other operation data are combined with the full furnace coordinate data and the randomly generated secondary air volume to form a new data set and input it into the boiler furnace temperature field distribution prediction model to obtain the boiler furnace temperature field distribution at 5 moments.

[0183] Step 7.3: Map the constructed objective function to the fitness function and calculate the fitness value;

[0184] Step 7.4: Select some excellent individuals from the current population as parents, and obtain a new offspring population through selection (roulette selection), crossover (single-point crossover), and mutation (probabilistic mutation);

[0185] Step 7.5: In each iteration, the individual with the highest fitness is selected from the new generation population obtained by the global search strategy of the genetic algorithm (GA) as the initial climbing point, some parameters of the optimal individual are retained, and the remaining parameters are appropriately adjusted to generate new individuals;

[0186] Step 7.6: If the maximum number of hill climbing times is reached, go to step 7.4, otherwise go to step 7.5 to continue the hill climbing operation;

[0187] Step 7.7: Determine whether the current fitness value is good enough, that is, the ideal furnace combustion temperature distribution is achieved or the maximum number of iterations is reached. If satisfied, the program terminates and outputs the current optimal solution to obtain the target secondary air volume.

[0188] The MA algorithm that integrates genetic algorithm and hill climbing search is used to solve the optimization function and obtain the optimal solution set of secondary wind that meets the optimization goal. Under different loads, the MA algorithm is used to solve the convergence curve as shown in the figure. Figure 3-Figure 5 As shown, Figure 3 The convergence curve is obtained by using the MA algorithm under full load (100%). Figure 4 The convergence curve is obtained by using the MA algorithm under 75% load. Figure 5 The convergence curve is solved by MA algorithm under 50% load. The average deviation of secondary air volume and temperature field distribution before and after optimization is shown in Table 4:

[0189] Table 4

[0190]

[0191] Although the present invention is described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. It should therefore be understood that many modifications may be made to the exemplary embodiments and that other arrangements may be devised without departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the various dependent claims and features described herein may be combined in a manner different from that described in the original claims. It should also be understood that features described in conjunction with individual embodiments may be used in other described embodiments.

Claims

1. A boiler temperature field distribution optimization control method based on multi-source data drive, characterized in that: It includes: S1. Collect and obtain the operating process parameter set of m typical operating conditions of the boiler; S2. According to the operating process parameter set, the combustion state in the boiler is numerically simulated m times to obtain the numerical simulation data of the furnace temperature field under typical operating conditions; S3, multi-source data fusion is performed on the numerical simulation data of furnace temperature field under typical boiler operating conditions and the operating process parameter set to obtain a multi-source data fusion data set; S4. Classify the data set after fusion of multi-source data according to boiler operation load to obtain a modeling data set; S5. Based on the modeling data set, the KNN algorithm is used to build a soft measurement model for furnace temperature distribution to predict the furnace temperature distribution of boilers under different working conditions; S6. Select controllable variables from the operating process parameters, optimize the controllable variables, and take the minimum deviation between the furnace temperature distribution and the ideal temperature field distribution as the optimization objective function; S7. Taking the ideal temperature field distribution as the optimization target, the genetic algorithm MA that integrates the genetic algorithm GA and the hill climbing search HC is used to calculate the optimal solution set of the secondary air volume that meets the optimization target, and the boiler is regulated and burned according to the optimal solution set of the secondary air volume.

2. The method for optimizing the distribution of boiler temperature field based on multi-source data drive according to claim 1 is characterized in that: S1 The operating process parameter set D includes boiler operating load Load, total air volume T A , total coal quantity T C , primary air volume P AV , primary air temperature P AT , Secondary air volume S AV , secondary air temperature S AT and burnout air volume SOFA AV ; The operating process parameter set D under the typical operating condition i of the boiler i Specifically: Wherein, i = 1, 2, ..., m represents the number of typical boiler operating conditions, and m represents the total number of typical boiler operating conditions; [Load i T Ai T Ci P AVi P ATi S AVi S ATi SOFA AVi ] characterizes the operating parameters of the typical boiler operating condition i, Among them, Load i represents the boiler operating load of the typical boiler operating condition i, T Ai represents the total air volume of the boiler in typical operating condition i, T Ci represents the total coal quantity of the boiler in typical operating condition i, P AVi represents the primary air volume of the boiler under typical operating condition i, P ATi represents the primary air temperature of the boiler under typical operating condition i, S AVi represents the secondary air volume of the boiler under typical operating condition i, S ATi represents the secondary air temperature of the boiler under typical operating condition i, SOFA AVi It represents the burnout air volume of the boiler under typical operating condition i.

3. The method for optimizing the distribution of boiler temperature field based on multi-source data drive according to claim 2 is characterized in that: The specific method of performing m numerical simulations of the combustion state in the boiler and obtaining the numerical simulation data of the furnace temperature field under typical operating conditions as described in S2 includes: The numerical simulation data C of the furnace temperature field under typical operating conditions includes the X coordinate, Y coordinate, Z coordinate representing the internal position of the furnace and the furnace temperature data T corresponding to the coordinate point; Numerical simulation data of furnace temperature field under typical operating condition i C i Specifically: Where j = 1, 2, ..., n represents the total number of data points obtained by numerical simulation under typical boiler operating condition i, X ij The X coordinate of the typical boiler operating condition i is represented by the furnace width, and Y ij The Y coordinate of the typical boiler operating condition i is represented by the furnace height, Z ij The Z coordinate of the typical boiler operating condition i is represented by the furnace depth. T ij The furnace temperature data corresponding to the coordinate point i of the typical operating condition of the boiler are shown.

4. The method for optimizing the distribution of boiler temperature field based on multi-source data drive according to claim 3 is characterized in that: The specific method of obtaining the data set after fusion of multi-source data in S3 includes: The numerical simulation data of furnace temperature field under typical boiler operating conditions C i With the running process parameter set D i Perform multi-source data fusion to obtain the multi-source data fusion dataset H i :

5. The method for optimizing the distribution of boiler temperature field based on multi-source data drive according to claim 4 is characterized in that: The specific method for obtaining the modeling data set described in S4 includes: S4-1. The boiler operating load is used as the classification basis. The operating conditions include full load, 75% load and 50% load. The dataset H after multi-source data fusion i Classify data with similar loads into a class of data G i ; S4-2. Randomly select one working condition data from each classified data as the model test data set, perform stratified sampling on the data of the remaining working conditions, and merge the stratified sampled data according to the parameter category to obtain the modeling data set; S4-3, use the max-min normalization method to preprocess the data, specifically: Among them, x i represents the sample value of the parameter, x * represents the data value after normalization, x min Indicates the minimum value of the parameter sample point, x max Indicates the maximum value of the parameter sample points.

6. The method for optimizing the distribution of boiler temperature field based on multi-source data drive according to claim 5 is characterized in that: S5 uses the KNN algorithm to construct a furnace temperature distribution soft measurement model, and the specific method for predicting the furnace temperature distribution of boilers under different working conditions includes: S5-1, construct characteristic variables, use the data set R as the model input characteristic variable, and use the furnace temperature data T as the model output value; R={Load,T A ,T C ,P AV ,T AT ,S AV ,S AT ,SOFA AV ,X,Y,Z}; S5-2, use the cross-validation method to obtain the optimal k value to achieve the best sample regression accuracy, where k represents the number of nearby points to be searched; S5-3, using Euclidean distance as the similarity judgment index, select samples close to the target to be predicted, and set R(T i ) is the characteristic variable of the point to be predicted, R(T J ) is the Jth historical feature variable, and the similarity between the two is d(R(T i ),R(T J )) is calculated as follows: Among them, J∈[1,N], N represents the number of historical data points, R f (T i ) represents R(T i ) The fth feature attribute in the feature variable set, R f (T J ) represents R(T J ) The fth characteristic attribute in the characteristic variable set, f = 1, 2, ..., F, F represents the total number of characteristic variables; S5-4. For the neighboring samples of the prediction point, the weighted and average calculation method is used to obtain the temperature prediction result: Among them, f KNN represents the predicted value of the KNN model, P(R s ) represents the neighboring points of the sample to be predicted, s = 1, 2, ..., k; S5-5. According to the temperature prediction result obtained in S5-4, the boiler furnace temperature distribution under different operating conditions is obtained.

7. The method for optimizing the distribution of boiler temperature field based on multi-source data drive according to claim 6 is characterized in that: The specific method of selecting controllable variables from the operating process parameters, optimizing the controllable variables, and taking the minimum deviation between the furnace temperature distribution and the ideal temperature field distribution as the optimization objective function as described in S6 specifically includes: S6-1, the secondary air volume S AV As a controllable variable, the secondary air volume S AV Optimize and adjust the boiler combustion state to reduce the deviation between the furnace temperature distribution and the ideal temperature field distribution, so that the furnace reaches the ideal temperature distribution state. Uncontrollable variables remain unchanged as fixed values ​​during the optimization process. S6-2. The minimum deviation between the furnace temperature distribution and the ideal temperature field distribution is taken as the optimization objective function. The optimization objective function is specifically: Among them, F1(u) represents the deviation of furnace temperature distribution from the ideal temperature field distribution, T p =[t p1 ,t p2 ,…,t pq ,…] represents the predicted value of the temperature field distribution of the entire furnace at time p, TI p =[ti p1 ,ti p2 ,…,ti pq ,…] represents the ideal temperature field distribution at time p, q=1,…,2776652 represents the data volume of the temperature field distribution of the entire furnace. Y pre (p) represents the real-time prediction model of furnace temperature field distribution at time p. S AV (p) represents the controllable variable at time p, the lower limit of the value is 380.286, and the upper limit of the value is 896.

188. x,y,z,T A (p)…,SOFA AV (p) represents the state quantities at time p, S AV (p) and x, y, z, T A (p)…,SOFA AV (p) Input samples that make up the model.

8. The method for optimizing the distribution of boiler temperature field based on multi-source data drive according to claim 7 is characterized in that: The specific method described in S7, taking the ideal temperature field distribution as the optimization target, using the genetic algorithm GA and the cultural gene algorithm MA that integrates the hill climbing search HC, and calculating the optimal solution set of the secondary air volume that meets the optimization target includes: S7-1, randomly generate Z individuals as the initial population, and the individuals in each population are composed of the secondary air volume at time p; S7-2, read the historical operation data at time p, the total air volume is composed of primary air volume and secondary air volume, update the total air volume when introducing the randomly generated individuals, construct the updated total air volume, other operation data, full furnace coordinate data and randomly generated secondary air volume into a new data set, input the new data set into the furnace temperature field distribution prediction model, and obtain the boiler furnace temperature field distribution at time p; S7-3, mapping the constructed objective function to a fitness function, and calculating the fitness value; S7-4, select excellent individuals from the population as parents, and obtain new offspring populations through selection, crossover and mutation; S7-5. In each round of iteration, the global search strategy of the genetic algorithm GA is adopted to select the individual with the highest fitness in the new generation population as the initial climbing point, retain the part of the individual with the highest fitness, adjust the remaining parameters, and generate new individuals; S7-6, determine whether the number of climbing times reaches the set number, if not, return to execute S7-5, if yes, execute S7-7; S7-7, determine whether the individual reaches the maximum value of the fitness setting, otherwise return to execute S7-5, if yes, the ideal furnace combustion temperature distribution is achieved, output the current optimal solution, and obtain the target secondary air volume.

9. The method for optimizing the distribution of boiler temperature field based on multi-source data drive according to claim 8, characterized in that: The m is 54; Said n is 2776652; The k is 29.

10. A boiler temperature field distribution optimization control method based on multi-source data drive according to claim 8 or 9, characterized in that: The value range of the boiler load Load is [175,350], Total air volume T A The value range is [603,1304], the total coal quantity T C The value range is [90,211], Primary air volume P AV The value range is [228,451], the primary air temperature P AT The value range is [315,425], Secondary air volume S AV The value range is [332,896], the secondary air temperature S AT The value range is [320,366], Burnout air volume SOFA AV The value range is [352,915]; The value range of the X coordinate is [-7.62, 7.62], the value range of the Y coordinate is [6.5, 64.9], and the value range of the Z coordinate is [-25.6, 7.62]; The value range of the furnace temperature data T corresponding to the coordinate point is [320,2372].

Images