Thermal power boiler combustion efficiency optimization method based on ball tree nearest neighbor search

Abstract

The invention discloses a thermal power boiler combustion efficiency optimization method based on ball tree nearest neighbor search, and the method comprises the steps: screening out key parameters affecting the combustion efficiency through correlation analysis, and then finding a historical optimal operation parameter in a multi-dimensional parameter space through a ball tree nearest neighbor search algorithm, thereby achieving the optimization of the combustion efficiency. According to the method, the historical optimal parameter combination can be rapidly positioned, the method adapts to rapid conversion of combustion working conditions, and the method has high adaptability and is suitable for boiler efficiency optimization of a thermal power plant.

Description

Technical Field

[0001] The invention relates to the field of boiler combustion efficiency optimization, in particular to a method for optimizing the combustion efficiency of a thermal power boiler based on a ball tree nearest neighbor search. Background Art

[0002] In the process of thermal power generation, boiler combustion efficiency is a key factor affecting energy utilization. Improving boiler combustion efficiency can significantly improve the economic benefits of power plants, so boiler combustion efficiency optimization is of great significance. However, affected by many factors, such as fuel type, combustion conditions, air supply, load fluctuations, etc., the boiler combustion process is complex and combustion efficiency optimization is very challenging.

[0003] Traditional boiler combustion efficiency optimization methods are mainly based on mechanism models, and in actual operation they rely heavily on the operator's experience. However, the mechanism model-based method is difficult to adapt to the dynamic changes in the combustion process, resulting in poor optimization results. Summary of the invention

[0004] The purpose of the present invention is to solve the real-time optimization problem of boiler efficiency in thermal power plants and to provide a method for optimizing the combustion efficiency of thermal power boilers based on ball tree nearest neighbor search. The method proposed in the present invention can quickly locate the historical optimal parameter combination, adapt to the rapid change of combustion conditions, has strong adaptability, and is suitable for optimizing the boiler efficiency of thermal power plants.

[0005] The specific technical solutions provided by the present invention are as follows:

[0006] A method for optimizing combustion efficiency of thermal power boilers based on ball tree nearest neighbor search comprises the following steps:

[0007] S1. Collect relevant characteristic parameter data of the boiler system from the DCS system of the power plant and establish a historical case database;

[0008] S2. Using the positive balance optimization method, taking the boiler outlet temperature as the target parameter for optimization, and using correlation analysis to select the operating parameters and operation parameters that are closely related to the boiler outlet temperature;

[0009] S3. Extract data from the historical case database to build a boiler combustion efficiency optimization case database based on the selected operating parameters, operation parameters and optimization target parameters;

[0010] S4. Normalize the data in the boiler combustion efficiency optimization case database and use the principal component analysis method to reduce the dimension of the data;

[0011] S5. Use the reduced-dimensional data to train the ball tree model. In the online monitoring phase, use the ball tree nearest neighbor search algorithm to retrieve the nearest n case data in the boiler combustion efficiency optimization case database according to the real-time operating parameters.

[0012] S6. Filter the data whose boiler outlet temperature is greater than the real-time boiler outlet temperature from the n case data, and perform weighted summation based on the distance score and the optimization score to obtain the optimal boiler combustion efficiency operating parameters.

[0013] Furthermore, in step S1, historical operation data of the boiler system is collected from the DCS, and the collected data is preliminarily cleaned to remove noise, outliers and invalid data, and a historical case database is constructed.

[0014] Furthermore, in step S2, a positive balance optimization method is adopted, the boiler outlet temperature is taken as the main optimization target, and the correlation coefficient between each operating parameter and the boiler outlet temperature is calculated using a correlation analysis method, from which parameters highly correlated with the boiler outlet temperature are selected.

[0015] Furthermore, in step S4, the data in the boiler combustion efficiency optimization case database is subjected to Min-Max normalization processing to eliminate the influence between different parameter dimensions.

[0016] Furthermore, in step S5, the ball tree model is trained using the reduced-dimensional data. Specifically, the process is divided into the following steps:

[0017] S51. Starting from the entire data set, find the centroid of the data set, and then determine the distance from the centroid to the farthest point in the data set as the radius of the sphere to form an initial sphere that contains all data points. The centroid calculation formula is as follows:

[0018]

[0019] Among them, x i is a point inside the sphere, and n is the number of points inside the sphere. The formula for calculating the radius of the sphere is as follows:

[0020]

[0021] Among them, ||x i -c|| means x i Euclidean distance to c;

[0022] S52. For each sphere, if the number of data points exceeds a certain threshold or reaches the maximum depth of the tree, the sphere is divided into two sub-spheres: the two farthest points are selected as splitting points; the data points are assigned to the spheres closer to the splitting points, thereby forming two new sub-spheres, each of which forms a new node; the above process is repeated for each sub-sphere until the number of data points in the sphere is less than the threshold or reaches the maximum depth of the tree;

[0023] S53, in the online monitoring stage, the operating parameters of the boiler are collected in real time, and these real-time parameters are input into the ball tree model, and the n case data most similar to the current operating conditions are retrieved from the boiler combustion efficiency optimization case database through the ball tree nearest neighbor search algorithm;

[0024] Furthermore, in S6, case data whose boiler outlet temperature is greater than the current real-time boiler outlet temperature are screened out from the retrieved n case data, and then a weighted sum calculation is performed based on the distance score and optimization score of each case data; the distance score is used to evaluate the similarity between the case data and the current operating conditions, and the optimization score measures the optimization effect of each case data; finally, based on the result of the weighted summation, the optimal boiler combustion efficiency operating parameters are selected to guide the actual operation adjustment and improve the boiler combustion efficiency and operation economy; the specific calculation formula is as follows:

[0025] score=d+α*(-target)

[0026] Among them, d is the distance between the real-time sample and the nearest neighbor sample, and target is the difference between the nearest neighbor sample and the real-time sample optimization target variable.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention proposes a method for optimizing the combustion efficiency of thermal power boilers based on correlation analysis and a ball tree nearest neighbor search algorithm, which can significantly improve the combustion efficiency of boilers and adapt to variable operating conditions. Compared with the traditional optimization method based on the mechanism model, the present invention selects key operating parameters through correlation analysis, and uses the ball tree nearest neighbor search algorithm to quickly locate the historically optimal parameter combination, thereby ensuring the efficiency and accuracy of the optimization process. The present invention reduces the dependence on the operator's experience, can also quickly respond to changes in combustion conditions, maintain an efficient combustion state, has strong adaptability, wide versatility, and has broad industrial application potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A flow chart of the steps of the present invention

[0030] Figure 2 Histogram for correlation analysis

[0031] Figure 3 To optimize the case effect diagram DETAILED DESCRIPTION

[0032] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0033] like Figure 1As shown, the present invention proposes a method for optimizing the combustion efficiency of a thermal power boiler based on a ball tree nearest neighbor search. The specific implementation steps in the embodiment are as follows:

[0034] S1. Collect relevant characteristic parameter data of the boiler system from the DCS system of the power plant, such as secondary hot air duct air flow, primary cooling water flow, low pre-pass flue gas oxygen content (left), low pre-pass flue gas oxygen content (right), boiler bed temperature 2, economizer inlet temperature, economizer outlet temperature, upper economizer inlet flue gas temperature (left), upper economizer inlet flue gas temperature (right), intermediate economizer inlet flue gas temperature (left), intermediate economizer inlet flue gas temperature (right), upper air preheater inlet flue gas temperature (left), upper air preheater inlet flue gas temperature (right), intermediate air preheater inlet flue gas temperature (left), intermediate air preheater inlet flue gas temperature (right), etc. For low pre-pass flue gas oxygen content (left), low pre-pass flue gas oxygen content (right) and other operating parameters that contain two left and right operating parameters, the left and right parameters are averaged, and only the low pre-pass flue gas oxygen content parameter is retained. Then clean the original data, remove missing data and abnormal data, and establish a historical case database;

[0035] S2. Use the positive balance optimization method, take the boiler outlet temperature as the target parameter for optimization, and use correlation analysis to select the operating parameters and operation parameters that are closely related to the boiler outlet temperature. The results of the correlation analysis are as follows: Figure 2 As shown, the top operating parameters are: boiler bed front temperature 2, high temperature superheater inlet flue gas temperature, upper economizer inlet flue gas temperature, gas header steam temperature, etc., and 22 parameters with correlation greater than 0.4 are selected for modeling;

[0036] S3. Extract data from the historical case database to build a boiler combustion efficiency optimization case database based on the selected operating parameters, operation parameters and optimization target parameters;

[0037] S4. Perform Min-Max normalization on the data in the boiler combustion efficiency optimization case database to eliminate the influence between different parameter dimensions. Use the principal component analysis method to reduce the dimension of the normalized data, retain the main components with a total contribution greater than 95%, and finally reduce the input parameters to 12 characteristic parameters;

[0038] S5. Use the reduced-dimensional data to train the ball tree model. During the online monitoring phase, use the ball tree nearest neighbor search algorithm to retrieve the 100 most recent case data in the boiler combustion efficiency optimization case database based on real-time operating parameters. The specific steps are as follows:

[0039] S51. Starting from the entire data set, find the centroid of the data set, and then determine the distance from the centroid to the farthest point in the data set as the radius of the sphere to form an initial sphere. This sphere contains all data points. The centroid calculation formula is as follows:

[0040]

[0041] Among them, x i is the point inside the sphere, and n is the number of points inside the sphere. The formula for calculating the radius of the sphere is as follows:

[0042]

[0043] Among them, ||x i -c|| means x i The Euclidean distance to c.

[0044] S52. For each sphere, if the number of data points exceeds 40, the sphere is divided into two sub-spheres: the two farthest points are selected as splitting points. The data points are assigned to the spheres of the splitting points closer to them, thereby forming two new sub-spheres, each of which forms a new node. The above process is repeated for each sub-sphere until the number of data points in the sphere is less than the threshold or the maximum depth of the tree is reached.

[0045] S53. During the online monitoring stage, the operating parameters of the boiler are collected in real time, and these real-time parameters are input into the ball tree model. Through the ball tree nearest neighbor search algorithm, the 100 case data most similar to the current operating conditions are retrieved from the boiler combustion efficiency optimization case database.

[0046] S6. Filter out the case data whose boiler outlet temperature is greater than the current real-time boiler outlet temperature from the 100 retrieved case data, and then perform weighted sum calculation based on the distance score and optimization score of each case data. The distance score is used to evaluate the similarity between the case data and the current operating conditions, and the optimization score measures the optimization effect of each case data. Finally, based on the result of the weighted summation, select the optimal boiler combustion efficiency operating parameters to guide actual operation adjustments and improve the boiler's combustion efficiency and operating economy. The specific calculation formula is as follows:

[0047] score=d+α*(-target)

[0048] Among them, d is the distance between the real-time sample and the nearest neighbor sample, and target is the difference between the nearest neighbor sample and the real-time sample optimization target variable. The final optimization result is as follows Figure 3 shown.

[0049] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for optimizing combustion efficiency of thermal power boilers based on ball tree nearest neighbor search, characterized in that: The steps include: S1. Collect relevant characteristic parameter data of the boiler system from the DCS system of the power plant and establish a historical case database; S2. Using the positive balance optimization method, taking the boiler outlet temperature as the target parameter for optimization, and using correlation analysis to select the operating parameters and operation parameters that are closely related to the boiler outlet temperature; S3. Extract data from the historical case database to build a boiler combustion efficiency optimization case database based on the selected operating parameters, operation parameters and optimization target parameters; S4. Normalize the data in the boiler combustion efficiency optimization case database and use the principal component analysis method to reduce the dimension of the data; S5. Use the reduced-dimensional data to train the ball tree model. In the online monitoring phase, use the ball tree nearest neighbor search algorithm to retrieve the nearest n case data in the boiler combustion efficiency optimization case database according to the real-time operating parameters. S6. Filter the data whose boiler outlet temperature is greater than the real-time boiler outlet temperature from the n case data, and perform weighted summation based on the distance score and the optimization score to obtain the optimal boiler combustion efficiency operating parameters.

2. The method for optimizing combustion efficiency of thermal power boilers based on ball tree nearest neighbor search according to claim 1, characterized in that: In step S1, historical operation data of the boiler system is collected from the DCS system, and the collected data is preliminarily cleaned to remove noise, outliers and invalid data, and a historical case database is constructed.

3. The method for optimizing combustion efficiency of thermal power boilers based on ball tree nearest neighbor search according to claim 1, characterized in that: In step S2, a positive balance optimization method is adopted, the boiler outlet temperature is taken as the main optimization target, and the correlation coefficient between each operating parameter and the boiler outlet temperature is calculated using a correlation analysis method, from which parameters highly correlated with the boiler outlet temperature are selected.

4. The method for optimizing combustion efficiency of thermal power boilers based on ball tree nearest neighbor search according to claim 1, characterized in that: In step S4, the data in the boiler combustion efficiency optimization case database is subjected to Min-Max normalization processing to eliminate the influence between different parameter dimensions.

5. The method for optimizing combustion efficiency of thermal power boilers based on ball tree nearest neighbor search according to claim 1, characterized in that: In step S5, the ball tree model is trained using the reduced-dimensional data, which is specifically divided into the following steps: S51. Starting from the entire data set, find the centroid of the data set, and then determine the distance from the centroid to the farthest point in the data set as the radius of the sphere to form an initial sphere that contains all data points. The centroid calculation formula is as follows: Among them, x i is a point inside the sphere, n is the number of points inside the sphere, and the radius of the sphere is calculated as follows: Among them, ||x i -c|| means x i Euclidean distance to c; S52. For each sphere, if the number of data points exceeds a certain threshold or reaches the maximum depth of the tree, the sphere is divided into two sub-spheres: the two farthest points are selected as splitting points; the data points are assigned to the spheres closer to the splitting points, thereby forming two new sub-spheres, each of which forms a new node; the above process is repeated for each sub-sphere until the number of data points in the sphere is less than the threshold or reaches the maximum depth of the tree; S53. During the online monitoring stage, the operating parameters of the boiler are collected in real time, and these real-time parameters are input into the ball tree model. Through the ball tree nearest neighbor search algorithm, the n case data most similar to the current operating conditions are retrieved from the boiler combustion efficiency optimization case database.

6. The method for optimizing combustion efficiency of thermal power boilers based on ball tree nearest neighbor search according to claim 1, characterized in that: In step S6, case data whose boiler outlet temperature is greater than the current real-time boiler outlet temperature are screened out from the retrieved n case data, and then a weighted sum calculation is performed based on the distance score and optimization score of each case data; the distance score is used to evaluate the similarity between the case data and the current working conditions, and the optimization score measures the optimization effect of each case data; finally, based on the result of the weighted summation, the optimal boiler combustion efficiency operating parameters are selected to guide the actual operation adjustment and improve the boiler combustion efficiency and operation economy; the specific calculation formula is as follows: score=d+α*(-target) Among them, d is the distance between the real-time sample and the nearest neighbor sample, and target is the difference between the nearest neighbor sample and the real-time sample optimization target variable.

Images