Intelligent combustion adjustment method and system based on power station boiler working condition identification and self-adaptive combustion characteristic model coupling

Abstract

The invention discloses an intelligent combustion adjustment method and system based on power station boiler working condition identification and self-adaptive combustion characteristic model coupling. The intelligent combustion adjustment method comprises the steps that data collection and quality judgment are conducted; multivariable historical data representing boiler combustion characteristics are extracted, and the multivariable historical data are divided into different operation condition combinations with large characteristic differences by adopting a classification algorithm based on the historical data; based on the working condition classification result, modeling is carried out on each working condition group of boiler operation; determining corresponding evaluation indexes for different working condition groups, and establishing a relationship between the evaluation indexes and the sub-working condition prediction model; after the model is put into operation, calculating the precision of the different-working-condition prediction model in real time, and if the precision does not meet a preset condition, triggering automatic remodeling to ensure that the prediction model is matched with the current characteristics of the boiler; aiming at the target function, a multi-target optimization algorithm is adopted to calculate the coal feeding amount of each mill, the primary air amount of each mill, the total oxygen amount and a secondary air door opening degree recommendation value in real time; and instruction output is optimized, so that the boiler efficiency is improved, and pollutant generation is reduced.

Description

Technical Field

[0001] The present invention relates to an intelligent combustion adjustment method and system, and in particular to an intelligent combustion adjustment method and system based on coupling of power station boiler operating condition identification and an adaptive combustion characteristic model. Background Art

[0002] Renewable energy generation is characterized by randomness, intermittency, and volatility, which leads to frequent and large load changes in thermal power units. Power station boilers are faced with problems such as rapid load changes, deep peak regulation, and variable coal quality. Since boiler combustion is a strongly coupled, nonlinear multi-input and multi-output process, operators generally fail to adjust combustion in a timely manner after rapid load changes and coal quality changes in the boiler, resulting in poor boiler operation economy and high pollutant emissions. Boiler combustion adjustment faces great challenges.

[0003] Existing boiler combustion adjustments generally use the best operating parameters under a given load and given coal quality determined by on-site adjustment tests. However, boiler loads and coal types are currently variable, and the combustion adjustment test results are difficult to guide the operation of the boiler under all operating conditions. Summary of the invention

[0004] Purpose of the invention: The purpose of the present invention is to provide an intelligent combustion adjustment method and system based on the coupling of power plant boiler operating condition identification and adaptive combustion characteristic model, to assist boiler operators to quickly make combustion adjustments, improve boiler efficiency and reduce nitrogen oxide emissions.

[0005] Technical solution: The present invention comprises the following steps:

[0006] Data collection and quality judgment: collect data and make quality judgments on the collected variable data;

[0007] Operating condition classification: extract multivariate historical data that characterizes boiler combustion characteristics, and use classification algorithms based on the historical data to classify them into different operating condition combinations with greatly different characteristics;

[0008] Modeling by operating condition: Based on the operating condition classification results, model each operating condition group of boiler operation separately;

[0009] Setting of objective function for different working conditions: Determine the corresponding evaluation index for different working condition groups, and establish the relationship between the evaluation index and the prediction model for different working conditions;

[0010] Online rolling model evaluation and online model correction: After the model is put into operation, the accuracy of the prediction model for each operating condition is calculated in real time. If the accuracy does not meet the pre-set conditions, automatic re-modeling is triggered to ensure that the prediction model matches the current characteristics of the boiler;

[0011] Optimizing the objective function: A multi-objective optimization algorithm is used to calculate the coal feed rate, primary air volume, total oxygen content, and recommended values ​​of secondary air door opening of each mill in real time based on the objective function;

[0012] Optimize command output: Automatically output command value based on the deviation between the recommended values ​​of coal feed rate for each mill, primary air volume for each mill, total oxygen content, and secondary air door opening obtained by optimizing the objective function and the current real-time value. If the deviation is within the adjustment dead zone, the current command value remains unchanged; if the deviation exceeds the adjustment dead zone, the instantaneous command value is set according to the recommended value of the objective function.

[0013] The data collection is to collect real-time data from DCS or SIS to obtain load, economizer CO concentration, air preheater outlet flue gas temperature, SCR inlet NOx concentration, coal feed to each mill, primary air volume of each mill, total air volume, and real-time value of secondary air door opening. At the same time, the elemental analysis and calorific value of the coal entering the furnace are obtained from the coal quality online monitoring system to establish a boiler operation history database.

[0014] The specific process of quality judgment is as follows:

[0015] Determine the time period, extract historical data, remove data that does not meet the requirements, and calculate each variable var i The mean μ i and standard deviation σ i ;

[0016] Data quality judgment, if the variable var i The value of i -3σ i , μ i +3σ i ], the data is normal, otherwise the data is abnormal.

[0017] The classification algorithm uses mechanism classification or machine learning clustering to determine the operating condition group that describes the boiler operating characteristics.

[0018] The classification results of the working conditions are as follows:

[0019] The working conditions are divided into n groups, namely GK1, GK2, ..., GK i , …, GK n ;

[0020] The historical data sets corresponding to each working condition group are data1, data2, ..., data i , …, data n .

[0021] Working Condition GK i The modeling process is as follows:

[0022] For historical data sets data i Preprocessing is performed, and the preprocessing model is modelpre,i , the preprocessed data set is data pre,i ;

[0023] After preprocessing, the data set pre,i The training set and test set are divided into data pre,train,i 、data pre,test,i ;

[0024] Determine input and output parameters;

[0025] In the training set data pre,train,i The artificial intelligence regression algorithm is used to establish a prediction model between the output parameters and the input parameters, and the model hyperparameters are optimized and the best hyperparameters are determined. pre,test,i The model effect is evaluated by adjusting the model hyperparameters until the preset evaluation index requirements are met. The final prediction model is model i .

[0026] The evaluation indexes include boiler combustion economy, environmental protection and safety effect evaluation indexes, and the operating condition group GK i The boiler combustion evaluation index is m i , respectively index i,1 、index i,2 , …, index i,j , …, Establish combustion evaluation index i,j And the prediction model for different working conditions i The relationship between .

[0027] The online rolling evaluation and online correction of the model are as follows: after the model is run, the deviation between the model simulation value and the actual value of the output parameter is calculated in real time, and a sliding time window is set to calculate the dynamic model accuracy evaluation index within the sliding time window. If the dynamic model accuracy evaluation index does not meet the pre-set conditions, automatic re-modeling is triggered, and the boiler degradation trend is tracked in real time to ensure that the prediction model matches the current characteristics of the boiler.

[0028] The objective function optimization process is:

[0029] According to the real-time value of the boiler operating characteristic parameters, the current operating condition group is determined to be GK i ;

[0030] Call the current working condition GK i Corresponding boiler combustion effect evaluation index i,1 、index i,2 , …, index i,j , …,

[0031] Multi-dimensional evaluation index i,1 、index i,2 , …, index i,j , …, Optimization is carried out to obtain recommended values ​​for the coal feed rate of each mill, the primary air volume of each mill, the total oxygen content, and the secondary air door opening.

[0032] An intelligent combustion adjustment system based on the coupling of power plant boiler operating condition identification and adaptive combustion characteristic model, comprising:

[0033] Data collection and quality judgment module: collect data and make quality judgment on the collected data;

[0034] Operating condition classification module: extracts multi-parameter historical data that characterizes boiler combustion characteristics, and uses a classification algorithm based on the historical data to classify it into different operating condition combinations;

[0035] Working condition modeling module: Based on the working condition classification results, each category is modeled separately;

[0036] Operating condition objective function setting module: Determine boiler combustion, environmental protection, and safety effect evaluation indicators for different operating condition groups, and establish the relationship between the evaluation indicators and the operating condition prediction model;

[0037] Model online rolling evaluation and model online correction module: After the model is put into operation, the accuracy of the prediction model for each working condition is dynamically calculated. If the accuracy does not meet the pre-set conditions, automatic re-modeling is triggered;

[0038] Objective function optimization module: Based on the objective function, a multi-objective optimization algorithm is used to calculate the coal feed rate of each mill, the primary air volume of each mill, the total oxygen content, and the recommended value of the secondary air door opening in real time;

[0039] Optimization command output module: Automatically output command values ​​based on the deviation between the recommended values ​​of coal feed rate for each mill, primary air volume for each mill, total oxygen content, and secondary air door opening obtained by optimizing the objective function and the current real-time values.

[0040] Beneficial effect: The present invention establishes a boiler variable load and variable coal quality optimization model to guide the boiler combustion adjustment in real time; by real-time optimization of the mill operation combination, mill ventilation volume, total oxygen content, and secondary air damper opening, it assists boiler operators to quickly adjust combustion, improve boiler efficiency, reduce nitrogen oxide emissions, and reduce pollutant generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of the present invention;

[0042] Figure 2This is the modeling process of the combustion characteristic model for different working conditions of the present invention. DETAILED DESCRIPTION

[0043] The present invention will be further described below in conjunction with the accompanying drawings.

[0044] Example 1

[0045] like Figure 1 As shown, the intelligent combustion adjustment method based on the coupling of power plant boiler operating condition identification and adaptive combustion characteristic model of the present invention includes the following steps:

[0046] S1. Data collection and quality judgment: Collect DCS or SIS real-time data through industrial communication protocols such as OPC, Modbus, and RESTful API to obtain load, economizer CO concentration, air preheater outlet exhaust temperature, SCR inlet NOx concentration, coal feed to each mill, primary air volume of each mill, total air volume, and secondary air door opening (including main burner and burnout air) real-time values. At the same time, obtain the elemental analysis (including C, H, O, N, S components) and calorific value of the coal entering the furnace from the coal quality online monitoring system to establish a boiler operation history database. The Laida criterion is used to judge the quality of the data for each variable data collected. The specific process is as follows:

[0047] S11. Determine the appropriate time period, extract historical data, remove missing values, out-of-range data, etc., and calculate each variable var i The mean μ i and standard deviation σ i ;

[0048] S12, Data quality judgment, if the variable var i The value of i -3σ i , μ i +3σ i ], the data is normal, otherwise the data is abnormal.

[0049] S2. Classification of operating conditions: Extract multivariate historical data that characterizes the combustion characteristics of the boiler, and use appropriate classification algorithms based on the historical data to divide it into different operating condition combinations with large differences in characteristics. Multivariate variables include load, economizer CO concentration, air preheater outlet exhaust gas temperature, SCR inlet NOx concentration, coal feed to each mill, primary air volume of each mill, total air volume, secondary air door opening (including main burner and burnout air), elemental analysis of coal entering the furnace (including C, H, O, N, S components) and calorific value. The classification algorithm uses mechanism classification or machine learning clustering to determine the operating condition group that describes the operating characteristics of the boiler. Mechanism classification can divide the boiler operating conditions into deep load operation condition group, low load operation condition group, medium load operation condition group, high load operation condition group, etc. according to certain rules. Machine learning clustering uses K-means (K-Means), hierarchical clustering (HierarchicalClustering), density-based clustering (Density-Based Spatial Clustering of ApplicationswithNoise) and other algorithms to divide historical data into different operating condition groups.

[0050] The final working condition classification results are as follows:

[0051] 1) The working conditions are divided into n groups, namely GK1, GK2, ..., GK i , …, GK n ;

[0052] 2) The historical data sets corresponding to each working condition group are data1, data2, σ…, data i ,σ…,data n .

[0053] S3. Modeling by operating condition: Based on the operating condition classification results, each operating condition group of the boiler is modeled separately, and an artificial intelligence regression algorithm is used to establish a prediction model for economizer CO concentration, air preheater outlet flue gas temperature, SCR inlet NOx concentration and load, coal quality, coal feed to each mill, primary air volume of each mill, total oxygen content, and secondary air door opening (including main burner and burnout air).

[0054] Working Condition GK i The modeling process is as follows:

[0055] 1) Historical data set data i Preprocessing is performed. The preprocessing algorithm uses normalization, regularization, standardization, etc. The preprocessing model is model pre,i , the preprocessed data set is data pre,i ;

[0056] 2) Preprocessed data set data pre,iDivide the data into training and test sets by leaving out the data or by time period. pre,i Divide into mutually exclusive training set and test set. After division, the training set and test set are data pre,train,i 、data pre,test,i ;

[0057] 3) Determine the input and output parameters. The input parameters include load, coal quality, coal feed rate of each mill, primary air volume of each mill, total oxygen content, secondary air door opening (including main burner and burnout air), etc. The output parameters include economizer CO concentration, air preheater outlet exhaust gas temperature, SCR inlet NOx concentration, etc.;

[0058] 4) In the training set data pre,train,i Artificial intelligence regression algorithms such as neural networks and support vector machines are used to establish a prediction model between output parameters and input parameters. The grid search method is used to optimize the model hyperparameters to determine the best hyperparameters. pre,test,i The model effect is evaluated, and the model evaluation indicators are root mean square error (RMSE), mean absolute error (MAE), R-squared, etc. By adjusting the model hyperparameters until the preset evaluation index requirements are met, the final prediction model is model i .

[0059] S4. Setting of objective function for different working conditions: Determine the corresponding evaluation index for different working condition groups, and establish the relationship between the evaluation index and the prediction model for different working conditions. Among them, the evaluation index includes boiler combustion economy, environmental protection and safety effect evaluation index. i The boiler combustion evaluation index is m i , respectively index i,1 、index i,2 , …, index i,j , …, And establish combustion evaluation index i,j And the prediction model for different working conditions i The mathematical relationship between .

[0060] S5, online rolling evaluation and online correction of the model: After the model is put into operation, the accuracy of the prediction model for each operating condition is calculated in real time. If the accuracy does not meet the pre-set conditions, it triggers automatic re-modeling and repeats S2 to S4 to ensure that the prediction model matches the current characteristics of the boiler. Specifically: After the model is running, the deviation between the simulated value and the actual value of the output parameter model is calculated in real time, and a certain sliding time window is set to calculate the dynamic model accuracy evaluation index within the sliding time window. The dynamic model accuracy evaluation index includes root mean square error (RMSE), mean absolute error (MAE), R-squared, etc. If the dynamic model accuracy evaluation index does not meet the pre-set conditions, it triggers automatic re-modeling and repeats S2 to S4. The boiler degradation trend can be tracked in real time to ensure that the prediction model matches the current characteristics of the boiler.

[0061] S6. Optimization of objective function: Multi-objective optimization algorithm is used to calculate the coal feed rate of each mill, the primary air volume of each mill, the total oxygen content, and the recommended value of the secondary air door opening in real time for objective functions such as boiler combustion economy and environmental protection effect evaluation. The optimization process is as follows:

[0062] 1) According to the real-time value of the boiler operating characteristic parameters, determine that the current operating condition group is GK i ;

[0063] 2) Call the current working condition GK determined by S4 i Corresponding boiler combustion effect evaluation index i,1 、index i,2 , …, index i,j , …,

[0064] 3) Multi-dimensional evaluation index i,1 、index i,2 , …, index i,j , …, Multi-objective genetic algorithm (MOGA), multi-objective particle swarm optimization algorithm (MOPSO), multi-objective grey wolf optimization algorithm (MOGWO), multi-objective artificial hummingbird algorithm (MOAHA) and other algorithms are used for optimization to obtain the recommended values ​​of coal feed rate, primary air volume, total oxygen content and secondary air door opening of each mill.

[0065] S7. Optimize command output: Automatically output command value based on the deviation between the recommended values ​​of coal feed rate for each mill, primary air volume for each mill, total oxygen content, and secondary air door opening obtained by optimizing the objective function and the current real-time value. If the deviation is within the adjustment dead zone, the current command value is kept unchanged; if the deviation exceeds the adjustment dead zone, the instantaneous command value is set at a certain rate of change.

[0066] like Figure 1 and Figure 2As shown, the intelligent combustion adjustment method based on the coupling of power plant boiler operating condition identification and adaptive combustion characteristic model in this embodiment includes the following steps:

[0067] Step 1: Data collection and quality judgment. Collect DCS real-time data through Modbus communication protocol to obtain load, economizer CO concentration, air preheater outlet exhaust gas temperature, SCR inlet NOx concentration, primary air temperature, secondary air temperature, total oxygen content, A mill coal feed and primary air volume, B mill coal feed and primary air volume, C mill coal feed and primary air volume, D mill coal feed and primary air volume, E mill coal feed and primary air volume, F mill coal feed and primary air volume, main burner secondary air door opening, OFA air door opening real-time value, and obtain elemental analysis (including C, H, O, N, S components) and calorific value of coal fed into each mill from the coal quality online monitoring system to establish a boiler operation history database. Use the Laida criterion to judge the quality of the data for each variable data collected. The specific process is as follows:

[0068] 1) Extract the historical data of the previous year, remove missing values, over-range, outage data, etc., and calculate the variables var i The mean μ i and standard deviation σ i ;

[0069] 2) Data quality judgment, if the variable var i The value of i -3σ i , μ i +3σ i ], the data is normal, otherwise the data is abnormal.

[0070] Step 2: Classify the operating conditions and extract the multivariate normal historical data that characterizes the combustion characteristics of the boiler. The multivariate includes load, CO concentration of the economizer, exhaust gas temperature at the air preheater outlet, NOx concentration at the SCR inlet, coal feed to each mill, primary air volume of each mill, total air volume, secondary air door opening (including main burner and burnout air), elemental analysis of coal entering the furnace (including C, H, O, N, S components) and calorific value. Based on the normal historical data, the K-means (K-Means) is used to divide the historical data into three types of operating condition groups, namely low load GK1, medium load GK2, and high load GK3. The classification results are as follows:

[0071] 1) The working conditions are divided into 3 groups, namely GK1, GK2 and GK3;

[0072] 2) The historical data sets corresponding to each working condition are data1, data2, and data3.

[0073] Step 3: Modeling by operating condition. Based on the boiler operating condition classification results determined in step 2, model the low load GK1, medium load GK2, and high load GK3 operating condition groups of the boiler respectively. Taking high load GK3 as an example, the modeling process is as follows:

[0074] 1) Preprocess the historical data set data3 using a normalized preprocessing algorithm. The preprocessing model is model pre,3 , the preprocessed data set is data pre,3 .

[0075] 2) Preprocessed data set data pre,3 The training set and test set are divided by using the random retention method. pre,3 It is divided into mutually exclusive training set and test set, of which the training set data accounts for 80%. After the division, the training set and the test set are data pre,train,3 、data pre,test,3 .

[0076] 3) Determine the input and output parameters. The input parameters include load, coal quality, coal feed rate of each mill, primary air volume of each mill, total oxygen content, secondary air door opening (including main burner and burnout air), etc. The output parameters include economizer CO concentration, air preheater outlet exhaust gas temperature, SCR inlet NOx concentration, etc.

[0077] 4) In the training set data pre,train,3 The BP neural network algorithm is used to establish a prediction model between output parameters and input parameters. The grid search method is used to determine the best hyperparameters for model hyperparameter optimization. pre,test,3 The model effect was evaluated, and the model evaluation index was that the root mean square error (RMSE) was no more than 5%. By adjusting the model hyperparameters, the number of hidden layers and the node data of each layer, the root mean square error (RMSE) was finally satisfied to be 4%, and the prediction model was model3.

[0078] The preprocessing and prediction models of working condition GK1 are model pre,1 、model1.

[0079] The preprocessing and prediction models of working condition GK2 are model pre,2 、model2.

[0080] Step 4: Setting the objective function. Determine the combustion effect evaluation index for low load GK1, medium load GK2, and high load GK3. The evaluation index for low load GK1 is the combustion stability index index 1,1 , combustion environmental protection index 1,2 , combustion economy index 1,3,, the evaluation index of medium load GK2 is combustion environmental index 2,1 , combustion economy index 2,2 The evaluation index of high-load GK3 is the combustion environmental index index 3,1 , combustion economy index 3,2 .

[0081] Step 5: Model online rolling evaluation and model online correction. After the model is put into real-time optimization, the deviation between the output parameter model simulation value and the actual value is calculated in real time, and the sliding time window is set to 24 hours. The dynamic model accuracy evaluation index is that the average root mean square error (RMSE) within the sliding window is not greater than 5%. Get the current real-time data and judge that the working condition is working condition group GK3, call the preprocessing algorithm model pre,3 After preprocessing, the real-time data is input into the prediction model model3. The average root mean square error (RMSE) of the past 24 hours is calculated to be 4%, which meets the requirements and can be used for real-time optimization but will not be updated for the time being.

[0082] Step 6: Optimize the objective function. The real-time optimization process is as follows:

[0083] 1) According to the real-time values ​​of the boiler operation characteristic parameters, determine that the current operating condition group is GK3;

[0084] 2) Call the boiler combustion effect evaluation index corresponding to the current working condition GK3 determined in step 4 as the combustion environmental protection index index 3,1 , combustion economy index 3,2 ;

[0085] 3) Multi-dimensional evaluation index 3,1 、index 3,2 The multi-objective particle swarm optimization algorithm (MOPSO) was used to find the optimal solution and obtain the recommended values ​​of coal feed rate, primary air volume, total oxygen content and secondary air door opening of each mill.

[0086] Step seven, optimize the command output: after calculation, the relative deviations between the recommended values ​​of coal feed of each mill, primary air volume of each mill, total oxygen content and secondary air door opening and the current command values ​​are 4%, 4.5%, 25% and 4.5% respectively; the relative deviations of the adjustment dead zones of coal feed of each mill, primary air volume of each mill, total oxygen content and secondary air door opening are within 5%, 5%, 5% and 5% respectively; the coal feed of each mill, primary air volume of each mill and secondary air door opening meet the adjustment dead zone conditions, and the current command remains unchanged; the total oxygen content does not meet the adjustment dead zone conditions and needs to be adjusted. The current command value is 4% and the recommended value is 5%. The total oxygen content command change rate is set to 1 / (%·min), and the oxygen content command value is adjusted to the recommended value after 1 minute.

[0087] Example 2

[0088] An intelligent combustion adjustment system based on the coupling of power plant boiler operating condition identification and adaptive combustion characteristic model, comprising:

[0089] Data collection and quality judgment module: collect DCS or SIS real-time data to obtain load, economizer CO concentration, air preheater outlet exhaust temperature, SCR inlet NOx concentration, each mill coal feed, each mill primary air volume, total air volume, secondary air door opening (including main burner and burnout air) real-time values, and obtain the element analysis (including C, H, O, N, S components) and calorific value of the coal entering the furnace from the coal quality online monitoring system. Use appropriate basis to judge the quality of the collected data.

[0090] Operating condition classification module: extracts multi-parameter historical data that characterizes boiler combustion characteristics, and uses a suitable classification algorithm based on the historical data to divide it into different operating condition combinations with greatly different characteristics.

[0091] Modeling module for different operating conditions: Based on the operating condition classification results, each type is modeled separately, and an artificial intelligence regression algorithm is used to establish a prediction model for economizer CO concentration, air preheater outlet flue gas temperature, SCR inlet NOx concentration and load, coal quality, coal feed to each mill, primary air volume of each mill, total oxygen content, and secondary air door opening (including main burner and burnout air).

[0092] Operating condition objective function setting module: determine the appropriate boiler combustion, environmental protection, and safety effect evaluation indicators for different operating condition groups, and establish the mathematical relationship between the evaluation indicators and the operating condition prediction model.

[0093] Model online rolling evaluation and model online correction module: After the model is put into operation, the accuracy of the prediction model for each operating condition is dynamically calculated. If the accuracy does not meet the pre-set conditions, automatic re-modeling is triggered to ensure that the prediction model matches the current characteristics of the boiler.

[0094] Objective function optimization module: Based on objective functions such as boiler combustion economy and environmental protection effect evaluation, a multi-objective optimization algorithm is used to calculate in real time the coal feed rate of each mill, the primary air volume of each mill, the total oxygen content, and the recommended value of the secondary air door opening.

[0095] Optimization command output module: Automatically output command values ​​based on the deviation between the recommended values ​​of coal feed rate for each mill, primary air volume for each mill, total oxygen content, and secondary air door opening obtained by optimizing the objective function and the current real-time values.

Claims

1. An intelligent combustion adjustment method based on the coupling of power plant boiler operating condition identification and adaptive combustion characteristic model, characterized in that: The following steps are involved: Data collection and quality judgment: collect data and make quality judgments on the collected variable data; Operating condition classification: extract multivariate historical data that characterizes boiler combustion characteristics, and use classification algorithms based on the historical data to classify them into different operating condition combinations with greatly different characteristics; Modeling by operating condition: Based on the operating condition classification results, model each operating condition group of the boiler operation separately; Setting of objective function for different working conditions: Determine the corresponding evaluation index for different working condition groups, and establish the relationship between the evaluation index and the prediction model for different working conditions; Online rolling model evaluation and online model correction: After the model is put into operation, the accuracy of the prediction model for each operating condition is calculated in real time. If the accuracy does not meet the pre-set conditions, automatic re-modeling is triggered to ensure that the prediction model matches the current characteristics of the boiler; Optimizing the objective function: A multi-objective optimization algorithm is used to calculate the coal feed rate, primary air volume, total oxygen content, and recommended values ​​of secondary air door opening of each mill in real time based on the objective function; Optimize command output: Automatically output command value based on the deviation between the recommended values ​​of coal feed rate for each mill, primary air volume for each mill, total oxygen content, and secondary air door opening obtained by optimizing the objective function and the current real-time value. If the deviation is within the adjustment dead zone, the current command value is kept unchanged; if the deviation exceeds the adjustment dead zone, the instantaneous command value is set.

2. According to claim 1, an intelligent combustion adjustment method based on coupling of power plant boiler operating condition identification and adaptive combustion characteristic model is characterized in that: The data collection is to collect real-time data from DCS or SIS to obtain load, economizer CO concentration, air preheater outlet flue gas temperature, SCR inlet NOx concentration, coal feed to each mill, primary air volume of each mill, total air volume, and real-time value of secondary air door opening. At the same time, the elemental analysis and calorific value of the coal entering the furnace are obtained from the coal quality online monitoring system to establish a boiler operation history database.

3. The intelligent combustion adjustment method based on coupling of power plant boiler operating condition identification and adaptive combustion characteristic model according to claim 2 is characterized in that: The specific process of quality judgment is as follows: Determine the time period, extract historical data, remove data that does not meet the requirements, and calculate each variable var i The mean μ i and standard deviation σ i ; Data quality judgment, if the variable var i The value of i -3σ i , μ i +3σ i ], the data is normal, otherwise the data is abnormal.

4. The intelligent combustion adjustment method based on coupling of power plant boiler operating condition identification and adaptive combustion characteristic model according to claim 1 is characterized in that: The classification algorithm uses mechanism classification or machine learning clustering to determine the operating condition group that describes the boiler operating characteristics.

5. The intelligent combustion adjustment method based on coupling of power plant boiler operating condition identification and adaptive combustion characteristic model according to claim 1 or 4, characterized in that: The classification results of the working conditions are as follows: The working conditions are divided into n groups, namely GK1, GK2, ..., GK i , …, GK n ; The historical data sets corresponding to each working condition group are data1, data2, ..., data i , …, data n .

6. The intelligent combustion adjustment method based on coupling of power plant boiler operating condition identification and adaptive combustion characteristic model according to claim 5 is characterized in that: Working Condition GK i The modeling process is as follows: For historical data sets data i Preprocessing is performed, and the preprocessing model is model pre,i , the preprocessed data set is data pre,i ; After preprocessing, the data set pre,i The training set and test set are divided into data pre,train,i 、data pre,test,i ; Determine input and output parameters; In the training set data pre,train,i The artificial intelligence regression algorithm is used to establish a prediction model between the output parameters and the input parameters, and the model hyperparameters are optimized and the best hyperparameters are determined. pre,test,i The model effect is evaluated by adjusting the model hyperparameters until the preset evaluation index requirements are met. The final prediction model is model i .

7. The intelligent combustion adjustment method based on coupling of power plant boiler operating condition identification and adaptive combustion characteristic model according to claim 6 is characterized in that: The evaluation indexes include boiler combustion economy, environmental protection and safety effect evaluation indexes, and the operating condition group GK i The boiler combustion evaluation index is m i , respectively index i,1 、index i,2 , …, index i,j , …, index i,mi , establish combustion evaluation index i,j And the prediction model for different working conditions i The relationship between .

8. The intelligent combustion adjustment method based on coupling of power plant boiler operating condition identification and adaptive combustion characteristic model according to claim 7 is characterized in that: The online rolling evaluation and online correction of the model are specifically as follows: after the model is running, the deviation between the simulated value and the actual value of the output parameter model is calculated in real time, and a sliding time window is set to calculate the dynamic model accuracy evaluation index within the sliding time window. If the dynamic model accuracy evaluation index does not meet the pre-set conditions, automatic remodeling is triggered.

9. The intelligent combustion adjustment method based on coupling of power plant boiler operating condition identification and adaptive combustion characteristic model according to claim 8 is characterized in that: The objective function optimization process is: According to the real-time value of the boiler operating characteristic parameters, the current operating condition group is determined to be GK i ; Call the current working condition GK i Corresponding boiler combustion effect evaluation index i,1 、index i,2 , …, index i,j , …, Multi-dimensional evaluation index i,1 、index i,2 , …, index i,j , …, Optimization is carried out to obtain recommended values ​​for the coal feed rate of each mill, the primary air volume of each mill, the total oxygen content, and the secondary air door opening.

10. An intelligent combustion adjustment system based on power plant boiler operating condition identification and adaptive combustion characteristic model coupling, characterized in that: include: Data collection and quality judgment module: collect data and make quality judgment on the collected data; Operating condition classification module: extracts multi-parameter historical data that characterizes boiler combustion characteristics, and uses a classification algorithm based on the historical data to classify it into different operating condition combinations; Working condition modeling module: Based on the working condition classification results, each type is modeled separately; Operating condition objective function setting module: Determine boiler combustion, environmental protection, and safety effect evaluation indicators for different operating condition groups, and establish the relationship between the evaluation indicators and the operating condition prediction model; Model online rolling evaluation and model online correction module: After the model is put into operation, the accuracy of the prediction model for each working condition is dynamically calculated. If the accuracy does not meet the pre-set conditions, automatic re-modeling is triggered; Objective function optimization module: Based on the objective function, a multi-objective optimization algorithm is used to calculate the coal feed rate of each mill, the primary air volume of each mill, the total oxygen content, and the recommended value of the secondary air door opening in real time; Optimization command output module: Automatically output command values ​​based on the deviation between the recommended values ​​of coal feed rate for each mill, primary air volume for each mill, total oxygen content, and secondary air door opening obtained by optimizing the objective function and the current real-time values.

Images