A combustor control system and method for reducing thermal shock under variable load of a water-cooled wall

Abstract

The present application provides a kind of combustor regulating system and method for reducing thermal shock of water-cooled wall under load change, the method comprises the following steps: S101, obtain boiler parameters, historical operation data of boiler and boiler variables;S102, build furnace combustion model based on boiler parameters and verify;S103, different states of boiler variables are arranged as different operation combinations using orthogonal test method, and corresponding furnace water-cooled wall temperature distribution data are obtained;S104, the operation combination of different boiler variables and its corresponding furnace water-cooled wall temperature distribution data are input into machine learning model for training;S105, predict furnace water-cooled wall temperature distribution data under load at a given time in the future;S106, screen out the operation combination of the boiler variable with over-temperature data;S107, screen out the combustor combination that minimizes the thermal shock of water-cooled wall under load change.The present application reduces the thermal shock caused by load change on water-cooled wall pipeline, and realizes the safe and stable operation of deep peak shaving boiler.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of boiler load variation regulation, and particularly relates to a combustor regulation system and method for reducing thermal shock of a water-cooled wall under load variation. BACKGROUND

[0002] With large-scale grid connection of clean energy such as wind power and photovoltaic power, the main task of a thermal power plant gradually changes to deep peak regulation and frequency modulation, so as to make up for the instability of new energy. However, frequent changes in the load of the boiler seriously affect the operation safety and service life of equipment and key components. The water-cooled wall, as the main heating surface, is most affected by the load variation of the boiler, and is more likely to cause a tube burst accident when subjected to thermal shock during boiler operating condition variation, thereby causing the boiler to shut down and causing huge economic losses and safety hazards. SUMMARY

[0003] In view of this, the present application aims to provide a combustor regulation system and method for reducing thermal shock of a water-cooled wall under load variation, so as to solve or at least partially solve the above-mentioned problems existing in the prior art.

[0004] To achieve the above-mentioned purpose, the present application provides a combustor regulation method for reducing thermal shock of a water-cooled wall under load variation, which comprises the following steps:

[0005] S101, obtaining boiler parameters, historical operation data of the boiler, and boiler variables;

[0006] S102, constructing a furnace combustion model based on the boiler parameters, and verifying the furnace combustion model according to the historical operation data;

[0007] S103, dividing the boiler variables into different states based on the historical operation data, and using an orthogonal test method to arrange the different states of the boiler variables into different operating combinations, and obtaining furnace water-cooled wall temperature distribution data corresponding to the different operating combinations through the furnace combustion model;

[0008] S104, inputting the different operating combinations of the boiler variables and the corresponding furnace water-cooled wall temperature distribution data into a machine learning model for training;

[0009] S105, predicting furnace water-cooled wall temperature distribution data corresponding to different operating combinations of the boiler variables at a given time in the future based on the trained machine learning model;

[0010] S106, comparing the predicted furnace water-cooled wall temperature distribution data with the temperature upper limit of the corresponding region, screening out operating combinations of the boiler variables with over-temperature data, and discarding them;

[0011] S107, compare the corresponding furnace water wall temperature distribution data of the operation combination of the boiler variables left after screening with the temperature distribution data of the previous historical moment of the furnace water wall, and screen out the burner combination that minimizes the thermal shock of the water wall under variable load.

[0012] Further, the boiler variables include boiler load, pulverized coal concentration, burner start-stop combination, burner angle, and burner coal supply amount.

[0013] Further, the machine learning model includes an input layer, a hidden layer, and an output layer, and the hidden layer is additionally provided with three gating mechanisms of a forget gate, an input gate, and an output gate.

[0014] Further, the training process of the hidden layer includes:

[0015] S41, delete part of the information through the forget gate, save the effective data, and represent as follows:

[0016]

[0017] wherein, is the forget gate, , is t input variable of the moment model, forget gate bias;

[0018] S42, determine the input information proportion through the input gate, and screen the input information, and represent as follows:

[0019]

[0020]

[0021]

[0022] wherein, is the input information proportion, is the input gate weight, i is the input gate, is the input gate bias, is t candidate state of the memory unit at the moment, is the activation function, is the alternative state weight, c is the alternative state, is the alternative state bias, is t memory unit state at the moment, is memory unit state at the moment;

[0023] S43, determine the output ratio through the output gate, and screen the output information to obtain an ideal output result, which is expressed as follows:

[0024]

[0025]

[0026] wherein, is the output ratio, is the model output, is the output gate weight, is the output gate bias.

[0027] Further, in step S105, the specific steps of obtaining the furnace water wall temperature distribution data corresponding to different combinations of boiler variables under the predicted load at a given future time are as follows:

[0028] S51, recombine the boiler load, the pulverized coal concentration, and the different states of the burner start-stop combination, the burner angle, and the burner coal supply amount under the load at a given future time into a new operation combination;

[0029] S52, input the new operation combination into the trained machine learning model;

[0030] S53, obtain the furnace water wall temperature distribution data corresponding to the new operation combination based on step S52.

[0031] Further, the specific steps of screening out the burner combination scheme that minimizes the thermal shock of the furnace water wall under variable load are as follows:

[0032] S71, obtain the temperature distribution data of the furnace water wall at a previous historical time;

[0033] S72, compare the furnace water wall temperature distribution data corresponding to the boiler variable operation combination remaining after step S106 with the temperature distribution data of the furnace water wall at a previous historical time;

[0034] S73, calculate the temperature difference of the corresponding grid points of the furnace water wall at the previous and subsequent times based on the comparison result of step S72;

[0035] S74, calculate the average error of the entire furnace water wall temperature distribution data based on the temperature difference of the corresponding grid points of the furnace water wall at the previous and subsequent times, and select the furnace water wall temperature distribution data with the smallest average error as the optimal adjustment result;

[0036] S75, take the combination of boiler variables corresponding to the furnace water wall temperature distribution data with the smallest average error as the burner combination that minimizes the thermal shock of the water wall under variable load.

[0037] The second aspect of the present application provides a combustor regulation system for reducing thermal shock of a water-cooled wall under variable load, comprising:

[0038] A data acquisition module is configured to acquire boiler parameters, historical operation data of the boiler, and boiler variables.

[0039] A modeling and simulation module is configured to construct a furnace combustion model according to the boiler parameters, verify the furnace combustion model according to the historical operation data, divide the boiler variables into different states based on the historical operation data, arrange the different states of the boiler variables using an orthogonal test method to obtain different operation combinations, and acquire furnace water-cooled wall temperature distribution data corresponding to the different operation combinations through the furnace combustion model.

[0040] A machine learning module is configured to input the different operation combinations of the boiler variables and the corresponding furnace water-cooled wall temperature distribution data into a machine learning model for training, predict furnace water-cooled wall temperature distribution data corresponding to different operation combinations of the boiler variables at a given time in the future based on the trained machine learning model, compare the predicted furnace water-cooled wall temperature distribution data with the temperature upper limit of the corresponding region, filter out the operation combination of the boiler variable with over-temperature data, discard the operation combination, and finally compare the furnace water-cooled wall temperature distribution data corresponding to the remaining operation combination of the boiler variable after filtering with the temperature distribution data of the previous historical time of the furnace water-cooled wall, and filter out the combustor combination that causes the least thermal shock to the water-cooled wall under variable load.

[0041] Compared with the prior art, the present application has the following advantages:

[0042] The present application proposes a combustor regulation system and method for reducing thermal shock of a water-cooled wall under variable load, which uses a machine learning model to find the relationship between the operation combination of the boiler variables and the furnace water-cooled wall temperature distribution, predicts the furnace water-cooled wall temperature distribution, filters the prediction results, optimizes the combustion organization under variable load of the boiler, actively prevents over-temperature of the furnace water-cooled wall, compares the furnace water-cooled wall temperature distribution data corresponding to the remaining operation combination of the boiler variable after filtering with the temperature distribution data of the previous historical time of the furnace water-cooled wall, and filters out the combustor combination that causes the least thermal shock to the water-cooled wall under variable load, realizes smooth change of the boiler load, reduces thermal shock caused by load variation to the water-cooled wall pipeline, prolongs the fatigue life of the heating surface, and helps to realize safe and stable operation of the deep peak shaving boiler. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description are only preferred embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0044] Figure 1 A flowchart of the combustor for reducing thermal shock of water-cooled wall under variable load is provided for the embodiments of the present application.

[0045] Figure 2 An optimization example diagram of the combustor start-stop combination is provided for the embodiments of the present application.

[0046] Figure 3 An orthogonal test example diagram is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0047] The principles and characteristics of the present application are described below in combination with the drawings, and the listed embodiments are only used to explain the present application and are not used to limit the scope of the present application.

[0048] The present embodiment provides a combustor control method for reducing thermal shock of water-cooled wall under variable load, characterized in that the method comprises the following steps:

[0049] As shown in Figure 1 , a flowchart of the combustor for reducing thermal shock of water-cooled wall under variable load is provided, which comprises:

[0050] S101, obtaining boiler parameters, historical operation data of the boiler and boiler variables, specifically comprising:

[0051] The boiler variables include: boiler load, pulverized coal concentration, combustor start-stop combination, combustor angle, and combustor coal supply amount.

[0052] S102, constructing a furnace combustion model based on the boiler parameters, and verifying the furnace combustion model according to the historical operation data.

[0053] S103, dividing the boiler variables into different states based on the historical operation data, and using an orthogonal test method to arrange the different states of the boiler variables into different operation combinations, and obtaining the furnace water-cooled wall temperature distribution data corresponding to the different operation combinations through the furnace combustion model, specifically comprising:

[0054] The working states of the burners do not affect each other, different start-stop combinations can be carried out, and the burner start-stop combination can be changed according to the characteristics of different boilers. Taking the opposed pulverized coal boiler with a front three and rear three arrangement as an example, six burners are arranged in each row, five rows of burners are started to carry out flame collision during normal operation, and the load is changed, which is easy to cause the center of the furnace flame to deviate. Figure 2 The burner start-stop combination optimization diagram, Figure 2 In (a), (b), (c), (d), the load is continuously reduced, but through the optimization of the burner start-stop combination and other parameters, the flame deviation can be minimized. The working state of the burner is coded as 0 and 1, 0 represents that the burner stops working, and 1 represents that the burner is started to operate, and a matrix is formed to input the combustion model, Figure 2 The burner start-stop combination optimization diagram shown is only an example, and other combinations can be carried out according to the site conditions;

[0055] Taking four levels of boiler variables as an example, an orthogonal test table is designed, as shown in Figure 3 The boiler load, pulverized coal concentration, burner start-stop combination, burner angle, and burner coal supply amount are sequentially marked as factors A, B, C, D, and E, and the combinations (a)-(d) shown in Figure 2 The combinations (a)-(d) are marked as levels 1, 2, 3, and 4 of factor C, respectively, to obtain 16 running combinations, which are input into the furnace combustion model to obtain 16 furnace water wall temperature distribution data under the corresponding running combinations, Figure 3 The orthogonal test table shown is only an example, and different numbers of levels can be set according to the change range of historical data during actual operation to obtain more rich and comprehensive running combinations.

[0056] S104, input different running combinations of boiler variables and their corresponding furnace water wall temperature distribution data into a machine learning model for training, specifically including:

[0057] Taking the running combination of the boiler load, pulverized coal concentration, burner start-stop combination, burner angle, and burner coal supply amount as input, and the furnace water wall temperature distribution data as output, the machine learning model can master the mapping relationship between the input and the output;

[0058] The machine learning model can select a long short-time memory network model (Long Short-Time Momery, LSTM), and the LSTM model mainly consists of an input layer, a hidden layer, and an output layer. To solve the gradient disappearance problem, the hidden layer adds three gating mechanisms of a forgetting gate, an input gate, and an output gate, which are suitable for solving long-term dependence problems;

[0059] The training process of the hidden layer includes:

[0060] S41, delete part of information through the forgetting gate, filter out the operation combination with different boiler variables and save the effective data of the corresponding furnace water wall temperature distribution data, as shown below:

[0061]

[0062] wherein, is the forgetting gate, , is t input variable of the moment model, is the forgetting gate bias;

[0063] S42, determine the input information ratio through the input gate, and filter the input information, as shown below:

[0064]

[0065]

[0066]

[0067] wherein, is the input information ratio, is the input gate weight, i is the input gate, is the input gate bias, is t candidate state of the memory unit at the moment, is the activation function, is the candidate state weight, c is the candidate state, is the candidate state bias, is t memory unit state at the moment, is memory unit state at the moment;

[0068] S43, determine the output ratio through the output gate, and filter the output information to obtain the ideal output result, as shown below:

[0069]

[0070]

[0071] wherein, is the output ratio, is the model output, is the output gate weight, is the output gate bias.

[0072] S105, predicting the furnace water wall temperature distribution data corresponding to different boiler variable operation combinations under the future given load based on the trained machine learning model, specifically comprising the following steps:

[0073] S51, recombining the boiler load, the pulverized coal concentration, and the different states of the burner start-stop combination, the burner angle, and the burner coal supply amount under the future given load into a new operation combination;

[0074] S52, inputting the new operation combination into the trained machine learning model;

[0075] S53, obtaining the furnace water wall temperature distribution data corresponding to the new operation combination based on step S52.

[0076] S106, comparing the predicted furnace water wall temperature distribution data with the temperature upper limit of the corresponding region, screening out the boiler variable operation combination with over-temperature data, and discarding it.

[0077] S107, comparing the furnace water wall temperature distribution data corresponding to the remaining boiler variable operation combination after screening with the temperature distribution data of the previous historical time of the furnace water wall, screening out the burner combination that minimizes the thermal shock of the furnace water wall under variable load, specifically comprising the following steps:

[0078] S71, obtaining the temperature distribution data of the previous historical time of the furnace water wall;

[0079] S72, comparing the furnace water wall temperature distribution data corresponding to the remaining boiler variable operation combination after step S106 with the temperature distribution data of the previous historical time of the furnace water wall;

[0080] S73, calculating the temperature difference of the corresponding grid points of the furnace water wall at the previous and subsequent times based on the comparison result of step S72;

[0081] S74, calculating the average error of the entire furnace water wall temperature distribution data based on the temperature difference of the corresponding grid points of the furnace water wall at the previous and subsequent times, and selecting the furnace water wall temperature distribution data with the smallest average error as the optimal adjustment result;

[0082] S75, taking the boiler variable operation combination corresponding to the furnace water wall temperature distribution data with the smallest average error as the burner combination that minimizes the thermal shock of the water wall under variable load.

[0083] Another embodiment of the present application provides a burner control system for reducing thermal shock of water wall under variable load, comprising:

[0084] The data acquisition module is used to acquire the boiler parameters, the historical operation data of the boiler, and the boiler variables;

[0085] Modeling simulation module: for constructing a furnace combustion model according to the boiler parameters, verifying the furnace combustion model according to the historical operation data, and dividing the boiler variables into different states based on the historical operation data, using the orthogonal test method to arrange the different states of the boiler variables, obtaining different operation combinations, and obtaining the furnace water wall temperature distribution data corresponding to the different operation combinations through the furnace combustion model;

[0086] Machine learning module: for inputting the different operation combinations of the boiler variables and the corresponding furnace water wall temperature distribution data into a machine learning model for training, predicting the furnace water wall temperature distribution data corresponding to the different operation combinations of the boiler variables under a given load at a future time based on the trained machine learning model, comparing the predicted furnace water wall temperature distribution data with the temperature upper limit of the corresponding region, screening out the operation combination of the boiler variables with over-temperature data, discarding the operation combination, and finally comparing the furnace water wall temperature distribution data corresponding to the remaining operation combination of the boiler variables after screening with the temperature distribution data of the previous historical time of the furnace water wall, and screening out the burner combination that minimizes the thermal shock of the water wall under variable load;

[0087] Data storage module: for storing the temperature distribution data of the previous historical time of the furnace water wall, the machine learning module can access the data stored in this module at any time and compare the prediction results, and the optimal result screened out by the machine learning module is also stored in the data storage module and compared with the next machine learning model prediction.

[0088] The above is only the preferred embodiment of the present application, and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method of burner regulation to reduce thermal shock in water-cooled walls under variable load, characterized in that, The method comprises the following steps: S101, obtaining boiler parameters, historical operation data of the boiler, and boiler variables; S102, constructing a furnace combustion model based on the boiler parameters, and verifying the furnace combustion model according to the historical operation data; S103, dividing the boiler variables into different states based on the historical operation data, and using an orthogonal test method to arrange the different states of the boiler variables into different operation combinations, and obtaining furnace water wall temperature distribution data corresponding to the different operation combinations through the furnace combustion model, specifically comprising: The burners can be combined in different start-stop combinations, which can be changed according to the characteristics of different boilers, and the working states of the burners are coded as 0 and 1, 0 indicating that the burners stop working, and 1 indicating that the burners are turned on and run, to form a matrix input into the combustion model; S104, inputting the different operation combinations of the boiler variables and the corresponding furnace water wall temperature distribution data into a machine learning model for training; S105, predicting the furnace water wall temperature distribution data corresponding to different operation combinations of the boiler variables under a given load at a future time based on the trained machine learning model, specifically comprising the following steps: S51, recombining the boiler load, the pulverized coal concentration, and the different states of the start-stop combination of the burner, the angle of the burner, and the powder supply amount of the burner under the given load at the future time into a new operation combination; S52, inputting the new operation combination into the trained machine learning model; S53, obtaining the furnace water wall temperature distribution data corresponding to the new operation combination based on step S52; S106, comparing the predicted furnace water wall temperature distribution data with the temperature upper limit of the corresponding region, screening out the operation combination of the boiler variable with over-temperature data, and discarding it; S107, comparing the furnace water wall temperature distribution data corresponding to the remaining operation combination of the boiler variable after screening with the temperature distribution data of the furnace water wall at the previous historical time, and screening out the burner combination that minimizes the thermal shock of the water wall under variable load, specifically comprising the following steps: S71, obtaining the temperature distribution data of the furnace water wall at the previous historical time; S72, comparing the furnace water wall temperature distribution data corresponding to the remaining operation combination of the boiler variable after step S106 with the temperature distribution data of the furnace water wall at the previous historical time; S73, calculating the temperature difference of the corresponding grid points of the furnace water wall at the previous and next times based on the comparison result of step S72; S74, calculating the average error of the entire furnace water wall temperature distribution data based on the temperature difference of the corresponding grid points of the furnace water wall at the previous and next times, and selecting the furnace water wall temperature distribution data with the minimum average error as the optimal adjustment result; S75, taking the operation combination of the boiler variable corresponding to the furnace water wall temperature distribution data with the minimum average error as the burner combination that minimizes the thermal shock of the water wall under variable load.

2. A method of regulating a combustor to reduce thermal shock under load changes in a water-cooled wall according to claim 1, wherein The boiler variables include the boiler load, the pulverized coal concentration, the start-stop combination of the burner, the angle of the burner, and the powder supply amount of the burner.

3. A method of regulating a combustor to reduce thermal shock under load changes in a water-cooled wall according to claim 1, wherein The machine learning model comprises an input layer, a hidden layer, and an output layer, and the hidden layer is additionally provided with three gating mechanisms of a forgetting gate, an input gate, and an output gate.

4. A method of regulating a combustor to reduce thermal shock under load changes in a water-cooled wall according to claim 3, wherein The training process of the implicit layer includes: S41, delete part of the information through the forgetting gate, filter out the effective data for saving, as shown below: f t = σ(w f [ h t-1 , x t ] + b f ) wherein f t is the information forgetting ratio, σ is the activation function, W f is the forgetting gate weight, f is the forgetting gate, h t-1 is the output of the memory cell at time t-1, x t is the input variable of the model at time t, b f is the forgetting gate bias; S42, determine the input information proportion through the input gate, and filter the input information, as shown below: i t = σ(W i [ h t-1 , x t ] + b i ) I t = tanh(W c [h t-1 ,x t ]+ b c ) C t = f t * C t-1 + i t I t wherein, i t is the input information ratio, W i is the input gate weight, i is the input gate, b i is the input gate bias, I t is the candidate state of the memory cell at time t, tanh is the activation function, w c is the candidate state weight, c is the candidate state, b c is the candidate state bias, C t is the memory cell state at time t, C t-1 is the memory cell state at time t-1; S43, determine the output proportion through the output gate, and filter the output information to obtain the ideal output result, as shown below: O t = σ(W o + b t-1 ) [h t , x o ] h t =O t *tanh(C t ) where O t is the output ratio, h t is the model output, W o is the output gate weight, and b o is the output gate bias.

5. A combustor control system to reduce thermal shock under water wall load changes, characterized in that, It includes: The data acquisition module is used for acquiring the boiler parameters, the historical operation data of the boiler and the boiler variables; The modeling simulation module is used for constructing the furnace combustion model according to the boiler parameters, verifying the furnace combustion model according to the historical operation data, and dividing the boiler variables into different states based on the historical operation data, arranging the different states of the boiler variables by using the orthogonal test method to obtain different operation combinations, obtaining the furnace water wall temperature distribution data corresponding to the different operation combinations through the furnace combustion model, and the burner can be combined in different start-stop combinations, the burner start-stop combination can be changed according to the characteristics of different boilers, the working state of the burner is coded as 0 and 1, 0 represents that the burner stops working, and 1 represents that the burner starts working, to form a matrix input into the combustion model; The machine learning module is used for inputting the different operation combinations of the boiler variables and the corresponding furnace water wall temperature distribution data into the machine learning model for training, predicting the furnace water wall temperature distribution data corresponding to the different operation combinations of the boiler variables under the future given load based on the trained machine learning model, recombining the different states of the boiler load, the pulverized coal concentration and the burner start-stop combination, the burner angle and the burner coal supply amount under the future given load into a new operation combination, inputting the new operation combination into the trained machine learning model to obtain the furnace water wall temperature distribution data corresponding to the new operation combination, comparing the predicted furnace water wall temperature distribution data with the temperature upper limit of the corresponding region, screening out the operation combination of the boiler variable with the over-temperature data, discarding the operation combination, and finally comparing the furnace water wall temperature distribution data corresponding to the remaining operation combination of the boiler variable with the temperature distribution data of the previous historical time of the furnace water wall, screening out the burner combination that causes the least thermal shock to the water wall under variable load, including acquiring the temperature distribution data of the previous historical time of the furnace water wall, comparing the furnace water wall temperature distribution data corresponding to the remaining operation combination of the boiler variable with the temperature distribution data of the previous historical time of the furnace water wall, calculating the temperature difference of the corresponding grid points of the furnace water wall at the previous and next times based on the comparison result, calculating the average error of the entire furnace water wall temperature distribution data based on the temperature difference of the corresponding grid points of the furnace water wall at the previous and next times, selecting the furnace water wall temperature distribution data with the minimum average error as the optimal adjustment result, and selecting the operation combination of the boiler variable corresponding to the furnace water wall temperature distribution data with the minimum average error as the burner combination that causes the least thermal shock to the water wall under variable load.

Images