A circulating fluidized bed boiler furnace outlet pressure control optimization method and system

Abstract

The present application discloses a circulating fluidized bed boiler furnace outlet pressure control optimization method and system, which relates to the technical field of boiler pressure control. The method includes obtaining historical operation data of the circulating fluidized bed boiler; determining the influencing factor characteristic data and furnace outlet pressure data according to the historical operation data; inputting the influencing factor characteristic data into the air supply system prediction model to obtain the air supply change corresponding to the factor characteristic data; inputting the air supply change into the air supply system and induced draft system relationship model to obtain the first induced draft change; determining the second induced draft change under the control of the PID controller according to the furnace outlet pressure data; determining the final induced draft change according to the first induced draft change and the second induced draft change, and the final induced draft change is the induced draft change that the induced draft system currently needs to adjust. The present application can improve the accuracy and effect of furnace outlet pressure control.

Description

Technical Field

[0001] The present application relates to the technical field of boiler pressure control, and in particular to a circulating fluidized bed boiler furnace outlet pressure control optimization method and system. Background Art

[0002] Circulating fluidized bed boilers generally include two major parts: a combustion chamber and a circulating furnace. They use fluidized combustion and are a highly industrialized clean coal combustion technology. They have high operating wind speeds, which strengthen heterogeneous reaction processes such as combustion and desulfurization. The boiler capacity can be expanded to a large capacity (600MW or above) that the power industry can accept. Circulating fluidized bed boilers have well solved basic problems such as thermodynamics, mechanics, and materials science, as well as engineering problems such as expansion, wear, and overheating, and have become an advanced technology for energy utilization of difficult-to-burn solid fuels (such as coal gangue, oil shale, urban garbage, sludge, and other wastes). However, for circulating fluidized bed boilers, if the furnace outlet pressure of the circulating fluidized bed boiler is too high or too low, it will have a significant impact on equipment life, safety, economic benefits, and other aspects. Therefore, it is very important to control the furnace outlet pressure of the circulating fluidized bed boiler to keep the furnace outlet pressure of the circulating fluidized bed boiler stable.

[0003] At present, the furnace outlet pressure control of circulating fluidized bed boilers is generally to integrate the PID (Proportional-Integral-Derivative) controller into the DCS (Distributed Control System), and use the PID controller in the DCS system to control the furnace outlet pressure. Specifically, when the operating conditions or external factors change, the furnace outlet pressure will be too high or too low. At this time, the PID controller adjusts the speed frequency of the induced draft fan in the induced draft system to pull the furnace outlet pressure back to the target pressure range, thereby achieving control of the furnace outlet pressure.

[0004] Although this furnace outlet pressure control method based on PID controller has a rapid control response, its control method is relatively simple and relies entirely on the PID controller for control. Moreover, the PID controller will only take action to control when the furnace outlet pressure is too high or too low. Therefore, this method has the problems of low accuracy and poor effect of furnace outlet pressure control. Summary of the invention

[0005] The purpose of this application is to provide a circulating fluidized bed boiler furnace outlet pressure control optimization method and system, which can improve the accuracy and effect of furnace outlet pressure control.

[0006] To achieve the above objectives, this application provides the following solutions.

[0007] In the first aspect, the present application proposes a method for optimizing the control of the furnace outlet pressure of a circulating fluidized bed boiler, wherein the circulating fluidized bed boiler comprises a furnace, an induced draft system, an air supply system and a PID controller, wherein the air outlet of the air supply system and the air inlet of the induced draft system are both connected to the interior of the furnace, and the PID controller is respectively connected to the induced draft system and the air supply system, and the method for optimizing the furnace outlet pressure of the circulating fluidized bed boiler comprises the following steps.

[0008] Acquire historical operation data of the circulating fluidized bed boiler; the historical operation data is the equipment operation data of the circulating fluidized bed boiler that has been running continuously in the past period of time.

[0009] According to the historical operation data, the influencing factor characteristic data and the furnace outlet pressure data are determined; the influencing factor characteristic data are data corresponding to various factors that affect the size of the air supply change of the air supply system; the influencing factor characteristic data include coal supply amount, oxygen amount and boiler bed temperature.

[0010] The influencing factor characteristic data is input into the air supply system prediction model to obtain the air supply change corresponding to the factor characteristic data; the air supply system prediction model is a neural network model established based on the air supply system and trained according to the influencing factor characteristic data and the actual air supply change.

[0011] The air supply change is input into the relationship model between the air supply system and the induced draft system to obtain the first induced draft change; the relationship model between the air supply system and the induced draft system is a neural network model established based on the air supply system and the induced draft system, and trained according to the air supply change and the actual induced draft change.

[0012] According to the furnace outlet pressure data, a second induced draft change amount under the control of the PID controller is determined.

[0013] A final induced air change amount is determined according to the first induced air change amount and the second induced air change amount; the final induced air change amount is the induced air change amount that the induced air system currently needs to adjust.

[0014] Optionally, before the step of inputting the influencing factor characteristic data into the air supply system prediction model to obtain the air supply change corresponding to the factor characteristic data, the circulating fluidized bed boiler furnace outlet pressure control optimization method further includes the following steps.

[0015] An air supply system prediction model is established, and the coal quantity, oxygen quantity and boiler bed temperature are used as inputs, and the corresponding actual air supply change amount is used as output, and the air supply system prediction model is trained.

[0016] Optionally, the air supply system prediction model is a TCN model, an LSTM model or a Transformer model.

[0017] Optionally, the loss function of the air supply system prediction model is as follows.

[0018]

[0019] Wherein, L1 represents the loss function of the air supply system prediction model, is the predicted value of the air supply change, and y1 is the actual value of the air supply change.

[0020] Optionally, before the step of inputting the supply air variation into the relationship model between the supply air system and the induced draft system to obtain the first induced draft variation, the circulating fluidized bed boiler furnace outlet pressure control optimization method further includes the following steps.

[0021] A relationship model between the air supply system and the induced draft system is established, and the air supply change is used as input and the corresponding actual induced draft change is used as output, and the relationship model between the air supply system and the induced draft system is trained.

[0022] Optionally, the relationship model between the air supply system and the induced air system is a linear model, a SVM model or a GBDT model.

[0023] Optionally, the loss function of the relationship model between the air supply system and the induced air system is as follows.

[0024]

[0025] Wherein, L2 represents the loss function of the relationship model between the air supply system and the induced air system, is the predicted value of the induced draft change, and y2 is the actual value of the induced draft change.

[0026] Optionally, determining a final induced air change amount according to the first induced air change amount and the second induced air change amount specifically includes the following steps.

[0027] The first induced air change amount and the second induced air change amount are summed to obtain a final induced air change amount.

[0028] Optionally, after the step of summing the first induced draft variation and the second induced draft variation to obtain a final induced draft variation, the circulating fluidized bed boiler furnace outlet pressure control optimization method further includes the following steps.

[0029] According to the final induced draft change, the rotation speed of the induced draft fan in the induced draft system is adjusted to keep the furnace outlet pressure stable.

[0030] In the second aspect, the present application proposes a circulating fluidized bed boiler furnace outlet pressure control optimization system, the circulating fluidized bed boiler includes a furnace, an induced draft system, an air supply system and a PID controller, the air outlet of the air supply system and the air inlet of the induced draft system are both connected to the interior of the furnace, the PID controller is respectively connected to the induced draft system and the air supply system, and the circulating fluidized bed boiler furnace outlet pressure control optimization system includes the following functional modules.

[0031] The historical operation data acquisition module is used to acquire the historical operation data of the circulating fluidized bed boiler; the historical operation data is the equipment operation data of the circulating fluidized bed boiler that has been running continuously in the past period of time.

[0032] A data determination module is used to determine the influencing factor characteristic data and the furnace outlet pressure data according to the historical operation data; the influencing factor characteristic data is the data corresponding to various factors affecting the size of the air supply change of the air supply system; the influencing factor characteristic data includes coal supply amount, oxygen amount and boiler bed temperature.

[0033] The air supply change acquisition module is used to input the influencing factor characteristic data into the air supply system prediction model to obtain the air supply change corresponding to the factor characteristic data; the air supply system prediction model is a neural network model established based on the air supply system and trained according to the influencing factor characteristic data and the actual air supply change.

[0034] The first induced draft change acquisition module is used to input the supply air change into the relationship model between the supply air system and the induced draft system to obtain the first induced draft change; the relationship model between the supply air system and the induced draft system is a neural network model established based on the supply air system and the induced draft system, and trained according to the supply air change and the actual induced draft change.

[0035] The second induced draft change amount acquisition module is used to determine the second induced draft change amount under the control of the PID controller according to the furnace outlet pressure data.

[0036] The final induced air change amount determination module is used to determine the final induced air change amount according to the first induced air change amount and the second induced air change amount; the final induced air change amount is the induced air change amount that the induced air system currently needs to adjust.

[0037] According to the specific embodiments provided in this application, this application discloses the following technical effects.

[0038] The present application provides a circulating fluidized bed boiler furnace outlet pressure control optimization method and system, based on the historical operation data of the circulating fluidized bed boiler for modeling and analysis, by combining PID control technology with neural network technology, on the one hand, by using the correlation between the supply air volume and the induced air volume, based on the supply air system prediction model and the supply air system and induced air system relationship model two models to predict the supply air change in advance and further determine the corresponding first induced air change; on the other hand, based on the traditional PID controller according to the furnace outlet pressure to obtain the second induced air change under PID control, through the supply air system prediction model, the supply air system and induced air system relationship model and the PID controller cooperation In combination, the first induced draft change and the second induced draft change are combined to determine the induced draft change that currently needs to be adjusted in the final induced draft system. Compared with the traditional control method of a single PID controller, the present application not only enriches the control means and improves the accuracy of furnace outlet pressure control, but also because the supply air change and the induced draft change are predicted in advance based on historical operating data, therefore, the present application can sense the changes in the supply air system in advance and control the induced draft system in advance to perform synchronous actions, thereby solving the problem that the PID controller will only take action to control when the furnace outlet pressure is already too high or too low in the traditional PID control method, effectively improving the effect of furnace outlet pressure control. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0040] Figure 1 A flow chart of a circulating fluidized bed boiler furnace outlet pressure control optimization method provided in one embodiment of the present application.

[0041] Figure 2 A schematic diagram of the principle of a circulating fluidized bed boiler furnace outlet pressure control optimization method provided in one embodiment of the present application.

[0042] Figure 3 A schematic diagram of the output of a first-order linear system corresponding to a unit step signal input provided in an embodiment of the present application.

[0043] Figure 4 A schematic diagram of the structure of an air supply system prediction model provided in one embodiment of the present application.

[0044] Figure 5 A schematic structural diagram of a circulating fluidized bed boiler furnace outlet pressure control optimization system provided in one embodiment of the present application. DETAILED DESCRIPTION

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

[0046] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0047] like Figure 1 As shown, this embodiment proposes a circulating fluidized bed boiler furnace outlet pressure control optimization method, the circulating fluidized bed boiler at least includes a furnace, an induced draft system, an air supply system and a PID controller and other structures, wherein the air outlet of the air supply system and the air inlet of the induced draft system are both connected to the inside of the furnace, the PID controller is respectively connected to the induced draft system and the air supply system, and the circulating fluidized bed boiler furnace outlet pressure control optimization method includes the following steps.

[0048] Step S1, obtaining historical operation data of the circulating fluidized bed boiler; the historical operation data is the equipment operation data of the circulating fluidized bed boiler that has been running continuously in the past period of time.

[0049] Step S2, determining the influencing factor characteristic data and the furnace outlet pressure data according to the historical operation data; the influencing factor characteristic data is the data corresponding to various factors affecting the air supply change amount of the air supply system; the influencing factor characteristic data includes coal supply amount, oxygen amount and boiler bed temperature, etc.

[0050] Step S3, input the influencing factor characteristic data into the air supply system prediction model to obtain the air supply change corresponding to the factor characteristic data; the air supply system prediction model is a neural network model established based on the air supply system and trained according to the influencing factor characteristic data and the actual air supply change.

[0051] Step S4, input the air supply change into the relationship model between the air supply system and the induced draft system to obtain the first induced draft change; the relationship model between the air supply system and the induced draft system is a neural network model established based on the air supply system and the induced draft system, and trained according to the air supply change and the actual induced draft change.

[0052] Step S5: determining the second induced draft change under the control of the PID controller according to the furnace outlet pressure data.

[0053] Step S6: determining a final induced air change amount according to the first induced air change amount and the second induced air change amount; the final induced air change amount is the induced air change amount that the induced air system currently needs to adjust.

[0054] In this embodiment, before inputting the influencing factor characteristic data into the air supply system prediction model in step S3 to obtain the air supply change corresponding to the factor characteristic data, the circulating fluidized bed boiler furnace outlet pressure control optimization method also includes the following steps.

[0055] An air supply system prediction model is established, and the coal quantity, oxygen quantity and boiler bed temperature are used as inputs, and the corresponding actual air supply change amount is used as output, and the air supply system prediction model is trained.

[0056] In this embodiment, the air supply system prediction model is a TCN model, an LSTM model, or a Transformer model.

[0057] In this embodiment, before inputting the supply air variation into the relationship model between the supply air system and the induced draft system to obtain the first induced draft variation in step S4, the circulating fluidized bed boiler furnace outlet pressure control optimization method further includes the following steps.

[0058] A relationship model between the air supply system and the induced draft system is established, and the air supply change is used as input and the corresponding actual induced draft change is used as output, and the relationship model between the air supply system and the induced draft system is trained.

[0059] In this embodiment, the relationship model between the air supply system and the induced air system is a linear model, a SVM model or a GBDT model.

[0060] like Figure 2 As shown, in this embodiment, step S6 determines the final induced air change amount according to the first induced air change amount and the second induced air change amount, and specifically includes the following steps.

[0061] The first induced air change amount and the second induced air change amount are superimposed and summed to obtain a final induced air change amount.

[0062] In this embodiment, after the step of summing the first induced draft variation and the second induced draft variation to obtain a final induced draft variation, the circulating fluidized bed boiler furnace outlet pressure control optimization method further includes the following steps.

[0063] According to the final induced draft change, the rotation speed of the induced draft fan in the induced draft system is adjusted to keep the furnace outlet pressure stable.

[0064] In order to make the technical solution of this embodiment clearer, the specific implementation process of the technical solution is described below.

[0065] (1) Construct a prediction model for the air supply system.

[0066] In this embodiment, when constructing the air supply system prediction model, data cleaning is required first. During data cleaning, the experience of relevant experts can be combined to preliminarily determine the characteristics related to the air supply system from a mechanism perspective, such as the total coal supply, main steam outlet flow, oxygen content, etc. In addition, the data of industrial scenes fluctuates to a certain extent due to the accuracy of equipment collection instruments or the data itself, and it is necessary to use certain filtering techniques to smooth the data.

[0067] In this embodiment, since the boiler has to be shut down for maintenance at regular intervals, the continuous data is generally one month's data, two months' data, or three months' data. This embodiment takes two months as an example, and the data selects the actual production data of the last two months. One piece of data is generated every 5 seconds, and there are 86,400 pieces of data every day. It is mainly for continuous data (such as total coal supply, main steam outlet flow, and oxygen content, etc.), and the data type is floating point data. Industrial scenarios often use the first-order lag filtering method to operate. For the data continuously generated in these months, smoothing filtering operations are performed in sequence from far to near. The specific filtering coefficient is determined according to the fluctuation of each feature combined with the timing diagram. After cleaning the data, data with noise fluctuations eliminated is obtained.

[0068] When performing data analysis, this embodiment analyzes the characteristics of factors that affect the air supply volume (including primary air volume, secondary air volume and return air volume) of the air supply system based on historical operating data combined with expert experience, and statistically analyzes the characteristics of each factor separately, checks the distribution of the historical values ​​of each characteristic, determines whether there are characteristics with unchanged values ​​or characteristics with abnormal data, and eliminates these characteristics. Perform a correlation analysis on the remaining factor characteristics and the actual values ​​of the primary air volume, secondary air volume, return air volume, etc., and check from a data-driven perspective whether there is a correlation between each factor characteristic and the primary air volume, secondary air volume, and return air volume. For factor characteristics with large time lags, they can be judged in conjunction with the timing diagram. Among them, the Pearson correlation coefficient can be used to judge the correlation. If it is uncertain whether there is a time lag, the correlation between each characteristic and the change in air supply can be determined after traversing the offset for multiple steps.

[0069] In this embodiment, for the air supply system of the circulating fluidized bed boiler, the primary air refers to the air sent into the furnace from below the air distribution plate to fully fluidize the material layer, control the boiler bed temperature and ensure the wind required for initial fluidized combustion. Usually, the gas coming out of the cyclone separator is pressurized by the powder discharger and then blown into the furnace together with the coal powder output by the powder feeder. This part of the gas is generally called primary air. The main function of the primary air is to transport coal powder. The secondary air mainly refers to the combustion-supporting air that does not participate in the fluidization at the bottom of the bed layer and is used for the combustion of material particles in the dense phase area and the suspended section. The return air is the source of power to push the high-temperature material separated by the circulating ash cyclone separator in the circulating fluidized bed boiler back to the furnace. It plays the role of conveying and fluidizing the circulating materials on the ash falling side below the separator material leg and the return material inclined leg side corresponding to the self-balancing return device.

[0070] In the time lag analysis of this embodiment, it is determined when the air supply system starts to change when the characteristics of various factors change. This part can be manually judged through a timing diagram or given through an optimization algorithm. The loss function of the optimization algorithm is the mean square error.

[0071] The prediction model of the air supply system to be constructed in this embodiment has a loss function as the mean of the square of the difference between the predicted value and the actual value. In order to obtain the predicted value, it is necessary to establish an air supply system prediction model. The core of the model is the model hyperparameters. For example, the hyperparameters of the DMC (Dynamic Matrix Control) prediction model are the amplification factor, time lag and time constant. The values ​​of these three parameters can be determined by an optimization algorithm. The purpose is to minimize the loss function by determining the values ​​of these three parameters. Taking the total coal feed as an example, when the total coal feed changes and other characteristics do not change, it is necessary to determine the change in the secondary air volume. When the total coal feed changes, the time lag time of the secondary air volume refers to when the secondary air volume begins to change, and the reaction time of the secondary air volume refers to when the secondary air volume completes the change. In this way, a time series diagram of the change in the total coal feed and the change in the secondary air volume can be drawn, and points can be discretized for it, such as Figure 3 shown.

[0072] like Figure 3 As shown in the figure, when the total coal supply changes by 1 unit and other characteristics remain unchanged, how should the secondary air volume change? Figure 3 Where t1 is the time lag, t2 is the reaction time, and k is the method coefficient. When the above three parameters are determined, the output change trend graph corresponding to the input change at a certain moment can be drawn, and the output change trend graph at each moment is superimposed to obtain the predicted value. The sum of the squares of the difference between it and the actual value y is the loss function. Different three parameters will result in different predicted values. Find the one that makes the predicted value The set of parameters (amplification factor, time lag and time constant) closest to the actual value will affect the input length of subsequent modeling. If the input length is less than the sum of the time lag and three times the time constant, it means that a certain moment in the past has an impact on the future prediction, but it is not taken into account, resulting in inaccurate prediction.

[0073] In this embodiment, by Figure 3 The impact time can be observed from the timing diagram in , or the impact time can be obtained by analyzing the parameters optimized based on the DMC model through historical operation data. For the three parameters of the randomly given DMC, the Figure 3 The unit step signal input in corresponds to the first-order linear system output diagram, which can determine the predicted value You can also determine the actual value y, or optimize the above three parameters in the direction of the gradient, or use a heuristic algorithm to determine the above three parameters.

[0074] In this embodiment, through the above data analysis combined with expert experience, the various factors affecting the primary air volume, secondary air volume and return air volume are determined, including the total coal supply, the actual value of the oxygen content, the target value of the oxygen content and the boiler bed temperature, as well as the time lag and reaction time of each feature on the air supply system. The above information is used as the input of the air supply system prediction model. The time lag and reaction time will affect the input length of the air supply system prediction model, and the relevant influencing factors will affect the number of input features of the air supply system prediction model. Here, it is mainly to determine each useful feature, and the time lag and reaction time of each feature on the air supply system. Assuming that the changes of a certain feature in the first 50 moments have an impact on the air supply volume of the air supply system, but if only the data of the first 30 moments of the feature are taken, the final prediction result will be inaccurate. This embodiment uses a timing diagram and DMC parameter optimization method to determine the changes of the feature in the first 50 moments, rather than only the changes in the first 30 moments, to ensure the accuracy of the prediction.

[0075] When constructing the model in this embodiment, the preliminary features of the input model and the time lag of the impact of each feature on the air supply system are obtained through the above analysis process, and a suitable time series prediction model is selected, such as TCN (Temporal Convolutional Network), LSTM (Long Short-Term Memory), Transformer, etc. The input features of the model may include the total coal supply, the main steam outlet flow, the oxygen content and the oxygen content target value, etc. The output is the air supply change, which includes the primary air volume, the secondary air volume and the return air volume. The relationship between the input features and the output results is analyzed by modeling analysis. The input length of the model is the time lag plus three times the time constant. The features of the model input are the features determined by the above data analysis process, and the input is constructed into an m*n matrix. Based on the input length of the model, the number of neurons in the model input layer is determined, and then the number of convolution layers, the number of bidirectional or unidirectional LSTM layers, and the number of encoders and decoders are constructed.

[0076] In this embodiment, the Transformer model is taken as an example and used as the prediction model of the air supply system. The traditional Transformer model has 6 encoders and 6 decoders, wherein the encoder includes a multi-head self-attention layer, a feedforward neural network, a residual connection and a normalization layer and a position encoding, and the decoder includes a multi-head self-attention layer, an encoder-decoder attention layer, a feedforward neural network, a residual connection and a normalization layer and a position encoding. Alternatively, the structure of the Transformer model can be improved to 2 encoders and 2 decoders, and this embodiment improves the structure of the decoder by adopting a GRU structure, such as Figure 4 As shown, the encoder structure is a traditional Transformer encoder structure. Considering the data magnitude of industrial scenarios and other factors, this embodiment improves 6 encoders to 2. Figure 4In , B represents the number of samples in each batch, T represents the length of the sample sequence, and N represents the number of features. The decoder structure is different from the traditional Transformer decoder structure. A GRU (Gated Recurrent Unit) module is added. The first decoder inputs the result of the GRU module into the multi-head attention part, sums its output, and then concatenates it with the data of the last step of the GRU module (cat), followed by a fully connected layer (Dense). The second decoder takes the output of the GRU module of the first decoder as input. The operation is the same as the first decoder. When doing the concatenation operation, the output of the fully connected layer of the first decoder is added, and then the three parts are concatenated and enter the fully connected layer. The output is the predicted output of the induced air system. Through the air supply system prediction model, the change trend of the air supply system (primary air, secondary air, and return air) in the future can be predicted. The prediction result will be needed after the relationship model between the air supply system and the induced air system is constructed.

[0077] In this embodiment, the goal of the air supply system prediction model is to minimize the loss function, and the loss function L1 is as shown in formula (1).

[0078]

[0079] Among them, L1 represents the loss function of the air supply system prediction model, is the predicted value of the air supply change, and y1 is the actual value of the air supply change.

[0080] (2) Construct a relationship model between the supply air system and the induced air system.

[0081] In this embodiment, the furnace outlet pressure needs to be stable, so the inlet and outlet air volume, that is, the supply air volume and the induced air volume, needs to be balanced. The air supply system determines how the supply air (inlet air) should change according to the combustion conditions. In order to maintain the furnace outlet pressure stable, the induced air (outlet air) needs to change accordingly. Regarding the correlation between the air supply system and the induced air system, the prediction results output by the prediction model of the air supply system are mainly used, that is, the supply air change (including primary air volume, secondary air volume, and return air volume) in the future is known to determine the induced air change to ensure the stability of the negative pressure.

[0082] In this embodiment, first, when cleaning the data, combined with the experience of relevant experts, the characteristics of primary air volume, secondary air volume, return air volume, induced draft air volume, load and negative pressure are determined. In the filtering operation, it is necessary to ensure that the filtering method and coefficient of the same feature are consistent in the entire project. Then, according to the time of the historical operation data, it is processed from far to near, and for a certain continuous period of time data (primary air volume, secondary air volume, return air volume, induced draft air volume, and load have not changed, and last for a certain time, such as 10 minutes, etc.), the negative pressure is averaged to construct a sample point. M data points are constructed, and the difference of the sample points of continuous time is taken (the load data does not take the difference), and N sample points are constructed. According to the above steps, N sample points are determined, and the characteristics are the change in primary air volume, the change in secondary air volume, the change in return air volume, the change in induced draft air volume, the change in load and the change in negative pressure.

[0083] In this embodiment, the modeling analysis is used to determine the changes in the primary air volume, secondary air volume, and return air volume under a certain load, and how the induced draft air volume should change at this time in order to ensure the stability of the negative pressure. Specifically, for the above processed samples, the data with constant negative pressure (within a certain range, such as the data corresponding to the negative pressure interval of -20Pa to 20Pa) is screened out. For this part of the data, the modeling analysis is performed to determine the changes in the primary air volume, secondary air volume, and return air volume under a certain load, and the changes in the induced draft air volume are further determined when the load is stable.

[0084] In this embodiment, through machine learning models, such as linear models, SVM (Support Vector Machine) or GBDT (GradientBoosting Decision Tree) models, the input of the model is mainly the change in supply air, specifically the change in load, primary air volume, secondary air volume and return air volume, and the output is the change in induced air. The goal of the relationship model between the supply air system and the induced air system is to minimize the loss function, and the loss function L2 is as shown in formula (2).

[0085]

[0086] Among them, L2 represents the loss function of the relationship model between the air supply system and the induced air system, is the predicted value of the induced draft change, and y2 is the actual value of the induced draft change.

[0087] In this embodiment, the changes in the primary air volume, the secondary air volume and the return air volume under a certain load are analyzed through historical operation data, and the changes in the induced air volume are determined to ensure the stability of the negative pressure. Then, the change in the supply air volume of the air supply system predicted by the air supply system prediction model is input into the relationship model between the air supply system and the induced air system to calculate the change in the induced air volume of the induced air system at the next moment. The change in the induced air volume here corresponds to the adjustment amount of the induced air system.

[0088] This embodiment takes into account the situation that when the characteristics in (1) and (2) have not changed, it is considered that the air supply and induced air systems maintain the status quo and the negative pressure can be stable. However, the actual situation is not the case. Often, the characteristics collected by the system have not changed, and the air supply and induced air volumes have not changed, but the negative pressure slowly deviates from the target range over time. Therefore, this embodiment requires a PID controller to ensure that the negative pressure is stable when the negative pressure changes due to changes in factors that have not been collected (such as raw material quality, air environment, etc.).

[0089] (3) Parameter optimization of PID control model.

[0090] This embodiment takes into account that regardless of whether the air supply system changes, the negative pressure may become too large or too small over time. Therefore, this embodiment is used in conjunction with a PID controller. The coefficients before the proportional term, differential term, and integral term of a traditional PID controller are adjusted on-site. This embodiment aims to obtain the above coefficients by modeling the historical operating data. The modeling and analysis method initializes the proportional coefficient of the PID controller, that is, a linear model is used to fit the relationship between the change in negative pressure and the change in induced draft. Specifically, for the above-mentioned processed samples, the data of the unchanged air supply system is screened out, and this part of the data is modeled and analyzed to determine the change in negative pressure and the change in induced draft. Such a relationship is fitted by linear regression, assuming that the function is Input x is the negative pressure change (error term in PID), output is the change in induced air flow (the output term in PID, i.e. the regulated quantity). The goal of the PID control model is to minimize the loss function, and the loss function L3 is as shown in formula (3).

[0091]

[0092] Among them, L3 represents the loss function of the PID control model, is the predicted value of the induced draft change under PID control, and y3 is the actual value of the induced draft change under PID control. We try to find the value of a that minimizes the loss function. From the perspective of machine learning, a is a parameter that needs to be trained. The formula is the same as the proportional link in the PID expression. According to the mapping relationship y=ax, a is used as the proportional coefficient of the PID controller, and the coefficients of the integral link and the differential link can be set to 0.

[0093] In another embodiment, a circulating fluidized bed boiler furnace outlet pressure control optimization system can also be provided, wherein the circulating fluidized bed boiler comprises a furnace, an induced draft system, an air supply system and a PID controller, wherein the air outlet of the air supply system and the air inlet of the induced draft system are both connected to the interior of the furnace, such as Figure 5 As shown, the PID controller is connected to the induced draft system and the air supply system respectively, and the circulating fluidized bed boiler furnace outlet pressure control optimization system includes the following functional modules.

[0094] The historical operation data acquisition module M1 is used to acquire the historical operation data of the circulating fluidized bed boiler; the historical operation data is the equipment operation data of the circulating fluidized bed boiler that has been running continuously in the past period of time.

[0095] The data determination module M2 is used to determine the influencing factor characteristic data and the furnace outlet pressure data according to the historical operation data; the influencing factor characteristic data is the data corresponding to various factors affecting the air supply change amount of the air supply system; the influencing factor characteristic data includes coal supply amount, oxygen amount and boiler bed temperature.

[0096] The air supply change acquisition module M3 is used to input the influencing factor characteristic data into the air supply system prediction model to obtain the air supply change corresponding to the factor characteristic data; the air supply system prediction model is a neural network model established based on the air supply system and trained according to the influencing factor characteristic data and the actual air supply change.

[0097] The first induced draft change acquisition module M4 is used to input the supply air change into the relationship model between the supply air system and the induced draft system to obtain the first induced draft change; the relationship model between the supply air system and the induced draft system is a neural network model established based on the supply air system and the induced draft system, and trained according to the supply air change and the actual induced draft change.

[0098] The second induced draft variation acquisition module M5 is used to determine the second induced draft variation under the control of the PID controller according to the furnace outlet pressure data.

[0099] The final induced draft change amount determination module M6 is used to determine the final induced draft change amount according to the first induced draft change amount and the second induced draft change amount; the final induced draft change amount is the induced draft change amount that the induced draft system currently needs to adjust.

[0100] In this embodiment, by combining the prediction model of the air supply system and the relationship model between the air supply system and the induced draft system, it is possible to determine the air supply change of the air supply system in a certain working condition in the future, and then how the induced draft change of the induced draft system should change to ensure the stability of the negative pressure. Due to the complexity of the on-site working conditions and the existence of some important factors that affect the negative pressure information and cannot be collected, such as coal quality, negative pressure fluctuations may still occur when the working conditions remain unchanged. At this time, it is necessary to use the PID controller in time to adjust it. That is to say, on the one hand, this embodiment uses the prediction model of the air supply system and the relationship model of the air supply system and the induced draft system to predict the induced draft change of the induced draft system based on the collected features. On the other hand, it also uses the PID controller to control the influence of the uncollected features on the negative pressure. It is necessary to adjust the induced draft system to ensure the stability of the negative pressure. Therefore, it is necessary to superimpose and sum the two induced draft change results output from both aspects to obtain the required adjustment amount of the final induced draft system, such as Figure 2 Assume that the first induced air change result output by the relationship model between the air supply system and the induced air system is OUT1, and the second induced air change result output by the PID controller is OUT2, then the final required adjustment amount of the induced air system is the sum of OUT1 and OUT2.

[0101] In the circulating fluidized bed boiler furnace port pressure control method of this embodiment, the air supply system prediction model and the air supply system and induced draft system relationship model are constructed through statistical analysis of historical operation data. When it is predicted that the air supply prediction system is about to change, such as changes in primary air, secondary air, and return air, the change amount of the induced draft system is predicted in combination with the air supply system and induced draft system relationship model, and the induced draft system changes synchronously with the air supply system, aiming to ensure the stability of the furnace outlet pressure target range.

[0102] At present, the furnace outlet pressure control technology of circulating fluidized bed boiler is relatively mature, and the time lag between the furnace outlet pressure and the air supply and induced draft system is small. The general method is to use a PID controller to adjust the speed frequency of the induced draft fan. When the furnace outlet pressure is high, the speed frequency of the induced draft fan is increased to reduce the furnace outlet pressure; when the furnace outlet pressure is low, the speed frequency of the induced draft fan is reduced to increase the furnace outlet pressure. Although this furnace outlet pressure control method based on the PID controller has a rapid control response, its control mode is relatively single, and only when the furnace outlet pressure has been high or low, the PID controller will take action to control. Therefore, this method cannot perceive the changes in the air supply system in advance, and cannot effectively avoid the problem of high or low furnace outlet pressure caused by changes in working conditions or interference from external factors. This embodiment analyzes and models the historical operation data, perceives the changes in the air supply system in advance, and synchronizes the induced draft system in advance, aiming to reduce or eliminate the situation where the furnace outlet pressure is high or low due to changes in working conditions or interference from external factors. The PID control strategy in the general technology is changed from the main control strategy and the only control strategy to the auxiliary control strategy and the bottom-up strategy.

[0103] This embodiment proposes a circulating fluidized bed boiler furnace outlet pressure control optimization method, firstly, analyze the historical operation data to determine the factors that affect the air supply system (including time lag and reaction time); then establish an air supply system prediction model, based on the useful influencing factors that have been determined, take the useful factors from the previous k moments to the current moment as input, and the air supply system values ​​of the next n times as output, train the model, and obtain the air supply system prediction model; then, in order to ensure the stability of the furnace outlet pressure, predict how much the future air supply will change, then the induced draft also needs to change, and the specific change is given by the relationship model between the air supply system and the induced draft system analyzed from the historical operation data. That is, the first induced draft change OUT1 of the induced draft system at the current moment is output; then, when the air supply system and the induced draft system remain unchanged, the furnace outlet pressure will also change over time, so a PID controller is needed to handle this part of the impact. For example, changes in the quality of raw materials, changes in air quality, or equipment problems are factors that cannot be collected or are not determined by expert experience and data analysis. In other words, there may occasionally be characteristic influences that affect the furnace outlet pressure but for which there is no data. At this time, the PID controller is used to output the second induced draft change OUT2; then the first induced draft change OUT1 output by the relationship model of the air supply system and the induced draft system and the second induced draft change OUT2 output by the PID controller are summed to obtain the final induced draft change OUT of the induced draft system, that is, the adjustment amount of the induced draft fan at the current moment.

[0104] This embodiment uses historical operation data for analysis and modeling to construct a prediction model for the air supply system, an induced draft system and an air supply system relationship model, and uses a PID controller to jointly determine the induced draft change of the induced draft system. When controlling the furnace outlet pressure of a circulating fluidized bed boiler, the control of the low-delay system is introduced into the predictive control to completely avoid the low-delay, eliminate the situation where the furnace outlet pressure deviates from the target range in a short period of time and then returns to normal, minimize or eliminate the situation where the furnace outlet pressure is high or low due to changes in operating conditions or interference from external factors, and improve the stability of the furnace outlet pressure and its proportion in the target range.

[0105] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0106] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0107] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A circulating fluidized bed boiler furnace outlet pressure control optimization method, the circulating fluidized bed boiler comprises a furnace, an induced draft system, an air supply system and a PID controller, the air outlet of the air supply system and the air inlet of the induced draft system are both connected to the interior of the furnace, the PID controller is connected to the induced draft system and the air supply system respectively, characterized in that: The circulating fluidized bed boiler furnace outlet pressure control optimization method comprises: Acquire historical operation data of the circulating fluidized bed boiler; the historical operation data is the equipment operation data of the circulating fluidized bed boiler that has been continuously operated in the past period of time; Determine the influencing factor characteristic data and furnace outlet pressure data according to the historical operation data; the influencing factor characteristic data is data corresponding to various factors affecting the air supply change amount of the air supply system; the influencing factor characteristic data includes coal supply amount, oxygen amount and boiler bed temperature; The influencing factor characteristic data is input into the air supply system prediction model to obtain the air supply change amount corresponding to the factor characteristic data; the air supply system prediction model is a neural network model established based on the air supply system and trained according to the influencing factor characteristic data and the actual air supply change amount; The air supply variation is input into the air supply system and induced draft system relationship model to obtain a first induced draft variation; the air supply system and induced draft system relationship model is a neural network model established based on the air supply system and the induced draft system and trained according to the air supply variation and the actual induced draft variation; Determining a second induced draft change under the control of the PID controller according to the furnace outlet pressure data; A final induced air change amount is determined according to the first induced air change amount and the second induced air change amount; the final induced air change amount is the induced air change amount that the induced air system currently needs to adjust.

2. The circulating fluidized bed boiler furnace outlet pressure control optimization method according to claim 1, characterized in that: Before the step of inputting the influencing factor characteristic data into the air supply system prediction model to obtain the air supply change amount corresponding to the factor characteristic data, the circulating fluidized bed boiler furnace outlet pressure control optimization method further includes: An air supply system prediction model is established, and the coal supply amount, oxygen amount and boiler bed temperature are used as inputs, and the corresponding actual air supply change amount is used as output, and the air supply system prediction model is trained.

3. The circulating fluidized bed boiler furnace outlet pressure control optimization method according to claim 1, characterized in that: The air supply system prediction model is a TCN model, an LSTM model or a Transformer model.

4. The circulating fluidized bed boiler furnace outlet pressure control optimization method according to claim 1, characterized in that: The loss function of the air supply system prediction model is: Wherein, L1 represents the loss function of the air supply system prediction model, is the predicted value of the air supply change, and y1 is the actual value of the air supply change.

5. The circulating fluidized bed boiler furnace outlet pressure control optimization method according to claim 1, characterized in that: Before the step of inputting the air supply variation into the air supply system and induced draft system relationship model to obtain the first induced draft variation, the circulating fluidized bed boiler furnace outlet pressure control optimization method further includes: A relationship model between the air supply system and the induced draft system is established, and the air supply change is used as input and the corresponding actual induced draft change is used as output, and the relationship model between the air supply system and the induced draft system is trained.

6. The circulating fluidized bed boiler furnace outlet pressure control optimization method according to claim 1, characterized in that: The relationship model between the air supply system and the induced air system is a linear model, a SVM model or a GBDT model.

7. The circulating fluidized bed boiler furnace outlet pressure control optimization method according to claim 1, characterized in that: The loss function of the relationship model between the air supply system and the induced draft system is: Wherein, L2 represents the loss function of the relationship model between the air supply system and the induced air system, is the predicted value of the induced draft change, and y2 is the actual value of the induced draft change.

8. The circulating fluidized bed boiler furnace outlet pressure control optimization method according to claim 1, characterized in that: Determining a final induced air change amount according to the first induced air change amount and the second induced air change amount specifically includes: The first induced air change amount and the second induced air change amount are summed to obtain a final induced air change amount.

9. The circulating fluidized bed boiler furnace outlet pressure control optimization method according to claim 8, characterized in that: After the step of summing the first induced draft variation and the second induced draft variation to obtain a final induced draft variation, the circulating fluidized bed boiler furnace outlet pressure control optimization method further includes: According to the final induced draft change, the rotation speed of the induced draft fan in the induced draft system is adjusted to keep the furnace outlet pressure stable.

10. A circulating fluidized bed boiler furnace outlet pressure control optimization system, the circulating fluidized bed boiler comprises a furnace, an induced draft system, an air supply system and a PID controller, the air outlet of the air supply system and the air inlet of the induced draft system are both connected to the interior of the furnace, the PID controller is connected to the induced draft system and the air supply system respectively, characterized in that: The circulating fluidized bed boiler furnace outlet pressure control optimization system comprises: A historical operation data acquisition module is used to acquire the historical operation data of the circulating fluidized bed boiler; the historical operation data is the equipment operation data of the circulating fluidized bed boiler that has been continuously operated in the past period of time; A data determination module, for determining the influencing factor characteristic data and the furnace outlet pressure data according to the historical operation data; the influencing factor characteristic data is data corresponding to various factors affecting the air supply change amount of the air supply system; the influencing factor characteristic data includes coal supply amount, oxygen amount and boiler bed temperature; An air supply change acquisition module is used to input the influencing factor characteristic data into the air supply system prediction model to obtain the air supply change corresponding to the factor characteristic data; the air supply system prediction model is a neural network model established based on the air supply system and trained according to the influencing factor characteristic data and the actual air supply change; A first induced air change amount acquisition module is used to input the supply air change amount into the relationship model between the supply air system and the induced air system to obtain a first induced air change amount; the relationship model between the supply air system and the induced air system is a neural network model established based on the supply air system and the induced air system and trained according to the supply air change amount and the actual induced air change amount; A second induced draft change amount acquisition module, used to determine the second induced draft change amount under the control of the PID controller according to the furnace outlet pressure data; The final induced air change amount determination module is used to determine the final induced air change amount according to the first induced air change amount and the second induced air change amount; the final induced air change amount is the induced air change amount that the induced air system currently needs to adjust.

Images