Circulating fluidized bed temperature intelligent prediction method based on physical information neural network

Abstract

The invention discloses a circulating fluidized bed temperature intelligent prediction method based on a physical information neural network, and mainly relates to the technical field of industrial process intelligent control and energy power engineering crossing. Comprising the following steps: S1, acquiring operation data of the circulating fluidized bed, and preprocessing the operation data to obtain a data set; s2, constructing a physical information residual neural network; s3, constructing a loss function of adaptive weight adjustment; s4, training the physical information residual neural network by using the data set, and performing iterative optimization in combination with the constructed physical information residual neural network and the loss function to obtain a bed temperature prediction model; s5, inputting operation data of the circulating fluidized bed into the bed temperature prediction model to obtain a bed temperature prediction value; the method can solve the problem of complexity and nonlinearity of dynamic prediction of the bed temperature of the circulating fluidized bed, can optimize the operation efficiency of the boiler, improves the combustion stability, and reduces the emission of pollutants.

Description

Technical Field

[0001] The present invention relates to the cross - technical field of industrial process intelligent control and energy power engineering, and specifically to an intelligent prediction method for the bed temperature of a circulating fluidized bed based on a physics - informed neural network. Background Art

[0002] As an efficient and clean combustion device, the circulating fluidized bed (CFB) boiler is widely used in fields such as electric power, chemical industry, and heat energy supply. Its core advantages lie in enabling efficient fuel combustion, low pollutant emissions, and good fuel adaptability. During the operation of a CFB boiler, the precise control of the bed temperature is a key factor affecting combustion efficiency, heat transfer efficiency, and pollutant generation. Therefore, the bed temperature control of a CFB boiler is particularly important.

[0003] Currently, the bed temperature control of a CFB boiler mainly relies on traditional control strategies, such as experience - based PID control or model - based predictive control. Although these methods can achieve stable control of the bed temperature to a certain extent, due to the complex physical and chemical processes inside the CFB boiler (such as fuel combustion, heat transfer, gas - solid flow, etc.), traditional control methods are difficult to accurately predict the dynamic changes of the bed temperature, especially under large - scale operating condition changes. In addition, traditional methods have high requirements for the accuracy of the model and the stability of parameters, and are difficult to adapt to complex non - linear systems.

[0004] With the development of industrial big data technology, data - driven machine learning methods have gradually been applied to the bed temperature prediction of CFB boilers. For example, methods such as regression analysis based on historical data, support vector machines, or artificial neural networks are used to establish bed temperature prediction models. However, these methods mainly rely on a large amount of historical data, have high requirements for the quality and quantity of data, and lack in - depth understanding and interpretability of physical processes. When the data is insufficient or the operating conditions change, the generalization ability and prediction accuracy of data - driven models will decrease significantly.

[0005] On the other hand, there are also modeling methods based on physical principles, such as thermodynamic equations and fluid mechanics equations, which can more accurately describe the physical processes of CFB boilers. However, these models are usually too complex to solve, and are difficult to meet the requirements of real - time prediction. In addition, the parameter identification and calibration processes of physical models are complex and have poor adaptability to actual operations.

[0006] In summary, although certain progress has been made in the existing technology for CFB boiler bed temperature control, the following key problems still exist: When traditional control methods and data-driven methods handle the complex non-linear dynamic characteristics of CFB boilers, the prediction accuracy is limited, making it difficult to meet the requirements of high-precision bed temperature control; data-driven models are highly dependent on data. When the operating conditions change or the data is insufficient, the model generalization ability is insufficient, resulting in a decline in prediction performance; pure data-driven methods are difficult to explain the prediction results of the model and cannot provide intuitive physical meanings for operators, restricting their application in actual industries; although physics-based models have high interpretability, due to the complexity of the models and slow solution speeds, it is difficult to meet the requirements of real-time prediction. Summary of the Invention

[0007] The purpose of the present invention is to solve the problems existing in the prior art. By combining artificial intelligence technology with the thermodynamic mechanism of circulating fluidized bed (CFB), an intelligent prediction method for the bed temperature of a circulating fluidized bed based on a physics-informed neural network is provided. Combining the powerful data fitting ability of deep learning algorithms and the description ability of physical models for complex physical processes, by introducing physical information constraints, the adaptability and prediction accuracy of the model to complex non-linear dynamic characteristics are enhanced, and it is used to solve the complex non-linear problems of dynamic prediction of the bed temperature of a circulating fluidized bed. At the same time, the present invention aims to solve the problems of insufficient prediction accuracy, poor generalization ability, lack of physical interpretability, and lack of real-time performance existing in the prior art, and provide an efficient and reliable solution for the intelligent control of CFB boilers, which can optimize the boiler operation efficiency, improve combustion stability, and reduce pollutant emissions.

[0008] To achieve the above object, the present invention is realized through the following technical solutions:

[0009] An intelligent prediction method for the bed temperature of a circulating fluidized bed based on a physics-informed neural network, comprising the steps of:

[0010] S1. Obtain the operation data of the circulating fluidized bed, and preprocess it to obtain a data set;

[0011] S2. Construct a physics-informed residual neural network. The physics-informed residual neural network includes an input layer, multiple hidden layers, and an output layer. The physics-informed residual neural network adopts a fully connected structure. Among them, the input layer receives the main steam flow rate, coal feeding amount, and total air volume, and the output layer outputs the predicted value of the bed temperature at the next moment;

[0012] S3. Construct a loss function with adaptive weight adjustment. The calculation formula of the loss function is:

[0013]

[0014] where L phy is the physical information loss; L datais the data-driven loss; λ is the weight coefficient, and the value range of λ is from 0 to 1. A threshold of the bed temperature change rate is set, and λ is adjusted according to the comparison result between the bed temperature change rate and the preset threshold. When the bed temperature change rate exceeds the threshold, λ is increased; otherwise, λ is decreased.

[0015] S4. Use the data set to train the physics-informed residual neural network, and perform iterative optimization by combining the constructed physics-informed residual neural network and the loss function to obtain the bed temperature prediction model.

[0016] S5. Input the operating data of the circulating fluidized bed into the bed temperature prediction model to obtain the bed temperature prediction value.

[0017] Preferably, in step S1, the operating data includes the main steam flow rate, the coal feeding rate, the total air volume, and the average bed temperature; the preprocessing process includes the steps:

[0018] S11. Perform preliminary processing on the operating data to remove outliers and missing values to obtain a preliminary processed data set.

[0019] S12. Perform normalization processing on the preliminarily processed data, and normalize the main steam flow rate, the coal feeding rate, the total air volume, and the average bed temperature to the interval [0, 1] to obtain the preprocessed data set. The normalization formula is as follows:

[0020]

[0021] where, x mim represents the minimum value in the preliminarily processed data set; x max represents the maximum value in the preliminarily processed data set; refers to the normalized data; x i is the original unprocessed data in the preliminarily processed data set;

[0022] S13. Divide the preprocessed data set into a training set and a test set, and then divide the training set into a training subset and a validation subset. Among them, the training subset is used for the training of the physics-informed residual neural network, the validation subset is used for optimizing the hyperparameters of the physics-informed residual neural network, and the test set is used for evaluating the performance of the trained physics-informed residual neural network.

[0023] Preferably, in step S13, the hyperparameters include the learning rate, the number of neurons in the hidden layer, and the initial value of the weight coefficient λ.

[0024] Preferably, in step S13, the evaluation indicators for performance evaluation include the root mean square error, the mean absolute error, and the mean absolute percentage error. Among them,

[0025]

[0026] Among them, RMSE represents the root mean square error; MAE represents the mean absolute error; MAPE represents the mean absolute percentage error; n represents the number of bed temperature prediction data points; represents the predicted value; represents the target value.

[0027] Preferably, in step S2, the physics-informed residual neural network consists of five fully connected layers, where:

[0028] Layer 1: Maps the input data to a high-dimensional space to provide rich data representation;

[0029] Layer 2 and Layer 3: Introduce non-linear features through the ReLU activation function to enhance the non-linear representation ability of the model;

[0030] Layer 4: Combines the residual connection structure to combine the original input data with the output of Layer 3, accelerating convergence and retaining linear features;

[0031] Layer 5: As the output layer, maps the output of Layer 4 to the specified dimension to generate the final prediction result.

[0032] Preferably, in step S3, the calculation formula for the physics-informed loss is:

[0033]

[0034] Among them, N represents the total number of data points used in the calculation; K represents the relevant input parameters at the k-th data point; represents the actual temperature at the k-th data point; represents the predicted temperature at the k-th data point; represents the time derivative of the predicted temperature at the k-th data point; G(·) is a function obtained from network training.

[0035] Preferably, in step S3, the calculation formula for the data-driven loss is:

[0036]

[0037] Among them, N represents the total number of data points used in the calculation; represents the actual temperature at the k-th data point; represents the predicted temperature at the k-th data point.

[0038] Preferably, in step S3, based on the energy balance equation of the circulating fluidized bed boiler bed temperature, the prediction result of the model is constrained by physical laws to enhance the physical interpretability of the model. The energy balance equation is as follows:

[0039]

[0040] Among them, C s represents the specific heat capacity of the bed material; m s represents the mass of the bed material; T k represents the bed temperature to be predicted; w c represents the coal feeding rate; represents the oxygen concentration in the furnace; k air represents the calorific value of the air entering the furnace; A ir is the total air flow; T1 represents the total air temperature entering the furnace; k fl represents the flue gas heat correction coefficient; k a represents the average heat transfer coefficient of the heating surface; S a refers to the total area of the heating surface; T f is the working fluid temperature.

[0041] Preferably, in step S5, the trained bed temperature prediction model is deployed to the control system of the circulating fluidized bed boiler to receive the input of operation parameters in real time and output the bed temperature prediction value.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] By combining physical information and deep learning algorithms, the present invention significantly improves the accuracy of bed temperature prediction in circulating fluidized bed boilers and can better adapt to complex non-linear dynamic characteristics; at the same time, the design of the loss function with adaptive weight adjustment enhances the generalization ability of the model, enabling it to maintain high prediction performance under different working conditions; on the other hand, embedding physical laws into the model makes the model interpretable, facilitating understanding and practical application. In short, the present invention has high computational efficiency, can meet the requirements of real-time prediction, can provide a quick response for the intelligent control of the boiler, optimize the combustion process by accurately predicting the bed temperature, improve the operation efficiency of the boiler, and reduce pollutant emissions, with significant economic and environmental benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is the flowchart of the method of the present invention;

[0045] Figure 2 is the structural diagram of the physical information residual neural network. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms also fall within the scope defined by this application.

[0047] Example 1: As shown in the appendix Figure 1As shown in the figure, the present invention provides an intelligent prediction method for the fluidized bed temperature of a circulating fluidized bed based on a physics-informed neural network, which combines the powerful data fitting ability of a deep learning algorithm and the description ability of a physical model for complex physical processes. By introducing physical information constraints, the adaptability and prediction accuracy of the model to complex non-linear dynamic characteristics are enhanced, including the following steps:

[0048] S1. Obtain the real-time operation data of the circulating fluidized bed, including the main steam flow rate, coal feeding rate, total air volume, and the fluidized bed temperature at each point. Calculate the average fluidized bed temperature based on the fluidized bed temperature at each point, and perform normalization processing on the obtained data. The normalization method is as follows:

[0049]

[0050] where x min and x max represent the minimum value and the maximum value in the dataset respectively; refers to the normalized data; x i is the original data without processing;

[0051] S2. Provide a Physics-Informed Residual Neural Network (PIRNN), and embed physical laws into the loss function of the deep learning model. By combining the thermodynamic equation of the CFB boiler and the deep learning algorithm, the non-linear mapping relationship between the main steam flow rate, coal feeding rate, and total air volume at the previous moment and the average fluidized bed temperature at the next moment can be accurately captured to achieve high-precision prediction of the fluidized bed temperature;

[0052] This network consists of five fully connected layers. The input data is first mapped to a high-dimensional space through the first fully connected layer to provide rich data representations for subsequent feature learning; then, the data is mapped back to the dimension of the input data through the third fully connected layer, where the ReLU activation function is applied to introduce non-linear features, thereby enhancing the representation ability of the model; after that, the ResNet structure is introduced to combine the original input data with the processed data from the third fully connected layer, so as to accelerate convergence during the training process while maintaining the model's ability to capture linear features; after the ResNet structure, a new fully connected layer, namely the fourth layer, is added to further extract and combine features to prepare more abstract and high-level feature representations for the final output layer; the fifth fully connected layer serves as the output layer, mapping the output of the fourth layer to the specified dimensional space to generate the final prediction;

[0053] S3. Provide a loss function with adaptive weight adjustment, which consists of two parts: physical information loss and data loss. When the bed temperature change rate exceeds the given threshold θ, the system will exhibit strong nonlinearity. In this case, the influence of the physical process becomes more complex. Therefore, it is necessary to increase λ to emphasize the importance of physical information in the model, thereby improving the model's ability to simulate complex nonlinear behaviors; otherwise, reduce λ. The calculation formula for the total loss is as follows:

[0054]

[0055] Among them, L phy represents the physical information loss, enabling the model to learn the necessary physical information through the energy balance equation of the bed layer temperature group parameters, thereby reducing the dependence on data. The calculation formula is as follows:

[0056]

[0057] Among them, N represents the total number of data points used in the calculation; K represents the relevant input parameters at the k-th data point; represents the actual temperature at the k-th data point; represents the predicted temperature at the k-th data point; represents the time derivative of the predicted temperature at the k-th data point; G(·) is a function obtained from network training.

[0058] The data loss term L data directly compares the predicted state and the true state

[0059]

[0060] Among them, N represents the total number of data points used in the calculation; represents the actual temperature at the k-th data point; represents the predicted temperature at the k-th data point.

[0061] Based on the energy balance equation of the CFB boiler bed temperature, the prediction results of the model are constrained by physical laws to enhance the physical interpretability of the model. The energy balance equation is as follows:

[0062]

[0063] Among them, C s represents the specific heat capacity of the bed material; m s represents the mass of the bed material; T k represents the bed temperature to be predicted; w c represents the coal feeding rate; represents the oxygen concentration in the furnace; k airRepresents the calorific value of the air entering the furnace; A ir is the total air flow rate; T1 represents the total air temperature entering the furnace; k fl represents the flue gas heat correction coefficient; k a represents the average heat transfer coefficient of the heating surface; S a refers to the total area of the heating surface; T f is the working fluid temperature.

[0064] S4. Use the processed historical operation data to train the physics-informed residual neural network. By optimizing the loss function for adaptive weight adjustment, the physics-informed residual neural network learns data features and physical laws simultaneously during the training process. After training, use the test data to verify the prediction accuracy and generalization ability of the physics-informed residual neural network to ensure its reliability under different working conditions.

[0065] S5. Deploy the trained physics-informed residual neural network to the control system of the CFB boiler, receive the input of operation parameters in real time, output the predicted value of the bed temperature, and provide decision support for the intelligent control of the boiler.

[0066] Example 2: As shown in the appendix Figure 1 The present invention provides an intelligent prediction method for the bed temperature of a circulating fluidized bed based on a physics-informed neural network, and the specific implementation method is as follows:

[0067] The first step is data collection and processing.

[0068] Collect samples from the historical operation data of a CFB boiler in a certain power plant. The data includes the main steam flow rate, coal feeding rate, total air volume, and the bed temperature at each point. The data acquisition time interval is 5 minutes, and a total of 1 month of historical data is collected, with a total of about 8500 sample points.

[0069] Then, perform preliminary processing on the obtained data to remove outliers and missing values. For missing values, use the linear interpolation method to fill them; for outliers, set reasonable thresholds to eliminate them. Then calculate the average temperature using the bed temperature at each point.

[0070] Normalize the main steam flow rate, coal feeding rate, total air volume, and the calculated average bed temperature to the interval [0, 1] to improve the training efficiency and prediction accuracy of the model. The normalization formula is as follows:

[0071]

[0072] Among them, x min and x max respectively represent the minimum value and the maximum value in the dataset; refers to the normalized data; x i is the original data without processing.

[0073] Finally, the dataset is divided into a training set and a test set at a ratio of 8:1. The training set is further divided into a training subset and a validation subset at a ratio of 9:1. The training subset, validation subset, and test set are used for training the physics-informed residual neural network, hyperparameter optimization, and performance evaluation, respectively.

[0074] In the second step, a physics-informed residual neural network (PIRNN) is constructed.

[0075] The physics-informed residual neural network (PIRNN) adopts a fully connected structure, as Figure 2 shown, including an input layer, four hidden layers, and an output layer. The input layer receives three parameters: main steam flow rate, coal feeding rate, and total air volume. The output layer outputs the predicted value of the bed temperature at the next moment. Table 1 shows the detailed configuration of the PIRNN.

[0076] Table 1 PIRNN Configuration Table

[0077]

[0078] In the third step, a loss function is constructed, and the physics-informed residual neural network (PIRNN) is trained and optimized.

[0079] The thermodynamic equation of the CFB boiler is embedded into the loss function as the calculation basis for the physics-informed loss. The specific equation is as follows:

[0080]

[0081] An adaptive weight coefficient λ is set, and the value of λ is dynamically adjusted according to whether the bed temperature change rate exceeds a given threshold. When the bed temperature change rate exceeds the threshold, the value of λ is increased to emphasize the importance of the physics information; otherwise, the value of λ is decreased.

[0082] The training subset data is used to train the physics-informed residual neural network (PIRNN), and the network parameters are updated by optimizing the loss function with adaptive weight adjustment. The total loss calculation formula of the loss function is:

[0083]

[0084] Among them, the data loss is calculated using the mean square error:

[0085]

[0086] The Adam optimizer is used for gradient descent, the learning rate is set to 0.01, and the number of training iterations is 500 times.

[0087] Optimize the hyperparameters of the physics-informed residual neural network (such as learning rate, number of neurons in the hidden layer, initial value of the weight coefficient λ) using the validation subset data to improve the generalization ability of the model.

[0088] Use the test set data to evaluate the performance of the trained physics-informed residual neural network. The main evaluation metrics include root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE), which are calculated as follows:

[0089]

[0090]

[0091] where n represents the number of bed temperature prediction data points; represents the predicted value; represents the target value.

[0092] On the local computer, use the test set data to verify the physics-informed residual neural network. During the verification process, the model inputs parameters such as main steam flow rate, pulverized coal feeding rate, and total air volume in the test set, outputs the predicted bed temperature value, and compares it with the actual bed temperature value. By calculating the error between the predicted value and the true value, evaluate the prediction performance of the model.

[0093] The results show that the physics-informed residual neural network has a high prediction accuracy on the test set, can effectively capture the dynamic changes of the bed temperature, and predict the rising trend of the bed temperature in advance. The excellent performance of the physics-informed residual neural network indicates its potential for application in the actual industrial environment. By deploying the physics-informed residual neural network into the control system of the CFB boiler, real-time intelligent prediction of the bed temperature can be achieved, providing decision-making support for the optimized operation of the boiler. By accurately predicting the bed temperature, the physics-informed residual neural network model helps to optimize the combustion process, improve the boiler operation efficiency, and reduce pollutant emissions, which can not only bring significant environmental benefits, but also improve energy utilization efficiency and reduce operating costs, bringing economic benefits.

Claims

1. An intelligent prediction method for the fluidized bed temperature of a circulating fluidized bed based on a physics-informed neural network, characterized in that, Including the steps: S1. Obtain the operation data of the circulating fluidized bed, and preprocess it to obtain a data set; S2. Construct a physics-informed residual neural network, which includes an input layer, multiple hidden layers, and an output layer. The physics-informed residual neural network adopts a fully connected structure. Among them, the input layer receives the main steam flow rate, coal feeding rate, and total air volume, and the output layer outputs the predicted value of the bed temperature at the next moment; S3. Construct a loss function with adaptive weight adjustment. The calculation formula of the loss function is: Among them, L phy is the physical information loss; L data is the data-driven loss; λ is the weight coefficient, and the value range of λ is from 0 to 1. Set the threshold of the bed temperature change rate, and adjust λ according to the comparison result between the bed temperature change rate and the preset threshold. When the bed temperature change rate exceeds the threshold, increase λ; otherwise, decrease λ. S4. Use the data set to train the physics-informed residual neural network, and perform iterative optimization by combining the constructed physics-informed residual neural network and the loss function to obtain a bed temperature prediction model; S5. Input the operation data of the circulating fluidized bed into the bed temperature prediction model to obtain the predicted value of the bed temperature.

2. The intelligent prediction method for the circulating fluidized bed bed temperature based on the physics-informed neural network according to claim 1, wherein In step S1, the operation data includes the main steam flow rate, coal feeding rate, total air volume, and average bed temperature; The preprocessing process includes the steps: S11. Perform preliminary processing on the operation data, remove outliers and missing values, and obtain a preliminarily processed data set; S12. Perform normalization processing on the preliminarily processed data, and normalize the main steam flow rate, coal feeding rate, total air volume, and average bed temperature to the interval [0, 1] to obtain a preprocessed data set. The normalization formula is as follows: Among them, x min represents the minimum value in the preliminarily processed dataset; x max represents the maximum value in the preliminarily processed dataset; refers to the data after normalization; x i is the raw data in the preliminarily processed dataset that has not been processed; S13. Divide the preprocessed data set into a training set and a test set, and then divide the training set into a training subset and a validation subset. Among them, the training subset is used for the training of the physics-informed residual neural network, the validation subset is used to optimize the hyperparameters of the physics-informed residual neural network, and the test set is used to evaluate the performance of the trained physics-informed residual neural network.

3. The intelligent prediction method for the circulating fluidized bed bed temperature based on the physics-informed neural network according to claim 2, characterized in that, In step S13, the hyperparameters include the learning rate, the number of neurons in the hidden layer, and the initial value of the weight coefficient λ.

4. The intelligent prediction method for the circulating fluidized bed bed temperature based on the physics-informed neural network according to claim 2, characterized in that, In step S13, the evaluation metrics for performance evaluation include the root mean square error, mean absolute error, and mean absolute percentage error. Among them, Among them, RMSE represents the root mean square error; MAE represents the mean absolute error; MAPE represents the mean absolute percentage error; n represents the number of bed temperature prediction data points; represents the predicted value; represents the target value.

5. The intelligent prediction method for the circulating fluidized bed bed temperature based on the physics-informed neural network according to claim 1, characterized in that, In step S2, the physics-informed residual neural network consists of five fully connected layers, where: The 1st layer: Maps the input data to a high-dimensional space to provide rich data representation; The 2nd and 3rd layers: Introduce non-linear features through the ReLU activation function to enhance the non-linear representation ability of the model; The 4th layer: Combines the residual connection structure, combines the original input data with the output of the 3rd layer, accelerates convergence, and retains linear features; The 5th layer: As the output layer, maps the output of the 4th layer to the specified dimension to generate the final prediction result.

6. The intelligent prediction method for the circulating fluidized bed bed temperature based on the physics-informed neural network according to claim 1, characterized in that, In step S3, the calculation formula of the physics-informed loss is: where N represents the total number of data points used in the calculation; K represents the relevant input parameter at the k-th data point; represents the actual temperature at the k-th data point; represents the predicted temperature at the k-th data point; represents the time derivative of the predicted temperature at the k-th data point; G(·) is a function obtained from network training.

7. The intelligent prediction method for the circulating fluidized bed bed temperature based on the physics-informed neural network according to claim 1, characterized in that, In step S3, the calculation formula of the data-driven loss is: Where N represents the total number of data points used in the calculation; represents the actual temperature of the k-th data point; represents the predicted temperature of the k-th data point.

8. The intelligent prediction method for the circulating fluidized bed bed temperature based on the physics-informed neural network according to claim 1, characterized in that In step S3, based on the energy balance equation of the bed temperature of the circulating fluidized bed boiler, the prediction result of the model is constrained by physical laws to enhance the physical interpretability of the model. The energy balance equation is as follows: Among them, C s represents the specific heat capacity of the bed material; m s represents the mass of the bed material; T k represents the bed temperature to be predicted; w c represents the coal feeding rate; represents the oxygen concentration in the furnace; k air represents the calorific value of the air entering the furnace; A ir is the total air flow rate; T1 represents the total air temperature entering the furnace; k fl represents the flue gas heat correction coefficient; k a represents the average heat transfer coefficient of the heating surface; S a refers to the total area of the heating surface; T f is the working fluid temperature.

9. The intelligent prediction method for the circulating fluidized bed bed temperature based on the physics-informed neural network according to claim 1, characterized in that, In step S5, deploy the trained bed temperature prediction model to the control system of the circulating fluidized bed boiler, receive the input of operation parameters in real time, and output the predicted value of the bed temperature.

Images