Boiler multi-region CFD-AI modeling method and system based on federated learning

Abstract

The present invention discloses a boiler multi-region CFD-AI modeling method and system based on federated learning, which belongs to the field of boiler monitoring technology. The specific steps of the modeling method are as follows: S1, simulating the heat transfer and fluid behavior of the boiler, establishing a multi-region CFD grid division, and collecting and preprocessing the dynamic operation data of each region from the actual operation of the boiler; S2, extracting the features of the preprocessed dynamic operation data, constructing a boiler multi-region CFD-AI model based on the extracted features, and globally optimizing the boiler multi-region CFD-AI model; S3, dynamically adjusting the structure and parameters of the globally optimized boiler multi-region CFD-AI model, and enhancing the key area decision-making of the boiler multi-region CFD-AI model; the present invention can improve the efficiency of model training, save a lot of communication resources, reduce the risk of privacy leakage, and promote the cross-region adaptability of the model, significantly improve the performance of the model under complex working conditions, can effectively handle multi-objective optimization problems, and can more accurately predict and control the performance of the boiler in different regions and working conditions.

Description

Technical Field

[0001] The present invention relates to the field of boiler monitoring technology, and in particular to a federated learning-based boiler multi-region CFD-AI modeling method and system. Background Art

[0002] With the continuous advancement of industrial automation and intelligence, boiler systems play a vital role in industries such as power, chemical engineering, and metallurgy. Boiler operating efficiency, stability, and emission levels directly impact production process optimization and energy conservation. Accurate dynamic modeling and prediction are essential to improving the safety and economic efficiency of boiler operation. Traditional boiler modeling methods rely on precise descriptions of physical models and control systems, but often struggle to meet high-precision requirements in complex, nonlinear, and multivariable real-world environments. In recent years, models combining computational fluid dynamics (CFD) and artificial intelligence (AI) have become a research hotspot. CFD-AI models, in particular, can effectively enhance boiler dynamic prediction capabilities. However, training CFD-AI models typically requires large amounts of high-quality data, the collection of which often faces challenges such as privacy protection and uneven data distribution. Therefore, developing a federated learning-based CFD-AI modeling method for multi-region boilers is crucial.

[0003] The existing multi-zone CFD-AI modeling method for boilers has poor model training efficiency, wastes a lot of communication resources, has a high risk of privacy leakage, and has poor cross-regional adaptability of the model; in addition, the existing multi-zone CFD-AI modeling method for boilers performs poorly under complex working conditions and cannot effectively handle multi-objective optimization problems. Summary of the Invention

[0004] The purpose of the present invention is to address the defects in the existing technology and propose a boiler multi-zone CFD-AI modeling method and system based on federated learning.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] The federated learning-based CFD-AI modeling method for multi-zone boilers includes the following steps:

[0007] S1. Simulate the heat transfer and fluid behavior of the boiler, establish a multi-region CFD grid, and collect and pre-process the dynamic operation data of each region from the actual operation of the boiler;

[0008] S2. Extracting features of the preprocessed dynamic operation data, constructing a boiler multi-zone CFD-AI model based on the extracted features, and globally optimizing the boiler multi-zone CFD-AI model;

[0009] S3. Dynamically adjust the structure and parameters of the globally optimized boiler multi-zone CFD-AI model to enhance the key regional decision-making of the boiler multi-zone CFD-AI model;

[0010] S4. Evaluate the deviation between the prediction results of the boiler multi-region CFD-AI model and the actual data, and adjust the decision path and model structure to improve the fitting effect of high-dimensional features;

[0011] S5. Repeatedly train and update the boiler multi-zone CFD-AI model, perform real-time predictions through the final model, and regularly collect operating data for online updates and optimization.

[0012] A further improvement of the present invention is that the specific steps of pre-processing the dynamic operation data of each region in S1 are as follows:

[0013] 1.1: Dynamic operation data is collected from actual boiler operation. The noise in each set of dynamic operation data is smoothed using a high-pass filter, and missing values ​​in the dynamic operation data are filled using the cubic spline interpolation method.

[0014] 1.2: Use the quartile algorithm to divide the dynamic running data after missing values ​​into 4 parts in ascending order, each accounting for 25% of the entire data set, and the lower quartile is recorded as , the median is recorded as , the upper quartile is recorded as Calculate the difference between the upper quartile and the lower quartile and record it as the interquartile range ,according to as well as Get inner limits for outliers [ , ] and remove the data outside the range;

[0015] 1.3: Combine the boiler mechanism model and CFD software to simulate the boiler's combustion, heat transfer, and fluid behavior, generate simulation data consistent with the actual operating status, align the preprocessed actual dynamic operating data with the CFD simulation data, and integrate each group of dynamic operating data through partition matching.

[0016] A further improvement of the present invention is that the specific steps of constructing the boiler multi-region CFD-AI model in S2 are as follows:

[0017] 2.1: Based on the boiler CFD grid division, the integrated dynamic operation data is partitioned into spatial regions. Each partitioned training data set is then normalized. Feature extraction formulas are used to obtain the feature information corresponding to each set of dynamic operation data. The dynamic operation data of each region and its corresponding feature information are then divided into training, validation, and test sets.

[0018] 2.2: Convert the training, validation, and test sets into time series format. Build a multi-zone boiler CFD-AI model based on the model architecture consisting of the input layer, LSTM layer, fully connected layer, and output layer. Then, use random initialization to initialize the weight and bias parameters of the multi-zone boiler CFD-AI model.

[0019] 2.3: The training set data is input into the boiler multi-zone CFD-AI model. The CFD-AI model uses the forward propagation algorithm to calculate the hidden state and output value of the LSTM layer. It recursively updates the hidden state and memory units at each time step to obtain the final output layer prediction value. The model's prediction value is compared with the actual value, and the corresponding loss value is calculated using the MSE loss function.

[0020] 2.4: Using the chain rule, the loss value is gradually propagated from the output layer to the input layer. The gradient of the loss value with respect to the parameters of each model layer is calculated, and the model parameters are updated using the gradient descent algorithm. After each round of training, the performance indicators of the boiler multi-zone CFD-AI model are evaluated using the validation set data, and the current evaluation results are compared with the results of the previous round. If there is underfitting or overfitting, the training is terminated early. Otherwise, the model hyperparameters are adjusted and the model is retrained.

[0021] 2.5: Repeat forward propagation, loss calculation, backpropagation, and parameter updates until the loss value of the boiler multi-zone CFD-AI model converges to the preset range or reaches the maximum number of iterations. Then, use the test set to evaluate the performance of the final model to verify its accuracy and robustness on unseen data, and store the trained model parameters and structure in the central server.

[0022] A further improvement of the present invention is that the specific steps of the global optimization of the boiler multi-region CFD-AI model in S2 are as follows:

[0023] 3.1: The central server sends the boiler multi-zone CFD-AI model to each client as the client's initial local model and initializes each client's local model using the trained model parameters and structure;

[0024] 3.2: Each client uses the locally collected dynamic running data to train the corresponding local model, and uses the backpropagation algorithm to calculate the gradient and update the model parameters with the goal of minimizing the loss function. After completing multiple rounds of local training, each client transmits the final model parameters updated by local training to the central server;

[0025] 3.3: The central server calculates the model parameters of each client using the weighted average method and aggregates them into a global model. It then updates the global model through mean iteration. After the global model update is completed, the central server broadcasts the new global model to all clients and updates the parameter information of each client's local model.

[0026] 3.4: Each client continues to train locally based on the new global model and performs the next round of mean iteration. Training ends when the change in the global model parameter update of the central server is less than the preset threshold, the change in the loss function for multiple consecutive rounds converges to the preset threshold, or the maximum number of training rounds is reached;

[0027] 3.5: Use the optimized global model to evaluate the test set data again to verify its performance on unseen data. Replace the original boiler multi-zone CFD-AI model parameters and structure with the global model parameters and structure, and deploy the updated boiler multi-zone CFD-AI model parameters on the actual boiler monitoring platform.

[0028] A further improvement of the present invention is that the specific steps of dynamically adjusting the structure and parameters of the globally optimized boiler multi-region CFD-AI model in S3 are as follows:

[0029] 4.1: Collect the structure and parameter information of the boiler multi-zone CFD-AI model and randomly generate multiple sets of optimization body positions to represent each set of solutions in the model space. The position of each optimization body represents a potential model structure and parameter combination. Substitute the parameters of each optimization body into the globally optimized boiler multi-zone CFD-AI model. Use the validation set to evaluate its performance, and use the MSE loss function result as the fitness of the optimization body.

[0030] 4.2: At the beginning of the iteration, select the three groups of optimized bodies with the highest fitness, and define them as Alpha body with the best fitness, Beta body with the second best fitness, and Gamma body with the third best fitness, and record the positions of the three groups of optimized bodies;

[0031] 4.3: Perform a global search by updating the positions of the remaining optimized bodies based on the position of the Alpha body. Adjust the positions of the remaining optimized bodies except the Alpha body based on the positions of the Beta body and the Gamma body to perform a local search. After the position update of each optimized body is completed, check whether the position of each optimized body meets the preset model structure and parameter boundary values. If it exceeds the preset boundary value, it is corrected to the boundary value. Then, the fitness calculation, optimized body selection and position update are repeated until the fitness value change of the Alpha body converges to the preset range;

[0032] 4.4: After the iteration, the model structure and parameter combination corresponding to the optimized body with the highest fitness, that is, the Alpha body, is selected as the optimal combination, and used as the optimized boiler multi-zone CFD-AI model parameters for final deployment.

[0033] A further improvement of the present invention is that the specific steps of enhancing the key area decision of the boiler multi-area CFD-AI model in S3 are as follows:

[0034] 5.1: The different states of each boiler region are treated as nodes, and the transition relationships between these states are generated based on the boiler's control system and environmental conditions. The current state of each boiler region predicted by the boiler multi-region CFD-AI model is used as the root node. Based on the transition relationship of this state, the corresponding state of each boiler region is used as a child node, and the state transition relationship is used as an edge. A corresponding search tree is constructed based on its root node and child nodes, and the number of visits and cumulative rewards of each group of nodes are initialized;

[0035] 5.2: Starting from the root node, use the UCT formula to calculate the upper confidence value of each child node. Select the child node with the highest upper confidence value along the search tree to build the current optimal path until reaching a leaf node. If the current leaf node has not been fully expanded, randomly select an unexplored action from the action space of the current node, calculate the new state using the state transition function, and add the new state as a new node to the search tree. Initialize the number of visits to the node and the accumulated reward;

[0036] 5.3: Starting from the newly expanded node, the state transition is repeated according to the multiple transfer paths generated by the boiler multi-zone CFD-AI model to generate a complete path until the preset maximum number of steps is reached. After the simulation is completed, the cumulative reward of the simulated path is calculated using the set reward function. The accumulated reward obtained from the simulation is back-tracked to all nodes on the path and the visit count of these nodes is updated until the root node is reached and the simulation stops.

[0037] 5.4: After the cumulative rewards and visit counts for each node are updated, repeat the selection, expansion, simulation, and backtracking until the cumulative reward changes for each path converge to the preset range. Then, starting from the root node, the child nodes with the most visits are selected in sequence as the optimal decisions for the key areas. Based on the decisions obtained from the search, the optimization direction and parameter adjustment strategy of the boiler multi-region CFD-AI model in the boiler's highly nonlinear area are determined.

[0038] The federated learning-based boiler multi-zone CFD-AI modeling system includes:

[0039] Data collection and preprocessing module, used to simulate the heat transfer and fluid behavior of the boiler, establish multi-region CFD grid division, and collect and preprocess the dynamic operation data of each region from the actual operation of the boiler;

[0040] The model building module is used to extract the features of the pre-processed dynamic operation data, build a boiler multi-zone CFD-AI model based on the extracted features, and globally optimize the boiler multi-zone CFD-AI model;

[0041] The model dynamic adjustment module is used to dynamically adjust the structure and parameters of the globally optimized boiler multi-zone CFD-AI model to enhance the key regional decision-making of the boiler multi-zone CFD-AI model;

[0042] An evaluation module is used to assess the deviation between the prediction results of the boiler multi-zone CFD-AI model and the actual data, and to adjust the decision path and model structure to improve the fitting effect of high-dimensional features;

[0043] The model real-time prediction module is used to repeatedly train and update the boiler multi-zone CFD-AI model, perform real-time predictions through the final model, and regularly collect operating data for online updates and optimization.

[0044] A further improvement of the present invention is that, in the data collection and preprocessing module, the specific steps of preprocessing the dynamic operation data of each region are as follows:

[0045] 1.1: Dynamic operation data is collected from actual boiler operation. The noise in each set of dynamic operation data is smoothed using a high-pass filter, and missing values ​​in the dynamic operation data are filled using the cubic spline interpolation method.

[0046] 1.2: Use the quartile algorithm to divide the dynamic running data after missing values ​​into 4 parts in ascending order, each accounting for 25% of the entire data set, and the lower quartile is recorded as , the median is recorded as , the upper quartile is recorded as Calculate the difference between the upper quartile and the lower quartile and record it as the interquartile range ,according to as well as Get inner limits for outliers [ , ] and remove the data outside the range;

[0047] 1.3: Combine the boiler mechanism model and CFD software to simulate the boiler's combustion, heat transfer, and fluid behavior, generate simulation data consistent with the actual operating status, align the preprocessed actual dynamic operating data with the CFD simulation data, and integrate each group of dynamic operating data through partition matching.

[0048] A further improvement of the present invention is that, in the model construction module, the specific steps of constructing the boiler multi-region CFD-AI model are as follows:

[0049] 2.1: Based on the boiler CFD grid division, the integrated dynamic operation data is partitioned into spatial regions. Each partitioned training data set is then normalized. Feature extraction formulas are used to obtain the feature information corresponding to each set of dynamic operation data. The dynamic operation data of each region and its corresponding feature information are then divided into training, validation, and test sets.

[0050] 2.2: Convert the training, validation, and test sets into time series format. Build a multi-zone boiler CFD-AI model based on the model architecture consisting of the input layer, LSTM layer, fully connected layer, and output layer. Then, use random initialization to initialize the weight and bias parameters of the multi-zone boiler CFD-AI model.

[0051] 2.3: The training set data is input into the boiler multi-zone CFD-AI model. The CFD-AI model uses the forward propagation algorithm to calculate the hidden state and output value of the LSTM layer. It recursively updates the hidden state and memory units at each time step to obtain the final output layer prediction value. The model's prediction value is compared with the actual value, and the corresponding loss value is calculated using the MSE loss function.

[0052] 2.4: Using the chain rule, the loss value is gradually propagated from the output layer to the input layer. The gradient of the loss value with respect to the parameters of each model layer is calculated, and the model parameters are updated using the gradient descent algorithm. After each round of training, the performance indicators of the boiler multi-zone CFD-AI model are evaluated using the validation set data, and the current evaluation results are compared with the results of the previous round. If there is underfitting or overfitting, the training is terminated early. Otherwise, the model hyperparameters are adjusted and the model is retrained.

[0053] 2.5: Repeat forward propagation, loss calculation, backpropagation, and parameter updates until the loss value of the boiler multi-zone CFD-AI model converges to the preset range or reaches the maximum number of iterations. Then, use the test set to evaluate the performance of the final model to verify its accuracy and robustness on unseen data, and store the trained model parameters and structure in the central server.

[0054] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the federated learning-based boiler multi-zone CFD-AI modeling method.

[0055] Compared with the prior art, the present invention has at least the following beneficial technical effects:

[0056] The boiler multi-area CFD-AI modeling method and system based on federated learning provided by the present invention sends the boiler multi-area CFD-AI model to each client through the central server as the initial local model of the client, and uses the model parameters and structure after training to initialize the local model of each client. Each client uses the local dynamic operation data collected to train the corresponding local model, and uses the back propagation algorithm to calculate the gradient and update the model parameters with the goal of minimizing the loss function. After completing multiple rounds of local training, each client transmits the final model parameters updated by the local training to the central server. The central server calculates the model parameters of each client by the weighted average method, aggregates them into a global model, and then updates the global model by mean iteration. After the global model is updated, the central server The new global model is broadcast to all clients, and the parameter information of the local model of each client is updated. Each client continues to train locally according to the new global model and performs the next round of mean iteration. When the change in the global model parameter update of the central server is less than the preset threshold, the loss function change for multiple consecutive rounds converges to the preset threshold or reaches the maximum number of training rounds, the training is terminated, and the optimized global model is used to evaluate the test set data again to verify its performance on unseen data. The original boiler multi-region CFD-AI model parameters and structure are replaced with the global model parameters and structure, and the updated boiler multi-region CFD-AI model parameters are deployed on the actual boiler monitoring platform. This can improve the efficiency of model training, save a lot of communication resources, reduce the risk of privacy leakage, and promote the cross-regional adaptability of the model.

[0057] The boiler multi-region CFD-AI modeling method based on federated learning in the present invention collects the structure and parameter information of the boiler multi-region CFD-AI model, randomly generates the positions of multiple groups of optimization bodies to represent each group of solutions in the model space, and the position of each optimization body represents a potential model structure and parameter combination. The parameters of each optimization body are substituted into the globally optimized boiler multi-region CFD-AI model, and its performance is evaluated using a validation set. The result of the MSE loss function is used as the fitness of the optimization body. At the beginning of the iteration, the three groups of optimization bodies with the highest fitness are selected, and global and local searches are performed based on the selected three groups of optimal bodies. After the position of each optimization body is updated, the detection Whether the position of each optimization body meets the preset model structure and parameter boundary values. If it exceeds the preset boundary values, it will be corrected to the boundary values. Then the fitness calculation, optimization body selection and position update are repeated until the fitness value change of the optimization body with the highest fitness converges to the preset range. After the iteration, the model structure and parameter combination corresponding to the optimization body with the highest fitness is taken as the optimal combination, and it is used as the optimized boiler multi-region CFD-AI model parameter and used for final deployment, which significantly improves the performance of the model under complex working conditions, can effectively handle multi-objective optimization problems, and can more accurately predict and control the performance of the boiler in different regions and working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0059] Figure 1 This is a flow chart of the federated learning-based boiler multi-zone CFD-AI modeling method proposed in this invention.

[0060] Figure 2 This is a structural block diagram of the federated learning-based boiler multi-zone CFD-AI modeling system proposed in this invention. DETAILED DESCRIPTION

[0061] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and description are to be considered as illustrative in nature and not restrictive.

[0062] It will be understood that when used in this specification and the appended claims, the terms “comprises” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0063] It should also be understood that the terms used in the present specification are only for the purpose of describing particular embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an", and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0064] It should be further understood that the term "and / or" used in the present description and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0065] The accompanying drawings illustrate various schematic diagrams of structures according to embodiments disclosed herein. These figures are not drawn to scale; for clarity, some details are exaggerated and some details may be omitted. The shapes of the various regions and layers shown in the figures, as well as their relative sizes and positional relationships, are merely exemplary and may deviate in practice due to manufacturing tolerances or technical limitations. Those skilled in the art may design regions / layers with different shapes, sizes, and relative positions as needed.

[0066] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0067] Example 1

[0068] Reference Figure 1 , a multi-zone CFD-AI modeling method for boilers based on federated learning. The specific steps of this modeling method are as follows:

[0069] Simulate the heat transfer and fluid behavior of the boiler, establish multi-region CFD mesh division, and collect and pre-process the dynamic operation data of each region from the actual operation of the boiler.

[0070] Specifically, dynamic operation data were collected from the actual operation of the boiler, and the noise in each group of dynamic operation data was smoothed using a high-pass filter. The missing values ​​in the dynamic operation data were filled using the cubic spline interpolation method. The dynamic operation data after missing value filling were divided into 4 parts in ascending order using the quartile algorithm, each accounting for 25% of the entire data set, and the lower quartile was recorded as , the median is recorded as , the upper quartile is recorded as Calculate the difference between the upper quartile and the lower quartile and record it as the interquartile range ,according to as well as Get inner limits for outliers [ , ] and eliminate data outside this range. Combining the boiler mechanism model and CFD software, the combustion, heat transfer and fluid behavior of the boiler are simulated to generate simulation data consistent with the actual operating status. The actual dynamic operation data after preprocessing is aligned with the CFD simulation data, and each group of dynamic operation data is integrated through partition matching.

[0071] The features of the preprocessed dynamic operation data are extracted, a boiler multi-zone CFD-AI model is constructed based on the extracted features, and the boiler multi-zone CFD-AI model is globally optimized.

[0072] Specifically, according to the boiler CFD grid division, the integrated dynamic operation data are partitioned into spatial regions, and then the partitioned training data sets are normalized. The feature information corresponding to each group of dynamic operation data is obtained through the feature extraction formula, and the dynamic operation data of each region and its corresponding feature information are divided into training set, validation set and test set. The training set, validation set and test set are converted into time series format. According to the model architecture of the input layer, LSTM layer, fully connected layer and output layer, a boiler multi-region CFD-AI model is established. Then, the random initialization method is used to initialize the weight parameters and bias parameters of the boiler multi-region CFD-AI model. The training set data is input into the boiler multi-region CFD-AI model. The CFD-AI model calculates the hidden state and output value of the LSTM layer through the forward propagation algorithm, and recursively updates the hidden state and memory unit time step by time to obtain the final output. Output the layer prediction value, compare the model prediction value with the true value, calculate the corresponding loss value through the MSE loss function, and gradually propagate the loss value from the output layer to the input layer through the chain rule. Calculate the gradient of the loss value to the parameters of each layer of the model, and use the gradient descent algorithm to update the model parameters. After each round of training, evaluate the performance indicators of the boiler multi-area CFD-AI model through the validation set data, and compare the current evaluation results with the evaluation results of the previous round. If there is fitting or overfitting, terminate the training early. Otherwise, adjust the hyperparameters of the model and retrain the model. Repeat forward propagation, loss calculation, back propagation and parameter update until the loss value of the boiler multi-area CFD-AI model converges to the preset range or reaches the maximum number of iterations. Then use the test set to evaluate the performance of the final model to verify its accuracy and robustness on unseen data, and store the trained model parameters and structure in the central server.

[0073] Specifically, the central server sends the boiler multi-area CFD-AI model to each client as the client's initial local model, and uses the model parameters and structure after training to initialize the local model of each client. Each client uses the local dynamic operation data collected to train the corresponding local model, and uses the back propagation algorithm to calculate the gradient and update the model parameters with the goal of minimizing the loss function. After completing multiple rounds of local training, each client transmits the final model parameters updated by local training to the central server. The central server calculates the model parameters of each client through the weighted average method, aggregates them into a global model, and then updates the global model through mean iteration. The global model update is completed Afterwards, the central server broadcasts the new global model to all clients and updates the parameter information of each client's local model. Each client continues to train locally according to the new global model and performs the next round of mean iteration. When the change in the global model parameter update of the central server is less than the preset threshold, the loss function change for multiple consecutive rounds converges to the preset threshold or reaches the maximum number of training rounds, the training is terminated, and the optimized global model is used to evaluate the test set data again to verify its performance on unseen data. The original boiler multi-zone CFD-AI model parameters and structure are replaced with the global model parameters and structure, and the updated boiler multi-zone CFD-AI model parameters are deployed on the actual boiler monitoring platform.

[0074] Example 2

[0075] Reference Figure 1 , a multi-zone CFD-AI modeling method for boilers based on federated learning. The specific steps of this modeling method are as follows:

[0076] Dynamically adjust the structure and parameters of the globally optimized boiler multi-zone CFD-AI model to enhance the key area decision-making of the boiler multi-zone CFD-AI model.

[0077] Specifically, the structure and parameter information of the boiler multi-region CFD-AI model is collected, and the positions of multiple groups of optimization bodies are randomly generated to represent each group of solutions in the model space. The position of each optimization body represents a potential model structure and parameter combination. The parameters of each optimization body are substituted into the globally optimized boiler multi-region CFD-AI model, and its performance is evaluated using a validation set. The result of the MSE loss function is used as the fitness of the optimization body. At the beginning of the iteration, the three groups of optimization bodies with the highest fitness are selected and defined as the Alpha body with the best fitness, the Beta body with the second best fitness, and the Gamma body with the third best fitness. At the same time, the positions of the three groups of optimization bodies are recorded, and the positions of the remaining optimization bodies are updated according to the position of the Alpha body. The optimization bodies are positioned for global search, and the positions of the remaining optimization bodies except the Alpha body are adjusted according to the positions of the Beta body and the Gamma body for local search. After the position of each optimization body is updated, it is detected whether the position of each optimization body meets the preset model structure and parameter boundary value. If it exceeds the preset boundary value, it is corrected to the boundary value. Then, the fitness calculation, optimization body selection and position update are repeated until the fitness value change of the Alpha body converges to the preset range. After the iteration, the optimization body with the highest fitness, that is, the model structure and parameter combination corresponding to the Alpha body is taken as the optimal combination, and it is used as the optimized boiler multi-region CFD-AI model parameter for final deployment.

[0078] Specifically, the different states of each area of ​​the boiler are taken as nodes, and the transfer relationship between each state is generated based on the control system and environmental conditions of the boiler. The current state of each area of ​​the boiler predicted by the boiler multi-area CFD-AI model is taken as the root node, and based on the transfer relationship of the state, the corresponding state of each area of ​​the boiler is taken as the child node, and its state transfer relationship is used as the edge. The corresponding search tree is constructed based on its root node and child nodes, and the number of visits and cumulative returns of each group of nodes are initialized. Starting from the root node, the UCT formula is used to calculate the upper confidence value of each child node, and the child node with the highest upper confidence value is selected along the search tree to construct the current optimal path until the leaf node is reached. If the current leaf node is not fully expanded, an unexplored action is randomly selected from the action space of the current node, and the new state is calculated using the state transfer function. The new state is added to the search tree as a new node to initialize the node. The number of visits and cumulative rewards of each node are calculated. Starting from the currently expanded new node, the state transfer is repeated according to the multiple transfer paths generated by the boiler multi-region CFD-AI model to generate a complete path until the preset maximum number of steps is reached and then stopped. After the simulation is completed, the cumulative reward of the simulation path is calculated by the set reward function, and the cumulative reward obtained by simulation is traced back to all nodes on the path, and the number of visits to these nodes is updated until it is traced back to the root node and then stopped. After the cumulative rewards and the number of visits of each node are updated, the selection, expansion, simulation and backtracking are repeated until the cumulative reward changes of each path converge to the preset range. Then, starting from the root node, the child nodes with the most visits are selected in turn as the best decision for the key area, and based on the decision obtained by the search, the optimization direction and parameter adjustment strategy of the boiler multi-region CFD-AI model in the high nonlinear area of ​​the boiler are determined.

[0079] Evaluate the deviation between the prediction results of the boiler multi-zone CFD-AI model and the actual data, and adjust the decision path and model structure to improve the fitting effect of high-dimensional features.

[0080] The boiler multi-zone CFD-AI model is repeatedly trained and updated, and real-time predictions are made through the final model. Operation data is regularly collected for online updates and optimization.

[0081] Example 2

[0082] like Figure 2 As shown, the boiler multi-region CFD-AI modeling system based on federated learning provided by the present invention includes:

[0083] Data collection and preprocessing module, used to simulate the heat transfer and fluid behavior of the boiler, establish multi-region CFD grid division, and collect and preprocess the dynamic operation data of each region from the actual operation of the boiler;

[0084] The model building module is used to extract the features of the pre-processed dynamic operation data, build a boiler multi-zone CFD-AI model based on the extracted features, and globally optimize the boiler multi-zone CFD-AI model;

[0085] The model dynamic adjustment module is used to dynamically adjust the structure and parameters of the globally optimized boiler multi-zone CFD-AI model to enhance the key regional decision-making of the boiler multi-zone CFD-AI model;

[0086] An evaluation module is used to assess the deviation between the prediction results of the boiler multi-zone CFD-AI model and the actual data, and to adjust the decision path and model structure to improve the fitting effect of high-dimensional features;

[0087] The model real-time prediction module is used to repeatedly train and update the boiler multi-zone CFD-AI model, perform real-time predictions through the final model, and regularly collect operating data for online updates and optimization.

[0088] Example 3

[0089] The present invention provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the federated learning-based boiler multi-zone CFD-AI modeling method.

[0090] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0091] The present application is described with reference to the flowcharts and / or block diagrams of the methods, systems, and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A system that specifies the functions of a box or boxes.

[0092] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0093] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0094] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, from all points of view, the embodiments should be regarded as illustrative and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes that fall within the meaning and range of equivalents of the claims are included in the present invention. Any reference signs in the claims should not be construed as limiting the claim to which they relate.

[0095] In addition, it should be understood that although this specification describes the embodiments, not every embodiment contains only one independent technical solution. This description is for clarity only. Those skilled in the art should consider the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for the purpose of illustrating the technical concept of the present invention and cannot be used to limit the scope of protection of the present invention. Any changes made based on the technical solution in accordance with the technical concept proposed by the present invention fall within the scope of protection of the claims of the present invention.

Claims

1. A multi-region CFD-AI modeling method for boilers based on federated learning, characterized by: The steps include: S1. Simulate the heat transfer and fluid behavior of the boiler, establish a multi-region CFD grid, and collect and pre-process the dynamic operation data of each region from the actual operation of the boiler; S2. Extract features of the pre-processed dynamic operation data, construct a boiler multi-zone CFD-AI model based on the extracted features, and globally optimize the boiler multi-zone CFD-AI model. The specific steps of globally optimizing the boiler multi-zone CFD-AI model are as follows: 3.1: The central server sends the boiler multi-zone CFD-AI model to each client as the client's initial local model and initializes each client's local model using the trained model parameters and structure; 3.2: Each client uses the locally collected dynamic running data to train the corresponding local model, and uses the backpropagation algorithm to calculate the gradient and update the model parameters with the goal of minimizing the loss function. After completing multiple rounds of local training, each client transmits the final model parameters updated by local training to the central server; 3.3: The central server calculates the model parameters of each client using the weighted average method and aggregates them into a global model. It then updates the global model through mean iteration. After the global model update is completed, the central server broadcasts the new global model to all clients and updates the parameter information of each client's local model. 3.4: Each client continues to train locally based on the new global model and performs the next round of mean iteration. Training ends when the change in the global model parameter update of the central server is less than the preset threshold, the change in the loss function for multiple consecutive rounds converges to the preset threshold, or the maximum number of training rounds is reached; 3.5: Use the optimized global model to re-evaluate the test data set to verify its performance on unseen data. The original boiler multi-zone CFD-AI model parameters and structure are replaced with the global model parameters and structure. The updated boiler multi-zone CFD-AI model parameters are deployed on the actual boiler monitoring platform. S3. Dynamically adjust the structure and parameters of the globally optimized multi-region CFD-AI model of the boiler to enhance key regional decision-making of the multi-region CFD-AI model of the boiler. The specific steps of dynamically adjusting the structure and parameters of the globally optimized multi-region CFD-AI model of the boiler are as follows: 4.1: Collect the structure and parameter information of the boiler multi-zone CFD-AI model and randomly generate multiple sets of optimization body positions to represent each set of solutions in the model space. The position of each optimization body represents a potential model structure and parameter combination. Substitute the parameters of each optimization body into the globally optimized boiler multi-zone CFD-AI model. Use the validation set to evaluate its performance, and use the MSE loss function result as the fitness of the optimization body. 4.2: At the beginning of the iteration, select the three groups of optimized bodies with the highest fitness, and define them as Alpha body with the best fitness, Beta body with the second best fitness, and Gamma body with the third best fitness, and record the positions of the three groups of optimized bodies; 4.3: Perform a global search by updating the positions of the remaining optimized bodies based on the position of the Alpha body. Adjust the positions of the remaining optimized bodies except the Alpha body based on the positions of the Beta body and the Gamma body to perform a local search. After the position update of each optimized body is completed, check whether the position of each optimized body meets the preset model structure and parameter boundary values. If it exceeds the preset boundary value, it is corrected to the boundary value. Then, the fitness calculation, optimized body selection and position update are repeated until the fitness value change of the Alpha body converges to the preset range; 4.4: After the iteration, the model structure and parameter combination corresponding to the optimized body with the highest fitness, that is, the Alpha body, is selected as the optimal combination and used as the optimized boiler multi-zone CFD-AI model parameters for final deployment; S4. Evaluate the deviation between the prediction results of the boiler multi-region CFD-AI model and the actual data, and adjust the decision path and model structure to improve the fitting effect of high-dimensional features; S5. Repeatedly train and update the boiler multi-zone CFD-AI model, perform real-time predictions through the final model, and regularly collect operating data for online updates and optimization.

2. The federated learning-based boiler multi-region CFD-AI modeling method according to claim 1 is characterized in that: The specific steps for pre-processing the dynamic operation data of each region in S1 are as follows: 1.1: Dynamic operation data is collected from actual boiler operation. The noise in each set of dynamic operation data is smoothed using a high-pass filter, and missing values ​​in the dynamic operation data are filled using the cubic spline interpolation method. 1.2: Use the quartile algorithm to divide the dynamic running data after missing values ​​into 4 parts in ascending order, each accounting for 25% of the entire data set. The lower quartile is recorded as Q1, the median is recorded as Q2, and the upper quartile is recorded as Q3. The difference between the upper quartile and the lower quartile is recorded as the interquartile range I QR , according to Q1-1.5I QR and Q3+1.5I QR Get the inner limit of the outlier [F L , F U ] and remove the data outside the range; 1.3: Combine the boiler mechanism model and CFD software to simulate the boiler's combustion, heat transfer, and fluid behavior, generate simulation data consistent with the actual operating status, align the preprocessed actual dynamic operating data with the CFD simulation data, and integrate each group of dynamic operating data through partition matching.

3. The federated learning-based boiler multi-region CFD-AI modeling method according to claim 2 is characterized in that: The specific steps for constructing the boiler multi-zone CFD-AI model described in S2 are as follows: 2.1: Based on the boiler CFD grid division, the integrated dynamic operation data is partitioned into spatial regions. Each partitioned training data set is then normalized. Feature extraction formulas are used to obtain the feature information corresponding to each set of dynamic operation data. The dynamic operation data of each region and its corresponding feature information are then divided into training, validation, and test sets. 2.2: Convert the training, validation, and test sets into time series format. Build a multi-zone boiler CFD-AI model based on the model architecture consisting of the input layer, LSTM layer, fully connected layer, and output layer. Then, use random initialization to initialize the weight and bias parameters of the multi-zone boiler CFD-AI model. 2.3: The training set data is input into the boiler multi-zone CFD-AI model. The CFD-AI model uses the forward propagation algorithm to calculate the hidden state and output value of the LSTM layer. It recursively updates the hidden state and memory units at each time step to obtain the final output layer prediction value. The model's prediction value is compared with the actual value, and the corresponding loss value is calculated using the MSE loss function. 2.4: Using the chain rule, the loss value is gradually propagated from the output layer to the input layer. The gradient of the loss value with respect to the parameters of each model layer is calculated, and the model parameters are updated using the gradient descent algorithm. After each round of training, the performance indicators of the boiler multi-zone CFD-AI model are evaluated using the validation set data, and the current evaluation results are compared with the results of the previous round. If there is underfitting or overfitting, the training is terminated early. Otherwise, the model hyperparameters are adjusted and the model is retrained. 2.5: Repeat forward propagation, loss calculation, backpropagation, and parameter updates until the loss value of the boiler multi-zone CFD-AI model converges to the preset range or reaches the maximum number of iterations. Then, use the test set to evaluate the performance of the final model to verify its accuracy and robustness on unseen data, and store the trained model parameters and structure in the central server.

4. The boiler multi-region CFD-AI modeling method based on federated learning according to claim 1 is characterized in that: The specific steps for enhancing the key area decision-making of the boiler multi-area CFD-AI model described in S3 are as follows: 5.1: The different states of each boiler region are treated as nodes, and the transition relationships between these states are generated based on the boiler's control system and environmental conditions. The current state of each boiler region predicted by the boiler multi-region CFD-AI model is used as the root node. Based on the transition relationship of this state, the corresponding state of each boiler region is used as a child node, and the state transition relationship is used as an edge. A corresponding search tree is constructed based on its root node and child nodes, and the number of visits and cumulative rewards of each group of nodes are initialized; 5.2: Starting from the root node, use the UCT formula to calculate the upper confidence value of each child node. Select the child node with the highest upper confidence value along the search tree to build the current optimal path until reaching a leaf node. If the current leaf node has not been fully expanded, randomly select an unexplored action from the action space of the current node, calculate the new state using the state transition function, and add the new state as a new node to the search tree. Initialize the number of visits to the node and the accumulated reward; 5.3: Starting from the newly expanded node, the state transition is repeated according to the multiple transfer paths generated by the boiler multi-zone CFD-AI model to generate a complete path until the preset maximum number of steps is reached. After the simulation is completed, the cumulative reward of the simulated path is calculated using the set reward function. The accumulated reward obtained from the simulation is back-tracked to all nodes on the path and the visit count of these nodes is updated until the root node is reached and the simulation stops. 5.4: After the cumulative rewards and visit counts for each node are updated, repeat the selection, expansion, simulation, and backtracking until the cumulative reward changes for each path converge to the preset range. Then, starting from the root node, the child nodes with the most visits are selected in sequence as the optimal decisions for the key areas. Based on the decisions obtained from the search, the optimization direction and parameter adjustment strategy of the boiler multi-region CFD-AI model in the boiler's highly nonlinear area are determined.

5. The boiler multi-region CFD-AI modeling system based on federated learning is characterized by: include: Data collection and preprocessing module, used to simulate the heat transfer and fluid behavior of the boiler, establish multi-region CFD grid division, and collect and preprocess the dynamic operation data of each region from the actual operation of the boiler; The model building module is used to extract features of the preprocessed dynamic operation data, build a boiler multi-zone CFD-AI model based on the extracted features, and globally optimize the boiler multi-zone CFD-AI model. The specific steps of globally optimizing the boiler multi-zone CFD-AI model are as follows: 3.1: The central server sends the boiler multi-zone CFD-AI model to each client as the client's initial local model and initializes each client's local model using the trained model parameters and structure; 3.2: Each client uses the locally collected dynamic running data to train the corresponding local model, and uses the backpropagation algorithm to calculate the gradient and update the model parameters with the goal of minimizing the loss function. After completing multiple rounds of local training, each client transmits the final model parameters updated by local training to the central server; 3.3: The central server calculates the model parameters of each client using the weighted average method and aggregates them into a global model. It then updates the global model through mean iteration. After the global model update is completed, the central server broadcasts the new global model to all clients and updates the parameter information of each client's local model. 3.4: Each client continues to train locally based on the new global model and performs the next round of mean iteration. Training ends when the change in the global model parameter update of the central server is less than the preset threshold, the change in the loss function for multiple consecutive rounds converges to the preset threshold, or the maximum number of training rounds is reached; 3.5: Use the optimized global model to re-evaluate the test data set to verify its performance on unseen data. The original boiler multi-zone CFD-AI model parameters and structure are replaced with the global model parameters and structure. The updated boiler multi-zone CFD-AI model parameters are deployed on the actual boiler monitoring platform. The model dynamic adjustment module is used to dynamically adjust the structure and parameters of the globally optimized multi-zone boiler CFD-AI model to enhance key regional decision-making of the multi-zone boiler CFD-AI model. The specific steps of dynamically adjusting the structure and parameters of the globally optimized multi-zone boiler CFD-AI model are as follows: 4.1: Collect the structure and parameter information of the boiler multi-zone CFD-AI model and randomly generate multiple sets of optimization body positions to represent each set of solutions in the model space. The position of each optimization body represents a potential model structure and parameter combination. Substitute the parameters of each optimization body into the globally optimized boiler multi-zone CFD-AI model. Use the validation set to evaluate its performance, and use the MSE loss function result as the fitness of the optimization body. 4.2: At the beginning of the iteration, select the three groups of optimized bodies with the highest fitness, and define them as Alpha body with the best fitness, Beta body with the second best fitness, and Gamma body with the third best fitness, and record the positions of the three groups of optimized bodies; 4.3: Perform a global search by updating the positions of the remaining optimized bodies based on the position of the Alpha body. Adjust the positions of the remaining optimized bodies except the Alpha body based on the positions of the Beta body and the Gamma body to perform a local search. After the position update of each optimized body is completed, check whether the position of each optimized body meets the preset model structure and parameter boundary values. If it exceeds the preset boundary value, it is corrected to the boundary value. Then, the fitness calculation, optimized body selection and position update are repeated until the fitness value change of the Alpha body converges to the preset range; 4.4: After the iteration, the model structure and parameter combination corresponding to the optimized body with the highest fitness, that is, the Alpha body, is selected as the optimal combination and used as the optimized boiler multi-zone CFD-AI model parameters for final deployment; An evaluation module is used to assess the deviation between the prediction results of the boiler multi-zone CFD-AI model and the actual data, and to adjust the decision path and model structure to improve the fitting effect of high-dimensional features; The model real-time prediction module is used to repeatedly train and update the boiler multi-zone CFD-AI model, perform real-time predictions through the final model, and regularly collect operating data for online updates and optimization.

6. The federated learning-based boiler multi-region CFD-AI modeling system according to claim 5 is characterized in that: In the data collection and preprocessing module, the specific steps of preprocessing the dynamic operation data of each area are as follows: 1.1: Dynamic operation data is collected from actual boiler operation. The noise in each set of dynamic operation data is smoothed using a high-pass filter, and missing values ​​in the dynamic operation data are filled using the cubic spline interpolation method. 1.2: Use the quartile algorithm to divide the dynamic running data after missing values ​​into 4 parts in ascending order, each accounting for 25% of the entire data set. The lower quartile is recorded as Q1, the median is recorded as Q2, and the upper quartile is recorded as Q3. The difference between the upper quartile and the lower quartile is recorded as the interquartile range I QR , according to Q1-1.5I QR and Q3+1.5I QR Get the inner limit of the outlier [F L , F U ] and remove the data outside the range; 1.3: Combine the boiler mechanism model and CFD software to simulate the boiler's combustion, heat transfer, and fluid behavior, generate simulation data consistent with the actual operating status, align the preprocessed actual dynamic operating data with the CFD simulation data, and integrate each group of dynamic operating data through partition matching.

7. The federated learning-based boiler multi-region CFD-AI modeling system according to claim 5 is characterized in that: In the model construction module, the specific steps of constructing the boiler multi-zone CFD-AI model are as follows: 2.1: Based on the boiler CFD grid division, the integrated dynamic operation data is partitioned into spatial regions. Each partitioned training data set is then normalized. Feature extraction formulas are used to obtain the feature information corresponding to each set of dynamic operation data. The dynamic operation data of each region and its corresponding feature information are then divided into training, validation, and test sets. 2.2: Convert the training, validation, and test sets into time series format. Build a multi-zone boiler CFD-AI model based on the model architecture consisting of the input layer, LSTM layer, fully connected layer, and output layer. Then, use random initialization to initialize the weight and bias parameters of the multi-zone boiler CFD-AI model. 2.3: The training set data is input into the boiler multi-zone CFD-AI model. The CFD-AI model uses the forward propagation algorithm to calculate the hidden state and output value of the LSTM layer. It recursively updates the hidden state and memory units at each time step to obtain the final output layer prediction value. The model's prediction value is compared with the actual value, and the corresponding loss value is calculated using the MSE loss function. 2.4: Using the chain rule, the loss value is gradually propagated from the output layer to the input layer. The gradient of the loss value with respect to the parameters of each model layer is calculated, and the model parameters are updated using the gradient descent algorithm. After each round of training, the performance indicators of the boiler multi-zone CFD-AI model are evaluated using the validation set data, and the current evaluation results are compared with the results of the previous round. If there is underfitting or overfitting, the training is terminated early. Otherwise, the model hyperparameters are adjusted and the model is retrained. 2.5: Repeat forward propagation, loss calculation, backpropagation, and parameter updates until the loss value of the boiler multi-zone CFD-AI model converges to the preset range or reaches the maximum number of iterations. Then, use the test set to evaluate the performance of the final model to verify its accuracy and robustness on unseen data, and store the trained model parameters and structure in the central server.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the federated learning-based boiler multi-zone CFD-AI modeling method according to any one of claims 1 to 4.

Images