Thermal power generation equipment control signal optimization method and system based on deep learning

Abstract

The invention discloses a thermal power generation equipment control signal optimization method and system based on deep learning, and the method comprises the steps: collecting the operation data of thermal power generation equipment in real time, extracting key features in combination with principal component analysis, and carrying out the modeling and prediction of the spatial features and time sequence features of the equipment through CNN and LSTM, thereby achieving the optimization of the control signal of the thermal power generation equipment. And according to a prediction result of the deep learning model, constructing a multi-objective optimization function of fuel consumption minimization, pollutant emission minimization and equipment efficiency maximization, generating an optimization control signal, and dynamically adjusting operation parameters of equipment, such as fuel flow, air supply quantity and boiler load. And meanwhile, the real-time performance and accuracy of the control signal are ensured by periodically updating the deep learning model. The method can significantly improve the operation efficiency of thermal power generation equipment, reduce fuel consumption and pollution emission, prolong the service life of the equipment, and is suitable for efficient intelligent operation control in the field of thermal power generation.

Description

Technical Field

[0001] The present invention relates to the technical field of thermal power generation, and particularly to a method and system for optimizing control signals of thermal power generation equipment based on deep learning. Background Art

[0002] As one of the most widely used power generation methods globally, thermal power generation has long occupied an important position in energy supply. Traditional thermal power plants usually rely on large amounts of fuels (such as coal, natural gas, etc.) to generate heat energy, driving boilers and generator sets to generate electricity. The operation process of a thermal power plant is affected by multiple variables, including equipment performance, fuel flow rate, air supply volume, boiler load, combustion temperature, and other factors. The changes in these factors directly affect the energy efficiency, pollution emissions, and operation stability of the power plant. If these parameters are not adjusted and optimized in real time and accurately, it may lead to excessive energy waste, higher pollutant emissions, and even equipment failures. Therefore, how to obtain equipment operation data in real time and timely adjust equipment operation parameters to achieve the goals of improving power generation efficiency and reducing emissions has become an important research direction in the field of thermal power generation.

[0003] Currently, the operation of thermal power generation equipment usually relies on empirical rules or automatic regulation based on control systems. Although these traditional methods can ensure the normal operation of equipment to a certain extent, their control strategies have certain limitations: 1. Lack of intelligence in control strategies: Most existing control systems are based on preset rules, such as PID controllers. These rules are usually adjusted based on equipment empirical parameters and do not have self-learning capabilities. The control effect of the equipment is relatively rough and cannot be automatically optimized and adjusted according to the real-time changes in the equipment state. 2. Lack of accurate data analysis and prediction: Traditional control methods for thermal power generation equipment usually ignore the comprehensive analysis and accurate prediction of real-time data. The operation state of the equipment is affected by multiple factors, and the existing technologies fail to effectively utilize the historical data and real-time data of the equipment for dynamic adjustment. 3. Insufficient optimization of energy efficiency and emission control: When traditional systems control operation parameters such as fuel flow rate, air supply volume, and boiler load, they often rely on empirical models and it is difficult to achieve global optimality. There are often problems of energy waste and excessive pollutant emissions during the thermal power generation process, especially when the load fluctuates greatly. Summary of the Invention

[0004] Aiming at the deficiencies of the existing control of thermal power generation equipment, the present invention discloses a method and system for optimizing control signals of thermal power generation equipment based on deep learning, which can predict the future state of the equipment according to the real-time collected equipment operation data, and adjust the operation parameters of the equipment through multi-objective optimization to achieve the minimization of fuel consumption, the minimization of pollutant emissions, and the maximization of equipment efficiency.

[0005] To achieve the above object, the present invention adopts the following technical solutions: A method for optimizing control signals of thermal power generation equipment based on deep learning, characterized by comprising the following steps: Collect the operation data of thermal power generation equipment in real time, and the operation data includes boiler temperature, boiler pressure, fuel flow rate, oxygen concentration, flue gas emission concentration, and load; Perform data preprocessing on the collected operation data, including denoising, filling missing data, and standardization; Extract the characteristics of the operation of thermal power generation equipment from the preprocessed operation data through principal component analysis. The characteristics include equipment performance, energy efficiency, and pollution emissions, and select the key characteristics that have a significant impact on optimization through a feature selection method; Use CNN to extract the spatial features of the preprocessed operation data, use LSTM to model the time series features, input the extracted spatial features and time series features into a deep learning model for training, and the trained deep learning model can predict the future operation status and performance changes of thermal power generation equipment according to the input data, and improve the prediction accuracy by optimizing the loss function; Construct a multi-objective optimization function according to the prediction results of the deep learning model, and the optimization objectives include minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency; Generate control signals and adjust the operation parameters of thermal power generation equipment according to the results of the multi-objective optimization function, including fuel flow rate, air supply volume, and boiler load; Regularly update the deep learning model and retrain the model based on new data.

[0006] A further improvement of the present invention is that data preprocessing is performed on the collected operation data, including: Apply a band-pass filter to the collected operation data for denoising; Fill in the missing data by polynomial interpolation to ensure the continuity and consistency of the data; Normalize the real-time data using standardization processing so that the mean of each feature is zero and the standard deviation is one.

[0007] A further improvement of the present invention is that the characteristics of the operation of thermal power generation equipment are extracted from the preprocessed operation data through principal component analysis. The characteristics include equipment performance, energy efficiency, and pollution emissions, and the key characteristics that have a significant impact on optimization are selected through a feature selection method, including: Reduce the dimension of the preprocessed operation data through principal component analysis. The steps of principal component analysis include calculating the covariance matrix of the data, performing eigenvalue decomposition, and selecting the component with the highest proportion of explained variance as the principal component; Based on the results of principal component analysis, a feature selection algorithm is used to screen the principal components. The feature selection algorithm is based on information gain, mutual information, or correlation coefficient metrics to screen out the key features that have a significant impact on the optimization of thermal power generation equipment; Output the set of screened key features, including boiler temperature, boiler pressure, fuel flow rate, emission concentration, and energy efficiency.

[0008] A further improvement of the present invention is that a CNN is used to extract spatial features of the preprocessed operation data, and an LSTM is used to model time series features. The extracted spatial features and time series features are input into a deep learning model for training, including: Use a CNN model to extract spatial features of the preprocessed operation data. Take the preprocessed operation data as the input of the CNN convolutional layer, and use convolutional operations to extract spatial features in the standardized real-time data. The convolutional operations extract different spatial features through multiple convolutional kernels; Reduce the dimension of the feature map output by the convolutional layer through a pooling layer. The pooling operations include max pooling or average pooling; Input the time series features of the thermal power generation equipment into the LSTM and use the LSTM to model the extracted time series features to predict the future operating state by learning the time series pattern of the historical data of the thermal power generation equipment; Combine the spatial features and time series features extracted by the CNN and LSTM into a deep learning model, and use the backpropagation algorithm to train and adjust the model weights of the deep learning model to improve the prediction accuracy by optimizing the loss function.

[0009] A further improvement of the present invention is that the formula of the loss function is:

[0010] where, is the actual value, that is, the true operating state or performance of the thermal power generation equipment, is the predicted value of the deep learning model, that is, the predicted future operating state of the thermal power generation equipment according to the input data, is the number of samples, representing the number of training data, is the sample index.

[0011] A further improvement of the present invention is that a multi-objective optimization function is constructed according to the prediction results of the deep learning model, including: Through the trained deep learning model, predict the operating state and performance of the thermal power generation equipment to obtain the future states of each thermal power generation equipment, including fuel consumption, pollutant emissions, and equipment efficiency; Based on the results of fuel consumption, pollutant emissions, and equipment efficiency predicted by the deep learning model, a multi-objective optimization function is constructed. The formula of the multi-objective optimization function is:

[0012] Wherein, represents the multi-objective optimization function, 、 and are the weight coefficients for minimizing fuel consumption, minimizing pollution emissions, and maximizing equipment efficiency respectively, is the predicted value of the fuel consumption of the thermal power generation equipment, is the predicted value of the pollutant emissions of the thermal power generation equipment, represents the predicted value of the equipment efficiency; The optimization algorithm is used to balance the optimization objectives to generate the optimal control parameters, so as to minimize fuel consumption, minimize pollutant emissions, and maximize equipment efficiency, thereby realizing the optimal operation strategy of the thermal power generation equipment.

[0013] A further improvement of the present invention is that, according to the results of the multi-objective optimization function, a control signal is generated and the operating parameters of the thermal power generation equipment are adjusted, including: Based on the results of the multi-objective optimization function, a control signal is generated for adjusting the operating parameters of the equipment, including 、 and where is the fuel flow rate, is the air supply, is the boiler load, 、 and are control algorithms that map the multi-objective optimization results into specific operating parameters; The control algorithm is used to apply the generated control signal to the control module of the thermal power generation equipment to adjust the operating parameters of the thermal power generation equipment, so that the equipment reaches the goals of minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency; Through the real-time feedback control system, the operating state of the adjusted thermal power generation equipment is monitored, and the feedback signal is transmitted back to the optimization module for the next round of optimization, so as to ensure that the thermal power generation equipment is always in the optimal operating state.

[0014] A further improvement of the present invention is that the deep learning model is updated regularly and the model is retrained based on new data, including: Regularly obtain the latest data of the equipment operation, including fuel flow rate, air supply, and boiler load, as training data; Preprocess the newly acquired data, perform normalization and denoising operations to ensure the quality and consistency of the data; Retrain the deep learning model using the latest data, where the deep learning model is used to predict the operating status of thermal power generation equipment and adjust internal parameters to adapt to the changes of thermal power generation equipment; Evaluate the performance of the trained deep learning model, use an independent validation dataset to test the prediction accuracy of the deep learning model and the effectiveness of the control signal. If the performance is insufficient, adjust the deep learning model structure or training parameters; Replace the original deep learning model with the newly trained deep learning model and apply the updated deep learning model to the control system of thermal power generation equipment to generate new control signals to adjust the operating parameters of thermal power generation equipment; During the entire update process, through incremental learning or other efficient algorithms, ensure the real-time nature of the control signal generation process, reduce the delay during the update process, and ensure that the thermal power generation equipment always operates in the optimal state.

[0015] A control signal optimization system for thermal power generation equipment based on deep learning, including: A data acquisition module for real-time collecting the operating data of the equipment from the thermal power generation equipment, including boiler temperature, boiler pressure, fuel flow rate, oxygen concentration, flue gas emission concentration, and load; A data preprocessing module for performing normalization processing on the collected operating data; A feature extraction module for extracting the key features of the operation of thermal power generation equipment from the normalized data through principal component analysis; A feature selection module for screening out the key features that have a significant impact on optimization according to the feature selection method; A deep learning module for inputting the extracted features into the deep learning model for training to predict the future operating status and performance changes of thermal power generation equipment; An optimization module for constructing a multi-objective optimization function based on the prediction results of the deep learning model, and the optimization objectives include minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency; A control module for adjusting the operating parameters of thermal power generation equipment according to the optimized control signal, including fuel flow rate, air supply volume, and boiler load; A model update module for regularly updating the deep learning model and retraining the model based on new data to ensure the real-time nature and effectiveness of the control signal.

[0016] A further improvement of the present invention lies in that the deep learning module includes: A CNN layer for extracting spatial features from the operating data of thermal power generation equipment; An LSTM layer for modeling the time series features of the operation data of thermal power generation equipment; A fully connected layer for fusing the extracted features and predicting the future operation status and performance changes of thermal power generation equipment.

[0017] Compared with the prior art, the present invention has at least the following beneficial technical effects: The present invention uses CNN to extract the spatial features of the equipment, and realizes the learning of complex spatial relationships through convolutional layers and pooling layers. At the same time, LSTM is used to model the time series features, learn the timing characteristics of the equipment, and use these features to predict the future operation status of the equipment. This method combines spatial features and time series features, can capture complex dynamic laws in high-dimensional data, the deep learning model can comprehensively consider various dynamic factors of the equipment, accurately predict the future state of the equipment, and enhance the prediction accuracy and response ability of the system; and by constructing a multi-objective optimization function based on the prediction results of the deep learning model, the optimization objectives include minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency. The optimization function uses the weighted summation method to uniformly process multiple optimization objectives, achieving a balance between energy efficiency and environmental protection, reducing pollution emissions while reducing fuel consumption, and improving equipment efficiency.

[0018] The present invention obtains the operation data of thermal power generation equipment in real time, including but not limited to parameters such as equipment temperature, pressure, flow rate, load, and energy efficiency. These real-time data provide a basis for subsequent data processing and training of the deep learning model. Accurate data collection ensures the accuracy and reliability of the subsequent optimization process. The present invention can reflect the equipment status in real time, avoiding the problems of data lag or inaccuracy in traditional monitoring systems, and improving the timeliness and accuracy of control decisions.

[0019] The present invention uses principal component analysis to extract key features from the standardized equipment data, including equipment performance, energy efficiency, pollutant emissions, etc.; principal component analysis removes redundant features through dimensionality reduction and focuses on the features that have the greatest impact on equipment performance. Through feature selection methods, key features that have a significant impact on the optimization control signal are screened out. The feature selection process improves the computational efficiency of the model, avoids interference from irrelevant data, reduces computational complexity and improves prediction accuracy, ensuring that the deep learning model can also achieve efficient operation with limited computational resources. Description of the Drawings

[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0021] Figure 1 It is a flowchart of the control signal optimization method of the present invention; Figure 2 It is an overall block diagram of the control signal optimization system of the present invention. Specific Embodiments

[0022] In the following text, only some exemplary embodiments are briefly described. As those skilled in the art can recognize, the described embodiments can be modified in various different ways without departing from the spirit or scope of the present invention. Therefore, the drawings and the description are considered to be exemplary in nature rather than restrictive.

[0023] Structural schematic diagrams according to the disclosed embodiments of the present invention are shown in the drawings. These figures are not drawn to scale, and for the purpose of clear expression, some details are enlarged and some details may be omitted. The shapes of various regions and layers shown in the figures, as well as their relative sizes and positional relationships, are only exemplary. In practice, there may be deviations due to manufacturing tolerances or technical limitations, and those skilled in the art can design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0024] The following will describe the embodiments of the present invention in detail with reference to the drawings.

[0025] Embodiment 1 As Figure 1 shown, the control signal optimization method for thermal power generation equipment based on deep learning provided by the present invention includes the following steps: S1. Real-time collect the operation data of the thermal power generation equipment, where the operation data includes boiler temperature, boiler pressure, fuel flow rate, oxygen concentration, flue gas emission concentration, and load.

[0026] S2. Perform data preprocessing on the collected operation data, including denoising, filling missing data, and normalization processing.

[0027] The data preprocessing of the operation data specifically includes the following steps: S21. Apply a band-pass filter to the collected operation data for denoising; In the original operation data, the readings of thermal power generation equipment may be affected by environmental changes (such as temperature, humidity) or sensor failures, resulting in noise. Low-pass filters or smoothing methods (such as moving average) can be used to remove high-frequency noise and retain the true operation trend of the equipment.

[0028] S22. Perform polynomial interpolation filling on the missing data to ensure data continuity and consistency; There may be missing values in the operation data of thermal power generation equipment due to equipment failures, sensor problems, etc. In this step, the missing values can be filled using the following methods: filling with the mean or median of the historical data of the thermal power generation equipment, filling with forward filling or backward filling methods, or using interpolation methods such as linear interpolation to fill the missing values.

[0029] At the same time, outliers usually lead to inaccurate predictions and can be processed through the following methods: Z-Score method: Identify data points with too large a difference from the mean based on the mean and standard deviation of the data.

[0030] Box plot method: Use box plots to detect outliers beyond the upper or lower limits, and values outside the range need to be further checked or removed.

[0031] S23. Use standardization processing to normalize the real-time data so that the mean of each feature is zero and the standard deviation is one.

[0032] The purpose of normalization is to standardize and normalize data with different dimensions and different units of measurement to ensure that the data is within the same scale range so that deep learning models can effectively process and train.

[0033] First, perform standardization by adjusting the mean of each feature to 0 and the standard deviation to 1. The formula is as follows:

[0034] Among them, is the value after standardization, is the original data value, is the mean of the feature, is the standard deviation of the feature.

[0035] Then perform normalization by scaling the data values to a specific range (usually [0, 1]). The formula is as follows:

[0036] Among them, is the value after normalization, is the original data value, and They are the minimum and maximum values of the feature data respectively.

[0037] Standardization and normalization ensure that different features are on the same scale, avoiding some features dominating in model training due to large dimensions, thus improving the training efficiency and accuracy of deep learning models.

[0038] S3. Extract the key features of the operation of thermal power generation equipment from the preprocessed operation data through principal component analysis, including equipment performance, energy efficiency, and pollution emissions, and screen out the key features that have a significant impact on optimization according to the feature selection method.

[0039] Step S3 specifically includes the following steps: S31. Reduce the dimension of the preprocessed operation data through principal component analysis. The principal component analysis steps include calculating the covariance matrix of the data, performing eigenvalue decomposition, and selecting the components with the highest proportion of explained variance as the principal components; Principal component analysis maps the data into a new feature space through linear transformation. The goal is to concentrate the variance of most data on a few principal components, reduce the dimension of the data by selecting the first few principal components, and retain most of the information at the same time. The specific process is as follows: Calculate the covariance matrix of the standardized data; Through eigenvalue decomposition of the covariance matrix, obtain the eigenvalues and eigenvectors; Sort the eigenvalues from large to small and select the first few eigenvectors as the principal components.

[0040] S32. Based on the results of principal component analysis, use the feature selection algorithm to screen the principal components. The feature selection algorithm is based on information gain, mutual information, or correlation coefficient indicators to screen out the key features that have a significant impact on the optimization of thermal power generation equipment; On the basis of dimension reduction by principal component analysis, use the feature selection method (such as the feature selection algorithm based on importance scoring) to screen out the key features that have a significant impact on the optimization of thermal power generation equipment. The following steps can be adopted: Calculate the proportion of explained variance of each principal component and select the principal components with a high proportion of explained variance; Further screen out the features that have a significant impact on the optimization objectives (such as energy efficiency, pollution emissions, equipment performance, etc.) through methods based on information gain, mutual information, correlation coefficient, etc.

[0041] S33. Output the set of screened key features, including boiler temperature, boiler pressure, fuel flow rate, emission concentration, and energy efficiency.

[0042] The finally obtained feature set will be the key features that have a significant impact on the optimization effect of thermal power generation equipment, including but not limited to boiler temperature, boiler pressure, fuel flow, emission concentration, energy efficiency, etc.

[0043] S4. Use CNN to extract the spatial features of the equipment data, use LSTM to model the time series features, and input the extracted spatial features and time series features into the deep learning model for training. The trained deep learning model can predict the future operating status and performance changes of the thermal power generation equipment according to the input data. The training of the deep learning model specifically includes the following steps: S41. Use the CNN model to extract the spatial features of the preprocessed operating data. Take the preprocessed operating data as the input of the CNN convolutional layer, and use the convolution operation to extract the spatial features in the standardized real-time data. The convolution operation extracts different spatial features through multiple convolutional kernels. First, collect the multi-dimensional data of the thermal power generation equipment, including equipment performance (such as temperature, pressure, etc.), energy efficiency (such as fuel consumption, power generation efficiency, etc.), pollution emissions (such as CO2, NOx emission concentration, etc.), and real-time data of the equipment operating status (such as rotation speed, load, etc.). Standardize the collected equipment data using the Z-score standardization method to ensure the comparability between different data dimensions and eliminate the influence of dimensions.

[0044] Take the standardized equipment data as the input of CNN, and use the convolution operation to extract the spatial features in the equipment data. For example, when processing data such as temperature and pressure, the convolutional kernel can capture the relationships between different regions or different equipment parameters. The formula for the convolution operation is:

[0045] where is the convolutional kernel, which is used to perform convolution calculation on the input data The size of the convolutional kernel will affect the type of features extracted. is the input data, is the output feature map, and are the sizes of the convolutional kernel, which are used to define the region range for feature extraction. and are the positions in the feature map.

[0046] Through multiple convolutional kernels of different sizes, CNN can extract different spatial features in the data, such as the correlations between various components of the thermal power generation equipment and the spatial distribution of the equipment operating status.

[0047] S42. Downsample the feature map output by the convolutional layer through a pooling layer. The pooling operation includes max pooling or average pooling; After the convolution operation, downsample through a pooling layer (such as max pooling or average pooling) to reduce the computational amount and retain important features. The pooling operation is as follows:

[0048] Among them, is the position in the pooled feature map At this point, the pooling operation takes the maximum value or average value within a certain area of the convolution.

[0049] Flatten the pooled feature map into a one-dimensional vector and further extract features through a fully connected layer.

[0050] S43. Input the time series features of the thermal power generation equipment into the LSTM and use the LSTM to model the extracted time series features, and predict the future operating state by learning the temporal pattern of the historical data of the thermal power generation equipment; The LSTM can effectively process time series data and capture the temporal changes during equipment operation. The LSTM is a neural network that can effectively learn long-term dependencies and can model the changes in input data in the time dimension. In this step, input the time series data of the equipment into the LSTM network, and predict the future operating state by learning the temporal pattern of the equipment's historical data. The calculation formula of the LSTM includes: Input gate :

[0051] Among them, is the activation value of the input gate, used to control the importance of the current input information, is the weight matrix of the input gate, is the hidden state of the previous moment, is the input at the current moment (such as the current parameters of the equipment), is the bias term of the input gate, is the Sigmoid activation function, used to output a value between 0 and 1, indicating the degree of information retention.

[0052] Forget gate :

[0053] Among them, is the activation value of the forget gate, used to control the degree of forgetting of the memory state of the previous moment, is the weight matrix of the forget gate, is the bias term of the forget gate.

[0054] Output gate :

[0055] Among them, is the activation value of the output gate, used to control the output information at the current moment, is the weight matrix of the output gate, is the bias term of the output gate.

[0056] Cell state :

[0057] Among them, is the cell state at the current moment, storing long-term memory, is the cell state at the previous moment, is the weight matrix of the cell state, is the bias term of the cell state.

[0058] Hidden state :

[0059] Among them, is the hidden state at the current moment, representing the output of the LSTM network, is the hyperbolic tangent function, used for non-linear mapping of the cell state.

[0060] S44. Combine the spatial features extracted by CNN and the time series features extracted by LSTM into a deep learning model, use the backpropagation algorithm to train and adjust the model weights, and improve the prediction accuracy by optimizing the loss function.

[0061] Combine the spatial features extracted by CNN and the time series features modeled by LSTM into a deep learning model for training. During the training process, use the backpropagation algorithm to adjust the model weights and improve the prediction accuracy by minimizing the loss function. The loss function can choose the mean squared error (MSE):

[0062] Among them, is the actual value, that is, the true operating state or performance of the thermal power generation equipment, is the predicted value of the deep learning model, that is, the predicted future operating state of the thermal power generation equipment according to the input data, is the number of samples, representing the number of training data, is the sample index.

[0063] After training, the validation set and the test set are used to evaluate the model, and evaluation metrics such as prediction accuracy and error are calculated. The trained model can predict the future operating status and performance changes of the thermal power generation equipment based on new input data (device parameters monitored in real time), and provide corresponding optimal control signals.

[0064] S5. Construct a multi-objective optimization function based on the prediction results of the deep learning model. The optimization objectives include minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency. Step S5 specifically includes the following steps: S51. Through the trained deep learning model, predict the operating status and performance of the thermal power generation equipment to obtain the future status of each thermal power generation equipment, including fuel consumption, pollutant emissions, and equipment efficiency. S52. Based on the results of fuel consumption, pollutant emissions, and equipment efficiency predicted by the deep learning model, construct a multi-objective optimization function. To achieve the optimization objectives, such as minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency, it is necessary to design a multi-objective optimization function. The goal of this function is to consider these three factors simultaneously and balance them. The weighted sum method can be used to construct the optimization function:

[0065] Among them, represents the multi-objective optimization function value. 、 and are the weight coefficients for minimizing fuel consumption, minimizing pollution emissions, and maximizing equipment efficiency respectively. They are used to adjust the importance of different optimization objectives to meet the requirements in different situations. is the predicted value of the fuel consumption of the thermal power generation equipment, indicating the amount of fuel consumed during the operation of the thermal power generation equipment. It is predicted by the deep learning model and reflects the energy use efficiency of the thermal power generation equipment. is the predicted value of the pollutant emissions of the thermal power generation equipment, indicating the amount of pollutants (such as CO2, NOx, etc.) emitted by the equipment. Through the prediction of the deep learning model, the impact of the operation of the thermal power generation equipment on the environment can be understood. represents the predicted value of the equipment efficiency, indicating the energy efficiency of the thermal power generation equipment, that is, the energy input required per unit of energy output. The larger this value is, the higher the efficiency of the thermal power generation equipment, and the fuel consumption and pollution emissions can be reduced. S53. Use an optimization algorithm to balance the optimization objectives and generate the optimal control parameters to minimize fuel consumption, minimize pollutant emissions, and maximize equipment efficiency, so as to achieve the best operation strategy of the thermal power generation equipment.

[0066] Optimization algorithms (such as genetic algorithms, particle swarm optimization, etc.) will calculate the optimization scheme according to the above objective function, and minimize and by adjusting the input parameters, and maximize , so as to achieve the efficient, low-pollution and low-energy-consumption operation of thermal power generation equipment.

[0067] The optimal parameter combination obtained through multi-objective optimization will be used in the control system to automatically adjust the operating parameters of the equipment (such as fuel flow, load, steam temperature, etc.) to achieve the optimization goal.

[0068] S6. Generate control signals according to the results of the multi-objective optimization function and adjust the operating parameters of the thermal power generation equipment, including fuel flow, air supply, and boiler load; Step S6 specifically includes the following steps: S61. Based on the results of the multi-objective optimization function, generate control signals for adjusting the operating parameters of the equipment, including , and , where is the fuel flow, is the air supply, is the boiler load, , and are control algorithms that map the multi-objective optimization results into specific operating parameters; Based on the foregoing steps, specific control signals are generated according to the optimization objective results calculated by the multi-objective optimization function. These control signals are used to adjust the operating parameters of the thermal power generation equipment to achieve the optimization goal, including fuel flow, air supply, and boiler load, etc.

[0069] The optimization results will be converted into actual control signals through control algorithms (such as PID control, fuzzy control or other appropriate control strategies). These control signals will be used to adjust the following operating parameters respectively: Fuel flow : Control the fuel input of the equipment to ensure the minimum fuel consumption in the optimization goal.

[0070] Air supply : Adjust the air input of the boiler to optimize the combustion process, thereby improving efficiency and reducing pollutant emissions.

[0071] Boiler load : Control the load of the boiler to ensure that a suitable power generation output can be maintained at the maximum equipment efficiency.

[0072] ​S62. Use a control algorithm to apply the generated control signal to the control module of the thermal power generation equipment, and adjust the operating parameters of the thermal power generation equipment to minimize fuel consumption, minimize pollutant emissions, and maximize equipment efficiency; After obtaining the specific control signal, the control system will transmit it to each control module of the thermal power generation equipment. According to the control signal, the operating parameters of the equipment will be adjusted accordingly to ensure that the equipment operates in an optimized state, achieving the goals of minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency.

[0073] S63. Through a real-time feedback control system, monitor the operating state of the adjusted thermal power generation equipment, and transmit the feedback signal back to the optimization module for the next round of optimization, so as to ensure that the thermal power generation equipment is always in the optimal operating state.

[0074] S7. Regularly update the deep learning model and retrain the model based on new data to ensure the real-time performance and effectiveness of the control signal.

[0075] Step S7 specifically includes the following steps: S71. Regularly obtain the latest data on the operation of the equipment, including but not limited to parameters such as fuel flow rate, air supply volume, and boiler load, as training data; S72. Preprocess the obtained latest data, perform standardization and denoising operations to ensure the quality and consistency of the data; S73. Retrain the deep learning model using the latest data, where the deep learning model is used to predict the operating state of the thermal power generation equipment and adjust the internal parameters to adapt to the changes of the thermal power generation equipment; S74. Evaluate the performance of the trained deep learning model, use an independent validation data set to test the prediction accuracy of the deep learning model and the effectiveness of the control signal. If the performance is insufficient, adjust the deep learning model structure or training parameters; S75. Replace the original deep learning model with the newly trained deep learning model, and apply the updated deep learning model to the control system of the thermal power generation equipment to generate new control signals to adjust the operating parameters of the thermal power generation equipment; S76. During the entire update process, through incremental learning or other efficient algorithms, ensure the real-time performance of the control signal generation process, reduce the delay during the update process, and ensure that the thermal power generation equipment always operates in the optimal state.

[0076] Embodiment 2 As Figure 2 shown, the control signal optimization system for thermal power generation equipment based on deep learning provided by the present invention includes: A data acquisition module, which is used to collect the operation data of the equipment in real time from the thermal power generation equipment, including boiler temperature, boiler pressure, fuel flow rate, oxygen concentration, flue gas emission concentration, and load.

[0077] The data acquisition module includes: A sensor network, which is used to monitor various operation parameters of the thermal power generation equipment in real time; A data communication module, which is used to transmit the collected data of the thermal power generation equipment to the data preprocessing module for subsequent processing.

[0078] A data preprocessing module, which is used to perform standardization processing on the collected operation data.

[0079] A feature extraction module, which is used to extract the key features of the operation of the thermal power generation equipment from the standardized data through principal component analysis.

[0080] The feature extraction module uses the principal component analysis method for feature extraction, including: Performing dimensionality reduction processing on the standardized data, and extracting key features reflecting equipment performance, energy efficiency, and pollution emissions, etc.; According to the principal component analysis results, screening out the most representative principal components and sending them as input features to the deep learning module.

[0081] A feature selection module, which is used to screen out the key features that have a significant impact on optimization according to the feature selection method.

[0082] A deep learning module, which is used to input the extracted features into the deep learning model for training, and predict the future operation status and performance changes of the thermal power generation equipment.

[0083] In this embodiment, the deep learning module includes: A CNN layer, which is used to extract spatial features from the operation data of the thermal power generation equipment; An LSTM layer, which is used to model the time series features of the operation data of the thermal power generation equipment; A fully connected layer, which is used to fuse the extracted features and predict the future operation status and performance changes of the thermal power generation equipment.

[0084] An optimization module, which is used to construct a multi-objective optimization function based on the prediction results of the deep learning model, and the optimization objectives include minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency.

[0085] In this embodiment, the optimization module includes: A multi-objective optimization function, which is used to optimize fuel consumption, pollutant emissions, and equipment efficiency according to the prediction results; An optimization algorithm, which is used to solve the optimal control strategy in the multi-objective optimization function; A control signal generation unit, configured to generate a control signal according to the optimization result and transmit the control signal to the control module.

[0086] A control module, configured to adjust the operating parameters of the thermal power generation equipment according to the optimized control signal, including fuel flow rate, air supply amount, and boiler load.

[0087] In this embodiment, the control module includes: A fuel flow regulator, configured to adjust the fuel flow rate of the thermal power generation equipment; An air supply amount regulator, configured to adjust the air supply amount of the thermal power generation equipment; A boiler load regulator, configured to adjust the boiler load of the thermal power generation equipment.

[0088] A model update module, configured to update the deep learning model regularly and retrain the model based on new data to ensure the real-time performance and effectiveness of the control signal.

[0089] In this embodiment, the model update module includes: A data receiving unit, configured to collect new operating data from the equipment; An incremental learning unit, configured to update the deep learning model based on the new data; A model evaluation unit, configured to evaluate the prediction effect of the updated model and determine whether to retrain.

[0090] The above has shown and described the basic principles, main features, and advantages of the present invention. For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic features of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be encompassed within the present invention. Any reference signs in the claims should not be regarded as limiting the claims involved.

[0091] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment only includes an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and 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 to illustrate the technical idea of the present invention and cannot be used to limit the protection scope of the present invention. Any modification made on the basis of the technical solution according to the technical idea proposed by the present invention falls within the protection scope of the claims of the present invention.

Claims

1. A control signal optimization method for thermal power generation equipment based on deep learning, characterized in that: The following steps are involved: Real-time collection of operating data of thermal power generation equipment, including boiler temperature, boiler pressure, fuel flow, oxygen concentration, flue gas emission concentration and load; Perform data preprocessing on the collected operating data, including denoising, missing data filling and standardization; The operation characteristics of thermal power generation equipment are extracted from the preprocessed operation data through principal component analysis, including equipment performance, energy efficiency and pollution emissions, and the key features that have a significant impact on optimization are screened out based on the feature selection method; Use CNN to extract the spatial features of the preprocessed operating data, use LSTM to model the time series features, and input the extracted spatial features and time series features into the deep learning model for training. The trained deep learning model can improve the prediction accuracy by optimizing the loss function according to the future operating status and performance changes of the thermal power generation equipment based on the input data; Construct a multi-objective optimization function based on the prediction results of the deep learning model. The optimization objectives include minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency. Generate control signals and adjust the operating parameters of thermal power generation equipment, including fuel flow, air supply, and boiler load, based on the results of the multi-objective optimization function; Update deep learning models regularly and retrain them based on new data.

2. The control signal optimization method for thermal power generation equipment based on deep learning according to claim 1 is characterized in that: Perform data preprocessing on the collected operation data, including: Apply a bandpass filter to the collected operating data for denoising; Perform polynomial interpolation to fill missing data to ensure data continuity and consistency; The real-time data is normalized using standardization so that each feature has a mean of zero and a standard deviation of one.

3. The control signal optimization method for thermal power generation equipment based on deep learning according to claim 1, characterized in that: The characteristics of thermal power generation equipment operation are extracted from the preprocessed operation data through principal component analysis. The characteristics include equipment performance, energy efficiency and pollution emissions. The key features that have a significant impact on optimization are screened out according to the feature selection method, including: The preprocessed running data is reduced in dimension by principal component analysis. The principal component analysis steps include calculating the covariance matrix of the data, performing eigenvalue decomposition and selecting the component with the highest explained variance as the principal component; Based on the results of principal component analysis, the principal components are screened using a feature selection algorithm. The feature selection algorithm is based on information gain, mutual information or correlation coefficient indicators to screen out key features that have a significant impact on the optimization of thermal power generation equipment; Outputs a filtered set of key features including boiler temperature, boiler pressure, fuel flow, emission concentration, and energy efficiency.

4. The method for optimizing control signals of thermal power generation equipment based on deep learning according to claim 1, characterized in that: Use CNN to extract the spatial features of the preprocessed running data, use LSTM to model the time series features, and input the extracted spatial features and time series features into the deep learning model for training, including: Use the CNN model to extract spatial features from the preprocessed running data. The preprocessed running data is used as the input of the CNN convolution layer. The convolution operation is used to extract spatial features from the standardized real-time data. The convolution operation extracts different spatial features through multiple convolution kernels. The feature map output by the convolutional layer is processed by the pooling layer for dimensionality reduction. The pooling operation includes maximum pooling or average pooling. The time series features of thermal power generation equipment are input into LSTM and the extracted time series features are modeled using LSTM. The future operating status is predicted by learning the time series pattern of the historical data of thermal power generation equipment. The spatial features and time series features extracted by CNN and LSTM are combined into a deep learning model. The deep learning model is trained using the back propagation algorithm to adjust the model weights, and the prediction accuracy is improved by optimizing the loss function.

5. The method for optimizing control signals of thermal power generation equipment based on deep learning according to claim 4, characterized in that: The formula of the loss function is: in, is the actual value, that is, the actual operating status or performance of the thermal power generation equipment. is the predicted value of the deep learning model, that is, the future operating status of the thermal power generation equipment predicted based on the input data. is the number of samples, indicating the number of training data, is the sample index.

6. The control signal optimization method for thermal power generation equipment based on deep learning according to claim 1, characterized in that: Construct a multi-objective optimization function based on the prediction results of the deep learning model, including: Through the trained deep learning model, the operating status and performance of thermal power generation equipment are predicted to obtain the future status of each thermal power generation equipment, including fuel consumption, pollutant emissions and equipment efficiency; Based on the fuel consumption, pollutant emissions and equipment efficiency results predicted by the deep learning model, a multi-objective optimization function is constructed. The formula of the multi-objective optimization function is: in, represents the multi-objective optimization function, , and are the weight coefficients for minimizing fuel consumption, minimizing pollution emissions, and maximizing equipment efficiency, respectively. is the predicted value of fuel consumption of thermal power generation equipment, is the predicted value of pollutant emissions from thermal power generation equipment, represents the predicted value of equipment efficiency; The optimization algorithm is used to weigh the optimization objectives and generate the optimal control parameters to minimize fuel consumption, minimize pollutant emissions, and maximize equipment efficiency, thereby achieving the best operating strategy for thermal power generation equipment.

7. The method for optimizing control signals of thermal power generation equipment based on deep learning according to claim 1, characterized in that: Based on the results of the multi-objective optimization function, control signals are generated and the operating parameters of the thermal power generation equipment are adjusted, including: Based on the results of the multi-objective optimization function, control signals are generated to adjust the operating parameters of the equipment, including , and ,in is the fuel flow rate, is the air supply, is the boiler load, , and For the control algorithm, the multi-objective optimization results are mapped into specific operating parameters; Using a control algorithm, the generated control signal is applied to a control module of the thermal power generation equipment to adjust the operating parameters of the thermal power generation equipment so that the equipment achieves the goals of minimizing fuel consumption, minimizing pollutant emissions and maximizing equipment efficiency; Through the real-time feedback control system, the operating status of the adjusted thermal power generation equipment is monitored, and the feedback signal is transmitted back to the optimization module for the next round of optimization, thereby ensuring that the thermal power generation equipment is always in the optimal operating state.

8. The method for optimizing control signals of thermal power generation equipment based on deep learning according to claim 1, characterized in that: Regularly update deep learning models and retrain them based on new data, including: Regularly obtain the latest data on equipment operation, including fuel flow, air supply, and boiler load, as training data; Preprocess the latest data obtained, perform standardization and denoising operations to ensure data quality and consistency; Retraining the deep learning model using the latest data, where the deep learning model is used to predict the operating status of thermal power generation equipment and adjust internal parameters to adapt to changes in thermal power generation equipment; Evaluate the performance of the trained deep learning model, use an independent validation dataset to test the prediction accuracy of the deep learning model and the effectiveness of the control signal, and adjust the deep learning model structure or training parameters if the performance is insufficient; Replacing the original deep learning model with a newly trained deep learning model, and applying the updated deep learning model to the control system of the thermal power generation equipment to generate a new control signal to adjust the operating parameters of the thermal power generation equipment; During the entire update process, incremental learning or other efficient algorithms are used to ensure the real-time nature of the control signal generation process, reduce delays in the update process, and ensure that thermal power generation equipment always operates in the optimal state.

9. A control signal optimization system for thermal power generation equipment based on deep learning, characterized in that: include: Data acquisition module, used to collect real-time operation data of equipment from thermal power generation equipment, including boiler temperature, boiler pressure, fuel flow, oxygen concentration, flue gas emission concentration and load; Data preprocessing module, used to standardize the collected operation data; A feature extraction module is used to extract key features of the operation of thermal power generation equipment from the standardized data through principal component analysis; Feature selection module, used to select key features that have a significant impact on optimization according to feature selection methods; A deep learning module is used to input the extracted features into a deep learning model for training and predict the future operating status and performance changes of thermal power generation equipment; The optimization module is used to build a multi-objective optimization function based on the prediction results of the deep learning model. The optimization objectives include minimizing fuel consumption, minimizing pollutant emissions, and maximizing equipment efficiency; A control module for adjusting operating parameters of the thermal power generation equipment, including fuel flow, air supply and boiler load, according to the optimized control signal; The model update module is used to regularly update the deep learning model and retrain the model based on new data to ensure the real-time and effectiveness of the control signal.

10. The control signal optimization system for thermal power generation equipment based on deep learning according to claim 9, characterized in that: The deep learning modules include: CNN layer, used to extract spatial features from the operating data of thermal power generation equipment; LSTM layer, used to model the time series characteristics of thermal power generation equipment operation data; The fully connected layer is used to fuse the extracted features and predict the future operating status and performance changes of thermal power generation equipment.

Images