An intelligent control system and method for power plant combustion power generation

Abstract

The present application relates to the technical field of thermal power generation, in particular to a power plant combustion power generation intelligent control system and method, the method comprising: collecting combustion power generation control, coal quality parameters and power generation data, constructing a time series data set, constructing and training a combustion power generation control model; obtaining power plant historical load and meteorological data, constructing a load data set, and training a power plant power generation demand prediction model based on a long short-term memory neural network (LSTM) model; combining the power generation demand prediction result and the combustion control model to solve the optimal combustion power generation control data, and controlling and adjusting the boiler combustion during power generation through a PID adjustment mechanism. The present application realizes intelligent management of power plant combustion power generation through prediction and control adjustment, and can improve the power generation efficiency of the power plant.

Description

Technical Field

[0001] This invention relates to the field of thermal power generation technology, specifically to an intelligent control system and method for power plant combustion power generation. Background Technology

[0002] With industrial development and the continuous growth of electricity demand, improving the operating efficiency and economy of thermal power plants has become an important issue for the power industry. Traditional boiler combustion control mainly relies on the operator's experience and simple feedback control, which is difficult to adapt to the complex operating conditions caused by coal quality fluctuations and load changes.

[0003] Generalized Regressive Neural Networks (GRNNs) are neural network models based on nonparametric regression. Trained on sample data, they establish a nonlinear mapping relationship between input and output variables. GRNN models are simple in structure, train quickly, and possess good generalization ability, and have been successfully applied in multiple fields. Applying GRNNs to combustion power generation control can accurately predict power generation performance based on real-time collected combustion control parameters, coal quality parameters, and power generation data, providing reliable model support for optimized control.

[0004] On the other hand, power plant load forecasting is a crucial aspect of power dispatch and operation optimization. Accurately predicting future electricity demand is essential for rationally allocating generating capacity, reducing reserve capacity, and improving economic efficiency. Traditional load forecasting methods, primarily based on statistical models and expert experience, struggle to adapt to the nonlinear and non-stationary characteristics of loads. In recent years, deep learning technology has made significant progress in time series forecasting. Long Short-Term Memory (LSTM) neural networks, a special type of recurrent neural network, can effectively learn and store long-term dependency information through the introduction of gating mechanisms, offering unique advantages in modeling and predicting time-series data. Applying LSTM to power plant load forecasting can fully leverage the temporal characteristics and external influencing factors (such as weather conditions) within load data, enabling accurate predictions of future electricity demand.

[0005] Boiler combustion control and load forecasting in thermal power plants are key issues for improving the efficiency and economy of power system operation. Applying advanced artificial intelligence technologies, such as generalized regression neural networks and long short-term memory neural networks, to boiler combustion power generation control and power plant load forecasting can significantly improve the system's intelligence level and optimization performance.

[0006] In view of this, the present invention proposes an intelligent control system and method for power plant combustion power generation. Summary of the Invention

[0007] To achieve the above objectives, the present invention provides an intelligent control system and method for power plant combustion power generation, the specific technical solution of which is as follows:

[0008] A smart control method for combustion power generation in power plants includes:

[0009] Collect power plant combustion power generation control data, obtain coal quality parameters and power generation data, and merge the combustion power generation control data, coal quality parameters and power generation data to obtain an augmented time series dataset;

[0010] A combustion power generation control model was constructed and trained based on the generalized regression neural network (GRNN) model.

[0011] Historical power load data of the power plant and meteorological data of the same period are obtained and merged with the load data to obtain an augmented power plant load dataset.

[0012] Based on the Long Short-Term Memory (LSTM) neural network model, a power plant power generation demand prediction model is constructed and trained to predict the power plant power generation demand in future time periods.

[0013] Solve the control data of power plant combustion power generation, construct a PID regulation mechanism for power generation combustion, and control combustion power generation.

[0014] Preferably, combustion power generation control data is collected, including coal feed rate of coal feeder, air volume of primary air fan and air volume of secondary air fan, and the sampling period is set to ; outlier removal and normalization processing are performed on the combustion power generation control data to obtain preprocessed combustion power generation control data;

[0015] Collect coal quality parameters, including lower heating value, total sulfur content, and volatile matter; collect power generation data;

[0016] The preprocessed combustion power generation control data, coal quality parameters, and power generation output are aligned and merged according to timestamps to obtain an augmented time series dataset.

[0017] Preferably, a generalized regression neural network (GRNN) model is constructed, which includes an input layer, a pattern layer, a summation layer, and an output layer; the input layer is used to input data, and the number of samples in the pattern layer corresponds to the number of neurons;

[0018] Based on the generalized regression neural network (GRNN) model, using the dataset as training samples, the GRNN model is trained as a combustion power generation control model.

[0019] Preferably, historical load data of the power plant and meteorological data of the same period are collected, wherein the meteorological data includes temperature and humidity; the historical load data of the power plant and the meteorological data of the same period are aligned and merged according to timestamps to obtain an augmented load dataset;

[0020] Based on the Long Short-Term Memory (LSTM) neural network model, the load dataset is processed through a sliding time window to construct training samples and train the power plant load prediction model.

[0021] Preferably, the sliding time window width and prediction step size are set, and samples are extracted from the load dataset using the sliding time window. Each sample includes multiple consecutive load values ​​and corresponding meteorological data as input, and future load values ​​as output.

[0022] The load dataset is obtained by extracting the load dataset through a sliding time window, and then the load dataset is classified into training set, validation set and test set.

[0023] Preferably, a Long Short-Term Memory (LSTM) neural network model is constructed to train the power plant load prediction model;

[0024] The Long Short-Term Memory (LSTM) neural network model includes an LSTM input layer, an LSTM hidden layer, and an LSTM output layer; the LSTM input layer is used to receive input data; the output layer includes a fully connected layer to predict the load value for future time steps.

[0025] Preferably, the training steps of the Long Short-Term Memory (LSTM) neural network model are as follows:

[0026] Input the training set data of the load dataset into the LSTM model for training; set the batch size and number of training epochs; evaluate the model performance using the validation set data after each training epoch; adjust the model structure and hyperparameters based on the validation set performance; use early stopping to prevent overfitting; evaluate the predictive performance of the trained LSTM model on the test set; calculate the root mean square error and mean absolute error evaluation metrics.

[0027] When the Long Short-Term Memory (LSTM) neural network model is trained, it is used as a power plant load prediction model to predict future power plant load values ​​for new input historical load values ​​and meteorological data.

[0028] Preferably, based on the power plant load forecasting model, the predicted power generation demand is substituted into the combustion power generation control model to solve for the power plant combustion power generation control data.

[0029] Using the trained power plant load forecasting model, input historical power plant load data and meteorological data, predict the power plant load for future time periods to obtain a load forecasting sequence;

[0030] Based on the load forecast sequence, the predicted load power value is input into the combustion power generation control model. According to the power generation demand and power generation combustion characteristics, the corresponding combustion power generation control data sequence is solved, including the data sequences of coal feed, primary air volume and secondary air volume.

[0031] Preferably, a PID controller is constructed to adjust the power generation and combustion control parameters; control variables for coal feed rate, primary air volume, and secondary air volume are set, and power generation is the controlled variable; and a PID controller structure is constructed.

[0032] By monitoring power generation and combustion status in real time, the actual values ​​are fed back to the PID controller. Based on the real-time feedback, the PID controller dynamically adjusts the power generation and combustion control parameters.

[0033] A smart control system for power plant combustion power generation, used to implement the aforementioned smart control method for power plant combustion power generation, includes: a first data acquisition module, a combustion power generation control module, a second data acquisition module, a power plant power generation demand prediction module, and a combustion control adjustment module;

[0034] The first data acquisition module collects power plant combustion power generation control data, obtains coal quality parameters and power generation data, and merges the combustion power generation control data, coal quality parameters and power generation data to obtain an augmented time series dataset;

[0035] The combustion power generation control module is based on the generalized regression neural network (GRNN) model, which is used to construct and train the combustion power generation control model.

[0036] The second data acquisition module acquires historical power load data of the power plant and meteorological data of the same period, and merges them with the load data to obtain an augmented power plant load dataset.

[0037] The power plant power generation demand forecasting module is based on the Long Short-Term Memory (LSTM) neural network model, which constructs and trains the power plant power generation demand forecasting model to predict the power plant power generation demand in future time periods.

[0038] The combustion control and regulation module solves the power plant combustion power generation control data, constructs a PID regulation mechanism for power generation combustion, and controls combustion power generation.

[0039] The beneficial effects of this invention are as follows: This invention provides a foundation for subsequent modeling and optimization by collecting real-time data on boiler combustion control; the preprocessing step can remove outliers in the data, improve data quality, and prepare for model training.

[0040] This invention constructs a dataset that comprehensively reflects the boiler's operating status by merging combustion power generation control data, coal quality parameters, and power generation data. The augmented time series dataset contains multiple key factors affecting boiler performance, laying the foundation for establishing an accurate combustion power generation control model.

[0041] This invention utilizes the powerful nonlinear mapping capability of the GRNN model to establish a complex relationship between boiler combustion control parameters and power generation performance. By training the GRNN model, power generation efficiency and output can be accurately predicted based on real-time combustion power generation control data and coal quality parameters, providing a basis for optimized control.

[0042] This invention establishes a dataset reflecting the characteristics of electricity demand by collecting historical power plant load data and corresponding meteorological data. Incorporating meteorological factors into the load dataset can more comprehensively depict the load change pattern and improve the accuracy of load forecasting.

[0043] This invention leverages the advantages of LSTM models in modeling time-series data to learn the long-term dependence and trend characteristics of power plant loads; by training the LSTM model, it is possible to accurately predict the power generation demand of power plants in the future.

[0044] This invention combines power plant demand forecasting results with a combustion power generation control model to achieve synergistic optimization of predictive regulation and feedback control. By solving for the optimal combustion power generation control data, it improves combustion efficiency and reduces energy consumption and emissions while meeting power generation demand. The invention also introduces a PID control mechanism to perform real-time feedback control of the boiler combustion process, ensuring that the boiler operates smoothly near the optimized setpoint and improving the robustness and stability of the control. Attached Figure Description

[0045] Figure 1 This is a flowchart of an intelligent control method for combustion power generation in a power plant according to the present invention;

[0046] Figure 2 This is a structural diagram of an intelligent control system for power plant combustion power generation according to the present invention. Detailed Implementation

[0047] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0048] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0049] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of this application. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0050] Example 1

[0051] Reference Figure 1 As a first embodiment of the present invention, a smart control method for power plant combustion power generation is provided, comprising:

[0052] S1: Collect combustion power generation control data from the power plant and preprocess the collected data; obtain coal quality parameters and power generation data; merge the combustion power generation control data, coal quality parameters, and power generation data to obtain an augmented time series dataset. .

[0053] Collect combustion power generation control data This includes the coal feeder's coal feed rate. Primary air volume and secondary air volume The sampling period is ;right Outlier removal and normalization are performed to obtain preprocessed combustion power generation control data. '.

[0054] Collect coal quality parameters The coal quality parameters Including low heat generation , total sulfur content and volatile matter Collect power generation data ; the pre-processed combustion power generation control data Coal quality parameters and power generation Align and merge according to timestamps to obtain an augmented time series dataset. Where t is time, within the sampling period. A total of N samples were collected, that is , For dataset The number of samples.

[0055] By collecting control data, coal quality parameters, and power generation data for combustion power generation, and performing preprocessing and merging, an augmented time series dataset comprehensively reflecting the combustion power generation process was obtained. This dataset contains multiple key factors affecting boiler combustion and power generation efficiency, laying a data foundation for subsequently building an accurate combustion power generation control model. Simultaneously, outlier removal and normalization were performed on the data, improving data quality and contributing to enhanced model performance and robustness.

[0056] S2: Based on the Generalized Regressive Neural Network (GRNN) model, using a dataset Using samples, a combustion power generation control model was constructed and trained.

[0057] A generalized regression neural network (GRNN) model is constructed, comprising an input layer, a pattern layer, a summation layer, and an output layer. The input layer is used to input data, and the pattern layer... The number of samples corresponds to One neuron.

[0058] Based on the Generalized Regression Neural Network (GRNN) model, with dataset Using training samples, a generalized regression neural network (GRNN) model is trained as a combustion power generation control model; the input to the GRNN model is... The output is combustion power generation control data. The mathematical expression for the GRNN model is:

[0059]

[0060] in, The input data for the t-th training sample is... For the newly entered number Input data for each training sample, For the corresponding power generation capacity, is the smoothing factor for GRNN.

[0061] Optimize the smoothing factor using cross-validation. The optimal GRNN combustion power generation model was obtained. .

[0062] Smoothing factor The performance of GRNN models is significantly affected and requires optimization through a data-driven approach. A common optimization method is cross-validation, with the following specific steps:

[0063] The training sample set is randomly divided into K mutually exclusive subsets (usually K=5 or 10);

[0064] Each time, select K-1 subsets as the training set and the remaining 1 subset as the validation set;

[0065] For different σ values, train the GRNN model on the training set and calculate the model's performance metrics (such as mean squared error, MSE) on the validation set.

[0066] Repeat this process K times to obtain the average of the K performance metrics;

[0067] The optimal σ value for average performance is selected as the final smoothing factor.

[0068] A combustion power generation control model was constructed and trained based on a generalized regressive neural network (GRNN) model, using an augmented time-series dataset as samples. The GRNN model effectively establishes a nonlinear mapping relationship between combustion control parameters, coal quality parameters, and power generation, fully utilizing the complex patterns inherent in the data. By optimizing the smoothing factor through cross-validation, the optimal GRNN combustion power generation model was obtained. This model can accurately predict power generation based on real-time combustion power generation control data and coal quality parameters, providing reliable model support for optimized control.

[0069] S3: Obtain historical power load data of the power plant, as well as meteorological data for the same period, including temperature and humidity, and merge them with the load data to obtain an augmented power plant load dataset D2.

[0070] Collect historical load data from power plants Meteorological data for the same period The meteorological data Including temperature and humidity ;Will and The dataset is merged and aligned according to timestamps to obtain the augmented load dataset. ,Right now ,in For dataset The number of samples.

[0071] Based on the Long Short-Term Memory (LSTM) neural network model, the load dataset is processed using a sliding time window. The data is processed to construct training samples and train the power plant load prediction model.

[0072] Set the sliding time window width and prediction step size to both k, and use the sliding time window to predict from the load dataset. Samples are extracted, and each sample includes k consecutive load values. and corresponding meteorological data As input, the load value for the next k steps As output, the sliding time window slides with a step size of 1 until it covers the entire historical load data, thus obtaining the load dataset. :

[0073]

[0074] in, The range is , load dataset The dataset is divided into training set, validation set, and test set.

[0075] An augmented power plant load dataset was constructed by acquiring historical power plant load data and concurrent meteorological data. A sliding time window was used to extract samples from the load dataset; each sample contains load and meteorological information within a specific time range, which helps capture the temporal characteristics of load changes. The load dataset was divided into training, validation, and test sets to prepare for training and evaluating the load forecasting model. The augmented load dataset comprehensively considers historical load trends and the influence of weather factors, providing rich information for building a high-quality load forecasting model.

[0076] S4: Based on the Long Short-Term Memory (LSTM) neural network model, construct and train a power plant power generation demand prediction model to predict the power plant power generation demand in future time periods.

[0077] A Long Short-Term Memory (LSTM) neural network model was constructed to train a power plant load prediction model.

[0078] The Long Short-Term Memory (LSTM) neural network model includes an LSTM input layer, an LSTM hidden layer, and an LSTM output layer; the LSTM input layer is used to receive signals of shape... The input data consists of k time steps, each containing load value, temperature, and humidity features; the LSTM hidden layer can be configured with one or more LSTM hidden layers as needed, and the number of neurons in each layer can be adjusted according to the complexity of the task; the output layer includes a fully connected layer with k neurons, used to predict the load value at the next k time steps.

[0079] The training steps for the Long Short-Term Memory (LSTM) neural network model are as follows:

[0080] Load dataset The training set data is input into the LSTM model for training; the batch size and number of training epochs are set simultaneously; after each training epoch, the model performance is evaluated using the validation set data; the model structure and hyperparameters are adjusted based on the validation set performance; early stopping is used to prevent overfitting; the predictive performance of the trained LSTM model is evaluated on the test set; and the root mean square error and mean absolute error are calculated as evaluation metrics.

[0081] Once the Long Short-Term Memory (LSTM) neural network model is trained, it is used as the power plant load forecasting model to predict the power plant load values ​​for the next k steps, given new input historical load values ​​and meteorological data. .

[0082] A power plant load forecasting model was constructed and trained based on a Long Short-Term Memory (LSTM) neural network. The LSTM model effectively learns and mines long-term dependencies and temporal features from load data, enabling accurate prediction of power generation demand over a future period. By setting appropriate LSTM model structure, training parameters, and early shutdown strategies, the model was trained on the training set, and its performance was monitored in real-time using a validation set to prevent overfitting. The resulting power plant load forecasting model can accurately predict future power plant demand based on historical load and meteorological data, providing a reliable target reference for optimizing combustion control.

[0083] S5: Based on the power generation demand prediction model of the power plant and the combustion power generation control model, solve the power plant combustion power generation control data, and at the same time construct a PID regulation mechanism for power generation combustion to control combustion power generation.

[0084] Based on the power plant load forecasting model, the predicted power generation demand is substituted into the combustion power generation control model to solve for the power plant combustion power generation control data.

[0085] Using the trained power plant load forecasting model, input historical power plant load data and meteorological data, predict future loads. The power plant load over a given time period is used to obtain the load forecast sequence:

[0086]

[0087] Based on the load forecast sequence, the predicted load power values ​​are input into the combustion power generation control model. According to the power generation demand and combustion characteristics, the corresponding combustion power generation control data sequence is solved, including the data sequences for coal feed rate, primary air volume, and secondary air volume. ;in:

[0088]

[0089]

[0090]

[0091] in, It is a coal feed rate control sequence. It is a primary air volume control sequence. It is a secondary air volume control sequence.

[0092] Construct a PID controller to adjust the power generation and combustion control parameters; set the control variable for coal feed rate as... The control variable for primary air volume is And the control variable for secondary air volume is Power generation It is the controlled variable.

[0093] The PID controller structure is as follows:

[0094]

[0095] in, To control the output of variables; Is the integral variable in Timing control error; It is the control error of the integral variable at time t; , and These are the gain coefficients for proportional, integral, and derivative operations, respectively.

[0096] By monitoring power generation and combustion status in real time, the actual values ​​are fed back to the PID controller. Based on the real-time feedback, the PID controller dynamically adjusts the power generation and combustion control parameters.

[0097] By combining power plant demand forecasting results with a combustion power generation control model, optimized combustion power generation control data, including control sequences for coal feed, primary air volume, and secondary air volume, is obtained. Simultaneously, a PID controller is introduced to adjust combustion control parameters in real time, dynamically optimizing the combustion process based on feedback from actual power generation. Through the coordinated optimization of predicted power generation demand and combustion control, boiler combustion efficiency can be maximized while meeting power demand. The introduction of the PID controller further improves the system's dynamic response performance and robustness, enabling boiler combustion control to quickly adapt to load changes and ensuring the stability and economy of the power generation process.

[0098] Example 2

[0099] Reference Figure 2 This is the second embodiment of the present invention, which provides an intelligent control system for power plant combustion power generation.

[0100] Specifically, the system includes: a first data acquisition module, a combustion power generation control module, a second data acquisition module, a power plant power generation demand forecasting module, and a combustion control and regulation module.

[0101] The first data acquisition module collects power plant combustion power generation control data, obtains coal quality parameters and power generation data, and merges the combustion power generation control data, coal quality parameters and power generation data to obtain an augmented time series dataset.

[0102] The combustion power generation control module is based on the generalized regression neural network (GRNN) model, and uses the dataset as samples to construct and train the combustion power generation control model.

[0103] The second data acquisition module acquires historical power load data of the power plant and meteorological data of the same period, and merges them with the load data to obtain an augmented power plant load dataset D2.

[0104] The power plant generation demand forecasting module is based on the Long Short-Term Memory (LSTM) neural network model, which constructs and trains a power plant generation demand forecasting model to predict the power plant generation demand in future time periods.

[0105] The combustion control and regulation module solves the power plant combustion power generation control data, constructs a PID regulation mechanism for power generation combustion, and controls combustion power generation.

[0106] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings or direct couplings or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0107] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments under the guidance of this application without departing from the spirit and scope of protection of the claims. All of these variations are within the protection scope of this application.

Claims

1. An intelligent control method for power plant combustion power generation, characterized by, The method comprises the following steps: Collecting power plant combustion power generation control data, obtaining coal quality parameters and power generation data, merging the combustion power generation control data, the coal quality parameters and the power generation data to obtain an augmented time series data set; The power plant combustion power generation control data includes coal feeder coal supply, primary air fan air volume and secondary air fan air volume, and the combustion power generation control data is preprocessed; the coal quality parameters include low calorific value, total sulfur content and volatile matter; power generation power data is collected; The preprocessed combustion power generation control data, coal quality parameters and power generation power are aligned and merged according to timestamps to obtain an augmented time series data set; A combustion power generation control model is constructed and trained based on a generalized regression neural network (GRNN) model; The generalized regression neural network (GRNN) model comprises an input layer, a pattern layer, a summation layer and an output layer; Historical power plant load data and contemporaneous meteorological data are obtained, and the load data is merged to obtain an augmented power plant load data set; A power plant power generation demand prediction model is constructed and trained based on a long short-term memory neural network (LSTM) model to predict the power plant power generation demand in a future time period; The power plant combustion power generation control data is solved, and a PID regulation mechanism during power generation combustion is constructed to control the combustion power generation; Based on the power generation demand predicted by the power plant load prediction model, the predicted power generation demand is substituted into the combustion power generation control model to solve the power plant combustion power generation control data; The trained power plant load prediction model is used to input the historical load data and meteorological data of the power plant to predict the load of the power plant in a future time period, and a load prediction sequence is obtained; According to the load prediction sequence, the predicted load power value is input into the combustion power generation control model to solve the corresponding combustion power generation control data sequence, including the data sequence of the coal supply, the primary air volume and the secondary air volume.

2. The intelligent control method for power plant combustion power generation according to claim 1, characterized in that, Collecting historical load data and contemporaneous meteorological data of the power plant, the meteorological data including temperature and humidity; aligning and merging the historical load data and contemporaneous meteorological data of the power plant according to timestamps to obtain an augmented load data set; Based on long short-term memory neural network LSTM model, the load data set is processed through sliding time window, training samples are constructed, and a power plant load prediction model is trained. Based on long short-term memory neural network LSTM model, the load data set is processed through sliding time window, training samples are constructed, and a power plant load prediction model is trained.

3. The intelligent control method for power plant combustion power generation according to claim 2, characterized in that, Setting the time window width and prediction step of the sliding time window, and extracting samples from the load data set using the sliding time window, each sample including multiple consecutive load values and corresponding meteorological data as input, and future load values as output; The load data set is obtained by extracting the load data set through the sliding time window, and the load data set is classified into a training set, a validation set and a test set.

4. The intelligent control method for power plant combustion power generation according to claim 3, characterized in that, A long short-term memory neural network (LSTM) model is constructed to train the power plant load prediction model; The long short-term memory neural network (LSTM) model comprises an LSTM input layer, an LSTM hidden layer and an LSTM output layer; the LSTM input layer is used to receive input data; the output layer comprises a full connection layer to predict the load value of the future time step.

5. The intelligent control method for power plant combustion power generation according to claim 4, characterized in that, The training steps of the long short-term memory neural network (LSTM) model are as follows: The training set data of the load data set is input into the LSTM model for training; the batch size and the number of training rounds are set; after each training round, the validation set data is used to evaluate the model performance; the model structure and hyperparameters are adjusted according to the validation set performance; Early stopping is used to prevent overfitting; the trained LSTM model is evaluated on the test set to assess its predictive performance; the root mean square error and mean absolute error evaluation metrics are calculated; When the long short-term memory neural network LSTM model is trained, for new input historical load values and meteorological data, the trained LSTM model is used as a power plant load prediction model to predict future power plant load values.

6. The intelligent control method for power plant combustion power generation according to claim 5, characterized in that, PID controller is constructed to adjust the power generation combustion control parameters; the control variables of coal supply, primary air volume and secondary air volume are set, and the power generation power is the controlled variable; the PID controller structure is constructed; By real-time monitoring of power generation power and combustion state, the actual value is fed back to the PID controller, and the PID controller dynamically adjusts the power generation combustion control parameters according to real-time feedback.

7. An intelligent control system for power plant combustion power generation, which is used to implement the intelligent control method for power plant combustion power generation according to any one of claims 1 to 6. It comprises: a first data acquisition module, a combustion power generation control module, a second data acquisition module, a power plant power generation demand prediction module, and a combustion control adjustment module; The first data acquisition module acquires power plant combustion power generation control data, obtains coal quality parameters and power generation data, combines the combustion power generation control data, coal quality parameters and power generation data, and obtains an augmented time series data set; The combustion power generation control module is based on a generalized regression neural network GRNN model to construct and train a combustion power generation control model; The second data acquisition module obtains historical power plant load data and contemporaneous meteorological data, and combines the load data to obtain an augmented power plant load data set; The power plant power generation demand prediction module is based on a long short-term memory neural network LSTM model to construct and train a power plant power generation demand prediction model to predict the power plant power generation demand in the future time period; The combustion control adjustment module solves the power plant combustion power generation control data, constructs a PID adjustment mechanism during power generation combustion, and controls the combustion power generation.

Images