Intelligent control system in coking process
Through the intelligent control system and the improved long short-term memory network model, the problems of resource waste and inefficiency caused by manual control in the coking process were solved, and the stability and efficiency of the coking process were improved.
Patent Information
- Application Number
- CN202510783071.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-12
- Publication Date
- 2025-09-26
AI Technical Summary
The control of the coking process in coking plants is affected by manual experience, resulting in waste of resources, low system efficiency and high costs.
An intelligent control system, including data acquisition, processing, model building and calculation subsystems, is adopted to predict and control the coking process using an improved long short-term memory network model.
Improve the stability and efficiency of the coking process, reduce resource waste and lower costs.
Smart Images
Figure CN120699646A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of coking intelligent control, and in particular relates to an intelligent control system in a coking process. Background Art
[0002] A coking plant is a production facility that transforms coal into chemical products. Coking plants use coal as their primary raw material. Through high-temperature dry distillation, coal gas is extracted. Further chemical processes produce products such as saccharin, tar, asphalt, coke, and numerous pharmaceutical raw materials. Coking plants provide a wide variety of coking products and are a vital component of social development.
[0003] However, at present, coking plants often use manual control to control the coking process. The control process is affected by the experience and technology of the controller, which easily leads to waste of resources and fails to optimize the overall system. Not only does it reduce the efficiency of the system, but the cost also remains high. Summary of the Invention
[0004] In order to solve the above technical problems, the present invention proposes an intelligent control system for a coking process to solve the problems existing in the above prior art.
[0005] To achieve the above object, the present invention provides an intelligent control system for a coking process, comprising:
[0006] Data acquisition subsystem, used to obtain parameter data of the coking process;
[0007] a data processing subsystem connected to the data acquisition subsystem, for preprocessing the parameter data summarized in the coking process to obtain a preprocessed data set;
[0008] A model building subsystem, connected to the data processing subsystem, is used to build an improved long short-term memory network model, train the improved long short-term memory network model using a historical data set, and output the model after learning the associations in the coking process to generate an intelligent prediction model;
[0009] A calculation subsystem, connected to the model building subsystem, for inputting the preprocessed data set into the intelligent prediction model for calculation to obtain a coking prediction result;
[0010] A control subsystem is connected to the computing subsystem and is used to intelligently control the coking process based on the coking prediction results.
[0011] Preferably, the data acquisition subsystem includes:
[0012] Equipment parameter acquisition module, used to obtain coke oven operating parameters;
[0013] Working condition acquisition module, used to obtain the quality indicators of coke oven coal production;
[0014] An integration module is used to integrate the coke oven operating parameters and the coke oven coal production quality indicators to generate parameter data of the coking process.
[0015] Preferably, the coke oven operating parameters include: furnace temperature, hearth pressure, gas cooling temperature, gas purification efficiency and gas composition;
[0016] The coke oven coal production quality indicators include the ash content, volatile matter content and fixed carbon content of the coke.
[0017] Preferably, the data processing subsystem includes:
[0018] A correlation calculation module is used to calculate the correlation of the parameter data of the coking process based on the Pearson correlation coefficient to obtain correlation data;
[0019] a labeling module, configured to label the correlation data to the parameter data of the coking process to generate a labeling data set;
[0020] The normalization module is used to perform normalization processing on the labeled data set to generate the preprocessed data set.
[0021] Preferably, the model building subsystem includes:
[0022] A network model construction module is used to construct a long short-term memory network model and improve the long short-term memory network model to generate an improved long short-term memory network model;
[0023] A training module is used to construct a historical data set, input the historical data set into the improved long short-term memory network model for training, and generate a training model;
[0024] A validation module is used to perform cross-validation on the training model using a test set to generate an evaluation result;
[0025] A fine-tuning module is used to adjust the structure and hyperparameters of the training model based on the evaluation results and then retrain to generate the intelligent prediction model.
[0026] Preferably, the network model building module includes:
[0027] Module building unit, used to build attention mechanism module and batch normalization module;
[0028] An improvement unit is used to connect the attention mechanism module and the time step of the long short-term memory network model, and at the same time connect the batch normalization module with the hidden layer of the long short-term memory network model to generate the improved long short-term memory network model.
[0029] Preferably, the training module includes:
[0030] A pre-training unit is used to train the improved long short-term memory network model using an RBMs algorithm, and terminate the training after the improved long short-term memory network model completes the weight distribution learning of the data to obtain an optimized long short-term memory network model;
[0031] The data set training unit is used to train the optimized long short-term memory network model through the historical data set until the model learns the association in the coking process and outputs it to generate the training model.
[0032] Preferably, the computing subsystem is further configured to generate a coking prediction result, generate an equipment control signal according to the coking prediction result and feed it back to the control subsystem, and the control subsystem intelligently controls the coking process based on the equipment control signal.
[0033] Compared with the prior art, the present invention has the following advantages and technical effects:
[0034] The LSTM network model has the advantage of modeling sequential data and can effectively capture the dynamic characteristics of time series data in the coking process. The present invention uses the LSTM network model to continuously model time series data such as furnace temperature, furnace pressure, and gas composition, thereby better understanding the dynamic changes in the coking process. Furthermore, the improved LSTM network model in the present invention has stronger information transfer and memory capabilities, effectively transferring and retaining important information between different time steps. This allows the model to better utilize historical data to guide current and future control decisions, thereby improving the stability and efficiency of the coking process. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:
[0036] Figure 1 Schematic diagram of an intelligent control system in a coking process according to an embodiment of the present invention;
[0037] Figure 2 Schematic diagram of a long short-term memory network model according to an embodiment of the present invention. DETAILED DESCRIPTION
[0038] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.
[0039] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.
[0040] Example 1
[0041] like Figure 1 and Figure 2 As shown, this embodiment provides an intelligent control system for a coking process, including:
[0042] Data acquisition subsystem, used to obtain parameter data of the coking process;
[0043] a data processing subsystem connected to the data acquisition subsystem, for preprocessing the parameter data summarized in the coking process to obtain a preprocessed data set;
[0044] A model building subsystem, connected to the data processing subsystem, is used to build an improved long short-term memory network model, train the improved long short-term memory network model using a historical data set, and output the model after learning the associations in the coking process to generate an intelligent prediction model;
[0045] A calculation subsystem, connected to the model building subsystem, for inputting the preprocessed data set into the intelligent prediction model for calculation to obtain a coking prediction result;
[0046] A control subsystem is connected to the computing subsystem and is used to intelligently control the coking process based on the coking prediction results.
[0047] Specifically:
[0048] The main task of the data acquisition subsystem is to monitor the coking process in real time and record and store relevant parameter data for subsequent processing and analysis.
[0049] Connected to the data acquisition subsystem is the data processing subsystem. The data processing subsystem preprocesses the coking process parameter data obtained from the data acquisition subsystem. The purpose of preprocessing is to convert the raw data into a format more suitable for subsequent processing, such as removing noise, filling missing values, and standardizing, thereby generating a preprocessed dataset.
[0050] The model building subsystem's primary function is to construct an improved Long Short-Term Memory (LSTM) network model. LSTM is a deep learning model capable of processing sequential data, particularly well-suited for processing data with temporal correlations. The model building subsystem trains the improved LSTM network model using historical datasets. Through continuous learning and adjustment, the model learns the correlations within the coking process. Once model training is complete, an intelligent predictive model is generated.
[0051] The computing subsystem inputs the preprocessed data set into the trained intelligent prediction model, and obtains the prediction results of the coking process through the calculation of the model.
[0052] The control subsystem is responsible for intelligently controlling the coking process based on the coking prediction results obtained by the calculation subsystem. This intelligent control can automatically adjust the parameters of the coking process to achieve the purpose of optimizing the coking effect.
[0053] Further optimizing the solution, the data acquisition subsystem includes:
[0054] Equipment parameter acquisition module, used to obtain coke oven operating parameters;
[0055] Working condition acquisition module, used to obtain the quality indicators of coke oven coal production;
[0056] An integration module is used to integrate the coke oven operating parameters and the coke oven coal production quality indicators to generate parameter data of the coking process.
[0057] The data acquisition subsystem consists of several key modules.
[0058] First, the equipment parameter acquisition module is one of the core components of this subsystem. Its primary responsibility is to obtain various operating parameters from the coke oven equipment in real time. These parameters include temperature, pressure, humidity, and other information directly related to coke oven operation.
[0059] The main goal of the working condition acquisition module is to obtain quality indicators of coke oven coal production, including coal calorific value, carbon content, sulfur content and other information directly related to coal quality.
[0060] Finally, the integration module is the key step in integrating the information obtained by the above two modules. The function of this module is to comprehensively analyze the coke oven operating parameters and the coke oven coal production quality indicators to generate parameter data for the coking process.
[0061] Further optimizing the scheme, the coke oven operating parameters include: furnace temperature, hearth pressure, gas cooling temperature, gas purification efficiency and gas composition;
[0062] The coke oven coal production quality indicators include the ash content, volatile matter content and fixed carbon content of the coke.
[0063] Specifically,
[0064] Furnace temperature: Furnace temperature refers to the temperature inside the coke oven during the coke production process. The appropriate furnace temperature can affect the quality and output of coke.
[0065] Gas composition: The composition of the gas produced in the coking process is one of the important parameters, which includes carbon monoxide (CO), hydrogen (H2), methane (CH4) and other components. These components have an impact on the quality of coke and the production process.
[0066] Hearth pressure: Hearth pressure refers to the pressure in the hearth area inside the coke oven, which has a certain impact on the formation and production process of coke.
[0067] Coke quality parameters: including coke ash content, volatile matter content, fixed carbon content, etc. These parameters directly affect the quality and use of coke.
[0068] Coke oven gas flow: refers to the generation rate and flow speed of coal gas in the coke oven, which plays an important role in the operation and control of the coke oven.
[0069] Gas cooling temperature: Gas cooling temperature affects the processing and utilization of gas, and is also related to the stability of coke production.
[0070] Gas purification efficiency: refers to the efficiency of the gas purification system in purifying the gas generated in the coking process, which is of great significance to environmental protection and resource utilization.
[0071] Further optimizing the solution, the data processing subsystem includes:
[0072] The correlation calculation module calculates the correlation of the parameter data of the coking process based on the Pearson correlation coefficient to obtain correlation data. The Pearson correlation coefficient is a commonly used method to measure the linear correlation between two continuous variables. The Pearson correlation coefficient can be calculated between various parameters, with a value ranging from -1 to 1, where a value close to 1 indicates a positive correlation, a value close to -1 indicates a negative correlation, and a value close to 0 indicates no correlation.
[0073] a labeling module, configured to label the correlation data to the parameter data of the coking process to generate a labeling data set;
[0074] The normalization module is used to perform normalization processing on the labeled data set to generate the preprocessed data set.
[0075] Further optimizing the solution, the model building subsystem includes:
[0076] A network model construction module is used to construct a long short-term memory network model and improve the long short-term memory network model to generate an improved long short-term memory network model;
[0077] A training module is used to construct a historical data set, input the historical data set into the improved long short-term memory network model for training, and generate a training model;
[0078] A validation module is used to perform cross-validation on the training model using a test set to generate an evaluation result;
[0079] A fine-tuning module is used to adjust the structure and hyperparameters of the training model based on the evaluation results and then retrain to generate the intelligent prediction model.
[0080] Specifically:
[0081] Network Model Construction Module: This module's primary purpose is to build a basic LSTM network model. LSTM is a special type of neural network capable of processing sequential data, such as time series or natural language text. In this module, this example not only builds a basic LSTM model but also makes a series of improvements to it. These improvements include adding more layers, changing activation functions, and adjusting weight initialization methods, all in the hope of achieving a LSTM model with even better performance.
[0082] Training Module: Once this embodiment has an improved LSTM network model, the next step is to train it using a historical dataset. This historical dataset is pre-collected and organized data used to train the model so that it can learn the mapping relationship from input to output. In this module, this embodiment inputs the historical dataset into the improved LSTM model. Through multiple iterations of training, the model parameters are continuously adjusted until the model performance reaches a satisfactory level. This process generates a preliminarily trained model, called a trained model.
[0083] Validation Module: After model training is complete, this embodiment needs to verify the model's performance. The validation module is responsible for this task, using an independent test set to cross-validate the trained model. Cross-validation is a method for evaluating a model's generalization ability. It divides a dataset into multiple parts, rotating between using one part as test data and the remaining part as training data to evaluate the model's performance. In this way, this embodiment can generate an evaluation result that reflects the model's performance on different data.
[0084] Fine-tuning module: Based on the evaluation results generated by the verification module, this embodiment requires further adjustments to the trained model. The fine-tuning module is used to accomplish this task. In this module, this embodiment adjusts the training model's structure (e.g., increasing or decreasing the number of layers) and hyperparameters (e.g., learning rate, batch size, etc.) based on the evaluation results. After these adjustments, this embodiment trains again in the hope of obtaining a model with even better performance. This model becomes the final intelligent prediction model of this embodiment.
[0085] In this embodiment, past coking process data is used as supervisory signals, such as time series data of furnace temperature, furnace pressure, gas composition, gas yield, etc. These data can be used to train the model to predict the current or future coking process status.
[0086] Further optimizing the solution, the network model building module includes:
[0087] Module building unit, used to build attention mechanism module and batch normalization module;
[0088] Residual connections are added between adjacent LSTMs, directly adding or concatenating the output of the previous layer with the input of the next layer, thereby constructing a shortcut path across layers. This connection method helps solve the problems of vanishing and exploding gradients during deep network training, making it easier for the model to learn the identity mapping and reducing information loss.
[0089] The attention mechanism dynamically assigns attention weights at each time step to capture important information at different positions in sequence data. In the self-attention mechanism, attention is applied to each position in the input sequence, allowing the model to simultaneously consider the relationships between different positions in the input sequence. The attention mechanism can be used to enhance a model's ability to model sequential data, improving its performance and generalization capabilities.
[0090] Batch normalization is typically placed after the fully connected or convolutional operations in each hidden layer and before the activation function. The purpose of this is to normalize each feature dimension in each layer of the network, making the input distribution more stable and helping to improve the training speed and stability of the network.
[0091] An improvement unit is used to connect the attention mechanism module and the time step of the long short-term memory network model, and at the same time connect the batch normalization module with the hidden layer of the long short-term memory network model to generate the improved long short-term memory network model.
[0092] Further optimizing the solution, the training module includes:
[0093] A pre-training unit is used to train the improved long short-term memory network model using an RBMs algorithm, and terminate the training after the improved long short-term memory network model completes the weight distribution learning of the data to obtain an optimized long short-term memory network model;
[0094] The data set training unit is used to train the optimized long short-term memory network model through the historical data set until the model learns the association in the coking process and outputs it to generate the training model.
[0095] According to a further optimization scheme, the computing subsystem is further used to generate a coking prediction result, generate an equipment control signal according to the coking prediction result and feed it back to the control subsystem, and the control subsystem intelligently controls the coking process based on the equipment control signal.
[0096] Based on the model's predictions, the computing subsystem further analyzes which process parameters require adjustment. For example, if the prediction indicates that excessively high temperatures could lead to reduced product quality, the system will generate a control signal to reduce the combustion air supply. The control strategy utilizes the model's output to determine the optimal equipment operation. Based on the control strategy, the computing subsystem generates specific control signals. These signals are instructions for specific equipment operations, such as adjusting furnace temperature, changing furnace atmosphere, or adjusting raw material feed rate.
[0097] The above are merely preferred embodiments of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.
Claims
1. An intelligent control system in a coking process, characterized in that: It is composed of the following subsystems: Data acquisition subsystem, used to obtain parameter data of the coking process; a data processing subsystem connected to the data acquisition subsystem, for preprocessing the parameter data summarized in the coking process to obtain a preprocessed data set; A model building subsystem, connected to the data processing subsystem, is used to build an improved long short-term memory network model, train the improved long short-term memory network model using a historical data set, and output the model after learning the associations in the coking process to generate an intelligent prediction model; A calculation subsystem, connected to the model building subsystem, for inputting the preprocessed data set into the intelligent prediction model for calculation to obtain a coking prediction result; A control subsystem is connected to the computing subsystem and is used to intelligently control the coking process based on the coking prediction results.
2. The intelligent control system in the coking process according to claim 1, characterized in that: The data acquisition subsystem includes the following modules: Equipment parameter acquisition module, used to obtain coke oven operating parameters; Working condition acquisition module, used to obtain the quality indicators of coke oven coal production; An integration module is used to integrate the coke oven operating parameters and the coke oven coal production quality indicators to generate parameter data of the coking process.
3. The intelligent control system in the coking process according to claim 2, characterized in that: The coke oven operating parameters include: furnace temperature, hearth pressure, gas cooling temperature, gas purification efficiency and gas; The coke oven coal production quality indicators include the ash content, volatile matter content and fixed carbon content of the coke.
4. The intelligent control system in the coking process according to claim 1, characterized in that: The data processing subsystem includes the following modules: A correlation calculation module calculates the correlation of the parameter data of the coking process based on the Pearson correlation coefficient to obtain correlation data; a labeling module, configured to label the correlation data to the parameter data of the coking process to generate a labeling data set; The normalization module is used to perform normalization processing on the labeled data set to generate the preprocessed data set.
5. The intelligent control system in the coking process according to claim 1, characterized in that: The model building subsystem includes the following modules: A network model construction module is used to construct a long short-term memory network model and improve the long short-term memory network model to generate an improved long short-term memory network model; A training module is used to construct a historical data set, input the historical data set into the improved long short-term memory network model for training, and generate a training model; A validation module is used to perform cross-validation on the training model using a test set to generate an evaluation result; A fine-tuning module is used to adjust the structure and hyperparameters of the training model based on the evaluation results and then retrain to generate the intelligent prediction model.
6. The intelligent control system in the coking process according to claim 5, characterized in that: The network model building module includes the following units: Module building unit, used to build attention mechanism module and batch normalization module; An improvement unit is used to connect the attention mechanism module and the time step of the long short-term memory network model, and at the same time connect the batch normalization module with the hidden layer of the long short-term memory network model to generate the improved long short-term memory network model.
7. The intelligent control system in the coking process according to claim 5, characterized in that: The training module includes the following units: A pre-training unit is used to train the improved long short-term memory network model using an RBMs algorithm, and terminate the training after the improved long short-term memory network model completes the weight distribution learning of the data to obtain an optimized long short-term memory network model; The data set training unit is used to train the optimized long short-term memory network model through the historical data set until the model learns the association in the coking process and outputs it to generate the training model.
8. The intelligent control system in the coking process according to claim 1, characterized in that: The computing subsystem is further configured to generate a coking prediction result, generate an equipment control signal according to the coking prediction result, and feed the signal back to the control subsystem. The control subsystem intelligently controls the coking process based on the equipment control signal.