Method for optimizing transmission power curve of graphitization furnace

Abstract

The invention relates to the technical field of industrial electric furnace control, and discloses a graphitization furnace power transmission power curve optimization method, which comprises the following steps: S1, data acquisition and preprocessing: acquiring historical operation data of a graphitization furnace, and carrying out cleaning and standardization processing on the data; the historical operation data comprises a power transmission power curve, in-furnace temperature and pressure data, material initial parameters and finished product quality indexes; s2, key factor analysis: determining key factors influencing the quality of a finished product from historical operation data through correlation analysis and principal component analysis methods; s3, constructing an optimization model; s4, performing model training and verification; s5, generating an optimization curve; and S6, applying and iterating. Compared with the prior art, the method has the advantages that the optimal transmission power curve is automatically generated through data driving and a machine learning model, so that the problems of unstable product quality, high energy consumption and process adjustment lagging caused by dependence on artificial experience are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial electric furnace control, and particularly relates to a method for optimizing power curve of graphitization furnace. BACKGROUND

[0002] The graphitization furnace is a key equipment for producing high-purity graphite, negative electrode material and other carbon products. The process thereof is to complete the graphitization transformation of the products in the furnace at high temperature by heating through power transmission. The power transmission curve (i.e. the curve of power change over time) directly determines the temperature field in the furnace and the heat process experienced by the products, and is a core factor affecting the quality (such as resistivity, bulk density and ash content) of the final products and the energy consumption of production.

[0003] At present, the power transmission curve of the graphitization furnace is mostly set by the experience of the operators, and lacks scientific and accurate optimization methods. This experience-dependent control method has the following significant shortcomings: 1) the production stability of different batches is poor, and the product quality fluctuates greatly; 2) it is difficult to optimize the energy consumption under the premise of ensuring the quality, and the production cost is high; 3) the process adjustment period is long, and it is unable to quickly adapt to the initial parameters of different materials.

[0004] There are also some methods in the prior art for trying to optimize the process, but most of them have limitations. For example, some methods only focus on a single parameter of temperature, and fail to comprehensively consider the coupling effects of multiple factors such as power change rate and constant temperature time; some methods use simple mathematical fitting models, and the prediction accuracy and generalization ability are insufficient, and are unable to handle complex nonlinear relationships in production.

[0005] Therefore, there is an urgent need in the field for an intelligent method that can self-adaptively optimize the power transmission curve, so as to realize stable product quality, reduce energy consumption and quickly respond to changes in materials. SUMMARY

[0006] The technical problem to be solved by the present application is to overcome the above technical difficulties, and to provide a method for optimizing the power transmission curve of the graphitization furnace. The method automatically generates an optimal power transmission curve through data driving and a machine learning model, so as to solve the problems of unstable product quality, high energy consumption and process adjustment lag caused by reliance on artificial experience.

[0007] To solve the above technical problems, the technical solution provided by the present application is as follows:

[0008] A method for optimizing the power transmission curve of a graphitization furnace, comprising the following steps:

[0009] S1, data acquisition and preprocessing: collecting historical operation data of the graphitization furnace, and performing cleaning and standardization processing on the data; the historical operation data includes the power transmission curve, the temperature and pressure data in the furnace, the initial parameters of the materials and the quality indicators of the finished products;

[0010] S2, key factor analysis: determine the key factors affecting the quality of the finished product from the historical operation data through correlation analysis and principal component analysis method; the key factors include the power rise rate, the constant power value of each stage, the power maintenance time length and the power drop rate;

[0011] S3, optimization model construction: taking the finished product quality meeting the standard and the lowest energy consumption as the objective function, taking the value range of the key factors as the constraint condition, and adopting machine learning algorithm to construct the optimization model;

[0012] S4, model training and verification: the historical operation data processed in step S1 is used to train and verify the optimization model until the prediction accuracy of the model reaches the preset standard;

[0013] S5, optimization curve generation: input the real-time obtained material initial parameters into the trained optimization model to generate the optimization power curve including the preheating stage, the heating stage, the constant temperature stage and the cooling stage;

[0014] S6, application and iteration: the optimization power curve is applied to the actual production of the graphitization furnace, and real-time operation data and finished product data are collected; the optimization model is iteratively adjusted according to the feedback data.

[0015] Further, the material initial parameters include one or more of the type, particle size, bulk density and moisture content of the material; the finished product quality indicators include one or more of the volume density, resistivity and ash content of the finished product.

[0016] Further, in step S1, the data cleaning and standardization processing specifically includes: eliminating abnormal values and missing values in the historical operation data, and normalizing the data.

[0017] Further, in step S3, the machine learning algorithm is random forest algorithm or gradient boosting decision tree algorithm.

[0018] Further, in step S4, the historical operation data is used to train and verify the optimization model, specifically: the data set is divided into training set and verification set according to the proportion; the training set is used to train the model, and the verification set is used to verify the prediction accuracy of the model.

[0019] Further, in step S4, the preset standard is that the prediction error of the model for the finished product quality is less than 5%, and the prediction error for the energy consumption is less than 3%.

[0020] Further, in step S5, the generated optimization power curve needs to meet the constraint range determined by the key factor analysis in each stage.

[0021] Further, in step S6, the real-time operation data includes one or more of the following: in-furnace temperature, pressure data; the finished product data includes actual measured finished product quality indicators and energy consumption data.

[0022] Further, in step S6, the iterative adjustment of the optimization model according to the feedback data is specifically: when the actual production of finished product quality or energy consumption is not up to standard, the collected real-time operation data and finished product data are fed back to the historical database for retraining and parameter adjustment of the optimization model.

[0023] The present application has the following advantages compared with the prior art:

[0024] 1. The present application adopts a data-driven method, uses a machine learning model to automatically generate an optimal power curve, reduces the dependence on artificial experience, and realizes the intelligentization and automation of process control.

[0025] 2. The present application takes improving product quality and reducing production energy consumption as dual optimization objectives, can significantly reduce electric energy consumption and save production cost under the premise of ensuring product pass rate.

[0026] 3. The present application can generate an optimal power curve matched with the initial parameters (such as type, particle size, etc.) of the input material, enhancing the adaptability of the method to different production batches and material changes.

[0027] 4. The present application introduces a feedback iteration mechanism, which can continuously correct and optimize the model using data in production practice, so that the system has the ability of self-learning and self-optimization, and can maintain the best running state for a long time.

[0028] 5. The present application uses advanced ensemble learning algorithms such as random forest, which has strong fitting ability for nonlinear relationships in complex industrial processes, high prediction accuracy and strong reliability. DETAILED DESCRIPTION

[0029] In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more. In addition, the term "includes" and any variation thereof is intended to cover non-exclusive inclusion.

[0030] The present application will be further described in detail below in combination with embodiments.

[0031] A method for optimizing the power curve of a graphitization furnace, comprising the following steps:

[0032] S1, Data collection and preprocessing: Collect data from the enterprise manufacturing execution system and historical database for 100 successful production batches in the past year. Each piece of data includes: complete power curve P(t), temperature curve T(t) of multiple temperature measurement points, furnace pressure data, material type (petroleum coke), particle size distribution D50, bulk density, moisture content, and quality indicators of finished products (resistivity, bulk density, ash content) and total energy consumption E. Use the Pandas library of Python to clean the data, remove abnormal jump values caused by sensor failure and incomplete batch data, and finally standardize all numerical features.

[0033] S2, Key factor analysis: For the preprocessed data, first perform Pearson correlation analysis to calculate the correlation coefficients of each process parameter (such as average heating rate, maximum power holding time, etc.) with product quality and energy consumption. On this basis, use principal component analysis to further reduce dimensionality and identify the most important comprehensive influencing factors. The key factors that have the greatest impact on the power curve are: power rise rate R_up in the 800-1500℃ interval, power setting value P_hold and its holding time T_hold in the constant temperature stage, and power drop rate R_down in the early cooling stage.

[0034] S3, Optimization model construction: This embodiment selects the gradient boosting decision tree (GBDT) algorithm to construct the optimization model. The objective function is set as: Minimize(E), while satisfying the constraint conditions: resistivity > target value * 0.95 and < target value * 1.05, bulk density > target value * 0.98.

[0035] S4, Model training and verification: Divide the 100 batches of data into a training set (70 batches) and a validation set (30 batches) randomly at a ratio of 7:3. Use the training set to train the GBDT model and adjust the depth of the tree, learning rate, and other hyperparameters. On the validation set, when the average relative error of the predicted resistivity and bulk density of the finished product is less than 5% and the predicted energy consumption error is less than 3%, the model training is considered complete and the model file is saved.

[0036] S5, Optimization curve generation: When there is a new production task, the operator inputs the initial parameters of the material for this batch (such as petroleum coke, particle size D50 = 15μm, bulk density 0.85g / cm 3 , moisture content 0.5%) on the system interface. The system calls the trained model to quickly predict the optimal key factor combination for this batch of material (such as R_up = 120kW / h, P_hold = 550kW, T_hold = 8h, R_down = 80kW / h), and automatically generates a complete optimized power curve that includes the preheating, heating, constant temperature, and cooling stages.

[0037] S6, application and iteration: the optimization curve is issued to the PLC control system of the graphitization furnace for execution. The system collects data such as furnace temperature and pressure in real time during the entire power transmission process, and collects product quality and energy consumption data after production is completed. If the resistivity of the finished product of this batch does not meet the standard or the energy consumption exceeds the expectation, the system will automatically include all data of this batch into the historical database. After accumulating a certain amount of new data (such as 5-10 batches), the system can automatically or manually trigger incremental learning or retraining of the model, update the model parameters, and thus realize continuous optimization of the model.

[0038] The above describes the present application and its embodiments, which are not limited. If a person skilled in the art is inspired by it, without departing from the purpose of the present application, similar embodiments of the technical solution can be designed without creativity, and all should belong to the protection scope of the present application.

Claims

1. A method for optimizing the power delivery curve of a graphitization furnace, characterized in that, Includes the following steps: S1. Data Acquisition and Preprocessing: Collect historical operating data of the graphitization furnace and clean and standardize the data; the historical operating data includes power supply curve, furnace temperature and pressure data, initial material parameters and finished product quality indicators; S2. Key Factor Analysis: Through correlation analysis and principal component analysis, key factors affecting the quality of finished products are identified from historical operating data; the key factors include the rate of increase of power transmission, constant power value at each stage, power maintenance duration, and power decrease rate. S3. Optimization Model Construction: With the objective function of achieving the product quality standard and minimizing energy consumption, and with the value range of the key factors as constraints, an optimization model is constructed using machine learning algorithms. S4. Model Training and Validation: The optimized model is trained and validated using the historical running data processed in step S1 until the model prediction accuracy reaches the preset standard. S5. Optimization curve generation: Input the real-time acquired initial material parameters into the trained optimization model to generate an optimized power delivery curve that includes the preheating stage, heating stage, isothermal stage and cooling stage. S6. Application and Iteration: Apply the optimized power supply curve to the actual production of the graphitization furnace and collect real-time operation data and finished product data; iterate and adjust the optimization model based on the feedback data.

2. The method for optimizing the power supply curve of a graphitization furnace according to claim 1, characterized in that: In step S1, the initial parameters of the material include one or more of the following: material type, particle size, bulk density, and moisture content; the quality indicators of the finished product include one or more of the following: bulk density, resistivity, and ash content.

3. The method for optimizing the power supply curve of a graphitization furnace according to claim 1, characterized in that: In step S1, the data cleaning and standardization process specifically includes: removing outliers and missing values ​​from historical operating data, and normalizing the data.

4. The method for optimizing the power supply curve of a graphitization furnace according to claim 1, characterized in that: In step S3, the machine learning algorithm is either the random forest algorithm or the gradient boosting decision tree algorithm.

5. The method for optimizing the power supply curve of a graphitization furnace according to claim 1, characterized in that: In step S4, the step of using historical running data to train and validate the optimization model specifically involves: dividing the dataset into a training set and a validation set proportionally; using the training set to train the model; and using the validation set to validate the model's prediction accuracy.

6. The method for optimizing the power supply curve of a graphitization furnace according to claim 1, characterized in that: In step S4, the preset standard is: the model's prediction error for the quality of the finished product is less than 5%, and its prediction error for energy consumption is less than 3%.

7. The method for optimizing the power supply curve of a graphitization furnace according to claim 1, characterized in that: In step S5, the power parameters of each stage of the generated optimized power transmission curve must meet the constraints determined by the key factor analysis.

8. The method for optimizing the power supply curve of a graphitization furnace according to claim 1, characterized in that: In step S6, the real-time operating data includes one or more of the following: furnace temperature and pressure data; the finished product data includes the actual measured finished product quality indicators and energy consumption data.

9. The method for optimizing the power supply curve of a graphitization furnace according to claim 1, characterized in that: In step S6, the iterative adjustment of the optimization model based on the feedback data specifically involves: when the quality or energy consumption of the actual produced product fails to meet the standards, the collected real-time operating data and finished product data are fed back to the historical database for retraining and parameter adjustment of the optimization model.