A method for predicting optimal energy consumption and intelligently diagnosing energy efficiency in a steel production process

Abstract

The application is a kind of steel production process optimal energy consumption prediction and energy efficiency intelligent diagnosis method, and belongs to the technical field of steel production process, which comprises the following steps: based on the steel production process, a steel production database is established in units of heats or batches, and the data in the steel production database is preprocessed; a steel production process mechanism model is established; a black box model is constructed based on correlation analysis and a neural network model, and the black box model is trained; the steel production process mechanism model and the trained black box model are coupled to establish an energy consumption prediction model for predicting the energy consumption of the energy consumption factors of the steel production process; the energy consumption of the energy consumption factors of the steel production process is determined based on the energy consumption prediction model through a machine learning method, and the optimal energy consumption of the steel production process is determined through an energy consumption calculation formula of the steel production process, and the historical optimal energy consumption is determined according to historical data.

Description

Technical Field

[0001] This invention belongs to the technical field of steel production process and relates to a method for optimal energy consumption prediction and intelligent energy efficiency diagnosis in steel production process. Background Technology

[0002] In today's industrial production environment, especially in energy-intensive industries like steel production, effective energy management and significant energy consumption reduction have become core issues for enhancing corporate competitiveness and promoting sustainable development. Traditionally, energy management relies heavily on manual experience and intuitive judgment, which is not only inefficient but also makes it difficult to accurately assess and optimize energy consumption levels. With the advancement of scientific research, the mechanistic analysis of typical industrial production processes—based on thermodynamics and material balance and other physicochemical methods—has been thoroughly studied. Based on these mechanisms, corresponding energy consumption models can be established. Furthermore, the rapid development of information technology, particularly the widespread application of industrial big data and artificial intelligence, has provided new solutions for energy management. Industrial big data encompasses massive amounts of data generated during production processes. This data contains rich information that can be used to reveal energy consumption patterns, optimize production parameters, and thus achieve precise energy management. Data-driven methods utilize algorithms such as machine learning and deep learning to efficiently process and analyze large amounts of data, discovering complex relationships between energy consumption and various production parameters, and enabling the prediction and optimization of energy consumption and efficiency.

[0003] Mechanistic analysis (such as physicochemical methods based on thermodynamics and material balance) and data-driven methods (such as analyzing production data using machine learning and deep learning algorithms) are both applied in industrial energy consumption analysis, but each has its limitations. While mechanistic analysis models can provide in-depth theoretical insights, their uncertainty increases significantly when faced with numerous "black box" aspects of the actual production process. Conversely, while data-driven methods can efficiently process and analyze large amounts of data and predict energy consumption and efficiency trends, the instability and reliability of the data often affect the accuracy and precision of the model, potentially causing predictions to deviate from reality. Therefore, how to couple mechanistic analysis and data-driven methods to construct a comprehensive model that can both deeply understand energy consumption mechanisms and efficiently utilize data is a key issue that urgently needs to be addressed in the field of industrial energy efficiency management. Summary of the Invention

[0004] To solve the above problems, the technical solution adopted by the present invention is: an optimal energy consumption prediction method for steel production processes, comprising the following steps:

[0005] S1: Based on the steel production process, establish a steel production database in units of furnaces or batches, and preprocess the data in the steel production database;

[0006] S2: Establish a mechanism model for the steel formation process;

[0007] S3: Construct a black box model based on correlation analysis and neural network model, and train the black box model;

[0008] S4: Based on the mutual complementation and coupling of the steel production process mechanism model and the trained black box model, an energy consumption prediction model is established to predict the energy consumption factors in the steel production process.

[0009] S5: Determine the energy consumption of energy-consuming factors in the steel production process based on an energy consumption prediction model using machine learning methods, then determine the optimal energy consumption of the steel production process using the energy consumption calculation formula for the steel production process, and determine the historical optimal energy consumption based on historical data.

[0010] Furthermore, the establishment of the steel production database includes material consumption data, product data, composition parameters, physical property parameters, operating parameters, and technical parameter information. The database construction requirements are that it corresponds one-to-one with the product furnace or batch.

[0011] Furthermore, the process of constructing a black-box model based on correlation analysis and neural network model, and training the black-box model, is as follows:

[0012] S301: Determine the parameters of the neural network model and optimize them.

[0013] S302: Construct a black-box model based on correlation analysis and neural network model to analyze the mechanism of steel formation process;

[0014] S303: Update and optimize the black box model based on on-site data of the steel production process to better adapt to changes in process conditions and uncertainties in industrial production, and obtain a well-trained black box model.

[0015] Furthermore, the process of establishing an energy consumption prediction model for predicting energy consumption factors in the steel production process by complementing and coupling the steel production process mechanism model and the trained black box model is as follows:

[0016] Identify the relationship between the steel production process mechanism model and the trained black box model, determine the input-output relationship, and identify the coupling point between the steel production process mechanism model and the trained black box model;

[0017] Based on the coupling point, a coupling mechanism between the steel production process mechanism model and the trained black box model is established, and the coupling constraints are determined.

[0018] By combining data from the steel production process mechanism model and the trained black-box model, and adjusting parameters in real time through a dynamic feedback mechanism, an energy consumption prediction model is obtained.

[0019] Furthermore, the energy consumption calculation formula for the steel production process is expressed as follows:

[0020]

[0021] Among them: Q all Energy consumption forecast; Q use_i Q represents the amount of energy consumed. rec_i Q: represents the amount of energy recovered, k: represents the number of types of energy consumed, m: represents the number of types of energy recovered; use_elec Power consumption; Q use_cog Coke oven gas consumption; Q use_lng Natural gas consumption; Q use_water : Water consumption; Q use_lqi Low-pressure steam consumption; Q use_mqi Medium-pressure steam consumption; Q use_n2 Nitrogen consumption; Q use_jn2 High-pressure nitrogen consumption; Q use_ar Argon consumption; Q use_o2 Oxygen consumption; Q rec_lqi Low-pressure steam recovery rate; Q rec_zha Slag heat recovery.

[0022] Furthermore, the machine learning method employs either particle swarm optimization or diploid genetic algorithm within genetic algorithms.

[0023] Furthermore: the energy consumption of energy-consuming factors in the steel production process is determined based on an energy consumption prediction model using machine learning methods. Then, the optimal energy consumption for the steel production process is determined using the energy consumption calculation formula, and the historical optimal energy consumption is determined based on historical data, as follows:

[0024] The energy consumption prediction model is optimized by using the particle swarm optimization algorithm or the diploid genetic algorithm in the genetic algorithm of machine learning. After determining the range of changes of different factors in the production process, the optimal energy consumption in the production process can be determined through iterative calculation.

[0025] By optimizing historical production data, the optimal energy consumption value that meets production conditions is found, and the historical optimal energy consumption is obtained.

[0026] An energy efficiency intelligent diagnosis method based on any one of the optimal energy consumption prediction methods for steel production processes includes the following steps:

[0027] Based on the characteristics of each process in steel production, optimal energy consumption, and energy consumption prediction models, an intelligent energy consumption diagnosis model is established that analyzes the primary, secondary, and tertiary influencing factors of the steel production process.

[0028] Based on the intelligent energy consumption diagnostic model, the energy consumption and energy efficiency in the actual steel production process can be analyzed to determine the reasons for the fluctuations in energy consumption in the actual production process.

[0029] Furthermore, the process of analyzing energy consumption and energy efficiency in the actual steel production process based on the intelligent energy consumption diagnostic model, and determining the causes of energy consumption fluctuations in the actual steel production process, is as follows:

[0030] Determine the optimal operating conditions for the steel production process. Based on the determined optimal energy consumption, set the values ​​of energy consumption factor variables in the energy consumption prediction model as benchmark values ​​to obtain the benchmark operating conditions under the optimal energy consumption state.

[0031] Actual energy efficiency calculations are performed by inputting collected data from actual production into the intelligent energy consumption diagnostic model to obtain the actual energy efficiency value.

[0032] By comparing the data of the optimal and actual operating conditions of the steel production process, the variables with differences are substituted into the intelligent energy consumption diagnosis model for calculation. Through sensitivity analysis and regression analysis, the impact on the consumption and recovery of various energy media and the impact on total energy consumption are realized to conduct a single-factor analysis of the energy consumption impact of the steel production process.

[0033] The results of single-factor analysis enable the analysis of the causes of energy consumption fluctuations in the actual steel production process.

[0034] This invention provides a method for optimal energy consumption prediction and intelligent energy efficiency diagnosis in steel production processes. It aims to optimize energy consumption in typical industrial production processes by combining industrial big data, and to improve process energy efficiency through intelligent energy efficiency analysis, thereby promoting digital transformation and energy conservation and emission reduction.

[0035] This invention, based on current industrial big data, establishes a production database by collecting relevant production data, builds a mechanistic model based on current theoretical research, and improves the shortcomings of the mechanistic model through machine learning and deep learning to establish an energy consumption prediction model that couples mechanism and data-driven approaches. Based on the energy consumption prediction model, the optimal energy consumption is determined as the benchmark energy consumption through intelligent optimization algorithms or historical optimization. Actual production data is compared with the optimal energy consumption, and factor analysis is performed in the energy consumption prediction model. Finally, intelligent diagnostic results for three levels of energy consumption and energy efficiency are obtained.

[0036] Compared with existing technologies, this invention proposes an optimal energy consumption prediction and intelligent energy efficiency diagnosis method for steel production processes. Starting from industrial big data, it avoids the shortcomings of simply analyzing energy consumption mechanisms or using data-driven approaches to build energy consumption models. It transforms historical data into valuable data resources, establishing an energy consumption prediction model that better reflects actual production conditions. Furthermore, this application proposes a method for determining optimal energy consumption and an intelligent energy efficiency diagnosis approach. The method of this invention enables well-reasoned energy efficiency analysis, supported by concrete data and reliable theoretical analysis, and also provides operational suggestions. Energy management units and on-site operators can clearly determine the current production situation, facilitating adjustments to achieve energy conservation, emission reduction, cost reduction, and efficiency improvement. This invention has a high level of intelligence and automation, good operability for production enterprises, and can produce good results in practical applications. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a schematic diagram of the process of the present invention;

[0039] Figure 2 This is a schematic diagram of the data processing flow of the present invention;

[0040] Figure 3 This is a flowchart illustrating the neural network prediction process.

[0041] Figure 4 This is a flowchart illustrating the diploid genetic algorithm.

[0042] Figure 5 This is an example of how to use an electric arc furnace in steelmaking. Detailed Implementation

[0043] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and embodiments. To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Figure 1 This is a schematic diagram of the process of the present invention;

[0045] The purpose of this invention is to establish an optimal energy consumption prediction method for steel production processes, comprising the following steps:

[0046] S1: Based on the steel production process, establish a steel production database in units of furnaces or batches, and preprocess the data in the steel production database;

[0047] S2: Based on the data in the preprocessed steel production database, establish a mechanism model for the steel production process;

[0048] S3: Construct a black box model based on correlation analysis and neural network model, and train the black box model;

[0049] S4: Based on the coupling of the steel production process mechanism model and the trained black box model, an energy consumption prediction model is established to predict the energy consumption factors in the steel production process.

[0050] S5: Determine the energy consumption of energy-consuming factors in the steel production process based on an energy consumption prediction model using machine learning methods, then determine the optimal energy consumption of the steel production process using the energy consumption calculation formula for the steel production process, and determine the historical optimal energy consumption based on historical data.

[0051] The steps S1 / S2 / S3 / S4 / S5 are executed sequentially;

[0052] Furthermore: Sampling points are set up for the steel production process to establish a production database. For the target production process, suitable production lines are selected, and sensors are deployed to collect data such as flow rate, temperature, and pressure required by the model, based on data including material consumption, product data, composition parameters, physical property parameters, operating parameters, and technical parameters. This data is then stored in the database. The database construction requires a one-to-one correspondence between each product furnace or batch.

[0053] Taking steel production as an example, it is necessary to establish a production database based on furnace batches or batches, and to match various parameters one by one.

[0054] Table 1 shows the material parameters for Furnace No. 1 in the steel production process.

[0055]

[0056] Table 2 shows the production parameters of the primary furnace in the steel production process.

[0057]

[0058]

[0059] Table 3 shows the operating parameters of the primary furnace in the steel production process.

[0060]

[0061] Furthermore, the data in the steel production database is preprocessed to obtain effective analytical data on the steel production process. Because noise and errors exist in the data collected from the steel production process, missing and outlier values ​​will occur. Therefore, it is necessary to handle missing and outlier values ​​in the raw data. For missing values, the median imputation method is used to complete the missing values; for outliers, the box plot method is used for processing. The process is as follows: Figure 2 As shown, the specific process is as follows:

[0062] Read detailed historical data of the steel production process;

[0063] Determine if the historical data of the steel production process is complete. If it is determined that the historical data of the steel production process is complete, collect the remaining complete data.

[0064] When it is determined that the historical data of the steel production process is incomplete, it is determined whether the missing values ​​can be supplemented. If the missing values ​​cannot be supplemented, the data is deleted. If the missing values ​​can be supplemented, they are supplemented based on the median method.

[0065] Based on the complete data, a threshold is set to determine whether there are outliers. When outliers are found, they are deleted.

[0066] When no abnormal data is found, all data is integrated to obtain preprocessed steel production data.

[0067] Step 2: Establishment of a Mechanism Model for Steel Production. Establishing a mechanism model for the steel production process is a crucial step in understanding and optimizing the energy consumption and efficiency of typical industrial production processes and their sub-processes. Domestic and international experts have conducted in-depth and extensive research on these mechanisms. This research has not only deepened our understanding of the essence of industrial production processes but also provided a solid theoretical foundation for constructing reliable mechanism models. Taking steel production as an example, its sub-processes—coking, sintering, ironmaking, and steelmaking—typically involve complex physical, chemical, and thermodynamic processes. Based on the latest relevant research, mechanism models can be established.

[0068] The expression for the mechanistic model of the steel production process is as follows:

[0069] aA+bB→cC+dDΔH

[0070] ΔU=QW

[0071] ∑Q in =∑Q out

[0072] Q = mcΔT

[0073] M material,element =M product,element

[0074] Where: A, B, C, D: reactants and products of chemical reaction; a, b, c, d: coefficients; ΔH: enthalpy change; U: internal energy; Q: thermal energy; W: work; Q in Input energy; Q out Output energy; m: mass; c: specific heat; T: temperature; M material,element : Raw material element mass; M product,element Product element quality.

[0075] To address the uncertainties involved in mechanistic models and the adaptability of black-box models, data-driven models are needed to refine them. A black-box model refers to a complex and unclear mechanism in steel production that is not yet fully understood; a mechanistic model is a model that can explain various processes.

[0076] Taking electric arc furnace steelmaking in the steel production process as an example, most of the reaction processes involved in this process have established material balance and heat balance analysis models in the latest literature. However, when combined with the production process, some parameters of the black box model are uncertain due to environmental influences and need to be further adjusted in conjunction with actual production.

[0077] Figure 3 This is a flowchart illustrating the neural network prediction process.

[0078] Mechanistic models often face simplification assumptions, parameter uncertainties, "black box" problems, and adaptability challenges when reflecting actual industrial production conditions. Neural network algorithms, through parameter estimation and optimization, construction of high-precision black box prediction models, anomaly detection and fault diagnosis, as well as adaptive learning and optimization, effectively solve these problems, realize the fusion and integration of mechanism and data-driven models, significantly improve the accuracy and practicality of mechanistic models, and provide strong support for the optimization of industrial production processes and the improvement of energy efficiency.

[0079] Taking electric arc furnace steelmaking in the steel production process as an example, key parameters and factors such as element burn-off rate, carbon content, tapping temperature, and carbon content of molten steel (input parameters of the energy consumption prediction model) can be better determined through neural network algorithms.

[0080] The process of constructing a black-box model based on correlation analysis and neural network model, and training the black-box model is as follows:

[0081] S301: Determine the parameters of the neural network model and optimize them.

[0082] S302: Construct a black-box model based on correlation analysis and neural network model to analyze the mechanism of steel formation process;

[0083] S303: Update and optimize the black box model based on on-site data of the steel production process to better adapt to changes in process conditions and uncertainties in industrial production, and obtain a well-trained black box model.

[0084] Furthermore, the neural network model parameters are determined, and the process of optimizing these parameters is as follows:

[0085] Neural network algorithms can learn the relationship between parameters and output variables from large amounts of data in the steel production process, thereby better determining the parameters in the mechanistic model and optimizing them based on real-time data. The parameter determination and optimization process involves regression and optimization algorithms, as shown below:

[0086] (1) Data preparation: Determine the independent and dependent variables of the neural network algorithm, and collect and preprocess the data of the steel production process;

[0087] (2) Model selection: Select a suitable regression model, such as linear regression, multinomial regression, logarithmic regression, etc., based on the data of the steel production process.

[0088] (3) Model training: Use the training dataset to train the regression model and find a suitable regression model.

[0089] (4) Model evaluation: Use the test dataset to evaluate the reliability of the regression model and calculate its error.

[0090] (5) Model optimization: Adjust the model parameters or select a more suitable regression model based on the results.

[0091] Black-box model construction. For the "black box" within a mechanistic model, machine learning algorithms can build high-precision predictive models based on data, thus compensating for the limitations of mechanistic models in terms of their inability to perform modeling and analysis, and improving the model's completeness. Black-box model construction requires the use of correlation analysis and neural network models.

[0092] Step 4021: Correlation analysis to explore the influencing factors of the model.

[0093] (1) Identify the factors for correlation analysis and determine the relevant data. Based on literature and production conditions, determine the range of influencing factors and comparison values. Define the comparison value sequence as the parent sequence and the influencing factor sequence as the child sequence, and establish the parent sequence matrix as Y = [y1, y2, y3, ..., y m ] T The subsequence matrix is ​​constructed as follows:

[0094]

[0095] Where: x: influencing factor; y: comparison value; n, m: number;

[0096] (2) Preprocess the data in the steel production database. Different elements have different dimensions and data ranges, so they need to be normalized to reduce their quantity and unify them into an approximate range, with a focus on their changes and trends.

[0097]

[0098] in: Mean; Normalized result.

[0099] (3) Calculate the grey relational coefficient. Calculate the correlation coefficient between each indicator in the subsequence and the parent sequence. Let:

[0100] a = min i min k |x0(k)-x i (k)|

[0101] b = max i max k |x0(k)-x i (k)|

[0102] structure:

[0103]

[0104] Where a, b: calculation reference coefficients; ρ: resolution coefficient; z k.j ξ j (k): Correlation coefficient.

[0105] (4) Calculate the correlation degree.

[0106]

[0107] Where, r j For correlation degree.

[0108] Step 4022: Neural Network Prediction. Neural network prediction is a popular prediction method in current industrial applications. Based on the grey relational analysis in Step 4021, after identifying relevant influencing factors, the neural network prediction algorithm can be used to obtain a mathematical model (product output, temperature prediction, composition prediction, etc.) that better reflects the actual production situation. Neural network prediction includes six steps: data normalization, data partitioning, network setup, data training, data validation, and testing generalization ability. The process is as follows: Figure 3 As shown.

[0109] Adaptive optimization. The establishment of mechanistic models needs to be aligned with actual production conditions. Differences exist between mechanistic analyses in the literature and actual production, necessitating optimization of the models in the literature to improve their applicability. Machine learning algorithms possess excellent adaptive capabilities, allowing for updates and optimization of mechanistic models based on on-site steel production data, thereby better adapting to changes in process conditions and uncertainties in industrial production.

[0110] Anomaly diagnosis and fault analysis. Machine learning is highly effective in analyzing data and promptly identifying anomalies and faults in industrial processes. Clustering algorithms within machine learning are particularly useful.

[0111] Step 5: Based on the mutual complementation and coupling of the steel production process mechanism model and the trained black box model, establish an energy consumption prediction model for predicting energy consumption factors in the steel production process.

[0112] For the steel production process, we summarized the current research findings through literature review and constructed a mechanistic model.

[0113] Based on the energy consumption calculation formula, and according to the needs of the production site, the source of media consumption data is determined, namely, data acquisition, mechanistic model prediction, or black box model prediction.

[0114] By constructing a black box model, we can address uncertainties in the mechanistic model and provide information that is needed for energy consumption prediction but lacks a mechanistic model. We can then use neural network prediction methods to establish a black box model that better reflects actual production conditions.

[0115] Identify the relationship between the steel production process mechanism model and the trained black box model, determine the input-output relationship, and identify the coupling point between the steel production process mechanism model and the trained black box model;

[0116] The input to the energy consumption prediction model comes from on-site collected data, i.e., the production database;

[0117] The output of the energy consumption prediction model is the predicted energy consumption value;

[0118] Coupling points: (1) Input data required by the mechanism model, which is difficult to collect on-site and lacks actual data, but can be predicted by the black box model; (2) Environmental parameters and reaction parameters required by the mechanism model, which are difficult to collect in practice but can be predicted by the black box model; (3) Complex mechanism models that require many parameters, but these parameters are difficult to collect or implement, but can be simplified by the black box model, etc.

[0119] Based on the coupling point, a coupling mechanism between the steel production process mechanism model and the trained black box model is established, and the coupling constraints are determined.

[0120] Coupling constraints include production site data, upper and lower limits of materialized processes, and equipment operating conditions.

[0121] By establishing a data combination between the iron production process mechanism model and the trained black box model, and adjusting the parameters in the coupled model in a timely manner through a dynamic feedback mechanism, energy consumption prediction is performed to obtain an energy consumption prediction model.

[0122] Through machine learning modeling, manufacturing enterprises can better establish relevant production models that are more consistent with actual production. During actual production, these models can be coupled to create energy consumption prediction models, allowing for prediction of energy consumption under known production conditions. Taking electric arc furnace steelmaking in the steel production process as an example, the calculation method of the energy consumption prediction model is shown below. The consumption and recovery of different energy media largely come from the model's predictions. The energy consumption model is a predictive model, not simply data collection. A complex and accurate energy consumption prediction model provides the analytical foundation for subsequent determination of optimal energy consumption and energy efficiency diagnosis.

[0123] The energy consumption calculation formula for the steel production process is expressed as follows:

[0124]

[0125] Among them: Qall Energy consumption forecast; Q use_i Q represents the amount of energy consumed. rec_i Q: represents the amount of energy recovered, k: represents the number of types of energy consumed, m: represents the number of types of energy recovered; use_elec Power consumption; Q use_cog Coke oven gas consumption; Q use_lng Natural gas consumption; Q use_water : Water consumption; Q use_lqi Low-pressure steam consumption; Q use_mqi Medium-pressure steam consumption; Q use_n2 Nitrogen consumption; Q use_hn2 High-pressure nitrogen consumption; Q use_ar Argon consumption; Q use_o2 Oxygen consumption; Q rec_lqi Low-pressure steam recovery rate; Q rec_zha Slag heat recovery;

[0126] This represents the energy consumption from the j-th type of energy consumption to the k-th type of energy consumption; This represents the amount of energy recovered from the nth type of recovered energy to the mth type of recovered energy.

[0127] Furthermore, the optimal energy consumption is determined. Through the coupling of mechanism and data-driven approaches, an energy consumption prediction model is established. Various factors affecting energy consumption during the production process are reflected in the relationships between variables within the model. The magnitude and rate of influence of different factors on energy consumption can be quantitatively analyzed. Machine learning algorithms such as particle swarm optimization and genetic algorithms (e.g., diploid genetic algorithms) can be used to optimize complex energy consumption models. After determining the range of variation for different factors during the production process, the optimal energy consumption during production can be determined through iterative computer calculations.

[0128] The optimal energy consumption determined in this way, i.e. the model's optimal energy consumption, may not be achievable in reality. For example, the optimal energy consumption may be difficult to meet the production operation requirements. In order to combine with actual production, historical optimal energy consumption can also be obtained from actual production. By optimizing historical production data, the optimal energy consumption value that can meet the production conditions can be found and used as the benchmark for analysis. Both the model's optimal energy consumption and the historical optimal energy consumption are benchmarks for analysis, and both meet the requirement of relative objectivity and can be selected by production personnel.

[0129] The diploid genetic algorithm is a model optimization algorithm that can be used as a reference. The outflow graph of the diploid genetic algorithm is shown below. Figure 4 As shown. The main process includes analyzing the problem, determining the chromosomes, determining the population, calculating fitness, crossover, mutation, dominant / recessive rearrangements, etc. Some key calculation formulas are shown below:

[0130] If the objective function is a maximization problem, then the fitness function can be taken as:

[0131] Fit(f(x)) = f(x)

[0132] If the objective function is a minimization problem, then the fitness function can be taken as:

[0133]

[0134] The fitness function is sized and transformed, and the new fitness function f′ is transformed as follows:

[0135] f′=af+b

[0136] in:

[0137] Monte Carlo method is the most commonly used selection method in genetic algorithms, and the probability of an individual being selected is:

[0138]

[0139] The most commonly used sorting model is the linear sorting model, which was proposed by Beck:

[0140]

[0141] An energy efficiency intelligent diagnostic method based on any one of the optimal energy consumption prediction methods for steel production processes includes the following steps:

[0142] Based on the characteristics of each process in steel production, optimal energy consumption, and energy consumption prediction models, an intelligent energy consumption diagnosis model is established that analyzes the primary, secondary, and tertiary influencing factors of the steel production process.

[0143] Based on the intelligent energy consumption diagnostic model, the energy consumption and energy efficiency in the actual steel production process can be analyzed to determine the reasons for the fluctuations in energy consumption in the actual production process.

[0144] Furthermore, the energy efficiency diagnostic analysis process involves comparing the current energy consumption and efficiency during production with optimal energy consumption. Intelligent energy efficiency diagnosis aims to identify the reasons for changes in current energy consumption and efficiency compared to optimal energy consumption, providing supporting data results and offering clear cause analysis and operational suggestions to energy management units and on-site operators. Intelligent energy efficiency diagnosis is based on an energy consumption prediction model, and its steps are as follows:

[0145] Step 701: Determine the optimal operating conditions for the steel production process. Based on the determined optimal energy consumption, the values ​​of each variable in the energy consumption prediction model are set as baseline values ​​to obtain the baseline operating conditions under the optimal energy consumption state.

[0146] Step 702: Actual Energy Efficiency Calculation. Based on the collected data from actual production, the actual energy efficiency value is calculated by inputting it into the model.

[0147] Step 703: Univariate analysis.

[0148] By comparing data from optimal and actual operating conditions, variables with discrepancies are input into the energy consumption intelligent diagnostic model for calculation. The impact of these variables on the consumption and recovery of various energy media and on total energy consumption is analyzed. Single-factor analysis can further refine the impact of individual factors on energy efficiency.

[0149] Taking electric arc furnace steelmaking in the steel production process as an example, the single-factor analysis includes the analysis of factors such as the temperature of molten iron entering the furnace, the preheating temperature of scrap steel, the composition of molten iron, the consumption of gas, the consumption of scrap steel, the consumption of molten iron, the amount of oxygen blown, and the steam recovery analysis.

[0150] Analysis of three levels of influencing factors. Based on the results of single-factor analysis, a three-level influencing factor analysis model can be established. The first, second, and third-level factors are determined according to the mutual influence between factors. Factors that directly affect energy consumption are first-level factors, while second and third-level factors are indirect factors affecting the first-level factors. Taking electric arc furnace steelmaking in the steel production process as an example, first-level factors include raw material factors, gas factors, tapping factors, and other power medium factors; second-level factors include molten iron composition, scrap steel consumption, molten iron ratio, gas consumption, and power medium consumption; and third-level factors include molten iron carbon content and silicon content.

[0151] Figure 5 This example demonstrates the method of using an electric arc furnace for steelmaking.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting optimal energy consumption in steel production processes, characterized in that: Includes the following steps: Establish a steel production database based on the steel production process, using furnaces or batches as units; Establish a mechanism model for steel formation process; A black-box model is constructed based on correlation analysis and neural network model, and the black-box model is trained. The process of constructing a black-box model based on correlation analysis and neural network model, and training the black-box model is as follows: S301: Determine the parameters of the neural network model and optimize them; S302: Construct a black-box model based on correlation analysis and neural network model to analyze the mechanism of steel formation process; S303: Update and optimize the black box model based on on-site data of the steel production process to adapt to changes in process conditions and uncertainties in industrial production, and obtain a well-trained black box model. An energy consumption prediction model is established based on the mutual complementation and coupling of the steel production process mechanism model and the trained black box model to predict the energy consumption factors in the steel production process. The process of establishing an energy consumption prediction model for predicting energy consumption factors in the steel production process by complementing and coupling a steel production process mechanism model and a trained black box model is as follows: Identify the relationship between the steel production process mechanism model and the trained black box model, determine the input-output relationship, and identify the coupling point between the steel production process mechanism model and the trained black box model; Based on the coupling point, a coupling mechanism between the steel production process mechanism model and the trained black box model is established, and the coupling constraints are determined. By combining data from the steel production process mechanism model and the trained black box model, and adjusting parameters in real time through a dynamic feedback mechanism, energy consumption prediction is obtained, thus producing an energy consumption prediction model. The energy consumption of energy-consuming factors in the steel production process is determined by using machine learning methods based on an energy consumption prediction model. Then, the optimal energy consumption of the steel production process is determined by the energy consumption calculation formula, and the historical optimal energy consumption is determined based on historical data.

2. The optimal energy consumption prediction method for steel production process according to claim 1, characterized in that: The establishment of the steel production database includes material consumption data, product data, composition parameters, physical property parameters, operating parameters, and technical parameter information. The database construction requirements are that each product furnace or batch corresponds one-to-one.

3. The optimal energy consumption prediction method for steel production process according to claim 1, characterized in that: The energy consumption calculation formula for the steel production process is expressed as follows: Among them: Q all Energy consumption forecast; Q use_i Q represents the amount of energy consumed. rec_i Q: represents the amount of energy recovered, k: represents the number of types of energy consumed, m: represents the number of types of energy recovered; use_elec Power consumption; Q use_cog Coke oven gas consumption; Q use_lng Natural gas consumption; Q use_water : Water consumption; Q use_lqi Low-pressure steam consumption; Q use_mqi Medium-pressure steam consumption; Q use_n2 Nitrogen consumption; Q use_hn2 High-pressure nitrogen consumption; Q use_ar Argon consumption; Q use_o2 Oxygen consumption; Q rec_lqi Low-pressure steam recovery rate; Q rec_zha Slag heat recovery.

4. The optimal energy consumption prediction method for steel production process according to claim 1, characterized in that: The machine learning method used is either particle swarm optimization or diploid genetic algorithm.

5. The optimal energy consumption prediction method for steel production process according to claim 1, characterized in that: The process of determining the energy consumption of energy-consuming factors in the steel production process using machine learning methods based on an energy consumption prediction model, determining the optimal energy consumption of the steel production process using the energy consumption calculation formula, and determining the historical optimal energy consumption based on historical data is as follows: The energy consumption prediction model is optimized by using the particle swarm optimization algorithm or the diploid genetic algorithm in the genetic algorithm of machine learning. The range of variation of different factors in the production process is determined, and the optimal energy consumption in the production process is determined by iterative calculation. By optimizing historical production data, the optimal energy consumption value that meets production conditions is found, and the historical optimal energy consumption is obtained.

6. The energy efficiency intelligent diagnosis method for the optimal energy consumption prediction method in steel production process according to any one of claims 1-5, characterized in that: Includes the following steps: Based on the characteristics of each process in steel production, optimal energy consumption, and energy consumption prediction models, an intelligent energy consumption diagnosis model is established that analyzes the primary, secondary, and tertiary influencing factors of the steel production process. Based on the intelligent energy consumption diagnostic model, the energy consumption and energy efficiency in the actual steel production process can be analyzed to determine the reasons for the fluctuations in energy consumption in the actual production process.

7. The energy efficiency intelligent diagnosis method for the optimal energy consumption prediction method in steel production process according to claim 6, characterized in that: The process of analyzing energy consumption and energy efficiency in actual steel production based on the intelligent energy consumption diagnostic model, and determining the causes of energy consumption fluctuations in actual steel production, is as follows: Determine the optimal operating conditions for the steel production process. Based on the determined optimal energy consumption, set the values ​​of energy consumption factor variables in the energy consumption prediction model as benchmark values ​​to obtain the benchmark operating conditions under the optimal energy consumption state. Actual energy efficiency calculations are performed by inputting collected data from actual production into the intelligent energy consumption diagnostic model to obtain the actual energy efficiency value. By comparing the data of the optimal and actual operating conditions of the steel production process, the variables with differences are substituted into the intelligent energy consumption diagnosis model for calculation. Through sensitivity analysis and regression analysis, the impact on the consumption and recovery of various energy media and the impact on total energy consumption are realized to conduct a single-factor analysis of the energy consumption impact of the steel production process. The results of single-factor analysis enable the analysis of the causes of energy consumption fluctuations in the actual steel production process.

Images