A furnace intelligent control system and method

Abstract

The present invention relates to a furnace intelligent control system and method, belonging to the field of furnace intelligent control technology, the system includes a data acquisition module, an automatic control module, a human-computer interaction module and a data storage module, the data acquisition module, the automatic control module, the human-computer interaction module and the data storage module are sequentially connected in communication; the automatic control module is used to automatically adjust the working parameters of the furnace according to the control decision through the collected real-time data, and issue control instructions to control the fuel supply and cooling water flow; the data acquisition module is used to collect real-time data through sensor equipment and perform preliminary processing on the data; the data storage module is used to store the collected real-time data and classify the data according to the different operating conditions established. The present invention realizes precise automatic control through control decision, realizes the optimization of the smelting process, and at the same time, by adopting a multi-objective optimization algorithm to obtain the fuel supply and cooling water flow, the working parameters are adjusted in real time to reduce energy use.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent control of melting furnaces, and in particular relates to an intelligent control system and method for a melting furnace. Background Art

[0002] The intelligent control system of the furnace plays a vital role in the metallurgical industry, especially in the smelting process of aluminum alloys, steel and other metals, which can realize efficient, precise and automated production.

[0003] However, the existing control system has difficulty in intelligent regulation under complex operating conditions of the furnace. When dealing with complex, nonlinear, and multi-variable control problems, insufficient control accuracy or delayed response may occur. At the same time, the control of control variables such as fuel supply and cooling water flow cannot achieve a balanced and coordinated state, resulting in inaccurate temperature control and energy waste. Summary of the invention

[0004] In order to solve the above problems existing in the prior art, the present invention provides a furnace intelligent control system and method, which realizes precise automatic control through control decision-making, realizes the optimization of the smelting process, and at the same time, obtains the fuel supply and cooling water flow by adopting a multi-objective optimization algorithm to adjust the working parameters in real time and reduce energy use.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A first aspect of the present disclosure provides a furnace intelligent control system, comprising a data acquisition module, an automatic control module, a human-computer interaction module and a data storage module, wherein the data acquisition module, the automatic control module, the human-computer interaction module and the data storage module are sequentially communicatively connected;

[0007] The automatic control module is used to automatically adjust the working parameters of the furnace according to the control decision through the collected real-time data, and issue control instructions to control the fuel supply and cooling water flow;

[0008] Among them, control decisions include:

[0009] Match target parameters according to process requirements, and generate control instructions based on the collected real-time data to obtain the optimal combination of fuel supply and cooling water flow that achieves the target parameters; the target parameters include the furnace target temperature and the type of operating condition;

[0010] According to the furnace operation data, identify the furnace operation status, determine whether the real-time data is within the normal threshold range under the current operating conditions, and issue an early warning when the data is abnormal;

[0011] Collect all operating data in stages and update the lag time of furnace temperature response for different operating conditions.

[0012] As a preferred technical solution of the present invention, the step of obtaining an optimal combination of fuel supply amount and cooling water flow rate to achieve target parameters comprises the following steps:

[0013] Define parameters: Fuel supply F(t) and cooling water flow C(t) are used as decision variables, and the current furnace temperature T(t) and target temperature T are used as decision variables. s is the state variable, and the furnace temperature response lag time is τ ;

[0014] Establish energy consumption model: According to the decision variables of furnace temperature control, the total energy consumption E is expressed as the weighted sum of fuel and cooling water consumption, expressed as:

[0015] ;

[0016] In the formula, a and b are the cost coefficients of unit fuel and cooling water respectively;

[0017] Set constraints: Set constraints based on the supply range of fuel and cooling water, the temperature range of safe operation, and the degree of fluctuation of dynamic response;

[0018] Establishing the objective function: Establishing the objective function based on the furnace temperature response lag prediction and temperature control energy consumption, the calculation formula is:

[0019] ;

[0020] Where w1 and w2 are the weight coefficients of temperature control accuracy and temperature control energy consumption, respectively. p (t+τ) is the predicted temperature value at time t+τ;

[0021] Solve the optimal solution: Use a multi-objective optimization algorithm to obtain the optimal combination of fuel supply and cooling water flow.

[0022] As a preferred technical solution of the present invention, the constraint conditions include:

[0023] F min ≤F(t)≤F max , C min ≤C(t)≤C max ;

[0024] T min ≤T(t)≤T max ;

[0025] |F(t)-F(t-Δt)|≤ΔF max , |C(t)-C(t-Δt)|≤ΔC max ;

[0026] In the formula, F min and Fmax Respectively represent the lower and upper limits of fuel supply, C min and C max Respectively represent the lower and upper limits of cooling water flow, ΔF max Indicates the maximum allowable value of fuel supply change, ΔC max It indicates the maximum allowable value of cooling water flow rate change, and Δt is the sampling interval.

[0027] As a preferred technical solution of the present invention, the furnace temperature response lag time τ is the delayed response of the system input change to the output, and the temperature prediction value at time t+τ is obtained by using the lag response described by the first-order dynamic characteristics:

[0028] ;

[0029] Where k is the response coefficient of temperature change to fuel and cooling water changes, F(t-τ) and C(t-τ) are the fuel supply and cooling water flow rates before the lag time, respectively.

[0030] As a preferred technical solution of the present invention, the method of finding the optimal solution comprises the steps of:

[0031] Initialize the population: randomly generate N groups of candidate solutions, each solution consists of decision variables [F(t), C(t)], and satisfies the constraints so that the initial population covers the range of solutions;

[0032] Evaluate fitness: Calculate the objective function J for each solution [F(t), C(t)] in the population;

[0033] Non-dominated sorting: define non-dominated solutions, find solutions that are not dominated by any solution, form the first-level frontier, remove the first-level frontier solutions, find the non-dominated solutions among the remaining solutions, form the second-level frontier, and so on, until all solutions are stratified;

[0034] Crowding distance calculation: Sort the solutions by each target and calculate the crowding distance d of each solution i in the target dimension i :

[0035] ;

[0036] In the formula, J i+1 and J i-1 is the neighboring solution of the current solution on the target dimension, J max and J min is the value range of the target;

[0037] The total crowding distance is the sum of the distances in all target dimensions;

[0038] Select the population: Sort by the level of the non-dominated layer, sort by descending crowding distance within the same layer, give priority to solutions with large crowding distance, and select the next generation population according to the limit of population size N;

[0039] Crossover and mutation: Use simulated binary crossover to generate two offspring, and use polynomial mutation to randomly add small perturbations to each decision variable;

[0040] Merge and update: merge the parent and offspring to form a temporary population of size 2N, re-evaluate the non-dominated level of the merged population, select from low to high according to the non-dominated level, and if it exceeds the population size N, sort by crowding distance and delete redundant individuals;

[0041] Output solution set: Set termination conditions, stop after reaching a predetermined number of iterations, and output the set of non-dominated solutions in the population.

[0042] As a preferred technical solution of the present invention, the updating of the hysteresis time of the furnace temperature response of different operating conditions includes the following steps:

[0043] Data collation: Create training and test data sets by matching fuel supply and cooling water flow according to the timestamps of the analysis data sets;

[0044] Construct lag variables: Determine the appropriate lag step size (n, m) to generate features with different time lags for the fuel supply F(t) and cooling water flow C(t):

[0045] Generate hysteresis characteristics (F(t-1), F(t-2), ..., F(tn)) for the fuel supply amount F(t);

[0046] Generate hysteresis characteristics (C(t-1), C(t-2), ..., C(tm)) for the cooling water flow rate C(t);

[0047] Modeling and training: Select the long short-term memory network (LSTM) model and establish input and output:

[0048] Input variables: [F(t), F(t-1), F(t-2), …, F(tn), C(t), C(t-1), C(t-2), …, C(tm), L(t), P(t)], where L(t) is the liquid level and P(t) is the pressure; output variable: T(t).

[0049] As a preferred technical solution of the present invention, the updating of the hysteresis time of the furnace temperature response of different operating conditions also includes the steps of:

[0050] Split the time series data into input and output pairs, and create training data using a sliding window approach; set LSTM hyperparameters, including the number of hidden layers, the number of neurons in each layer, the learning rate, and the time step length, and perform training. Use mean square error as the loss function to optimize the model, and use regularization and Dropout methods to reduce overfitting;

[0051] Lag time extraction: Input the disturbance signal to the trained LSTM model, observe the furnace temperature change trend output by the model, and record the time delay between the input variable change and the furnace temperature response;

[0052] Record the time point t0 when the input variable changes;

[0053] Record the time point t when the furnace temperature begins to change significantly r ;

[0054] The lag time is: .

[0055] As a preferred technical solution of the present invention, the data acquisition module is used to collect real-time data through sensor equipment and perform preliminary processing on the data, wherein the sensor equipment includes a temperature sensor, a liquid level sensor, a gas sensor and a pressure sensor;

[0056] The preliminary processing of the data includes:

[0057] Classify furnace operating conditions according to processing technology and set normal threshold ranges for liquid level, gas composition and pressure under different operating conditions;

[0058] Mark abnormal data according to the operating condition type and normal threshold range, retain normal data and establish analysis data set according to timestamp;

[0059] The sliding average method was used to remove high-frequency noise in the signal, and the timestamps of the collected variables were corrected through cross-correlation analysis to complete data alignment.

[0060] As a preferred technical solution of the present invention, the human-computer interaction module is used to display the real-time data, status information and historical records of the furnace, and allows the operator to manually intervene and set, and when an abnormal situation is monitored, an audible and visual alarm is issued in real time;

[0061] The data storage module is used to store the collected real-time data and classify the data according to different established operating condition types.

[0062] A second aspect of the present disclosure provides a furnace intelligent control method, which is applied to a furnace intelligent control system as described above, comprising the following steps:

[0063] S1, obtain temperature, liquid level, gas composition and pressure data in real time through the data acquisition module, and save the acquired data into the data storage module after preliminary processing;

[0064] S2. Obtain the target temperature and the lag time of the furnace temperature response according to the current processing technology and operating condition type;

[0065] S3, obtaining an optimal combination of fuel supply and cooling water flow rate through a multi-objective optimization algorithm according to the target temperature and the current temperature, and generating a control instruction to adjust the fuel supply and cooling water flow rate;

[0066] S4, periodically collect fuel supply, cooling water flow, liquid level and pressure data as input, and use temperature as output to establish an input-output relationship through a machine learning model, and update the lag time of furnace temperature response under different working conditions;

[0067] S5. Execute the above steps in a loop, monitor the operation status of the furnace in real time and adjust the fuel supply and cooling water flow.

[0068] The beneficial effects of the present invention are:

[0069] The present invention first collects the real-time data of the furnace and monitors the operating status to issue a warning, then based on the collected real-time data, combined with the current operating condition type and the furnace temperature response time to compare the target temperature, a multi-objective optimization algorithm is used to solve the optimal combination of fuel supply and cooling water flow, and the adjustment of fuel supply and cooling water flow is reasonably coordinated, thereby improving the accuracy of furnace temperature control and minimizing energy consumption; at the same time, a machine learning model is established through all the collected variables to obtain the lag time of the furnace temperature response, thereby improving the control accuracy and response speed, thereby reducing over-adjustment and under-adjustment, and avoiding system oscillation.

[0070] The present invention realizes precise automatic control through control decision, avoids mistakes caused by human operation, optimizes the smelting process and improves production efficiency; comprehensively monitors the furnace status and issues early warning in time to reduce the risk of accidents and ensure the life safety of operators; and optimizes the smelting process and adjusts working parameters in real time by adopting a multi-objective optimization algorithm to obtain the fuel supply and cooling water flow, thereby reducing energy use. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0072] Figure 1 A schematic diagram of the structure of a furnace intelligent control system provided by an embodiment of the present invention;

[0073] Figure 2 A schematic flow chart of a furnace intelligent control method provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0074] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation method, structure, characteristics and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0075] The intelligent control system and method of a furnace are important technical means to achieve efficient, accurate and automated production in the modern metallurgical industry. Specifically, the intelligent control system and method of a furnace of the present invention are described in detail through the following content:

[0076] This embodiment provides a furnace intelligent control system, such as Figure 1 As shown, it includes a data acquisition module, an automatic control module, a human-computer interaction module and a data storage module, and the data acquisition module, the automatic control module, the human-computer interaction module and the data storage module are communicatively connected in sequence.

[0077] The data acquisition module is used to collect real-time data through sensor devices and perform preliminary processing on the data, wherein the sensor devices include temperature sensors, liquid level sensors, gas sensors and pressure sensors;

[0078] The preliminary processing of the data includes:

[0079] Classify furnace operating conditions according to processing technology and set normal threshold ranges for liquid level, gas composition and pressure under different operating conditions;

[0080] Mark abnormal data according to the operating condition type and normal threshold range, retain normal data and establish analysis data set according to timestamp;

[0081] The sliding average method was used to remove high-frequency noise in the signal, and the timestamps of the collected variables were corrected through cross-correlation analysis to complete data alignment.

[0082] It should be noted that the timestamp correction is to find the lag time corresponding to the maximum value by calculating the cross-correlation function when the system dynamics or sensor response introduces a delay, and verify the lag by comparing the time series of the main variable and the auxiliary variable, such as the main variable is the feed flow rate and the auxiliary variable is the liquid level. By correcting the variable time lag caused by differences in acquisition equipment, signal transmission time or other systematic delays, the accuracy of system response and the reliability of data analysis can be improved.

[0083] It can be understood that the temperature sensor is used to monitor the temperature changes in the furnace in real time to ensure that the smelting process is carried out within the optimal temperature range; the liquid level sensor is used to detect the liquid level of the molten aluminum pool to prevent safety accidents caused by excessively high or low liquid levels; the gas sensor is used to monitor the gas composition in the furnace (such as , , NOx, etc.), to ensure environmental safety during the smelting process; the pressure sensor is used to monitor the pressure changes inside the furnace to ensure the stability of the smelting process. It should be noted that in the process of preliminary data processing, liquid level data, gas composition and pressure data are used as auxiliary variables to eliminate abnormal data. Among them, gas composition is mainly used for the classification of working conditions and abnormal warning, and liquid level data and pressure data are used as auxiliary variables to obtain the furnace temperature response lag time.

[0084] The automatic control module is used to automatically adjust the working parameters of the furnace according to the control decision through the collected real-time data, and issue control instructions to control the fuel supply and cooling water flow.

[0085] Among them, control decisions include:

[0086] The target parameters are matched according to the process requirements, and the control instructions are generated based on the collected real-time data to obtain the optimal combination of fuel supply and cooling water flow to achieve the target parameters; the target parameters include the furnace target temperature and the operating condition type.

[0087] According to the furnace operation data, identify the furnace operation status, determine whether the real-time data is within the normal threshold range under the current operating conditions, and issue an early warning when the data is abnormal;

[0088] Collect all operating data in stages and update the lag time of furnace temperature response for different operating conditions.

[0089] It should be noted that, in this embodiment, the furnace temperature response lag time of different operating conditions is first obtained and applied to the control decision of actual fuel supply and cooling water flow, thereby improving the control accuracy and response speed, thereby reducing over-regulation and under-regulation and avoiding system oscillation.

[0090] The step of obtaining the optimal combination of fuel supply and cooling water flow to achieve the target parameters includes the following steps:

[0091] Define parameters: Fuel supply F(t) and cooling water flow C(t) are used as decision variables, and the current furnace temperature T(t) and target temperature T are used as decision variables. s is the state variable, and the furnace temperature response lag is τ Unit time;

[0092] Establish energy consumption model: According to the decision variables of furnace temperature control, the total energy consumption E is expressed as the weighted sum of fuel and cooling water consumption, expressed as:

[0093] ;

[0094] Where a and b are the cost coefficients of unit fuel and cooling water respectively.

[0095] Set constraints: Set constraints based on the supply range of fuel and cooling water, the temperature range of safe operation, and the degree of fluctuation of dynamic response, including:

[0096] F min ≤F(t)≤F max , C min ≤C(t)≤C max ;

[0097] T min ≤T(t)≤T max ;

[0098] |F(t)-F(t-Δt)|≤ΔF max , |C(t)-C(t-Δt)|≤ΔC max ;

[0099] In the formula, F min and F max Respectively represent the lower and upper limits of fuel supply, C min and C max Respectively represent the lower and upper limits of cooling water flow, ΔF max Indicates the maximum allowable value of fuel supply change, ΔC max It indicates the maximum allowable value of cooling water flow rate change, and Δt is the sampling interval.

[0100] Establishing the objective function: Establishing the objective function based on the furnace temperature response lag prediction and temperature control energy consumption, the calculation formula is:

[0101] ;

[0102] Where w1 and w2 are the weight coefficients of temperature control accuracy and temperature control energy consumption, respectively. p (t+τ) is the predicted temperature value at time t+τ.

[0103] It should be noted that the lag time τ describes the delayed response of the system input change to the output, and the first-order dynamic characteristics are used to describe the lag response:

[0104] ;

[0105] Where k is the response coefficient of temperature change to fuel and cooling water changes, F(t-τ) and C(t-τ) are the fuel supply and cooling water flow rates before the lag time, respectively.

[0106] Solving the optimal solution: Use a multi-objective optimization algorithm to obtain the optimal combination of fuel supply and cooling water flow, including the following steps:

[0107] Initialize the population: randomly generate N groups of candidate solutions, each solution consists of decision variables [F(t), C(t)], and satisfies the constraints so that the initial population covers the range of solutions;

[0108] Evaluate fitness: Calculate the objective function J for each solution [F(t), C(t)] in the population;

[0109] Non-dominated sorting: define non-dominated solutions, find solutions that are not dominated by any solution, form the first-level frontier, remove the first-level frontier solutions, find the non-dominated solutions among the remaining solutions, form the second-level frontier, and so on, until all solutions are stratified;

[0110] Among them, the definition of non-dominated solution, a solution A is dominated by another solution B under the condition:

[0111] (1) A is no worse than B in all objectives:

[0112] For the objective functions J1, J2, ..., J in the minimization problem m :

[0113] ;

[0114] Or, for the maximization problem:

[0115] ;

[0116] (2) A is superior to B in at least one objective:

[0117] There exists some target j:

[0118] J j (A) <J j (B) (or J j (A)>J j (B) For maximization problems);

[0119] If solution B satisfies these two conditions, then solution B is considered to dominate solution A.

[0120] Crowding distance calculation: Within the same layer, measure the sparsity of the solution, sort the solutions by each target, and calculate the crowding distance d of each solution i in the target dimension i :

[0121] ;

[0122] In the formula, J i+1 and J i-1 are the two nearest neighbors of the current solution on the target dimension, J max and J min is the value range of the target;

[0123] The total crowding distance is the sum of distances in all target dimensions.

[0124] It should be noted that J i+1 -J i-1 represents the difference in target values ​​between adjacent individuals, reflecting local sparsity; J max -J min Indicates the overall range of target values, used for normalization. The larger the crowding distance, the sparser the individuals are, and they are preferred for retention.

[0125] Select the population: Sort by the level of the non-dominated layer, sort by descending order of crowding distance within the same layer, give priority to solutions with large crowding distance, and select the next generation population based on the limit of population size N.

[0126] Crossover and mutation: Use simulated binary crossover (SBX) to generate two offspring, and use polynomial mutation to randomly add small perturbations to each decision variable.

[0127] Merge and update: merge the parent and child generations to form a temporary population of size 2N, re-evaluate the non-dominated level of the merged population, select from low to high according to the non-dominated level, and if it exceeds the population size N, sort by crowding distance and delete redundant individuals.

[0128] Output solution set: Set termination conditions, stop after reaching a predetermined number of iterations, and output the set of non-dominated solutions in the population.

[0129] It is understandable that the working parameters of the furnace include temperature, pressure, flow, etc. The working parameters are adjusted to maintain the optimal operating state, and the working state of the actuator is adjusted by issuing control instructions to control the fuel supply and cooling water flow.

[0130] Further, updating the hysteresis time of the furnace temperature response of different operating conditions includes the following steps:

[0131] Data collation: Create training and test data sets by matching fuel supply and cooling water flow according to the timestamps of the analysis data sets;

[0132] Construct hysteresis variables: Analyze the correlation between variables and outputs from historical data (can also be determined by testing different step ranges), determine the appropriate hysteresis step (n, m), and generate characteristics of different time lags for the fuel supply F(t) and cooling water flow C(t):

[0133] Generate hysteresis characteristics (F(t-1), F(t-2), ..., F(tn)) for the fuel supply amount F(t);

[0134] A hysteresis characteristic (C(t-1), C(t-2), ..., C(tm)) is generated for the cooling water flow rate C(t).

[0135] Modeling and training: Select the long short-term memory network (LSTM) model and establish input and output:

[0136] Input variables: [F(t), F(t-1), F(t-2), …, F(tn), C(t), C(t-1), C(t-2), …, C(tm), L(t), P(t)], where L(t) is the liquid level and P(t) is the pressure; output variable: T(t).

[0137] The time series data is split into input and output pairs, and the training data is created using a sliding window approach; for example, the input data is the features of the past five time steps, and the output is the furnace temperature of the current time step.

[0138] Set LSTM hyperparameters, including the number of hidden layers (such as 1-3 layers), the number of neurons in each layer, the learning rate (such as 0.001), and the time step length (such as 5), perform training, use the mean square error (MSE) as the loss function, optimize the model, and reduce overfitting through regularization (such as L2 regularization) and Dropout methods.

[0139] Lag time extraction: Input the disturbance signal to the trained LSTM model, observe the furnace temperature change trend output by the model, and record the time delay between the input variable change and the furnace temperature response.

[0140] Record the time point t0 when the input variable changes;

[0141] Record the time point t when the furnace temperature begins to change significantly r ;

[0142] The lag time is: .

[0143] It is understandable that the Long Short-Term Memory Network (LSTM) can effectively extract the lag time between input variables and output variables by capturing the long-term and short-term dependencies in time series data. Combined with feature engineering and response analysis methods, the extracted lag time can be applied to obtain the optimal combination of fuel supply and cooling water flow to achieve the target parameters, thereby improving the energy saving effect, stability and production efficiency of the temperature control system.

[0144] The human-computer interaction module is used to display the real-time data, status information and historical records of the furnace, and allows the operator to manually intervene and set. When an abnormal situation is detected, an audible and visual alarm is issued in real time to ensure that the operator responds in time.

[0145] The data storage module is used to store the collected real-time data and classify the data according to different established operating condition types.

[0146] This embodiment also provides a furnace intelligent control method for implementing the above-mentioned furnace intelligent control system. Figure 2 As shown, the following steps are included:

[0147] S1, obtain temperature, liquid level, gas composition and pressure data in real time through the data acquisition module, and save the acquired data into the data storage module after preliminary processing;

[0148] S2. Obtain the target temperature and the lag time of the furnace temperature response according to the current processing technology and operating condition type;

[0149] S3, obtaining an optimal combination of fuel supply and cooling water flow rate through a multi-objective optimization algorithm according to the target temperature and the current temperature, and generating a control instruction to adjust the fuel supply and cooling water flow rate;

[0150] S4, periodically collect fuel supply, cooling water flow, liquid level and pressure data as input, and use temperature as output to establish an input-output relationship through a machine learning model, and update the lag time of furnace temperature response under different working conditions;

[0151] S5. Execute the above steps in a loop, monitor the operation status of the furnace in real time and adjust the fuel supply and cooling water flow.

[0152] The present invention first collects the real-time data of the furnace and monitors the operating status to issue a warning, then based on the collected real-time data, combined with the current operating condition type and the furnace temperature response time to compare the target temperature, a multi-objective optimization algorithm is used to solve the optimal combination of fuel supply and cooling water flow, and the adjustment of fuel supply and cooling water flow is reasonably coordinated, thereby improving the accuracy of furnace temperature control and minimizing energy consumption; at the same time, a machine learning model is established through all the collected variables to obtain the lag time of the furnace temperature response, thereby improving the control accuracy and response speed, thereby reducing over-adjustment and under-adjustment, and avoiding system oscillation.

[0153] The present invention realizes precise automatic control through control decision, avoids mistakes caused by human operation, optimizes the smelting process and improves production efficiency; comprehensively monitors the furnace status and issues early warning in time to reduce the risk of accidents and ensure the life safety of operators; and optimizes the smelting process and adjusts working parameters in real time by adopting a multi-objective optimization algorithm to obtain the fuel supply and cooling water flow, thereby reducing energy use.

[0154] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A furnace intelligent control system, characterized in that: It includes a data acquisition module, an automatic control module, a human-computer interaction module and a data storage module, wherein the data acquisition module, the automatic control module, the human-computer interaction module and the data storage module are sequentially connected in communication; The automatic control module is used to automatically adjust the working parameters of the furnace according to the control decision through the collected real-time data, and issue control instructions to control the fuel supply and cooling water flow; Among them, control decisions include: Match target parameters according to process requirements, and generate control instructions based on the collected real-time data to obtain the optimal combination of fuel supply and cooling water flow that achieves the target parameters; the target parameters include the furnace target temperature and the type of operating condition; According to the furnace operation data, identify the furnace operation status, determine whether the real-time data is within the normal threshold range under the current operating conditions, and issue an early warning when the data is abnormal; Collect all operating data in stages and update the lag time of furnace temperature response for different operating conditions; The step of obtaining an optimal combination of fuel supply amount and cooling water flow rate to achieve target parameters comprises the following steps: Define parameters: Fuel supply F(t) and cooling water flow C(t) are used as decision variables, and the current furnace temperature T(t) and target temperature T are used as decision variables. s is the state variable, and the furnace temperature response lag time is τ ; Establish energy consumption model: According to the decision variables of furnace temperature control, the total energy consumption E is expressed as the weighted sum of fuel and cooling water consumption, expressed as: E(t)=a·F(t)+b·C(t); In the formula, a and b are the cost coefficients of unit fuel and cooling water respectively; Set constraints: Set constraints based on the supply range of fuel and cooling water, the temperature range of safe operation, and the degree of fluctuation of dynamic response; Establishing the objective function: Establishing the objective function based on the furnace temperature response lag prediction and temperature control energy consumption, the calculation formula is: J=w1·∣T p (t+τ)-T s ∣+w2·E(t); Where w1 and w2 are the weight coefficients of temperature control accuracy and temperature control energy consumption, respectively. p (t+τ) is the predicted temperature value at time t+τ; Solve the optimal solution: Use a multi-objective optimization algorithm to obtain the optimal combination of fuel supply and cooling water flow.

2. The intelligent control system for a melting furnace according to claim 1, characterized in that: The constraints include: F min ≤F(t)≤F max ,C min ≤C(t)≤C max ; T min ≤T(t)≤T max ; ∣F(t)-F(t-Δt)∣≤ΔF max ,∣C(t)-C(t-Δt)∣≤ΔC max ; In the formula, F min and F max Respectively represent the lower and upper limits of fuel supply, C min and C max Respectively represent the lower and upper limits of cooling water flow, ΔF max Indicates the maximum allowable value of fuel supply change, ΔC max It indicates the maximum allowable value of cooling water flow rate change, and Δt is the sampling interval.

3. The intelligent control system for a melting furnace according to claim 1, characterized in that: The furnace temperature response lag time τ is the delayed response of the system input change to the output. The temperature prediction value at time t+τ is obtained by using the lag response described by the first-order dynamic characteristics: T p (t+τ)=T(t)+k·(F(t-τ)-C(t-τ)); Where k is the response coefficient of temperature change to fuel and cooling water changes, F(t-τ) and C(t-τ) are the fuel supply and cooling water flow rates before the lag time, respectively.

4. The intelligent control system for a melting furnace according to claim 1, characterized in that: The method for finding the optimal solution comprises the steps of: Initialize the population: randomly generate N groups of candidate solutions, each solution consists of decision variables [F(t), C(t)], and satisfies the constraints so that the initial population covers the range of solutions; Evaluate fitness: Calculate the objective function J for each solution [F(t), C(t)] in the population; Non-dominated sorting: define non-dominated solutions, find solutions that are not dominated by any solution, form the first-level frontier, remove the first-level frontier solutions, find the non-dominated solutions among the remaining solutions, form the second-level frontier, and so on, until all solutions are stratified; Crowding distance calculation: Sort the solutions by each target and calculate the crowding distance d of each solution i in the target dimension i : ； In the formula, J i+1 and J i-1 is the neighboring solution of the current solution on the target dimension, J max and J min is the value range of the target; The total crowding distance is the sum of the distances in all target dimensions; Select the population: Sort by the level of the non-dominated layer, sort by descending crowding distance within the same layer, give priority to solutions with large crowding distance, and select the next generation population according to the limit of population size N; Crossover and mutation: Use simulated binary crossover to generate two offspring, and use polynomial mutation to randomly add small perturbations to each decision variable; Merge and update: merge the parent and offspring to form a temporary population of size 2N, re-evaluate the non-dominated level of the merged population, select from low to high according to the non-dominated level, and if it exceeds the population size N, sort by crowding distance and delete redundant individuals; Output solution set: Set termination conditions, stop after reaching a predetermined number of iterations, and output the set of non-dominated solutions in the population.

5. The intelligent control system for a melting furnace according to claim 1, characterized in that: The method of updating the hysteresis time of the furnace temperature response of different operating conditions includes the following steps: Data collation: Create training and test data sets by matching fuel supply and cooling water flow according to the timestamps of the analysis data sets; Construct hysteresis variables: Set the hysteresis step size (n, m) to generate features with different time lags for the fuel supply F(t) and cooling water flow C(t): Generate hysteresis characteristics (F(t-1), F(t-2), ..., F(tn)) for the fuel supply amount F(t); Generate hysteresis characteristics (C(t-1), C(t-2), ..., C(tm)) for the cooling water flow rate C(t); Modeling and training: Select the LSTM model and establish input and output: Input variables: [F(t), F(t-1), F(t-2), …, F(tn), C(t), C(t-1), C(t-2), …, C(tm), L(t), P(t)], where L(t) is the liquid level and P(t) is the pressure; output variable: current temperature T(t).

6. The intelligent control system for a melting furnace according to claim 5, characterized in that: The updating of the hysteresis time of the furnace temperature response of different operating conditions also includes the steps of: Split the time series data into input and output pairs, and create training data using a sliding window approach; set LSTM hyperparameters, including the number of hidden layers, the number of neurons in each layer, the learning rate, and the time step length, and perform training. Use mean square error as the loss function to optimize the model, and use regularization and dropout methods to reduce overfitting; Lag time extraction: Input the disturbance signal to the trained LSTM model, observe the furnace temperature change trend output by the model, and record the time delay between the input variable change and the furnace temperature response; Record the time point t0 when the input variable changes; Record the time point t when the furnace temperature begins to change significantly r ; The lag time is: τ=t r -t0.

7. The intelligent control system for a melting furnace according to claim 1, characterized in that: The data acquisition module is used to collect real-time data through sensor devices and perform preliminary processing on the data, wherein the sensor devices include temperature sensors, liquid level sensors, gas sensors and pressure sensors; The preliminary processing of the data includes: Classify furnace operating conditions according to processing technology and set normal threshold ranges for liquid level, gas composition and pressure under different operating conditions; Mark abnormal data according to the operating condition type and normal threshold range, retain normal data and establish analysis data set according to timestamp; The sliding average method was used to remove high-frequency noise in the signal, and the timestamps of the collected variables were corrected through cross-correlation analysis to complete data alignment.

8. The intelligent control system for a melting furnace according to claim 1, characterized in that: The human-computer interaction module is used to display the real-time data, status information and historical records of the furnace, and allows the operator to manually intervene and set up. When an abnormal situation is detected, an audible and visual alarm is issued in real time; The data storage module is used to store the collected real-time data and classify the data according to different established operating condition types.

9. A furnace intelligent control method, applied to a furnace intelligent control system as claimed in any one of claims 1 to 8, characterized in that: The following steps are involved: The data acquisition module acquires temperature, liquid level, gas composition and pressure data in real time, and performs preliminary processing on the acquired data before saving it in the data storage module; According to the current processing technology and operating conditions, obtain the target temperature and the lag time of the furnace temperature response; According to the target temperature and the current temperature, the optimal combination of the fuel supply amount and the cooling water flow rate is obtained through a multi-objective optimization algorithm, and a control instruction is generated to adjust the fuel supply amount and the cooling water flow rate; The fuel supply, cooling water flow, liquid level and pressure data are collected periodically as input, and the temperature is used as output to establish the input-output relationship through the machine learning model, and update the lag time of furnace temperature response of different working conditions; The above steps are executed cyclically to monitor the operation status of the furnace in real time and adjust the fuel supply and cooling water flow.

Images