Cement firing production real-time optimization method based on artificial intelligence

Abstract

The invention discloses a cement firing production real-time optimization method based on artificial intelligence, and belongs to the field of cement firing optimization, and the method comprises the steps: complementing missing data caused by maintenance or faults in a firing process through employing an artificial intelligence algorithm, and obtaining a complete and normalized time series data set; then, based on a variable coupling control matrix and a sliding time window mechanism, a nonlinear coupling relation among an operation variable, an information variable, a disturbance variable and an optimization target variable is converted into a trainable input-output structure, and an artificial intelligence prediction model is adopted to carry out minute-hour level advanced prediction on future working conditions; and finally, optimal equipment operation parameters which can be directly issued to a DCS (Distributed Control System) or a PLC (Programmable Logic Controller) are generated in real time through weighted limiting conditions and gradient descent optimization, so that closed-loop control of the firing system is realized. According to the method, the control precision and robustness of the cement sintering system under the working conditions of frequent fluctuation of raw materials, alternative fuel doping and large lag are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of cement firing optimization, and in particular relates to a real-time optimization method for cement firing production based on artificial intelligence. Background Technology

[0002] The clinker calcination process is a core component of cement production, and its operational quality directly determines clinker quality, energy consumption, and system stability. This process involves multiple thermal units, including preheaters, decomposers, rotary kilns, and grate coolers, requiring simultaneous real-time monitoring and control of hundreds of process parameters. In recent years, with the promotion of technologies for co-processing municipal solid waste, hazardous waste, biomass, and various alternative fuels in cement kilns, fluctuations in raw material composition and calorific value have become increasingly frequent. The coupling relationships between system variables exhibit highly nonlinear, strong coupling, and large lag characteristics, making traditional control methods insufficient to meet the requirements of refined and stable production.

[0003] Existing technologies mainly employ methods such as PID control, expert rule control, model predictive control (MPC), and advanced process control (APC). PID and DCS systems rely on manual experience to set parameters, resulting in lag in adjustments when faced with raw material fluctuations or operational disturbances. This leads to low control accuracy, poor robustness, and a long-term operational rate of less than 60%. Traditional APC systems are mostly based on MPC, requiring a large amount of stable historical data for offline modeling. When the source of raw materials such as limestone and siliceous clay changes or the proportion of alternative fuels changes, the original model quickly becomes invalid, causing the system to frequently switch to manual operation, making it difficult to sustain energy-saving and consumption-reducing effects. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a real-time optimization method for cement firing production based on artificial intelligence, comprising:

[0005] The raw operational data collected during the cement firing process are preprocessed to generate a complete and normalized time-series dataset.

[0006] Based on the time series dataset, the coupling relationship between the operating variables, information variables, disturbance variables and optimization target variables in the firing system is determined in order to construct a variable coupling control matrix;

[0007] Based on the aforementioned variable coupling control matrix, an artificial intelligence prediction model is used to predict the operating state of the calcination system in future time steps, so as to output the predicted values ​​of the operating condition indicators.

[0008] Based on the predicted values ​​of the operating condition indicators and the preset constraints, a weighted constraint condition and a cost function are constructed.

[0009] The optimal equipment operating parameters for the calcination system are generated by performing gradient descent optimization on the cost function.

[0010] Optionally, the preprocessing includes:

[0011] Macro-state markers are set based on the normal operation dataset to remove data from periods of kiln shutdown;

[0012] Multiple sets of variables associated with the missing time periods in the remaining data are constructed into a time series dataset, and the time series dataset is normalized to generate a normalized time series dataset.

[0013] For periods of missing data, artificial intelligence algorithms are used to fill in the missing data using complete sequences preceding the missing period, thus forming a complete time-series dataset.

[0014] Optionally, the construction of the variable coupling control matrix includes:

[0015] Based on thermal principles and firing experience, a variable coupling relationship diagram is constructed to preliminarily determine the relationship between the operating variables, information variables, disturbance variables and optimization target variables;

[0016] A sliding time window is introduced to perform time-series feature matching on the operational variables, information variables, and disturbance variables to form a training data set with time-series features;

[0017] Based on correlation matrix analysis, variables with an absolute correlation coefficient greater than 0.8 or a significant response to the operated variables are selected from the training data set to determine the final input and output variable sets of the artificial intelligence prediction model.

[0018] Optionally, the training of the artificial intelligence prediction model includes:

[0019] The complete time-series dataset is divided into training set, validation set and test set according to the proportions;

[0020] Construct an artificial intelligence-based prediction model, taking the final input variable set as input and the predicted value of the working condition index as output;

[0021] The prediction model is trained using the mean squared error loss function, and the model weights are updated through backpropagation to obtain the optimal set of prediction parameters for the working conditions.

[0022] Optionally, the construction of the weighted constraints and cost function includes:

[0023] The current state value of each constraint is compared with its upper and lower limits to calculate the weighted constraints;

[0024] The weighted constraints are mapped to the penalty function curve to form a set of penalty scores;

[0025] The sub-items in the penalty score set are summed to construct the cost function.

[0026] Optionally, the gradient descent optimization includes:

[0027] Calculate the gradient of the cost function with respect to the features of each sub-item;

[0028] The optimization coefficients are iteratively updated based on the learning rate and gradient value until the gradient norm threshold or the maximum number of iterations is met, in order to determine the optimal device operating parameters.

[0029] Optionally, the optimal equipment operating parameters are transmitted to the equipment controller of the firing system via the Industrial Internet or wireless network to achieve real-time optimized control of the firing system.

[0030] On the other hand, the present invention also provides an electronic device including a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein the processor implements the method when executing the computing program.

[0031] On the other hand, the present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method.

[0032] Compared with the prior art, the present invention has the following advantages and technical effects:

[0033] This invention utilizes artificial intelligence to adaptively complete missing data, dynamically model variable coupling relationships, and predict large lag conditions. Even with frequent fluctuations in raw materials, the system can still generate optimal equipment operating parameters online and execute them in a closed loop, thereby stabilizing clinker quality, reducing coal and electricity consumption simultaneously, and maintaining a long-term operating rate of over 90%. Attached Figure Description

[0034] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0035] Figure 1 This is a schematic diagram of the penalty function curve in an embodiment of the present invention;

[0036] Figure 2 This is a structural diagram of an artificial intelligence-based real-time optimization system for cement production, according to an embodiment of the present invention. Detailed Implementation

[0037] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0038] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0039] Example 1

[0040] like Figure 1-2 As shown, this embodiment provides a real-time optimization method for cement firing production based on artificial intelligence, including:

[0041] The raw operational data collected during the cement firing process are preprocessed to generate a complete and normalized time-series dataset.

[0042] Based on the time series dataset, the coupling relationship between the operating variables, information variables, disturbance variables and optimization target variables in the firing system is determined in order to construct a variable coupling control matrix;

[0043] Based on the aforementioned variable coupling control matrix, an artificial intelligence prediction model is used to predict the operating state of the calcination system in future time steps, so as to output the predicted values ​​of the operating condition indicators.

[0044] Based on the predicted values ​​of the operating condition indicators and the preset constraints, a weighted constraint condition and a cost function are constructed.

[0045] The optimal equipment operating parameters for the calcination system are generated by performing gradient descent optimization on the cost function.

[0046] Step 1: Determine the Macrostate based on the factory's normal operation dataset. The Macrostate contains several key operational parameters, some of which are shown below:

[0047]

[0048]

[0049]

[0050]

[0051] Step 2: Use Macrostates to remove entire data segments during kiln shutdown. The logical relationship is as follows:

[0052]

[0053] Step 3: Assuming the Macrostate calculation result is true, select multiple sets of variables related to the missing data, such as kiln current, feed rate, hot raw material temperature, and head coal rotor scale feedback, under normal operating conditions to construct a time-series dataset. The dataset is determined based on the actual working conditions and operating status. The number of elements is normalized using the following formula:

[0054] ;

[0055] in For the i-th time series data The normalized value of the j-th parameter. For the i-th time series data The j-th parameter. After computation, a new time-series dataset is generated: This serves as a normalized dataset related to the composition of gas in the smoke chamber under normal operating conditions.

[0056] Step 4: Define the time window for data loss due to unforeseen circumstances as (the time window is missing in the original text). The complete sequence of the same length preceding the missing part Let's assume... time, The time continues in this manner until... The weight matrices at time points are as follows: , ,… The weight matrix at time step is a randomly initialized matrix. Through... At each sampling time starting from time 1, artificial intelligence algorithms are continuously used to obtain the last valid hidden state that corresponds one-to-one with the missing part. .

[0057] Step 5: The following formula can be used to... Configure the relevant algorithm parameters and perform predictions within the time window:

[0058] ;

[0059] get Time weight matrix , For carrying The hidden state of compressed information in the historical sequence can reflect parameters such as the thermal inertia of the kiln system. The bias vector under actual working conditions. Input the normalized dataset for the relevant data at the current time step. This is the Sigmoid activation function.

[0060] Step 6: Using the formula:

[0061] ;

[0062] accomplish Input information for time-related data normalized dataset.

[0063] Step 7: Using the formula:

[0064] ;

[0065] accomplish Filtering information from time-related data normalized datasets.

[0066] Step 8: Using the formula:

[0067] ;

[0068] Enable information updates.

[0069] Step 9: Using the formula:

[0070] ;

[0071] ;

[0072] Obtain the first set of predicted data within the relevant data missing time window. By using the data missing time window By reusing the aforementioned artificial intelligence algorithm and processing it using the above normalization formula, a normalized dataset containing predicted data for time series can be generated. The normalized dataset provides a data foundation for training subsequent real-time optimization methods for artificial intelligence.

[0073] The real-time optimization method based on artificial intelligence includes the following steps:

[0074] Step 1: Determine the coupling relationship between variables;

[0075] (1) Using the variable coupling relationship diagram summarized based on thermal principles and firing professional experience, thermal theoretical calculations and preliminary variable relationship coupling were performed to determine the initial set of operating variables. Information variable group I = Perturbation variable group and optimization of the target variable set The interrelationships between these variables form a variable coupling control matrix, which initially determines the relationship between the optimization target variable and the operational variables, information variables, and disturbance variables.

[0076] (2) Since cement burning is a system with a large time lag, a sliding time window is introduced for the initial operating variables, information variables, and disturbance variables determined in (1). This enables the timing feature matching of each data set, forming a training data set with characteristics of the firing system, targeting different process control parameter time windows. The length is given by a combination of thermal theory calculations and process experience.

[0077] (3) Using the correlation matrix method, analyze the set of operational variables obtained in the second step. Information variable group I = Perturbation variable group and optimization of the target variable set Perform correlation analysis, selecting variables with an absolute correlation coefficient greater than 0.8 or those showing a significant response to the manipulated variable in the scatter plot, to obtain the optimal transformation parameters and sliding time window. This forms the input variable set for the final artificial intelligence prediction model. and output variables .

[0078] Step 2: Training the artificial intelligence model;

[0079] (1) Constructing the dataset: The normalized training dataset obtained above using artificial intelligence-based data processing methods is used to construct the dataset. The total number of sets is divided into training set, validation set and test set in a ratio of 6:2:2.

[0080] (2) Build an artificial intelligence-based prediction model for the coupling relationship between working conditions and variables in the cement production process, determine the input and output of the optimization variable model, and introduce a weighted set of constraints and a set of cost functions according to actual working conditions.

[0081] (3) Prepare training datasets for the above artificial intelligence prediction models under different working conditions based on the working conditions of the production site.

[0082] (4) Use the above-described training set, validation set and test set to train, optimize and evaluate the artificial intelligence prediction model.

[0083] (5) The weight matrix is ​​initialized using the Kaiming method, the matrix variance is adjusted, and the model convergence speed and training stability are improved.

[0084] (6) The hidden layer uses the ReLU activation function to enhance the model's nonlinear expressive power. This is achieved through the formula:

[0085] ;

[0086] ;

[0087] ;

[0088]

[0089] ;

[0090] in As input to the normalized optimization model, This is the weight matrix. For bias vectors, This is the output of the i-th layer.

[0091] (7) Calculate the input and output of each hidden layer. Map the output to a single operating condition index prediction value through a fully connected layer.

[0092] (8) Using the mean square error (MSE), the formula is:

[0093] ;

[0094] Calculate the loss function and perform backpropagation to update the weight parameters. Select an appropriate number of training epochs to train the operating condition index prediction model, and evaluate the model performance using test set data. Finally, obtain the optimal set of operating condition prediction parameters and the corresponding equipment prediction parameters.

[0095] Step 3: Weighted constraint conditions, cost function construction, and equipment parameter optimization;

[0096] (1) Compare the predicted parameters generated in the above steps with the corresponding constraints, and construct the weighted constraints and cost function using the following formula:

[0097] ;

[0098] ;

[0099] in The weighted constraint is formed by weighting the i-th constraint. The current state value under this constraint. This is the lower limit of the constraint. This is the upper limit of the constraint. These are the weighting coefficients.

[0100] (2) Map the calculation results of all constraints to, as shown in the figure Figure 1 The penalty function curve shown provides a set of penalty scores with a uniform standard:

[0101] ;

[0102] The magnitude of the penalty score reflects the impact of the constraint on the optimization result.

[0103] (3) The cost function CF of the optimization logic can be obtained by summing all the sub-items in the penalty score set:

[0104] ;

[0105] (4) Optimize the gradient descent parameters of the cost function CF using the gradient descent method;

[0106] ;

[0107] Where m is the number of possible cost function result samples. These are the characteristics of each sub-item of the cost function.

[0108] (5) Update coefficients through iteration:

[0109] ;

[0110] in For learning rate, This represents the partial derivative of the loss function with respect to the coefficients. It is calculated iteratively and using the gradient norm threshold. The maximum number of iterations is limited, and the optimal solution and corresponding equipment operating parameters are sought within the feasible range.

[0111] In addition to the methods described above, this invention also discloses a real-time optimization control hardware system based on artificial intelligence, including a real-time optimization server running on an industrial server, intelligent sensors, an analyzer, a data acquisition module, a data processing module, a variable relationship coupling module, an AI real-time optimization module, and an optimization parameter output module. Through the Industrial Internet, it connects bidirectionally with the DCS control system using the OPC communication protocol or uploads on-site production operation data through a wireless network and the HTTP protocol.

[0112] Intelligent temperature and pressure sensors are added to key components of the cement calcination system. Sensor data, raw material fluorescence analysis results, clinker detection results, fuel detection results, and results from alternative fuels and solid / hazardous waste detection are uploaded in real-time to a real-time database within an AI-based real-time optimization system, serving as the data source for optimization algorithms. By utilizing the real-time data in the database, an effective real-time production optimization control strategy and optimal equipment operating parameters can be formed. These optimized parameters are then fed back to the relevant equipment controllers via an optimization parameter output module, achieving intelligent production control and optimization. The overall system structure is as follows: Figure 2 As shown.

[0113] The data acquired by the data acquisition module includes supplementary data entered manually, data from the system's built-in smart sensors, existing sensors, online analyzers, equipment controllers, etc., collected through the Industrial Internet and OPC services. Among them, the data collected from sensors, online analyzers, and equipment controllers are various process and equipment operating parameters and experimental data on site, which together constitute the required source data.

[0114] The data processing module is connected to the data acquisition module. It uses the artificial intelligence data processing method mentioned above to correct the lack of real-time data on operating conditions and equipment caused by faults and other reasons, and stores the data that can be used for training, prediction and optimization in the form of a normalized dataset into the real-time database.

[0115] The variable coupling module is connected to the real-time database. It clarifies the interaction between variables through variable coupling control matrix and correlation analysis based on thermal principles and firing experience, and provides usable training sets, validation sets and test sets for the AI ​​real-time optimization module.

[0116] The real-time AI optimization module is connected to the variable relationship coupling module. Its main function is to read various coupling parameters input in real time through artificial intelligence machine learning algorithms and comprehensively consider various constraints to obtain a prediction of the operating status of the firing system in the next time step. The algorithm mentioned above is used to optimize the parameters for the operation of the process equipment.

[0117] The optimized parameter output module is connected to the AI ​​real-time optimization module and the human-machine interaction device. Its main functions are to send back the equipment operation parameters generated by the AI ​​real-time optimization module to the industrial internet and to receive the constraint information from the human-machine interaction device.

[0118] The human-computer interaction device is connected to the AI ​​real-time optimization module and the optimization parameter output module. Its main functions are: to activate and deactivate the real-time optimization system, to display the current predicted values, to display the optimization results of the equipment operating parameters, and to receive the constraint parameter settings from the operator. Tables 1, 2, and 3 are point tables for the operational variables, objective variables, and constraint conditions.

[0119] Table 1

[0120] Serial number Type RTO Point Description System Point 20100 OP RTO_KilnFeed.Rec Kiln feed RTO SP RTO_KilnFeed_Rec.VALUE 20200 OP RTO_IDFanSpeed.Rec ID fan speed RTO SP RTO_IDFanSpeed_Rec.VALUE 20300 OP RTO_CalcinerExitTemp.Rec Calciner exit temperature RTO SP RTO_CalcinerExitTemp_Rec.VALUE 20400 OP RTO_KilnSpeed.Rec Kiln speed RTO SP RTO_KilnSpeed_Rec.VALUE 20500 OP RTO_KilnCoal.Rec Kiln coal RTO SP RTO_KilnCoal_Rec.VALUE 20600 OP RTO_BypassExhaustFanSpeed.Rec Bypass exhaust fan speed RTO SP RTO_BypassExhaustFSp_Rec.VALUE 20700 OP RTO_CoolerGrate1Speed.Rec Cooler grate 1 speed RTO SP RTO_CoolerGrate1SpeedRec.VALUE 20900 OP RTO_CoolerFan1Speed.Rec Cooler Fan 1 speed RTO SP RTO_CoolerFan1Speed_Rec.VALUE 21000 OP RTO_CoolerFan1NewSpeed.Rec Cooler Fan 1 (New) speed RTO SP RTO_CoolerFan1NewSp_Rec.VALUE 21100 OP RTO_CoolerFan2Speed.Rec Cooler Fan 2 speed RTO SP RTO_CoolerFan2Speed_Rec.VALUE 21200 OP RTO_CoolerFan3Speed.Rec Cooler Fan 3 speed RTO SP RTO_CoolerFan3Speed_Rec.VALUE 21300 OP RTO_CoolerFan4Speed.Rec Cooler Fan 4 speed RTO SP RTO_CoolerFan4Speed_Rec.VALUE 21400 OP RTO_CoolerFan5Speed.Rec Cooler Fan 5 speed RTO SP RTO_CoolerFan5Speed_Rec.VALUE 21500 OP RTO_CoolerFan6Speed.Rec Cooler Fan 6 speed RTO SP RTO_CoolerFan6Speed_Rec.VALUE 21600 OP RTO_CoolerFan7Speed.Rec Cooler Fan 6 speed RTO SP RTO_CoolerFan7Speed_Rec.VALUE

[0121] Table 2

[0122] Serial number Type Description Point Unit Maximum value Minimum value 15 INF Cooler_grate_section_2_speed_SP PN_471LQ02LP01_SV.VALUE 1 / min 0 10 16 INF Cooler_grate_section_2_speed_PV PN_471LQ02LP01_AI6.PV 1 / min 0 10 51 OBJ Q_Clinker_Lab_f_CaO MES_ZY1_FC.VALUE % 53 OBJ T_Kiln_hood PA_471CN1TT02.PV °C 1180 1250 54 OBJ T_clinker_cooler_outlet 12TT02B.PV °C 0 150 62 OBJ P_Cooler_Fan_G12_2_Outlet_Pa PA_471FN02PT01.PV Pa 0 9800 63 OBJ Q_Clinker_Lab_Cl QCX_ZY1_CL.VALUE % 103 OBJ I_Kiln1_drive_A AI461MD02II01.PV A 104 OBJ [[ID=-44]]I_Kiln2_drive_A AI461MD01II01.PV A 112 OBJ Calciner_coal_SP_tph Z_PN_761RS01KP_AO6.VALUE tph 113 OBJ ​ ​ ​ 114 ​ ​ ​ ​ -11000 1000 115 ​ ​ ​ °C 0 1500 116 ​ ​ ​ ​ -11000 1000 117 ​ ​ ​ ​ -11000 1000 118 ​ ​ ​ °C 0 1500 119 ​ ​ ​ ​ -11000 1000 120 ​ ​ ​ °C 0 1500 121 ​ ​ ​ ​ -7000 1000 122 ​ ​ ​ °C 0 1500

[0123] Table 3

[0124] ​ ​ ​ ​ ​ 55 ​ CO_InletChamberppm AI451PR1AT01_CO.PV ppm 56 CONST O2_InletChamber% AI451PR1AT01_O2.PV % 57 CONST NOx_InletChamber_mg_m3 451PR1AT01_NOx.VALUE mg / m³ 58 CONST CO_behindCAL% AI431RS01FT02_CO.PV % 59 CONST O2_behindCAL% A431RS01FT02_O2.PV % 60 CONST O2_behindPH% AI431RS01O201.PV % 105 CONST T_C1A1Outlet PA_441CN7TT01.PV °C 106 CONST T_C1B1Outlet PA_441CN1TT03.PV °C 107 CONST CO_behindPH% AI431RS01CO01.PV % 108 CONST NOx_behindPH_mg_m3 431RS01_NOx.VALUE mg / m³ 109 CONST NOx_behindCAL_mg_m3 431RS01FT02_NOx.VALUE mg / m³ 110 CONST O2_stack% PN_TXXT_AI005ZS.VALUE % 111 CONST NOx_stack_mg_m3_10phO2 PN_TXXT_AI003.PV mg / m³

[0125] On the other hand, this embodiment also provides an electronic device, including a memory, a processor, and a computing program stored in the memory and executable on the processor, wherein the processor implements the method when executing the computing program.

[0126] On the other hand, this embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method.

[0127] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A real-time optimization method for cement firing production based on artificial intelligence, characterized in that, include: The raw operational data collected during the cement firing process are preprocessed to generate a complete and normalized time-series dataset. Based on the time series dataset, the coupling relationship between the operating variables, information variables, disturbance variables and optimization target variables in the firing system is determined in order to construct a variable coupling control matrix; Based on the aforementioned variable coupling control matrix, an artificial intelligence prediction model is used to predict the operating state of the calcination system in future time steps, so as to output the predicted values ​​of the operating condition indicators. Based on the predicted values ​​of the operating condition indicators and the preset constraints, a weighted constraint condition and a cost function are constructed. The optimal equipment operating parameters for the calcination system are generated by performing gradient descent optimization on the cost function.

2. The method according to claim 1, characterized in that, The preprocessing includes: Macro-state markers are set based on the normal operation dataset to remove data from periods of kiln shutdown; Multiple sets of variables associated with the missing time periods in the remaining data are constructed into a time series dataset, and the time series dataset is normalized to generate a normalized time series dataset. For periods of missing data, artificial intelligence algorithms are used to fill in the missing data using complete sequences preceding the missing period, thus forming a complete time-series dataset.

3. The method according to claim 1, characterized in that, The construction of the variable coupling control matrix includes: Based on thermal principles and firing experience, a variable coupling relationship diagram is constructed to preliminarily determine the relationship between the operating variables, information variables, disturbance variables and optimization target variables; A sliding time window is introduced to perform time-series feature matching on the operational variables, information variables, and disturbance variables to form a training data set with time-series features; Based on correlation matrix analysis, variables with an absolute correlation coefficient greater than 0.8 or a significant response to the operated variables are selected from the training data set to determine the final input and output variable sets of the artificial intelligence prediction model.

4. The method according to claim 1, characterized in that, The training of the artificial intelligence prediction model includes: The complete time-series dataset is divided into training set, validation set and test set according to the proportions; Construct an artificial intelligence-based prediction model, taking the final input variable set as input and the predicted value of the working condition index as output; The prediction model is trained using the mean squared error loss function, and the model weights are updated through backpropagation to obtain the optimal set of prediction parameters for the working conditions.

5. The method according to claim 1, characterized in that, The construction of the weighted constraints and cost function includes: The current state value of each constraint is compared with its upper and lower limits to calculate the weighted constraints; The weighted constraints are mapped to the penalty function curve to form a set of penalty scores; The sub-items in the penalty score set are summed to construct the cost function.

6. The method according to claim 1, characterized in that, The gradient descent optimization includes: Calculate the gradient of the cost function with respect to the features of each sub-item; The optimization coefficients are iteratively updated based on the learning rate and gradient value until the gradient norm threshold or the maximum number of iterations is met, in order to determine the optimal device operating parameters.

7. The method according to claim 1, characterized in that, The optimal equipment operating parameters are transmitted to the equipment controller of the firing system via the Industrial Internet or wireless network to achieve real-time optimized control of the firing system.

8. An electronic device comprising a memory, a processor, and a computing program stored in the memory and executable on the processor, characterized in that, When the processor executes the computing program, it implements the method of any one of claims 1-7.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1-7.

Images