A coke oven header pressure control method based on multivariable constraint optimization

Abstract

This invention relates to the field of coking industry technology, and particularly to a method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization. The method includes collecting, filtering, and preprocessing historical operating data of the coke oven gas collecting pipe system under different operating conditions; designing a multivariable nonlinear constraint optimization control algorithm; introducing constraints reflecting the actual operation of the coke oven system into the nonlinear constraint optimization problem, and adding penalty terms for control input constraints and gas collecting pipe pressure constraints to the optimization objective function; and at the beginning of each control cycle, collecting key parameters of the coke oven system in real time, and predicting and calculating these key parameters using a NARX neural network model. The advantages of this invention are: it ensures that the generated dataset comprehensively covers the state of the coke oven gas collecting pipe system in various actual operating scenarios, enabling the subsequently trained NARX neural network model to fully learn the dynamic behavior patterns of the system under different operating conditions.

Description

Technical Field

[0001] This invention relates to the field of coking industry technology, and in particular to a method for controlling the pressure of coke oven gas collecting pipes based on multivariate constraint optimization. Background Technology

[0002] In the coking industry, the coke oven is the core equipment for producing coke, and the pressure control of its gas collecting system is crucial for improving production efficiency and ensuring stable equipment operation. During operation, the coke oven produces large amounts of gases containing CO2, SO2, and NO. x Coal gas contains harmful components. Industrially, coke oven gas collection systems are typically used to recover and treat this gas, improving energy efficiency, reducing direct emissions of harmful gases, and contributing to environmental protection and resource recycling.

[0003] With the advancement of the national energy conservation and emission reduction strategy, the coking industry is committed to optimizing control systems and improving the accuracy of coke oven gas collecting pipe pressure regulation to meet the requirements of energy conservation and efficiency improvement. The stability of gas collecting pipe pressure is crucial for improving coking process quality, ensuring safe production, and protecting the environment. However, in practical applications, gas collecting pipe systems face many complex problems, such as numerous variables, strong coupling, significant nonlinearity, large external disturbances, and dynamic time-varying characteristics. Inconsistent operating states between the coke oven and the carbonization chamber can lead to uneven pressure distribution; the carbonization chamber experiences severe pressure disturbances during critical operations such as coke pushing, coal charging, and reversing heating; and unreasonable pressure setpoints can cause serious environmental pollution, energy waste, and safety hazards.

[0004] Currently, single-variable PID control strategies are commonly used for complex pressure systems in industrial settings, but this method has significant limitations. On one hand, the complexity and nonlinear characteristics of coke oven systems make PID control poorly adaptable under multi-loop coupling conditions, making it difficult to meet the demands of multi-variable dynamic adjustment. On the other hand, existing equipment designs generally reserve large redundancy spaces to ensure safety, resulting in low equipment utilization and significant resource waste. Furthermore, the diversity in structure and topology among different coke ovens and carbonization chambers makes it difficult to construct a unified overall model, further limiting the optimization capabilities of the control system. Summary of the Invention

[0005] The purpose of this invention is to provide a coke oven gas collecting pipe pressure control method based on multivariable constraint optimization, which can significantly improve control accuracy, system response speed and stability. By applying data-driven neural network modeling and multivariable nonlinear predictive control algorithm, high-precision dynamic regulation of coke oven gas collecting pipe pressure is achieved. At the same time, the gas collecting pipe pressure control process is optimized, thereby reducing energy consumption, improving coke quality, and ultimately ensuring the economic efficiency and safety of the coke oven gas collecting pipe system.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization includes:

[0008] S1. Collect, filter and preprocess historical operating data of the coke oven gas collecting pipe system under different operating conditions, generate a dataset for training, validating and testing the nonlinear autoregressive NARX neural network model with external input, train and validate the nonlinear autoregressive NARX neural network model with external input, and obtain the multivariate coupling characteristics describing the coke oven gas collecting pipe system.

[0009] S2. Based on the NARX neural network model, design a multivariable nonlinear constraint optimization control algorithm;

[0010] S3. In the nonlinear constraint optimization problem, constraints reflecting the actual operation of the coke oven system are introduced, including the upper / lower limit of the coke oven gas collecting pipe pressure, the safe operating range of valve opening and blower speed, and energy consumption limits. Penalty terms for control input constraints and gas collecting pipe pressure constraints are added to the optimization objective function to punish violations.

[0011] S4. At the beginning of each control cycle, key parameters of the coke oven system are collected in real time, and the key parameters of the coke oven system are predicted and calculated using the NARX neural network model to obtain the predicted value of the coke oven gas collecting pipe system state in the future prediction time domain. The control parameters are adjusted according to the optimization results of the constraint optimization control output. The first value of the optimal control input increment sequence is applied to the control input to update the control input value and complete the control of the gas collecting pipe pressure.

[0012] In S1, preprocessing includes merging data with aligned time by taking the intersection of the time axes, and mapping the data to the normalized interval using the maximum and minimum value normalization method to eliminate the impact of differences in the magnitude of different variables on the training and prediction of the nonlinear autoregressive NARX neural network model with external input.

[0013] The screening adopts a dynamic segmentation method, which is based on the identification results of the coke oven operating conditions, and divides the coke oven gas collection pipe system operating data into multiple subsets.

[0014] The dataset is obtained by merging data with aligned time by taking the intersection of the time axes and performing preprocessing such as normalization of maximum and minimum values.

[0015] In S2, the multivariable nonlinear constraint optimization control algorithm is as follows:

[0016] Within each control cycle, a nonlinear constrained optimization problem is constructed to predict the dynamic behavior of the coke oven gas collecting pipe system at multiple future moments. The objective function is to minimize the sum of squares of the tracking error between the gas collecting pipe pressure and the setpoint in the future prediction time domain, the weighted sum of the sum of squares of the incremental changes in control input, and includes constraints on the nonlinear system dynamic equations based on a nonlinear autoregressive NARX neural network model with external input. The control variables are dynamically adjusted through rolling optimization based on sequential quadratic programming, so that the coke oven gas collecting pipe system can follow the gas collecting pipe pressure setpoint, while suppressing pressure fluctuations in the carbonization chamber caused by different operating conditions.

[0017] The multivariable nonlinear constraint optimization control algorithm combines the dynamic changes of coke oven operating conditions with the optimization of control variables. The dynamic changes are predicted by a nonlinear autoregressive NARX neural network model with external input.

[0018] In S3, the upper / lower limits of the coke oven gas collecting pipe pressure are constrained. The constraint range is set according to the coke oven safety production requirements and process requirements. Valve opening and blower speed are introduced into safety operation constraints.

[0019] Adding penalty terms for control input constraints and gas collection pipe pressure constraints to the optimization objective function is achieved by introducing relaxation variables to allow small-scale violations of hard constraints, and then applying penalty terms to the relaxation variables in the optimization objective function.

[0020] In S4, the real-time collected gas collecting pipe pressure is fed back to the control system and compared with the set value of the gas collecting pipe pressure. The control parameters are adjusted by the optimization result output by the multivariable nonlinear constraint optimization control algorithm to form a feedback loop.

[0021] The nonlinear autoregressive NARX neural network model with exogenous input is obtained through offline training. The inputs to the nonlinear autoregressive NARX neural network model include current and historical key parameters, and the output of the nonlinear autoregressive NARX neural network with exogenous input is the predicted value of the gas collection tube pressure in multiple future prediction time domains.

[0022] The rolling optimization method of sequential quadratic programming transforms the nonlinear constraint optimization problem into multiple quadratic programming subproblems for solving;

[0023] The quadratic programming subproblem is solved using a quadratic programming solver. The Gauss-Newton method is used to approximate the Hessian matrix when solving the quadratic programming subproblem.

[0024] The optimization objective function includes minimizing the sum of squares of the tracking error between the gas collecting pipe pressure and the set value in the future prediction time domain, the weighted sum of the sum of squares of the incremental changes in control input, and also includes a constraint term on the control input. The constraint term is dynamically adjusted according to the coke oven gas production and pressure fluctuation limit requirements to ensure the energy optimization operation of the coke oven system under the premise of stable gas collecting pipe pressure.

[0025] Adjusting control parameters based on the optimization results of the constraint optimization control output includes applying the first value of the optimal control input increment sequence to the control input and updating the control input value through incremental increments.

[0026] Compared with the prior art, the beneficial effects of the present invention are:

[0027] 1. It ensures that the generated dataset comprehensively covers the state of the coke oven gas collection pipe system in various actual operating scenarios. This allows the subsequently trained NARX neural network model to fully learn the dynamic behavior patterns of the system under different operating conditions, avoiding inaccurate or unpredictable predictions in certain operating conditions due to missing data. The screening process removes abnormal, erroneous, and irrelevant data, retaining the data that is valuable for model training. Preprocessing (such as data alignment and normalization) eliminates the adverse effects of differences in the units of different variables on model training, making the data more standardized and orderly, meeting the requirements of neural network models for input data, thereby improving the efficiency and accuracy of model training and laying a good foundation for subsequent model training, validation, and testing.

[0028] 2. The coke oven gas collection pipe system has strong multivariate coupling and nonlinear dynamic characteristics. Traditional linear models are difficult to accurately describe its complex behavior. As a nonlinear model, the NARX neural network can effectively capture and learn the complex coupling relationships between multiple variables in the system through its nonlinear mapping capability. After training and validation, the NARX neural network model can accurately describe the dynamic characteristics of the coke oven gas collection pipe system under different operating conditions, providing high-precision system model support for subsequent pressure control. By training and validating on datasets containing data from different operating conditions, the NARX neural network model can learn the general laws of the system, rather than being limited to the behavior patterns under a specific operating condition. When the model is applied to the actual control of the coke oven gas collection pipe system, it has good adaptability and generalization ability to unseen operating conditions or changes in system state, ensuring the reliability of prediction and control effects under various actual operating conditions.

[0029] 3. Since the NARX neural network model can accurately describe the multivariable coupling characteristics of the coke oven gas collecting pipe system, the multivariable nonlinear constraint optimization control algorithm designed based on this model can fully consider the mutual influence and coupling relationship between the variables in the system. This enables the control algorithm to more accurately coordinate and optimize the control of multiple control variables (such as valve opening, blower speed, etc.), avoiding the system imbalance or poor control effect caused by single variable control, thereby improving the performance and stability of the gas collecting pipe pressure control. The operation of the coke oven gas collecting pipe system is limited by a variety of constraints, including upper and lower pressure limits, valve opening range, blower speed range, and energy consumption limits. The optimization control algorithm based on the NARX neural network model can incorporate these complex constraints into the control strategy. By constructing and solving a nonlinear constraint optimization problem, it ensures that the control results meet both the requirements for stable system operation and the actual constraints of industrial production, achieving comprehensive optimization of the system in terms of safety, economy, and efficiency.

[0030] 4. By introducing constraints such as the upper and lower limits of coke oven gas collecting pipe pressure, valve opening, and safe operating range of blower speed, the control variables and system state can be strictly limited to changes within a safe range. This prevents equipment damage, production accidents, or environmental pollution caused by excessively high or low pressure, excessively large or small valve opening, or abnormal blower speed, ensuring the safe and stable operation of the coke oven system. Considering energy consumption constraints and adding penalty terms for control input constraints and gas collecting pipe pressure constraints to the optimization objective function, the optimization algorithm can minimize energy waste and unnecessary consumption while meeting the system pressure control requirements. At the same time, the penalty term can improve the robustness of the optimization problem to measurement noise and model errors. It allows for minor violations of constraints in certain situations, but by penalizing violations, the system can be optimized as much as possible within a safe and economical range, reducing production costs and improving the economic benefits of the enterprise.

[0031] 5. Real-time acquisition of key parameters of the coke oven system and input into the NARX neural network model for prediction allows for the prediction of the coke oven gas collecting pipe system's state over a certain time range (prediction time domain). This enables the control system to understand potential development trends and changes in the system, such as pressure rise or fall trends and possible pressure fluctuations, thus providing a basis for taking corresponding control measures in advance and enhancing the control system's foresight and initiative. Since the operating conditions of the coke oven system are dynamically changing, real-time prediction can promptly reflect the latest changes in the system's state. The control system can quickly adjust its control strategy based on these real-time prediction results, enabling control parameters to adapt to changes in operating conditions in a timely manner. This improves the system's response speed and adaptability to dynamic disturbances (such as pressure fluctuations caused by coal charging and coke pushing operations), ensuring that the coke oven gas collecting pipe pressure remains stable within the set range and maintaining the normal operation of the coking process.

[0032] 6. By promptly applying the first value of the optimal control input increment sequence from the constraint optimization control output to the control input, the optimization result can be quickly transformed into actual control actions, such as adjusting valve opening or changing blower speed. This timely update of control parameters ensures that the system can control according to the current optimal solution in each control cycle, avoiding control lag and improving the real-time performance and effectiveness of the gas collecting pipe pressure control. By continuously updating the control input value, the control system can continuously fine-tune and optimize the coke oven gas collecting pipe pressure, keeping the system in its optimal operating state. At the same time, since the optimization algorithm is based on system constraints and dynamic characteristics, each adjustment of control parameters is within a safe and reasonable range, thus ensuring the stability and continuity of the control process and avoiding system oscillations or instability caused by excessive or inappropriate adjustment of control parameters. This ensures precise control of the coke oven gas collecting pipe pressure and long-term stable operation of the system. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of the coke oven gas collecting pipe pressure control system based on multivariable constraint optimization.

[0034] Figure 2 This is a modeling effect diagram of the neural network model established in the embodiment.

[0035] Figure 3 This is a schematic diagram of the constraint optimization control structure based on neural networks established in this invention.

[0036] Figure 4 This is a diagram illustrating the effect of pressure control in the gas collecting pipe in the embodiment. Detailed Implementation

[0037] The present invention will now be described in detail with reference to the accompanying drawings, but it should be noted that the implementation of the present invention is not limited to the following embodiments.

[0038] The following embodiments are implemented based on the technical solution of the present invention, providing detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments. Unless otherwise specified, the methods used in the following embodiments are conventional methods.

[0039] Example 1

[0040] A method for controlling the pressure of coke oven gas collecting pipes based on multivariable constraint optimization, the specific contents of which include:

[0041] S1. Collect, filter and preprocess historical operating data of the coke oven gas collecting pipe system under different operating conditions, generate a dataset for training, validating and testing the nonlinear autoregressive NARX neural network model with external input, train and validate the nonlinear autoregressive NARX neural network model with external input, and obtain the multivariate coupling characteristics describing the coke oven gas collecting pipe system.

[0042] Preprocessing includes merging data to align time by taking the intersection of time axes, and mapping the data to a normalized interval using the maximum-minimum normalization method to eliminate the impact of differences in the magnitude of different variables on the training and prediction of the nonlinear autoregressive NARX neural network model with external input.

[0043] The screening adopts a dynamic segmentation method. The dynamic segmentation is based on the identification results of the coke oven operating conditions. The coke oven gas collection pipe system operation data is divided into multiple subsets. The coke oven operating conditions include, but are not limited to, coal charging, coke pushing and reversing conditions, so as to more accurately capture the dynamic characteristics of the system under different operating conditions. The identification results are used to construct a representative dataset.

[0044] The dataset is obtained by merging data to align time by taking the intersection of time axes and preprocessing it by normalizing the maximum and minimum values. The dataset is used to train a NARX neural network model describing the multivariate strongly coupled and nonlinear autoregressive coke oven gas gathering pipe system with external input.

[0045] S2. Based on the NARX neural network model, design a multivariable nonlinear constraint optimization control algorithm, as follows:

[0046] Within each control cycle, a nonlinear constrained optimization problem is constructed to predict the dynamic behavior of the coke oven gas collecting pipe system at multiple future moments. The objective function is to minimize the sum of squares of the tracking error between the gas collecting pipe pressure and the setpoint in the future prediction time domain, the weighted sum of the sum of squares of the incremental changes in control inputs, and includes constraints on the nonlinear system dynamic equations based on a nonlinear autoregressive NARX neural network model with external inputs. Control variables such as valve opening and blower speed are dynamically adjusted through a rolling optimization method based on sequential quadratic programming, enabling the coke oven gas collecting pipe system to follow the gas collecting pipe pressure setpoint while suppressing pressure fluctuations in the carbonization chamber under different operating conditions and controlling the incremental changes in inputs such as valve opening and blower speed. The rolling optimization method of sequential quadratic programming transforms the nonlinear constrained optimization problem into multiple quadratic programming subproblems for efficient real-time optimization. The quadratic programming subproblems are solved using a quadratic programming solver. When solving the quadratic programming subproblems, the Gauss-Newton method is used to approximate the Hessian matrix to reduce the computational burden of each iteration and accelerate the optimization process.

[0047] The optimization objective function includes minimizing the sum of squares of the tracking error between the gas collecting pipe pressure and the set value in the future prediction time domain, the weighted sum of the sum of squares of the incremental changes in control input, and also includes a constraint term on the control input. The constraint term is dynamically adjusted according to the coke oven gas production and pressure fluctuation limit requirements to ensure the stable gas collecting pipe pressure, thereby achieving energy optimization operation of the coke oven system and reducing electricity consumption.

[0048] The nonlinear autoregressive NARX neural network model with external input is obtained through offline training. The inputs for the nonlinear autoregressive NARX neural network model include current and historical key parameters, including the displayed value of the gas collection pipe pressure, the optimized setpoint of the gas collection pipe regulating valve, the optimized setpoint of the blower inlet regulating valve, the optimized setpoint of the blower return valve, the optimized setpoint of the blower speed, the front / rear suction of the blower, and the high-pressure ammonia flow rate. The output of the nonlinear autoregressive NARX neural network with external input is the predicted value of the gas collection pipe pressure in multiple future prediction time domains.

[0049] The multivariable nonlinear constraint optimization control algorithm combines the dynamic changes of coke oven operating conditions with the optimization of control variables such as valve opening and blower speed. The dynamic changes are predicted by a nonlinear autoregressive NARX neural network model with external input. This not only effectively suppresses pressure fluctuations, but also reduces the system's energy consumption by optimizing control variables while ensuring pressure stability.

[0050] S3. In the nonlinear constraint optimization problem, constraints reflecting the actual operation of the coke oven system are introduced, including the upper / lower limits of the coke oven gas collecting pipe pressure, the safe operating range of valve opening and blower speed, and energy consumption limits. This ensures that the control results meet the safety and economic requirements of industrial operation. Penalty terms for control input constraints and gas collecting pipe pressure constraints are added to the optimization objective function. By introducing slack variables, small-scale violations of hard constraints are allowed, and penalty terms are applied to the slack variables in the optimization objective function to improve the solvability of the optimization problem and its robustness to measurement noise and model errors. The constraint range of the upper / lower limits of the coke oven gas collecting pipe pressure is set according to the safety and process requirements of the coke oven. For example, if the coke oven gas collecting pipe pressure is too low (negative pressure), air will enter the furnace, causing coke combustion and affecting coke quality; if the coke oven gas collecting pipe pressure is too high, it will lead to leakage of raw coal gas, causing environmental pollution and reducing the recovery rate of raw coal gas. Therefore, the upper / lower limits of the coke oven gas collecting pipe pressure should be within the safe range that meets the safety and process requirements of the coke oven. Safety operation constraints are introduced for the opening of valves such as the gas collecting pipe regulating valve, the blower inlet regulating valve, and the blower return valve, as well as the speed of the blower used to generate negative pressure to draw coal gas. For example, limits are set on the rate of change of control variables such as valve opening and blower speed, or minimum values ​​are set for operating parameters such as gas collecting pipe pressure and blower speed, to prevent instability caused by the coke oven gas collecting pipe system responding too quickly or too slowly, and to avoid a decrease in coal gas conveying efficiency caused by low-speed operation.

[0051] S4. At the beginning of each control cycle, key parameters such as the coke oven gas collecting pipe pressure of the coke oven system are collected in real time. The key parameters of the coke oven system are predicted and calculated using the NARX neural network model to obtain the predicted state value of the coke oven gas collecting pipe system in the future prediction time domain. Based on the optimization results of the constraint optimization control output, the control parameters such as the valve opening of the gas collecting pipe regulating valve, the blower inlet regulating valve, the blower return valve, etc., and the blower operating speed are adjusted. The first value of the optimal control input increment sequence is applied to the control input to update the control input value to adapt to changes in operating conditions, improve the robustness of the system to dynamic disturbances, and complete the control of the gas collecting pipe pressure. The real-time collected gas collecting pipe pressure is fed back to the control system and compared with the set value of the gas collecting pipe pressure. The control parameters are adjusted through the optimization results output by the multivariable nonlinear constraint optimization control algorithm to form a feedback loop, ensuring that the pressure regulation system can respond quickly to external disturbances, while improving the stability and robustness of the control. Adjusting control parameters based on the optimization results of the constraint optimization control output includes applying the first value of the optimal control input increment sequence to the control input and updating the control input value through incremental increments. The control input includes, for example, valve opening setpoints and booster fan operating speed setpoints.

[0052] Example 2

[0053] Taking the gas collecting pipe at a coal and coke chemical plant as an example, the pressure control method for coke oven gas collecting pipe based on multivariable constraint optimization provided by the present invention is described in detail. The specific steps are as follows:

[0054] Step A: Collect and preprocess historical operating data of the gas collection pipe system, and construct a nonlinear autoregressive neural network prediction model with external input;

[0055] See Figure 1 Long-term operational data was collected from the integrated measurement and control system of the coke oven system. In order to construct a dataset that can accurately capture the dynamic characteristics of the system, key process variables under typical operating conditions were selected from the data. Key parameters include, but are not limited to: gas collecting pipe pressure (e.g., the pressure PR of gas collecting pipe No. 1). 1a Pressure PR of No. 2 gas collecting pipe 1b Pressure PR of No. 3 gas collecting pipe 2a Pressure PR of No. 4 gas collecting pipe 2b ), Gas main pipe pressure PR before the primary cooler 41 The valve opening of the gas manifold regulating valve (e.g., the opening PVI of regulating valve No. 1) 1a No. 2 regulating valve opening PVI 1b No. 3 regulating valve opening PVI 2a No. 4 regulating valve opening PVI 2b ), the valve opening degree (PVI) of the blower inlet regulating valve 41 ), the valve opening degree of the blower return valve (PVI) 42 ), Blower speed (SI), Blower front / rear pressure (PR) 42 PR 43 ), high-pressure ammonia flow rate (e.g., high-pressure ammonia flow rate FR1 for No. 1 and high-pressure ammonia flow rate FR2 for No. 2).

[0056] The collected raw data undergoes multi-step preprocessing:

[0057] First, perform data cleaning, including removing outliers and filling in missing values;

[0058] Secondly, to address potential time discrepancies between different variables, a time axis intersection method is used to merge data with aligned times, ensuring the temporal synchronization of data from different variables and facilitating subsequent data analysis and modeling.

[0059] Furthermore, for different operating conditions such as coal loading, coke pushing, and reversing, a dynamic segmentation method is adopted to divide the system operation data into multiple subsets in order to more comprehensively cover and reflect the dynamic characteristics of the system under different operating conditions.

[0060] Finally, the processed data is standardized, with maximum and minimum value normalization being preferred. This maps the data to a unit interval, effectively eliminating the problem of imbalanced variable weights during the modeling process caused by differences in units, and improving the training efficiency and prediction accuracy of the model.

[0061] Finally, the preprocessed high-quality dataset is divided into training, validation, and test sets in appropriate proportions for training and evaluating nonlinear autoregressive neural network models with exogenous inputs.

[0062] A lightweight nonlinear autoregressive neural network with exogenous input is used to approximate the nonlinear dynamics of the gas collecting tube. The input layer of the model (i.e., the nonlinear autoregressive NARX neural network model with exogenous input) receives the gas collecting tube pressure (PR) at the current time and historical time. 1a PR 1b PR 2a PR 2b ) and the optimized given value of the opening of the gas manifold regulating valve (PV) 1a PV 1b PV 2a PV 2b ), Optimization of the given value for the opening of the blower inlet regulating valve (PV) 41 ), Optimization of the given value for the opening of the blower return valve (PV) 42 The system uses control input variables such as the optimized setpoint (SP) for blower speed, and the output layer predicts the pressure of the gas collecting pipe over multiple prediction time domains. In the neural network modeling process, a multi-layer neural network model with an appropriate number of layers and neurons can be constructed to capture the complex multivariate coupling relationships and nonlinear characteristics of the system. The Adam optimizer, a commonly used neural network training algorithm, is used to train the model, and the model's performance on the validation set is evaluated using cross-validation. The network model with the best prediction accuracy and generalization ability is selected for the subsequent design of the control system. In this embodiment, the trained nonlinear autoregressive neural network model with exogenous input achieves high prediction accuracy for each variable, especially a high fit for the gas collecting pipe pressure. Figure 2 The actual field measurement data is close to the neural network prediction results, and the high-precision prediction model can be used as a reliable prediction model for nonlinear model prediction controllers.

[0063] Step B: Design a multivariable nonlinear constraint optimization control algorithm based on a neural network model.

[0064] See Figure 3 A constrained optimization control architecture is used to construct a predictive model using a nonlinear autoregressive neural network with external input, and dynamic and precise control of the coke oven gas collecting pipe pressure is achieved through rolling optimization.

[0065] The algorithm consists of two phases:

[0066] Offline phase

[0067] Based on historical operating data of the coke oven system, a multivariate prediction model is constructed using neural networks, and the prediction accuracy of the model is optimized through training. Once the model is trained, it can be used for online control.

[0068] Online phase

[0069] Based on the trained prediction model, a nonlinear constrained optimization problem is constructed at each sampling time, and an efficient optimization solver is used to solve it in real time, dynamically adjusting the control variables.

[0070] Based on the multivariable nonlinear characteristics of the coke oven gas collecting pipe system, a neural network model with a nonlinear autoregressive structure is constructed, and the prediction formula is as follows:

[0071] y k =f(y k-1 ,u k-1 ) ①

[0072] In formula ①, y k y represents the pressure value at time k. k-1 u represents the pressure value at the previous moment. k-1 This represents the control input at the previous moment, including the optimized setpoint (PV) of the manifold regulating valve. 1a PV 1b PV 2a PV 2b Optimization of the setpoint for the blower inlet regulating valve (PV) 41 Optimization of the return valve for blowers (PV) 42 ), the blower speed optimization given (SP), considering the influence of historical data, the model is expressed as:

[0073]

[0074] In formula ②, y k+i|k This represents the predicted pressure in the gas collecting pipe at sampling time k with respect to future time k+i; y k+i-1 The pressure vector of the gas collecting pipe at time k+i-1 and multiple historical times prior can be represented as [y k+i-1 ,y k+i-2 ,…,y k+i-p ], p represents the delay order of the historical pressure data, u k+i-1 The control input vector representing time k+i-1 and multiple previous historical times can be represented as [u k+i-1 ,u k+i-2 ,…,u k+i-q ], q is the delay order of the control input historical data; d k+i-1represents the known perturbation input vector at time k+i-1 and multiple previous historical times, such as high-pressure ammonia spraying, coke pushing, coal loading, environmental factors, etc.; f represents the nonlinear mapping relationship implemented by the trained neural network model.

[0075] The neural network adopts a multilayer perceptron structure, uses ReLU as the activation function, and has linear activation for the output layer. The mean squared error is used as the loss function. The network parameters are trained using the Adam optimization algorithm, and cross-validation is used to select the optimal network structure and hyperparameters. The resulting neural network model can predict stress at multiple future moments with a prediction error controlled within ±2%, demonstrating good dynamic prediction performance.

[0076] Based on the obtained lightweight neural network model, a nonlinear model predictive controller is constructed with this model as the predictive subsystem. The controller's objective is to minimize the deviation between the system's predicted output and the target value within a finite prediction time domain, considering the smoothness of the control input and the actual constraints of the system. The controller's optimization objective function is as follows:

[0077]

[0078] The objective function weight matrix Q of the optimization problem N Q and R are used to weigh the impact of pressure tracking error on system performance against changes in control input. N represents the preset control time domain, and it is assumed that the prediction time domain is equal to the control time domain; y k+N This represents the predicted value at time k+N; y r The setpoint indicating pressure fluctuation; y k+i This represents the predicted value at time k+i in the future, calculated at time k; Q represents the weight matrix of pressure fluctuations; and R represents the weight matrix of control inputs.

[0079] To improve the computational efficiency and solvability of the algorithm, hard constraints are transformed into penalty terms in the objective function, allowing for penalties to be imposed for minor violations of hard constraints. By introducing slack variables, the optimization problem is rewritten as follows:

[0080]

[0081] In formula ④, ε is the slack variable, S is the penalty weight matrix, used to control the tolerance for constraint violations; Δu k+i It is the incremental form of the control input, representing the increment of the control input at time k+i predicted at time k.

[0082] The above nonlinear constrained optimization problem is solved using a sequential quadratic programming method. The nonlinear objective function and constraints are linearized and quadratically approximated, respectively, transforming the problem into multiple quadratic programming subproblems to be solved iteratively. The derivative information of the neural network is used to construct the gradient of the optimization problem, and the Gauss-Newton method is used to approximate the Hessian matrix, i.e., for the form ||J(Δu)|| 2 The Hessian approximation of the terms is: The Hessian matrix is ​​approximated by the product of Jacobian matrices, which avoids the explicit calculation of the second derivative, significantly reduces the computational burden of each iteration, and speeds up the optimization process.

[0083] Step C: Introduce system constraints during the optimization process to achieve edge control in constraint optimization.

[0084] Multiple constraints are introduced to limit system operation within a safe range and optimize energy efficiency, preventing abnormal behavior caused by parameter settings or operating condition fluctuations. Specifically, the pressure of the coke oven gas collecting pipe system is limited to a set range. By introducing upper and lower limit constraints, it is ensured that the pressure control sequence generated during the optimization process always meets the following requirements:

[0085] y min ≤y k+i ≤y max ,i=1,2,…N ⑤

[0086] In formula ④, to protect the stable operation of control equipment and systems, safe operating ranges are set for key operating variables such as valve opening and fan speed to avoid overreaction or loss of control caused by complete closure or full opening. The constraint is expressed as follows:

[0087] u min ≤u k+i ≤u max ,i=0,1,…N-1 ⑥

[0088] Δu min ≤Δu k+i ≤Δu max ,i=0,1,…N-1 ⑦

[0089] In formula ⑥, u k+i This represents the control input predicted at time k+i from time k; u min This represents the minimum value of the control input; u max Indicates the maximum value of the control input;

[0090] In formula ⑦, Δu k+i Δu represents the increment of the control input at time k+i predicted at time k; minThis represents the minimum value of the control input increment; Δu max This indicates the maximum value of the control input increment;

[0091] The system dynamic constraints are the dynamic characteristics predicted by the neural network model. By using multi-step prediction to simulate the future system dynamics, dynamic disturbances can be identified in advance, thereby enhancing the system's adaptability and robustness to dynamic disturbances.

[0092] Step D: Use a neural network model to predict pressure and dynamically adjust valve opening and fan speed based on the optimization results to achieve closed-loop control of the coke oven gas collection pipe system.

[0093] Key operating parameters of the coke oven system are collected in real time using a sensor network and used as input data for predictive calculations. These parameters include current gas collecting pipe pressure, valve opening, and blower speed. The real-time collected parameters are then input into an offline-trained neural network model to predict the pressure change trend y over the next N control time domains. k+1 y k+2 ... y k+N The neural network model quickly obtains the prediction results through forward computation and passes them to the optimization module.

[0094] Based on the prediction results and the system's dynamic constraints, the optimization results are calculated using a nonlinear constraint optimization control algorithm, obtaining the final result after meeting the accuracy requirements of the KKT conditions. Within each control cycle, after solving the optimization problem, the first value of the control sequence is extracted. As the control command at the current sampling moment, the valve opening and blower speed are dynamically adjusted to achieve precise control of the gas collecting pipe pressure. The remaining parts are recalculated in the next cycle to adapt to dynamic changes in the system. The control command is sent to the execution unit in real time to dynamically adjust the valve position of the gas collecting pipe regulating valve and the speed of the blower, thereby achieving high-precision closed-loop control of the coke oven gas collecting pipe pressure. At the same time, during the control process, the system can sense changes in operating status in real time and adaptively respond to possible disturbances, operation switching or model deviations, improving the robustness and stability of the system.

[0095] The proposed coke oven gas collecting pipe pressure control method based on neural network multivariate constraint optimization was validated using measured data from coke oven gas collecting pipe systems at domestic coal and coke chemical sites. During validation, the typical setpoint for the coke oven gas collecting pipe pressure was 150 Pa. The coke oven gas collecting pipe system was controlled using the neural network-based nonlinear model predictive controller designed in this invention. The control effect obtained is shown in [see figure]. Figure 4As can be seen, under the control of the multivariable constrained optimization control algorithm of this invention, the pressure of all four gas collecting pipes can accurately track the setpoint of 150 Pa, reflecting the expected goal of multivariable control. To further demonstrate the tracking control performance of the designed controller, the setpoint of the gas collecting pipe pressure was changed from 150 Pa to 149 Pa at the 1000th sampling time of the simulation verification. Figure 4 The response curves show that the pressure of all four gas collection pipes can quickly respond to changes in the set value and rapidly stabilize at around 149 Pa, indicating that the method of the present invention has good set value tracking capability. Figure 4 The local region of pressure fluctuation response of the gas collecting pipe system under dynamic disturbances was further amplified. As shown in the figure, despite the presence of disturbances, the control method designed in this invention can still effectively control the gas collecting pipe pressure, maintaining it within a small fluctuation range. Statistical analysis of the verified control results shows that, near the setpoint of 150 Pa, the proportion of gas collecting pipe pressure fluctuations within ±20 Pa is as high as 99.98%. In comparison with existing technologies, current coke oven gas collecting pipe pressure control typically employs a DCS-based PID control system. Due to the significant multivariable coupling, nonlinear characteristics, and dynamic disturbances caused by coke oven reversing heating issues in the coke oven gas collecting pipe system, traditional PID control struggles to achieve high-precision control. According to statistical analysis of measured operating data of existing DCS PID control systems under the same operating conditions, only 44.5% of the gas collecting pipe pressure fluctuations are within ±20 Pa of 150 Pa, and 24.3% of the data fluctuations exceed ±50 Pa. The comparative analysis above shows that, compared with the traditional PID control method, the multivariable constraint optimization control method based on neural networks proposed in this invention significantly reduces the control fluctuation of the gas collecting pipe pressure and greatly improves the system's control accuracy, dynamic performance, and robustness. Higher control accuracy helps improve the operational stability of the coke oven, reduces gas emissions, thereby indirectly contributing to reduced energy consumption and having a positive impact on improving coke quality.

[0096] In summary, the coke oven gas collecting pipe pressure control method based on multivariate constraint optimization of this invention provides a novel industrial control solution for the coking industry. This invention first utilizes a neural network to construct a multivariate nonlinear prediction model of the coke oven gas collecting pipe system, and then combines this with a sequential quadratic programming algorithm to design an efficient constraint optimization control architecture. By introducing constraints such as upper and lower pressure limits and safe operating ranges, precise dynamic pressure control is achieved through rolling optimization. This method can significantly improve the accuracy and stability of system pressure tracking, reduce energy consumption fluctuations, and improve coke quality and production efficiency. The method provided by this invention is applicable to pressure control in complex industrial environments, providing important technical support for the intelligent development of the coking industry and possessing broad engineering application value.

[0097] This invention ensures that the generated dataset comprehensively covers the state of the coke oven gas collection pipe system in various actual operating scenarios. This allows the subsequently trained NARX neural network model to fully learn the dynamic behavior patterns of the system under different operating conditions, avoiding inaccurate or unpredictable predictions in certain conditions due to missing data. The screening process removes abnormal, erroneous, and irrelevant data, retaining only the data valuable for model training. Preprocessing (such as data alignment and normalization) eliminates the adverse effects of differences in the units of different variables on model training, making the data more standardized and orderly, meeting the requirements of neural network models for input data, thereby improving... The efficiency and accuracy of model training lay a solid foundation for subsequent model training, validation, and testing. The coke oven gas collecting pipe system exhibits strong multivariate coupling and nonlinear dynamic characteristics, making it difficult for traditional linear models to accurately describe its complex behavior. The NARX neural network, as a nonlinear model, can effectively capture and learn the complex coupling relationships between multiple variables in the system through its nonlinear mapping capabilities. After training and validation, the NARX neural network model can accurately describe the dynamic characteristics of the coke oven gas collecting pipe system under different operating conditions, providing high-precision system model support for subsequent pressure control. Through training and validation on datasets containing data from different operating conditions, NAR… The NARX neural network model can learn the general laws of a system, rather than being limited to behavioral patterns under a specific operating condition. When applied to the actual control of a coke oven gas collecting pipe system, it demonstrates good adaptability and generalization ability to unseen operating conditions or changes in system state, ensuring the reliability of prediction and control effects under various actual operating conditions. Because the NARX neural network model can accurately describe the multivariable coupling characteristics of the coke oven gas collecting pipe system, the multivariable nonlinear constraint optimization control algorithm designed based on this model can fully consider the mutual influence and coupling relationships between variables in the system. This allows the control algorithm to more accurately control multiple variables (such as valve opening, blower speed, etc.). Coordinated optimization control (such as speed) is performed to avoid system imbalance or poor control effect caused by single variable control, thereby improving the performance and stability of gas collecting pipe pressure control. The operation of the coke oven gas collecting pipe system is subject to various constraints, including upper and lower pressure limits, valve opening range, blower speed range, and energy consumption limits. The optimization control algorithm based on the NARX neural network model can incorporate these complex constraints into the control strategy. By constructing and solving a nonlinear constraint optimization problem, it ensures that the control results meet both the requirements for stable system operation and the actual constraints of industrial production, achieving comprehensive optimization of the system in terms of safety, economy, and efficiency.Introducing constraints such as the upper and lower limits of coke oven gas collecting pipe pressure, valve opening, and safe operating range of blower speed can strictly limit the changes of control variables and system state within safe ranges, preventing equipment damage, production accidents, or environmental pollution caused by excessively high or low pressure, excessively large or small valve opening, or abnormal blower speed, thus ensuring the safe and stable operation of the coke oven system. Considering energy consumption constraints and adding penalty terms for control input constraints and gas collecting pipe pressure constraints to the optimization objective function can encourage the optimization algorithm to minimize energy waste and unnecessary consumption while meeting system pressure control requirements. Furthermore, the penalty terms can improve the optimization problem's resilience to measurement noise and model errors. Its robustness allows for minor violations of constraints under certain circumstances, but by penalizing these violations, the system can optimize its operation as much as possible within safe and economical limits, reducing production costs and improving the company's economic efficiency. Real-time acquisition of key parameters of the coke oven system and input into the NARX neural network model for prediction allows for advance prediction of the coke oven gas collecting pipe system's state over a certain time range (prediction time domain). This enables the control system to anticipate potential development trends and changes, such as pressure rise or fall trends and possible pressure fluctuations, thus providing a basis for taking appropriate control measures in advance and enhancing the control system's foresight and proactivity. Because of the coke oven's robustness, the system can optimize its operation within safe and economical limits, reducing production costs and improving the company's economic efficiency. Furthermore, real-time acquisition of key parameters of the coke oven system and input into the NARX neural network model for prediction allows for advance prediction of the coke oven's gas collecting pipe system's state over a certain time range (prediction time domain). This allows the control system to understand potential development trends and changes, such as pressure rise or fall trends and possible pressure fluctuations, providing a basis for taking appropriate control measures in advance and enhancing the control system's foresight and proactivity. The operating conditions of the coke oven system are dynamically changing. Real-time forecasting can promptly reflect the latest changes in the system state. The control system can quickly adjust its control strategy based on these real-time forecasts, enabling control parameters to adapt to changes in operating conditions in a timely manner. This improves the system's response speed and adaptability to dynamic disturbances (such as pressure fluctuations caused by coal charging and coke pushing operations), ensuring that the coke oven gas collecting pipe pressure remains stable within the set range and maintaining the normal operation of the coking process. Applying the first value of the optimal control input increment sequence from the constrained optimization control output to the control input in a timely manner can quickly translate the optimization results into actual control actions, such as adjusting valve openings or changing blower speeds. This timely update of control parameters ensures the system... Within each control cycle, control is performed based on the current optimal solution, avoiding control lag and improving the real-time performance and effectiveness of the coke oven gas collecting pipe pressure control. By continuously updating the control input values, the control system can continuously fine-tune and optimize the coke oven gas collecting pipe pressure, ensuring the system always remains in its optimal operating state. Furthermore, because the optimization algorithm is based on system constraints and dynamic characteristics, each adjustment of control parameters is made within a safe and reasonable range, thus guaranteeing the stability and continuity of the control process. This avoids system oscillations or instability caused by excessive or inappropriate adjustments to control parameters, ensuring precise control of the coke oven gas collecting pipe pressure and long-term stable system operation.

Claims

1. A method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization, characterized in that, include: S1. Collect, filter and preprocess historical operating data of the coke oven gas collecting pipe system under different operating conditions, generate a dataset for training, validating and testing the nonlinear autoregressive NARX neural network model with external input, train and validate the nonlinear autoregressive NARX neural network model with external input, and obtain the multivariate coupling characteristics describing the coke oven gas collecting pipe system. S2. Based on the NARX neural network model, design a multivariable nonlinear constraint optimization control algorithm; S3. In the nonlinear constraint optimization problem, constraints reflecting the actual operation of the coke oven system are introduced, including the upper / lower limit of the coke oven gas collecting pipe pressure, the safe operating range of valve opening and blower speed, and energy consumption limits. Penalty terms for control input constraints and gas collecting pipe pressure constraints are added to the optimization objective function to punish violations. S4. At the beginning of each control cycle, key parameters of the coke oven system are collected in real time, and the key parameters of the coke oven system are predicted and calculated through the NARX neural network model to obtain the predicted value of the coke oven gas collecting pipe system state in the future prediction time domain. The control parameters are adjusted according to the optimization results of the constraint optimization control output. The first value of the optimal control input increment sequence is applied to the control input to update the control input value and complete the control of the gas collecting pipe pressure. The multivariable nonlinear constraint optimization control algorithm is described below: Within each control cycle, a nonlinear constrained optimization problem is constructed to predict the dynamic behavior of the coke oven gas collecting pipe system at multiple future moments. The objective function is to minimize the sum of squares of the tracking error between the gas collecting pipe pressure and the set value in the future prediction time domain, the weighted sum of the sum of squares of the incremental changes in the control input, and includes constraints on the nonlinear system dynamic equations based on a nonlinear autoregressive NARX neural network model with external input. The control variables are dynamically adjusted through rolling optimization based on sequential quadratic programming, so that the coke oven gas collecting pipe system can follow the gas collecting pipe pressure set value, while suppressing pressure fluctuations in the carbonization chamber caused by different operating conditions. The nonlinear autoregressive NARX neural network model with exogenous input is obtained through offline training. The inputs to the nonlinear autoregressive NARX neural network model include current and historical key parameters. The output of the nonlinear autoregressive NARX neural network with exogenous input is the predicted value of the gas collection pipe pressure in multiple future prediction time domains.

2. The method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization according to claim 1, characterized in that, In S1, the preprocessing includes merging data with aligned time by taking the intersection of the time axes, and mapping the data to the normalization interval using the maximum and minimum value normalization method, in order to eliminate the influence of differences in the magnitude of different variables on the training and prediction of the nonlinear autoregressive NARX neural network model with external input. The screening adopts a dynamic segmentation method, which is based on the identification results of the coke oven operating conditions, and divides the coke oven gas collecting pipe system operating data into multiple subsets. The dataset described is obtained by merging data with aligned time by taking the intersection of the time axes and performing preprocessing such as normalization of maximum and minimum values.

3. The method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization according to claim 1, characterized in that, The multivariable nonlinear constraint optimization control algorithm combines the dynamic changes of coke oven operating conditions with the optimization of control variables. The dynamic changes are predicted by a nonlinear autoregressive NARX neural network model with external input.

4. The method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization according to claim 1, characterized in that, In S3, the upper / lower limits of the coke oven gas collecting pipe pressure are constrained. The constraint range is set according to the coke oven safety production requirements and process requirements. Valve opening and blower speed are introduced into safety operation constraints. The addition of penalty terms for control input constraints and gas collection pipe pressure constraints to the optimization objective function is achieved by introducing relaxation variables to allow small-scale violations of hard constraints and imposing penalty terms on the relaxation variables in the optimization objective function.

5. The method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization according to claim 1, characterized in that, In S4, the real-time collected gas collecting pipe pressure is fed back to the control system and compared with the set value of the gas collecting pipe pressure. The control parameters are adjusted by the optimization result output by the multivariable nonlinear constraint optimization control algorithm to form a feedback loop.

6. The method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization according to claim 1, characterized in that, The rolling optimization method of the sequential quadratic programming described above transforms the nonlinear constraint optimization problem into multiple quadratic programming subproblems for solution. The quadratic programming subproblem is solved using a quadratic programming solver; When solving the quadratic programming subproblem, the Gauss-Newton method is used to approximate the Hessian matrix.

7. The method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization according to claim 1, characterized in that, The optimization objective function includes minimizing the sum of squares of the tracking error between the gas collecting pipe pressure and the set value in the future prediction time domain, and the weighted sum of the sum of squares of the incremental changes in the control input. It also includes a constraint term on the control input, which is dynamically adjusted according to the coke oven gas production and pressure fluctuation limit requirements to ensure the stable gas collecting pipe pressure and complete the energy optimization operation of the coke oven system.

8. The method for controlling the pressure of a coke oven gas collecting pipe based on multivariable constraint optimization according to claim 1, characterized in that, The method of adjusting control parameters based on the optimization results of constraint optimization control output includes applying the first value of the optimal control input increment sequence to the control input and updating the control input value by incremental increment.

Images