A method, device and system for automatic control of temperature stability in a zinc smelting roasting process

Abstract

A kind of zinc smelting roasting process temperature stabilizing automatic control method, device and system, the roasting furnace includes roasting furnace body, setting in the belt speed sensor of roasting furnace, temperature sensor being arranged at roasting furnace, feed belt scale being arranged at roasting furnace, DCS system, automatic control method includes: collecting the historical data of roasting furnace operation save in database, after pre-processing to historical data, system model is identified, design model predictive control algorithm optimization controller, based on predictive model calculation control quantity Δµ Directly write in DCS and act on roasting furnace, to this cycle.The present application provides an automatic control method, device and system for the complex and changeable situation of industrial roasting furnace operation condition, by predictive model update to adapt to different working conditions, based on intelligent algorithm, predict future operating state and solve accurate optimal control quantity, to replace manual input control instruction, reduce human factor, improve the stability of roasting furnace temperature control.

Description

Technical Field

[0001] This invention relates to the field of zinc smelting technology, specifically to an automatic control method, apparatus, and system for stabilizing the temperature during the zinc smelting roasting process. Background Technology

[0002] Zinc smelting primarily employs hydrometallurgy. Zinc roasting, the first step in this process, converts zinc concentrate into zinc roasted ore, and the quality of this ore is crucial for downstream production. Industrially, the quality of zinc roasted ore is determined by testing the content of zinc oxide and insoluble sulfides. However, in industrial settings, testing is only conducted once a day, making it impossible to obtain real-time quality data during the roasting process, thus impacting real-time control. To assess the real-time quality of zinc roasting, industrial processes typically use the furnace standard temperature as an indirect indicator of the roasting effect. However, due to the complex and variable operating conditions of the roasting furnace influenced by factors such as feed rate and blast volume, the furnace standard temperature fluctuates drastically, leading to instability in the quality of the zinc roasted ore.

[0003] Currently, the standard temperature fluctuation control of roasting is controlled by operators on duty 24 hours a day. Based on industrial field test data, the operators manually adjust the feed belt speed, and then adjust the feed rate and blower frequency settings. However, manual control is affected by the operator's experience and subjective factors. The standard temperature fluctuation in the roasting furnace usually reaches ±25℃, which affects the quality of zinc roasted sand. Moreover, manual control is time-consuming, labor-intensive, and poses safety hazards.

[0004] The patent with publication number CN110471285A provides a fuzzy control method for zinc smelting roasting process based on trend event driving. It improves the problem of difficulty in evaluating working conditions and the decline in the control performance of fuzzy controllers caused by the dynamic characteristics of the roasting process and the constraints of the on-site environment. The disadvantage is that the control quantity is not precise enough and cannot adapt to changing working conditions.

[0005] The patent with publication number CN202110133546.4 provides a method for stabilizing the temperature of a fluidized bed roasting furnace, which is beneficial to improving the stability of the temperature control of the fluidized bed roasting furnace. The disadvantage is that the control rules are relatively simple and rely on manual operation experience, and the control accuracy is not high when facing complex working conditions. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide an automatic temperature control method, device and system for zinc smelting roasting process that addresses the complex and variable operating conditions of industrial roasting furnaces. This method uses predictive model updates to adapt to different operating conditions and, based on intelligent algorithms, predicts future operating states and solves for precise optimal control quantities. This replaces manual input of control commands, reduces human factors, and improves the stability of roasting furnace temperature control.

[0007] The present invention also provides a temperature stabilization device and system for the zinc smelting roasting process.

[0008] The technical solution adopted in this invention is as follows:

[0009] An automatic temperature stabilization control method for a zinc smelting roasting process, characterized in that the roasting furnace includes a roasting furnace body, a belt speed sensor installed on the roasting furnace belt to acquire the speed of the feed belt, a temperature sensor installed in the roasting furnace to acquire the standard temperature of the boiling layer, a feed belt scale installed on the roasting furnace belt, and the data collected by the belt speed sensor and temperature sensor are directly transmitted to a DCS system, with control commands issued to the roasting furnace by the DCS system; the zinc smelting roasting process temperature stabilization control method includes:

[0010] The OPC communication protocol is used to communicate between the calcining furnace DCS system and the industrial control computer;

[0011] Historical data on the operation of the roasting furnace are collected and stored in a database; the historical data includes feed belt speed, standard temperature, and feed rate.

[0012] The historical data is preprocessed;

[0013] Based on the preprocessed data, the system model is identified using the subspace method, and an optimized controller for the model predictive control algorithm is designed. The control quantity Δμ is calculated based on the identified predictive model, where the control quantity is the change in belt speed. The system model identification uses belt speed as the control input and standard temperature as the predictive output.

[0014] The control quantity Δμ is corrected using auxiliary rules;

[0015] The control quantity Δμ is directly written into the DCS and applied to the calcining furnace;

[0016] This cycle is repeated to maintain a stable standard temperature throughout the entire roasting process in the roasting furnace.

[0017] Furthermore, if the prediction model is suitable, the control quantity Δμ is directly written into the DCS system and applied to the roasting furnace; if the prediction model is not suitable or an abnormal situation occurs, the control quantity Δμ is corrected according to the auxiliary rules, and then the corrected control quantity Δμ is written into the DCS system and applied to the roasting furnace. The control method is to transform the actual roasting furnace into a data-driven prediction model and apply it to the control algorithm. Since the actual operating state of the roasting furnace is nonlinear, while the identified model is linear, it can only simulate the operating state for a certain period of time. In other time periods, the model may not be consistent with reality and may not be suitable.

[0018] Furthermore, the data preprocessing method is as follows:

[0019] Differential operation: Perform differential operation on belt speed and standard temperature. If a large number of zero data points are found, perform secondary sampling and differential operation to obtain the corresponding change value.

[0020] Standardization: Let S = (s1, s2, ..., s...) p+q To model the overall input and output dataset, standardization is performed.

[0021]

[0022] Where s i Let N be a column vector with a length equal to the number of samples in the dataset. The mean of the column vectors is calculated and then expanded to the original column vector size, denoted as N. The function sum() sums the elements of a column vector, where σ is the standard deviation.

[0023] Filtering preprocessing: Let the size of the forward sliding window be k. F s(t) is a row vector in the dataset S.

[0024]

[0025] Furthermore, the specific method for identifying the system model using the subspace method is as follows:

[0026] Based on the output and input vectors, considering k past and future moments, let U = (u1, u2, ..., u...). p Y = (y1, y2, ..., y3) represents the observable variable data at the input end. q ) represents the given modeling output variable data;

[0027]

[0028] in n u n y and n x The dimensions are the control variable, the output variable, and the state variable, respectively. N is calculated based on the dataset size (size) and the hyperparameter k, i.e., N = size - 2 * k + 1. From the input-output matrix, we can define the past and future system states:

[0029]

[0030] QR decomposition can transform the input-output matrix into the following form:

[0031]

[0032] The R matrix on the left is a lower triangular matrix, and the Q matrix on the right is an orthogonal matrix. Observing the third column of the R matrix, we can see that past inputs and outputs were zero, and future inputs are also zero. Therefore, we can conclude that future outputs will also be zero. Hence, R... 33 The value is 0. The relationship between the future output and the known past and future inputs / outputs, obtained from the QR decomposition, is expressed as follows:

[0033]

[0034] remember get

[0035] Y f =L w W p +L u U f (7)

[0036] The subspace predictor coefficient matrix Then, the obtained subspace predictor model is fed into the model predictive control algorithm as the prediction model;

[0037] The model predictive control algorithm is as follows: The optimization problem in the controller can be constructed in the following form:

[0038]

[0039] stY f =L w W p +L u U f (8)

[0040] L x ≤u(i)≤L s

[0041]

[0042] In the formula, N p For predicting the time domain, N c To control the time domain, N is typically used. p ≥N c Select N p =N c =N, Furthermore, P≥0 is positive semi-definite, and R>0 is positive definite. For the set control objective, L x and L s These are the upper and lower limits of the control quantity, LCX x and LCX sThese are the upper and lower limits of the increase / decrease overlimit constraint. The increase / decrease overlimit constraint refers to the cumulative control change over a certain period not exceeding a certain range. Directly solving this optimization problem has high complexity, so we introduce...

[0043]

[0044] Based on the subspace predictor (7) and auxiliary variables (9), the optimization objective function of MPC is transformed into the following form:

[0045]

[0046] The optimization problem of MPC is transformed into a quadratic programming (QP) optimization problem, which is solved using the active set method.

[0047] A temperature stabilization control device for zinc smelting roasting process, the device comprising:

[0048] Data acquisition module 1 is used to collect historical data of the roasting furnace operation and store it in the database, and read real-time operation data at 10-second intervals;

[0049] Data preprocessing module 2 is used to resample, differentially process, standardize, and filter the belt speed data and standard temperature data from the calcining furnace sensor data to obtain the corresponding change values; when the data enters the database, it is real-time data, and when it is stored in the database, it is historical data, and the historical data and real-time data are the same.

[0050] Model Predictive Control Module 3 is used to identify the predictive model based on the preprocessed data using the subspace method, and to adapt it to the roasting process at each moment by updating the model. Based on the identified predictive model, the belt speed control quantity Δμ is calculated. If the predictive model is suitable, it is directly written into the DCS system through the OPC communication protocol and applied to the roasting furnace.

[0051] The auxiliary rule correction module 4 is used to correct the belt speed control amount Δμ when the prediction model is not suitable or an abnormal situation occurs.

[0052] Furthermore, the auxiliary rule correction module includes a material fluctuation module, a blower fluctuation module, an anomaly handling module, and a long-term temperature difference handling module.

[0053] Furthermore, the system includes a roasting furnace DCS system and an industrial control computer, and a temperature stabilization control device for the zinc smelting roasting process. The roasting furnace DCS system and the industrial control computer communicate with the temperature stabilization control device for the zinc smelting roasting process via the OPC communication protocol.

[0054] Furthermore, it includes a temperature stabilization control device for the zinc smelting roasting process, wherein the roasting furnace body is communicatively connected to the temperature stabilization control device for the zinc smelting roasting process.

[0055] The beneficial effects of this invention are:

[0056] This invention addresses the complex and variable operating conditions of industrial roasting furnaces by providing an automatic temperature control method, device, and system for zinc smelting roasting processes. This system utilizes predictive model updates to adapt to different operating conditions and employs intelligent algorithms to predict future operating states and solve for precise optimal control quantities. This replaces manual input of control commands, reduces human error, and improves the stability of roasting furnace temperature control.

[0057] This invention performs resampling, differential processing, standardization, and filtering preprocessing on historical data of the roasting furnace stored in the database. Based on the subspace method, it identifies the roasting furnace system model and transmits it to the control layer as a prediction model. The model predictive control algorithm is used to solve for the optimal control quantity. The optimal control quantity, which is calculated and corrected in real time, is directly written into the DCS system and applied to the roasting furnace. This process is repeated to replace manual input of control commands, reduce human factors, and improve the stability of temperature control throughout the roasting process of the roasting furnace.

[0058] The effectiveness of the model predictive control algorithm of this invention depends on the quality of the predictive model. When the predictive model is not suitable or an abnormal situation occurs, it will cause the MPC to solve for incorrect control quantities. At this time, auxiliary rules are added to correct the control quantities to ensure the effectiveness of real-time control of the entire zinc smelting roasting process. Attached Figure Description

[0059] Figure 1 This is a structural block diagram of the automatic control method of the present invention;

[0060] Figure 2 This is a diagram of the automatic control device of the present invention;

[0061] Figure 3 This is a diagram showing the results of manual control during the roasting process in the roasting furnace according to an embodiment of the present invention.

[0062] Figure 4 This is a diagram showing the automatic control results of the roasting process in the roasting furnace according to an embodiment of the present invention;

[0063] Figure 5 The algorithm flowchart of the automatic control method according to an embodiment of the present invention. Detailed Implementation

[0064] Take the zinc roasting process of a large zinc smelter in Yunnan as an example. This roasting process has a 109m³... 2The fluidized bed roasting furnace, built and put into operation in May 2005, has been in production for over ten years. However, the deterioration of its internal structure and equipment has made stable control increasingly difficult. Furthermore, the furnace's level of automation lags behind that of newly commissioned roasting furnaces across the country in recent years, resulting in low production efficiency. In actual industrial production, due to the reliance on manual input of control commands and high dependence on operator experience, the furnace's standard temperature fluctuates by ±25℃. Figure 3 As shown.

[0065] This embodiment employs the calcination process temperature stabilization control device and system of the present invention to improve the temperature stability and intelligence level of the calcination process. The specific solution is as follows: Figure 1 , Figure 2 As shown, the roasting furnace uses Emerson's Delta VDCS system. Sensor data from the physical layer is directly transmitted to the DCS system, and control commands are also issued to the roasting furnace from the DCS system. This embodiment adds a temperature stabilization control device for the roasting process of the present invention, and adopts the automatic temperature stabilization control method of the present invention to replace manual input of control commands.

[0066] Install an automatic temperature control device for the roasting process in the roasting furnace:

[0067] Data Acquisition Module 1

[0068] The DCS system and industrial control computer are communicated using the OPC communication protocol to collect data, store historical data of the calcining furnace operation in the database, and read real-time operation data at 10-second intervals.

[0069] Data Preprocessing Module 2

[0070] The control algorithm uses data including the temperature of the southwest 1st fluidized bed in the calcining furnace (standard temperature), the feed belt speed, the feed rate, the temperature setting, the time, and historical suggested values. The standard temperature and feed belt speed are used to identify the system model; the feed belt speed is the control input, and the standard temperature is the predicted output.

[0071] First, for the prediction model, since the mapping relationship between the change in feed belt speed and the change in standard temperature is better than the mapping relationship between the original speed and temperature, and the calcining furnace is a complex nonlinear system, the linear model obtained can only fit the operating conditions within a short time range. Therefore, when the data of the identified model is at a higher temperature and the current temperature is lower, the model identified based on the original data may not be applicable, but the model identified based on the change is still applicable. Therefore, the first step in data processing is to perform a difference operation on the data of feed belt speed and standard temperature to obtain the corresponding change value.

[0072] Since the control cycle T is 2 minutes, the feed belt speed is constant within 2 minutes, resulting in a large number of zero data points after differential sampling. This can significantly negatively impact model identification. Therefore, secondary sampling of the data is necessary, with a sampling interval of 2 minutes (i.e., 12 data points). In practical applications, errors may occur when reading data from the DCS system, leading to missed or incorrect readings. In such cases, the interval between changes may not be 12 data points. To address this, secondary sampling can be performed before differential sampling to avoid modeling data errors.

[0073] Considering the difference in magnitude between the change in feed belt speed and the change in standard temperature, to improve the accuracy of the identification model, let S = (s1, s2, ..., s...). p+q To model the overall input and output dataset, it will be standardized:

[0074]

[0075] Where s i Let N be a column vector with a length equal to the number of samples in the dataset. The mean of the column vectors is calculated and then expanded to the original column vector size, denoted as N. The function sum() sums the elements of a column vector, where σ is the standard deviation.

[0076] The calcining furnace is a high-noise system, and the data collected on-site contains noise information, which needs to be filtered to suppress the interference of external noise. This paper adopts the moving average filtering algorithm

[10] , and the size of the forward sliding window is denoted as k. F s(t) is a row vector in the dataset S:

[0077]

[0078] After data standardization and filtering, the data collected on site has been integrated into a suitable dataset, and subsequent modeling and control are based on this dataset.

[0079] Model Predictive Control Module 3

[0080] For the complex reaction processes inside the roasting furnace, simple linear models can only fit the furnace operation process over a short period, while complex nonlinear models lead to a huge computational load in the control stage, affecting the real-time performance of the algorithm. This invention uses a subspace predictor as the system model and adapts it to the roasting process at each moment through model updating methods;

[0081] The specific method is as follows: Based on the output vector and input vector, considering k past and future moments, the Hankel matrix of the dataset is written in the following form: Let U = (u1, u2, ..., u... pY = (y1, y2, ..., y3) represents the observable variable data at the input end. q ) is the given modeling output variable data.

[0082]

[0083] in n u n y and n x These represent the dimensions of the control variable, output variable, and state variable, respectively. N is calculated based on the dataset size (size) and the hyperparameter k, i.e., N = size - 2 * k + 1. The past and future system states are defined by the input-output matrix.

[0084]

[0085] QR decomposition can transform the input-output matrix into the following form:

[0086]

[0087] The R matrix on the left is a lower triangular matrix, and the Q matrix on the right is an orthogonal matrix. Observing the third column of the R matrix, we can see that past inputs and outputs were zero, and future inputs are also zero. Therefore, we can conclude that future outputs will also be zero. Hence, R... 33 The value is 0. From the QR decomposition results, we can obtain the expression relating the future output to the known past inputs and outputs, and the future inputs:

[0088]

[0089] remember get

[0090] Y f =L w W p +L u U f (7)

[0091] The subspace predictor coefficient matrix Then, the obtained subspace predictor model is fed into the model predictive control algorithm as the prediction model.

[0092] The principle of model predictive control (MMC) algorithm design is as follows: within each control cycle, based on the currently detectable system state value, a finite-time open-loop optimization problem is solved online. The first value of the solved control sequence is then applied to the controlled system to update the system state. In subsequent control cycles, the new state is used to update the optimization problem online, and new control variables are calculated to achieve rolling optimization. For the calcining furnace, the optimization problem in the controller is constructed as follows:

[0093]

[0094] stY f =L w W p +L u U f (8)

[0095] L x ≤u(i)≤L s

[0096]

[0097] In the formula, N p For predicting the time domain, N c To control the time domain, N is typically used. p ≥N c To balance accuracy and complexity, N is chosen. p =N c =N, Furthermore, P≥0 is positive semi-definite, and R>0 is positive definite. For the set control objective, L x and L s These are the upper and lower limits of the control quantity, LCX x and LCX s These are the upper and lower limits of the increase / decrease overlimit constraint. The increase / decrease overlimit constraint refers to the cumulative control change over a past period not exceeding a certain range. Directly solving this optimization problem is highly complex; therefore, we introduce...

[0098]

[0099] Based on the subspace predictor (7) and auxiliary variables (9), the optimization objective function of MPC can be transformed into the following form:

[0100]

[0101] At this point, the optimization problem of MPC has been transformed into a quadratic programming (QP) optimization problem. Solving the QP problem can be handled by mature toolkits or built-in functions, among which the active set method is generally used.

[0102] Auxiliary rule module 4

[0103] The effectiveness of model predictive control algorithms depends on the quality of the predictive model. When the predictive model is not suitable or anomalies occur, it often leads to MPC (Model Predictive Control Method) solving for incorrect control values. In this case, auxiliary rules need to be added to correct the control values ​​to ensure the effectiveness of control at all times.

[0104] Based on the operating characteristics of the calcining furnace, the added auxiliary rule modules include a material fluctuation module, an air fluctuation module, an anomaly handling module, and a long-term temperature difference handling module. Among them:

[0105] Material fluctuation module: The feed to the roasting furnace is directly fed from the concentrate silo onto the feed conveyor belt, where it is roughly distributed in an arch shape. However, due to the varying sizes of the concentrate particles, the upper edge of the arch has an irregular serrated shape, resulting in uneven feed distribution and fluctuations in the feed rate at the same feed conveyor speed. In the DCS system, the feed rate variable "Feed" refers to the tons of concentrate fed per hour, derived from the weight of concentrate per unit area on the feed conveyor belt, and the feed conveyor speed is denoted as "speed". The material fluctuation correction rules are as follows:

[0106] if E[Feed(-2<t<0)]-E[Feed(-4<t<-2)]>1, then speed=speed-1

[0107] if E[Feed(-2<t<0)]-E[Feed(-4<t<-2)]-1, then speed=speed+1

[0108] In the formula, the function E(x) represents the average value. This rule means that when the average feed rate of the past two minutes is 1 greater than the average of the two minutes before that, it is considered that the feed rate is increasing while the rotation speed is the same, and the rotation speed needs to be reduced; when the average feed rate of the past two minutes is 1 less than the average of the two minutes before that, it is considered that the feed rate is decreasing while the rotation speed is the same, and the rotation speed needs to be increased.

[0109] Blast Fluctuation Module: With fluctuations in the power grid, the actual blast volume will change even without altering the blast frequency setpoint. Significant blast volume fluctuations can noticeably impact the reaction within the roasting furnace. It is believed that frequent changes to the feed speed during periods of severe blast fluctuation will further exacerbate reaction volatility within the roasting furnace, making furnace operation difficult to control. Therefore, to address blast fluctuations, an auxiliary rule is added, denoted as blast volume:

[0110] if E[max 10 (blast)]-E[min 10(blast)]>4000, then T=4min

[0111] Considering the blower volume (PV) over the past 10 minutes, the 10 largest and 10 smallest blower volumes were selected. The difference in their average values ​​exceeded 4000 m³ / s. 3 If / h is used, the control cycle T will be changed from 2 minutes to 4 minutes.

[0112] Anomaly Handling Module: When the prediction model is not well-suited, the control values ​​calculated by MPC may exhibit significant anomalies. Directly inputting these anomalies into the DCS can cause the calcining furnace to evolve towards a larger temperature difference. To prevent the impact of outliers, anomaly handling rules are added:

[0113]

[0114]

[0115] In the formula, y(i) represents the standard temperature inside the furnace, and u(i) represents the change in belt speed. This rule is explained as follows: if the standard temperature is lower than the target setpoint, the cumulative change in belt speed over the past 20 minutes is less than zero, and the current control quantity is the minimum control quantity allowed by the MPC constraint, then the control quantity is an outlier, and the control quantity is set to -1 to reduce the impact of the outlier; if the standard temperature is higher than the target setpoint, the cumulative change in belt speed over the past 20 minutes is greater than zero, and the current control quantity is the maximum control quantity allowed by the MPC constraint, then the control quantity is an outlier, and the control quantity is set to +1 to reduce the impact of the outlier.

[0116] Long-term temperature difference handling module: If the temperature inside the roasting furnace remains significantly different from the set value for an extended period, it indicates insufficient adjustment of the feed belt speed. This is because normal control involves fine-tuning the speed, and the MPC also has over-limit constraints restricting speed changes over a certain period. In abnormal situations, the fine-tuning cannot promptly bring the temperature back to the set value, leading to excessive overshoot or prolonged large temperature differences. To address this, a long-term temperature difference handling module is added: If a large temperature difference persists for the past 30 minutes (threshold 5℃), the temperature difference is used as an additional control factor, and the time is recalculated; when the temperature difference returns to a smaller range, this additional factor is deducted back.

[0117] Algorithm Implementation

[0118] like Figure 5As shown, based on the data preprocessing module, model predictive control module, and auxiliary rule module, the roasting process control scheme is implemented using an algorithm. The control quantity Δμ, representing material fluctuations, is written into the DCS and applied to the control belt speed of the roasting furnace. This cycle is repeated to maintain the standard temperature inside the roasting furnace near the set value. From 9:30 AM to 9:30 PM on August 14, 2022, a segment of the real-time automatic control effect was captured at the industrial site. Figure 4 During the 12-hour period shown in the figure, the temperature remained stable within ±15℃ for the entire time period, and within ±10℃ for 98.3% of the time. During this period, the operating temperature exceeded ±10℃ twice, with the longest single exceedance time being 12 minutes (the cause of this overshoot was manual operation inside the furnace). The operator only performed manual intervention once a day, and the automatic control connection time was over 99%.

[0119] The automated control method, device and system of the present invention can be used to carry out intelligent transformation and upgrading of the roasting furnace.

[0120] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. An automatic temperature control method for zinc smelting roasting process, characterized in that, The roasting furnace includes a roasting furnace body, a belt speed sensor installed on the roasting furnace belt to obtain the speed of the feed belt, a temperature sensor installed on the roasting furnace to obtain the standard temperature of the boiling layer of the roasting furnace, and a feed belt scale installed on the roasting furnace belt. The data collected by the belt speed sensor and the temperature sensor are directly transmitted to the DCS system, and the control commands are issued to the roasting furnace by the DCS system. The automatic temperature control method for the zinc smelting roasting process includes: The OPC communication protocol is used to communicate between the calcining furnace DCS system and the industrial control computer; Historical data on the operation of the roasting furnace are collected and stored in a database; the historical data includes feed belt speed, standard temperature, and feed rate. The historical data is preprocessed; Based on the preprocessed data, the system model is identified using the subspace method, and an optimized controller for the model predictive control algorithm is designed. The control quantity Δµ is calculated based on the identified predictive model, where Δµ is the change in belt speed. The system model identification uses belt speed as the control input and standard temperature as the predictive output. Determine whether the prediction model is suitable for the current operating conditions. If it is suitable, the control quantity Δµ is directly written into the DCS and applied to the roasting furnace. If it is not suitable, auxiliary rules are used to correct the control quantity Δµ, and the corrected control quantity Δµ is written into the DCS and applied to the roasting furnace. The auxiliary rules include at least material fluctuation rules, airflow fluctuation rules, abnormal handling rules, and long-term temperature difference handling rules; This cycle is repeated to maintain a stable standard temperature throughout the entire roasting process in the roasting furnace.

2. The automatic temperature stabilization control method for zinc smelting roasting process according to claim 1, characterized in that, The data preprocessing method is as follows: Differential operation: Perform differential operation on belt speed and standard temperature. If a large number of zero data points are found, perform secondary sampling and differential operation to obtain the corresponding change value. Standardization: Note To model the overall input and output dataset, standardization is performed. in Let N be a column vector with a length equal to the number of samples in the dataset. The mean of the column vectors is calculated and then expanded to the original column vector size, denoted as N. The function sum() sums the elements of a column vector. Standard deviation; Filtering preprocessing: Let the size of the forward sliding window be... , For a given row vector in the S dataset matrix 。 3. The automatic temperature stabilization control method for zinc smelting roasting process according to claim 1, characterized in that, The specific steps for identifying the system model using the subspace method are as follows; Based on the output and input vectors, considering k past and future moments, let... It is the input observable variable data. These are the given modeling output variable data; in , , , , , and The dimensions are the control variable, the output variable, and the state variable, respectively. N is obtained based on the dataset size and the hyperparameter k, i.e., N = size - 2 * k + 1. From the input-output matrix, we can define the past and future system states: QR decomposition can transform the input-output matrix into the following form: The R matrix on the left is a lower triangular matrix, and the Q matrix on the right is an orthogonal matrix. , , , , Observing the third column of the R matrix, we can see that the past input and output were zero, and the future input is also zero. Therefore, we can conclude that the future output will also be zero. The value is 0; the relationship between the future output and the known past input / output and future input is obtained from the QR decomposition: remember , ,get The subspace predictor coefficient matrix , Then, the obtained subspace predictor model is used as the prediction model and fed into the model predictive control algorithm. The model predictive control algorithm is as follows: The optimization problem in the controller can be constructed in the following form: In the formula, To predict the time domain, To control the time domain, generally Select , , And P≥0 is positive semi-definite, R>0 is positive definite. For the set control objectives, and These are the upper and lower limits of the control quantity. and These are the upper and lower limits of the increase / decrease overlimit constraint. The increase / decrease overlimit constraint refers to the cumulative control change over a certain period not exceeding a certain range. Directly solving this optimization problem has high complexity, so we introduce... , Based on subspace predictor (7) and auxiliary variables and The objective function of the model predictive control algorithm is transformed into the following form: Model Predictive Control Algorithm The optimization problem of the model predictive control algorithm is transformed into a quadratic programming (QP) optimization problem, and the active set method is adopted.

4. A temperature stabilization control device for zinc smelting roasting process, characterized in that, The device includes: The data acquisition module is used to collect historical data of the roasting furnace operation and store it in the database, and read real-time operation data at 10-second intervals and store it in the historical data. The real-time operation data includes at least the feed belt speed, standard temperature, and feed rate. The data preprocessing module is used to resample, differentially process, standardize, and filter the belt speed data and standard temperature data to obtain the corresponding changes. The model predictive control module is used to identify the predictive model based on the preprocessed data using the subspace method, and to adapt it to the roasting process at each moment by updating the model. Based on the identified predictive model, the change in belt speed Δμ is calculated. If the predictive model is suitable, it is directly written into the DCS system through the OPC communication protocol and applied to the roasting furnace. Auxiliary rule correction module: used to correct the belt speed change Δµ calculated by the model prediction control module based on preset auxiliary rules. The auxiliary rules include at least correction rules based on material fluctuations, correction rules based on air fluctuations, anomaly handling rules, and long-term temperature difference handling rules.

5. A temperature stabilization control system for zinc smelting roasting process, characterized in that, The invention includes a calcining furnace DCS system and an industrial control computer, and a temperature stabilization control device for the zinc smelting calcining process as described in claim 4. The calcining furnace DCS system and industrial control computer communicate with the zinc smelting calcining process temperature stabilization control device via the OPC communication protocol. The calcining furnace includes a calcining furnace body, a belt speed sensor installed on the calcining furnace belt to obtain the speed of the feed belt, a temperature sensor installed on the calcining furnace to obtain the standard temperature of the boiling layer, and a feed belt scale installed on the calcining furnace belt. The data collected by the belt speed sensor and the temperature sensor are directly transmitted to the DCS system, and the control commands are issued to the calcining furnace by the DCS system.

6. A zinc smelting roasting furnace, characterized in that, The invention includes a temperature stabilization control device for a zinc smelting roasting process as described in claim 4, wherein the roasting furnace is communicatively connected to the temperature stabilization control device for a zinc smelting roasting process.

Images