Multi-time-scale intelligent control method for steel production equipment

Abstract

The invention discloses a multi-time-scale intelligent control method for steel production equipment, and belongs to the technical field of intelligent control in the steel production process. Comprising the steps that historical multi-modal data are collected and preprocessed, and multi-modal time sequence data are formed; constructing an equipment digital twinborn body fusing the physical model and the data driving model, predicting equipment response, and training the equipment digital twinborn body by adopting multi-modal time sequence data; collecting real-time multi-modal data, preprocessing the multi-modal data, and inputting the preprocessed multi-modal data into the trained equipment digital twinborn body to obtain a predicted value of equipment response; performing multi-time scale layered optimization control on control parameters through an improved multi-target particle swarm optimization algorithm and a constrained Bayesian optimization algorithm based on the predicted value of the equipment response; and based on the optimization control result, adjusting the control parameters of the equipment through a PLC execution mechanism, and feeding back the adjusted equipment state variable to a sensor to realize closed-loop control.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent control in the iron and steel production process, and relates to a multi-time scale intelligent control method for iron and steel production equipment. Background Art

[0002] In the iron and steel production process, the operating state and maintenance requirements of equipment change over time. Traditional equipment control methods are difficult to adapt to the degradation characteristics of equipment and complex production conditions. Digital twin technology can accurately simulate the physical state of equipment and establish a prediction model based on a large amount of data. However, existing methods lack flexible control strategies to cope with the dynamically changing production environment. The model-based intelligent optimization method can achieve dynamic adjustment of equipment parameters, but it fails to effectively combine multi-dimensional degradation modeling with cross-production line knowledge sharing. Therefore, an intelligent control method integrating digital twin, intelligent optimization, and adaptive control is needed to improve the accuracy and stability of equipment control. Summary of the Invention

[0003] To solve the above technical problems, the object of the present invention is to provide a multi-time scale intelligent control method for iron and steel production equipment, which can realize the autonomous perception, accurate prediction, and hierarchical optimization of key control parameters of the equipment, thereby improving the control accuracy and operation stability.

[0004] The present invention provides a multi-time scale intelligent control method for iron and steel production equipment, including:

[0005] Step 1: Collect historical multi-modal data and perform preprocessing to form multi-modal time series data;

[0006] Step 2: Build a device digital twin integrating a physical model and a data-driven model, predict the device response, and train the device digital twin with multi-modal time series data;

[0007] Step 3: Collect real-time multi-modal data, input it into the trained device digital twin after preprocessing, and obtain the predicted value of the device response;

[0008] Step 4: Based on the predicted value of the device response, perform multi-time scale hierarchical optimization control on the control parameters through an improved multi-objective particle swarm optimization algorithm and a constrained Bayesian optimization algorithm;

[0009] Step 5: Based on the optimized control result of Step 4, adjust the control parameters of the device through a PLC actuator, and feedback the adjusted device state variables to the sensor to achieve closed-loop control.

[0010] The multi-time scale intelligent control method for iron and steel production equipment of the present invention has the following beneficial effects:

[0011] The control method of the present invention constructs a device digital twin integrating a physical model and LSTM based on multi-modal data to improve prediction accuracy; by introducing a dual-time-scale optimization algorithm, it realizes the integration of long-term and short-term dynamic control; solves the problems of slow response and difficult adjustment in the traditional control system of steel production equipment, and can dynamically optimize equipment control parameters according to real-time changes during the production process. In particular, by combining digital twin technology, optimization algorithms and multi-modal perception, it enhances the adaptive ability, flexibility and real-time response ability of the system under complex production conditions; it has strong portability and is applicable to the adaptive operation control of steel manufacturing equipment. Brief Description of the Drawings

[0012] Figure 1 It is a flowchart of a multi-time-scale intelligent control method for a steel production equipment of the present invention. Detailed Embodiment

[0013] The following combines the drawings and embodiments to further describe in detail the specific embodiments of the present invention. In this embodiment, the equipment object is the F1-F7 stands of the 2250mm hot continuous rolling finishing mill.

[0014] As Figure 1 shown, a multi-time-scale intelligent control method for a steel production equipment of the present invention includes:

[0015] Step 1: Collect historical multi-modal data and perform preprocessing to form multi-modal time series data. Specifically:

[0016] Step 1.1: Deploy a multi-modal sensor array at key parts of the target steel production equipment to collect historical state variables, including: the temperature inside the heating furnace in the heating furnace process; the rolling force and bending roll force in the rough rolling process; the rolling force, bending roll force, inlet speed, outlet speed, strip inlet temperature, strip outlet temperature, stand vibration acceleration, inter-stand tension and cooling water flow of all stands in the hot rolling process; and the strip inlet thickness and strip outlet thickness of each process.

[0017] In this embodiment, the sensor deployment form is: each infrared temperature measurement array consists of 10 infrared sensors, which respectively monitor the temperatures of the strip head, middle and tail, and the sampling period is 50ms; the pressure sensors are installed on the rolling mill hydraulic cylinders, and the sampling period is 50ms; the accelerometers are arranged on the drive side and the operation side of the rolling mill, and the sampling period is 50ms; the X-ray thickness gauge, and the sampling period is 50ms. Some data are shown in Table 1:

[0018] Table 1: Some Sampling Data

[0019] Step 1.2: Each sensor is connected to a unified data acquisition system, and the historical state variables are subjected to data cleaning to remove outliers, denoising, and normalization processing to ensure the consistency and reliability of the multi-modal time series data.

[0020] Step 1.3: Construct multi-modal time series data based on the preprocessed historical state variables:

[0021]

[0022] Among them, represents the sampling value of the i-th state variable at time t.

[0023] Step 2: Construct a device digital twin that integrates a physical model and a data-driven model to predict the device response, and use the multi-modal time series data to train the device digital twin.

[0024] Specifically, the device response includes the strip exit temperature and the strip exit thickness. The physical model reflects the actual physical properties of the device, and the data-driven model supplements the state prediction of the device. The specific steps of Step 2 are as follows:

[0025] Step 2.1: Establish a physical model, and based on the multi-modal time series data, use the least squares fitting method to determine the static weight coefficients of the model:

[0026]

[0027]

[0028] Among them, represents the predicted value of the physical model, represents the weighted sum of , is the static weight coefficient corresponding to , which is obtained by fitting the historical multi-modal data through the least squares method; represents the weighted sum considering the change rate, is the dynamic weight coefficient, represents the actual value of the device response under a specific state, that is, the actual value of the strip exit temperature and the actual value of the strip exit thickness.

[0029] Step 2.2: Establish an LSTM network as the data-driven model, and use the residual between the predicted value of the physical model and the actual value of the device response under a specific state as the input for training the LSTM data-driven model, and the output is the residual compensation value .

[0030] Step 2.3: Construct a digital twin of the device through a physical model and an LSTM network. The LSTM network compensates for the errors of the physical model under different working conditions by learning historical residual sequences, making the prediction results of the device response output by the digital twin of the device closer to the actual value of the device response. The prediction results are expressed as 。

[0031] The trained digital twin of the device is used to predict the device response during the real-time operation phase. The prediction results will be used as inputs for the subsequent improved multi-objective particle swarm optimization algorithm (IMOPSO) and constrained Bayesian optimization algorithm (CBO) to optimize the control parameters of the device and ensure the optimal operating state of the device.

[0032] In this embodiment, to predict the strip steel outlet temperature, the following physical model is constructed:

[0033] At this time, the static weights , , are obtained by fitting 3,500 groups of historical data (3,000 groups for the training set and 500 groups for the test set) through the least squares method; the dynamic weights are , , 。

[0034] The data-driven model uses an LSTM network to compensate for the difference predicted by the physical model. The input layer is the temperature prediction residual sequence of the past 60 time steps ; the hidden layer is a stacked LSTM (64 units × 2 layers), the activation function is tanh, and the Dropout rate is 0.2; the output layer is the residual compensation value; the learning rate is 0.001, the training period is 200, and the loss function is MSE.

[0035] Step 3: Collect real-time multi-modal data, preprocess it, and input it into the trained digital twin of the device to obtain the predicted value of the device response.

[0036] Step 4: Based on the predicted value of the device response, perform multi-time scale hierarchical optimization control on the control parameters through the improved multi-objective particle swarm optimization algorithm and the constrained Bayesian optimization algorithm, specifically:

[0037] Step 4.1: Obtain the predicted value of the device response with hours as the large cycle, and perform hourly periodic global search optimization on the control parameter set through the improved multi-objective particle swarm optimization algorithm. The control parameters in the control parameter set constitute the particles in the improved multi-objective particle swarm optimization algorithm, specifically:

[0038] Step 4.1.1: Initialize the particle swarm, and each particle represents a set of control parameters , randomly generate particle positions represents the mth particle Position in dimension n; particle velocity express The particle speed is initialized to 10% of the control parameter range.

[0039] In specific implementation, the control parameter set P represents a subset combination selected from multimodal data collected in real time that has a positive regulatory effect on equipment operation and is adjustable; such as the temperature setting of the heating furnace, the rolling speed setting, and the cooling water flow setting.

[0040] In this embodiment, the control parameters are set as:

[0041] .

[0042] in, , The value range is constrained by security Sure.

[0043] Step 4.1.2: Construct the objective function of the improved multi-objective particle swarm optimization algorithm:

[0044]

[0045] Among them, the objective function represents the prediction result of the device digital twin With actual value The error is minimal.

[0046] Step 4.1.3: Determine whether the particle is in the global optimal position by calculating the fitness function value. If it is not the global optimal position, update the particle speed and particle position according to the following formula:

[0047]

[0048]

[0049] in, Represents the static weight coefficient of the input variable; Represents the inertia weight attenuation coefficient, which is used to balance global search and local search; and represents two random factors, , controlling the balance of the search; Represents particles The best historical position in dimension n; Represents the current global best position.

[0050] In this embodiment, the particle swarm contains 100 particles, and the initial velocity is set to 10% of the search range; the initial inertia weight is set ; the attenuation coefficient , which satisfies ; , . After 40 generations of iteration, the optimal solution is an outlet velocity of 14.2 m / s, a reduction amount of 23.5%, and a cooling water flow rate of 2000 L / min.

[0051] Step 4.1.4: Re-obtain the predicted value of the equipment response based on the equipment digital twin to ensure that the optimization search direction is always consistent with the actual physical response of the equipment.

[0052] Step 4.1.5: According to the set termination conditions, determine whether to terminate the optimization process. If the termination conditions are met, output the global best position of the particles as the optimal control parameters for the next major cycle; if the termination conditions are not met, return to Step 4.1.2 to continue the iteration.

[0053] In this embodiment, to achieve the dynamic adaptive adjustment of the control parameters in the long cycle and ensure that the system operates with high performance under multiple working conditions, an objective function of the improved multi-objective particle swarm optimization algorithm is constructed based on the prediction results output by the equipment digital twin, and this objective function is used as the fitness evaluation index to guide the improved multi-objective particle swarm optimization algorithm to perform global optimization of the particle swarm. This process uses the equipment digital twin as a predictor of the equipment behavior, running through the fitness calculation and update process of the particles.

[0054] Step 4.2: Obtain the predicted value of the equipment response in minutes as a small cycle, and use the constrained Bayesian optimization algorithm to perform minute-level local optimization on the control parameter set , specifically:

[0055] Step 4.2.1: Construct a surrogate model of the control objective function :

[0056]

[0057] where and respectively represent the mean and variance of the predicted results of the equipment response corresponding to the control parameters in the control parameter set P when fluctuations occur of

[0058] Step 4.2.2: Establish the following expected improvement expression, where the expected improvement measures the improvement potential of the current control parameters relative to the historical optimal control parameter set:

[0059]

[0060] in, Represents the historical minimum control objective function value; represents the probability density function of the standard normal distribution, Represents the cumulative distribution function of the standard normal distribution.

[0061] Step 4.2.3: In the device safety area In, get The corresponding control parameters when reaching the maximum , as the optimal control parameter for the next small cycle:

[0062]

[0063] in, , p min Indicates the minimum setting value of the control parameter at which fluctuation occurs, p max Indicates the maximum set value of the control parameter at which fluctuations occur.

[0064] In specific implementation, when the control parameters fluctuate, the above steps will be repeated every minute or shorter time interval to continuously optimize the control parameters through the constrained Bayesian optimization algorithm.

[0065] In this example, the control parameters fluctuated. Specifically, when the cooling water flow rate was originally 2000 L / min and the cooling water pressure increased by 0.5 bar, this affected the cooling water flow rate's control over the finishing temperature. The predicted value of the equipment response after the fluctuation, i.e., the predicted strip outlet temperature, was obtained through the equipment digital twin. The following proxy model was established:

[0066]

[0067] In the safe zone The optimal cooling water flow rate is obtained by maximizing the EI criterion:

[0068]

[0069] in, , numerically solved to get .

[0070] After optimization using the constrained Bayesian optimization algorithm, the optimal solution is an outlet velocity of 14.2 m / s, a reduction of 23.5%, and a cooling water flow rate of 1980 L / min.

[0071] Step 5: Based on the optimized control results of step 4, the control parameters of the equipment are adjusted through the PLC actuator, and the adjusted equipment state variables are fed back to the sensor to achieve closed-loop control.

[0072] The above are only the preferred embodiments of the present invention and are not intended to limit the idea of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A multi-time-scale intelligent control method for steel production equipment, characterized in that Including: Step 1: Collect historical multi-modal data and preprocess it to form multi-modal time series data; Step 2: Build a device digital twin that integrates a physical model and a data-driven model to predict the device response, and use the multi-modal time series data to train the device digital twin; Step 3: Collect real-time multi-modal data, preprocess it, and input it into the trained device digital twin to obtain the predicted value of the device response; Step 4: Based on the predicted value of the device response, perform multi-time scale hierarchical optimal control on the control parameters through an improved multi-objective particle swarm optimization algorithm and a constrained Bayesian optimization algorithm; Step 5: Based on the optimal control result of Step 4, adjust the control parameters of the device through a PLC actuator, and feedback the adjusted device state variables to the sensor to achieve closed-loop control.

2. The multi-time scale intelligent control method for steel production equipment according to claim 1, characterized in that The specific content of Step 1 is as follows: Step 1.1: Deploy a multi-modal sensor array on the target steel production equipment to collect historical state variables, including: the temperature inside the heating furnace in the heating furnace process; the rolling force and the work roll bending force in the rough rolling process; the rolling force, the work roll bending force, the inlet speed, the outlet speed, the strip inlet temperature, the strip outlet temperature, the rack vibration acceleration, the inter-stand tension, and the cooling water flow of all stands in the hot rolling process; as well as the strip inlet thickness and the strip outlet thickness of each process; Step 1.2: Preprocess the historical state variables. The preprocessing includes data cleaning, denoising, and normalization to ensure the integrity and accuracy of the data; Step 1.3: Form multi-modal time series data according to the preprocessed historical state variables; Among them, represents the sampled value of the i-th state variable at time t.

3. The multi-time scale intelligent control method for steel production equipment according to claim 2, wherein The device response includes the strip outlet temperature and the strip outlet thickness. The specific content of Step 2 is as follows: Step 2.1: Establish a physical model, and based on the multi-modal time series data, use the least squares fitting method to determine the static weight coefficient of the model; in, represents the predicted value of the physical model, express The weighted sum of It corresponds to The static weight coefficient of is obtained by fitting historical multimodal data using the least squares method; Express consideration The weighted sum of the rates of change, is the dynamic weight coefficient, Indicates the actual value of the equipment response, that is, the actual value of the strip outlet temperature and the actual value of the strip outlet thickness; Step 2.2: Establish an LSTM network as a data-driven model, and use the predicted value of the physical model and the actual value of the device response The residual between is used as the input for training the LSTM network, and the output is the residual compensation value ; Step 2.3: Construct a device digital twin through a physical model and an LSTM network. The LSTM network compensates for the errors of the physical model under different working conditions by learning historical residual sequences. The predicted result of the device response output by the device digital twin is expressed as .

4. The multi-time scale intelligent control method for steel production equipment according to claim 2, wherein The specific content of Step 4 is as follows: Step 4.1: Obtain the predicted values of the device response with hours as the large cycle, and perform global search and optimization on the control parameter set at the hourly level through the improved multi-objective particle swarm optimization algorithm. The control parameters in the control parameter set form the particles in the improved multi-objective particle swarm optimization algorithm. ​ Step 4.2: Obtain the predicted value of the device response with minutes as the small cycle, and use the constrained Bayesian optimization algorithm to perform minute-level local optimization on the control parameter set for minute-level local optimization.

5. The multi-time-scale intelligent control method for steel production equipment according to claim 4, characterized in that The specific content of Step 4.1 is as follows: Step 4.1.1: Initialize the particle swarm, where each particle represents a set of control parameters , randomly generate the particle positions represents the position of the m-th particle in dimension n; the particle velocity represents the adjustment value of, and the particle velocity is initialized to 10% of the value range of the control parameters; Step 4.1.2: Build the objective function of the improved multi-objective particle swarm optimization algorithm; Among them, the objective function represents the prediction result of the device digital twin and the actual value with the minimum error; Step 4.1.3: Determine whether the particle is in the global optimal position by calculating the fitness function value. If it is not in the global optimal position, update the particle velocity and the particle position according to the following formula: Among them, represents the static weight coefficient of the input variable; represents the inertia weight decay coefficient, which is used to balance global search and local search; and represent two random factors, , controlling the balance of the search; represents the particle 's historical best position in dimension n; represents the current global best position; Step 4.1.4: Re-obtain the predicted value of the device response based on the device digital twin to ensure that the optimization search direction is always consistent with the actual physical response of the device; Step 4.1.5: According to the set termination condition, determine whether to terminate the optimization process. If the termination condition is met, output the global best position of the particle as the optimal control parameter for the next major cycle; if the termination condition is not met, return to Step 4.1.2 to continue the iteration.

6. The multi-time scale intelligent control method for the steel production equipment according to claim 4, characterized in that The specific content of Step 4.2 is as follows: Step 4.2.1: Construct the surrogate model of the control objective function : Among them, and respectively represent the mean and variance of the predicted results of the device responses corresponding to the control parameters in the control parameter set P when fluctuations occur. ​ Step 4.2.2: Establish the following expected improvement expression: Among them, represents the historical minimum control objective function value; represents the probability density function of the standard normal distribution, represents the cumulative distribution function of the standard normal distribution; Step 4.2.3: In the device safety area obtain the control parameter corresponding to the maximum value of as the optimal control parameter for the next small cycle: ​ Among them, , p min represents the minimum set value of the control parameter with fluctuations, and p max represents the maximum set value of the control parameter with fluctuations.

Images