Model prediction and temperature control method for reaction calorimetry system based on online correction of mechanism parameters

Abstract

The present invention discloses a reaction calorimetry system model prediction and temperature control method based on online correction of mechanism parameters. The present invention establishes an equivalent mechanism parameter model of the reaction calorimetry system based on heat transfer analysis for predicting output calculations. The general parameter calibration process of the reaction calorimeter is used to synchronously update the initial values ​​of the model parameters, and the model parameters are adjusted online in the process of model prediction and temperature control in combination with real-time parameter relationship comparison and other logic. The temperature control stability of the present invention is significantly improved, overshoot is effectively suppressed, and the isothermal control convergence speed is increased. Under the action of the online correction of the model parameters, the adaptability to a wide range of working conditions of the reaction system is improved, the influence of serious mismatch of the prediction model is reduced, and a reference change trend of the mechanism parameters can be given after the experiment is completed. Combined with the online parameter calibration process of the reaction calorimetry system, the initial model parameters can be automatically updated without the need for complex or time-consuming system parameter identification.

Description

Technical Field

[0001] The invention belongs to the technical field of fine chemical reaction safety testing, and relates to a reaction calorimetry system model prediction and temperature control method based on online correction of mechanism parameters. Background Art

[0002] Controlled object - reaction calorimeter [1-2] A reaction calorimeter is an automated instrument that simulates semi-batch reaction processes in a kettle under laboratory conditions. During the process steps, it measures changes in the sample temperature and feed mass in real time, calculates reaction heat release, and predicts temperature changes. This information provides a basis for reaction safety risk assessment, process development, and optimization, making it a key scientific instrument in the fields of fine chemical reaction safety assessment and automated chemistry research. Single-step or continuously variable isothermal control are the fundamental control functions required by reaction calorimeters to achieve their measurement objectives.

[0003] The principle structure of the reaction calorimeter is as follows Figure 1 As shown: It mainly includes a reactor 4 with a circulating oil jacket, temperature sensors 9, 10, 11, a circulating oil bath system 12, a correction heater 2 (driven by a heater power driver 5), a stirrer 1, a stirring motor 6 and a monitoring system 8. The principle of the system's feedback control of the temperature of the reactant sample 3 in the reactor is as follows: the monitoring system 8 continuously monitors the temperature of the sensors 9, 10, 11 through the signal communication server 7, and adjusts the set temperature of the circulating oil bath thermostat 12 in real time according to the output of the control algorithm. The circulating oil bath thermostat 12 heats or cools the oil according to the instructions, and drives the temperature-controlled oil to circulate in the oil pipeline 13 and the reactor jacket, so that the oil and the sample transfer heat through the jacket wall to achieve sample temperature regulation. The block diagram of the temperature control system is shown as follows Figure 2 shown.

[0004] The temperature control of the reaction calorimetry system faces the following bottlenecks:

[0005] First, this system is a typical hysteresis system. Calorimetric reactors typically use volumes ranging from several hundred mL to several liters. The sample itself has a large heat capacity, and the circulating oil also has a large thermal mass. The oil bath system controls the sample temperature through a series of processes, including oil temperature fluctuations, oil flow, heat exchange between the jacket wall, and sample temperature changes. This results in low overall heat transfer efficiency. Consequently, the system's temperature control response exhibits significant hysteresis, prone to overshoot, and poor control speed and stability.

[0006] Second, when the system structure and hardware performance remain unchanged, temperature control performance depends solely on the temperature control algorithm in the monitoring system. Optimizing the algorithm under certain operating conditions is not difficult—for example, parameter tuning of the PID algorithm. However, the sample properties in a reaction calorimetry system are often complex and uncertain, and the thermodynamic properties of different reaction systems vary greatly, making it difficult for methods like PID to adapt parameters to meet requirements. Finally, even within the same measurement experiment, reaction calorimetry systems often require large temperature changes or continuous multi-step isothermal control based on process requirements. Large temperature changes can also significantly alter the heat transfer coefficient and heat capacity of the sample system, placing high demands on the adaptability of the temperature control method.

[0007] Model predictive control is an advanced control method that has been successfully applied to complex industrial processes such as petroleum, chemical, metallurgy and electric power. [3-5] , its basic feature is that the control process includes three links: model prediction, finite time domain rolling optimization and feedback correction. Compared with the "after-the-fact adjustment" algorithms such as PID, the model prediction idea has inherent advantages for the control of hysteresis systems. The optimization process of its control output is repeatedly performed online within a limited moving time, and is suitable for complex systems with changing dynamic characteristics. At the same time, due to the existence of the feedback correction process, the algorithm has strong robustness and can overcome the uncertainty caused by factors such as system nonlinearity, time variation, and model mismatch to a large extent. Therefore, the present invention adopts this algorithm as the basic structure of the method. Nevertheless, the problem with its application is that, as mentioned above, the properties of the controlled object-sample will change significantly in experiments of different reaction systems or in large temperature changes in the same experiment. The output of the model used for prediction may have a large error with the actual process output, that is, serious model mismatch, which will lead to deterioration of the control effect, so the calibration and online correction of the prediction model parameters are very important.

[0008] Regarding the online calibration of system characteristics, one of the basic links of mature reaction calorimetry technology is the parameter calibration method before and after the reaction. [6-8] Taking the general measurement method as an example, in the stage where no chemical reaction occurs, the sample is stepped through the reactor jacket to obtain two thermal equilibrium states. In each thermal equilibrium state, the calibration heater outputs a fixed power to measure the heat transfer factor, and the system heat capacity is calculated using the heating process. The temperature and power curves of the general measurement method are as follows: Figure 3 As shown. In this process, assuming that the change of heat loss is not considered, taking the temperature equilibrium state of the target temperature 1 as the benchmark, the incremental heat balance equation of the system is:

[0009] P H =Q accum +Q f (1)

[0010] Among them, P HTo correct for changes in heater power, Qf is the change in heat flow from the sample to the jacket, and the heat transfer factor (UA) is related to Q f The relationship can be expressed by the following formula:

[0011] Q f =(UA)(ΔT r -ΔT j ) (2)

[0012] Where, ΔT r and ΔT j are the changes of sample and jacket oil temperature relative to the baseline. accum is the total heat capacity of the reaction system (C p M) Heat accumulation with temperature:

[0013] Q accum =(C p M)·dT r / dt (3)

[0014] Here, t represents time.

[0015] exist Figure 3 In the 0-1 and 4-5 stages, if the correction heater outputs a fixed power, the control system keeps the temperature of sample 3 constant by lowering the jacket oil temperature. The process begins with the heat accumulation Q accum =0, so (UA) is calculated by the following formula (taking the 0-1 stage as an example):

[0016]

[0017] During the temperature rise process of stage 2-3, the correction heater is turned off. The energy required to raise the sample from target temperature 1 to target temperature 2 is obtained by heat transfer from the jacket circulating oil to the sample. Therefore, the heat capacity of the system can be calculated by the following formula:

[0018]

[0019] The change of (UA) is obtained by interpolating the measured values ​​in the 0-1 and 4-5 stages.

[0020] References

[0021] [1] Benoît Zuffrei, Francis Stossel, Urs Grote. Methods for simulating process plants at laboratory scale: , 2007.

[0022] [2]Jacobsen J P.Reaction calorimeter-a useful tool in chemical engineering[J].Thermochimica Acta, 1990, 160(1):13-23.

[0023] [3] Chen Bensong. Predictive functional control based on modern industrial process control [J]. Journal of Hunan University of Arts and Science: Natural Science Edition, 2007, 19(2): 3.

[0024] [4] Xu Zuhua. Research on Model Predictive Control Theory and Applications[D]. Zhejiang University, 2004.

[0025] [5]Xu VV; Zhang J, Zhang R.Application of multi-modelswitchingpredictive functional control on the temperaturesystem of an electric heatingfurnace[J].lsa Trans, 2017: 287-292.

[0026] [6]Singh J. Reaction calorimetry for process development: Recentadvances[J]. Process Safety Progress, 2010, 16(1): 43-49.

[0027] [7] Schildknecht J. Reaction calorimeter for applications in chemical process industries: Performance and calibration [J]. Thermochimica Acta, 1981, 49(1): 87-100.

[0028] [8] Zhuang Zhong. Study on the hazard of thermal runaway in acetic anhydride hydrolysis reaction[D]. Nanjing University of Science and Technology, 2009. Summary of the Invention

[0029] The purpose of the present invention is to provide a reaction calorimetry system model prediction and temperature control method based on online correction of mechanism parameters.

[0030] The technical solution of the present invention is:

[0031] (1) Equivalent mechanism parameter model output

[0032] The equivalent mechanism parameter model is used for prediction output, and the system is simplified to a first-order inertial system, with the internal temperature of the oil bath T o is the input, the sample temperature T r For output, assume the equivalent heat transfer factor (UA)e , the equivalent thermal inertia is (C p M) e And the equivalent heat dissipation coefficient is α, with T s is the sampling time, and the discretized model output of the future P step is expressed as follows:

[0033]

[0034] Where k represents the current sampling moment, and factors A and B are functions of the equivalent mechanism parameters:

[0035]

[0036] (2) Actual output delay estimation correction and predicted output deviation correction

[0037] Considering the existence of a pure lag link in the system with a delay time of τ, the actual sample temperature is estimated and corrected:

[0038] T r_pav (k) = T r (k)+T r_m (k+τ / T s )-T r_m (k) (8)

[0039] Calculate the deviation between the sample temperature after delay estimation correction and the current model prediction output value:

[0040] e(k)=T r_pav (k)-T r_m (k) (9)

[0041] (3) Calculation of optimal control quantity

[0042] Considering the optimization of the control objective in the future P step, the optimal control quantity at the current moment is:

[0043]

[0044] Among them, T traj represents the sample temperature reference trajectory.

[0045] (4) Model parameter initialization and online correction

[0046] Before each chemical reaction occurs, the calibration process is performed to initialize the parameters online to obtain α0, (UA) e_0 and (C p M) e_0 Initial values ​​of three parameters and model mechanism parameters (UA) are implemented in each step of isothermal control during the experiment e 、(C p M) e, α and delay time τ online correction, with (UA) e For example, the parameter adjustment method is:

[0047] (UA) e =(UA) e_0 +∫d(UA) e (11)

[0048] Furthermore, in order to suppress the impact of model mismatch, an online parameter adjustment strategy is designed. In each sampling control cycle, the parameter α is adjusted as follows:

[0049] dα=-k1r'(T r -T r_m ) / |T r | (12)

[0050] Where k1 is the proportional coefficient and r' represents the adjustment speed factor related to the control process stage. Parameter (UA) e and (C p M) e The adjustment amount is:

[0051]

[0052] Where subscript m represents the estimated value of the parameter relationship, and subscript s represents the set value of the relationship. k2 and k3 are proportional coefficients. The adjustment amount of the delay time τ is:

[0053]

[0054] Where k4 is the proportional coefficient, sign is the sign judgment function, T r_set is the target temperature, T ini is the initial temperature of the sample, T r_pav It is the actual temperature after delay estimation correction, and t represents time.

[0055] Furthermore, at the beginning of each reaction experiment, the universal calibration process of the reaction calorimetry system is used to calculate the equivalent parameters and initialize the model parameters. The curve of the calibration process is as follows: Figure 3 As shown, the following calculations are completed synchronously during this process:

[0056]

[0057] α=(UA) e (T o -T r ) / T r (17)

[0058] Where P H To correct the heater power, ΔT r and ΔTo are the relative changes of sample and oil bath temperature respectively. The above calculation results are used as α0, (UA) e_0 and (C p M) e_0 Update the initial parameter values ​​and set the model benchmark operating point to the process temperature during calibration.

[0059] The present invention has the following beneficial effects: The present invention utilizes a model predictive control method, significantly improving the smoothness and rapidity of temperature control. The model based on mechanism parameters is easy to understand, and the proposed online parameter adjustment logic is simple, avoiding complex parameter identification. Combined with the universal calibration process for reaction calorimetry systems, it can quickly achieve online initialization of model parameters for different experimental systems, significantly reducing the impact of model mismatch. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is the principle structure of the reaction calorimetry system;

[0061] 1. Agitator, 2. Calibration heater, 3. Reactant sample, 4. Jacketed calorimetric reactor, 5. Heater power driver, 6. Stirring motor, 7. Signal communication server, 8. Monitoring system, 9. Sample temperature sensor, 10. Jacket oil temperature sensor, 11. Oil bath temperature sensor, 12. Circulating oil bath thermostat, 13. Oil pipeline connecting the oil bath and reactor jacket.

[0062] Figure 2 This is the block diagram of the reactor oil bath temperature control system;

[0063] Figure 3 It is the temperature and power curve of the universal parameter calibration method;

[0064] Figure 4 It is a model prediction temperature control method based on online correction of mechanism parameters;

[0065] Figure 5 Example of temperature control effect during heating process;

[0066] Figure 6 Example of model parameter initialization and online correction effect. DETAILED DESCRIPTION

[0067] The present invention will be further described below with reference to the accompanying drawings.

[0068] 1. Equivalent mechanism parameter model prediction

[0069] The heat transfer mechanism of the reaction calorimetry system is simplified to the greatest extent to establish an equivalent model for predictive control: the participation of the correction heater, the exothermic reaction of the chemical reaction and other control processes are not considered, and only the heat exchange between the oil bath system and the sample through the reactor jacket is focused on; the working process of the heating and cooling components inside the oil bath, the temperature change process of the oil, the flow of the oil, the heat conduction and convection on both sides of the reactor jacket and other link models are ignored. It is assumed that the controlled object is a first-order inertial system, and the internal temperature of the oil bath T is used as the reference. o is the input, the sample temperature T r is the output, and the equivalent heat transfer factor between the two is (UA) e , the equivalent thermal inertia is (C p M) e , considering the heat loss Q of the sample to the external environment through phase change convection, insert conduction and other links loss , and it is considered to be proportional to the sample temperature, and the equivalent heat dissipation coefficient is α, then the equivalent mechanism model is as follows:

[0070]

[0071] T s is the sampling time, and the system difference equation is obtained by discretization. The output of the discretized model in the future P step can be expressed as follows by mathematical induction:

[0072]

[0073] Where k represents the current sampling moment, factors A and B are functions of the equivalent mechanism parameters, and the relationship is organized as follows:

[0074]

[0075] In addition, considering the existence of a pure lag link in the system with a delay time of τ, the Smith prediction idea is used to perform a delay prediction correction on the actual sample temperature. That is, the following formula is used instead of the actual sample temperature for subsequent calculations:

[0076] T r_pav (k) = T r (k)+T r_m (k+τ / T s )-T r_m (k) (21)

[0077] The equivalent mechanism parameter model of the reaction calorimetry system shown in Equation (19) is applied to predictive control in the following way:

[0078] First, the predicted value is corrected according to the calculated error between the current sample temperature and the model output. The deviation is calculated as follows:

[0079] e(k)=T r_pav(k)-T r_m (k) (22)

[0080] The temperature reference trajectory for predictive control is calculated as follows:

[0081]

[0082] Among them, T ref Represents the reference trajectory time constant. Considering the optimization of the control target in the future P steps, the optimization target is:

[0083] J=min[T r_m (k+P)+e(k)-T traj (k+P)] 2 (twenty four)

[0084] Therefore, the optimal control at the current moment is:

[0085]

[0086] After the calibration process at the beginning of each experiment, the model parameters are initialized and updated, and the model mechanism parameters are modified online in each step of the isothermal control during the experiment. The complete method logic structure diagram is shown in the following figure. Figure 4 shown.

[0087] 2. Parameter online correction

[0088] In order to suppress the model mismatch effect caused by the difference between different systems or the change of working conditions in the reaction calorimetry system, an online parameter adjustment strategy is designed to adjust the heat transfer factor (UA) e 、Thermal inertia (C p M) e The three equivalent mechanism parameters and heat dissipation coefficient α and delay time τ are corrected. e For example, the parameter adjustment method is:

[0089] (UA) e =(UA) e_0 +∫d(UA) e (26)

[0090] Among them (UA) e_0 It is the initial value of the parameter. Its setting method is related to the corresponding reference working point and is described in Part 3 of the technical solution. e Indicates the parameter adjustment amount for each sampling period during the online control process. The calculation method for the sampling period adjustment amount of each parameter is as follows:

[0091] Adjustments to the equivalent mechanism parameters can still be combined with an understanding of the system's heat transfer. First, when the temperature control enters the steady-state and near-steady-state stages, the oil bath temperature tends to be stable. Equation (18) shows that the steady-state sample temperature is related to the heat dissipation coefficient and the heat transfer factor. Considering the heat dissipation coefficient as the primary influencing factor, the deviation between the model output and the actual value of the steady-state sample temperature can reflect the direction of the heat dissipation coefficient deviation correction. Therefore, the parameter α adjustment is set proportionally to it, and the stage of the single-step isothermal control process is considered:

[0092] dα=-k1r'(T r -T r_m ) / |T r | (27)

[0093] Where k1 is the proportional coefficient, r' represents the adjustment speed factor related to the control process stage. The closer the process is to steady state, the larger its value is. It can be expressed as follows:

[0094]

[0095] Where T r_set is the target temperature, T ini is the initial temperature of the sample in the temperature control step.

[0096] (UA) e and (C p M) e The correction is made by measuring the parameter relationship in the sliding time domain during the temperature control process. Assuming that the system equivalent parameters remain unchanged in a relatively short time domain, two consecutive short time windows can be taken in this time domain to perform sliding average on the input and output data. The sliding average value of the temperature in each time window still satisfies formula (18):

[0097]

[0098] where ΔT or_i represents the average temperature difference between the oil bath and the sample within the i-th group sliding window, T r_i and (dT r / dt) i are the average values ​​of the sample temperature and its rate in the i-th group sliding window. Assuming that the equivalent parameters remain unchanged, the two equations can be obtained (UA) e and (C p M) e Estimated relationship with α:

[0099]

[0100] There is a real-time deviation between the estimated value of the short-term sliding window and the parameter relationship setting value in the control model. This deviation is used to determine the parameter correction direction and set the adjustment amount proportional to it:

[0101]

[0102] Where subscript m represents the estimated value of the relationship, subscript s represents the set value of the relationship, and k2 and k3 are proportional coefficients. Note that as the temperature control approaches stability, the calculation of equation (30) gradually becomes ineffective. Therefore, this correction process mainly acts on the initial and middle stages of the single-step isothermal control process and is inversely proportional to the factor r' representing the control process.

[0103] Finally, the delay time estimation also has a significant impact on the accuracy of the model prediction. When the sample temperature rate is high and stable during the temperature control process, the direction and magnitude of the τ correction can be determined by comparing the model output with the actual sample temperature after delay estimation correction. The adjustment amount is expressed as follows:

[0104]

[0105] The sign of the adjustment amount is determined by the direction of temperature control, the relationship between the model output and the temperature of the calibration sample. Taking the heating process as an example, if the temperature rate tends to be stable and the model output is higher than the actual output at a certain moment, the delay time should be increased, and vice versa.

[0106] In summary, it is only necessary to adjust the four proportional coefficients k1 to k4 in the sampling period adjustment formula to achieve directional correction of the parameters of the mechanism model through the above simple logic, thereby reducing the impact of model mismatch on control performance.

[0107] 3. Model parameter initialization

[0108] Another significance of using mechanism parameters to design prediction models is that the universal calibration process of the reaction calorimetry system can be used to calculate equivalent parameters and update model parameters, which significantly improves the accuracy of model predictions in the subsequent temperature control process.

[0109] Due to the need to conduct online reaction exothermic measurement, for different sample systems, each reaction calorimetric experiment requires online calibration of thermodynamic parameters before and after the chemical reaction process. As described in the general calibration method of the background technology, formula (4) and formula (5) are used as calculation formulas for heat transfer factor and system heat capacity. The present invention fully utilizes this calibration process and uses the additional input information provided by the correction heater to identify the system parameters, thereby avoiding the application of other complex or time-consuming system identification methods. When calculating the equivalent mechanism parameters, the oil bath temperature T o Replace the jacket temperature T j , using the equivalent heat transfer factor (UA) e 、Equivalent thermal inertia is (C p M) eThe heat transfer factor between the sample and the jacket and the total heat capacity of the system are respectively replaced by equations (4) and (5) for calculation. Therefore, there is no need to perform additional experimental operations. The online calculation of the two equivalent mechanism parameters can be realized synchronously during the calibration process, namely:

[0110]

[0111]

[0112] In addition, in the isothermal control steady-state stage without heating rod power input, as Figure 3 In the 1-2 stage or the 3-4 stage, the heat transfer from the jacket to the sample is balanced with the heat dissipation from the sample to the environment. Therefore, the equivalent heat dissipation coefficient can also be estimated from the data measured in these two stages:

[0113] α=(UA) e (T o -T r ) / T r (35)

[0114] The above equivalent parameter calculation is performed in the calibration process before the chemical reaction of each experiment, and the prediction model parameters are initialized online to obtain α0, (UA) e_0 and (C p M) e_0 And setting the model reference working point to the process temperature during calibration can effectively reduce the impact of model mismatch in the subsequent temperature control process.

[0115] The present invention will be further described below.

[0116] 1. Issues such as data filling time and zero crossing point

[0117] Correction (UA) by measuring the relationship between parameters in the sliding time domain during the temperature control process e and (C p M) e When , since it is necessary to take two sets of short-time sliding windows to perform sliding average on the input and output data, the calculation of formula (31) should be avoided in the initial stage of a temperature control step and before the sliding short-time window data is filled, let (UA) e and (C p M) e The adjustment amount is 0.

[0118] When using formula (30) to calculate the parameter relationship, the denominator of the formula may approach 0 during the dynamic process of temperature control and when the temperature tends to be stable, which will cause the calculated value to be distorted. Therefore, the algorithm sets a lower limit for the absolute value of the denominator. When the calculated value of the denominator is less than or equal to the lower limit, the corresponding parameter adjustment amount is 0.

[0119] 2. Parameter boundary constraints and adjustment termination conditions

[0120] Considering the physical meaning of the equivalent mechanism parameters, the heat transfer factor, thermal inertia and system delay time must be non-negative. Therefore, in the process of online parameter adjustment, in order to avoid excessive directional adjustment of parameters due to long temperature control time, the method has the following logic: Constraint (UA) e and (C p M) e When the real-time calculated value of the parameter is less than the boundary condition and the real-time calculated value of the delay time is less than or equal to 0, the adjustment amount of the corresponding parameter is set to 0.

[0121] The online adjustment method essentially performs directional parameter corrections. Without a termination condition, the real-time correction process will continue, causing the predicted output to constantly change and slowing control convergence. Therefore, the method employs the following logic: It simultaneously determines the difference between the actual sample temperature and the set temperature, as well as the rate of temperature change. When both are less than a set threshold, the temperature control is considered stable, and the online parameter adjustment is terminated, returning all parameter adjustments to zero.

[0122] 3. Comparison of temperature control effects

[0123] The control effect of the present invention is illustrated by combining the reaction calorimetry system simulation temperature control process curve. Figure 5 The isothermal control process curves of the reaction calorimetry system are shown using different temperature control methods to increase the sample temperature from room temperature to 150°C. The thickness of the curves distinguishes the parameters. The thin black line represents the oil bath set temperature, which is the controller output operation quantity T o_set , the thick black line represents the sample temperature T r ; Line type distinction method, the dotted line represents PID control, the dotted line represents the MPC control without using parameter initialization and online correction, and only using the mechanism parameter model of the present invention, and its model parameters are set according to general experience, and the solid line is the control method implemented by the complete control logic of the present invention, represented by PIM-MPC, as shown Figure 5 At the same time, the figure compares the convergence speed of each method of isothermal control, using the threshold of the difference between the sample temperature and the target temperature and the real-time temperature rate as the judgment, and the time for each method to reach steady state is marked in the figure.

[0124] From the comparison of overshoot, oscillation trend and marking time, it can be seen that the temperature control stability and convergence speed of MPC and PIM-MPC are significantly better than PID. It should be noted that although PID may achieve a similar control effect on the target temperature after parameter tuning, within the operating temperature range of the reaction calorimetry system, the parameter adaptability of PID in other working conditions will still be limited. The control effect of the PIM-MPC method adopted in the present invention is also better than that of ordinary MPC, especially in terms of convergence speed. This is because after parameter initialization and online correction, the accuracy of model prediction is higher. Figure 6 Shown Figure 5 The model predicts the output T during the control process r_m Corrected sample temperature T with delay estimation r_pav Comparison of the curves shows that the model output results in the PIM-MPC method are more consistent with the estimated and corrected sample temperature. Especially during the dynamic adjustment process, the feedback error is kept at a small level, reducing the impact of model mismatch, thereby significantly improving the rapidity of temperature control.

[0125] In summary, this paper proposes a model-based predictive temperature control method for a reaction calorimetry system based on online correction of mechanism parameters. This method establishes an equivalent mechanism parameter model for the reaction calorimetry system based on heat transfer analysis, uses a universal parameter calibration method to set initial model parameter values, and then adjusts the model parameters online during the model-based predictive temperature control process by combining real-time parameter relationship comparison and other logic. This effectively suppresses model mismatch and achieves excellent temperature control performance. This method does not rely on other complex online system parameter identification algorithms, and changes in model parameters can provide a reference for actual parameter change trends.

Claims

1. A reaction calorimetry system model prediction and temperature control method based on online correction of mechanism parameters, characterized by: (1) Equivalent mechanism parameter model output The equivalent mechanism parameter model is used for prediction output, and the reaction calorimetric system is simplified to a first-order inertial system, with the internal temperature of the oil bath T o is the input, the sample temperature T r For output, assume the equivalent heat transfer factor (UA) e 、Equivalent thermal inertia is (C p M) e And the equivalent heat dissipation coefficient is α, T s is the sampling time, and the discretization model of the future P step is as follows: Where k represents the current sampling moment, and factors A and B are functions of the equivalent mechanism parameters: (2) Actual output delay estimation correction and predicted output deviation correction Considering the existence of a pure lag link in the system with a delay time of τ, the actual sample temperature is estimated and corrected: T r_pav (k)=T r (k)+T r_m (k+τ / T s )-T r_m (k) Calculate the deviation between the sample temperature after delay estimation correction and the current model prediction output value: e(k)=T r_pav (k)-T r_m (k) (3) Calculation of optimal control quantity Considering the optimization of the control objective in the future P step, the optimal control quantity at the current moment is: Among them, T traj represents the sample temperature reference trajectory; (4) Model parameter initialization and online correction Before each chemical reaction occurs, the calibration process is performed to initialize the parameters online to obtain α0, (UA) e_0 and (C p M) e_0 Initial values ​​of three parameters and model mechanism parameters (UA) are implemented in each step of isothermal control during the experiment e 、(C p M) e , α and delay time τ online correction; In order to suppress the influence of model mismatch, an online parameter adjustment strategy is designed; in each sampling control cycle, the adjustment amount of the equivalent heat dissipation coefficient α is: dα=-k1r’(T r -T r_m ) / |T r | Where k1 is the proportional coefficient, r' represents the adjustment speed factor related to the control process stage; parameter (UA) e and (C p M) e The adjustment amount is: Where subscript m represents the estimated value of the parameter relationship, subscript s represents the set value of the relationship; k2 and k3 are proportional coefficients; the adjustment amount of the delay time τ is: dτ=k4sign[(T r_set -T ini )(T r_m -T r_pav )]...·|(T r_m -T r_pav )dt / dT r | Where k4 is the proportional coefficient, sign is the sign judgment function, T r_set is the target temperature, T ini is the initial temperature of the sample, T r_pav It is the actual temperature after delay estimation correction, and t represents time.

2. The reaction calorimetry system model prediction and temperature control method based on online correction of mechanism parameters according to claim 1 is characterized in that: At the beginning of each reaction experiment, the universal calibration process of the reaction calorimetry system is used to calculate the equivalent parameters and initialize the model parameters. The following calculations are completed simultaneously during the calibration process: α=(UA) e (T o -T r ) / T r Where P H To correct the heater power, △T r and △T o are the relative changes of sample and oil bath temperature respectively; the above calculation results are used as α0, (UA) e_0 and (C p M) e_0 Update the initial parameter values ​​and set the model benchmark operating point to the process temperature during calibration.

Images