A method and system for dynamic determination of heat transfer coefficient with regularization constraint

Abstract

The application discloses a kind of regularized constraint heat transfer coefficient dynamic determination method and system.The application introduces sine oscillation in jacket temperature, data is divided into sliding window;Linear model containing low-frequency term and simple harmonic term of the same frequency is constructed in window, and the energy balance equation is established by extracting the same frequency complex amplitude;Further, the least square target function with regularization term is constructed with the estimation value of the previous window as prior, and the non-negative constraint based on the interpolation of the boundary before and after reaction is applied, and the UA dynamic evaluation is realized by cyclic solution.The application effectively decouples non-steady-state drift and noise, suppresses numerical mutation, obtains smoother and more physical law heat transfer coefficient and heat release rate result, and significantly improves the robustness of complex reaction process calorimetry.

Description

Technical Field

[0001] This invention belongs to the field of fine chemical reaction safety testing technology, and relates to a regularized constraint dynamic measurement method and system for heat transfer coefficient, which is used to dynamically measure the heat transfer coefficient in the processing of chemical reaction calorimetric experimental data and to correct the calculation of heat release. Background Technology

[0002] The main purpose of reaction calorimetry is to continuously estimate the reaction exothermic rate from temperature measurements and energy balance equations around the reactor, providing a basis for reaction safety risk assessment, process development and optimization. The accuracy of heat transfer coefficient assessment is one of the key technologies for accurately estimating the reaction exothermic rate.

[0003] The thermal balance in the reactor is a crucial basis for calculating the exothermic reaction. The exothermic reaction is calculated using data obtained from an automated reaction calorimeter. The principle and structure of the automated reaction calorimeter and the heat transfer between the reactor sample are as follows: Figure 1 As shown.

[0004] Based on the derivation of the system's heat balance, we can obtain:

[0005] (1)

[0006] Among them, Q r Q represents the real-time exothermic reaction rate of sample 1 to be measured. flow The heat flow transferred from sample 1 to jacket 3 is expressed as Q. flow =UA( ), The temperature of sample 1 inside the reactor. The temperature of jacket 3 is obtained by temperature sensor 4, Q. acc The heat accumulated by reactor 2, liquid feed device 9, or stirring blades 8 is reflected in the temperature change, Q. loss The heat loss in reactor 2, Q, is the heat flow lost to the surrounding air through phase change convection via the lid 7, the stirring rod 5 inserted inside the reactor, the correction heater 6, and material conduction. dos It is the heat flow lost when passing through the liquid feed device 9, Q c It is the heat power released by the calibration heater 6 of the reaction calorimeter.

[0007] To obtain the real-time reaction exothermic rate Oscillating temperature calorimetry is popular due to its ease of operation and low cost. This method is based on the temperature of its jacket. Introducing frequency into control signals The sinusoidal oscillation with period P is derived using signal processing and mathematical derivation, resulting in the following formula for calculating temperature calorimetry using the traditional oscillatory method:

[0008] (2)

[0009] in, and The sample temperature under temperature oscillation and jacket temperature The amplitude, It is the system's residual heat capacity, where C p The specific heat capacity of the insert is M, and the total mass of the reactor is M. The values ​​are usually calculated using the calibrated values. Formula (2) is a dynamic evaluation of the reaction process by calculating UA for each oscillation cycle.

[0010] The effectiveness of the traditional oscillatory temperature calorimetry method (Equation (2)) depends on the assumption that the system is in an ideal sinusoidal harmonic steady state. However, in actual chemical calorimetry processes, the intense exothermic reaction can cause the sample temperature to produce a nonlinear distorted waveform, resulting in the sample temperature no longer having a single sinusoidal characteristic, which in turn leads to calculation distortion.

[0011] Furthermore, the algorithm of formula (2) exhibits a significant ill-conditioned range in numerical calculation. When the system's heat transfer efficiency is extremely high or the environmental noise is significant, causing the amplitude ratio to approach 1, the square root term in the denominator of the calculation formula tends to zero, which will cause the estimated UA value to tend to infinity or produce drastic numerical abrupt changes. Conversely, when the system's heat transfer efficiency is extremely poor or the material's thermal inertia is extremely high, causing the amplitude ratio to approach 0, the estimated UA value will be severely underestimated or even tend to zero. The aforementioned ill-conditioned range is essentially caused by insufficient information on the same frequency response, which limits the accuracy and stability of the dynamic estimation of traditional methods under unsteady conditions. Therefore, a dynamic evaluation method with physical constraints is needed. Summary of the Invention

[0012] To address the shortcomings of existing technologies, this invention proposes a method and system for dynamically determining the heat transfer coefficient under regularized constraints.

[0013] This invention employs sinusoidal oscillation excitation to realize a complete reaction process, and acquires and obtains sample temperature in real time. With jacket temperature The original time-domain sequence data was obtained, and the same-frequency response information was accurately decoupled from the complex unsteady temperature signal. By introducing a calculation formula with regularization constraints, a robust, continuous and dynamic evaluation of the heat transfer coefficient UA was achieved under complex dynamics, intense heat release and low signal-to-noise ratio conditions.

[0014] The present invention provides a method for dynamically determining the heat transfer coefficient under regularized constraints, used for the dynamic evaluation of the heat transfer coefficient during the reaction process of an automated reaction calorimeter, comprising the following steps:

[0015] Step (1) Introduce sinusoidal oscillation excitation into the jacket temperature control signal of the automatic reaction calorimeter controlled by isothermal method, execute complete reaction process control, and acquire the original sequence data of sample temperature and jacket temperature in real time.

[0016] Step (2) Divide the original sequence data into a series of sliding windows of length W, and set the heat transfer coefficient corresponding to the center time of each window as the quantity to be estimated;

[0017] Step (3) Within each sliding window k, construct a linear parameter model containing low-frequency terms and harmonic terms of the same frequency, and use the least squares method to identify the coefficients of the low-frequency terms and the coefficients of the harmonic terms of the same frequency in the model; wherein, the low-frequency terms are used to characterize the quantity with a slow changing trend, and the harmonic terms of the same frequency are used to characterize the same frequency response to the jacket sinusoidal excitation.

[0018] Step (4) Based on the coefficients of the harmonic terms identified in step (3), construct the complex amplitude values ​​of the sample temperature and jacket temperature components in the k-th window, and substitute these complex amplitude values ​​into the energy balance relationship under the same frequency components to establish the same frequency energy balance equation for the heat transfer coefficient of the current window; thereby mapping the same frequency temperature response to the heat transfer coefficient equation to be estimated.

[0019] Step (5) uses the same frequency energy balance equation obtained in step (4) as the data fitting term, the heat transfer coefficient estimate of the previous window as the prior reference term, constructs a least squares objective function with regularization term, obtains the heat transfer coefficient of the current window by finding the extreme value of the objective function, and applies physical constraints. The result is used as the prior reference input of the next sliding window for continuous updating of the heat transfer coefficient calculation of subsequent windows.

[0020] Step (6) Repeat steps (3) to (5) in each window to achieve dynamic evaluation of the heat transfer coefficient of the reaction process.

[0021] A regularized constraint-based dynamic measurement system for heat transfer coefficients according to the present invention includes:

[0022] The reactor is used to contain the reaction sample and is equipped with a temperature sensor and a calibration heater;

[0023] A jacket is provided around the reactor, and a circulating heat transfer medium is introduced into the jacket to control the reaction temperature, which is used to introduce sinusoidal oscillations into the temperature control signal.

[0024] The data acquisition module is connected to the temperature sensor and is used to acquire raw sequence data of sample temperature and jacket temperature in real time.

[0025] The sliding window processing module is used to divide the original sequence data into a series of sliding windows of equal length, and set the heat transfer coefficient corresponding to the center time of each window as the quantity to be estimated.

[0026] The temperature signal decomposition module is used to construct a linear parameter model containing low-frequency terms and harmonic terms within each sliding window, and to identify the coefficients of the low-frequency terms and harmonic terms using the least squares method. The low-frequency terms are used to characterize quantities with slow changing trends, and the harmonic terms are used to characterize the same-frequency response to the jacket sinusoidal excitation.

[0027] The energy balance mapping module is used to construct the complex amplitude values ​​of the sample temperature and jacket temperature in each window based on the same frequency harmonic term coefficients, and substitute the same frequency complex amplitude values ​​into the energy balance relationship under the same frequency components to establish the same frequency energy balance equation about the heat transfer coefficient of the current window.

[0028] The regularized solution module is used to construct a least-squares objective function with a regularization term, using the same-frequency energy balance equation as the data fitting term and the heat transfer coefficient estimate of the previous window as the prior reference term. The heat transfer coefficient of the current window is obtained by finding the extremum of the objective function and applying physical constraints. The result is used as the prior reference input for the next sliding window.

[0029] The cyclic evaluation module is used to repeatedly call the temperature signal decomposition module, energy balance mapping module, and regularization solution module within each sliding window to achieve dynamic evaluation of the heat transfer coefficient of the reaction process.

[0030] The beneficial effects of this invention are:

[0031] This invention does not simply filter and calculate the data, but decomposes the data into low-frequency terms and harmonic terms of the same frequency to establish a clear and identifiable mathematical model. The low-frequency terms are used to characterize the slow drift caused by exothermic reactions, changes in heat capacity, etc., while the harmonic terms of the same frequency are used to characterize the same-frequency response caused by the sinusoidal excitation of the jacket. This reduces the impact of unsteady-state drift on the calculation of the heat transfer coefficient at the data level and improves the robustness of the heat transfer coefficient estimation results of the reaction process.

[0032] This invention introduces a regularization constraint in the heat transfer coefficient solution, using the above window estimation result as a reference, so that the estimated values ​​of adjacent windows remain reasonably continuous in time. When the same frequency response is weak or the noise is large, causing the solution to be ill-timed, this constraint can suppress the estimation jitter and abrupt changes caused by noise amplification, thereby obtaining a smoother dynamic result that is more in line with the actual evolution law of the heat transfer coefficient.

[0033] This invention obtains the boundary values ​​of the heat transfer coefficient using a traditional oscillatory method before and after the reaction, and performs linear interpolation of the boundary values ​​according to the sliding window position during the reaction process. This serves as a non-negative constraint based on the boundary interpolation before and after the reaction in the dynamic evaluation process of the heat transfer coefficient, avoiding non-physical solutions such as negative heat transfer coefficients that violate the heat transfer mechanism, and improving the reliability of the dynamic evaluation results in the engineering scenario of automatic reaction calorimeter. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the principle structure of an automatic reaction calorimeter and the heat transfer of the reactor sample.

[0035] Figure 2 The flowchart shows the algorithm for dynamically evaluating the heat transfer coefficient under regularized constraints during the reaction process of an automatic reaction calorimeter.

[0036] Figure 3 This is a schematic diagram of a complete experimental process;

[0037] Figure 4 These are reaction curves for two simulated operating conditions in an embodiment of the present invention;

[0038] Figure 5 This is a comparison chart of UA estimation between the traditional oscillating temperature calorimetry method and the method of the present invention under two working conditions in an embodiment of the present invention.

[0039] Figure 6 This invention provides two examples of the traditional oscillating temperature calorimetry method (Qr) under different operating conditions in this embodiment. toc ) and the method of the present invention (Qr k Comparison chart of Qr estimations. Detailed Implementation

[0040] The invention will now be further described in conjunction with the accompanying drawings.

[0041] like Figure 2 As shown in the figure, this application provides a method for dynamically determining the heat transfer coefficient under regularization constraints. The specific steps are as follows:

[0042] (1) Jacket oscillation excitation and raw data acquisition

[0043] The reaction calorimeter is controlled by the isothermal method, and the jacket temperature T is controlled at a constant temperature. j The frequency introduced in the control signal is A sinusoidal oscillation excitation with a period of P is applied. The period P is determined based on the system time constant. This process controls the entire reaction process and collects the sample temperature T in real time. r With jacket temperature T j The original sequence data.

[0044] Furthermore, such as Figure 3 A complete chemical reaction experiment includes the following steps: Step 1: Raise the sample temperature to the process temperature; Step 2: Start the oscillation before the reaction begins; Step 3: The reaction process; Step 4: Oscillation after the reaction ends; Step 5: Return to the process temperature setpoint and end the reaction.

[0045] (2) Sliding window division

[0046] The sample temperature T will be obtained. r With jacket temperature T j The data is divided into a series of sliding windows of length W according to the sampling time, where the window length W is an integer multiple of the oscillation period, and the heat transfer coefficient UA corresponding to the center time of the window is set. k This is a quantity to be estimated.

[0047] (3) Temperature signal decomposition and model identification

[0048] Within each sliding window k, a linear parameter model as shown in formula (3) is constructed. The linear model decomposes the temperature signal (including the sample temperature signal and the jacket temperature signal) into a low-frequency term and a harmonic term of the same frequency.

[0049] The sample temperature T within each window was determined using the least squares method. r With jacket temperature T j The normal equations are solved, and the parameters of the linear model are identified simultaneously. The purpose of this step is to use polynomial terms to absorb the unsteady-state drift caused by the exothermic reaction, ensuring that the subsequently extracted harmonic coefficients are not affected by waveform distortion.

[0050] (3)

[0051] In the formula: The coefficients of the low-frequency terms, The coefficients of the harmonic terms of the same frequency, The center time of the k-th sliding window The sampling time of the temperature signal is . For the jacket temperature T j A frequency with the same frequency is introduced into the control signal.

[0052] (4) Sample temperature T r With jacket temperature T j Complex amplitude construction and energy balance mapping

[0053] Using the identified harmonic coefficients The complex amplitude values ​​of sample temperature Tr and jacket temperature Tj are constructed as shown in formulas (4) and (5). Combined with the total system heat capacity... According to the heat balance equation (1), an energy balance equation based on the same frequency component can be established, as shown in formula (6).

[0054] (4)

[0055] (5)

[0056] (6)

[0057] In the formula: and for and The complex amplitude value of the same frequency component under the k-th sliding window. The sample temperature T is under the k-th sliding window. r The coefficients of the harmonic terms of the same frequency, The jacket temperature T is under the k-th sliding window. j The coefficients of the harmonic terms with the same frequency.

[0058] (5) Solving the computational formula with regularization constraints

[0059] The heat transfer coefficient estimate UA in the above window k-1 As a priori reference, a least-squares objective function with a regularization term is constructed as shown in equation (7). In addition to the data fitting term in the first term, equation (7) also includes a physical penalty term (i.e., a regularization term) that reflects the heat transfer mechanism. This is because in actual chemical reactions, the heat transfer coefficient of the reactor changes continuously and slowly, and there should be no sudden changes in temperature.

[0060] This embodiment introduces a regularization term to constrain excessive differences in heat transfer coefficients between adjacent sliding windows, thereby obtaining smoother dynamic results that better reflect the actual physical evolution of the heat transfer coefficient. By minimizing the objective function and applying physical constraints, the current window's heat transfer coefficient is calculated using the update formula (8). .

[0061] (7)

[0062] (8)

[0063] In the formula, , , This refers to taking the real value. and for and The complex amplitude value of the same frequency component under the k-th sliding window. for The conjugate of complex numbers.

[0064] Furthermore, regularization parameters Calculated according to formula (9), by automatically adjusting parameters (A value between 0 and 1) is used to automatically adjust the estimated smoothness. The formula for calculating this parameter is:

[0065] (9)

[0066] By taking the median To obtain the baseline level of signal strength throughout the experiment, the estimated sequence was minimized using historical data. and benchmark value Calculate the root mean square error between them Values, thereby enabling each window to... Automatic adjustment.

[0067] (6) Baseline value Calculation

[0068] First, calculations were performed before and after the reaction using the traditional oscillatory method (Formula (2)). and Then, during the reaction process, linear interpolation is performed based on a sliding window, with the reference value... The calculation formula is as follows:

[0069] (10)

[0070] in Let k be the total number of sliding windows, and k be the current window position.

[0071] (7) Iterative evaluation within a sliding window of the full sequence data

[0072] By repeating steps (3) to (5) within each sliding window, the numerical oscillations generated during the solution process by the traditional oscillatory algorithm can be eliminated, thereby achieving accurate dynamic evaluation of the heat transfer coefficient throughout the reaction process.

[0073] This application also provides a system for dynamically measuring the heat transfer coefficient under regularization constraints, including:

[0074] The reactor is used to contain the reaction sample and is equipped with a temperature sensor and a calibration heater;

[0075] A jacket is provided around the reactor, and a circulating heat transfer medium is introduced into the jacket to control the reaction temperature, which is used to introduce sinusoidal oscillations into the temperature control signal.

[0076] The data acquisition module is connected to the temperature sensor and is used to acquire raw sequence data of sample temperature and jacket temperature in real time.

[0077] The sliding window processing module is used to divide the original sequence data into a series of sliding windows of equal length, and set the heat transfer coefficient corresponding to the center time of each window as the quantity to be estimated.

[0078] The temperature signal decomposition module is used to construct a linear parameter model containing low-frequency terms and harmonic terms within each sliding window, and to identify the coefficients of the low-frequency terms and harmonic terms using the least squares method. The low-frequency terms are used to characterize quantities with slow changing trends, and the harmonic terms are used to characterize the same-frequency response to the jacket sinusoidal excitation.

[0079] The energy balance mapping module is used to construct the complex amplitude values ​​of the sample temperature and jacket temperature in each window based on the same frequency harmonic term coefficients, and substitute the same frequency complex amplitude values ​​into the energy balance relationship under the same frequency components to establish the same frequency energy balance equation about the heat transfer coefficient of the current window.

[0080] The regularized solution module is used to construct a least-squares objective function with a regularization term, using the same-frequency energy balance equation as the data fitting term and the heat transfer coefficient estimate of the previous window as the prior reference term. The heat transfer coefficient of the current window is obtained by finding the extremum of the objective function and applying physical constraints. The result is used as the prior reference input for the next sliding window.

[0081] The cyclic evaluation module is used to repeatedly call the temperature signal decomposition module, energy balance mapping module, and regularization solution module within each sliding window to achieve dynamic evaluation of the heat transfer coefficient of the reaction process.

[0082] Verification example:

[0083] This verification example introduces two operating conditions:

[0084] (1) Constant low-power heat generation condition: This condition is as follows Figure 4 As shown in (a), a constant small power of about 5W was simulated (lasting for about 180 min) to simulate the dynamic process of a stable heat release rate.

[0085] (2) Simulation of time-varying heat release rate: such as Figure 4 In (b), this operating condition utilizes a Gaussian thermal pulse with a peak value of approximately 50W to simulate the thermal flow evolution caused by a typical constrained reaction process or semi-batch feeding in fine chemicals.

[0086] Among them, UA set The input setting value is:

[0087] (11)

[0088] Figure 5This invention provides two examples of the traditional oscillating temperature calorimetry method (UA) under different operating conditions. toc ) and the method of the present invention (UA) k The comparison chart shows the estimated UA values ​​under constant low-power heat production conditions. The results indicate that, under constant low-power heat production conditions, UA... toc The average relative error is 1.16%, while UA k The average relative error was reduced to 0.52%; under the time-varying heat release rate simulation condition, UA toc The average relative error is 2.90%, while UA k It dropped to 0.42%.

[0089] As can be seen from the curves, the heat transfer coefficient calculated by traditional methods is easily affected by noise and drift when the reaction is intense and the temperature fluctuates drastically, resulting in irregular fluctuations or local abrupt changes in the curve. In contrast, this application introduces frequency separation and energy balance modeling within a window, along with physical constraints, to make the output curve more stable and natural throughout the entire reaction cycle.

[0090] Figure 6 This invention provides two examples of the traditional oscillating temperature calorimetry method (Qr) under different operating conditions. toc ) and the method of the present invention (Qr k A comparison chart of Qr estimates. Experiments show that under constant low-power heat production, the reference value Qr... set The sum is 875J. The traditional method estimates the sum to be 881.85J (error 0.78%), while this application corrects it to 873.77J (error only -0.14%), which is closer to the actual physical quantity.

[0091] Under time-varying exothermic conditions, the estimation deviation of traditional methods is as high as -12.79%, while this application successfully converges it to -5.52%. From the curve evolution pattern, since the calculation of the exothermic rate is extremely sensitive to fluctuations in the heat transfer coefficient, even minor fluctuations in the heat transfer coefficient in traditional methods are significantly amplified, leading to unrealistic fluctuations or peak distortion in the exothermic rate curve. In contrast, the curve evolution obtained by the method of this invention is smoother, accurately capturing the entire process of the exothermic peak's initiation, peak attainment, and decay, effectively avoiding spurious peak interference.

[0092] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for dynamically determining the heat transfer coefficient under regularization constraints, used for the dynamic evaluation of the heat transfer coefficient during the reaction process of an automatic reaction calorimeter, characterized in that, Includes the following steps: Step (1) Introduce sinusoidal oscillation excitation into the jacket temperature control signal of the automatic reaction calorimeter controlled by isothermal method, execute complete reaction process control, and acquire the original sequence data of sample temperature and jacket temperature in real time. Step (2) Divide the original sequence data into a series of sliding windows of length W, and set the heat transfer coefficient corresponding to the center time of each window as the quantity to be estimated; Step (3) Within each sliding window k, construct a linear parameter model containing low-frequency terms and harmonic terms of the same frequency, and use the least squares method to identify the coefficients of the low-frequency terms and the coefficients of the harmonic terms of the same frequency in the model; wherein, the low-frequency terms are used to characterize the quantity with a slow changing trend, and the harmonic terms of the same frequency are used to characterize the same frequency response to the jacket sinusoidal excitation. Step (4) Based on the coefficients of the harmonic terms identified in step (3), construct the complex amplitude values ​​of the sample temperature and jacket temperature components in the k-th window, and substitute these complex amplitude values ​​into the energy balance relationship under the same frequency components to establish the same frequency energy balance equation for the heat transfer coefficient of the current window. Step (5) uses the same frequency energy balance equation obtained in step (4) as the data fitting term, the heat transfer coefficient estimate of the previous window as the prior reference term, constructs a least squares objective function with regularization term, obtains the heat transfer coefficient of the current window by finding the extreme value of the objective function, and applies physical constraints. The result is used as the prior reference input of the next sliding window for continuous updating of the heat transfer coefficient calculation of subsequent windows. Step (6) Repeat steps (3) to (5) in each window to achieve dynamic evaluation of the heat transfer coefficient of the reaction process.

2. The method for dynamically determining the heat transfer coefficient under regularized constraints according to claim 1, characterized in that, The linear parameter model described in step (3) decomposes the temperature signal into two parts: a low-frequency trend term and a harmonic oscillation term. The low-frequency trend term is represented by a polynomial of a time variable with the window center time as the origin, and is used to characterize the slow temperature drift caused by the exothermic reaction. The harmonic oscillation term is represented by a linear combination of cosine and sine functions with the oscillation frequency as the angular frequency, and is used to characterize the same-frequency temperature response caused by the jacket sinusoidal excitation.

3. The method for dynamically determining the heat transfer coefficient under regularized constraints according to claim 1, characterized in that, The complex amplitude of the same frequency component mentioned in step (4) is constructed from the cosine coefficient and sine coefficient of the same frequency simple harmonic term. Specifically, the cosine coefficient of the sample temperature or jacket temperature is taken as the real part, and the negative sine coefficient is taken as the imaginary part, and the amplitude is combined to form a complex form.

4. The method for dynamically determining the heat transfer coefficient under regularized constraints according to claim 3, characterized in that, The same-frequency energy balance equation in step (4) is expressed as follows: the product of the current window heat transfer coefficient and the complex amplitude of the same-frequency component of the jacket temperature minus the complex amplitude of the same-frequency component of the sample temperature is equal to the product of the cumulative heat capacity of the system with the imaginary unit, the oscillation angular frequency and the complex amplitude of the same-frequency component of the sample temperature.

5. The method for dynamically determining the heat transfer coefficient under regularized constraints according to claim 1, characterized in that, The least squares objective function with regularization in step (5) contains two terms: The first term is the data fitting term, which characterizes the difference between the complex amplitude values ​​of the same frequency components of the sample temperature and the jacket temperature and the degree of deviation from the cumulative heat capacity of the system. The second term is the regularization term, which represents the value obtained by multiplying the square of the difference between the heat transfer coefficient of the current window and the heat transfer coefficient of the previous window by the regularization parameter, wherein the regularization parameter is used to adjust the smoothness of the estimation.

6. The method for dynamically determining the heat transfer coefficient under regularized constraints according to claim 1, characterized in that, The objective function in step (5) is solved by an iterative update method. The heat transfer coefficient of the current window is expressed as a functional relationship that includes the heat transfer coefficient of the previous window, the regularization parameter, the difference between the complex amplitudes of the same frequency components of the sample temperature and the jacket temperature, the conjugate of the complex amplitudes of the same frequency components of the sample temperature, and the cumulative heat capacity of the system. The real part of the calculation result of this functional relationship is extracted as the estimated value of the heat transfer coefficient of the current window.

7. The method for dynamically determining the heat transfer coefficient under regularized constraints according to claim 5, characterized in that, The regularization parameter is determined based on the signal strength, and its value is jointly determined by the intensity reference of the difference between the adjustment coefficient and the complex amplitude values ​​of the same-frequency components of the sample temperature and the jacket temperature; wherein: The strength benchmark is the median of the square modulus of the difference between the complex amplitude of the jacket temperature component and the complex amplitude of the sample temperature component in each sliding window. The adjustment coefficient, ranging from zero to one, is used to adjust the strength of the regularization constraint.

8. The method according to claim 7, characterized in that, The adjustment coefficient is determined based on a reference heat transfer coefficient sequence, and is specifically calculated in the following manner: First, the initial and final boundary values ​​of the heat transfer coefficient were determined using the traditional oscillating temperature calorimetry method before and after the reaction. Then, during the reaction process, the initial boundary value and the final boundary value are linearly interpolated according to the time position of the sliding window to obtain the reference heat transfer coefficient sequence corresponding to each window; Furthermore, the adjustment coefficient is calculated by minimizing the root mean square error between the estimated sequence and the benchmark heat transfer coefficient sequence based on historical data, thereby achieving self-adjustment of the regularization parameter.

9. The method according to claim 1, characterized in that, The physical constraints mentioned in step (5) include non-negative constraints based on boundary interpolation before and after the reaction, which are implemented in the following way: The initial and final boundary values ​​of the heat transfer coefficient were obtained using the traditional oscillating temperature calorimetry method before and after the reaction. During the reaction process, the initial boundary value and the ending boundary value are linearly interpolated according to the time position of the sliding window to obtain the reference heat transfer coefficient of the current window; The calculated result of the heat transfer coefficient of the current window is constrained to a value greater than or equal to zero, and the benchmark heat transfer coefficient is used as a reference boundary for judging physical rationality.

10. A system for dynamically measuring the heat transfer coefficient under regularization constraints, characterized in that, include: The reactor is used to contain the reaction sample and is equipped with a temperature sensor and a calibration heater; A jacket is provided around the reactor, and a circulating heat transfer medium is introduced into the jacket to control the reaction temperature, which is used to introduce sinusoidal oscillations into the temperature control signal. The data acquisition module is connected to the temperature sensor and is used to acquire raw sequence data of sample temperature and jacket temperature in real time. The sliding window processing module is used to divide the original sequence data into a series of sliding windows of equal length, and set the heat transfer coefficient corresponding to the center time of each window as the quantity to be estimated. The temperature signal decomposition module is used to construct a linear parameter model containing low-frequency terms and harmonic terms within each sliding window, and to identify the coefficients of the low-frequency terms and harmonic terms using the least squares method. The low-frequency terms are used to characterize quantities with slow changing trends, and the harmonic terms are used to characterize the same-frequency response to the jacket sinusoidal excitation. The energy balance mapping module is used to construct the complex amplitude values ​​of the sample temperature and jacket temperature in each window based on the same frequency harmonic term coefficients, and substitute the same frequency complex amplitude values ​​into the energy balance relationship under the same frequency components to establish the same frequency energy balance equation about the heat transfer coefficient of the current window. The regularization solution module is used to construct a least squares objective function with a regularization term, using the same frequency energy balance equation as the data fitting term and the heat transfer coefficient estimate of the previous window as the prior reference term. The heat transfer coefficient of the current window is obtained by finding the extreme value of the objective function and applying physical constraints. The result is used as the prior reference input of the next sliding window. as well as The cyclic evaluation module is used to repeatedly call the temperature signal decomposition module, energy balance mapping module, and regularization solution module within each sliding window to achieve dynamic evaluation of the heat transfer coefficient of the reaction process.

Images