A method for adaptive temperature control of a triple-effect evaporator

Abstract

This invention relates to the field of evaporation process control technology, and discloses a temperature adaptive control method and system for a triple-effect evaporator. The method includes acquiring and preprocessing temperature signals and steam valve position signals for each effect to obtain operating data; using a preset Kalman observer to perform state estimation on the operating data to obtain thermal state values; updating preset heat transfer parameters based on the thermal state values ​​and a smoothing error sequence to obtain updated state results; determining valve position adjustment amounts based on the updated state results and a preset temperature target, and generating steam valve position commands. This method achieves dynamic correction and closed-loop regulation of the temperature control parameters of the triple-effect evaporator.

Description

Technical Field

[0001] This invention relates to the field of evaporation process control technology, and in particular to a temperature adaptive control method and system for a triple-effect evaporator. Background Technology

[0002] Currently, in industrial production, especially in the operation of triple-effect evaporators, the temperature stability between each effect directly affects evaporation efficiency, product quality, and system energy consumption. With increasing demands for automated manufacturing and energy conservation, temperature control technology for triple-effect evaporators has become an important research direction in industrial control systems.

[0003] In a current technology, temperature sensors are typically used to collect temperature signals from each effect, which are then fed back to a controller. The controller adjusts the steam valve position according to preset control parameters to achieve closed-loop control of the evaporation temperature. Some improved methods incorporate operating parameters such as valve opening degree to filter the temperature signal before performing the adjustment calculation, thereby reducing the impact of sampling noise on control accuracy. However, in the actual operation of a triple-effect evaporator, the heat transfer between each effect exhibits a significant dynamic correlation, and the heat transfer coefficient and temperature response relationship change under different operating conditions. Most current technologies rely on fixed model parameters or single temperature feedback for control, lacking a unified representation of the multi-effect heat transfer coupling relationship. This makes it difficult to adaptively correct control parameters based on actual operating conditions, easily leading to accumulated temperature estimation errors, valve position adjustment lag, and decreased control accuracy.

[0004] Therefore, existing technologies suffer from the technical problems of low temperature regulation accuracy and insufficient dynamic adaptability in triple-effect evaporators. Summary of the Invention

[0005] This invention provides a temperature adaptive control method and system for a triple-effect evaporator, solving the technical problems of low temperature regulation accuracy and insufficient dynamic adaptability in existing triple-effect evaporators. In a first aspect, to solve the above-mentioned technical problems, this invention provides a temperature adaptive control method for a triple-effect evaporator, comprising: Acquire the temperature signals and steam valve position signals of each effect, and preprocess the temperature signals and steam valve position signals of each effect to obtain operating data; The thermal state value is obtained by using a preset Kalman observer to perform state estimation on the operating data; Based on the thermal state value and the preset reference state value, an error sequence is generated, and the error sequence is smoothed to obtain a smoothed error sequence. If the smoothing error sequence meets the preset update conditions, historical data of the corresponding time period is extracted as correction data, and the preset heat transfer parameters of the triple-effect evaporator are updated according to the correction data to obtain the updated heat transfer parameters. The preset Kalman observer is updated based on the updated heat transfer parameters and the state estimation is performed again to obtain the updated state result. Based on the updated status results and the preset temperature target, the valve position adjustment amount is determined using a preset control strategy, and a steam valve position command is generated based on the valve position adjustment amount to perform closed-loop temperature regulation on the triple-effect evaporator.

[0006] Secondly, the present invention provides a temperature adaptive control system for a triple-effect evaporator, comprising: The data processing module is used to acquire the temperature signals of each effect and the steam valve position signals, and to preprocess the temperature signals of each effect and the steam valve position signals to obtain operating data; The state observation module is used to perform state estimation on the operating data using a preset Kalman observer to obtain thermal state values; The error analysis module is used to generate an error sequence based on the thermal state value and a preset reference state value, and to smooth the error sequence to obtain a smoothed error sequence. The update determination module is used to extract historical data of the corresponding time period as correction data if the smoothing error sequence meets the preset update conditions, and update the preset heat transfer parameters of the triple-effect evaporator according to the correction data to obtain the updated heat transfer parameters. The parameter update module is used to update the preset Kalman observer according to the updated heat transfer parameters and re-evaluate the state to obtain the updated state result. The control output module is used to determine the valve position adjustment amount based on the updated status result and the preset temperature target using a preset control strategy, and generate a steam valve position command based on the valve position adjustment amount to perform closed-loop temperature regulation of the triple-effect evaporator.

[0007] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention acquires the temperature signals and steam valve position signals of each effect, preprocesses these signals, and then uses a preset Kalman observer to estimate the state of the operating data to obtain thermal state values. This provides a unified characterization of the dynamic coupling relationship in the multi-effect heat transfer process of the triple-effect evaporator. Based on these thermal state values, the temperature regulation process is no longer limited to a single temperature feedback or fixed model parameters, reducing the accumulation of state estimation deviations caused by changes in operating conditions and improving the accuracy of the temperature state characterization of the triple-effect evaporator.

[0008] (2) This invention generates an error sequence based on the thermal state value and a preset reference state value, and smooths the error sequence to obtain a smoothed error sequence. When the smoothed error sequence meets a preset update condition, historical data for the corresponding time period is extracted as correction data, and the preset heat transfer parameters of the triple-effect evaporator are updated based on the correction data to obtain the updated heat transfer parameters. This allows the preset heat transfer parameters to be dynamically corrected according to changes in the actual operating state. Compared with control methods based on fixed heat transfer parameters, this technical solution improves the consistency between the parameters and the current operating conditions and reduces the impact of heat transfer parameter mismatch on the temperature adjustment process.

[0009] (3) This invention updates the preset Kalman observer based on the updated heat transfer parameters and re-estimates the state to obtain the updated state result. Then, based on the updated state result and the preset temperature target, a preset control strategy is used to determine the valve position adjustment amount, and a steam valve position command is generated based on the valve position adjustment amount to perform closed-loop temperature regulation of the triple-effect evaporator, ensuring that the steam valve position regulation process is consistent with the updated thermal state characterization. This technical solution reduces the valve position regulation lag phenomenon and improves the dynamic adaptability and control accuracy of the triple-effect evaporator temperature control process.

[0010] (4) By introducing an intercept term during the process of correcting the initial valve position adjustment using a first-order linear regression correction algorithm, and combining offline statistical significance test and online amplitude limit constraint to screen the effectiveness and limit the amplitude of the intercept term, this invention can provide a basic bias amount to overcome the valve dead zone in the steady state stage when the rate of change of control deviation approaches zero, avoid ineffective adjustment lag caused by valve stem static friction, eliminate the steady-state micro-amplitude oscillation phenomenon in temperature closed-loop control, and solve the technical problems of reduced adjustment accuracy and insufficient stability of existing control strategies under the nonlinear characteristics of valve actuator dead zone. Attached Figure Description

[0011] Figure 1 This is a schematic flowchart of a temperature adaptive control method for a triple-effect evaporator provided in the first embodiment of the present invention; Figure 2 This is a temperature adaptive control system for a triple-effect evaporator provided in the second embodiment of the present invention. Structural diagram. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Reference Figure 1 The first embodiment of the present invention provides a temperature adaptive control method for a triple-effect evaporator, comprising the following steps: S11, acquire the effect temperature signal and steam valve position signal, and preprocess the effect temperature signal and steam valve position signal to obtain operating data; S12, use a preset Kalman observer to perform state estimation on the operating data to obtain thermal state values; S13, Based on the thermal state value and the preset reference state value, an error sequence is generated, and the error sequence is smoothed to obtain a smoothed error sequence; S14. If the smoothing error sequence meets the preset update conditions, extract the historical data of the corresponding time period as correction data, and update the preset heat transfer parameters of the triple-effect evaporator according to the correction data to obtain the updated heat transfer parameters. S15, Update the preset Kalman observer according to the updated heat transfer parameters and re-evaluate the state to obtain the updated state result; S16. Based on the updated state results and the preset temperature target, the valve position adjustment amount is determined using a preset control strategy, and a steam valve position command is generated based on the valve position adjustment amount to perform closed-loop temperature regulation on the triple-effect evaporator.

[0014] In step S11, the effect temperature signals and steam valve position signals are acquired, and the effect temperature signals and steam valve position signals are preprocessed to obtain operating data, including: Acquire the temperature signals of each effect from the temperature sensors and the steam valve position signal from the steam regulating valve, and arrange the temperature signals of each effect and the steam valve position signal in time sequence to generate the original data sequence. The original data sequence is subjected to sliding window mean filtering to obtain a filtered data sequence. The window mean and window standard deviation are calculated based on the filtered data sequence, and the anomaly detection interval is determined based on the window mean and window standard deviation. The data points in the filtered data sequence that fall outside the anomaly detection interval are interpolated and corrected to obtain the running data.

[0015] In one implementation, the temperature signals of each effect and the steam valve position signals are read from the field data acquisition module of the triple-effect evaporator. The temperature signals of each effect are acquired by temperature sensors installed in the first, second, and third effect evaporation chambers, and are respectively the temperature values ​​of the first, second, and third effects. The steam valve position signal is acquired by the valve position feedback unit of the steam regulating valve, representing the valve opening value of the steam regulating valve at the current sampling time. The valve opening value is expressed as a percentage, ranging from 0 to 100, where 0 corresponds to the steam regulating valve being fully closed and 100 corresponds to the steam regulating valve being fully open. For each sampling time, the first, second, and third effect temperature values ​​and the valve opening value corresponding to the same timestamp are combined into a sampling record, and written into a buffer queue in ascending order of timestamps to generate the original data sequence. If there are duplicate timestamps in the original data sequence, the sampling record with the later acquisition time is retained; if there are records written out of order, they are reordered according to timestamps before being written into the original data sequence.

[0016] The first-effect temperature value, second-effect temperature value, third-effect temperature value, and valve opening value in the original data sequence are subjected to center sliding window mean filtering along the time axis. For any data point corresponding to a sampling time, a local data segment of a preset window length is extracted centered on that data point. The arithmetic mean of all data points within the local data segment is calculated, and this arithmetic mean is used as the filtering result for that sampling time. If the data point is located at the boundary of the original data sequence, resulting in a local data segment length that is less than the preset window length, the arithmetic mean is calculated based on the actual available data points, and this arithmetic mean is used as the filtering result for that sampling time. The filtering results corresponding to each sampling time are arranged in the original chronological order to obtain the filtered data sequence.

[0017] The preset window length is determined using an offline calibration method. During the offline calibration phase, historical operating data of the triple-effect evaporator is collected. Multiple continuous data segments are extracted from this historical data, showing the steam valve position signal transitioning from one stable opening interval to another. For each continuous data segment, the temperature changes of the first, second, and third effect temperatures after the steam valve position signal changes are statistically analyzed. The duration corresponding to when the temperature change reaches 80% of the final stable change is determined as the characteristic response duration of the corresponding temperature channel under that continuous data segment. The sampling period is set to 1 second. The characteristic response duration of the three temperature channels under each continuous data segment is divided by the sampling period to obtain the number of candidate window points. The median of all candidate window points is rounded down and determined as the preset window length. For example, when the number of candidate window points obtained during the offline calibration phase is 5, 7, 7, 9, and 7 respectively, the median of 7 is determined as the preset window length.

[0018] For each data point in the filtered data sequence, reuse the local data segment corresponding to the data point when performing center sliding window mean filtering, and calculate the window mean and window standard deviation of all data points within the local data segment; subtract the product of the preset standard deviation multiple and the window standard deviation from the window mean to determine the lower boundary of the anomaly detection interval; add the product of the preset standard deviation multiple and the window standard deviation to the window mean to determine the upper boundary of the anomaly detection interval, thus obtaining the anomaly detection interval corresponding to the current data point.

[0019] The preset standard deviation multiple is determined statistically from historical stable operating data. Data segments where the steam valve position signal is within a stable range and the first, second, and third effect temperature values ​​change continuously and smoothly are selected from historical operating data to form a stable operating sample. A central sliding window mean filter is applied to the stable operating sample, and the deviation between each data point and the corresponding window mean is calculated. The deviation is divided by the corresponding window standard deviation to obtain a standardized deviation sequence. The 99th percentile of the absolute value of the standardized deviation sequence is calculated, and this percentile is rounded up to obtain the preset standard deviation multiple. For example, when the 99th percentile is 2.86, 3 is determined as the preset standard deviation multiple, resulting in the anomaly judgment interval corresponding to the current data point.

[0020] In one implementation, for a single abnormal data point, the two closest normal data points before and after the abnormal data point are found. Linear interpolation is performed based on the relative position of the abnormal data point's timestamp within the two normal data point timestamps to obtain the corrected value for the abnormal data point. For an abnormal data segment consisting of multiple consecutive abnormal data points, the nearest normal data point before the beginning of the abnormal data segment and the nearest normal data point after the end of the abnormal data segment are used as interpolation boundary points. The corresponding corrected value is calculated point-by-point based on the relative position of each timestamp within the abnormal data segment within the time interval of the two interpolation boundary points. If the abnormal data segment is located at the beginning of the filtered data sequence, the value of the nearest normal data point after it is used to fill the gap; if the abnormal data segment is located at the end of the filtered data sequence, the value of the nearest normal data point before it is used to fill the gap. After correcting all abnormal data points, the corrected first-effect temperature values, second-effect temperature values, third-effect temperature values, and valve opening feedback values ​​are recombine in chronological order to obtain the operating data.

[0021] In step S12, a preset Kalman observer is used to perform state estimation on the operating data to obtain thermal state values, including: Based on the operational data, the heat transfer temperature difference characteristics and steam valve position change characteristics between each effect are extracted to construct a system state vector; Calculate the prior state prediction value based on the system state vector and the steam valve position change characteristics; The deviation is calculated based on the prior state prediction value and the running data to obtain the state deviation. Based on the state deviation, the prior state prediction value is corrected and updated using a preset Kalman observer to obtain the posterior state estimate value. Extract the state component representing the amount of heat storage inside the triple-effect evaporator from the posterior state estimate, and determine the extracted state component as the thermal state value.

[0022] In one implementation, the preset Kalman observer is a Kalman state observer constructed based on the heat dissipation equilibrium relationship of the triple-effect evaporator. The preset Kalman observer is constructed using offline calibration data before the system goes online. Historical operating data of the triple-effect evaporator under multiple sets of different steam valve opening ranges are collected. After the aforementioned preprocessing, the operating data is obtained. A heat dissipation equilibrium identification dataset is constructed based on the temperature change and valve position signal change at adjacent sampling times. Least square fitting is performed on the equivalent heat transfer coefficients and total equivalent heat capacity parameters between each effect. The parameter whose sum of the squared residuals between the one-step predicted temperature value and the actual temperature value is minimized is determined as the preset heat transfer parameter.

[0023] The heat dissipation balance equation for the triple-effect evaporator is constructed based on the energy conservation relationship during evaporator operation. The sampling period is fixed at 1 second, consistent with the data acquisition time interval in the embodiment. The state variables involved in the construction process are, in order, the temperature difference between the first and second effect evaporator chambers, the temperature difference between the second and third effect evaporator chambers, the temperature difference between the first and third effect evaporator chambers, and the heat accumulation inside the triple-effect evaporator. The heat transfer parameters involved are, in order, the equivalent heat transfer coefficient between the first and second effects, the equivalent heat transfer coefficient between the second and third effects, the equivalent heat transfer coefficient between the first and third effects, and the total equivalent heat capacity of the triple-effect evaporator. The control input is the steam valve position change characteristic, i.e., the steam valve opening value at the current sampling moment. The continuous domain heat balance equation is as follows: The meanings of each letter symbol are as follows. Total equivalent heat capacity of evaporator (unit: ); Evaporator average temperature equivalent (i.e., original) (Unit: ℃) : Saturated temperature of live steam (unit: °C); Actual equivalent heat transfer coefficient for single-effect / double-effect / triple-effect systems (unit: ); Steam valve position opening (dimensionless, 0~100%). : Thermal power per unit valve opening of live steam (unit: ).

[0024] The discrete-domain heat balance equation is discretized based on a 1s sampling period, and its expression is as follows: .

[0025] And it satisfies: The formula symbols are represented as follows: The discrete recursive relationship of the heat transfer temperature difference between each effect is consistent with the temperature transfer characteristics of the evaporator. The heat transfer temperature difference between any two effects at the current sampling time is equal to the corresponding heat transfer temperature difference at the previous sampling time, plus the difference in the temperature change of each effect. The temperature change is calculated by multiplying the ratio of the heat transfer rate of the corresponding effect to the equivalent heat capacity by the sampling period. Based on this relationship, combined with the aforementioned discrete recursive relationship of heat accumulation, a complete discrete state-space equation is constructed: X(k)=A·X(k-1)+B·u(k-1), where X is the system state vector, composed of four state variables arranged in a fixed order, A is the state transition matrix, and B is the control input matrix.

[0026] The discretization derivation of the state transition matrix A is entirely based on the aforementioned thermal balance relationship. The matrix has a dimension of 4 rows and 4 columns, and the calculation of all elements is completed using heat transfer parameters obtained from offline calibration and a fixed sampling period. The diagonal elements of the first to third rows have a fixed value of 1, corresponding to the self-sustaining characteristic of the heat transfer temperature difference. The remaining off-diagonal elements are calculated by multiplying the ratio of the equivalent heat transfer coefficient between adjacent effects to the total equivalent heat capacity by the sampling period. The first three elements of the fourth row are, respectively, the equivalent heat transfer coefficients of the first and second effects multiplied by the sampling period, the equivalent heat transfer coefficients of the second and third effects multiplied by the sampling period, and the equivalent heat transfer coefficients of the first and third effects multiplied by the sampling period. The fourth element of the fourth row has a fixed value of 1, corresponding to the self-sustaining characteristic of heat accumulation.

[0027] The discretization derivation of the control input matrix B is completed simultaneously with the state transition matrix, and the matrix has a dimension of 4 rows and 1 column. The first three elements are calculated by multiplying the influence coefficient of the heat flow per unit valve position on the effective temperature by the sampling period, and the fourth element is the value of the heat flow per unit valve position multiplied by the sampling period, corresponding to the direct influence of the control input on the heat accumulation.

[0028] The offline calibration process for the preset heat transfer parameters is completely consistent with the original scheme. Thirty historical operating data segments with steam valve openings uniformly distributed within the range of 0 to 100 were collected. Each data segment contained 600 seconds of continuous operating data before and after the valve opening change, with a sampling interval of 1 second. A central sliding window mean filtering process was performed on each data segment, with a filtering window length of 7 sampling points and a step size of 1 sampling point. The filtered data was used to construct the identification dataset. The input to the identification dataset consisted of three heat transfer temperature differences and steam valve position change characteristics at each sampling time. The output was the heat accumulation at the next sampling time. The least squares algorithm was used to fit the equivalent heat transfer coefficient and the total equivalent heat capacity. The parameter combination with the minimum sum of the squared residuals of the one-step predicted temperature value and the measured temperature value of all 30 data segments was determined as the initial preset heat transfer parameters.

[0029] After calibration, the exemplary preset heat transfer parameters used in this embodiment have the following numerical ratios: the equivalent heat transfer coefficients between the first and second effects, the equivalent heat transfer coefficients between the second and third effects, and the equivalent heat transfer coefficients between the first and third effects are 30:28:42. The corresponding weights are calculated using the L1 normalization method, i.e., a single equivalent heat transfer coefficient divided by the sum of the three equivalent heat transfer coefficients, resulting in weights of 0.3, 0.28, and 0.42 respectively, which are completely consistent with the weighting rules for constructing the system state vector in the original scheme. The exemplary calibrated value for the total equivalent heat capacity is 12.5 kJ / ℃, and the exemplary calibrated value for the heat flow per unit valve position is 0.8 kJ / (%·s). These values ​​are calibrated based on operating parameters of a rated evaporator capacity of 5 t / h and a rated live steam pressure of 0.6 MPa.

[0030] Based on the exemplary calibration parameters and a 1-second sampling period, the exemplary state transition matrix A is obtained: Weights are only used for weighted accumulation during state vector construction and must not be directly substituted into the state matrix. Example control input matrix B: B=[[0.02·T],[0.015·T],[0.03·T],[ In the formula, T = 1s.

[0031] After the heat transfer parameters are updated online, the update process of the state transition matrix and the control input matrix is ​​completely consistent with the aforementioned derivation rules. The updated equivalent heat transfer coefficient, total equivalent heat capacity, and unit valve position heat flow are substituted into the calculation formula of the corresponding elements to recalculate the updated matrix values. The original matrix parameters in the preset Kalman observer are replaced to complete the state model update of the observer, which is completely consistent with the execution process of step S15 of the original scheme.

[0032] The observation matrix of the Kalman state observer is constructed based on the correspondence between state variables and measured temperatures. It has a dimension of 3 rows and 4 columns. The diagonal elements of the first three column vectors are fixed at 1, and the remaining elements are fixed at 0. All elements in the fourth column are fixed at 0, corresponding to the three measured heat transfer temperature differences as the observed output quantities, which is completely consistent with the construction rules of the actual observation vector in the original scheme. The process noise covariance matrix and the observation noise covariance matrix are obtained based on the residual statistics of offline calibration data. The process noise covariance matrix is ​​a 4th-order diagonal matrix, with exemplary values ​​of diagonal elements of 0.01, 0.01, 0.01, and 0.05, respectively. The observation noise covariance matrix is ​​a 3rd-order diagonal matrix, with exemplary values ​​of diagonal elements of 0.001, corresponding to the technical parameter of ±0.1℃ measurement accuracy of the temperature sensor. The Kalman gain matrix is ​​obtained by iteratively solving the discrete Riccati equation. The iteration termination condition is that the root mean square error of the state estimation in the corresponding step between two adjacent iterations decreases by less than 0.1℃, which is completely consistent with the iteration rules of the original scheme.

[0033] At the same timestamp, the heat transfer temperature difference between the first and second effects is obtained by subtracting the second effect temperature from the first effect temperature value; the heat transfer temperature difference between the second and third effects is obtained by subtracting the third effect temperature from the second effect temperature value; and the heat transfer temperature difference between the first and third effects is obtained by subtracting the third effect temperature from the first effect temperature value. The steam valve position signal at the current sampling time is subtracted from the steam valve position signal at the previous sampling time, and then divided by the sampling period to obtain the steam valve position change characteristic. Based on the relative magnitudes of the equivalent heat transfer coefficients between each effect in the preset heat transfer parameters, the heat transfer temperature difference characteristics between the first and second effects, the heat transfer temperature difference characteristics between the second and third effects, and the heat transfer temperature difference characteristics between the first and third effects are weighted, normalized, and weighted to obtain the initial state component characterizing the heat accumulation inside the three-effect evaporator. The heat transfer temperature difference characteristics between the first and second effects, the heat transfer temperature difference characteristics between the second and third effects, the heat transfer temperature difference characteristics between the first and third effects, and the initial state component are concatenated in a fixed order to construct the system state vector. The weights in the weighted accumulation are obtained by dividing the equivalent heat transfer coefficients between each effect in the preset heat transfer parameters by the sum of the equivalent heat transfer coefficients between all effects. Then, the heat transfer temperature difference characteristics between each effect are weighted and accumulated according to the weighted normalization result. When the values ​​of the equivalent heat transfer coefficients between the first and second effects, the second and third effects, and the first and third effects account for 30%, 28%, and 42% of the total equivalent heat transfer coefficients, respectively, the three heat transfer temperature difference characteristics are weighted and accumulated according to 30%, 28%, and 42%, respectively, and the accumulated result is determined as the initial state component.

[0034] In one implementation, the system state vector is written to the state input of a preset Kalman observer, and the steam valve position change characteristics are written to the control input of the preset Kalman observer. Discrete state propagation operations are performed based on the state transition matrix and the control input matrix to obtain the prior state prediction value corresponding to the current sampling time. The order of the state components in the prior state prediction value is consistent with the system state vector, corresponding sequentially to the predicted heat transfer temperature difference between the first and second effects, the predicted heat transfer temperature difference between the second and third effects, the predicted heat transfer temperature difference between the first and third effects, and the predicted component characterizing the heat accumulation inside the triple-effect evaporator.

[0035] Extract the first, second, and third effect temperature values ​​at the current timestamp from the operational data. Construct the actual observation vector based on the differences between the first and second effect temperature values, the differences between the second and third effect temperature values, and the differences between the first and third effect temperature values. Calculate the predicted observation vector based on the prior state prediction value and the observation matrix, with each component of the predicted observation vector corresponding one-to-one with each component of the actual observation vector. Subtract the actual observation vector from the predicted observation vector component by component to obtain the state bias. Perform matrix multiplication of the state bias with the Kalman gain matrix and add the result to the prior state prediction value to obtain the posterior state estimate.

[0036] In one implementation, the posterior state estimate output by a preset Kalman observer is received, and each state component in the posterior state estimate is parsed in the order of its arrangement. When the state component representing the heat accumulation inside the triple-effect evaporator is parsed, the value corresponding to the state component is read and separated and output. The separated state component value is written into the thermal state buffer unit corresponding to the current sampling time, and the written value is determined as the thermal state value.

[0037] In step S13, an error sequence is generated based on the thermal state value and a preset reference state value, and the error sequence is smoothed to obtain a smoothed error sequence, including: The thermal state deviation value corresponding to each sampling time is obtained by performing an algebraic difference operation on the preset reference state value and the thermal state value. The thermal state deviation values ​​are arranged sequentially according to the sampling time sequence to generate an error sequence; The error sequence is obtained by performing a moving average filter on the error sequence using a preset smoothing window.

[0038] In one implementation, historical operating data of the triple-effect evaporator during normal production is acquired. The absolute value of the difference between steam valve position signals at adjacent sampling times is calculated to obtain a valve position change amplitude sequence. The first quartile of the valve position change amplitude sequence is then calculated. Sampling times with valve position change amplitudes not greater than the first quartile are selected, and the corresponding historical thermal state values ​​are extracted to form a reference state sample set. The historical thermal state values ​​in the reference state sample set are sorted by numerical value, and the value in the middle position after sorting is extracted and determined as the preset reference state value. When the number of samples in the reference state sample set is even, the arithmetic mean of the two middle values ​​is taken, and this arithmetic mean is determined as the preset reference state value.

[0039] For any given sampling time, a pre-defined reference state value is used as the minuend, and the corresponding thermal state value is used as the subtrahend. Algebraic subtraction is then performed to obtain the thermal state deviation value for that sampling time. The thermal state deviation values ​​for each sampling time are then sequentially written into the error buffer in ascending order of sampling time to generate an error sequence.

[0040] Historical error sequences are generated from historical thermal state values ​​and preset reference state values. Candidate window lengths of 3, 5, 7, 9, and 11 sampling points are used, with a step size of 1 sampling point. Central moving average filtering is applied to each historical error sequence to obtain multiple sets of candidate smoothing results. The standard deviation of each set of candidate smoothing results is calculated, along with the reduction in the number of sign changes compared to the historical error sequence. The number of sign changes represents the number of times adjacent sampling points in the error sequence switch between positive and negative, or vice versa. Candidate window lengths are first sorted by standard deviation from smallest to largest, and then by the reduction in the number of sign changes from smallest to largest. The ranking of each candidate window length in the two sorts is summed to obtain a comprehensive ranking value. The candidate window length with the smallest comprehensive ranking value is determined as the preset smoothing window. If two or more candidate window lengths have the same comprehensive ranking value, the smaller value is selected as the preset smoothing window. For example, after sorting and calculating the candidate window lengths 3, 5, 7, 9, and 11, the window length 7 has the smallest comprehensive sorting value, so 7 is determined as the preset smooth window.

[0041] In one implementation, for each thermal state deviation value in the error sequence, the current value is used. A local error sequence containing the current value and with a length equal to a preset smoothing window is extracted, with a step size of 1 sampling point. The arithmetic mean of all thermal state deviation values ​​in the local error sequence is calculated, and this arithmetic mean is determined as the smoothing error value corresponding to the current sampling time. If the current sampling time is located at the beginning or end of the error sequence, resulting in the number of data points that can be extracted being less than the preset smoothing window, the arithmetic mean of all the thermal state deviation values ​​that can be extracted is calculated, and this arithmetic mean is determined as the smoothing error value corresponding to the current sampling time. The smoothing error values ​​corresponding to each sampling time are arranged sequentially according to the sampling time sequence to obtain the smoothing error sequence.

[0042] In step S14, if the smoothing error sequence meets the preset update conditions, historical data for the corresponding time period is extracted as correction data, and the preset heat transfer parameters of the triple-effect evaporator are updated according to the correction data to obtain the updated heat transfer parameters, including: The error amplitude within a continuous sampling period is statistically analyzed based on the smoothed error sequence. If the error amplitude meets the preset update conditions, historical data within a preset time period before the trigger time is extracted as the correction data. A heat transfer parameter identification model is constructed based on the correction data, and the heat transfer parameter identification model is iteratively solved using the recursive least squares algorithm to obtain the updated heat transfer parameters.

[0043] In one implementation, the absolute value of the smoothed error value at each sampling time in the smoothed error sequence is taken to obtain the error amplitude at that sampling time. Using the current sampling time as the endpoint, the error amplitudes corresponding to several sampling points over a preset consecutive sampling period are extracted to form the current statistical window. All error amplitudes in the current statistical window are sorted by value, and the value in the middle position after sorting is extracted and determined as the statistical value of the window error amplitude at the current sampling time. When the number of sampling points in the current statistical window is even, the arithmetic mean of the two middle values ​​is taken, and this arithmetic mean is determined as the statistical value of the window error amplitude at the current sampling time.

[0044] Historical smoothed error sequences during stable equipment operation and during periods of heat transfer parameter shift are acquired. The aforementioned error amplitude statistical process is performed on both types of historical smoothed error sequences to obtain a set of stable sample statistics and a set of shifted sample statistics. Candidate values ​​for the preset continuous sampling period number are set to 3, 5, 7, and 9. Candidate criterions for the window error amplitude statistics are set to the 90%, 95%, and 97.5% quantiles of the stable sample statistics set. For each candidate combination, the number of false triggers in the stable samples and the number of correct triggers in the shifted samples are counted. The combinations are first sorted by the number of false triggers from smallest to largest, and then second sorted by the number of correct triggers from largest to smallest. The rankings of each candidate combination in the two sorts are summed to obtain the corresponding comprehensive ranking value. The candidate combination with the smallest comprehensive ranking value is determined as the preset update condition. The preset update condition is composed of the preset continuous sampling period number and the criterion corresponding to the window error amplitude statistics. When the window error amplitude statistics at the current sampling time are not less than the criterion, the smoothed error sequence is determined to meet the preset update condition, and the current sampling time is determined as the trigger time. For example, the candidate combination with the smallest comprehensive ranking value when the number of consecutive sampling periods is 7 and the judgment benchmark is the 95th percentile is determined as the preset update condition.

[0045] Multiple candidate historical data segments of varying lengths are extracted from historical operational data. The candidate lengths for these segments are set to 5 min, 10 min, 15 min, and 20 min. For each candidate segment length, a heat transfer parameter identification model is constructed and a recursive least squares solution is performed to obtain the corresponding identification results. The coefficient of variation of the identified heat transfer parameters for each candidate segment length is calculated across different data segments, and the mean absolute error of the one-step temperature prediction for each segment length is also calculated. The segments are first sorted by coefficient of variation from smallest to largest, and then secondly sorted by the mean absolute error of the one-step temperature prediction from smallest to largest. The ranking of each candidate segment length in both sorts is summed to obtain a comprehensive ranking value. The candidate segment length with the smallest comprehensive ranking value is selected as the preset duration. For example, a candidate length of 10 min has the smallest comprehensive ranking value, and 10 min is selected as the preset duration.

[0046] In one implementation, a historical buffer is established during operation to continuously store the operational data and thermal state values ​​corresponding to each sampling moment according to the sampling time sequence. After the current sampling moment is determined as the trigger moment, the operational data and thermal state values ​​corresponding to all sampling points within a preset time period before the trigger moment are read from the historical buffer, arranged in chronological order, to obtain the correction data.

[0047] The first-effect temperature, second-effect temperature, third-effect temperature, steam valve position signal, and thermal state value are sequentially read from the calibration data at two adjacent sampling times. The thermal state value at the next sampling time is determined as the identification target value. The differences between the first-effect temperature and the second-effect temperature, the differences between the second-effect temperature and the third-effect temperature, the changes in the steam valve position signal, and the thermal state value at the previous sampling time are sequentially used to form the identification input vector. The linear mapping relationship between the identification input vector and the identification target value is determined as the heat transfer parameter identification model.

[0048] The linear relationship of the heat transfer parameter identification model is as follows: the target value at the current sampling time is equal to the first parameter to be identified multiplied by the heat transfer temperature difference between the first and second effects at the previous sampling time, plus the second parameter to be identified multiplied by the heat transfer temperature difference between the second and third effects at the previous sampling time, plus the third parameter to be identified multiplied by the heat transfer temperature difference between the first and third effects at the previous sampling time, plus the fourth parameter to be identified multiplied by the steam valve position change characteristic at the previous sampling time, plus the fifth parameter to be identified multiplied by the thermal state value at the previous sampling time. The expression of the heat transfer parameter identification model is as follows: z(k) = θ1·φ1(k) + θ2·φ2(k) + θ3·φ3(k) + θ4·φ4(k) + φ5(k), where the symbols in the formula have the following meanings: The regression vector used in the recursive least squares algorithm is constructed point-by-point according to the sampling time sequence. The elements in the vector are arranged in a fixed order, namely, the heat transfer temperature difference between the first and second effects at the previous sampling time, the heat transfer temperature difference between the second and third effects at the previous sampling time, the heat transfer temperature difference between the first and third effects at the previous sampling time, the steam valve position change characteristics at the previous sampling time, and the thermal state value at the previous sampling time. The purely numerical symbolic expression of the regression vector is φ(k)=[φ1(k),φ2(k),φ3(k),φ4(k),φ5(k)]^T, and the order of the elements is completely consistent with the aforementioned arrangement.

[0049] The order of elements in the parameter vector to be identified used in the recursive least squares algorithm corresponds one-to-one with the order of elements in the regression vector, namely, the first parameter to be identified, the second parameter to be identified, the third parameter to be identified, the fourth parameter to be identified, and the fifth parameter to be identified. The parameter vector to be identified is expressed as θ = [θ1, θ2, θ3, θ4, θ5]^T, where the first three parameters correspond to the product of the equivalent heat transfer coefficient between the three-effect transistors and the sampling period, the fourth parameter corresponds to the product of the heat flow per unit valve position and the sampling period, and the fifth parameter corresponds to the self-holding coefficient of the thermal state value. All parameters completely correspond to the heat transfer parameters in the aforementioned heat dissipation equilibrium equation. The input-output correspondence of the identification model perfectly matches the temporal sequence of the correction data. The input of a single identification sample is the regression vector constructed at the previous sampling time, corresponding to sampling time number k-1. The output of a single identification sample is the identification target value at the current sampling time, corresponding to sampling time number k. The temporal interval between input and output is fixed at one sampling period, perfectly consistent with the system sampling period of 1 second. The correction data extracts all sampling data within a preset time period before the trigger time. The preset time period is calibrated to 10 minutes using historical data, corresponding to 600 consecutive identification samples. The samples are arranged in ascending order of sampling time, with no disordered or duplicate data. The thermal state values ​​in the correction data are processed using a central sliding window mean filter. The filter window length is 7 sampling points, and the step size is 1 sampling point, perfectly consistent with the window parameters used in the error sequence smoothing process.

[0050] The iterative process of the recursive least squares algorithm is executed point by point according to the time sequence of the correction data. All preset parameters used are determined through offline calibration. The specific parameters and their setting basis are as follows: The forgetting factor, with an exemplary value of 0.99, is set based on the slow time-varying operating characteristics of the heat transfer parameters of the triple-effect evaporator. Four candidate values ​​of 0.95, 0.98, 0.99, and 0.995 are selected and used to identify 30 sets of offline calibration data segments. The evaluation criteria are the minimum one-step prediction mean square error and the lowest parameter fluctuation range of the identification results. 0.99 is determined to be the optimal value. The initial parameter values ​​are taken from the preset heat transfer parameters currently in use by the system. The example values ​​are 0.3, 0.28, 0.42, 0.8, and 1, which are completely consistent with the initial preset heat transfer parameters obtained from offline calibration. The initial covariance matrix is ​​a 5th order diagonal matrix, with the following example values ​​for the diagonals: 10, 10, 10, 10, 1. This setting is based on 10 times the variance of each parameter sample in the 30 sets of identification results during the offline calibration stage, ensuring the rapid convergence capability of the parameters in the initial stage of iteration. The iteration termination condition is that all 600 sets of samples in the correction data have been traversed and iterated. There are no additional early termination rules to ensure that all running information in the correction data participates in parameter updates.

[0051] The iterative execution steps of the recursive least squares algorithm completely correspond to the weight requirements of the original scheme. It reads the identification samples in the correction data group by group according to the sampling time sequence, calculates the predicted value of the identification target based on the parameter value at the current time, calculates the difference between the predicted value and the measured identification target value to obtain the identification residual, calculates the parameter correction amount based on the regression vector, identification residual, current covariance matrix, and forgetting factor, and adds the parameter correction amount to the current parameter value to obtain the updated parameter value. The covariance matrix is ​​updated synchronously, and the iteration process for the next group of samples begins. After all samples have been iterated, the final value of the parameter vector to be identified is the updated heat transfer parameter. The first three parameters, divided by the sampling period of 1 second, yield the updated equivalent heat transfer coefficient between the three effects. The fourth parameter, divided by the sampling period of 1 second, yields the updated heat flow rate per unit valve position. The fifth parameter is directly used for the recursive calculation of the thermal state. All parameter update rules completely correspond to the update process of the state observer in step S15 of the original scheme, with no characteristic bias.

[0052] The weights of the equivalent heat transfer coefficients between the three effects are calculated using the L1 normalization method. The updated three equivalent heat transfer coefficients are divided by the sum of the three coefficients to obtain the corresponding weights, which are used to construct the system state vector. The weight calculation rules are completely consistent with the system state vector construction rules in step S12 of the original scheme.

[0053] In one implementation, the currently used preset heat transfer parameters are determined as the initial parameter values. An initial covariance matrix is ​​constructed based on the sample variances of each heat transfer parameter in the multiple identification results during the offline calibration phase. The identification input vector and identification target value in the calibration data are read sequentially over time, and the corresponding target predicted value is calculated based on the current parameter value. The target predicted value is subtracted from the identification target value to obtain the identification residual at the current time. The parameter correction is calculated based on the identification input vector, identification residual, and the current covariance matrix. The parameter correction is added to the current parameter value, and the covariance matrix is ​​updated synchronously to obtain the parameter value for the next iteration. After all sampling points in the calibration data have been traversed, the parameter value after the final iteration is determined as the updated heat transfer parameters.

[0054] In step S15, the preset Kalman observer is updated according to the updated heat transfer parameters, and the state estimation is performed again to obtain the updated state result, including: The updated heat transfer parameters are input into the preset Kalman observer to obtain the updated state model; Based on the updated state model and the running data at the current sampling time, the state observer is re-estimated to obtain the posterior state estimate. The thermal state component is extracted from the posterior state estimate, and the extracted thermal state component is determined as the updated state result.

[0055] In one implementation, the equivalent heat transfer coefficients between the first and second effects, the second and third effects, and the total equivalent heat capacity are read. These parameters are then used to replace the corresponding heat transfer parameters in the state model used when constructing the preset Kalman observer. The replaced parameters are substituted into the heat dissipation equilibrium relationship of the three-effect evaporator. Combined with a 1-second sampling period, the transfer coefficients of each state component between adjacent sampling times are calculated. The transfer coefficients of each state component to the state component at the previous sampling time are written into the corresponding positions of the state transition matrix. The influence coefficients of the steam valve position signal change characteristics on each state component are written into the corresponding positions of the control input matrix. After updating the matrix elements, the updated state transition matrix and the updated control input matrix are written into the preset Kalman observer. Together with the unchanged observation matrix, they constitute the state model corresponding to the current sampling stage. This state model is then determined as the updated state model. The observation matrix is ​​determined by the correspondence between the actual observation vector and the state components. Since only the heat transfer parameters are updated without changing the composition of the state components, the observation matrix remains unchanged.

[0056] Extract the temperature values ​​of the first, second, and third effects, along with the steam valve position signal, from the operating data at the current sampling time. Using the method described earlier for constructing the system state vector, extract the heat transfer temperature difference characteristics and steam valve position change characteristics between each effect. Determine the posterior state estimate corresponding to the most recent sampling time before parameter updates as the initial re-estimation state. Input the steam valve position change characteristics into the updated state model to calculate the prior state prediction value corresponding to the current sampling time. Construct the actual observation vector based on the operating data at the current sampling time. Calculate the predicted observation vector based on the prior state prediction value and the observation matrix. Subtract the actual observation vector from the predicted observation vector component by component to obtain the state deviation corresponding to the current sampling time. Use the state deviation to correct and update the prior state prediction value to obtain the posterior state estimate.

[0057] In the updated state model, the order of the state components remains unchanged. The state component representing the heat accumulation inside the triple-effect evaporator is located from the posterior state estimate. The value corresponding to the state component is read and recorded with the current sampling time. This value is then determined as the updated state result.

[0058] In step S16, based on the updated state result and the preset temperature target, a preset control strategy is used to determine the valve position adjustment amount, and a steam valve position command is generated according to the valve position adjustment amount to perform closed-loop temperature regulation of the triple-effect evaporator, including: Based on the updated state results and the preset temperature target, the deviation is calculated to obtain the control deviation. The control deviation is adjusted using a proportional-integral control algorithm to obtain the initial valve position adjustment. The rate of change is calculated based on the control deviation, and the initial valve position adjustment is corrected using a first-order linear regression algorithm based on the rate of change to obtain the valve position adjustment; wherein, the first-order linear regression algorithm includes an intercept term; Based on the valve position adjustment amount, a limiting process is performed and a cycle verification is issued to obtain a steam valve position command. The temperature of the triple-effect evaporator is then adjusted in a closed loop according to the steam valve position command.

[0059] In one implementation, historical qualified batch data consistent with the current product specifications, feed concentration range, and process stage are obtained. The temperature values ​​of the corresponding controlled object within the stable operating range are extracted. The extracted temperature values ​​are sorted by numerical value, and the median is taken as the preset temperature target. When the sample size is even, the arithmetic mean of the two middle values ​​is taken as the preset temperature target.

[0060] A linear mapping relationship between the updated status results and the controlled object temperature value is established based on historical qualified batch data. The updated status results and controlled object temperature values ​​corresponding to the same sampling time are paired, and the linear mapping relationship from status value to temperature value is obtained using least squares fitting. During online operation, the updated status result at the current sampling time is substituted into the linear mapping relationship to obtain the current estimated temperature value. The preset temperature target is used as the minuend, and the current estimated temperature value is used as the subtrahend; algebraic subtraction is then performed to obtain the control deviation.

[0061] The proportional and integral coefficients are identified through historical closed-loop operation data. The control deviation, cumulative control deviation, and actual valve position adjustment for each sampling moment are extracted. The cumulative control deviation is obtained by accumulating the control deviations prior to the current sampling moment according to the sampling sequence. The actual valve position adjustment is obtained by subtracting the steam valve position signal from the previous sampling moment's steam valve position signal at the current sampling moment. Using the control deviation and cumulative control deviation as input samples and the actual valve position adjustment as output samples, a linear correspondence is calculated using a least-squares fitting method. The fitting coefficient corresponding to the control deviation is determined as the proportional coefficient, and the fitting coefficient corresponding to the cumulative control deviation is determined as the integral coefficient. The control deviation at the current sampling moment and the cumulative deviation integral value saved at the previous sampling moment are read. The control deviation is multiplied by the proportional coefficient to obtain the proportional adjustment component. The control deviation is multiplied by the 1-second sampling period and accumulated to the deviation integral value to obtain a new deviation integral value. The new deviation integral value is multiplied by the integral coefficient to obtain the integral adjustment component. The proportional adjustment component and the integral adjustment component are algebraically summed to obtain the initial valve position adjustment. The new deviation integral value is written back to the integral buffer unit. For example, the proportional coefficient is 1.2 and the integral coefficient is 0.05 obtained by fitting historical data.

[0062] The control deviation at the current sampling time is subtracted from the control deviation at the previous sampling time, and the difference is divided by the 1-second sampling period to obtain the control deviation change rate.

[0063] In the first-order linear regression algorithm, the slope coefficient and intercept coefficient are determined using historical closed-loop operation data. The initial valve position adjustment is calculated based on this data. The valve position correction requirement is obtained by subtracting the initial adjustment from the actual adjustment. The control deviation change rate is used as the input sample, and the valve position correction requirement is used as the output sample. The slope coefficient and intercept coefficient are obtained using a least-squares fitting method. During online operation, the control deviation change rate at the current sampling moment is substituted into the first-order linear regression model to obtain the valve position correction. The valve position correction is then algebraically summed with the initial valve position adjustment to obtain the final valve position adjustment.

[0064] In one implementation, the intercept term of the first-order linear regression algorithm is determined through offline statistical testing. If the statistical significance test result of the intercept term is less than the preset minimum valve increment threshold, then the intercept term is included in the correction of the initial valve adjustment amount.

[0065] It should be noted that industrial steam control valves generally exhibit mechanical dead zones and valve position preload characteristics. When the control deviation rate of change is zero and the PI controller output tends to stabilize, the actual operation of the control valve often requires crossing a minimum opening change to overcome the static friction of the valve stem and the hysteresis of the positioner. Therefore, an intercept coefficient is introduced. This intercept coefficient characterizes the minimum operating offset, and its physical meaning is, based on valve manufacturer data or field step tests, the minimum valve position increment required for the valve to generate an effective flow change from rest, for example, ±0.5% opening. Near the steady state, the intercept term provides a basic offset consistent with the direction of deviation change, allowing the valve to cross the dead zone earlier, thereby eliminating steady-state micro-oscillations and improving the smoothness of temperature control.

[0066] To avoid introducing meaningless noise bias into the intercept coefficient, this embodiment performs further verification when determining the intercept coefficient offline. Only when the p-value corresponding to the intercept coefficient, i.e., the statistical significance test result, is less than 0.05, is it included in the online control law; otherwise, the intercept coefficient is forcibly set to zero, and a first-order linear model with zero intercept is adopted. Simultaneously, based on the physical direction of heat transfer in the evaporator, the sign of the intercept coefficient must be consistent with the valve position action direction of the dominant heat transfer direction; if the fitting result has the opposite sign, it is determined to be data noise, and the intercept coefficient is set to zero.

[0067] When applied online, the valve position correction amount contributed by the intercept term is limited separately, and its absolute value does not exceed twice the valve dead zone setting value, in order to prevent excessive offset when the operating conditions change drastically.

[0068] The p-value, i.e. the statistical significance test result P-value, is calculated through a standard t-test process. In this embodiment, this calculation process does not need to be run in real time, but is part of the offline parameter tuning process.

[0069] The upper and lower limits of the valve position adjustment are determined based on the difference in steam valve position signals between adjacent sampling times in historical qualified batches. The 95th quantile of the difference sample is determined as the upper limit of the valve position adjustment, and the 5th quantile is determined as the lower limit. Using 1, 2, and 3 sampling periods as candidate issuance periods, the average absolute temperature deviation and the number of steam valve position signal changes are statistically analyzed for each candidate issuance period. The candidate issuance period with the smaller average absolute temperature deviation and the fewer steam valve position signal changes is determined as the issuance period. For example, the upper limit of the valve position adjustment is determined to be 5, the lower limit to be -5, and the issuance period to be 1 second.

[0070] During online operation, if the valve position adjustment is greater than the upper limit, the adjustment is truncated to the upper limit to obtain the limited valve position adjustment. If the adjustment is less than the lower limit, the adjustment is truncated to the lower limit to obtain the limited valve position adjustment. If the adjustment is within the upper or lower limit range, it is directly determined as the limited valve position adjustment. If the sampling point interval between the current sampling time and the sampling time corresponding to the previous issued steam valve position command is not less than the issuance cycle, the current steam valve position signal and the limited valve position adjustment are algebraically summed to obtain the steam valve position command. If the sampling point interval is less than the issuance cycle, the previous issued steam valve position command is determined as the steam valve position command corresponding to the current sampling time. The steam valve position command is written to the steam regulating valve execution port to perform closed-loop temperature regulation of the triple-effect evaporator.

[0071] In summary, this invention discloses a temperature adaptive control method for a triple-effect evaporator, which realizes dynamic correction and closed-loop adjustment of temperature control parameters during the operation of the triple-effect evaporator. It effectively solves the technical problems in the prior art, such as the difficulty of fixed heat transfer parameters adapting to changes in operating conditions, the accumulation of temperature state estimation deviations, and the lag in valve position adjustment.

[0072] Reference Figure 2 The second embodiment of the present invention provides a temperature adaptive control system for a triple-effect evaporator, comprising: The data processing module is used to acquire the temperature signals of each effect and the steam valve position signals, and to preprocess the temperature signals of each effect and the steam valve position signals to obtain operating data; The state observation module is used to perform state estimation on the operating data using a preset Kalman observer to obtain thermal state values; The error analysis module is used to generate an error sequence based on the thermal state value and a preset reference state value, and to smooth the error sequence to obtain a smoothed error sequence. The update determination module is used to extract historical data of the corresponding time period as correction data if the smoothing error sequence meets the preset update conditions, and update the preset heat transfer parameters of the triple-effect evaporator according to the correction data to obtain the updated heat transfer parameters. The parameter update module is used to update the preset Kalman observer according to the updated heat transfer parameters and re-evaluate the state to obtain the updated state result. The control output module is used to determine the valve position adjustment amount based on the updated status result and the preset temperature target using a preset control strategy, and generate a steam valve position command based on the valve position adjustment amount to perform closed-loop temperature regulation of the triple-effect evaporator.

[0073] It should be noted that the temperature adaptive control system for a triple-effect evaporator provided in this embodiment of the invention is used to execute all the process steps of the temperature adaptive control method for a triple-effect evaporator in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0074] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0075] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A temperature adaptive control method for a triple-effect evaporator, characterized in that, include: Acquire the temperature signals and steam valve position signals of each effect, and preprocess the temperature signals and steam valve position signals of each effect to obtain operating data; The thermal state value is obtained by using a preset Kalman observer to perform state estimation on the operating data; Based on the thermal state value and the preset reference state value, an error sequence is generated, and the error sequence is smoothed to obtain a smoothed error sequence. If the smoothing error sequence meets the preset update conditions, historical data of the corresponding time period is extracted as correction data, and the preset heat transfer parameters of the triple-effect evaporator are updated according to the correction data to obtain the updated heat transfer parameters. The preset Kalman observer is updated based on the updated heat transfer parameters and the state estimation is performed again to obtain the updated state result. Based on the updated status results and the preset temperature target, the valve position adjustment amount is determined using a preset control strategy, and a steam valve position command is generated based on the valve position adjustment amount to perform closed-loop temperature regulation on the triple-effect evaporator.

2. The temperature adaptive control method for a triple-effect evaporator according to claim 1, characterized in that, The process of acquiring the effect temperature signals and steam valve position signals, and preprocessing the effect temperature signals and steam valve position signals to obtain operating data includes: Acquire the temperature signals of each effect from the temperature sensors and the steam valve position feedback signal, and arrange the temperature signals of each effect and the steam valve position feedback signal in time sequence to generate the original data sequence. The original data sequence is subjected to sliding window mean filtering to obtain a filtered data sequence. The window mean and window standard deviation are calculated based on the filtered data sequence, and the anomaly detection interval is determined based on the window mean and window standard deviation. The data points in the filtered data sequence that fall outside the anomaly detection interval are interpolated and corrected to obtain the running data.

3. The temperature adaptive control method for a triple-effect evaporator according to claim 1, characterized in that, The step of using a preset Kalman observer to perform state estimation on the operating data to obtain thermal state values ​​includes: Based on the operational data, the heat transfer temperature difference characteristics and steam valve position change characteristics between each effect are extracted to construct a system state vector; Calculate the prior state prediction value based on the system state vector and the steam valve position change characteristics; The deviation is calculated based on the prior state prediction value and the running data to obtain the state deviation. Based on the state deviation, the prior state prediction value is corrected and updated using a preset Kalman observer to obtain the posterior state estimate value. Extract the state component representing the amount of heat storage inside the triple-effect evaporator from the posterior state estimate, and determine the extracted state component as the thermal state value.

4. The temperature adaptive control method for a triple-effect evaporator according to claim 1, characterized in that, The step of generating an error sequence based on the thermal state value and a preset reference state value, and smoothing the error sequence to obtain a smoothed error sequence, includes: The thermal state deviation value corresponding to each sampling time is obtained by performing an algebraic difference operation on the preset reference state value and the thermal state value. The thermal state deviation values ​​are arranged sequentially according to the sampling time sequence to generate an error sequence; The error sequence is obtained by performing a moving average filter on the error sequence using a preset smoothing window.

5. The temperature adaptive control method for a triple-effect evaporator according to claim 1, characterized in that, If the smoothed error sequence meets the preset update conditions, historical data for the corresponding time period is extracted as correction data, and the preset heat transfer parameters of the triple-effect evaporator are updated according to the correction data to obtain the updated heat transfer parameters, including: The error amplitude within a continuous sampling period is statistically analyzed based on the smoothed error sequence. If the error amplitude meets the preset update conditions, historical data within a preset time period before the trigger time is extracted as the correction data. A heat transfer parameter identification model is constructed based on the correction data, and the heat transfer parameter identification model is iteratively solved using the recursive least squares algorithm to obtain the updated heat transfer parameters.

6. The temperature adaptive control method for a triple-effect evaporator according to claim 1, characterized in that, The step of updating the preset Kalman observer based on the updated heat transfer parameters and re-estimating the state to obtain the updated state result includes: The updated heat transfer parameters are input into the preset Kalman observer to obtain the updated state model; Based on the updated state model and the running data at the current sampling time, the state observer is re-estimated to obtain the posterior state estimate. The thermal state component is extracted from the posterior state estimate, and the extracted thermal state component is determined as the updated state result.

7. The temperature adaptive control method for a triple-effect evaporator according to claim 1, characterized in that, The step of determining the valve position adjustment amount using a preset control strategy based on the updated state result and the preset temperature target, and generating a steam valve position command based on the valve position adjustment amount to perform closed-loop temperature regulation of the triple-effect evaporator includes: Based on the updated state results and the preset temperature target, the deviation is calculated to obtain the control deviation. The control deviation is adjusted using a proportional-integral control algorithm to obtain the initial valve position adjustment. The rate of change is calculated based on the control deviation, and the initial valve position adjustment is corrected using a first-order linear regression algorithm based on the rate of change to obtain the valve position adjustment; wherein, the first-order linear regression algorithm includes an intercept term; Based on the valve position adjustment amount, a limiting process is performed and a cycle verification is issued to obtain a steam valve position command. The temperature of the triple-effect evaporator is then adjusted in a closed loop according to the steam valve position command.

8. The temperature adaptive control method for a triple-effect evaporator according to claim 7, characterized in that, The intercept term of the first-order linear regression algorithm is determined through offline statistical testing. If the statistical significance test result of the intercept term is less than the preset minimum valve position increment threshold, the intercept term is included in the correction of the initial valve position adjustment amount.

9. A temperature adaptive control system for a triple-effect evaporator, characterized in that, include: The data processing module is used to acquire the temperature signals of each effect and the steam valve position signals, and to preprocess the temperature signals of each effect and the steam valve position signals to obtain operating data; The state observation module is used to perform state estimation on the operating data using a preset Kalman observer to obtain thermal state values; The error analysis module is used to generate an error sequence based on the thermal state value and a preset reference state value, and to smooth the error sequence to obtain a smoothed error sequence. The update determination module is used to extract historical data of the corresponding time period as correction data if the smoothing error sequence meets the preset update conditions, and update the preset heat transfer parameters of the triple-effect evaporator according to the correction data to obtain the updated heat transfer parameters. The parameter update module is used to update the preset Kalman observer according to the updated heat transfer parameters and re-evaluate the state to obtain the updated state result. The control output module is used to determine the valve position adjustment amount based on the updated status result and the preset temperature target using a preset control strategy, and generate a steam valve position command based on the valve position adjustment amount to perform closed-loop temperature regulation of the triple-effect evaporator.

Images