A DCS-based coordination control method for chemical production processes

Abstract

The present application belongs to the field of chemical engineering automatic control technology, and particularly relates to a chemical production process coordinated control method based on DCS, which comprises the following steps: collecting the operation data of the control loop participating in coordination, and sequentially splicing the operation data according to the process flow direction to form a unit contribution vector; aligning and rearranging the unit contribution vector along the time axis according to the material transmission lag to obtain a two-dimensional matrix, calculating the correlation between the cumulative contribution curve and the reference template to obtain the drift significance; rewriting the corresponding gain column in the static allocation matrix by using a correction factor to generate an adaptive allocation instruction; at the same time of issuing the adaptive allocation instruction, calculating a feedforward compensation signal based on the correction factor, and injecting the feedforward compensation signal into the feedforward channel of the downstream rectifying unit with the material transmission lag as the waiting period. The present application can capture the implicit integral mismatch in advance, and realize the smooth transition of the material holding amount of the whole device.

Description

Technical Field

[0001] This invention relates to the field of chemical automation control technology. More specifically, this invention relates to a DCS-based coordinated control method for chemical production processes. Background Technology

[0002] Distributed control systems (DCS) serve as the automation hub of continuous chemical production plants, responsible for real-time regulation of key process variables such as reactor temperature and distillation column pressure. In fine chemical and new material production plants, multiple reaction units, heat exchange units, and separation units are strongly interconnected through material pipelines and energy loops. Load fluctuations in any unit are transmitted downstream along the process flow. Therefore, DCS not only undertakes single-loop PID regulation but also needs to coordinate and allocate the operating variables among multiple units. To maintain the material and energy balance of the entire plant, existing technologies typically employ a hierarchical coordinated control architecture. The bottom layer of this architecture relies on independent control loops to perform single-variable tracking, while the upper layer distributes the total load command to each reaction unit according to a fixed proportion based on a pre-tuned static gain matrix. This static allocation architecture can maintain stable system operation under ideal steady-state conditions. However, in actual intermittent and continuous mixed production scenarios, batch switching of raw materials often causes slight drifts in the feed composition, leading to deviations in the actual gain of the bottom control loop from the initial tuning state and resulting in fluctuations in the output performance of the bottom execution units.

[0003] To address the system fluctuation issue caused by deviations in the actual gain of the aforementioned underlying units, the improved existing technology introduces feedforward cascade regulation and interference compensation into the hierarchical architecture. This scheme configures a decoupling matrix at the control layer to attempt to counteract dynamic coupling interference between adjacent loops, while deploying loop alarms based on transient amplitude and conventional oscillation detection algorithms at the system monitoring layer. When the output deviation of a local unit exceeds a preset safety narrowband threshold, the monitoring system triggers an over-limit alarm. Subsequently, operators manually intervene and adjust key parameters such as the reflux ratio of the downstream distillation column based on the alarm prompts, restoring the material balance of the entire unit through large-scale manual compensation operations afterward.

[0004] While improved existing technologies can handle some sudden disturbances, they have limitations in dealing with implicit mismatches caused by slow, in-direction gain drift. Because the underlying loop of the preceding reaction unit can still control transient errors within a narrow band of its set value during gain drift, the control state appears healthy. The upper-level static allocation matrix continues to issue load commands to the unit according to a fixed nominal gain. This small and persistent in-direction mismatch between the control command and the actual discharge flow rate constitutes an objective fact that existing technologies cannot recognize. This small mismatch is amplified step-by-step along the process flow to the downstream heat exchange and separation units as the material holding volume increases. Existing alarm systems based on a one-dimensional instantaneous snapshot perspective are only sensitive to absolute amplitude and lack the ability to discriminate against material transport lag and slow, monotonous accumulation trends, resulting in the system losing its ability to proactively detect implicit integral mismatches. Therefore, this monitoring and allocation mechanism inevitably leads to the accumulation and outbreak of mismatch errors at the top condensate reflux tank of the downstream distillation column in the form of a slow and monotonous rise or fall in liquid level. This forces operators to only be able to make delayed and passive large-scale interventions after intermittent over-limit alarms, ultimately resulting in the serious consequences of frequent over-limits in the purity of the top product and instability of the entire unit. Summary of the Invention

[0005] To address the technical problem of implicit integral mismatch in downstream material inventory caused by slow drift of unit loop gain in the static allocation architecture of existing technologies, this invention provides a DCS-based coordinated control method for chemical production processes.

[0006] In a first aspect, the present invention provides a DCS-based coordinated control method for chemical production processes, comprising: S1, collecting control loop operation data involved in coordination, calculating basic descriptive quantities of the control loop operation data based on a sliding time window and performing standardization processing, and splicing them according to the process flow order to form a unit contribution vector; S2, aligning and rearranging the unit contribution vectors obliquely along the time axis according to the material transport lag to obtain a two-dimensional matrix, accumulating the two-dimensional matrix along the column direction to obtain a cumulative contribution curve, and calculating the correlation value between the cumulative contribution curve and the reference template as the drift significance; S3, for drift units whose drift significance exceeds the trigger threshold, calculating a correction factor based on the drift significance, and rewriting the corresponding gain column in the static allocation matrix using the correction factor to generate an adaptive allocation instruction; S4, while issuing the adaptive allocation instruction, calculating a feedforward compensation signal based on the correction factor, and injecting the feedforward compensation signal into the feedforward channel of the downstream distillation unit with the material transport lag as the waiting period.

[0007] This invention identifies and extracts the slow drift characteristics of control loop gain that cannot be captured by conventional alarms by using multi-dimensional sliding time window features and two-dimensional matrix analysis along the physical time sequence of material transport. Combined with dynamic allocation matrix directional rewriting and synchronous calculation of the injected feedforward compensation signal, it can intervene in advance when the mismatch effect reaches the downstream front end, solving the problem of implicit integral mismatch caused by the lack of time sequence accumulation perception in static architecture, and realizing a smooth and autonomous transition of the material inventory of the entire device.

[0008] Preferably, the acquisition of control loop operation data involved in coordination, and the calculation of basic descriptive quantities of the control loop operation data based on a sliding time window and the standardization processing, include: acquiring the setpoint, process value, controller output, and valve position feedback of the control loop involved in coordination; calculating the offset of the controller output, the steady-state residual of the process value, and the hysteresis deviation of the valve position feedback within the sliding time window; and performing a division operation on the offset, steady-state residual, and hysteresis deviation using historical standard deviation to complete the standardization processing, obtaining a dimensionless operating state triplet as the basic descriptive quantity.

[0009] This invention uses the standardized splicing of ternary descriptors within a sliding time window to uniformly map the operating data of control loops from different process units with different dimensions into a dimensionless state that can be compared laterally. This provides a complete data foundation for identifying unit gain drift on the entire process scale and eliminates the problem that the contributions of each unit cannot be directly compared due to incompatible dimensions.

[0010] Preferably, the step of forming a unit contribution vector by splicing together the control loops in the process flow order includes: extracting the basic descriptive quantities of all control loops belonging to the same process unit; and vertically stacking and splicing the basic descriptive quantities of all control loops belonging to the same process unit in the order of the process flow to form a unit contribution vector that reflects the net extra output or intake of materials and energy of the process unit.

[0011] Preferably, the step of obliquely aligning and rearranging the unit contribution vectors along the time axis according to the material transport lag to obtain a two-dimensional matrix includes: obtaining the pipeline volume and current flow rate from the upstream unit to the downstream unit; dividing the pipeline volume by the current flow rate to calculate the material transport lag; aligning the unit contribution vector of the upstream unit at the time corresponding to the material transport lag time before the current time of the downstream unit with the unit contribution vector of the downstream unit at the current time in the same column to obtain a two-dimensional matrix in which the row direction represents the spatial position and the column direction represents the time trajectory.

[0012] This invention transforms the originally isolated one-dimensional snapshot state into a two-dimensional cumulative trajectory perspective according to the physical time sequence of material transport, so that the weak but continuous same-direction error upstream can be aligned and revealed on the real spatial flow time axis, enhancing the sensitivity to capture hidden mismatches.

[0013] Preferably, the calculation of the correlation value between the cumulative contribution curve and the reference template as the drift significance includes: constructing a linearly increasing sequence with equal step size as a reference template for pure monotonic climbing; calculating the correlation value between the cumulative contribution curve and the reference template using the Pearson correlation coefficient algorithm, and using the correlation value as the drift significance characterizing the degree of latent mismatch.

[0014] Preferably, after calculating the correlation value between the cumulative contribution curve and the reference template as the drift significance, the method further includes: collecting historical sequences of drift significance of each process unit under all healthy operating conditions during the device commissioning and adjustment period; calculating the mean and standard deviation of the historical sequences; adding the product of the mean and standard deviation of the historical sequences and a preset multiple to obtain the trigger boundary; and using the trigger boundary as the trigger threshold to determine whether the drift significance exceeds the limit.

[0015] Preferably, the step of calculating the correction factor based on drift significance includes: obtaining the mean absolute value of each component of the unit contribution vector of the drifting unit within the current sliding window, and obtaining the contribution baseline statistically archived during the data acquisition phase; calculating the ratio of the mean absolute value to the contribution baseline, truncating the ratio to 1 when the ratio is greater than 1, and retaining the original ratio calculation result when the ratio is less than or equal to 1; and multiplying the drift significance by the truncated or retained ratio to obtain the correction factor used to constrain the pullback depth.

[0016] This invention introduces a bounded correction factor constrained by both drift significance and contribution ratio. Without affecting the allocation relationship of healthy units, it implements directional adaptive withdrawal and replenishment of the load allocation share of identified drifting units, suppressing the continuous accumulation of material mismatch in the entire unit due to unit gain drift. At the same time, the bounded design of the correction factor eliminates the possibility of physical illegal states such as command reversal.

[0017] Preferably, the step of rewriting the corresponding gain column in the static allocation matrix using a correction factor to generate an adaptive allocation instruction includes: proportionally reducing the elements of the fixed gain column corresponding to the drifting unit in the static allocation matrix using a correction factor; normalizing the reduced load share according to the elements of the fixed gain column of the remaining healthy units and redistributing it; and generating an adaptive allocation instruction to be issued to each process unit by multiplying the rewritten gain column elements with the total load instruction of the entire unit.

[0018] This invention precisely limits the adaptive allocation correction action to the source of mismatch, while compensating for the deviation of the mismatch unit and maintaining the conservation of the overall control load of the entire device, thus minimizing the impact of local coordination adjustments on the other steady-state healthy loops of the entire device.

[0019] Preferably, the step of calculating the feedforward compensation signal based on the correction factor includes: reading the material balance conversion factor from the drift unit to the downstream distillation unit; extracting the absolute value mean of the unit contribution vector of the drift unit within the current sliding time window; and multiplying the material balance conversion factor, the correction factor, and the absolute value mean together to obtain the feedforward compensation signal.

[0020] Preferably, the step of injecting a feedforward compensation signal into the feedforward channel of the downstream distillation unit with the material transport lag as the waiting period includes: starting a timer with the material transport lag time as the waiting period; superimposing the feedforward compensation signal into the feedforward channels of the reflux loop and the reboiler steam flow loop of the downstream distillation unit at the beginning of the waiting period; and synchronously returning the feedforward compensation signal to zero after the waiting period ends, so that each control loop of the downstream distillation unit returns to the normal automatic adjustment state.

[0021] This invention injects a compensation signal based on material balance calculation into the feedforward channel of the downstream distillation column at the same time as the upstream correction command is issued, and uses the pure lag in material transportation as the effective timing reference for the feedforward signal, thereby eliminating the risk of continuous liquid level drift and human error caused by the upstream correction effect not reaching the downstream during the transmission gap period.

[0022] The technical solution of the present invention has the following beneficial technical effects: This invention collects control loop operation data involved in coordination, calculates the basic descriptive quantities of the control loop operation data based on a sliding time window and performs standardization processing, and splices them in the process flow order to form a unit contribution vector, so that the control loop operation data is standardized and spliced, providing a data foundation for calculating correlation.

[0023] A two-dimensional matrix is ​​obtained by obliquely aligning and rearranging the unit contribution vectors along the time axis according to the material transport lag. The cumulative contribution curve is obtained by accumulating the two-dimensional matrix along the column direction. The correlation between the cumulative contribution curve and the reference template is calculated to obtain the drift significance. The drift significance can be accurately obtained based on the correlation, thereby obtaining the drift unit.

[0024] For drifting units whose drift significance exceeds the trigger threshold, a correction factor is calculated based on the drift significance. The corresponding gain column in the static allocation matrix is ​​rewritten using the correction factor to generate an adaptive allocation instruction. This method can rewrite the corresponding gain column in the static allocation matrix based on the correction factor, thus preventing the drifting unit from continuously causing the drifting of the downstream distillation unit.

[0025] By calculating the feedforward compensation signal based on the correction factor while issuing the adaptive allocation command, and injecting the feedforward compensation signal into the feedforward channel of the downstream distillation unit with the material transport lag as the waiting period, the downstream distillation unit can be stabilized by combining the material transport lag injection into the feedforward channel of the downstream distillation unit. Attached Figure Description

[0026] Figure 1 This is a flowchart of a DCS-based coordinated control method for chemical production processes; Figure 2 This is a comparison chart of the cumulative contribution curve and the reference template of a DCS-based coordinated control method for chemical production processes in an embodiment of the present invention. Detailed Implementation

[0027] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0028] This invention discloses a DCS-based coordinated control method for chemical production processes, referring to... Figure 1 This includes the following steps: S1. Collect loop data and splice them to form the unit contribution vector.

[0029] In an optional embodiment, a coordination layer data aggregation service is added to the distributed control system engineer station. This service retrieves all control loop operation data involved in the coordination from each control station at a fixed 5-second interval via the OPC UA interface. The retrieved data items cover set values. Process value Controller output and the valve position feedback of the corresponding actuator. The scope of capture covers the feed flow loop of the front-end reaction unit, the reactor temperature loop, the heat transfer medium flow loop of the middle heat exchange unit, and all loops related to the material and energy flow of the entire unit, such as the reflux flow, reboiler steam flow, and column top and bottom liquid levels of the back-end distillation column, ensuring that the coordination layer has a complete grasp of the entire process operation trajectory.

[0030] Furthermore, the data aggregation service maintains a sliding time window for each acquisition loop, with the window length being 3 to 5 times the dominant time constant of that loop. The dominant time constant is directly read from the existing proportional-integral-derivative (PID) tuning parameters in the distributed control system configuration, eliminating the need for additional loop identification tests. Within the sliding time window, the data aggregation service calculates three basic descriptors for each loop: The offset relative to its time window mean Compared to The steady-state residuals, and Compared to The lag bias. Due to and Dimensionless percentages from 0 to 100, while For process quantities with engineering units, such as flow rate in t / h and temperature in ℃, direct splicing would result in incomparable numerical scales for each component. Therefore, standardization is necessary for each item before splicing. Using the historical standard deviation of the corresponding descriptive quantities during the commissioning period of the circuit as divisors, the three items are converted into standardized deviations of pure numbers before splicing. The historical standard deviation during the commissioning period is objectively obtained from the distributed control system's historical database without manual assignment. After standardization, the three descriptive quantities combine to form a dimensionless operating state triplet for the circuit in the current time window, serving as the basic descriptive quantities. ; in, The controller outputs the offset. For a moment The controller output, The controller outputs the average value. Output the historical standard deviation for the controller.

[0031] ; in, The steady-state residual of the process value; For a moment The process value; For a moment The set value; This represents the historical standard deviation of the process value.

[0032] ; in, This refers to valve position hysteresis deviation; For a moment Valve position feedback; This represents the historical standard deviation of valve position feedback.

[0033] Next, the basic descriptive quantities of all control loops belonging to the same process unit are extracted. The dimensionless operating state triples of all loops belonging to the same process unit are then vertically stacked and spliced ​​in the process flow order to form the unit contribution vector. ,in For unit numbering, This is the end time of the time window. This represents the active output of the executing agency beyond the equilibrium state. A positive value indicates that the process variable is consistently higher than the set value. A positive value indicates that the actual valve opening is ahead of the controller's expected output. When all three are positively biased in the same direction, they comprehensively reflect the net additional material and energy output of the unit relative to the historical equilibrium state within the current time window. The unit contribution vector as a whole serves as a multi-dimensional standardized representation of the unit's net output or intake share. It reflects the degree of deviation of the current active output of the implementing agency from its own time window average; It reflects the residual steady-state tracking error between process variables and setpoints; The dynamic lag between valve position feedback and controller output is captured, and these three factors together constitute a multi-dimensional characterization of the actual output efficiency of the loop. No single piece of information is sufficient to independently determine the true contribution state of the unit. (The entire device...) The contribution vectors of each unit are stacked vertically in the order of flow to form a unit contribution snapshot matrix that is refreshed in real time, which serves as the original input for subsequent coordination processing.

[0034] During the commissioning period of the device, the data aggregation service synchronously performs historical statistics on the absolute mean of each component of the contribution vector of each unit, and archives the statistical mean as the contribution baseline of the corresponding unit. The contribution baseline is obtained from the historical database of the distributed control system through objective statistics and is used for subsequent correction factor calculation.

[0035] It should be noted that OPC UA is only an optional data interface protocol. Those skilled in the art can choose other interface methods such as Modbus TCP, DDE, or vendor-specific communication protocols according to the distributed control system manufacturer and the field network conditions. This invention does not impose any restrictions on this.

[0036] S2. Obtain the two-dimensional matrix and drift significance, and determine the trigger threshold.

[0037] In an optional embodiment, the traditional coordination layer only focuses on whether the lateral proportion of the contributions of each unit at the same moment conforms to the static allocation matrix. This one-dimensional snapshot perspective is insensitive to the slow, monotonically accumulating micro-drifts. Although the loop appears to be tracking normally within its own set value narrow band, the actual product flow rate ejected by the loop has been continuously deviating from the expected value in the same direction. The magnitude of this deviation on any instantaneous snapshot is extremely small, far below the conventional alarm threshold. Only by tracking the entire contribution trajectory of the same batch of material in each unit along the material transfer direction can the small, continuous deviation be accumulated and displayed on the time axis. To this end, this step introduces a dimensional transformation process, which aligns and rearranges the constructed unit contribution vector obliquely along the time axis according to the material transfer lag in the process flow direction: obtaining the pipeline volume and current flow rate from the upstream unit to the downstream unit, and calculating the material transfer lag by dividing the pipeline volume by the current flow rate. ;The upstream unit will advance the material transfer lag time The unit contribution vector and the downstream unit at the current time The unit contribution vectors are aligned in the same column; the final result is a two-dimensional matrix where the row direction represents the spatial location and the column direction represents the time trajectory. .

[0038] In a two-dimensional matrix The cumulative contribution curve for each unit is obtained by summing the data row by row along the column direction, within the same batch of materials. If a unit's PID loop is healthy but its underlying gain drifts, it will only appear as a small fluctuation near the mean in the original one-dimensional view of the unit contribution snapshot matrix, which cannot be detected by the threshold alarm; however, in the diagonally aligned two-dimensional matrix... In this case, the cumulative contribution curve of this unit will show a monotonous, consistent, and stable climbing pattern, which is significantly geometrically different from the zero-mean random fluctuations of other healthy units.

[0039] Next, a linearly increasing sequence with equal step size is constructed for the cumulative contribution curve of each unit as a reference template for pure monotonic climbing. The Pearson correlation coefficient algorithm is used to calculate the correlation between the cumulative contribution curve and the reference template, and the correlation value is used as the drift significance to characterize the degree of latent mismatch. . The closer it is to 1, the more likely the unit has a persistent, unidirectional hidden mismatch. The closer the value is to 0, the more normal the random fluctuation of the unit is. The correlation coefficient is used here instead of directly calculating the slope because the correlation coefficient is naturally insensitive to the amplitude and can identify slow drifts with extremely small slopes but consistent directions, which is the target pattern that this invention aims to capture. If the slope is used instead, the extremely small drift slope will be submerged in numerical noise, leading to missed detections.

[0040] Then, during the device commissioning and commissioning period, historical sequences of drift significance for each unit under all healthy operating conditions are collected; the mean and standard deviation of the historical sequences are statistically analyzed, and the trigger boundary is calculated in conjunction with a preset multiple. The trigger boundary is used as the trigger threshold to determine whether the drift significance exceeds the limit. The trigger boundary satisfies the following relationship: ; in, To trigger the boundary, The historical series mean. For preset multiples, This represents the standard deviation of the historical series. Preset multiple. The engineering trade-off between the costs of false triggering and missed detection of the device is determined by actual measured data during the commissioning phase, rather than by a given fixed value; for example, The value of is 3. , and finally All data are archived in the distributed control system's historical database using objective statistical data, and are updated in the same way as the device enters a new steady-state operating condition, without introducing any artificial empirical constants.

[0041] It should be noted that those skilled in the art can adjust the form of the reference template according to the actual working conditions. For example, for units known to have periodic fluctuations, the reference template can be replaced with a composite template with a monotonic trend superimposed with periodic components. This invention does not impose any hard restrictions on the specific selection of the template form.

[0042] Thus, by converting the one-dimensional instantaneous snapshot perspective into a two-dimensional cumulative perspective that tracks material batches, and combining this with the Pearson correlation coefficient for highly sensitive identification of monotonic ramp patterns, mismatched units that are healthy but whose gain has slowly drifted can be accurately separated from the normal fluctuation background before the downstream distillation column level enters the alarm range, thereby achieving proactive capture of implicit integral mismatch.

[0043] S3. Calculate the correction factor based on the drift significance and rewrite the gain column.

[0044] In an optional embodiment, the elements of the conventional static allocation matrix represent the share that a unit should bear when the total load command changes by one unit. This is calibrated and fixed through steady-state testing during plant commissioning. The original static proportional allocation satisfies the following relationship: ; in, For a moment Distributed to unit Load command, This refers to the solidified gain column element corresponding to this unit in the static allocation matrix. For a moment The total load command for the entire unit. Because... Once tuned, it becomes a constant. When the actual output efficiency of a cell deviates from the nominal value at the tuning time due to gain drift, the static formula cannot detect and automatically compensate for the output efficiency deviation. As a result, it continuously issues load commands that do not match the actual response capability to the drifted process cells, causing implicit mismatch to be amplified step by step.

[0045] To address the aforementioned problems, this invention focuses on identifying the significance of drift. Exceeding the trigger threshold drift unit The original static allocation relation is improved into an adaptive allocation based on drift significance: ; in, For a moment Distributed to unit Load command, For drift unit The corresponding solidified gain column element in the static allocation matrix For a moment Correction factor; For a moment Total load command for the entire unit.

[0046] The correction factor is constructed by obtaining the mean absolute value of the unit contribution vector of the drifting unit within the current time window. And the contribution baseline that has been archived in S1. ;when If the calculated result exceeds 1, it is truncated to 1; if the calculated result does not exceed 1, the original calculated result is retained; the drift significance is multiplied by the truncated ratio to obtain the correction factor used to constrain the drawdown depth: ; in, For a moment The correction factor To identify the significance of the drift, the value range is: ; The average absolute value of each component of the contributing vector within the current time window of this unit; The contribution baseline for this unit during the commissioning period is objectively obtained from the historical database of the distributed control system. This is a saturation truncation operation. Incorporating drift amplitude information into the correction factor: when a unit's current contribution is higher than the baseline, the correction factor increases, and the unit's allocation is correspondingly reduced; when a unit's contribution is lower than the baseline, i.e. As the correction factor decreases, the unit's allocation share is replenished accordingly. Saturation cutoff ensures that when a unit's contribution is already significantly higher than the baseline, further increasing the contribution ratio will not increase the retracement depth, preventing over-correction. Its own The interval constitutes an upper limit constraint on the retracement depth, which, together with the saturation cutoff, ensures the boundedness of the correction factor, thereby guaranteeing... Thus ensuring This means that no matter how deep the drift is, the load command sent to the unit will always be positive, there will be no physical illegal state of command reversal, and no manually set weight constants need to be introduced throughout the process.

[0047] Next, the fixed gain column elements corresponding to the drifting units in the static allocation matrix are proportionally reduced using a correction factor. The reduced load share is then normalized and redistributed according to the fixed gain column elements of the remaining healthy units, ensuring that the gain column elements of each healthy unit are proportionally amplified from their original values. This guarantees that the relative gain ratio between healthy units remains unchanged before and after the rewrite. Based on the multiplication of the rewritten gain column elements with the total load command for the entire device, an adaptive allocation command is finally generated and issued to each unit. This rewrite action is only triggered by the identified drifting units, and the relative gain ratio between healthy units remains unchanged after the rewrite, thus pinpointing the mismatch source and avoiding additional disturbances to the relative allocation relationship of the remaining steady-state loops of the entire device.

[0048] S4. Calculate the feedforward compensation signal and inject it into the downstream distillation unit.

[0049] In an optional embodiment, the material flow rate change caused by the upstream unit allocation ratio correction needs to reach the downstream distillation column due to material transport lag. During this transport lag period, if no feedforward notification is given to the downstream control loop, the liquid level in the top reflux tank will continue to drift along the original trend during the gap period when the correction has been initiated but the effect has not yet arrived. The operator will still see the liquid level continuously and monotonously rising from the human-machine interface, which may easily trigger large-scale manual intervention again, causing the upstream correction action that has already been initiated to be superimposed and interfered with by manual operation.

[0050] To eliminate misjudgments during this gap period, the coordination layer, while issuing adaptive allocation instructions, injects feedforward compensation signals into the feedforward channels of the downstream distillation unit's reflux loop and reboiler steam loop. This invention deeply correlates dynamic feedforward calculation with the distillation column material transport lag scenario, dynamically calculating the feedforward compensation signal based on real-time drift significance and historical contribution baseline. The specific calculation process is as follows: The material balance conversion factor from the drifting unit to the downstream distillation unit is read; the absolute mean of the unit contribution vector of the drifting unit within the current time window is extracted; the material balance conversion factor, correction factor, and absolute mean are multiplied together to finally obtain the feedforward compensation signal. ; in, For a moment The feedforward compensation signal, upstream unit The allocation correction amount is converted to the downstream circuit. Material balance conversion factor, For a moment The correction factor This is the average absolute value of each component of the contribution vector within the current time window of the unit. The material balance conversion factor is directly read from the input-output ratio of the existing device material balance calculation block in the distributed control system configuration under steady-state conditions. This is inherent engineering data of the device and requires no additional detection instruments, nor is it assigned manually based on experience. The dimensions of the feedforward signal are consistent with the operating variables of the downstream loop, and can be directly superimposed on the feedforward channel without causing dimensional misalignment.

[0051] Furthermore, a feedforward compensation signal is injected into the feedforward channel of the downstream distillation unit using the material transport lag as the waiting period. Specifically: a timer is started with the material transport lag time as the waiting period; at the beginning of the waiting period, the feedforward compensation signal is superimposed on the feedforward channels of the reflux loop and reboiler steam loop of the downstream distillation unit; after the waiting period ends, the feedforward compensation signal is synchronously reset to zero, allowing each loop of the downstream distillation unit to return to normal automatic adjustment. The feedforward compensation signal does not change the final set value of the column top liquid level, but rather fine-tunes the reflux ratio and reboiler load in advance within the lag delay window, so that the column top liquid level maintains its original steady-state trend before the upstream corrective material arrives. After the upstream corrective material arrives, the feedforward signal is synchronously reset to zero, and the column top loop seamlessly returns to normal automatic adjustment. The entire transition process appears to the operator as a stable liquid level rather than a jump, eliminating the triggering conditions for accidental manual intervention.

[0052] Next, the coordination layer will number the drift units identified in this round, value, The values, along with the triggered allocation correction and feedforward compensation values, are written into the distributed control system event history database and presented as information bars in a non-alarm format on the operation screen for process personnel to trace and manually verify. However, manual adjustment by operators is no longer required, thus realizing a shift from a passive remedial action after an alarm to an active correction during drift.

[0053] It should be noted that the injection channel of the feedforward compensation signal is not limited to the reflux flow loop and the reboiler steam flow loop. Those skilled in the art can extend the feedforward channel to other operating variables directly related to material and energy balance, such as the tower pressure regulation loop or the side-stream sampling loop, according to the specific distillation column structure and control scheme of the device. This invention does not impose any restrictions on this.

[0054] Thus, by simultaneously issuing the upstream correction command and injecting a compensation signal based on material balance calculations into the downstream distillation column feedforward channel, and with material transport being purely delayed... As the effective timing reference for the feedforward signal, it eliminates the risk of continuous level drift and human error caused by the upstream correction effect not being able to reach the downstream during the transmission gap period. This makes the entire set of identification, correction, and feedforward processing links form a complete coordinated processing loop with tight timing. The coordinated control of the whole unit has achieved autonomous and smooth transition without human intervention under typical operating conditions such as raw material batch switching.

[0055] like Figure 2 As shown, minute same-direction deviations of units that cannot be detected under traditional instantaneous snapshot perspectives exhibit significant geometric differentiation after oblique alignment rearrangement and row-by-row accumulation according to this invention: the cumulative contribution curve of healthy units maintains zero-mean random fluctuations without showing any monotonic trend; while the cumulative contribution curve of drifting units with latent mismatches reveals a monotonic, stable, and continuously amplified climbing pattern, highly consistent with the geometric direction of the reference template with linearly increasing step size. This morphological difference allows this invention to identify latent mismatch sources from the background of normal fluctuations early on, through correlation coefficient quantitative calculation, when the deviation just begins to show a climbing trend. Compared to the traditional method of passively entering the alarm range due to integral mismatch, this method successfully breaks through the blind zone of conventional threshold alarms. This advanced and highly sensitive positioning effect provides a reliable and conclusive basis for subsequent control systems to implement directional adaptive correction only for the gain column of the drifting unit, avoiding the cascading disturbance of correction actions to the remaining healthy units of the entire device.

[0056] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for coordinated control of a chemical production process based on DCS, characterized in that, include: S1. Collect the control loop operation data involved in coordination, calculate the basic descriptive quantities of the control loop operation data based on the sliding time window and perform standardization processing, and splice them in the process flow order to form the unit contribution vector. S2. Align and rearrange the unit contribution vectors diagonally along the time axis according to the material transport lag to obtain a two-dimensional matrix. Accumulate the two-dimensional matrix along the column direction to obtain the cumulative contribution curve. Calculate the correlation value between the cumulative contribution curve and the reference template as the drift significance. S3. For drift units whose drift significance exceeds the trigger threshold, calculate the correction factor based on the drift significance, and use the correction factor to rewrite the corresponding gain column in the static allocation matrix to generate an adaptive allocation instruction. S4. While issuing the adaptive allocation instruction, calculate the feedforward compensation signal based on the correction factor, and inject the feedforward compensation signal into the feedforward channel of the downstream distillation unit with the material transmission lag as the waiting period.

2. The DCS-based coordinated control method for chemical production processes according to claim 1, characterized in that, The process of collecting control loop operation data involved in coordination, calculating basic descriptive quantities of control loop operation data based on a sliding time window and performing standardization processing includes: obtaining setpoints, process values, controller outputs and valve position feedback of the control loop involved in coordination; The offset of the controller output, the steady-state residual of the process value, and the hysteresis deviation of the valve position feedback are calculated within the sliding time window. The offset, steady-state residual, and hysteresis deviation are standardized by performing division operations on the historical standard deviation to obtain a dimensionless operating state triplet as the basic descriptor.

3. The DCS-based coordinated control method for chemical production processes according to claim 1, characterized in that, The process of assembling unit contribution vectors according to process flow order includes: extracting basic descriptive quantities of all control loops belonging to the same process unit; The basic descriptive quantities of all control loops belonging to the same process unit are vertically stacked and spliced ​​in the order of process flow to form a unit contribution vector that reflects the net additional output or intake of materials and energy of the process unit.

4. The DCS-based coordinated control method for chemical production processes according to claim 1, characterized in that, The step of obliquely aligning and rearranging the unit contribution vector along the time axis according to the material transport lag to obtain a two-dimensional matrix includes: obtaining the pipeline volume and current flow rate from the upstream unit to the downstream unit, and dividing the pipeline volume by the current flow rate to calculate the material transport lag. Align the unit contribution vector of the upstream unit at a time that is one material transport lag time ahead of the current time of the downstream unit with the unit contribution vector of the downstream unit at the current time in the same column to obtain a two-dimensional matrix in which the row direction represents the spatial position and the column direction represents the time trajectory.

5. The DCS-based coordinated control method for chemical production processes according to claim 1, characterized in that, The correlation value between the calculated cumulative contribution curve and the reference template is used as the drift significance, including: constructing a linearly increasing sequence with equal step size as a reference template for pure monotonic climbing; The correlation coefficient algorithm of Pearson is used to calculate the correlation value between the cumulative contribution curve and the reference template, and the correlation value is used as the drift significance to characterize the degree of latent mismatch.

6. The DCS-based coordinated control method for chemical production processes according to claim 1, characterized in that, After calculating the correlation value between the cumulative contribution curve and the reference template as the drift significance, the method further includes: collecting the historical sequence of drift significance of each process unit under all healthy operating conditions during the unit commissioning and adjustment period. The trigger boundary is obtained by calculating the historical series mean and standard deviation, and then adding the products of the historical series mean and standard deviation to a preset multiple. The trigger boundary is used as the trigger threshold to determine whether the drift significance exceeds the limit.

7. The DCS-based coordinated control method for chemical production processes according to claim 1, characterized in that, The calculation of the correction factor based on drift significance includes: obtaining the mean absolute value of each component of the unit contribution vector of the drifting unit within the current sliding window, and obtaining the contribution baseline statistically archived during the data acquisition phase; Calculate the ratio of the absolute mean to the contribution baseline. If the ratio is greater than 1, the ratio is truncated to 1. If the ratio is less than or equal to 1, the original ratio calculation result is retained. Multiplying the drift significance by the ratio after truncation or retention yields a correction factor used to constrain the depth of drawdown.

8. The DCS-based coordinated control method for chemical production processes according to claim 1, characterized in that, The step of rewriting the corresponding gain column in the static allocation matrix using a correction factor to generate an adaptive allocation instruction includes: scaling down the elements of the solidified gain column corresponding to the drift unit in the static allocation matrix using a correction factor. The reduced load share is normalized and redistributed according to the fixed gain column elements of the remaining healthy units; Based on the multiplication of the rewritten gain column elements with the total load command of the entire unit, an adaptive allocation command is generated and issued to each process unit.

9. The DCS-based coordinated control method for chemical production processes according to claim 1, characterized in that, The calculation of the feedforward compensation signal based on the correction factor includes: reading the material balance conversion factor from the drift unit to the downstream distillation unit; Extract the mean absolute value of the contribution vector of the drifting unit within the current sliding window; The feedforward compensation signal is obtained by multiplying the material balance conversion factor, the correction factor, and the mean absolute value.

10. The DCS-based coordinated control method for chemical production processes according to claim 1, characterized in that, The step of injecting a feedforward compensation signal into the feedforward channel of the downstream distillation unit with the material transport lag time as the waiting period includes: starting a timer with the material transport lag time as the waiting period. At the start of the waiting period, the feedforward compensation signal is superimposed onto the feedforward channels of the reflux loop of the downstream distillation unit and the reboiler steam flow loop. After the waiting period ends, the feedforward compensation signal will be synchronously reset to zero, so that each control loop of the downstream distillation unit returns to the normal automatic adjustment state.

Images