Cooperative operation control method and system for multi-effect evaporators

Abstract

The invention relates to the technical field of industrial equipment automatic control, in particular to a cooperative operation control method and system for a multi-effect evaporator. The method comprises the following steps: training a GRU-based reference feedforward model by using multi-source time sequence data in a clean state of equipment; secondly, an observation total heat exchange coefficient and a theoretical total heat exchange coefficient are obtained in real time, and a cleanliness evaluation factor is calculated through a filter to represent the scaling degree; secondly, performing nonlinear gain correction on the reference feed-forward signal based on the factor to obtain a self-adaptive feed-forward signal for compensating the scaling influence; and finally, generating a dynamic cooperative weight according to the volatility of the current working condition, and carrying out weighted fusion on the self-adaptive feedforward signal and the feedback control signal. According to the method, the problem that the static model is mismatched along with equipment scaling is solved, and stable and efficient control of the whole period is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic control of industrial equipment, and in particular to a collaborative operation control method and system of a multi-effect evaporator. BACKGROUND

[0002] A multi-effect evaporator (MEE) is applied to high-salinity wastewater treatment in the chemical industry, pesticide industry, and new energy industry. The core principle of the MEE is to use the secondary steam of a previous effect as a heat source of a subsequent effect, so as to realize step-by-step utilization and high-efficiency energy saving of heat energy. In an actual industrial site, the feed flow rate and concentration of the MEE often experience dramatic mutations due to intermittent discharge of upstream processes. Since a traditional PID (proportional-integral-derivative controller) feedback control has a large hysteresis, it is difficult to cope with such rapid disturbances. In related technologies, a feedforward control strategy based on a neural network such as a gated recurrent unit (GRU) is usually introduced. The GRU is used to capture the nonlinear characteristics of time series data, and a mapping between process disturbances and control variables is established, so as to realize early compensation for mutated working conditions.

[0003] However, the above-mentioned control method based on the GRU has significant defects in actual application. The physical characteristics of the MEE system have strong time-varying properties. The heat exchange tube wall of the MEE system will produce fouling as the running time elapses, resulting in continuous attenuation of the overall heat transfer coefficient. The existing GRU feedforward model is usually trained offline based on clean data after cleaning of the equipment. Once the model parameters are determined, they remain unchanged. When the equipment enters a fouling state, the static GRU model cannot perceive the decline in heat transfer efficiency. The control amount (such as the steam compensation amount) output by the model is still based on the clean standard and is less than the actual demand under the current working condition. This static mismatch between the static parameters of the model and the dynamic fouling of the object will result in insufficient feedforward compensation, causing the control strategy to fail in the later running period. SUMMARY

[0004] To solve the problem of static mismatch during equipment operation in related technologies, resulting in failure of the control strategy, the present application provides a collaborative operation control method and system of a multi-effect evaporator.

[0005] In a first aspect, the present application provides a collaborative operation control method of a multi-effect evaporator, which adopts the following technical solution: A collaborative operation control method of a multi-effect evaporator, comprising: collecting multi-source time series data when the multi-effect evaporator is in a clean state, and training a baseline feedforward model for outputting a baseline feedforward control signal according to real-time working condition data; Real-time acquisition of an observed overall heat transfer coefficient and a theoretical overall heat transfer coefficient of the multi-effect evaporator, calculation of a clean evaluation factor by a filter, and use of the clean evaluation factor to represent the degree of fouling of the multi-effect evaporator; The benchmark feedforward control signal of the benchmark feedforward model is nonlinearly gain corrected based on a clean evaluation factor to obtain a corrected adaptive feedforward control signal, and the clean evaluation factor is used to compensate for the decrease in heat exchange efficiency caused by fouling. The fluctuation of the current working condition is evaluated in real time, and a dynamic coordination weight is generated; the feedback control signal of the multi-effect evaporator is obtained, and the corrected adaptive feedforward control signal and the feedback control signal are weighted and fused according to the dynamic coordination weight to obtain a final coordinated control output signal.

[0006] The fouling degree is represented by real-time calculation of the clean evaluation factor, and the benchmark feedforward model trained in the clean state is nonlinearly gain corrected based on the clean evaluation factor, effectively solving the static mismatch problem that the static model cannot adapt to the decrease in heat exchange efficiency caused by fouling during equipment operation. At the same time, a dynamic coordination weight is constructed based on the working condition fluctuation, and the adaptive feedforward is used to quickly respond in the mutation working condition, and the feedback control signal is used to eliminate the error in the steady state working condition, so that stable and efficient coordinated control of the multi-effect evaporator in the whole operation cycle is realized.

[0007] Optionally, under the clean state of the multi-effect evaporator, a mapping model of working condition data to total heat exchange coefficient is established; According to the real-time working condition data, the theoretical total heat exchange coefficient of the multi-effect evaporator in the clean state under the current working condition is predicted through the mapping model.

[0008] The mapping model is established to predict the theoretical total heat exchange coefficient under the current working condition in real time, which can eliminate the interference of changes in feed flow, temperature and other working conditions on the heat exchange coefficient, and ensure the reliability and accuracy of the calculated clean evaluation factor.

[0009] Optionally, in the step of calculating the clean evaluation factor through the filter, the ratio of the observed total heat exchange coefficient to the theoretical total heat exchange coefficient is first-order exponential smoothing filtered to obtain the clean evaluation factor.

[0010] Introducing first-order exponential smoothing filtering of the ratio of the observed value to the theoretical value can effectively filter out high-frequency noise and random interference generated in the process of industrial field sensor collection, prevent the clean evaluation factor from jumping due to measurement fluctuation, and ensure the smoothness of the clean evaluation factor and the operation stability of the control system.

[0011] Optionally, the filtering coefficient of the first-order exponential smoothing filter is greater than or equal to 0.99 and less than 1.

[0012] The filter is given a low-pass characteristic, so that it can accurately extract the extremely slow fouling trend signal from real-time data containing a large amount of noise.

[0013] Optionally, the step of modifying the benchmark feedforward control signal of the benchmark feedforward model based on the cleanliness evaluation factor to obtain the modified adaptive feedforward control signal comprises: subtracting the difference between 1 and the cleanliness evaluation factor to obtain a fouling degree, exponentially amplifying the fouling degree, then adding 1 to obtain a correction coefficient, and multiplying the correction coefficient by the benchmark feedforward control signal to obtain the adaptive feedforward signal.

[0014] The nonlinear exponential amplification correction strategy based on the fouling degree is more in line with the physical law of the influence of heat exchanger fouling on heat transfer efficiency, and ensures that the control signal has sufficient gain strength in the later stage of fouling Optionally, the dynamic coordination weight includes a feedforward coordination weight and a feedback coordination weight, the feedforward coordination weight is used as the weight of the adaptive feedforward signal, and the feedback coordination weight is used as the weight of the feedback control signal, and the sum of the feedforward coordination weight and the feedback coordination weight is 1.

[0015] Optionally, the step of obtaining the feedforward coordination weight comprises: obtaining historical data of a key working condition parameter; performing fast exponential moving average and slow exponential smoothing filtering on the historical data, respectively; taking the absolute difference between the two filtering results as the fluctuation degree of the working condition parameter; and inputting the fluctuation degree into a hyperbolic tangent function to obtain the feedforward coordination weight.

[0016] When the working condition is stable, the result obtained by the slow exponential moving average is close to the result obtained by the fast exponential moving average; when the working condition is suddenly changed, the result obtained by the fast exponential moving average is greater than the result obtained by the slow exponential moving average, so that the feedforward weight is automatically increased to quickly suppress the disturbance when the working condition fluctuates sharply, and the feedback is automatically switched to the dominant state to accurately maintain the set value when the working condition is stable, thereby significantly improving the anti-interference ability and dynamic response speed of the system Optionally, the feedback control signal adopts a PID controller and is calculated according to the deviation between the output concentration of the last effect of the multi-effect evaporator and the set value.

[0017] Optionally, the benchmark feedforward model is a gated recurrent unit (GRU) network; the input sequence of the benchmark feedforward model includes the feed flow, the feed concentration, the liquid level of each effect, and the pressure of each effect of a plurality of historical time steps; and the output sequence of the benchmark feedforward model includes the control amount of the first effect steam and the control amount of the last effect discharge pump at the target time.

[0018] In a second aspect, the application provides a coordinated operation control system of a multi-effect evaporator, which adopts the following technical scheme: A coordinated operation control system of a multi-effect evaporator, comprising: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a coordinated operation control method of a multi-effect evaporator according to the above is implemented.

[0019] The computer program generated by the multi-effect evaporator cooperative operation control method is stored in the memory and loaded and executed by the processor, so that the system is made according to the memory and the processor, and the use is facilitated.

[0020] The application has the following technical effects: The clean evaluation factor is introduced to perform nonlinear gain correction on the output of the benchmark feedforward model, and accurately compensate for the efficiency loss caused by internal fouling of the equipment. At the same time, combined with the dynamic cooperative weight based on the signal fluctuation rate, the fusion ratio of the adaptive feedforward control signal and the feedback control signal is adjusted in real time. Effectively solve the pain points of traditional control in equipment aging and sudden change of working condition. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 is a method flowchart of a multi-effect evaporator cooperative operation control method according to an embodiment of the application.

[0022] Figure 2 is a clean evaluation factor change graph output in real time in the method of a multi-effect evaporator cooperative operation control method according to an embodiment of the application.

[0023] Figure 3 is a change graph of the correction coefficient in the method of a multi-effect evaporator cooperative operation control method according to an embodiment of the application.

[0024] Figure 4 is a change graph of the feedforward cooperative weight and the feedback cooperative weight in the method of a multi-effect evaporator cooperative operation control method according to an embodiment of the application. DETAILED DESCRIPTION

[0025] The embodiment of the application discloses a multi-effect evaporator cooperative operation control method, which dynamically corrects the output of the static GRU feedforward model by real-time evaluation of heat exchange attenuation caused by fouling, and adaptively adjusts the cooperative weight of feedforward and feedback based on working condition fluctuation, thereby solving the technical problems that the traditional feedforward model is mismatched with the equipment aging and the control system is easily oscillated under sudden working condition, and making the multi-effect evaporator maintain stable and efficient cooperative control in the whole operation cycle.

[0026] REFERENCE Figure 1 A multi-effect evaporator cooperative operation control method includes steps S1-S4.

[0027] S1: Collecting multi-source time series data when the multi-effect evaporator is in a clean state, training a benchmark feedforward model for outputting a benchmark feedforward control signal according to real-time working condition data.

[0028] First, data collection is carried out in the state of just completing chemical cleaning of the multi-effect evaporator (MEE) system, at this time it is determined that there is no fouling phenomenon in the multi-effect evaporator. The collected data includes multiple sets of multi-source time series data at the time of working condition mutation. The input sequence is a multi-dimensional vector, which includes N working condition data of a number of historical time steps (for example, N=10), specifically including feed flow, feed concentration, liquid level of each effect, and pressure of each effect. The output sequence is the control target value at the future time, for example, the control amount of the first effect steam and the control amount of the last effect discharge pump.

[0029] A multi-input multi-output (MIMO) gated recurrent unit (GRU) neural network is trained using the above data set. The update gate and reset gate structure inside the GRU network is suitable for capturing the time lag and non-linear characteristics in the chemical process. After training, a static benchmark feedforward model is obtained. In subsequent real-time operation, the model will output the benchmark feedforward control signal according to the real-time collected working condition data.

[0030] By training the benchmark feedforward model in a clean state, the system can master the optimal control strategy of the equipment in the ideal state to deal with working condition mutations, providing a benchmark reference for subsequent adaptive correction.

[0031] S2: Real-time acquisition of the observed total heat transfer coefficient and the theoretical total heat transfer coefficient of the multi-effect evaporator, and calculation of the clean evaluation factor by a filter, the clean evaluation factor being used to represent the fouling degree of the multi-effect evaporator.

[0032] The ratio of the observed total heat transfer coefficient to the theoretical total heat transfer coefficient is subjected to first-order exponential smoothing filtering to obtain the clean evaluation factor.

[0033] Specifically, the calculation formula of the clean evaluation factor can be represented as: ; in the formula, represents the clean evaluation factor at time ; represents the clean evaluation factor at time ; represents the observed total heat transfer coefficient at time which is acquired in real time, and is calculated in real time according to the standard thermodynamic formula through the temperature, pressure, flow and other sensor data of each effect in the multi-effect evaporator, that is, the ratio between the current heat transfer amount and the product of the heat transfer area and the logarithmic mean temperature difference, wherein the current heat transfer amount can be obtained through the steam flow, the heat transfer area is provided by the equipment developer, and the logarithmic mean temperature difference is obtained from the difference between the steam side temperature and the liquid side temperature, the acquisition of the logarithmic mean temperature difference being a conventional technical means in the art, which will not be described here; represents the theoretical total heat transfer coefficient at time The theoretical total heat exchange coefficient of the device under the working condition state is obtained by the following steps: establishing a mapping model of working condition data to the total heat exchange coefficient under the clean state of the multi-effect evaporator; and predicting the theoretical total heat exchange coefficient of the multi-effect evaporator under the current working condition and the clean state through the mapping model according to real-time working condition data. The filter coefficient of the first-order exponential smoothing filter.

[0034] Because Easily disturbed by high-frequency noise, if the instantaneous ratio is directly used, the curve will fluctuate sharply. When is set to a constant close to , the formula shows filtering characteristics. Combined with Figure 2 , it is assumed that due to measurement noise, it will jump sharply between and , and after iterative calculation by the above formula, will present a smooth downward curve starting from . For example, when the system runs for a period of time, if drops to , it clearly indicates that the physical fouling has caused the heat exchange performance to irreversibly decay , thereby providing a stable and reliable basis for subsequent compensation.

[0035] S3: Nonlinear gain correction is performed on the baseline feedforward control signal of the baseline feedforward model based on the clean evaluation factor to obtain a corrected adaptive feedforward control signal. The clean evaluation factor is used to compensate for the decrease in heat exchange efficiency caused by fouling.

[0036] The difference between 1 and the clean evaluation factor is taken as the fouling degree, the fouling degree is exponentially amplified, then added to 1 to obtain a correction coefficient, and the product of the correction coefficient and the baseline feedforward control signal is taken as the adaptive feedforward signal Specifically, the calculation formula of the adaptive feedforward control signal can be expressed as: ; Wherein, represents the adaptive feedforward control signal at time ; represents the baseline feedforward control signal output by the baseline feedforward model; is a correction gain coefficient, which is set by a person skilled in the art based on experience, and is mainly used to adjust the adjustment strength of the clean evaluation factor on the baseline feedforward control signal, and improve the flexibility of the system; represents the fouling degree at time ; is a nonlinear factor, which is mainly used for nonlinear amplification of the fouling degree, and is set by a person skilled in the art based on experience.

[0037] represents the correction coefficient. It is assumed that the current system runs for a period of time, decreases to 0.7. Set . The correction coefficient is: .

[0038] If the reference model output (unit steam amount), the corrected signal . This means that the system automatically increases the control amount by about 35.4% to compensate for the thermal resistance caused by fouling. As shown in Figure 3 , over time, the correction coefficient shows a nonlinear upward trend.

[0039] In this way, by means of nonlinear correction based on the degree of fouling, it is possible to ensure that the feedforward control signal remains sufficient and accurate throughout the entire cycle of the device from clean to dirty, reducing the problem of model static mismatch.

[0040] S4: Real-time evaluation of the volatility of the current working condition, and generation of a dynamic coordination weight; obtaining a feedback control signal of the multi-effect evaporator, and weighting and fusing the corrected adaptive feedforward control signal and the feedback control signal according to the dynamic coordination weight to obtain a final coordinated control output signal.

[0041] The dynamic coordination weight includes a feedforward coordination weight and a feedback coordination weight, the feedforward coordination weight is taken as the weight of the adaptive feedforward signal, and the feedback coordination weight is taken as the weight of the feedback control signal, and the sum of the feedforward coordination weight and the feedback coordination weight is 1.

[0042] For the feedforward coordination weight, the obtaining step includes: obtaining historical data of a key working condition parameter; performing fast exponential moving average and slow exponential smoothing filtering on the historical data respectively; taking the absolute difference value of the two filtering results as the fluctuation degree of the working condition parameter; and inputting the fluctuation degree into a hyperbolic tangent function to obtain the feedforward coordination weight.

[0043] The calculation formula of the feedforward coordination weight can be expressed as: ; in the formula, is the feedforward coordination weight at time ; represents the result obtained by performing fast exponential moving average on the feed concentration , and in the process, the smoothing coefficient in the exponential moving average tends to 1, which can be set to 0.9; represents the feed concentration of the feed liquid entering the multi-effect evaporator from the feed end at time ; represents the feed concentration of the feed liquid entering the multi-effect evaporator from the feed end at time The result obtained by performing a slow exponential moving average is that the smoothing coefficient in the exponential moving average approaches 0 during the process, and can be set to 0.1; Represents the hyperbolic tangent function; This represents the gain coefficient, which is set based on the experience of those skilled in the art and is mainly used to improve the flexibility of the system.

[0044] The difference between 1 and the feedforward collaborative weight is used as the feedback collaborative weight.

[0045] In the step of weighting and fusing the corrected adaptive feedforward control signal and feedback control signal according to the dynamic collaborative weights to obtain the final collaborative control output signal, the calculation formula for the collaborative control output signal can be expressed as: In the formula, Indicates time The coordinated control output signal; Indicates time Feedforward collaborative weights; Indicates the feedback collaboration weight; Indicates time The adaptive feedforward control signal; Indicates time The feedback control signal, i.e. the output of the parallel PID controller, is calculated based on the deviation between the final output concentration of the multi-effect evaporator and the set value. This part is a conventional technique in this field, so the specific output process will not be described in detail.

[0046] For example, combining Figure 4 When operating conditions are stable, the results obtained from the slow exponential moving average are nearly equal to those obtained from the fast exponential moving average. When the value approaches 0, the output signal of the coordinated control approaches the feedback control signal, and the system mainly eliminates steady-state error through PID regulation.

[0047] When the operating conditions change abruptly As the weight approaches 1, the feedforward collaborative weight approaches 1, and the collaborative control output signal approaches the adaptive feedforward control signal. The system then utilizes the adaptive feedforward signal to respond quickly.

[0048] This application also discloses a collaborative operation control system for a multi-effect evaporator, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a collaborative operation control method for a multi-effect evaporator according to this application is implemented.

[0049] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0050] The above are all preferred embodiments of the present application, and do not limit the protection scope of the present application, so that: all equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method for coordinated operation control of a multi-effect evaporator, characterized in that, Multi-source time-series data were collected when the multi-effect evaporator was in a clean state, and a reference feedforward model was trained to output a reference feedforward control signal based on real-time operating data. The observed total heat transfer coefficient and theoretical total heat transfer coefficient of the multi-effect evaporator are obtained in real time. The cleanliness assessment factor is calculated through a filter and is used to characterize the degree of scaling of the multi-effect evaporator. The baseline feedforward control signal of the baseline feedforward model is nonlinearly modified based on the cleanliness assessment factor to obtain the modified adaptive feedforward control signal. The cleanliness assessment factor is used to compensate for the decrease in heat exchange efficiency caused by scaling. The system continuously assesses the volatility of the current operating conditions and generates dynamic collaborative weights; it acquires the feedback control signal of the multi-effect evaporator, and then weights and fuses the corrected adaptive feedforward control signal and the feedback control signal according to the dynamic collaborative weights to obtain the final collaborative control output signal.

2. The method for coordinated operation control of a multi-effect evaporator according to claim 1, characterized in that, Under clean conditions, a mapping model from operating data to the overall heat transfer coefficient is established for the multi-effect evaporator. Based on real-time operating data, the theoretical total heat transfer coefficient of the multi-effect evaporator in a clean state under the current operating conditions is predicted using the mapping model.

3. The method for coordinated operation control of a multi-effect evaporator according to claim 1, characterized in that, In the step of calculating the cleanliness assessment factor through a filter, the ratio of the observed total heat transfer coefficient to the theoretical total heat transfer coefficient is subjected to a first-order exponential smoothing filter to obtain the cleanliness assessment factor.

4. The method for coordinated operation control of a multi-effect evaporator according to claim 3, characterized in that, The filter coefficients of a first-order exponential smoothing filter are greater than or equal to 0.99 and less than 1.

5. The method for coordinated operation control of a multi-effect evaporator according to claim 1, characterized in that, The steps for nonlinearly correcting the baseline feedforward control signal of the baseline feedforward model based on the cleanliness assessment factor to obtain the corrected adaptive feedforward control signal include: subtracting the cleanliness assessment factor from 1 as the fouling degree, exponentially amplifying the fouling degree, adding it to 1 to obtain the correction coefficient, and multiplying the correction coefficient with the baseline feedforward control signal as the adaptive feedforward signal.

6. The method for coordinated operation control of a multi-effect evaporator according to claim 1, characterized in that, The dynamic collaborative weights include feedforward collaborative weights and feedback collaborative weights. The feedforward collaborative weights are used as the weights of the adaptive feedforward signal, and the feedback collaborative weights are used as the weights of the feedback control signal. The sum of the feedforward collaborative weights and the feedback collaborative weights is 1.

7. The method for coordinated operation control of a multi-effect evaporator according to claim 6, characterized in that, The steps for obtaining the feedforward collaborative weights include: acquiring historical data of key operating parameters; performing fast exponential moving average and slow exponential smoothing filters on the historical data respectively; using the absolute difference between the two filtering results as the fluctuation level of the operating parameters; and inputting the fluctuation level into the hyperbolic tangent function to obtain the feedforward collaborative weights.

8. The method for coordinated operation control of a multi-effect evaporator according to claim 1, characterized in that, The feedback control signal uses a PID controller, which calculates the value based on the deviation between the final output concentration of the multi-effect evaporator and the set value.

9. The method for coordinated operation control of a multi-effect evaporator according to claim 1, characterized in that, The baseline feedforward model is a gated cyclic unit (GRU) network. The input sequence of the baseline feedforward model includes feed flow rate, feed concentration, liquid level and pressure of each effect at several historical time steps. The output sequence of the baseline feedforward model includes the steam control quantity of the first effect and the discharge pump control quantity of the last effect at the target time.

10. A collaborative operation control system for a multi-effect evaporator, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a collaborative operation control method for a multi-effect evaporator according to any one of claims 1-9.

Images