A falling film evaporator liquid level setting optimization method and system

Abstract

This invention relates to the field of falling film evaporator technology, specifically to a method and system for optimizing liquid level setting in falling film evaporators. The invention acquires process data and actual liquid level at each moment, uses any process index as the target index, and obtains target parameters and reference data. It further classifies and obtains all reference groups and the process reference degree for each reference group; it constructs time-series sequences of parameters and liquid level for each reference group, analyzes the fluctuation correlation between the two sequences, and, combined with the process reference degree, obtains the index sensitivity of the target index; it further obtains the optimal recommended process data for the current process data to adjust the liquid level. By analyzing the influence of different process indices on liquid level changes and obtaining the corresponding index sensitivity, this invention further combines simulation models and recommendation models to obtain the optimal recommended process data. Based on the comprehensive adjustment of multiple process parameters, it improves the stability of the liquid level optimization setting process in falling film evaporators.

Description

Technical Field

[0001] The invention relates to the technical field of falling film evaporators, and in particular to a falling film evaporator liquid level setting optimization method and system. Background Art

[0002] Falling film evaporator is a heat transfer device used for evaporation, concentration or separation and purification of liquids. It is widely used in chemical, pharmaceutical and food industries. Its working principle is that the liquid forms a thin film on the surface of the heat exchange tube, which improves the heat exchange efficiency and accelerates evaporation. The liquid level of the falling film evaporator refers to the height of the liquid in the evaporator on the surface of the heat exchange tube. A reasonable liquid level can ensure the continuity of liquid evaporation and improve efficiency to reduce energy consumption. At the same time, it can also avoid safety accidents caused by overheating of the equipment or excessive liquid; therefore, the setting of the liquid level is very important.

[0003] At present, when setting the liquid level of a falling film evaporator, the optimal liquid level is usually determined based on the evaporation process parameters such as the liquid properties, evaporator design parameters, and evaporation experimental conditions, and based on experience, experiments, or related mathematical models. However, after determining the optimal liquid level, the liquid level is usually controlled by adjusting the flow rate of the feed and discharge; however, this adjustment method has a certain time lag and is prone to cause liquid level fluctuations, which may affect the stability and efficiency of the falling film evaporator. Summary of the invention

[0004] In order to solve the technical problem that the existing liquid level adjustment leads to poor stability of the falling film evaporator, the object of the present invention is to provide a falling film evaporator liquid level setting optimization method and system, and the technical scheme adopted is as follows:

[0005] The present invention provides a method for optimizing liquid level setting of a falling film evaporator, the method comprising:

[0006] Obtaining process data and actual liquid level of the falling film evaporator at each moment, wherein the process data includes process parameters of each process index; taking any process index as a target index, taking the process parameter of the target index as a target parameter, and taking process data without the target parameter as reference data;

[0007] Classify all reference data according to the similarity between the reference data, and obtain all reference groups and the process reference degree of each reference group; in each reference group, construct a parameter time series sequence and a liquid level time series sequence according to the target parameters and the actual liquid level at the corresponding time of all the reference data;

[0008] According to the fluctuation correlation between the parameter time series and the liquid level time series of each reference group, the time-delay liquid level of each target parameter in the parameter time series is obtained; according to the change difference and corresponding time difference between each target parameter and the corresponding time-delay liquid level, combined with the process reference degree of the reference group to which the reference data at the corresponding time of the target parameter belongs, the indicator sensitivity of the target indicator is obtained;

[0009] The ideal liquid level of the current process data is obtained based on a preset simulation model; all recommended process data of the ideal liquid level are obtained based on a preset process recommendation model; the best recommended process data is obtained based on the current process data at the current moment and the difference between each of the recommended process data, combined with the indicator sensitivity of each process indicator; the liquid level of the current falling film evaporator is adjusted according to the best recommended process data.

[0010] Furthermore, the method for obtaining the process reference degree includes:

[0011] Clustering all the reference data based on a clustering algorithm, and using each obtained cluster as a reference group;

[0012] In each of the reference groups, according to the data difference between each of the reference data and all the other reference data, an in-group outlier parameter of each of the reference data is obtained; the data difference is positively correlated with the in-group outlier parameter;

[0013] The outlier parameters within the group of all the reference data are integrated to obtain the process reference degree of each reference group; the outlier parameters within the group are negatively correlated with the process reference degree.

[0014] Furthermore, the method for acquiring the parameter time series and the liquid level time series includes:

[0015] In each of the reference groups, all the reference data are sorted in chronological order to obtain a reference data sequence; a parameter timing sequence is constructed according to the target parameters at the corresponding moment of each reference data in the reference data sequence; a liquid level timing sequence is constructed according to the actual liquid level at the corresponding moment of each reference data in the reference data sequence; the timing order of the parameter timing sequence and the liquid level timing sequence are the same as the timing order of the reference data sequence.

[0016] Furthermore, the method for obtaining the time-delay liquid level includes:

[0017] Perform DTW matching on the parameter time series sequence of each reference group and the liquid level time series sequence to obtain all matching liquid levels of each target parameter in each parameter time series sequence in the liquid level time series sequence;

[0018] A matching liquid level is selected from all matching liquid levels of each target parameter as a time-delay liquid level, the time corresponding to the time-delay liquid level is greater than the time corresponding to the target parameter, and the difference between the time corresponding to the time-delay liquid level and the time corresponding to the target parameter is the smallest.

[0019] Furthermore, the method for obtaining the index sensitivity includes:

[0020] According to the local time series variation difference between each target parameter and the corresponding time-delay liquid level, combined with the process reference degree of the reference group to which the reference data at the time corresponding to the target parameter belongs, the liquid level adjustment scale of the target indicator is obtained;

[0021] The negative correlation mapping result between each target parameter and the time difference of the corresponding time lag liquid level is used as the time lag parameter of each target parameter; the process reference degree of the reference group to which the reference data at the time corresponding to the target parameter belongs is used as the first reference weight of the time lag parameter of the target parameter; the first reference weight is used to weight the corresponding time lag parameter and the normalized result of the weighted average result is used as the adjustment response indicator of the target indicator;

[0022] The index sensitivity is obtained according to the liquid level adjustment scale and the adjustment response index, and the liquid level adjustment scale and the adjustment response index are both positively correlated with the index sensitivity.

[0023] Furthermore, the method for obtaining the liquid level adjustment scale includes:

[0024] According to the difference between each target parameter in the parameter time series and a preset number of adjacent target parameters, the parameter variation amplitude of each target parameter is obtained; according to the difference between each actual liquid level in the liquid level time series and a preset number of adjacent actual liquid levels, the liquid level variation amplitude of each actual liquid level is obtained;

[0025] The parameter variation of the target parameter plus a preset minimum non-zero positive parameter is used as the denominator, the liquid level variation of the time-delay liquid level corresponding to the target parameter plus a preset minimum non-zero positive parameter is used as the numerator, and the absolute value of the fractional ratio is used as the synchronous variation of the target parameter;

[0026] The process reference degree of the reference group to which the reference data at the moment corresponding to the target parameter belongs is used as the second reference weight of the synchronous amplitude variation of the target parameter; the corresponding synchronous amplitude variation is weighted averaged using the second reference weight, and the normalized result of the weighted average result is used as the liquid level adjustment scale of the target indicator.

[0027] Furthermore, the method for obtaining all recommended process data of the ideal liquid level based on a preset process recommendation model includes:

[0028] A training sample set is constructed according to the process data and the corresponding actual liquid level, and a neural network model is trained based on the training sample set to obtain a process recommendation model; the ideal liquid level is input into the trained process recommendation model, and all recommended process data of the ideal liquid level are output.

[0029] Furthermore, the method for obtaining the best recommended process data includes:

[0030] Under each process indicator, the difference between the current process data and the process parameters corresponding to each of the recommended process data is used as the process deviation; the corresponding process deviation is weighted and averaged after the indicator sensitivity negative correlation mapping of each process indicator, and the weighted average result is used as the adjustment difficulty of the corresponding recommended process data;

[0031] The negative correlation mapping result of the adjustment difficulty is used as the recommendation degree corresponding to the recommended process parameter; and the recommended process parameter corresponding to the maximum recommendation degree is used as the best recommended process data.

[0032] Furthermore, after obtaining the best recommended process data, the step further includes:

[0033] The ideal liquid level of the best recommended process data is obtained, and new best recommended process data is obtained based on the ideal liquid level of the best recommended process data; it is continuously iterated until the preset conditions are met, and the latest obtained best recommended process data is used as the best recommended process data for the ideal liquid level of the current process data at the current moment.

[0034] The present invention also proposes a falling film evaporator liquid level setting optimization system, the system comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the falling film evaporator liquid level setting optimization method when executing the computer program.

[0035] The present invention has the following beneficial effects:

[0036] The present invention first obtains the process data and actual liquid level of the falling film evaporator at each moment, and the process data includes the process parameters of each process indicator; with the idea of ​​controlling variables, any process indicator is taken as the target indicator, and the target parameters and reference data are obtained to prepare for the subsequent analysis of the influence of each process indicator on the liquid level change; then all the reference data are classified to obtain the process reference degree of each reference group, and the process reference degree reflects the analytical reference value of the influence of the target indicator on the liquid level change under the reference group; further constructs the parameter time series sequence and liquid level time series sequence corresponding to each reference group, and then evaluates the time lag of the liquid level change according to the fluctuation correlation of the sequence, and obtains the time-lag liquid level of each target parameter; according to the correlation between each target parameter and the corresponding time-lag liquid The change difference between the two positions and the corresponding time difference are combined with the process reference degree of the reference group to which the reference data at the corresponding time of the target parameter belongs, and the index sensitivity of the target index is obtained. The index sensitivity quantifies the efficiency of the adjustment of the target parameter on the liquid level change; the ideal liquid level of the current process data is obtained based on the preset simulation model; all recommended process data for the ideal liquid level are obtained based on the preset process recommendation model; according to the difference between the current process data and each recommended process data, combined with the index sensitivity of each process index, the best recommended process data is obtained. The best recommended process data is the recommended process data corresponding to the least parameter change and the fastest response during the liquid level adjustment process; finally, the liquid level of the current falling film evaporator is adjusted according to the best recommended process data. The present invention analyzes the influence of different process indicators on the liquid level change, obtains the index sensitivity of each process indicator, further combines the simulation model and the recommendation model, obtains the optimal recommended process data, and improves the stability of the falling film evaporator in the process of optimizing the liquid level setting based on the comprehensive adjustment of multiple process parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 A flow chart of a method for optimizing liquid level setting of a falling film evaporator provided by one embodiment of the present invention;

[0039] Figure 2 A schematic diagram of DTW matching between a parameter time series and a liquid level time series in a reference group provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a falling film evaporator liquid level setting optimization method and system proposed by the present invention, its specific implementation, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0041] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0042] The specific scheme of a falling film evaporator liquid level setting optimization method and system provided by the present invention is described in detail below with reference to the accompanying drawings.

[0043] See also Figure 1 , which shows a flow chart of a falling film evaporator liquid level setting optimization method provided by an embodiment of the present invention, specifically comprising:

[0044] Step S1, obtaining the process data and actual liquid level of the falling film evaporator at each moment, wherein the process data includes process parameters of each process index; taking any process index as a target index, taking the process parameters of the target index as target parameters, and taking the process data without the target parameters as reference data.

[0045] In one embodiment of the present invention, various sensors are installed at the feed port, the discharge port and the steam inlet of the falling film evaporator to collect the process data of the falling film evaporator at each moment during the evaporation of the liquid material, wherein the process data is composed of the process parameters under each process index collected by various sensors, that is, the process data is a multi-dimensional process parameter; the process index at least includes the liquid inlet flow rate , Inlet temperature , Liquid density , Influent concentration , hydraulic pressure Specific heat capacity of liquid , Liquid flow , liquid outlet temperature , liquid density , liquid concentration , hydraulic pressure , specific heat capacity of liquid , Steam flow , Steam temperature ; Then set a capacitive liquid level sensor in the liquid level observation window to collect the actual liquid level at each moment.

[0046] Taking into account that changes in each process parameter in the process data will cause changes in the liquid level of the falling film evaporator, but the changes in process parameters under different process indicators have different effects on the liquid level changes; in order to analyze the impact of each process indicator on the liquid level change, and thus facilitate the subsequent analysis and adjustment of the process parameters, the embodiment of the present invention will use the idea of ​​control variables, take any process indicator as the target indicator, take the process parameters of the target indicator as the target parameters, and take the process data without the target parameters as reference data; then change the target indicator, obtain the impact of each process indicator on the liquid level change, and thus analyze and adjust the process parameters.

[0047] As an example, the process data collected at time i for( , , , , , , , , , , , , , ), corresponding to the actual liquid level ; Among them, the liquid flow rate As the target indicator, is the target parameter at the i-th moment, then the reference data at the i-th moment for( , , , , , , , , , , , , ); that is, assuming that the process data is an n-dimensional data, the reference data is the n-1-dimensional data excluding the target index.

[0048] It should be noted that the collection frequency of all sensors is once per second, which can be set by the implementer, but the collection frequency of all sensors must be consistent. The implementer can also set the form of process data by himself, such as changing the number and type of process indicators, but it must cover feed-related parameters, discharge-related parameters, and steam or heat exchange medium-related parameters, which are not limited or elaborated here.

[0049] It should be noted that the analysis process of each process indicator is the same. Here, only one process indicator is used as the target indicator for analysis and expression. The analysis process of other process indicators is the same as the analysis process of the target indicator, and they will not be described one by one here.

[0050] Step S2, classify all reference data according to the similarity between the reference data, obtain all reference groups and the process reference degree of each reference group; in each reference group, construct the parameter time series sequence and the liquid level time series sequence respectively according to the target parameters and actual liquid level at the corresponding time of all reference data.

[0051] Considering that one process indicator is fixed, if the process parameters under the other process indicators are more similar, it can be considered that the change in liquid level is mainly affected by the change in the process parameters of the current fixed process indicator; therefore, after determining the target indicator and the reference data, the embodiment of the present invention first classifies all reference data according to the similarity of the reference data, obtains all reference groups, and obtains the process reference degree of each reference group according to the classification result; wherein, the reference group is the classification result of all similar reference data, and the more similar the reference data in each reference group, the higher the confidence in analyzing the impact of the target indicator on the liquid level change, and the process reference degree reflects the analytical reference value of the impact of the target indicator on the liquid level change under the reference group.

[0052] Preferably, in one embodiment of the present invention, considering that the clustering algorithm can cluster similar data into a cluster, and the more similar the data in the cluster is, the better the clustering effect is, the greater the reference value of the clustering result is, and further, the clustering has a greater reference value for the subsequent analysis of the impact of the target index on the liquid level change; therefore, the method for obtaining the process reference degree includes:

[0053] All reference data are clustered based on the clustering algorithm, and each cluster obtained is used as a reference group;

[0054] In each reference group, the in-group outlier parameter of each reference data is obtained according to the data difference between each reference data and all other reference data; the data difference is positively correlated with the in-group outlier parameter;

[0055] The outlier parameters within the group of all reference data are combined to obtain the process reference degree of each reference group; the outlier parameters within the group are negatively correlated with the process reference degree.

[0056] As an example, firstly, the process data at all times are clustered based on the density clustering algorithm DBSCAN, where the neighborhood radius ε in the DBSCAN algorithm is obtained based on the K-distance graph, the minimum number of data points in the cluster MinPts is set to 2, and each cluster in the clustering result is used as a reference group; then the intra-cluster similarity of each reference group is analyzed to obtain the process reference degree of each reference group; the calculation formula of the process reference degree is:

[0057] ;in, is the target indicator number, is the serial number of the reference group; For target indicators The next The process reference degree of each reference group; is an exponential function with the natural constant e as base; For target indicators The next The serial number of the reference data within a reference group; For target indicators The next The total number of reference data in each reference group; For target indicators The next Within the reference group, The mean of the Euclidean distances between a reference data and all other reference data is also the target indicator. The next Reference group The within-group outlier parameter of the reference data.

[0058] In the calculation formula of the process reference degree, the larger the mean of the Euclidean distance between the reference data and all other reference data, the more the reference data deviates from the other reference data in the group, and the larger the outlier parameter in the group; then after the outlier parameters in the group of all reference data are accumulated and averaged, the negative correlation is mapped to the exponential function to adjust the logic. The larger the mean of the deviation parameter in the group, the lower its reference value and the lower the process reference degree; in other examples, the implementer can also use other negative correlation mapping methods such as inverse operation, which will not be repeated here.

[0059] It should be noted that, considering that the number of process indicators corresponding to the reference data is large, there may be a dimensionality disaster when clustering. Therefore, the implementer can choose to perform dimensionality reduction analysis on the reference data according to the actual situation, such as using the principal component analysis method to reduce the dimensionality, select some principal components such as the top 50% of the principal components in the ranking to construct the dimensionality reduction result, and then use the dimensionality reduction result as the new reference data, and obtain the Euclidean distance between different new reference data points, and then use the DBSCAN algorithm for clustering to obtain all clustering results and all reference groups; the implementer can also obtain all reference groups based on the distance clustering algorithm, which will not be repeated here.

[0060] It should be noted that DBSCAN algorithm clustering, neighborhood radius ε, minimum number of data points MinPts and principal component analysis are all existing technologies well known to those skilled in the art and will not be described in detail here.

[0061] In order to analyze the impact of the target index on the liquid level change, the embodiment of the present invention further constructs a parameter time series sequence and a liquid level time series sequence in each reference group according to the target parameters and the actual liquid level at the corresponding moment of all reference data. This facilitates the subsequent analysis and evaluation of the impact of the target index on the liquid level change according to the fluctuation correlation between the two time series corresponding to each reference group.

[0062] Preferably, in one embodiment of the present invention, the method for acquiring the parameter time series and the liquid level time series comprises:

[0063] In each reference group, all reference data are sorted in chronological order to obtain a reference data sequence; a parameter chronological sequence is constructed according to the target parameters at the corresponding moment of each reference data in the reference data sequence; a liquid level chronological sequence is constructed according to the actual liquid level at the corresponding moment of each reference data in the reference data sequence; the chronological order of the parameter chronological sequence and the liquid level chronological sequence is the same as the chronological order of the reference data sequence.

[0064] As an example, for ease of understanding, assume that the target indicator is the liquid flow rate , the reference data within a reference group include , , … , , the reference data sequence constructed by sorting the reference data in chronological order is { , , … , }, the corresponding parameter timing sequence is { , , … , }, the corresponding liquid level timing sequence is { , , … , }; Further, the parameter time series and liquid level time series of each reference group can be constructed.

[0065] Step S3, according to the fluctuation correlation between the parameter timing sequence and the liquid level timing sequence of each reference group, obtain the time-delay liquid level of each target parameter in the parameter timing sequence; according to the change difference and corresponding moment difference between each target parameter and the corresponding time-delay liquid level, combined with the process reference degree of the reference group to which the reference data at the corresponding moment of the target parameter belongs, obtain the indicator sensitivity of the target indicator.

[0066] Taking into account that there is a certain time lag in the change of the liquid level of the falling film evaporator with the change of the target parameter, that is, the liquid level will not change immediately after the target parameter is adjusted, directly analyzing the change of the liquid level with the change of the target parameter to determine the influence of the target indicator on the liquid level change may have a certain time lag deviation effect; therefore, after obtaining the parameter time series sequence and the liquid level time series sequence of each reference group, the embodiment of the present invention can analyze the time lag influence of the parameter change and the liquid level change according to the fluctuation correlation between the sequences, and obtain the time lag liquid level of each target parameter in the parameter time series sequence, so as to subsequently analyze the influence of the target indicator on the liquid level change under the premise of avoiding the time lag influence.

[0067] Preferably, in one embodiment of the present invention, considering that Dynamic Time Warping (DTW) is an algorithm for calculating the similarity between two time series, it is particularly effective when there are different speeds or time delays in the time series data; and considering that when DTW aligns two time series, there may be a one-to-many or many-to-many possibility, it is necessary to further screen and analyze in combination with the corresponding time of the target parameter; therefore, the method for obtaining the time-delay liquid level includes:

[0068] Perform DTW matching on the parameter time series and liquid level time series of each reference group, and obtain all matching liquid levels of each target parameter in each parameter time series in the liquid level time series;

[0069] A matching liquid level is selected from all matching liquid levels of each target parameter as the time-delay liquid level, the time corresponding to the time-delay liquid level is greater than the time corresponding to the target parameter, and the difference between the time corresponding to the time-delay liquid level and the time corresponding to the target parameter is the smallest.

[0070] As an example, see Figure 2 , which shows a schematic diagram of DTW matching between a parameter time series sequence and a liquid level time series sequence in a reference group provided by an embodiment of the present invention; Figure 2 In the figure, the numbers under the scale axis are the element numbers of each sequence element in the two time series, which are not corresponding moments, but can be used to approximately replace moments in the analysis process; in the DTW matching relationship between the parameter time series and the liquid level time series, there are three matching relationships: one-to-many, many-to-one, and one-to-one. From the DTW matching relationship, all matching liquid levels of each target parameter in the liquid level time series can be obtained. However, due to the influence of time lag, the corresponding sequence number of the target parameter's time lag level should be greater than the sequence number of the target parameter, so further screening is required;

[0071] For example, in the matching relationship of A1—B1, B2, and B3, since the serial number of the matching liquid level B2 is greater than the target parameter A1 and the serial number difference with A1 is the smallest, the matching liquid level B2 is selected as the time-delay liquid level of the target parameter A1; for another example, in the matching relationship of A7, A8, A9, and A10—B9, since the serial number of the matching liquid level B2 is greater than the target parameters A7 and A8, and the target parameters A7 and A8 have only one matching liquid level, the matching liquid level B9 is selected as the time-delay liquid level of the target parameters A7 and A8, while the target parameters A9 and A10 have no corresponding time-delay liquid level; for another example, in the matching relationships of A11, A12—B10 and A15—B13, B14, and B15, the corresponding time-delay liquid level cannot be screened out.

[0072] In other embodiments of the present invention, the mutual function can be used to calculate the mutual correlation coefficient between the parameter timing sequence and the liquid level timing sequence of each reference group with the help of the correlation function in MATLAB, and the peak position of the mutual coefficient can be found to obtain the time lag between the two timing sequences, and then align them to obtain the time lag liquid level of each target parameter in the parameter timing sequence.

[0073] It should be noted that in the process of matching or aligning two time series, there may be a situation where some target parameters have no corresponding time-delay liquid level. Such target parameters may not be used for subsequent analysis; or, after obtaining all target parameters and corresponding time-delay liquid levels that meet the time-delay matching conditions according to the above matching method, the time difference between each target parameter and the corresponding time-delay liquid level is calculated, and the mode, median or mean of all time differences is used as the time-delay amount, and the remaining target parameters without corresponding time-delay liquid level are further forced to match. For example, if the calculated time-delay amount is 7, the actual liquid level corresponding to the time sequence number obtained by adding 7 to the corresponding time of the target parameter is used as the forced matching time-delay liquid level of the target parameter, and no further details are given.

[0074] At this point, the time-delayed liquid level of each target parameter in each parameter time series can be obtained.

[0075] Considering the difference between the change of the target parameter and the change of the corresponding time-delay liquid level, it can be used to measure the influence of the target index on the liquid level. The smaller the change of the target parameter and the larger the change of the time-delay liquid level, the greater its influence. Therefore, when adjusting the liquid level, it is only necessary to fine-tune the process parameters under the target index to improve the efficiency of the falling film evaporator. Considering that the smaller the difference between the corresponding moment of the target parameter and the corresponding moment of the time-delay liquid level, the smaller the influence of the target index on the time-delay of changing the liquid level, so the falling film evaporator can be quickly brought into an ideal state by adjusting the process parameters under the target index.

[0076] It is also considered that when analyzing the time lag effect of the target parameter change and the target index, the process reference degree of the reference group to which the reference data at the corresponding moment of the target parameter belongs will provide a certain reference value. Therefore, the embodiment of the present invention can further obtain the index sensitivity of the target index according to the change difference and corresponding moment difference between each target parameter and the corresponding time lag liquid level, combined with the process reference degree of the reference group to which the reference data at the corresponding moment of the target parameter belongs. The index sensitivity reflects the efficiency of the adjustment of the target parameter under the target index on the liquid level change of the falling film evaporator. The greater the index sensitivity, the lower the efficiency of the change of the target parameter under the target index for changing the liquid level, and the lower the weight of considering the adjustment of the target parameter under the target index when adjusting the liquid level.

[0077] Preferably, in one embodiment of the present invention, firstly, based on the difference between each target parameter and the corresponding time-delay liquid level, a liquid level adjustment scale is obtained, and the liquid level adjustment scale reflects the influence of the target index on the liquid level, that is, the smaller the change of the target parameter under the target index, the greater the change of the liquid level can be caused; then, based on the difference between each target parameter and the corresponding time of the corresponding time-delay liquid level, an adjustment response index is obtained, and the adjustment response index reflects the time-delay influence of the target index on the change of the liquid level; and then, the two can be combined to evaluate the influence efficiency of the adjustment of the target parameter under the target index on the liquid level change of the falling film evaporator, that is, the index sensitivity; based on this, the method for obtaining the index sensitivity includes:

[0078] According to the local time series variation difference between each target parameter and the corresponding time-delay liquid level, combined with the process reference degree of the reference group to which the reference data at the corresponding moment of the target parameter belongs, the liquid level adjustment scale of the target indicator is obtained;

[0079] The negative correlation mapping result between each target parameter and the corresponding time difference of the time lag liquid level is used as the time lag parameter of each target parameter; the process reference degree of the reference group to which the reference data at the corresponding time of the target parameter belongs is used as the first reference weight of the time lag parameter of the target parameter; the corresponding time lag parameter is weighted and averaged using the first reference weight, and the normalized result of the weighted average result is used as the adjustment response indicator of the target indicator;

[0080] The index sensitivity is obtained according to the liquid level adjustment scale and the adjustment response index. The liquid level adjustment scale and the adjustment response index are positively correlated with the index sensitivity.

[0081] Preferably, in one embodiment of the present invention, the method for obtaining the liquid level adjustment scale includes:

[0082] According to the difference between each target parameter in the parameter time series and a preset number of adjacent target parameters, the parameter variation of each target parameter is obtained; according to the difference between each actual liquid level in the liquid level time series and a preset number of adjacent actual liquid levels, the liquid level variation of each actual liquid level is obtained;

[0083] The parameter variation of the target parameter plus the preset minimum non-zero positive parameter is used as the denominator, the liquid level variation of the target parameter corresponding to the time-delay liquid level plus the preset minimum non-zero positive parameter is used as the numerator, and the absolute value of the fractional ratio is used as the synchronous variation of the target parameter;

[0084] The process reference degree of the reference group to which the reference data at the corresponding moment of the target parameter belongs is used as the second reference weight of the synchronous amplitude variation of the target parameter; the corresponding synchronous amplitude variation is weighted and averaged using the second reference weight, and the normalized result of the weighted average result is used as the liquid level adjustment scale of the target indicator.

[0085] The calculation formula for the liquid level adjustment scale is: ;in, is the serial number of the target indicator; Liquid level adjustment scale for target indicator; is the serial number of the reference group; For target indicators The total number of reference groups below; For target indicators The next The serial number of the reference data within the reference group is also the target indicator The next The reference group corresponds to the serial number of the target parameter in the parameter time series; For target indicators The next The total number of reference data in the reference group, target indicator The next The total number of target parameters in the parameter time series corresponding to the reference group; is the hyperbolic tangent function; For target indicators The next In the parameter time series corresponding to the reference group, The parameter variation of the target parameter; For target indicators The next In the parameter time series corresponding to the reference group, The target parameter corresponds to the level variation of the time-delayed liquid level; is the preset minimum non-zero positive parameter; For target indicators The next In the parameter time series corresponding to the reference group, Synchronous amplitude variation of target parameters; For target indicators The next The process reference degree of each reference group is also the second reference weight; is a linear normalization function.

[0086] It should be noted that the method for obtaining the parameter variation is to take any target parameter as the parameter to be analyzed in each parameter time series sequence, select a preset number of adjacent target parameters from the parameter time series sequence as reference parameters, calculate the difference between the parameter to be analyzed and each reference parameter, further calculate the average absolute value of all the differences, and take the average absolute value as the parameter variation of the parameter to be analyzed; wherein, the preset number is 4, and the implementer can also determine it according to actual conditions; the parameter variation of each target parameter can be obtained by changing the parameter to be analyzed; similarly, the liquid level variation of each actual liquid level or the liquid level variation of each target parameter corresponding to the time-delay liquid level can be obtained; it will not be repeated here.

[0087] As an example, when screening a preset number of adjacent target parameters, the implementer can take each target parameter as the center and screen two adjacent target parameters in the parameter time series from the center to the two directions as adjacent target parameters; the target parameters at the beginning and end of the parameter time series may not be screened, and they need to be filled in by themselves; for example, a parameter time series is {1, 2, 3, 4, 5, 6, 7, 8, 9}, where the adjacent target parameters of the target parameter corresponding to 2 are 1, 3, 4, 5, and the adjacent target parameters of the target parameter corresponding to 9 are 5, 6, 7, 8, and so on. It should be noted that the implementer can also determine the screening method of the adjacent target parameters by himself, so as to obtain its local change characteristics through subsequent analysis.

[0088] In the calculation formula of the liquid level adjustment scale, the preset minimum non-zero positive parameter is 0.001 to avoid the denominator being 0 without affecting the calculation result. The implementer can also set it according to the actual situation. The larger the synchronous amplitude, the greater the liquid level change when the target parameter changes less, which indirectly shows the influence of the target indicator on the adjustment of the liquid level. The synchronous amplitude variation is normalized, and then the process reference degree is used as the weight of the corresponding normalized value, and then the weighted average is calculated to obtain the liquid level adjustment scale of the target indicator through comprehensive analysis.

[0089] The calculation formula of the adjustment response index is: ;in, is the serial number of the target indicator; It is the adjustment response indicator of the target indicator; is the serial number of the reference group; For target indicators The total number of reference groups below; For target indicators The next The serial number of the reference data within the reference group is also the target indicator The next The reference group corresponds to the serial number of the target parameter in the parameter time series; For target indicators The next The total number of reference data in the reference group, target indicator The next The total number of target parameters in the parameter time series corresponding to the reference group; is an exponential function with the natural constant e as base; For target indicators The next In the parameter time series corresponding to the reference group, The target parameters correspond to the moment; For target indicators The next In the parameter time series corresponding to the reference group, The time-delayed liquid level corresponding to the target parameter; For target indicators The next The process reference degree of each reference group is also the first reference weight; For target indicators The next In the parameter time series corresponding to the reference group, The time lag parameter of the target parameter; is a linear normalization function.

[0090] In the calculation formula of the adjustment response index, the absolute value of the time difference between the target parameter and the corresponding time-delay liquid level is mapped to In the formula, x is the independent variable, so as to adjust the logic of negative correlation mapping and normalize it; in other implementations, the implementer can also add the absolute value of the moment difference to the minimum positive parameter and then perform a reciprocal operation to achieve negative correlation mapping, or use other negative correlation mapping methods, which are not described here; the larger the absolute value of the moment difference, the slower the response of the liquid level change to the change of the target parameter, and then use the process reference degree as the weight of the corresponding negative correlation normalized value, and perform weighted averaging to obtain the adjustment response index of the target index through comprehensive analysis.

[0091] It should be noted that the first reference weight and the second reference weight are both process reference degrees, which are set to facilitate the distinction between the weighting process in the calculation formula of the adjustment response index and the calculation formula of the liquid level adjustment scale.

[0092] After obtaining the adjustment response index and liquid level adjustment scale of the target index, the index sensitivity can be further obtained; specifically, the adjustment response index and the liquid level adjustment scale are multiplied and combined, and the product is used as the index sensitivity; the larger the liquid level adjustment scale and the liquid level adjustment scale, the greater the impact of the target index adjustment on the liquid level change, and the faster the liquid level changes with the parameters.

[0093] In other embodiments of the present invention, the implementer may also use basic mathematical operations such as addition or weighted correlation or correlation mapping to merge the two, which are existing technical means and will not be described in detail here.

[0094] Step S4, obtaining the ideal liquid level of the current process data based on the preset simulation model; obtaining all recommended process data of the ideal liquid level based on the preset process recommendation model; obtaining the best recommended process data based on the current process data at the current moment and the difference between each recommended process data, combined with the indicator sensitivity of each process indicator; adjusting the liquid level of the current falling film evaporator according to the best recommended process data.

[0095] Taking into account that the actual liquid level corresponding to each process data may not be the ideal liquid level corresponding to the optimal evaporation efficiency, when subsequently adjusting and optimizing the liquid level, it is necessary to adjust the process parameters in the process data in the direction of the optimal evaporation efficiency; therefore, the present invention first needs to obtain the ideal liquid level of the current process data at the current moment, and the ideal liquid level is the theoretical liquid level when the current process data corresponds to the optimal evaporation efficiency.

[0096] It should be noted that the preset simulation model is a simulation mathematical model provided in the liquid level control method of a tubular falling film evaporator based on optimal energy consumption, which can automatically simulate and calculate the corresponding ideal liquid level according to the current process data; it is also a well-known technology and will not be elaborated here. For details, please refer to "Liquid Level Control Method of Tubular Falling Film Evaporator Based on Optimal Energy Consumption" in Volume 32, Issue 5 of "Journal of Chemical Engineering of Colleges and Universities", and the DOI of the article is: 10.3969 / j.issn.1003-9015.2018.05.023.

[0097] After obtaining the ideal liquid level of the current process data, all recommended process data of the ideal liquid level can be further obtained based on the preset process recommendation model, so as to evaluate the best recommended process data according to the adjustment complexity and efficiency when the falling film evaporator is adjusted to all recommended process data.

[0098] Preferably, in one embodiment of the present invention, considering that the liquid level is affected by multiple process parameters, one liquid level may correspond to multiple process data; considering that changes in process parameters may have a complex mutual influence relationship, and the neural network model can effectively handle the complex mutual relationship between the liquid level and multiple process parameters, and provide data decision support through model training; therefore, the method for obtaining all recommended process data of the ideal liquid level based on the preset process recommendation model includes:

[0099] A training sample set is constructed according to the process data and the corresponding actual liquid level, and a neural network model is trained based on the training sample set to obtain a process recommendation model; the ideal liquid level is input into the trained process recommendation model, and all recommended process data of the ideal liquid level are output.

[0100] As an example, implementers can collect all historical process data and the corresponding actual liquid levels of falling film evaporators of the same model and type when evaporating the same liquid material; build a training sample set based on these historical process data and actual liquid levels, train a convolutional neural network, and then use the trained convolutional neural network model as a process recommendation model, that is, by inputting one liquid level data, you can get recommended solutions for multiple process data; then input the ideal liquid level of the current process data into the trained process recommendation model, and output all recommended process data for the ideal liquid level.

[0101] It should be noted that the training and application of convolutional neural network models are existing technologies well known to those skilled in the art and will not be elaborated here. In other examples, implementers may also use other network models or deep learning algorithms to obtain process recommendation models, which will not be elaborated here.

[0102] It should be noted that, considering that the inherent properties of the liquid material cannot be changed, such as process indicators such as specific heat capacity and concentration cannot be changed, because under normal circumstances the liquid material will be directly injected into the evaporator feed port as the mother liquor raw material, changing its concentration or specific heat capacity will increase the difficulty of evaporation, so changing the inherent properties of the liquid material is basically not considered; therefore, the ideal liquid level of the current process data corresponds to the inherent process indicators in all recommended process data, which should be consistent with the current process data. Implementers can do specific screening when constructing the sample training set, or they can do screening after obtaining all recommended process data for the ideal liquid level.

[0103] Taking into account the complexity and efficiency of adjusting the falling film evaporator from the working conditions corresponding to the current process data to the working conditions corresponding to all recommended process data; therefore, the embodiment of the present invention obtains the best recommended process data based on the difference between the current process data and each recommended process data, combined with the indicator sensitivity of each process indicator; the best recommended process data is the recommended process data corresponding to the least parameter change and the fastest response during the liquid level adjustment process.

[0104] Preferably, in one embodiment of the present invention, the method for obtaining the best recommended process data includes:

[0105] Under each process index, the difference between the current process data and the process parameters corresponding to each recommended process data is taken as the process deviation; the corresponding process deviation is weighted and averaged after the negative correlation mapping of the index sensitivity of each process index, and the weighted average result is taken as the adjustment difficulty of the corresponding recommended process data;

[0106] The negative correlation mapping result of the adjustment difficulty is used as the recommendation degree of the corresponding recommended process parameters; and the best recommended process data is obtained according to the recommendation degree.

[0107] As an example, the formula for calculating the recommendation degree is: ;in, It is the serial number of the recommended process data; For the The recommended degree of each recommended process data; is an exponential function with the natural constant e as base; is the serial number of the process indicator; is the total number of process indicators; For the The recommended process data Process parameters under each process index; The current process data Process parameters under each process index; For the The index sensitivity of each process index; The current process data and The recommended process data is in Process deviation under each process index; For the The difficulty of adjusting the recommended process data.

[0108] In the recommendation calculation formula, The function is used for negative correlation mapping. The implementer may also adopt other negative correlation mapping methods, which will not be elaborated here. The larger the process deviation is and the smaller the index sensitivity under the corresponding index is, the greater the difficulty of adjustment is, that is, the liquid level change requires a larger parameter change and a longer response time.

[0109] In one embodiment of the present invention, the recommended process data corresponding to the maximum recommendation degree is taken as the best recommended process data.

[0110] At this point, the best recommended process data corresponding to the current process data can be obtained, and the corresponding components of the falling film evaporator can be adjusted to the final recommended process data, thereby optimizing and adjusting the liquid level of the current falling film evaporator.

[0111] It should be noted that adjusting the falling film evaporator is an existing technology well known to those skilled in the art and will not be described in detail here.

[0112] At this point, the current liquid level of the falling film evaporator is optimized and adjusted, and the stability of the liquid level optimization setting process of the falling film evaporator is improved based on the comprehensive adjustment of multiple process parameters.

[0113] In another embodiment of the present invention, considering that there may be a certain deviation between theory and practice, the ideal liquid level corresponding to the current best recommended process data may still have a large deviation from the ideal liquid level of the current process data, so it is necessary to perform certain repeated fine-tuning to make the ideal liquid level corresponding to the adjusted process data infinitely close to the ideal liquid level corresponding to the current process data; based on this, the embodiment of the present invention will further obtain the ideal liquid level of the best recommended process data, and obtain new best recommended process data based on the ideal liquid level of the best recommended process data; iterate continuously until the preset conditions are met, and use the latest obtained best recommended process data as the best recommended process data for the ideal liquid level of the current process data at the current moment; and finally adjust the liquid level of the current falling film evaporator based on the final best recommended process data.

[0114] The preset condition is that the difference between the ideal liquid level of the new best recommended process data and the ideal liquid level of the current process data is less than a preset threshold. The preset threshold is 5% of the ideal liquid level of the current process data, which can be set by the implementer, but should not be too large to avoid invalid adjustments.

[0115] As an example, after obtaining the best recommended process data for the first time, the best recommended process data is used as the first recommended data, and the ideal liquid level of the first recommended data is continuously obtained based on the preset simulation model, and the ideal liquid level is used as the first ideal liquid level; the liquid level difference between the first ideal liquid level and the ideal liquid level of the current process data is evaluated. If the liquid level difference is greater than the preset threshold, the best recommended process data for the first ideal liquid level is continuously obtained based on the preset process recommendation model, and the best recommended process data is used as the second recommended data; the ideal liquid level of the second recommended data is continuously obtained based on the preset simulation model, and the ideal liquid level is used as the second ideal liquid level; the liquid level difference between the second ideal liquid level and the ideal liquid level of the current process data is evaluated. If the liquid level difference is greater than the preset threshold, the above iterative steps are repeated until the liquid level difference between the ideal liquid level corresponding to the new best recommended process data and the ideal liquid level of the current process data is less than the preset threshold; the latest obtained best recommended process data is used as the adjustment target of the current process data, and then the liquid level of the falling film evaporator is adjusted using the final best recommended process data.

[0116] The present invention also proposes a falling film evaporator liquid level setting optimization system, the system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of a falling film evaporator liquid level setting optimization method described in steps S1-S4 are implemented.

[0117] In summary, the present invention first obtains the process data and actual liquid level of the falling film evaporator at each moment, takes any process indicator as the target indicator, obtains the target parameters and reference data; further classifies and obtains all reference groups, and obtains the process reference degree of each reference group; further constructs the parameter time series and liquid level time series corresponding to each reference group, and then analyzes the fluctuation correlation of the two sequences, and combines the process reference degree to obtain the indicator sensitivity of the target indicator; further obtains the best recommended process data of the current process data to adjust the current liquid level of the falling film evaporator. The present invention analyzes the impact of different process indicators on liquid level changes, obtains the indicator sensitivity of each process indicator, further combines the simulation model and the recommendation model, obtains the optimal recommended process data, and improves the stability of the falling film evaporator in the process of optimizing the liquid level setting based on the comprehensive adjustment of multiple process parameters.

[0118] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0119] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for optimizing liquid level setting of a falling film evaporator, characterized in that: The method comprises: Obtaining process data and actual liquid level of the falling film evaporator at each moment, wherein the process data includes process parameters of each process index; taking any process index as a target index, taking the process parameter of the target index as a target parameter, and taking process data without the target parameter as reference data; Classify all reference data according to the similarity between the reference data, and obtain all reference groups and the process reference degree of each reference group; in each reference group, construct a parameter time series sequence and a liquid level time series sequence according to the target parameters and the actual liquid level at the corresponding time of all the reference data; According to the fluctuation correlation between the parameter time series and the liquid level time series of each reference group, the time-delay liquid level of each target parameter in the parameter time series is obtained; according to the change difference and corresponding time difference between each target parameter and the corresponding time-delay liquid level, combined with the process reference degree of the reference group to which the reference data at the corresponding time of the target parameter belongs, the indicator sensitivity of the target indicator is obtained; Obtaining the ideal liquid level of the current process data based on the preset simulation model; obtaining all recommended process data of the ideal liquid level based on the preset process recommendation model; obtaining the best recommended process data based on the difference between the current process data at the current moment and each of the recommended process data, combined with the indicator sensitivity of each process indicator; adjusting the liquid level of the current falling film evaporator according to the best recommended process data; The method for obtaining the process reference degree includes: Clustering all the reference data based on a clustering algorithm, and using each obtained cluster as a reference group; In each of the reference groups, according to the data difference between each of the reference data and all the other reference data, an in-group outlier parameter of each of the reference data is obtained; the data difference is positively correlated with the in-group outlier parameter; The outlier parameters within the group of all the reference data are integrated to obtain the process reference degree of each reference group; the outlier parameters within the group are negatively correlated with the process reference degree; The method for obtaining the time-delay liquid level comprises: Perform DTW matching on the parameter time series sequence of each reference group and the liquid level time series sequence to obtain all matching liquid levels of each target parameter in each parameter time series sequence in the liquid level time series sequence; Select a matching liquid level from all matching liquid levels of each target parameter as a time-delay liquid level, the time corresponding to the time-delay liquid level is greater than the time corresponding to the target parameter, and the difference between the time corresponding to the time-delay liquid level and the time corresponding to the target parameter is the smallest; The method for obtaining the index sensitivity includes: According to the local time series variation difference between each target parameter and the corresponding time-delay liquid level, combined with the process reference degree of the reference group to which the reference data at the time corresponding to the target parameter belongs, the liquid level adjustment scale of the target indicator is obtained; The negative correlation mapping result between each target parameter and the time difference of the corresponding time lag liquid level is used as the time lag parameter of each target parameter; the process reference degree of the reference group to which the reference data at the time corresponding to the target parameter belongs is used as the first reference weight of the time lag parameter of the target parameter; the corresponding time lag parameter is weighted and averaged using the first reference weight, and the normalized result of the weighted average result is used as the adjustment response indicator of the target indicator; Acquiring an index sensitivity according to the liquid level adjustment scale and the adjustment response index, wherein the liquid level adjustment scale and the adjustment response index are both positively correlated with the index sensitivity; The method for obtaining the liquid level adjustment scale includes: According to the difference between each target parameter in the parameter time series and a preset number of adjacent target parameters, the parameter variation amplitude of each target parameter is obtained; according to the difference between each actual liquid level in the liquid level time series and a preset number of adjacent actual liquid levels, the liquid level variation amplitude of each actual liquid level is obtained; The parameter variation of the target parameter plus a preset minimum non-zero positive parameter is used as the denominator, the liquid level variation of the time-delay liquid level corresponding to the target parameter plus a preset minimum non-zero positive parameter is used as the numerator, and the absolute value of the fractional ratio is used as the synchronous variation of the target parameter; The process reference degree of the reference group to which the reference data at the time corresponding to the target parameter belongs is used as the second reference weight of the synchronous amplitude variation of the target parameter; the corresponding synchronous amplitude variation is weighted and averaged using the second reference weight, and the normalized result of the weighted average result is used as the liquid level adjustment scale of the target indicator; The method for obtaining the best recommended process data includes: Under each process indicator, the difference between the current process data and the process parameters corresponding to each of the recommended process data is used as the process deviation; the corresponding process deviation is weighted and averaged after the indicator sensitivity negative correlation mapping of each process indicator, and the weighted average result is used as the adjustment difficulty of the corresponding recommended process data; The negative correlation mapping result of the adjustment difficulty is used as the recommendation degree corresponding to the recommended process parameter; and the recommended process parameter corresponding to the maximum recommendation degree is used as the best recommended process data.

2. The method for optimizing liquid level setting of a falling film evaporator according to claim 1, characterized in that: The method for obtaining the parameter time series and the liquid level time series comprises: In each of the reference groups, all the reference data are sorted in chronological order to obtain a reference data sequence; a parameter timing sequence is constructed according to the target parameters at the corresponding moment of each reference data in the reference data sequence; a liquid level timing sequence is constructed according to the actual liquid level at the corresponding moment of each reference data in the reference data sequence; the timing order of the parameter timing sequence and the liquid level timing sequence are the same as the timing order of the reference data sequence.

3. The method for optimizing liquid level setting of a falling film evaporator according to claim 1, characterized in that: The method for obtaining all recommended process data of the ideal liquid level based on a preset process recommendation model includes: A training sample set is constructed according to the process data and the corresponding actual liquid level, and a neural network model is trained based on the training sample set to obtain a process recommendation model; the ideal liquid level is input into the trained process recommendation model, and all recommended process data of the ideal liquid level are output.

4. The method for optimizing liquid level setting of a falling film evaporator according to claim 1, characterized in that: After obtaining the best recommended process data, the following further includes: The ideal liquid level of the best recommended process data is obtained, and new best recommended process data is obtained based on the ideal liquid level of the best recommended process data; it is continuously iterated until the preset conditions are met, and the latest obtained best recommended process data is used as the best recommended process data for the ideal liquid level of the current process data at the current moment.

5. A falling film evaporator liquid level setting optimization system, characterized in that: The system includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of a falling film evaporator liquid level setting optimization method as described in any one of claims 1 to 4 are implemented.

Images