Modeling method of risk diagnosis model of RO membrane

Abstract

The present application relates to the technical field of membrane operation, and particularly relates to a modeling method of a RO membrane risk diagnosis model, comprising: establishing a marker event library of RO membrane operation, and marking according to the time sequence of each marker event input; constructing a basic model according to RO membrane operation data, water quality biochemical test data, expert analysis diagnosis logic and general judgment rules; inputting the preprocessed data set into the basic model to obtain a risk diagnosis result, wherein one timestamp corresponds to one risk diagnosis event in the risk diagnosis result; according to the corresponding timestamp, the marker event in the marker event library is called, and one risk diagnosis event corresponds to at least one marker event; comparing the risk diagnosis event with the corresponding marker event to obtain a deviation result, and iteratively processing the model according to the deviation result until the deviation result meets the requirement, and obtaining an optimized RO membrane risk diagnosis model. The present application improves the risk diagnosis efficiency and operation management efficiency of the RO membrane.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of membrane operation, in particular to a modeling method of an RO membrane risk diagnosis model. BACKGROUND

[0002] Currently, the risk diagnosis in the RO membrane operation process of reclaimed water mainly comes from the judgment of experts on operation data such as differential pressure, water production, water conductivity, etc., or the judgment based on standardized data and test data. Such diagnosis method has many limitations. The most obvious one is that the risk analysis diagnosis is based on the experience of professionals. When the relevant professionals of an enterprise leave, the diagnosis ability of this part will be lost. In addition, when the experience of professionals needs to be transferred from one RO membrane of a site to another RO membrane of a different working condition of another site, the learning of the operation characteristics of the RO membranes of different sites is slow, and the migration of the diagnosis and analysis knowledge of the RO membranes of different sites takes a long time. It also needs a period of time to understand the operation status of the RO membranes. In addition, similar RO membrane representations in different scenes of different sites may correspond to different risk events and different actions that need to be taken, which also needs time accumulation, resulting in that the general diagnosis method cannot be quickly applied to different scenes, greatly reducing the operation and management efficiency of the RO membranes.

[0003] Therefore, how to quickly diagnose the risks in the long-term operation process of the RO membranes is crucial for prolonging the operation life of the RO membranes and improving the operation and management efficiency of the RO membranes. SUMMARY

[0004] The present application aims to provide a modeling method of an RO membrane risk diagnosis model to solve the problem of quick diagnosis of various risks that may occur in the long-term operation process of the RO membranes.

[0005] In order to achieve the above-mentioned purpose, the present application provides a modeling method of an RO membrane risk diagnosis model, comprising the following steps:

[0006] Establishing a marker event library of RO membrane operation, and labeling according to the time sequence of the input of each marker event;

[0007] Constructing a basic model according to the RO membrane operation data, water quality biochemical test data, expert analysis diagnosis logic and general judgment rules;

[0008] Inputting the preprocessed data set into the basic model to obtain a risk diagnosis result, in which one timestamp corresponds to one risk diagnosis event;

[0009] According to the corresponding timestamp, the marker events in the marker event library are called, and one risk diagnosis event corresponds to at least one marker event;

[0010] The risk diagnosis event is compared with the corresponding mark event, a deviation result is obtained, and model iteration is performed according to the deviation result until the deviation result meets the requirement, so that an optimized RO membrane risk diagnosis model is obtained.

[0011] Optionally, the content labeled according to the time sequence of each mark event input includes a time stamp, a risk severity label, an event type label, and a keyword label.

[0012] Optionally, according to the main risks occurring in the RO membrane operation process, for each risk, a three-dimensional coordinate system is established based on three dimensions of operation data, water quality biochemical test data, and data operation characteristics, and the expert analysis and diagnosis logic is constructed based on the point position area in the three-dimensional coordinate system.

[0013] The step of obtaining a risk diagnosis result specifically includes:

[0014] According to the position of the preprocessed real-time data in the three-dimensional coordinate system, the expert analysis and diagnosis logic is called to judge the severity of the risk, and the risk diagnosis event is output.

[0015] Optionally, if the input data of the basic model comes from the same time point, the time point is taken as the time stamp of the risk diagnosis event output by the model; if the input data of the basic model comes from different times, the last time point in all input data is taken as the time stamp of the risk diagnosis event output by the model.

[0016] Optionally, when the risk diagnosis event is compared with the corresponding mark event, if the risk diagnosis event fails to correspond to all the mark events, it means that the comparison fails, otherwise, it means that the comparison succeeds.

[0017] Optionally, the number of risk diagnosis events that fail to compare is divided by the total number of risk diagnosis events to obtain a deviation value, and the basic model is iterated based on the deviation value.

[0018] Optionally, the iteration mechanism includes:

[0019] When the deviation value is greater than a first preset value, model iteration is triggered;

[0020] When the deviation value is greater than a second preset value, sequentially traversing the adjustable parameters of the basic model, wherein the second preset value is greater than the first preset value;

[0021] When the deviation value is greater than a first preset value and less than a third preset value, preferentially adjusting the adjustment parameters related to the weakest class, wherein the second preset value is greater than the third preset value.

[0022] When the deviation value is between the third preset value and the third preset value, manual tuning is performed.

[0023] Optionally, the iterative mechanism further comprises:

[0024] When the deviation value is less than the fourth preset value after the base model is continuously tuned for a preset number of times, the tuning is rolled back to before the last tuning, and the next order tunable parameter is selected according to the current rule to perform tuning.

[0025] When the deviation value still does not reach the set target after the entire tuning process is completed according to the above iterative mechanism, the iteration is terminated.

[0026] Optionally, when the deviation value is greater than the second preset value, the step of sequentially traversing the tunable parameters of the base model comprises:

[0027] Three different types of diagnostic logic are set as f1(x), f2(y), and f3(z), where x, y, and z respectively represent the RO membrane running data, water quality biochemical test data, and data analysis features.

[0028] The adjustment parameters for each index of the three diagnostic logics are set as a1, a2, a3, a4...a n , b1, b2, b3, b4...b n , and c1, c2, c3, c4...c n .

[0029] The adjustment functions for the first two diagnostic logics f1(x) and f2(y) are set as g1(z1, z2, z3, z4...z n ) and g2(z1, z2, z3, z4...z n ).

[0030] The base model is recorded as:

[0031] {g1(z1, z2, z3, z4...z n ) × F1(a1f1(x1), a2f1(x2), a3f1(x3)...a n f1(x n )),

[0032] g2(z1, z2, z3, z4...z n ) × F2(b1f2(y1), b2f2(y2), b3f2(y3)...b n f2(y n )),

[0033] F3(c1f3(z1),c2f3(z2),c3f3(z3)…c n f3(z n ))}

[0034] According to the parameter adjustment range [AL, AH] of each adjustable parameter, the parameter adjustment interval is divided into N equal parts, wherein N is a positive integer greater than 1;

[0035] Starting from a1, all adjustable parameters including a1-a n , b1-b n , c1-c n in the parameter adjustment range are traversed in all combinations;

[0036] After traversal, the risk diagnosis event is compared with the corresponding mark event, and a deviation result is obtained. If the deviation result meets the requirements, the tuning is completed, and the step is terminated. Otherwise, a number of optimal tuning models are selected, and for each adjustable parameter of the tuning models, the existing optimal value of each adjustable parameter is set as a set point, the interval is the parameter adjustment range, and the parameter adjustment interval is divided into N equal parts again. All combinations are traversed;

[0037] The above steps are repeated until the deviation result meets the requirements.

[0038] Optionally, for the running mature RO membrane, the model iteration is performed according to the set time or the membrane washing period;

[0039] For the RO membrane with a running time exceeding a certain period, when a mark event defined as an important level is added to the mark event library, the model iteration is performed;

[0040] For the newly running RO membrane, the model iteration is performed once a week or once every two weeks in the period defined as the trial operation period.

[0041] In the modeling method of the RO membrane risk diagnosis model provided by the application, at least one of the following beneficial effects is achieved:

[0042] 1) Through the method of dynamic simulation and model parameter tuning, the problems of slow learning of RO operation characteristics in different fields, long migration time of RO operation diagnosis analysis knowledge between different fields, and inability of a general judgment method to quickly adapt to different scenes are solved, the debugging time cost and labor cost of analysis and diagnosis logic algorithm are reduced, and the operation management efficiency of the RO membrane is greatly improved;

[0043] 2) By constructing a database of RO membrane operation marker events and using expert analysis and diagnostic logic, the system achieves a correspondence between accumulated field experience and expert experience. This solves the efficiency problems, potential risks (such as the inability of experts to be on duty in case of emergencies) and cost problems (hiring) that may arise from relying on experts for risk assessment during the long-term operation of RO membranes. It improves the efficiency of RO membrane risk assessment, which can better extend the membrane's operating life, reduce operating costs, and improve water production efficiency.

[0044] 3) For new problems arising under new operating conditions, events can be quickly entered into the database and risks can be quickly compared to achieve more accurate diagnosis, ensure membrane life, reduce membrane risks, and improve operational efficiency. Attached Figure Description

[0045] Those skilled in the art will understand that the accompanying drawings are provided to better understand the invention and do not constitute any limitation on the scope of the invention. Wherein:

[0046] Figure 1 A flowchart illustrating a modeling method for an RO membrane risk diagnosis model provided in an embodiment of the present invention;

[0047] Figure 2 This is a schematic diagram illustrating the extraction of a flag event according to an embodiment of the present invention;

[0048] Figure 3 This is a schematic diagram comparing risk diagnostic events and marker events according to an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, advantages, and features of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the drawings are in a very simplified form and use non-precise proportions, and are only used to facilitate and clearly illustrate the purpose of the embodiments of this invention. Please refer to the accompanying drawings to make the objectives, features, and advantages of this invention more apparent and understandable. It should be understood that the structures, proportions, sizes, etc., depicted in the accompanying drawings are only used to complement the content disclosed in the specification, for those skilled in the art to understand and read, and are not intended to limit the implementation conditions of this invention. Any modifications to the structure, changes in proportions, or adjustments to the size, if they are the same as or similar to the effects and objectives achieved by this invention, should still fall within the scope of the technical content disclosed in this invention.

[0050] As used in the present application, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. As used in the present application, the term "or" is generally employed in its sense of "and / or" unless the content clearly dictates otherwise. As used in the present application, the term "several" is generally employed in its sense of "at least one" unless the content clearly dictates otherwise. As used in the present application, the term "at least two" is generally employed in its sense of "two or more" unless the content clearly dictates otherwise. In addition, the terms "first," "second," "third," etc., are used only to describe a particular feature and do not imply relative importance or imply the number of features indicated. Thus, features defined with "first," "second," "third," etc., can explicitly or implicitly include one or at least two of the features.

[0051] In the description of the present application, unless otherwise clearly and definitely specified and limited, the terms "mounting", "connection", "connecting", "fixing" should be understood in a broad sense, for example, can be fixed connection, or detachable connection, or integral; can be mechanical connection, or electrical connection; can be direct connection, or indirect connection through intermediate medium; can be internal communication of two elements or interaction relationship between two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0052] Please refer to Figure 1 The embodiment provides a modeling method of a RO membrane risk diagnosis model, including the following steps:

[0053] S1, a flag event library of RO membrane operation is established, and each flag event is labeled according to the time sequence of input;

[0054] S2, a basic model is constructed according to the RO membrane operation data, water quality biochemical test data, expert analysis diagnosis logic and general judgment rules;

[0055] S3, the preprocessed data set is input into the basic model to obtain a risk diagnosis result, in which one timestamp corresponds to one risk diagnosis event;

[0056] S4, according to the corresponding timestamp, the flag event in the flag event library is called, and one risk diagnosis event corresponds to at least one flag event;

[0057] S5, the risk diagnosis event and the corresponding flag event are compared to obtain a deviation result, and the model is iterated according to the deviation result until the deviation result meets the requirement, and an optimized RO membrane risk diagnosis model is obtained.

[0058] Firstly, S1 is performed to establish a RO membrane operation flag event library and mark according to the time sequence of each flag event input. In this embodiment, the flag event library comes from various events such as service records, operation condition records, inspection situation summaries, reagent treatment situation summaries, on-site trend analysis, etc. in the field, such as: a certain operation index shows an upward trend / downward trend / relatively stable, a certain test index is relatively stable, the system's weekly processing volume is relatively stable, a certain test index is lower / higher, the system fouling trend slows down after adding scale inhibitor, the system's scale inhibitor consumption is lower this week, the system's salt permeability continuously abnormally rises, the system's water production continuously abnormally drops, etc.

[0059] Preferably, the content of marking according to the time sequence of each flag event input includes a time stamp, a risk severity label, an event type label and a keyword label. In this embodiment, the events in the flag event library are marked with C1, C2, C3...Cn according to the time sequence of input, and all events are provided with a severity label, a type label and a keyword label. The severity label is divided into low, medium and high; the type label is divided into operation event, water quality event, risk event, maintenance event, threshold event, etc., which can be added or deleted according to the needs of the field; the threshold event refers to events such as membrane cleaning, change of reagent addition type, change of reagent addition concentration, change of maintenance method, etc. which will affect the operation state or health state of the membrane; the keyword label includes index name, risk name such as fouling, membrane oxidation, microbial risk, etc.

[0060] The editing users of the flag event library are divided into class I users and class II users. The permission of the class I users is to mark existing events and add time stamps, as well as to submit applications for new events, new types and new labels, and the class II users are responsible for approving new events, new types and new labels.

[0061] The establishment of the flag event library can be directly copying the existing RO membrane operation flag event library, or establishing a new flag event library according to various event summary tables in the field, or adding new events in the field based on the existing RO membrane operation flag event library.

[0062] The collection of the RO membrane operation mark event comes from the historical report and the inspection record, and also comes from the information collected by the I-class user in time; in the input process of the collection, the I-class user can determine whether the input mark event already exists by the existing event entry displayed by the automatic search through the keyword. When the I-class user determines that the input event is not the existing event in the event library, the I-class user submits the new event; the I-class user submits the definition of the severity, the type and the keyword at the same time; after the submission, the II-class user confirms the event description, the severity, the type and the keyword, and after the confirmation, the incremental construction of the mark event library is realized, and the new event is arranged in sequence in the existing event. In addition, the II-class user periodically cleans up, reorganizes or deletes the event in the event library.

[0063] In the embodiment, the marking process of the mark event is as follows:

[0064] Online recording of the RO membrane operation event: the online recording of the RO membrane operation event refers to that when the user logs in the RO membrane risk diagnosis model, the original paper record or Excel record of the RO membrane operation event is recorded in the system;

[0065] RO membrane operation event searching and comparison: after the online recording is completed, the system automatically searches the keyword and recommends the three event records with the highest matching degree

[0066] RO membrane operation event confirmation and marking: after the system automatically compares and recommends the most matched event record, the I-class user manually confirms and records the marking record matched by the event.

[0067] Then, S2 is performed, a basic model is constructed according to the RO membrane operation data, the water quality biochemical test data, the expert analysis diagnosis logic and the general judgment rule. And S3 is performed, the preprocessed data set is input into the basic model, and a risk diagnosis result is obtained, in which one time stamp corresponds to one risk diagnosis event.

[0068] In the embodiment, according to the main risks occurring in the RO membrane operation process, a three-dimensional coordinate system is established based on the operation data, the water quality biochemical test data and the data operation characteristics for each risk, and the expert analysis diagnosis logic is constructed based on the point position area in the three-dimensional coordinate system;

[0069] The step of obtaining the risk diagnosis result specifically includes:

[0070] According to the position of the preprocessed real-time data in the three-dimensional coordinate system, the expert analysis diagnosis logic is called to judge the severity of the risk, and the risk diagnosis event is output.

[0071] In this embodiment, the main risks that may occur during the operation of the RO membrane are assessed primarily from the perspectives of membrane damage and membrane operating efficiency. Risks related to membrane damage include membrane oxidation and scaling; risks related to membrane operating efficiency include a decrease in permeate flow and salt permeability, and the accumulation of microorganisms.

[0072] The three dimensions are: operational data such as water production, conductivity, and pressure; water quality biochemical test data such as iron, aluminum, silica, and SDI; and data operational characteristics. The data operational characteristics refer to the mathematical meaning of the operational data and water quality biochemical test data over time, such as increases, decreases, continuous increases, continuous decreases, user-defined fluctuations, and correlations among multiple data points.

[0073] The preprocessing includes data filtering, effective data screening, data overbalancing, and data normalization.

[0074] The expert analysis and diagnosis logic refers to the expert analysis and diagnosis logic constructed by experts for RO membrane risk assessment based on the point area in the three-dimensional coordinate system. It includes the analysis and judgment of risk phenomena, the speculation of the causes of risks, and suggestions for further screening or maintenance actions required for existing risks. The logic is invoked according to the location of the influencing factors constructed in the expert analysis and diagnosis logic, and outputs textual reference opinions, which is the risk diagnosis.

[0075] In this embodiment, if the input data of the basic model comes from the same point in time, then that point in time is used as the timestamp of the risk diagnosis event output by the model; if the input data of the basic model comes from different times, then the last point in time among all the input data is used as the timestamp of the risk diagnosis event output by the model; if the input data of the basic model comes from a time range, then the last point in time of the time range of all the input data is used as the timestamp of the risk diagnosis event output by the model.

[0076] Next, S4 is executed, and according to the corresponding timestamp, the flag events in the flag event library are retrieved. Each risk diagnosis event corresponds to at least one flag event.

[0077] Finally, S5 is executed to compare the risk diagnosis event with the corresponding flag event to obtain the deviation result, and the model is iterated based on the deviation result until the deviation result meets the requirements to obtain the optimized RO membrane risk diagnosis model.

[0078] For example, such as Figures 2-3 As shown, from risk diagnosis event D nThe corresponding timestamp YYYY-MM-DD HH:mm:ss starts the time retrieval, and the first threshold event C is retrieved k The timestamp YYYY1-MM1-DD1 HH1:mm1:ss of the threshold event C is extracted k All marker events C between YYYY-MM-DD HH:mm:ss and YYYY1-MM1-DD1 HH1:mm1:ss1 are extracted (not including C k ), and the obtained marker events are C n , C n+1 , C n+2 , C n+3 ...

[0079] As shown in Figure 3 , then compare the risk diagnosis event D n and the corresponding marker event C n , C n+1 , C n+2 , C n+3 ... If any of D n and C n , C n+1 , C n+2 , C n+3 ... corresponds successfully, the comparison of D n is successful, and if all of D n and C n , C n+1 , C n+2 , C n+3 ... do not correspond successfully, the comparison fails.

[0080] Then divide the number of risk diagnosis events that failed by the total number of risk diagnosis events to obtain a deviation value, and iterate the base model based on the deviation value.

[0081] Further, the iteration mechanism includes:

[0082] When the deviation value is greater than a first preset value, triggering model iteration;

[0083] When the deviation value is greater than a second preset value, sequentially traversing the adjustable parameters of the base model, wherein the second preset value is greater than the first preset value;

[0084] When the deviation value is greater than a first preset value and less than a third preset value, preferentially adjusting the adjustment parameters related to the weakest class diagnosis, wherein the second preset value is greater than the third preset value;

[0085] When the deviation value is between the third preset value and the third preset value, manual tuning is performed.

[0086] Further, the iterative mechanism further comprises:

[0087] When the deviation value is between the third preset value and the third preset value, manual tuning is performed.

[0088] When the deviation value is between the third preset value and the third preset value, manual tuning is performed.

[0089] In the embodiment, the iterative mechanism comprises:

[0090] Rule 1: Model iteration triggering rule: when the deviation is greater than 5%, the model iteration is triggered;

[0091] Rule 2: Model sequential iteration rule: when the deviation meets a certain preset rule, such as being greater than or equal to 50%, the adjustable parameters of the base model are sequentially traversed;

[0092] Rule 3: Prioritized reinforcement iteration rule: when the deviation meets a certain preset rule, such as being less than or equal to 30%, the related tuning parameters are prioritized according to the weakest class diagnosis;

[0093] Rule 4: Manual intervention iteration rule: when the deviation meets a certain preset rule, such as being between 30% and 50%, automatic iteration is stopped, and manual tuning rule setting is performed;

[0094] Rule 5: Model storage rule: model version saving and rollback mechanism;

[0095] Rule 6: Model iteration rollback rule: when the model is continuously tuned for a certain number of times (such as 100 times) and the deviation optimization is less than a fourth preset value (such as 5%), the tuning is rolled back to the previous time, and the next order adjustable parameter is selected according to the current rule for tuning.

[0096] Rule 7: Model iteration termination rule: in addition to when the tuning meets the preset deviation requirement, when the entire tuning process is completed according to the sequential rule and the rollback rule, but the deviation requirement is still not met, the model iteration is terminated.

[0097] Preferably, when the deviation value is greater than the second preset value, the step of sequentially traversing the adjustable parameters of the base model comprises:

[0098] Set three different types of diagnostic logic for: f1(x), f2(y), f3(z), wherein x, y, z respectively represent the RO membrane running data, water quality biochemical test data and data analysis characteristics;

[0099] Set the adjustment parameters for each index of the three diagnostic logics respectively: a1, a2, a3, a4...a n ; b1, b2, b3, b4...b n ; c1, c2, c3, c4...c n ;

[0100] Set the adjustment functions for the first two diagnostic logics f1(x) and f2(y) respectively: g1(z1, z2, z3, z4...z n ), g2(z1, z2, z3, z4...z n );

[0101] Then the basic model is recorded as:

[0102] {g1(z1, z2, z3, z4...z n )×F1(a1f1(x1), a2f1(x2), a3f1(x3)...a n f1(x n )),

[0103] g2(z1, z2, z3, z4...z n )×F2(b1f2(y1), b2f2(y2), b3f2(y3)...b n f2(y n )),

[0104] F3(c1f3(z1), c2f3(z2), c3f3(z3)...c n f3(z n )}

[0105] When the deviation value triggers the model sequential iteration rule, the following steps are performed:

[0106] According to the adjustment range [AL, AH] of each adjustable parameter, the adjustment interval is divided into N equal parts (such as 10 equal parts), wherein N is a positive integer greater than 1;

[0107] Starting from a1, all adjustable parameters including a1-a n , b1-b n , c1-c n are traversed in all combinations within the adjustment range;

[0108] After the traversal, the risk diagnosis event is compared with the corresponding mark event to obtain a deviation result. If the deviation result meets the requirement, the tuning is completed and the step is terminated. Otherwise, the optimal several sets of tuning models are selected. For each tunable parameter of the optimal several sets of tuning models, the existing optimal value of each tunable parameter is taken as a set point, a tuning range is taken as a tuning interval, and the tuning interval is divided into N equal parts again to perform traversal of all combinations.

[0109] The above steps are repeated until the deviation result meets the requirement. If the requirement of the deviation value cannot be met after continuous multiple times (such as 3 times), the model iteration is terminated, and manual intervention is performed.

[0110] In the embodiment, when the deviation value triggers the model priority reinforcement iteration rule, the following steps are performed:

[0111] When the generated results D1, D2, D3... and C1, C2, C3... of the running model correspond one by one, the deviation degree scoring calculation is performed on the dynamic simulation deviation performance of f1(x), f2(y), and f3(z) respectively.

[0112] For the diagnosis logic with the lowest deviation degree score, the tuning of the tunable parameters related to the diagnosis logic is performed according to the tuning steps of the model order iteration rule, and the dynamic simulation is performed in the complete model.

[0113] If the tuning of the diagnosis logic with the lowest score can meet the requirement of the overall deviation degree after the tuning, the model iteration is terminated.

[0114] If the requirement of the deviation value is not met, the tuning of the tunable parameters in the diagnosis logic with the second-lowest score is performed according to the tuning steps of the model order iteration rule and the dynamic simulation.

[0115] If the requirement of the deviation value is not met, the tuning of the tunable parameters related to the diagnosis logic is performed according to the tuning steps of the model order iteration rule from low to high according to the score.

[0116] If the requirement of the deviation value is not met, the tuning of all tunable parameters in the model is performed according to the tuning steps of the model order iteration rule until the iteration is completed or terminated.

[0117] After the optimized RO membrane risk diagnosis model is obtained, it can be used for rapid diagnosis of risks in the long-term operation of the RO membrane. For new problems that occur in new working conditions, the event can also be quickly entered into the database and compared with the risk problem to achieve more accurate diagnosis, protect the service life of the membrane, reduce the risk of the membrane, and improve the operation efficiency.

[0118] Preferably, for the mature RO membrane, the model iteration is performed according to the set time or the membrane washing period.

[0119] For the RO membrane with a running time exceeding a certain time, when each new sign event defined as an important level in the sign event library is added, the model iteration is performed;

[0120] For the newly running RO membrane, the model iteration is performed once a week or once every two weeks in the period defined as the trial operation period.

[0121] In summary, the embodiment of the present application provides a modeling method of the RO membrane risk diagnosis model, through the method of dynamic simulation and model parameter optimization, the problems of slow learning of RO operation characteristics in different fields, long migration time of RO operation diagnosis analysis knowledge between different fields, and the problem that the general judgment method cannot be quickly applied to different scenes are solved, the debugging time cost and the labor cost of the analysis diagnosis logic algorithm are reduced, and the operation management efficiency of the RO membrane is greatly improved.

[0122] In addition, it should be recognized that, although the present application has been disclosed as above with preferred embodiments, the above embodiments are not intended to limit the present application. For any person skilled in the art, many possible changes and modifications or equivalent embodiments of the above disclosed technical content can be made without departing from the scope of the present application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, without departing from the content of the present application, still belongs to the protection scope of the present application.

Claims

1. A method for modeling a RO membrane risk diagnosis model, the method comprising: The method comprises the following steps: a library of marker events of RO membrane operation is established, and each marker event is labeled according to the time sequence of input; a basic model is constructed according to the RO membrane operation data, water quality biochemical test data, expert analysis and diagnosis logic and general judgment rules; the preprocessed data set is input into the basic model to obtain a risk diagnosis result, in which one timestamp corresponds to one risk diagnosis event; the marker events in the library are retrieved according to the corresponding timestamp, and one risk diagnosis event corresponds to at least one marker event; the risk diagnosis event and the corresponding marker event are compared to obtain a deviation result, and the basic model is iterated according to the deviation result until the deviation result meets the requirement, and an optimized RO membrane risk diagnosis model is obtained.

2. The modeling method of the RO membrane risk diagnosis model according to claim 1, wherein, The labeling according to the time sequence of input of each marker event includes a timestamp, a risk severity label, an event type label and a keyword label.

3. The modeling method of the RO membrane risk diagnosis model according to claim 1, wherein, According to the main risks occurring in the RO membrane operation process, a three-dimensional coordinate system is established based on the operation data, water quality biochemical test data and data operation characteristics for each risk, and the expert analysis and diagnosis logic is constructed based on the point area in the three-dimensional coordinate system; The step of obtaining a risk diagnosis result specifically comprises: the expert analysis and diagnosis logic is called to judge the severity of the risk according to the position of the preprocessed real-time data in the three-dimensional coordinate system, and the risk diagnosis event is output.

4. The modeling method of a RO membrane risk diagnosis model according to claim 1, wherein, If the input data of the basic model comes from the same time point, the time point is taken as the timestamp of the risk diagnosis event output by the model; if the input data of the basic model comes from different times, the last time point in all input data is taken as the timestamp of the risk diagnosis event output by the model.

5. The modeling method of a RO membrane risk diagnostic model according to claim 1, wherein, When the risk diagnosis event and all the marker events fail to correspond, it means that the comparison fails, otherwise, it means that the comparison succeeds.

6. The modeling method of the RO membrane risk diagnosis model according to claim 5, wherein, The number of risk diagnosis events with failed comparison is divided by the total number of risk diagnosis events to obtain a deviation value, and the basic model is iterated based on the deviation value.

7. The modeling method of a RO membrane risk diagnostic model according to claim 1, wherein, The iteration mechanism comprises: when the deviation value is greater than a first preset value, model iteration is triggered; when the deviation value is greater than a second preset value, the adjustable parameters of the basic model are sequentially traversed, wherein the second preset value is greater than the first preset value; when the deviation value is greater than the first preset value and less than a third preset value, the weakest class diagnosis related adjustable parameter is preferentially adjusted, wherein the second preset value is greater than the third preset value; when the deviation value is between the third preset value and the third preset value, manual adjustment is performed.

8. The modeling method of the RO membrane risk diagnosis model according to claim 7, characterized in that, The iteration mechanism further comprises: when the basic model is continuously adjusted for a certain preset number of times, the optimization of the deviation value is less than a fourth preset value, the model is rolled back to the previous adjustment, and the next order adjustable parameter is selected according to the current rule for adjustment. When the deviation value does not reach the set target after the whole tuning process is completed according to the above iteration mechanism, the iteration is terminated. 9.The modeling method of the RO membrane risk diagnosis model according to claim 7, wherein, The step of sequentially traversing the tunable parameters of the basic model when the deviation value is greater than the second preset value specifically comprises: Three different types of diagnostic logic are set as f1(x), f2(y) and f3(z), wherein x, y and z respectively represent the RO membrane operation data, water quality biochemical test data and data analysis features; Set the adjustment parameters for each indicator of the three diagnostic logics respectively: a1, a2, a3, a4...a n ; b1, b2, b3, b4...b n ; c1, c2, c3, c4...c n ; Let the adjustment functions for the first two diagnostic logics f1(x) and f2(y) be set as g1(z1, z2, z3, z4...z n ), g2(z1, z2, z3, z4...z n ) respectively. The basic model is recorded as: {g1(z1,z2,z3,z4…z n )×F1(a1f1(x1),a2f1(x2),a3f1(x3)…a n f1(x n )), g2(z1,z2,z3,z4...z n ) x F2(b1f2(y1), b2f2(y2), b3f2(y3)...b n f2(y n )), F3 (c1f3(z1), c2f3(z2), c3f3(z3)... c n f3(z n ))} According to the tuning range [AL, AH] of each tunable parameter, the tuning interval is equally divided into N equal parts, wherein N is a positive integer greater than 1; From a1, all adjustable parameters including a1-a n b1-b n c1-c n in the tuning range, all combinations are traversed; After the traversal, the risk diagnosis event is compared with the corresponding flag event to obtain a deviation result. If the deviation result meets the requirement, the tuning is completed and the step is terminated. Otherwise, the optimal several groups of tuning models are selected. For each tunable parameter of the optimal several groups of tuning models, the existing optimal value of each tunable parameter is taken as a set point, the interval is taken as a tuning range, the tuning interval is equally divided into N equal parts again, and the traversal of all combinations is performed. The above steps are repeated until the deviation result meets the requirement.

10. The modeling method of a RO membrane risk diagnostic model according to claim 1, wherein, For the mature RO membrane, model iteration is performed according to the set time or membrane washing period; For the RO membrane with a running age exceeding a certain age, model iteration is performed when each flag event defined as an important level is added to the flag event library; For the newly running RO membrane, model iteration is performed once a week or once every two weeks during the period defined as the trial operation period.

Images