Apparatus and method for analyzing influence variable on membrane fouling of seawater desalination system

Patent metadata

Field	Value
Publication number	US-10472254-B2
Application number	US-201715702669-A
Country	US
Kind code	B2
Filing date	Sep 12, 2017
Priority date	Sep 12, 2016
Publication date	Nov 12, 2019
Grant date	Nov 12, 2019

Abstract

This disclosure relates to an apparatus and method for analyzing an influence variable on membrane fouling of a seawater desalination system, wherein influence variables other than variables having a low degree of influence, among variables affecting the membrane, are selected, and the influence thereof on membrane fouling is used to derive an equation. The apparatus includes a variable storage unit configured to store variables affecting membrane fouling of a seawater desalination system, a dominant variable selection unit configured to select at least one dominant variable among the variables through at least one algorithm, and an equation derivation unit configured to derive a specific equation based on a correlation between the selected dominant variable and the membrane fouling.

First claim

What is claimed is:

1. An apparatus for analyzing membrane fouling of a seawater desalination system, the apparatus comprising:
a variable reception unit configured to receive variables affecting membrane fouling of a seawater desalination system collected by a plurality of sensors included in the seawater desalination system;
a memory that includes a variable storage unit configured to store the received variables affecting membrane fouling of a seawater desalination system;
a processor that includes a dominant variable selection unit configured to select at least one dominant variable among the stored variables by
performing a multiple linear regression in which each dependent variable (y) of a single feature variable (x j ) belonging to a feature matrix (X) of the variables affecting the membrane fouling is determined using a PME (Progressive Model Enhancement) algorithm,
repeating a process of at least one of sequentially including the variables affecting the membrane fouling in the multiple linear regression until improvement in the multiple linear regression is no longer obtained, and sequentially excluding the variables affecting the membrane fouling in the multiple linear regression until improvement in the multiple linear regression is no longer obtained, and
determining the improvement based on a change in a likelihood of the multiple linear regression;
a derivation unit configured to determine a correlation between the selected dominant variable and the membrane fouling; and
a variable control unit configured to control the seawater desalination system to control the dominant variable.

2. The apparatus of claim 1 , wherein the variables include at least one variable measured during an intake process, at least one variable measured after a UF (Ultra Filtration) process, at least one variable measured during a DAF (Dissolved Air Flotation) process, at least one variable measured before an RO (Reverse Osmosis) process, and at least one variable measured after an RO process.

3. The apparatus of claim 1 , wherein the processor is configured to select the at least one dominant variable using at least one of SPI (Single-Predictor Identification), PME (Progressive Model Enhancement), BMA (Bayesian Model Averaging), and GA (Genetic Algorithm).

4. The apparatus of claim 1 , wherein the processor is configured to select a variable chosen in common by at least two of SPI (Single-Predictor Identification), PME (Progressive Model Enhancement), BMA (Bayesian Model Averaging) and GA (Genetic Algorithm) as the dominant variable.

5. The apparatus of claim 1 , wherein the derivation unit is configured to determine the correlation based on a PME algorithm.

6. The apparatus of claim 1 , wherein the dominant variable selection unit is configured to select, as an influence variable candidate, a feature variable that has a p-value is lower than a preset significance level (α) in a simple regression that describes each dependent variable (y) of a single feature variable (x j ) belonging to a feature matrix (X) of the variables affecting the membrane fouling.

7. The apparatus of claim 1 , wherein the dominant variable selection unit is further configured to
apply a penalty to a number of variables contained in a model used by the processor to select the at least one dominant variable using a likelihood-based AIC (Akaike Information Criterion).

8. The apparatus of claim 1 , wherein the dominant variable selection unit is configured to
perform a multiple linear regression in which each dependent variable (y) of a single feature variable (x j ) belonging to a feature matrix (X) of the variables affecting the membrane fouling is determined using a BMA algorithm, and
determine a posterior probability of a model produced by M k , which is any subset of feature variables (x 1 , . . . , x (p+q) ).

9. The apparatus of claim 1 , wherein the dominant variable selection unit is configured to
perform a multiple linear regression in which each dependent variable (y) of a single feature variable (x j ) belonging to a feature matrix (X) of the variables affecting the membrane fouling is determined using a GA algorithm,
input a MSE (Mean Squared Error) of a multiple linear regression residual to a fitness function f(⋅), and
repeat a genetic operation process including selection, crossover and mutation in an initial population M P ={M 1 , . . . , M g }.

10. The apparatus of claim 1 ,
wherein the dominant variable selected unit is configured to select the feature variable using each of four algorithms including SPI (Single-Predictor Identification), PME (Progressive Model Enhancement), BMA (Bayesian Model Averaging), and GA (Genetic Algorithm) in order to select a set of feature variables S min and a set of feature variables S max ,
wherein the feature variables S min and the feature variables S max are selected from a set of influence variable candidates S=(S SPI , S PME , S BMA , S GA ),
wherein the selected feature variables S min are included among the influence variable candidates S at least H times, and
wherein the selected feature variables S max are included among the influence variable candidates S at least once.

11. A method of analyzing membrane fouling of a seawater desalination system, the method comprising:
a variable reception step of receiving variables affecting membrane fouling of a seawater desalination system collected by a plurality of sensors included in the seawater desalination system;
a variable storage step of storing, in a memory, the received variables affecting membrane fouling of a seawater desalination system;
a dominant variable selection step of selecting, using a processor, at least one dominant variable among the stored variables by
performing a multiple linear regression in which each dependent variable (y) of a single feature variable (x j ) belonging to a feature matrix (X) of the variables affecting the membrane fouling is determined using a PME (Progressive Model Enhancement) algorithm,
repeating a process of at least one of sequentially including the variables affecting the membrane fouling in the multiple linear regression until improvement in the multiple linear regression is no longer obtained, and sequentially excluding the variables affecting the membrane fouling in the multiple linear regression until improvement in the multiple linear regression is no longer obtained, and
determining the improvement based on a change in a likelihood of the multiple linear regression;
determining a correlation between the selected dominant variable and the membrane fouling; and
a variable control step of controlling the seawater desalination system to control dominant variable.

12. The method of claim 11 , wherein the variables include at least one variable measured during an intake process, at least one variable measured after a UF process, at least one variable measured during a DAF process, at least one variable measured before an RO process, and at least one variable measured after an RO process.

13. The method of claim 11 , wherein the determining step includes determining the correlation using at least one of SPI (Single-Predictor Identification), PME (Progressive Model Enhancement), BMA (Bayesian Model Averaging), and GA (Genetic Algorithm).

14. The method of claim 11 , wherein the dominant variable selection step includes selecting a dominate variable identified chosen in common by at least two of SPI (Single-Predictor Identification), PME (Progressive Model Enhancement), BMA (Bayesian Model Averaging) and GA (Genetic Algorithm).

15. The method of claim 11 , wherein the determining

Assignees

• Doosan Heavy Ind & Construction Co Ltd
• Doosan Heavy Ind Contstruction Co Ltd

Inventors

• Kim Seongjoon
• Moon Jonghyuk
• Park Sungju
• Hwang Juwon

Classifications

• B01D2321/40

  Automatic control of cleaning processes · CPC title

• C02F2209/11

  Turbidity · CPC title

• G05B23/0229

  knowledge based, e.g. expert systems; genetic algorithms · CPC title

• C02F2209/10

  Solids, e.g. total solids [TS], total suspended solids [TSS] or volatile solids [VS] · CPC title

• B01D2311/04

  Specific process operations in the feed stream; Feed pretreatment · CPC title