Method of predicting residual chlorine in water supply systems

• US 5,675,504 A
• Filed: 01/23/1996
• Issued: 10/07/1997
• Est. Priority Date: 12/15/1995
• Status: Expired due to Fees

• Current Assignee: Universite Laval
• Original Assignee: Universite Laval
• Inventors: Rodriguez, Manuel, Serodes, Jean
• Primary Examiner(s): Trammell, James P.
• Assistant Examiner(s): KEMPER, MELANIE A
• Application Number:
  US08/589,869
• Time in Patent Office: 623 Days
• Field of Search: 364/496, 364/497, 364/499, 364/500, 364/550, 364/578, 395/914, 395/902, 395/903, 395/911, 395/21, 210/754, 210/739, 210/742
• US Class Current: 702/25
• CPC Class Codes: G01N 33/182 specific anions in water electrochemical analysis G01N27/26; detection of ions by colorimetry G01N31/22
• CPC Class Codes: G06N 3/049 Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
• CPC Class Codes: Y10S 706/911 Nonmedical diagnostics

Abstract

A method of configuring an artificial neural network for predicting residual chlorine concentration in water contained in a storage tank of a water supply system, the storage tank having an inlet for admitting water into the tank and an outlet for discharging water from the tank. The method of the invention comprises the steps of (a) collecting historical data representative of selected operational and water quality parameters associated with chlorine demand in the tank; (b) scaling the data collected in step (a); (c) organizing the data scaled in step (b) in the form of a set of time-lagged data; and (d) processing the data organized in step (c) through an artificial neural network by scanning a window over the set of data, the window having a size corresponding to a sub-set of the data, to associate a predicted value with a respective one of the sub-sets scanned by the window, the predicted value being representative of the residual chlorine concentration at the tank outlet. The artificial neural network is thus configured to recognize a set of data processed therethrough as corresponding substantially to one of the sub-sets of data and to associate with the recognized set of data the optimized predicted value associated with the sub-set of data.

Claims

1. A method of configuring an artificial neural network for predicting residual chlorine concentration in water contained in a storage tank of a water supply system, said storage tank having an inlet for admitting water into the tank and an outlet for discharging water from the tank, said method comprising the steps of:

a) collecting historical data representative of selected operational and water quality parameters associated with chlorine demand in said tank;

b) scaling the data collected in step (a);

c) organizing the data scaled in step (b) in the form of a set of time-lagged data;

d) processing the data organized in step (c) through an artificial neural network having a plurality of nodes arranged to define an input layer, an output layer and at least one hidden layer therebetween, by scanning a window over said set of data, said window having a size corresponding to a sub-set of said data, to associate a predicted value with a respective one of the sub-sets scanned by said window, said predicted value being representative of the residual chlorine concentration at the tank outlet; and

e) repeating step (d) while varying the size of said window and the number of nodes of said at least one hidden layer so as to optimize said predicted value;

whereby said artificial neural network is configured to recognize a set of data processed therethrough as corresponding substantially to one of said sub-sets of data and to associate with the recognized set of data the optimized predicted value associated with said one sub-set of data.

2. A method as claimed in claim 1, wherein said selected operational and water quality parameters include flowrate of water admitted into said tank, water temperature, chlorine concentration in water at the tank inlet, dose of chlorine injected in water and residual chlorine concentration in water at the tank outlet, and wherein the water has a predetermined mean residence time in said tank.

3. A method as claimed in claim 2, wherein the historical data is collected in step (a) at a frequency ranging from one-fourth to one-half of said mean residence time.

4. A method as claimed in claim 3, wherein said predicted value is determined at a time delay relative to a selected time reference, said time delay corresponding to the frequency of data collecting.

5. A method as claimed in claim 1, wherein the collected data is scaled in step (b) linearly between 0 and 1.

6. A method as claimed in claim 1, wherein said artificial neural network comprises one hidden layer of nodes.

7. A method as claimed in claim 2, wherein said window has a size such as to initially scan a sub-set of data having a predetermined number of time intervals corresponding to said mean residence time and wherein the window size is thereafter varied in step (e).

8. A method as claimed in claim 7, wherein the input layer of said artificial neural network has a number of nodes at least equal to the number of said parameters multiplied by said predetermined number of time intervals.

9. A method as claimed in claim 8, wherein said artificial neural network comprises one hidden layer having a number of nodes ranging from 1 to the number of nodes of said input layer.

10. A method as claimed in claim 1, wherein said artificial neural network includes a plurality of connection weights each associated with a respective one of said nodes of said at least one hidden layer, and wherein step (d) is repeated in step (e) a number of times using the same set of historical data while adjusting said connection weights according to a backpropagation algorithm until a root mean squarred error equal to or less than 0.15 is obtained between a desired predicted value and the predicted value provided at said output layer.

11. A method as claimed in claim 10, wherein said backpropagation algorithm is based on a gradient-descent delta rule and has a constant learning rate of 0.3 and a constant momentum of 0.9.

12. A method as claimed in claim 10, wherein the number of times that step (d) is repeated is about 70.

13. A method as claimed in claim 10, wherein the number of nodes of said at least one hidden layer is varied in step (e) until the smallest root mean squarred error is obtained.

14. A method as claimed in claim 10, wherein the size of said window is varied in step (e) until the smallest root mean squared error is obtained.

15. A method as claimed in claim 10, wherein about 90% of said set of historical data is used for training said artificial neural network and about 10% of said set of historical data is used for evaluating same.

16. A method of predicting residual chlorine concentration in water contained in a storage tank of a water supply system, said storage tank having an inlet for admitting water into the tank and an outlet for discharging water from the tank, said method comprising the steps of:

a) collecting present and past data representative of selected operational and water quality parameters associated with chlorine demand in said tank;

b) scaling the data collected in step (a);

c) organizing the data scaled in step (b) in the form of a set of time-lagged data; and

d) processing the data organized in step (c) through an artificial neural network configured by a method as defined in claim 1, to provide at the output layer of said network a predicted value representative of the residual chlorine concentration at the tank outlet.

17. A method as claimed in claim 16, wherein said selected operational and water quality parameters include flowrate of water admitted into said tank, water temperature, chlorine concentration in water at the tank inlet, dose of chlorine injected in water and residual chlorine concentration in water at the tank outlet, and wherein the water has a predetermined mean residence time in said tank.

18. A method as claimed in claim 17, wherein the present and past data are collected in step (a) at a frequency ranging from one-fourth to one-half of said mean residence time.

19. A method as claimed in claim 18, wherein prediction of the residual chlorine at the tank outlet is performed at a time delay relative to present time, said time delay corresponding to the frequency of data collecting.

20. A method as claimed in claim 18, wherein the collected data is scaled in step (b) linearly between 0 and 1.

21. A method as claimed in claim 16, wherein said storage tank is located downstream of a post-chlorination station.

22. A method as claimed in claim 16, wherein said storage tank is located downstream of a re-chlorination station.