Water distribution amount predicting system

Details Water distribution amount predicting system Water distribution amount predicting system

2002-10-03

2004-09-21

Cecil, Terry K. (Department: 1723)

Liquid purification or separation

With alarm, indicator, register, recorder, signal or...

Responsive to fluid flow

C210S143000, C700S283000, C702S180000, C702S182000

Reexamination Certificate

active

06793806

ABSTRACT:

TECHNICAL FIELD
The present a water distribution amount predicting system and, more particularly, to a water distribution amount predicting system which predicts the amount of water to be distributed from water purification facilities to water customers.
BACKGROUND ART
In water purification facilities, in distributing water to water customers such as factories, companies, stores, and homes, a predetermined water purification process is performed first, and then the water is distributed to each water customer. In general, in a water purification process, for example, a coagulant is charged into raw water taken in from a river or the like to form impurities into floc (particles) so as to make it settle, and the water is filtered. Since this process includes floc formation, sedimentation, and the like, it takes three to six hours for the entire water purification process. In addition, the amount of water that can be purified in a unit time is limited by the equipment size and the like. Furthermore, since a water purification process requires enormous power, it is required that the water purification process be done at nighttime when the power cost is comparatively low.
Owing to such limitations, in order to stably supply water to water customers, it is required that a water distribution amount be accurately predicted, and a necessary amount of water be stored in a distributing reservoir upon a water purification process before the water demand begins to increase, for example, at previous midnight.
In this case, water demand greatly varies depending on day, week, month, season, and the like. In addition, since the manner in which floc is formed varies depending on variation factors such as temperature and the pH value of raw water, the floc formation speed and sedimentation speed varies, and the time required for a water purification process also changes.
Conventionally, when a water distribution amount including such variation factors is to be predicted, a water demand pattern is plotted daily, and the amount of water to be prepared is predicted by referring to the pattern. Alternatively, a water distribution amount is predicted by using a Kalman filter, neural network, or memory-based learning.
In such a conventional water distribution amount predicting method, however, a water distribution amount cannot be accurately predicted in consideration of the time required for a water purification process. For example, according to the method of predicting a water distribution amount from graphed water demand patterns, all patterns associated with variation factors and water distribution amounts must be prepared, and an optimal pattern must be accurately selected from these enormous quantities of patterns. In practice, however, a pattern having some similarity to an optimal pattern is selected from a limited number of patterns, and hence only a rough tendency can be predicted. It is therefore inevitable that a large water distribution amount is estimated, resulting in wasting of a coagulant, drainage of purified water, and the like. Consequently, uneconomical operation becomes unavoidable.
In the method of predicting a water distribution amount by using a Kalman filter, a prediction can be made one hour before the water distribution at best but cannot be made in anticipation of the time required for a water purification process, i.e., three to six hours. According to the method of predicting a water distribution amount by using a neural network, although a prediction can be made in anticipation of a certain period of time, an optimal model is difficult to generate by learning. In addition, model updating requires a period of time similar to the time required for a water purification process, and it is difficult to evaluate a model itself and parameters. This method therefore lacks in practicability and cannot obtain sufficient reliability.
In the method of predicting a water distribution amount by using memory-based learning (e.g., Tamada et al., “Water Distribution Amount Prediction by Memory-Based Learning”, 1992 National Convention Record I.E.E. Japan), since variables in different units and ranges are handled on the same level, proper distances must be set between input data and sampled data of the respective variables. Such distances must be determined for each actual water purification facilities by cut and try. As variables become multidimensional, enormous work is required to determine distances. This method therefore lacks in practicability. Furthermore, in order to improve prediction precision, all sampled data must be stored without being aggregated and changed. This requires an enormous storage capacity.
The present invention has been made to solve such problems, and has as its object to provide a water distribution amount prediction system which has high practicability and can predict a future water distribution amount with high reliability in anticipation of the time required for a water purification process.
DISCLOSURE OF INVENTION
In order to achieve the above object, a water distribution amount predicting system according to the present invention includes a case base generating unit which receives many actual data constituted by sets of actual water distribution amounts obtained in advance from water purification equipment and variables required for prediction, stores the data as history data, quantizing an input space of a case base in accordance with a desired output allowable error, and generates a case representing at least one history data by placing each history data in the quantized input space, thereby generating a case base, and a similar case retrieving unit which retrieves a similar case corresponding to a newly input variable from the case base, and estimates a water distribution amount corresponding to the newly input variable from the similar case retrieved by the similar case retrieving unit.
As the variables, time data associated with a time position at which the actual water distribution amount was obtained and environment data associated with an environment where the actual water distribution amount was obtained may be used. In addition, a past water distribution amount associated with the actual water distribution amount may be used.
In addition, the present invention may include a flowmeter which measures an actual amount of water distributed from the water purification equipment, and may update the case base by revising only a predetermined case in the case base using a set of the actual water distribution amount obtained by the flowmeter and a variable corresponding to the water distribution amount.


REFERENCES:
patent: 03202515 (1991-09-01), None
patent: 0830581 (1996-02-01), None

Affiliated with

Also associated with

Rating

Water distribution amount predicting system does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Water distribution amount predicting system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Water distribution amount predicting system will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3269385