摘要 |
PURPOSE:To increase the working efficiency of distributed water, by learning the performance data and estimating the daily distributing water volume at every season and the hourly distributing water volume in a day. CONSTITUTION:Weather observation data in every season are extracted by a data- processor 21 and daily distributing water volume in every season and characteristic figures of varying patterns of the hourly distributing water volume are extracted by a data-processor 22. The daily distributing water volume and the characteristic figures of varying patterns of the hourly distributing water volume are estimated by use of these distributing water volume data as instruction signals by means of a learning means 23 learning significance coefficients by the back propagation through a neural network inputting weather observation data and informations classified into week days, holydays, etc. Next, the weather data of the day and the informations such as week day, holyday, or others are input to select an estimating model. The daily distributing water volume and the varying pattern of hourly distributing water volume are estimated, basing on the selected model. These two estimated figures are multiplied to estimate the daily distributing water volume of the day. In this way, the water distributing plan through a day can be submitted and we can also cope with the variation for a long or middle term.
|