Operators in the Syndicat des Eaux de l'Ile-de-France (SEDIF) have been using consumption forecasting tools or, to be more precise, tools to forecast distributed volume, to resolve the complex problem of optimally distributing the volume both geographically and in time. This paper defines the problem: how to forecast consumption and then to operate the production and distribution facilities to satisfy demand. It then describes forecasting tool requirements for the complete network, for production plants and their networks, and for lift-pump stations. Three forecasting systems are then described: (1) the water movement center global daily consumption forecasting system, based on a statistical study that resulted in a linear regression statistical model; (2) the sub-network local management hourly forecast system, which profiles the hourly consumption throughout a day; and (3) neural network forecasting systems, which involves simulating, on a computer, networks of neural cells that operate in a similar way to the neurons in the human brain. Neural networks--how they work, their advantages, models produced, and a comparison with statistical models--are the focus of the article.
| File Size : | 1
file |
| ISBN(s) : | 0898676347 |
| Note : | This product is unavailable in Ukraine, Russia, Belarus |
| Number of Pages : | 12 |
| Published : | 01/01/1992 |
