Title:
【発明の名称】需要予測方法
Document Type and Number:
Japanese Patent JP3039943
Kind Code:
B2
Abstract:
PURPOSE:To obviate the need of analysis of a plenty of data or interview with operators by a method wherein a neural network model learns a plurality of different demand achievements and disturbance variables, in the past, of a power system. CONSTITUTION:A neural network model 71 is configured hierarchically of an input layer 710 and at least one intermediate layer 720 and output layer 730. The neural network model 71 is made to learn the history of disturbance factor and demand pattern. At this time, distribution of the weight factor of the neural network model 71 varies. Furthermore, a plurality of demand achievements corresponding to the inputs are provided, as teacher signals, to the neural network model 71 to be leant thereby. Thus learnt neural network model 71 is then provided with a plurality of predicted values of yet-learnt disturbance variables and demand achievements in order to predict corresponding demands. By such arrangement, an operator can simulate the operation according to an example without requiring analysis of a plenty of data or interview with operators.
Inventors:
Shigeru Tamura
Hiroyuki Kudo
Kenichi Morita
Hiroyuki Kudo
Kenichi Morita
Application Number:
JP2829790A
Publication Date:
May 08, 2000
Filing Date:
February 09, 1990
Export Citation:
Assignee:
株式会社日立製作所
International Classes:
H02J3/00; G05B13/02; G06F15/18; G06N3/00; (IPC1-7): H02J3/00; G05B13/02
Domestic Patent References:
| JP60102822A | ||||
| JP60106325A | ||||
| JP6173519A | ||||
| JP1233579U |
Attorney, Agent or Firm:
Yasuo Sakuta