| 摘要 |
PURPOSE:To provide an electric power demand prediction device with high learning efficiency. CONSTITUTION:An output unit is divided into 14 units so that event data in accordance with every output unit stored in a data base 2 can be classified to 14(7X2) partial sets and correspond to such classification based on the day of the week (Sunday, Monday, Tuesday, Wednesday, Thursday, Friday, and Saturday) and seasons (October to March, April to September) of each event. In this way, it follows that a neural network 51a can be constituted of 14 modules of a neural net. Also, learning to be applied to the corresponding module out of 14 modules of the neural net is performed on each of the 14 partial sets of the data base 2. |
