To sharply reduce the ventilation power quantity by determining the operation quantity of a ventilation quantity based on the ventilation state index value estimated with a neural network.
The VI value, CO value, traffic quantity, the notch number of a ventilation facility, and environmental index value are fed to the first neural network 10 as the input and learning data. Networks 11, 12, 13 receive the environmental index values at certain points in the past based on the accumulated data in the past fixed period, the future VI values from these points are estimated, VI actual values are given from the output side for learning, and the internal weight is adjusted. The VI estimated values are inputted to a notch change judging means 20, the set upper and lower limit values are compared with VI10, VI15, VI20, and whether the change of the notch number of a ventilation facility is required or not is judged. The notch change is implemented by a notch selecting means 60 which is an action quantity selecting means and a notch determining means 70.
SASAMOTO TOSHIJI
KAWAKAMI KAZUMI
KARASAWA TAKASHI
YAMAMOTO MASANORI
FUJI ELECTRIC CO LTD