1997 年 62 巻 502 号 p. 43-50
A Recent researches show the thermal load prediction using Neural Networking (NNW) models are more accurate than other methods. However.those results are conducted by simulation under specific building conditions. Network architecture such as input parameters and a number of hidden layer units influences the prediction accuracy, yet they have so far been decided by the trial-and-error method. In this paper, we examined several methods proposed for constructing NNW models theoretically and evaluated effect of heat load characteristics on the accuracy of heat load prediction using NNW models under the experimental analysis.