Abstract
In this study, in order to establish a method to decide a learning period efficiently for precision hourly heat and power load prediction by neural network. First, the author suggested the index of the "Accordance Ratio" showing the ratio that explanation variable accorded training data and predicted the prediction precision became highest at the learning period when the index became highest. Second, the effectiveness of the index was inspected by carrying out the prediction at the actual office building in several learning periods and analysis the relation among the index, prediction precision and learning period.
