BUILDING ENERGY USE PREDICTION AND SYSTEM-IDENTIFICATION USING RECURRENT NEURAL NETWORKS

Following several successful applications of feedforward neural networ ks (NNs) to the building energy prediction problem (Wang and Kreider, 1992; JCEM, 1992, 1993; Curtiss et al., 1993, 1994; Anstett and kreide r, 1993; Kreider and Haberl, 1994) a more difficult problem has been a ddressed recently: namely, the prediction of building energy consumpti on well into the future without knowledge of immediately past energy c onsumption. This paper will report results on a recent study of six mo nths of hourly data recorded at the Zachry Engineering Center (ZEC) in College Station TX. Also reported are results on finding the R and C values for buildings from networks trained on building data.