BN-97-16-3 -- Operational Data as the Basis for Neural Network Prediction of Hourly Electrical Demand PDF

Describes the development of a neural network model for predicting electricity use by large hotel in San Francisco. The network was trained with hourly occupation data, and the corresponding electricity consumption was recorded over a year. The project 1) identified input and output parameters for the energy model, 2) identified candidate variables and surrogates for predicting energy use, 3) Defined the structure of a neural network capable of predicting electricity use, 4) trained neural network models and 5) tested the model for accuracy in both recursive and non-recursive modes. Finds that for the hotel, which is isolated from the environment, the operational data alone were sufficient to predict electrical demand with accuracy adequate for assessing HVAC system retrofits.

Units: I-P

KEYWORDS: year 1997, Hotels, electricity consumption, calculating, computer programs, expert systems, USA, case studies, research