NONLINEAR DYNAMICAL SYSTEMS APPROACH TOBUILDING ENERGY PREDICTION PROBLEMS

摘要

Given a time series data, model dynamical sys-rntems are built using a hierarchical Bayesianrnscheme with feedforward neural nets and then thernmodels are compared in terms of marginal likeli-rnhood. The model with the highest marginal like-rnlihood is used for predictions. The algorithm isrnapplied to building air-conditioning load predic-rntion.