Prediction of energy consumption in buildings with artificial intelligent techniques and Chaos time series analysis

This paper introduces a new approach for theprediction of hourly energy consumption inbuildings. The proposed method uses nonlineartimeseries analysis techniques for the reconstructionof energy consumption timeseries andthe estimation of the dynamic invariants, and artificialneural networks as a nonlinear modelingtool.Among the several neural network modelingfactors that affect time-series prediction, themost important are the window-size and thesampling lags for the data. Relevant theoreticalresults related to the reconstruction of a dynamicalsystem are analyzed and the relationshipbetween a correct embedding dimensionand network performance is investigated.The problem is examined initially for theunivariate case and is extended to include additionalcalendar parameters, in the process of estimatingthe optimum model.Different network topologies are considered,as well as existing approaches for solving multistepahead prediction problems. The predictiveperformance of short-term predictors is also examinedwith regard to prediction horizon.The performance of the predictors is evaluatedusing measured data from real scale buildings,showing promising results for the developmentof accurate prediction tools.