Application of Artificial Neuron Networks and Hurst Exponent to Forecasting of Successive Values of a Time Series Representing Environmental Measurements in an Intelligent Building

Abstract

Control systems for intelligent buildings based on environmental measurements. The information contained in the measurement data in many cases are independent and chaotic. You can not interact with the control system, but only to monitor. In many cases, the use of measurement data for the purpose of building automation control, requires the use of forecasting systems. For the needs forecasting this type of measurement data apply artificial neural networks. Learning provides a mechanism to adjust its internal parameters of artificial neural network to characterize the trend of the time series reflects the measurement data. For time series with greater variability of smoothing is necessary. The intention initial classification time series smoothing and allows the use of artificial neural networks to forecast the next value of the time series irrespective of their volatility.