Modeling and Forecast of Time Series by using a Harmonic Decomposition Approach

This paper proposes a harmonic decomposition approach for modeling and forecasting of time series, with the novelty that a state-space representation of the harmonic content of a non-stationary signal is established from a basic Fourier series analysis. Based on the time series state-space model, wh...

Full description

Saved in:
Bibliographic Details
Published in2019 IEEE International Autumn Meeting on Power, Electronics and Computing (ROPEC) pp. 1 - 6
Main Authors Garibo-Morante, A. Agustin, Ornelas-Tellez, Fernando
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.11.2019
Subjects
Analytical models
Forecasting
Fourier series
Harmonic analysis
Mathematical model
Predictive models
Time series analysis
Online AccessGet full text

Cover

More Information
Summary:This paper proposes a harmonic decomposition approach for modeling and forecasting of time series, with the novelty that a state-space representation of the harmonic content of a non-stationary signal is established from a basic Fourier series analysis. Based on the time series state-space model, which becomes a linear time-varying system, an optimal estimator (Kalman-Bucy filter) is designed to estimate the harmonic content of the signal (time series). The analyzed time series are temperature and wind speed. Once the harmonic model is stated for the time series, the forecast is obtained. Forecasting errors are quantitatively analyzed to evaluate the effectiveness of the proposed methodology, which additionally is compared with respect to classic ARIMA models.
ISSN:2573-0770
DOI:10.1109/ROPEC48299.2019.9057107