Un modelo dinámico bayesiano para pronóstico de energía diaria

  • Marisol Valencia Cárdenas Universidad Nacional de Colombia, Sede Medellín. Medellín, Colombia.
  • Juan Carlos Correa Morales Universidad Nacional de Colombia, Sede Medellín.
Palabras clave: Modelos dinámicos bayesianos, Monte Carlo por Cadenas de Markov, pronósticos, Colombia.

Resumen

Los modelos dinámicos bayesianos son una alternativa útil para elaborar pronósticos con pocos datos históricos, o que ayudan a complementar la poca información que se tenga. En este trabajo se propone el diseño de un algoritmo para realizar pronósticos usando un modelo dinámico bayesiano basado en Filtro de Kalman. Se ilustra el procedimiento aplicándolo al pronóstico de demanda de energía diaria de Colombia, lo cual puede ser útil en sistemas que presenten fallas, o en regiones donde apenas inicia el abastecimiento energético. La eficiencia se determina con el indicador de error absoluto medio (MAPE) de ajuste y de pronóstico; este último resulta menor del 3%, valor adecuado para mostrar validez del método propuesto.

Dynamic Bayesian Models are a useful alternative to make forecasts with few historical data, or that permits to complement the little information you have.  In this paper we propose an algorithm design to forecast using a dynamic Bayesian model based on Kalman Filter. We illustrate the procedure by applying it to the prediction of daily energy demand of Colombia, which can be useful in systems experiencing failures, or in regions where energy supply is just beginning. The efficiency is determined by the Average Absolute Error Indicator (MAPE), of adjustment and forecast; the last one is less than 3%, appropriate value to show validity of the proposed method.

Citas

ANDERSSON, Michael., and KARLSON, Sune. Bayesian forecast combination for VAR models. Sveriges Risk banc-working paper series. [en línea]. 2007, p. 1–17. [Consultado en agosto, 2013]. Disponible en internet: http://www.riksbank.se/Upload/Dokument_riksbank/Kat_publicerat/WorkingPapers/2007/wp216.pdf..

ANDRIEU, Christophe., et al. An Introduction to MCMC for machine learning. Machine learning [en línea]. 2003, vol. 50, p. 5–43. [Consultado en Octubre, 2013]. Disponible en internet: http://link.springer.com/article/10.1023/A:1020281327116.

AT&T. The R Project for Statistical Computing [en línea]. Institute for Statistics and Mathematics. 2013. [Consultado en agosto, 2013]. Disponible en internet: http://www.r-project.org/foundation/.

BARRERA, Carlos Javier., and CORREA, Juan Carlos. Distribución predictiva bayesiana para modelos de pruebas de vida vía MCMC. Revista Colombiana de Estadística, 2008, vol. 31, no 2, p. 145-155.

BOLSTAD, William. Harrison-Stevens Forecasting and the Multiprocess Dynamic Linear Model. The American Statistician, 1986, vol. 40, no. 2, p. 129–135.

CRAIG, Peter, et al. Bayesian forecasting for complex systems using computer simulators. Journal of the American Statistical Association, Jun 2001, vol. 96, no 454, p.717-729.

DE ALBA, Enrique., and MENDOZA, Manuel. Bayesian forecasting methods for short time series. The International Journal of Applied Forecasting. International Institute of Forecasters, Diciembre, 2007, p. 1–10.

DUNCAN, George, GORR, Wilpen., and SZCZYPULA, Janusz. Bayesian Unrelated Time Forecasting Series: for Seemingly to Local Forecasting Application Government Revenue. Management science, 1993, vol. 39, no 3, p. 275–293.

GEWEKE, John., and WHITEMAN, Charles. Handbook of economicForecast [en ligne]. Graham Ell. [Consultado en enero, 2014]. ISBN 978-0-444-62732-2. Disponible en : http://www.sciencedirect.com/science/handbooks/15740706.

GEYER, Charles., and THOMPSON, Elizabeth. Annealing Markov Chain Monte Carlo with applications to ancestral inference. Journal of the American Statistical Association [En línea]. 1995, vol. 90, no. 431, p. 909–920. [Consultado en octubre, 2013]. Disponible en internet: http://amstat.tandfonline.com/doi/full/10.1080/01621459.1995.10476590.

GILL, Jeff. Bayesian Methods-A social and Behavioral Sciences Approach, 2002. ISBN 978-1584885627.

HARRISON, Jeff., and STEVENS, C. Bayesian Forecasting. Journal of the Royal Statistical Society, 1976. vol. 38, no. 3, p. 205–247.

HARRISON, Jeff., and WEST, Mike. Dynamic Linear Model Diagnostics. Biometrika, 1991, vol. 78, no. 4, p. 797–808.

HERR, Henry., and KRZYSZTOFOWICZ, Roman. Bayesian ensemble forecast of river stages and ensemble size requirements. Journal of Hydrology, June 2010, vol. 387, no. 3-4, p. 151-164. ISSN 0022-1694. Disponible en: http://dx.doi.org/10.1016/j.jhydrol.2010.02.024.

MANDEL, Jan. A Brief Tutorial on the Ensemble Kalman Filter. [en línea]. 2009. n° January, p. 1–7. [Consultado en Agosto, 2013]. Disponible en internet: http://arxiv.org/pdf/0901.3725.pdf.

MARTIN, Andrew, QUINN, Kevin., and PARK, Jong Hee. MCMCpack: MarkovChain Monte Carlo. R. Journal of Statistical Software [en línea], 2011, vol. 42, no. 9, p. 1–21. [Consultado en diciembre, 2013]. Disponible en internet: https://www.law.berkeley.edu/files/jstatsoftMCMCpack.pdf.

MEDINA, S., and GARCÍA, J. Predicción de demanda de energía en Colombia mediante un sistema de inferencia difuso neuronal. Energética, Julio 2005, no. 33. p. 15–24.

MEINHOLD, Richard J., and SINGPURWALLA, Nozer D. Understanding the Kalman Filter. The American Statistician, 1983, vol. 37, no. 2, p. 123–127.

MOL, Christine De., GIANNONE, Domenico., and REICHLIN, Lucrezia. Forecasting using a large number of predictors: Is Bayesian shrinkage a valid alternative to principal components?,Journal of Econometrics [en línea], 2008, vol. 146, no. 2, p. 318–328. [Consultado en enero, 2014].
Disponible en internet: http://www.sciencedirect.com/science/article/pii/S0304407608001103.

NEELAMEGHAM, Ramya., and CHINTAGUNTA, Pradeep. A Bayesian model to forecast new product performance in domestic and international markets. Marketing Science [en línea],1999, vol. 18, no. 2, p. 115–136. [Consultado en Diciembre, 2013]. Disponible en internet : http://bear.warrington.ufl.edu/centers/mks/articles/684541.pdf.

PEDROZA, Claudia. A Bayesian forecasting model: predicting U.S. male mortality. Biostatistics, October 2006, vol. 7, no. 4, p. 530–550.

PEZZULLI, Sergio., et al. The seasonal forecast of electricity demand: a simple Bayesian model with climatological weather generator. [en línea], Octubre 2004, p. 1–16. [Consultado en agosto, 2013]. Disponible en internet: http://empslocal.ex.ac.uk/people/staff/dbs202/publications/2005/pezzullib.pdf.

RUEDA, Viviana, VELÁSQUEZ, Juan David., and FRANCO, Carlos. Avances recientes en la predicción de la demanda de electricidad usando modelos no lineales. Revista DYNA, 2011, vol. 78, no. 167, p. 36–43.

WANG, Shuchun. Exponential Smoothing for Forecasting and Bayesian Validation of Computer Models. A Thesis presented In Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy. Industrial and Systems Engineering. Georgia Institute of Technology, 2006.

WEINBERG, Jonathan., et al. Bayesian Forecasting of an In homogeneous Poisson Process with Applications to Call Center Data. [en línea], Junio 2006, p. 1–27. [Consultado en enero, 2014]. Disponible en: http://www-stat.wharton.upenn.edu/~lbrown/Papers/2007a%20Bayesian%20forecasting%20of%20an%20inhomogeneous%20Poisson%20process,%20with%20applications%20to%20call%20center%20data.%20(with%20J.%20Stroud%20and%20J.%20Weinberg).pdf.
Publicado
2013-07-31
Sección
Artículos

Agencias de apoyo