Short term load forecasting: two stage modelling

Abstract

This paper studies the hourly electricity load demand in the area covered by a utility situated in the Seattle, USA, called Puget Sound Power and Light Company. Our proposal is put into proof with the famous dataset from this company. We propose a stochastic model which employs ANN (Artificial Neural Networks) to model short-run dynamics and the dependence among adjacent hours. The model proposed treats each hour's load separately as individual single series. This approach avoids modeling the intricate intra-day pattern (load profile) displayed by the load, which varies throughout days of the week and seasons. The forecasting performance of the model is evaluated in similiar mode a TLSAR (Two-Level Seasonal Autoregressive) model proposed by Soares (2003) using the years of 1995 and 1996 as the holdout sample. Moreover, we conclude that non linearity is present in some series of these data. The model results are analyzed. The experiment shows that our tool can be used to produce load forecasting in tropical climate places.

Keywords

• Neural networks
• nonlinear models
• short-term load forecasting
• statistical model building.