Efficient forecast of daily demand of the interconnected electrical system of Peru through ARIMA stochastic analysis with external events

Abstract

The daily electric demand in Peruvian National Interconnected System-SEIN- has very particular trend, seasonality and characteristics external effects, a situation that complicates the process of estimating the short-term forecast. The aim of this paper is to formulate and calculate ARIMA models with External Events Analysis to achieve efficient forecasts of electricity demand each day, at total level and broken down by areas of the SEIN. The methodology is based on treating each time series using appropriate statistical-mathematical transformations to achieve stability in variance as regular seasonal averages, parallel external events to try to reach an optimal predictive model ARIMA each area of the electrical system of Peru (Central, South and North) and for each day of the week. The results demonstrate the predictive efficiency. Taking as a quality indicator forecast the Mean Absolute Percent Error (MAPE), have obtained values lower than 1% by the projections of the total daily demand SEIN versus 2% obtained with existing deterministic techniques.