Future estimation for electricity interruption in Dohuk governorate (Kurdistan-Iraq)‎ using SARIMA model

Abstract EN

Seasonality, in time series, refers to a regular pattern of changes that repeat for S time period, where S defines the numbers of timer periods until the pattern repeats again.

Seasonality, of course, usually causes the time series to be non stationary [4], seasonal differencing is defined as a difference between a value and a value with lag that is a multiple of S.

In this paper, we will discuss the daily average of hours of electricity interruption per month which represents the gap between the amount of energy available and energy required for consumption.

We took the data of Dohuk governorate (within Kurdistan region of Iraq) for the period from Jan.

2010 to Dec.

2016.

Since this series have seasonal affects, we used SARIMA model with S = 12 and the seasonal and non-seasonal differences are 0 or 1, to choose the appropriate model, that gives least value of BIC, RMSE, MAE, MAPE and largest value of R2 to forecast the periods of daily interruption for the next months.

We concluded that the best model is SARIMA(1,1,1)(0,1,0)12, and the expected values of the daily average of hours of electricity interruption per month in Dohuk Governorate are increasing and it is expected that the entirely lack of electricity supply during the month of December 2018 if the time series continues this pattern.

Key words: Time series, SARIMA model, BIC, RMSE, MAE, MAPE, R2, ACF, PACF, Box-Ljung statistics, electricity interruption