Utilizing of seasonal time series forms in predicting the quantity of electric power demand in Erbil city between the year (2010-2019)

الملخص

In this study, the researchers took the phenomenon of the quantities of demand for electrical energy represented by a time series of monthly rates of the phenomenon for the city of (Erbil - Iraq) for the period (2010 - 2019) from (General Directorate of Electricity of the city of Erbil - the control department) and the series consisted of (111) values, Where the researchers noted when analyzing the chain, there is an increasing general trend of the phenomenon, and this indicates an increased demand for electrical energy by city residents with a few concavities and protrusions of the string values, and this indicates a case of non-stationary counting around the average and the variance, respectively, and it was also observed the presence of Seasonal compound, in other words, there are changes that repeat themselves every regular period (12 months) of the phenomenon, in other words that the need or demand increases its intensity for the mentioned period, and for the purpose of converting the chain to a state of stability around the average, the first difference was taken to remove the chain from the influence of the general trend either to convert the chain to be stable On the variance, the natural logarithm transformation of the time series was taken after the first difference was taken. Either to remove the compound or seasonal changes, the first seasonal difference was taken for the seasonal period amounting to 12 months. For the purpose of future forecasting of the phenomenon, the Box-Jenkins methodology was applied to the stable chain. For this purpose, the researchers selected the seasonal and non-seasonal ranks of the mentioned methodological models (p, q, P, Q = 0,1,2,3,4). To reach the best seasonal model, double appropriate, with data from among the significant models for the purpose of predicting the future where the three criteria were used for the purpose of diagnosing the significant models, and they are as follows: (Mean Square Error (MSE)), Akaike information criterion AIC, The standard (Akaike information criterion AICc) used in the research, and we concluded that the SARIMA (0.1,1) (1,1,2) 12 time series represented the best representation because it had the lowest values ​​for the above criteria with a fit to pass a test. The stationary of the models, meaning that the errors of the personalized model were random, and after testing the stationary of the proposed model, we predicted the quantities of demand for electrical energy for a period of 12 months (2019-2020) in order to benefit from them in the future planning process.