Page - 112 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 3442 Ϭ ϭϬϬϬϬϬ ϮϬϬϬϬϬ ϯϬϬϬϬϬ ϰϬϬϬϬϬ ϱϬϬϬϬϬ ϲϬϬϬϬϬ ϳϬϬϬϬϬ ϴϬϬϬϬϬ ϵϬϬϬϬϬ &ŽƌĞĐĂƐƚ ϯD &ŽƌĞĐĂƐƚ ϰD &ŽƌĞĐĂƐƚ ϱD dŽƚĂů ůĞĐƚƌŝĐŝƚLJ ŽŶƐƵŵĞĚ Figure9.Chart formovingaveragemethod(1971–2014). Movingaverage isonemethodwhich isverymuchsuitable forshort-termloadforecasting,STLF. Theforecastof the fourthyear is theaverageof thefirst threeyearsandsoon. 3.6.WeightedMovingAverage The three years’ and four years’weightedmoving average for the time period 1971–2015 is calculatedhere. Thevalueswere foundtobe786,587.1and765,421.5MWrespectively for theyear 2015 inFigure10. Ϭ ϭϬϬϬϬϬ ϮϬϬϬϬϬ ϯϬϬϬϬϬ ϰϬϬϬϬϬ ϱϬϬϬϬϬ ϲϬϬϬϬϬ ϳϬϬϬϬϬ ϴϬϬϬϬϬ ϵϬϬϬϬϬ &ŽƌĞĐĂƐƚ ϯD &ŽƌĞĐĂƐƚ ϰD &ŽƌĞĐĂƐƚ ϱD dŽƚĂů ůĞĐƚƌŝĐŝƚLJ ŽŶƐƵŵĞĚ Figure10.Chart forWeightedmovingaveragemethod(1971–2015). For the threeyearsweightedmovingaveragetheαvaluefor theprevious threeyearswere0.5, 0.3and0.2respectively. Thehigherαvalue isallottedto the immediatemonthsince it influences the outcomemore thanthatof thepreviousvalues. For the fouryearsweightedmovingaverage theα value for theprevious fouryearsassignedwere0.4,0.3,0.2and0.1respectively. It ismadesure that theαvaluesaddupto1. 112