PERFORMANCE OF BACKPROPAGATION NEURAL NETWORKS FOR CULTIVATION DAILY LOADS IN JAW A CENTRAL-DIY

Abstract

Goals   of  this  research   are  implementing    Artificial   Neural   Network (ANN)  algorithm for  load forecasting   and getting  its performance.   The training  data  was takenfrom     UPB  Ungaran.  The performance   can be got through  comparing  ANN test result with the real load at that time. The   research    methodology    usc   experimental     and   design    models approach.    The     phases    of  this   research    were:    I.   analyzing    and identifying    of  need    2.  developing    of  load  forecasting     application software  with  C programming.   3. entering  and  training  the data to get data pattern.
The  result  of  this  research.   the  load forecasting  result  by ANN  was close with UPB loadforecasting.    but several  ANN  test result have more deviation  than  UPB. because  number  of training  data  was  less. so the forecasting     pattern    111as not   too   accurate     Beside    that.   another possibility   was  the number  of iteration  must  be more  than  / ()()(J  times
iterations   in  order  to get  more  less  error.   There  was  33,3% of ANN result  that  has  more  less  deviation,   although   the  number   of  training data  was  not  different,  because  that  data  has  no  extrem variation,  so
the pattern   was faster   to be recognized.   Generally,   ANN  will  give  an accurate  pattern  recognation    if the data is valid and the number  of the data is quite enough.