Abstract :
Electric power should be provided in amount or magnitude to meet the
requirement and also in the right time. Excess of the requirement electric power
may cause loss. On the contrary, lacking electric power supply, will cause blacking
out. Thus, to provide adequate electric power that meet the requirement, there
should be an electric power?s plan performed by making a prediction or electric
load forecasting. Therefore, matter of electric load forecasting become much
important in efficient electric power supply.
To predict electric load needs, PLN currently using load coefficient method.
Such computing method is based on empirical experience of electric power?s
planning division which relatively harder to complete especially in several
correction needed for change of load. Therefore, a better method is still needed
than load coefficient method.
In this research, the author attempted to build a prediction model for
shortterm electric load by using artificial neural network (ANN) with
backpropagation learning algorithm and sigmoid activation function. Research
data collection scope was limited by electric load in work region of Yogyakarta.
The result showed that the prediction of ANN electric load on January 1,
2018 to December 31, 2020 shows that the average load of 277 A increases by 334A
per day and the average percentage of ANN error is 19.6%.