Short Term Prediction of Electric Load Demand of University of Benin Using Artificial Neural Network

Abstract

Electricity load demand prediction is an integral part of Power
System Management. The importance cannot be over-emphasized in
Electricity Generation, Transmission, Distribution and Marketing.
Accurate electric power load forecasting is essential to the
operation, expansion and planning of a utility company. In this work
short term load demand prediction of University of Benin, Ugbowo
Campus, was carried out for a period of one week using 30 days data
between 1st to 30th September, 2019. The forecasting technique used
in actualizing this task is the artificial neural network (ANN) which
was modeled using the MATLAB R2013a toolbox. The actual load
demand predicted value are presented in graphical form. It was
observed that the ANN architecture 5-35-1 gave the optimal
performance of 0.0021% Average Mean Absolute Percentage Error
(MAPE). Considering the effect of time delay vector, the time delay
vector of [01] also gave the optimal performance of 0.0021% MAPE
with a time delay of 123 seconds and two length of vectors which
validates the correctness of the simulation using MAPE.
Consideration of the ratio parameter indicator in the model
prediction, it was observed that the ratio parameter of 0.01 also gave
an optimal performance with 0.0021% MAPE. Model adequacy is
achieved since the same optimal result was achieved in all the
various constraints applied.