Abstract:
Electric load forecast is very vital for the reliable operation of power systems. In order to
overcome the limitations of conventional statistical forecast algorithms, this work proposes a
Neuro-Arima model for the forecast of electric load in semi-built up areas. The design carried
out entailed the integration of an Autoregressive Integrated Moving Average (ARIMA) model
with an Artificial Neural Network (ANN) model for the forecast of electric load. The digital
model of the Abakpa Enugu Nigeria distribution substation (the case study distribution station)
was created and the Neuro – Arima model was implemented using MATLAB/SIMULINK
software. The Neuro – Arima model was trained and validated using historic load time series
data obtained from the case study distribution substation. Results from simulation carried out
showed high accuracy using the model to make hourly, three – hours ahead and weekly load
forecasts. The high forecast accuracy is indicated by the high values of coefficients of multiple
determinations of 0.9304, 0.9146 and 0.9192 for hourly, three – hours ahead and weekly load
forecasts respectively. Comparative evaluation carried out showed that the proposed Neuro –
Arima model achieved 60%, 24.83% and 39.40% improvements over the conventional ARIMA
forecast model for hourly, three – hours ahead and weekly load forecasts respectively.
