A PREDICTIVE MODEL FOR ELECTRICITY CONSUMPTION IN MODIBBO ADAMA UNIVERSITY OF TECHNOLOGY, YOLA, USING ARTIFICIAL NEURAL NETWORKS (ANN)
A predictive model for monthly electricity energy consumption with good accuracy is still a great challenge in the University. Considering the fact that the University consumes huge amount of electricity for numerous operations (academics, research, social, economic etc.). Energy availability, consumption and costs can present a great challenge of either reduction or increase in energy consumption if not properly conserved and utilized. This research proposed an Artificial Neural Networks model that is used to predict efficiently and effectively the monthly electricity consumption in the University using previous consumption data. Three years (2016-2018) datasets from Works department and Students affairs division of Modibbo Adama University of Technology (MAUTech) Yola was collected. Regression Analysis and ANN was used to project electricity consumption. The model was developed and tested in Matlab (R2016b) with feedforward neural network architectures which was optimized with two algorithms (Levenberg-Marquardt and resilient back-propagation). Sigmoid and Linear activation functions were used in the hidden and outer layers respectively. The model has two input neurons and three output neurons. The number of hidden layer was obtained via an experiment with 10 neurons at the hidden layer. This reveals that the regression which is 0.9337 is the highest out of the thirty experiments carried out. The study shows that ANN can sufficiently predict future electricity consumption which in a long run will assist the University community in effective planning and utilization towards power consumption. The study recommended the use of different predicting tools which can also be used and compared for further studies.
Keywords: Electricity consumption; artificial neural network; regression analysis.