Open AccessELECTRICITY CONSUMPTION PREDICTION SYSTEM USING A RADIAL BASIS FUNCTION NEURAL NETWORK
Author(s) -
Opeyemi Lateef Usman,
Olusegun Folorunso,
Olanrewaju Alaba
Publication year - 2017
Publication title -
journal of natural sciences, engineering and technology/journal of natural science, engineering and technology
Language(s) - English
Resource type - Journals
eISSN - 2315-7461
pISSN - 2277-0593
DOI - 10.51406/jnset.v15i1.1660
Subject(s) - backpropagation , artificial neural network , electricity , computer science , consistency (knowledge bases) , test data , function (biology) , radial basis function , mains electricity , radial basis function network , bridge (graph theory) , data mining , energy consumption , consumption (sociology) , artificial intelligence , machine learning , engineering , voltage , evolutionary biology , electrical engineering , biology , programming language , medicine , social science , sociology
The observed poor quality of service being experienced in the power sector of Nigeria economy has been traced to non-availability of adequate model that can handle the inconsistencies associated with traditional statistical models for predicting consumers’ electricity need, so as to bridge the gap between the demand and supply of the energy. This research presents Electricity Consumption Prediction System (ECPS) based on the principle of radial basis function neural network to predict the country’s electricity consumption using the historical data sourced from Central Bank of Nigeria (CBN) annual statistical bulletin. The entire datasets used in the study were divided into train, validation and test sets in the ratio of 13:3:4. By the above, 65% of the entire data were used for the training, 15% for validation and 20% for testing. The train data was presented to the constructed models to approximate the function that maps the input patterns to some known target values. The models were also used to simulate both validation and the test datasets as case data on the consistency of results obtained from the training session through the train data. Experimental results showed that RBF network model performs better than equivalent Backpropagation (BP) network models that were compared with it and provides the best platform for developing a forecast system.