A Comparative Evaluation of Conventional and Computational Intelligence Techniques for Forecasting Electricity Outage

• Published in: 2021 Southern African Universities Power Engineering Conference/Robotics and Mechatronics/Pattern Recognition Association of South Africa (SAUPEC/RobMech/PRASA) pp. 1 - 6
• Main Authors: Onaolapo, A. K., Pillay-Carpanen, R., Dorrell, D. G., Ojo, E. E.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 27.01.2021
• Subjects: Analytical models; artificial neural network; Artificial neural networks; back propagation algorithms; Climate change; Computational modeling; exponential smoothing; Forecasting; Meteorology; outage forecasting; Power system reliability; Predictive models; weather events
• DOI: 10.1109/SAUPEC/RobMech/PRASA52254.2021.9377243

Reliability of the electric grid structure for the transmission and distribution of power from the generating plants to the consumers, is an essential requirement for the reliability of electric supply. The components of the grid is exposed to weather events which cause faults to the grid's structures. There is an expectation that as climate change alters the severity and number of weather events, electricity supplies and electricity grid reliability are expected to be affected. Computational intelligence models have been proven to be excellent predictive models for electricity reliability problems in previous studies. But there has not been sufficient literature reporting their applications in weather related electricity outage forecasting problems. In this study, weather related electricity outage was forecasted using artificial neural networks (ANNs) with back-propagation algorithm. Real-life data sets of the city of Pietermaritzburg, South Africa, was used to investigate the performance of the ANN model and was compared with a conventional model - exponential smoothing (ES). The ANN model gave satisfactory results as compared to the ES model. The result is a demonstration of the robustness of computation techniques.