Prediction of Weather-Related Electric Power Interruptions on the 33 kV Bonsa Feeder Using Artificial Neural Networks

Abstract

The objective of this study is to establish the relationship between weather parameters such as rainfall, temperature, wind speed and relative humidity and power interruption of a 33 kV feeder and to develop an Artificial Neural Network (ANN) model for the prediction of these weather-related power interruptions. Four years data spanning 2013 to 2016 on the weather parameters for the geographical area and number of recorded outages on the feeder were taken. These data were used to develop the prediction model. The data were used to train, validate and test the performance of the network and that of 2016 were used to predict the number of outages. The Levenberg Marquardt algorithm was used to train the network. Different models were developed to predict the occurrences of the outages based on a total of nine scenarios. This was also done to investigate which parameters had the most influence on the outage events. The weather data for 2016 were used as new inputs (sample) to the networks, and all the networks were simulated to predict the number of outages. The results showed that the ANN model was able to predict the number of outages with a reasonable level of accuracy. Rainfall and wind speed were established as the critical causes of the outage events while temperature and humidity had minimal influence on the outage events.