Abstract
Load forecasting, which is the act of anticipating future loads, has been shown to be important in power system network planning, operations and maintenance. Artificial Intelligence (AI) techniques have been shown to be good tools for load forecasting. Load forecasting can assist power distribution utilities maximise their revenue through optimising maintenance planning. With the dawn of the smart grid, first world countries have moved past the customer’s point of supply and use smart meters to forecast customer loads. These recent studies also utilise recent state of the art AI techniques such as deep learning techniques. Weather parameters are such as temperature, humidity and rainfall are usually used as parameters in these studies. South African load forecasting studies are outdated and recent studies are limited. Most of these studies are from 2010, and dating backwards to 1999. Hence they do not use recent state of the art AI techniques. The studies do not focus at distribution level load forecasting for optimal maintenance planning. The impact of adjusting power consumption data when there are spikes and dips in the data was not investigated in all these South African studies. These studies did not investigate the impact of weather parameters on different South African loads and hence load forecasting performance.
D.Ing. (Electrical and Electronic Engineering)