Forecasting electricity demand in South Africa using artificial intelligence

D.Phil. (Electrical and Electronic Engineering)This thesis introduces a novel artificial intelligence technique called extreme learning machines (ELM) and structural causal models (SCM) for forecasting electricity consumption using time series and causality approaches. Time series data is used to construct univariate models for forecasting a one step ahead electricity consumption on a monthly basis. For causal analysis, the study is novel in that it mathematically models the relationship between electricity consumption and production levels in the manufacturing sector and mining sector in South Africa..