**Título:** Forecasting Manufacturing sector electricity consumption in South Africa: A Probabilistic Causality approach using bayesian neural networks

**Autores:** Marwala, Lufuno; Twala, Bhekisipho

**Resumo:** The South African manufacturing industry is a significant consumer of electricity. Energy consumption in this sector increases with increased production levels and decreases with decreased production levels. It can be asserted, therefore that there is a probabilistic causal relationship between manufacturing production level and electricity demand. The aim of this work is to develop a mathematical model for this causal relationship to assist in forecasting the future energy demand in the manufacturing sector. Neural networks are used to build the causal model. It is assumed that in the causal relationship between these two variables the cause occurs before the effect. In line with this assumption, lagged values of the production index are used to build a neural networks model which is used for assessing its effect on the demand. The results show the causality neural network models created can be used to predict electricity demand with accuracy.

**Palavras-chave:** Causality; Manufacturing production index; Electrictity demand; neural networks; markov chain monte carlo

**Páginas:** 6

**Código DOI:** 10.21528/CBIC2013-269

**Artigo em pdf:** bricsccicbic2013_submission_269.pdf

**Arquivo BibTex:** bricsccicbic2013_submission_269.bib