Applying fuzzy logic and neural networks to forecasting in efficiency programs

Abstract

This paper addresses the design and implementation of Fuzzy Inference System (FIS) and Artificial Neural Network (ANN) to determine daily demand curves of residential Electric Showers (ESs). To determine the daily curves were used two inputs: shower duration and number of showers. In Brazil the residential electricity corresponds to 25% of all demand. The use of ESs is widespread, it represents about 22% of the total residential consumption. This work evaluates the impacts of Energy Efficiency Programs (EEPs) in low-income communities located in the state of Rio de Janeiro in Brazil. Additionally, two different ESs devices are compared: the ES Temperature Control (ESTC) and the ES Heat Recovery (ESHR). This study was based on measurements made in 60 households in different low-income neighbourhoods. The results showed that ANN makes better predictions, however both FIS and ANN have the capacity to determine rapid changes in peak demand. These tools can be used in small and mediumsized areas with similar socio-economic features which allow determining the impact of EEPs in the communities in advance. Furthermore, the application of these techniques can be of help in the actions of Demand Side Management (DSM) mainly during the maximum demand period.