Application of fuzzy neural network on the electricity consumption forecasting

The electricity consumption is highly related to the consumption's pattern. Thus, analyzing the electricity consumption's pattern becomes an important issue in order to reduce the total electricity consumption. This paper aims to analyze the electricity consumption in Taipei Bus Station located in Taipei City, Taiwan. This building is one of the busiest bus station in Taiwan. The analysis of the electricity consumption in this building is conducted based on the historical data of total electricity consumed by some cooling equipment installed in the building and the total electricity consumptions. In addition, the building temperatures, humidity and CO2 are also considered. In order to obtain an accurate forecasting, some data preprocessing approaches are conducted. They include the missing value prediction, data normalization, feature selection and discretization. Furthermore, a fuzzy neural network is applied to obtain the forecasting model. The experiments results show that the proposed research framework can obtain an accurate forecasting model with very small error. The result also reveals that the total electricity consumption is only highly related to some of the features studied in this paper while other factors do not significantly influence the total electricity consumption.