Electricity consumption prediction for buildings using multiple adaptive network-based fuzzy inference system models and gray relational analysis

The rise in environmental awareness has increased the significance of controlling and monitoring electricity consumption. The efficiency of power management is directly affected by the accuracy of predicting electricity consumption. It is easy to estimate the electricity consumption if the electricity status is predicted. Therefore, this study proposes a method to predict the electricity consumption of public buildings by using an adaptive network-based fuzzy inference systems (ANFISs) and weather conditions. ANFIS combines the interpretability of fuzzy inference systems and the learning ability of neural networks. Gray relational analysis (GRA) is used to analyze the relationship between weather conditions and electricity consumption. In this study, a multi-ANFISs approach is introduced to estimate the electricity consumption by weather conditions and human activities. An alarm system was also developed using the estimation errors. The results show that the proposed multi-ANFISs achieves a greater performance with less number of parameters, and the GRA can evaluate the magnitude of relation between the factors and a specific output.