Physical Rules Based Adaptive Neuro-Fuzzy Inferential Sensor Model Design and Analysis in Predicting the Indoor Temperature in Heating System

摘要

The previous research on adaptive neuro-fuzzy inferential systems (ANFIS) presented an approach to estimating the average indoor temperature in the building environment. However, the restriction on robustness limited the energy efficiency and indoor comfort ratio. An accurate and robust prediction model is proposed in this paper. Comparing to the previous unphysical rules based ANFIS prediction model, the improvement of the physical rules based ANFIS prediction model will be presented and the reason of better performance of this new model will be discussed. Three performance measures (RMSE, RMS, and R2) are using in evaluating the proposed prediction model.