Keywords

ANFIS GP, ANFIS SC, energy consumption prediction, leisure center

Start Date

26-6-2018 10:40 AM

End Date

26-6-2018 12:00 PM

Abstract

Leisure centers are growing in popularity as health and physical fitness awareness is becoming an integral part of human society. Leisure centers consume more energy compared to most office buildings but are less studied in the area of non-residential energy consumption prediction. This work presents an energy consumption prediction effort for a leisure center using a class of two ANFIS based adaptive networks: ANFIS GP and ANFIS SC and multi-linear regression. Climatic, periodicity and energy use data collected over a period of eight months were pre-processed, normalized and split into training and testing sets before being presented to the adaptive networks for neuro-fuzzy inferencing. The results were compared to those of the multi-linear regression models and showed that adaptive networks were superior in performance and there was only a small difference between the two ANFIS algorithms. The combination which gave the best results comprised temperature, the hour of the day and relative humidity with MAE, RMSE and R2 values of 0.69 kWh, 0.70 kWh and 0.73 respectively, represented by ANFIS SC (0.5) model. This is a good predictive method offering an opportunity for better attainment of efficient energy management.

Stream and Session

B2

Share

COinS
 
Jun 26th, 10:40 AM Jun 26th, 12:00 PM

Energy Consumption Prediction for Recreation Facilities using hybrid Neuro-Fuzzy Inference Systems

Leisure centers are growing in popularity as health and physical fitness awareness is becoming an integral part of human society. Leisure centers consume more energy compared to most office buildings but are less studied in the area of non-residential energy consumption prediction. This work presents an energy consumption prediction effort for a leisure center using a class of two ANFIS based adaptive networks: ANFIS GP and ANFIS SC and multi-linear regression. Climatic, periodicity and energy use data collected over a period of eight months were pre-processed, normalized and split into training and testing sets before being presented to the adaptive networks for neuro-fuzzy inferencing. The results were compared to those of the multi-linear regression models and showed that adaptive networks were superior in performance and there was only a small difference between the two ANFIS algorithms. The combination which gave the best results comprised temperature, the hour of the day and relative humidity with MAE, RMSE and R2 values of 0.69 kWh, 0.70 kWh and 0.73 respectively, represented by ANFIS SC (0.5) model. This is a good predictive method offering an opportunity for better attainment of efficient energy management.