Evaluation of Regression Analysis Based Building Hourly Thermal Load Prediction Algorithms Under Climate Change

Sarwar, R., & Cho, H. (2014). Evaluation of Regression Analysis Based Building Hourly Thermal Load Prediction Algorithms Under Climate Change. The ASME 2014 International Mechanical Engineering Congress & Exposition. Montreal, Quebec, Canada: The American Society of Mechanical Engineers.

Abstract

This study presents performance of two ARX (auto-regressive with exogenous, i.e., external, inputs) building hourly thermal load prediction models under influence of climate change. ARX building thermal load prediction models are of particular interest not only for their higher prediction accuracy but also for computational efficiency. Precise thermal load prediction is a significant aspect of energy planning for mixed energy distribution systems in regional and national level as well as in intelligent building energy management and control. However, in order to ensure reliability, prediction models should be able to account for influence from climate change in the upcoming years. Performance and robustness of the models over five consecutive years have been evaluated. A case study with medium office reference building in Atlanta, GA has been carried out to demonstrate prediction accuracy of both models over the period utilizing a widely accepted building energy simulation software. Results have been evaluated using statistical criteria to quantify the effect of climate change on prediction accuracy.