As energy concerns continue to grow, the need to create more efficient building systems using accurate modeling techniques has increased. Most modern buildings are equipped with electric power meters that record electric power data that can be used for model accuracy improvements. This paper discuses typical data-based building energy models and proposes new improvements by using data classifications. Six different data-based models for estimating subhourly and hourly electric energy consumption are presented and discussed. These models are three typical single to multiple regression models, two proposed regression models, and one artificial neural network (ANN) model with recommended classifications. Power data collected from existing buildings at 15 min intervals are used to build and test the models. Additional hourly energy data obtained from a well-known energy simulation program are also used for detailed analysis. The results show that the proposed regression models and ANN model with recommended data classifications can provide very accurate results compared to traditional modeling techniques. Significant improvements in statistic index R-squared values are a result of using the proposed regression and ANN models for all tested buildings.
| File Size : | 1
file
, 5.6 MB |
| Note : | This product is unavailable in Russia, Belarus |
| Number of Pages : | 9 |
| Product Code(s) : | D-HO-18-007 |
| Published : | 2018 |
| Units of Measure : | Dual |
