| Keywords |
External dynamic shading device; Multi-objective optimization; Random Forest; Genetic Algorithm; TOPSIS; MCDM(Multi-Criteria Decision Making) |
| Abstract |
This paper suggests a multi-objective optimization process for a dynamic shading device to minimize cooling load and maintain the required illuminance. We have developed a Random Forest model based on the data from IDA ICE simulation. A non-dominated sorting genetic algorithm (NSGA-II) is applied to optimize the extrude length of the upper 2-axis of the dynamic shading device at each hour. The developed Random Forest model could predict the cooling load and illuminance with appropriate accuracy (CVRMSE of 1.55% and 3.58%, respectively) compared to the ground truth from the IDA ICE simulation. The optimal shape of the shading device is determined among several alternatives using the TOPSIS method that is one of the multi-criteria decision making methods. In this process, two objectives (cooling load and illuminance) could have different weights to reflect the priority or preference of decisionmakers.
As a result, the optimal shape of the dynamic shading device derived from the multi-objective optimization process could provided improved performances both on cooling load and visual comfort. Besides, our process was able to consider different weights of each objective depending on the decision makers’ preferences. |