Are Optimized Design Solutions Robust A Case in Point - Industrial Halls with Varying Process Loads and Occupancy Patterns
Through building performance simulation and optimization, this paper will identify the most optimal combinations of values of demand side parameters that will minimize the total energy consumption of ventilation, heating, and lighting for a typical industrial hall.
Optimized design solutions for industrial halls projected for a particular process load and occupancy pattern might not perform as predicted due to anticipated but unascertainable changes. To take into account such possible changes, uncertainty analysis can be performed to determine if the optimized design solutions are in fact robust to such changes and to identify solutions that are less susceptible to uncertainty. The case study includes a hypothetical building, which represents a typical industrial hall in Amsterdam, the Netherlands. The building is built with steel cladding on a steel frame. Workers are assumed to perform light work. For an industrial hall kind of environment, current guideline (ARAB 2006) recommends that the temperature of the space has to be maintained under 30°C to protect workers from heat stress and heating has to be provided only if the space drops below 18°C during occupied hours.
The building energy performance simulation program TRNSYS is used to perform the energy analysis for cooling and heating demands. Energy demand by the hour is evaluated and aggregated for the year. DAYSIM is used to evaluate the illuminance level on the work surface at each of the hour due to daylighting. Based on the illuminance level, lighting energy consumption is then calculated by a proprietary program written in MATLAB according to the dimmable lighting characteristics. Optimization is deployed to search for the optimized design solutions that consume the least amount of total energy for cooling, heating, and lighting. Out of the many available algorithms in modeFRONTIER, MOGA is chosen as the optimization algorithm. Though it is commonly deployed for multi-objective optimization, its efficiency in searching for global optimum makes it a good candidate, even though the case study is a single objective optimization that minimizes the total energy consumption. An initial search space of 40 configurations is generated with Latin Hypercube Sampling. The energy consumption of design solution case 3 is less than 0.03% different from the most optimized design solution. And the worst variation of having two-shift operation consumes only 13.4 Wh/m2h. Even though, solution 3 is not the most optimized solution, it is in fact the more robust one.