Back to results list
Please use this identifier to cite or link to this item:
|Title:||Developing a new passive design approach for green building assessment scheme with integrated statistical analysis and optimization||Authors:||Chen, Xi||Advisors:||YANG, Hongxing||Keywords:||Sustainable buildings -- China -- Hong Kong -- Design and construction
Environmental impact analysis -- China -- Hong Kong
Buildings -- Energy conservation
Dwellings -- Environmental aspects
|Issue Date:||2017||Publisher:||The Hong Kong Polytechnic University||Abstract:||This thesis aims to develop a novel passive design assessment system for green building rating tools with an example application to the building environmental assessment method (BEAM) in Hong Kong. Current passive design approach is criticized for arguable criteria allocation, unjustified weighting system and incompatibility with the traditional whole building energy simulation method. Therefore, a modelling based statistical analysis and optimization framework is proposed to improve the integration of passive designs for green building assessment in this thesis, which can address the synergy of energy and indoor environmental requirements under both free-running and artificially controlled conditions. This holistic design approach can account for the interactive effect between different passive strategies and enable decision-makers to understand the relative importance of each strategy and deploy them appropriately at the first opportunity for achieving the optimum building performance in future. The developed new passive design evaluation system can achieve consistent rating with the traditional assessment approach as proved by a successful application in a green building case study. The building orientation, external obstruction angle, external wall thermal resistance, wall specific heat, window transmittance, window U-value, window to ground ratio, overhang projection ratio and infiltration air mass flowrate coefficient were determined as representative input design factors highly-concerned in early design stages based on a comprehensive literature review of passive strategies related to the building layout, envelop thermophysics, building geometry and infiltration & air-tightness. After determining input variables and their distribution functions, a generic building model was developed and on-site measurement methods were specified to obtain the ventilation, daylight, thermal comfort and corresponding energy use indicators. The influence of selected passive design parameters on multiple indoor environmental indicators was initially quantified by standard regression analysis. After the verification of independence between input parameters, control variables including the running periods, internal loads, ventilation control methods and weather conditions were modulated to observe their impacts on both the coefficient of determination and sensitivity indices. Two regression model indices were found to vary greatly with different control variable settings when the humidity ratio, PMVSET and TSENS were taken as building performance outputs for the cooling period. Despite the observed variation, the window transmittance and geometry features were constantly the most influential design factors over other indoor environmental indices. In addition, the ASHRAE adaptive model with 90% acceptability was proved to be the most suitable comfort assessment model for assessing naturally ventilated buildings during the whole cooling seasonin hot and humid climates.
On top of standard regression analysis, the required minimum sampling size, effectiveness of rank transformations, uncertainty of sensitivity indices and prediction accuracy of different meta-models were evaluated by the statistical modeling and comparative analysis. A large sample size over 100 per regression coefficient was recommended to obtain stable statistical estimations of the prediction error and goodness of fit. The rank transformation managed to calibrate sensitivity coefficients by up to 6%, which even altered the rankings between certain design inputs. Uncertainties of sensitivity indices were validated to be within 2% with adequate bootstrap repetitions. Prediction accuracy of the ventilation and comfort model were also improved to an acceptable level with the non-parametric regression analysis. Succeeding to above statistical analyses, on-site measurements and simulations were conducted on an existing Public Rental Housing (PRH) development to verify the compliance with green building requirements and explore the performance improving potential with different ventilation strategies. An unoccupied naturally ventilated flat was found to fulfill the acceptability limits in the ASHRAE adaptive model for 67.5% of the time in July, while the required minimum ventilation rate of 2.0 ACH was achieved. More than 95.6% of the habitable area, subject to different obstruction levels and window to floor ratios, complied with the related daylight assessment criteria. In addition, full-day ventilation proved to be the best control strategy for low thermal mass buildings, while the simulated indoor conditions were partly verified by on-site measurement results. The energy saving potential with combined natural ventilation and daylight strategies was also anticipated to be 51.9% for air-conditioning and 8.3% for lighting compared to a baseline building. Based on the above modelling and experimental studies, a holistic design optimization process was proposed by integrating robust sensitivity analysis into multi-objective optimizations. Global sensitivity indices and corresponding significance statistics were first calculated to prune the optimization problem space. The NSGA-II based optimization was then conducted to obtain the Pareto frontier, which presented a trade-off between comfort and daylight objectives when the minimum ventilation rate was fulfilled. After a further post optimization analysis and comparison of decision-making methods, the final optimum solution for the passive design achieved a 11.2% reduction of the total unmet time. In addition, different population sizes, crossover probabilities and mutation rates were examined to find the most suitable setting of NSGA-II by balancing between the computation efficiency and optimization productivity. This simulation-based optimization process was further applied to four other cities in hot and humid areas. Finally, a new passive design assessment system as an equivalent alternative to the traditional whole building energy simulation was established with the application of the holistic design approach. Extensive global sensitivity analysis with regression, variance-based and screening-based methods were performed on the prototype building to constantly readjust the criteria coverage based on feedbacks from statistical and optimization analysis. With the five design variables finalized for the new assessment framework, upper and lower performance scales were derived from baseline simulations and NSGA-II based optimizations. The grading coefficient was further determined by a local sensitivity analysis for a pro-rata credit awarding scheme. Weighting systems transformed from different sensitivity indices were validated by modelling experiments, where FAST first-order indices outstood by accurately predicting 73.3% of test cases. The new system has also been successfully applied to a registered green building project for obtaining consistence with the traditional approach. Such a systematic approach detailed in this study can also be exploited to develop alternative approachesfor all performance-based criteria ina green building rating scheme.
|Description:||xxxviii, 272 pages : color illustrations
PolyU Library Call No.: [THS] LG51 .H577P BSE 2017 Chen
|Award:||FCE Awards for Outstanding PhD Theses||URI:||http://hdl.handle.net/10397/79318||Rights:||All rights reserved.|
|Appears in Collections:||Outstanding Work by Students|
Show full item record
Files in This Item:
|991021962222603411_link.htm||For PolyU Users||208 B||HTML||View/Open|
|991021962222603411.pdf||For All Users (Non-printable)||11.13 MB||Adobe PDF||View/Open|
Citations as of Nov 12, 2018
Citations as of Nov 12, 2018
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.