Optimization of ventilation system design and operation in office environment

With growing concern about the impact of indoor environment quality on office workers' well-being and productivity, coupled with the concern over the rising energy costs for space heating and cooling in office building sector, ventilation principles that integrate flexible and responsive elements have grown in popularity in office buildings. Such advanced elements as Underfloor air distribution (UFAD), passive swirl diffusers, and demand control on ventilation rate pose challenges to system design and operation.; This thesis is concerned with the development and implementation of a practical and robust optimization scheme, with the goal of aiding the office building designers and operators to enhance the thermal comfort and indoor air quality (IAQ without sacrificing energy costs of ventilation. The path taken is a simulation-based optimization approach by using computational fluid dynamics (CFD) techniques in conjunction with genetic algorithm (GA), with the integration of an artificial neural network (ANN) for response surface approximation (RSA) and for speeding up fitness evaluations inside the GA loop. It breaks the problem into three sequential steps.; First, the performance of various ventilation systems was predicted and evaluated by CFD simulations for an assumed set of indoor/outdoor environmental conditions. By varying the external temperature, the internal heat load, the geometric configuration in the office, the supply air states, and the placement of air terminals in the CFD model and examining the consequent effects, the influential parameters significantly affecting the objectives of interest can be identified and examined. Though CFD is often quoted as a method of acquiring detail and accuracy, the excessive computational costs retards the direct conflation of it into the optimization underway. It is then a worthy effort establishing a low fidelity model for RSA, which can be then used in the place of CFD to evaluate fitness during optimization search. In the second step, an ANN model was trained and tested for this purpose by using data obtained from pre-conducted CFD simulations. When created properly, such a model significantly decreases the computing time for optimization objectives and constraints calculation without compromising accuracy. Finally, a GA was applied in the third step to search for the near-optimal combinations of the controlled variables, using the pre-trained ANN model for fitness evaluations inside searching loops. The objective function is formulated in a way attempting to integrate and weight indicators such as predicted mean vote (PMV for thermal comfort assessment), ventilation effectiveness (epsilon v for IAQ evaluation), and energy usage by space cooling and a supply fan into one performance index.; The CFD simulations in this study are pre-validated using experimental data from baseline cases with both UFAD system and ceiling mounted mixing system (MS). Good agreements between the measured and the predicted air velocity/temperature profiles provide the justification for the current choice of turbulence model and the present specification of boundary conditions. It can be observed that the ANN model obtained and used cuts down the execution time from 17 hours per CFD simulation (thus, per fitness evaluation in GA originally) to a time scale of a few minutes for the whole GA search (invoking approximate 5000 fitness evaluations in total). Within a particular office space with a given indoor pollutant emission rate and internal/external thermal conditions, the final optimization solution contains a set of near optimal ventilation system design/operation parameters, including the ventilation system type, diffuser type, number of diffusers, supply air temperature, amount of supply air, as well as the location of diffusers and return grilles, which can substantially enhance the thermal comfort level and IAQ with saving in the energy costs simultaneously. Such optimization results in...