Nowadays, the ever-increasing world electricity generation by renewable energy sources has brought about changes in conventional power plants, especially in those ones where large steam turbines work, which were widely used to meet the world’s energy needs by operating mostly at fixed conditions. Now, instead, they have to be capable to operate with greater flexibility, including rapid load changes and quick starts as well, in order to make the most of the renewable resources while guaranteeing the coverage of any shortcomings of the latter with traditional fossil fuel systems. Such service conditions are particularly challenging for the exhaust hoods, which have a great influence on the overall turbine performance, especially at off-design conditions. In fact, the complex and high rotational 3D flow generated within the diffuser and the exhaust hood outer casing can cause an increase in aerodynamic losses along with the detriment of the hood recovery performance. For these reasons, an optimized design and adequate prediction of the exhaust hood performance under all the machine operating conditions is mandatory. Since it has been widely proven that the exhaust hood flow strongly interacts with the turbine rear stage, the necessity to model this as well into a CFD modeling becomes crucial, requiring a remarkable computational effort, especially for full transient simulations. Even if adopting simplified approaches to model the last stage and exhaust hood interfaces, such as the so-called Frozen Rotor and the Mixing Plane ones, helps to keep the computational cost low, it can be not for an exhaust hood optimization process, which requires a significant number of CFD simulations to identify the most performing geometry configuration. For these reasons, a simplified model of the exhaust hood must be adopted to analyse all the possible design variants within a feasible time.
The purpose of this work is to present a strategy for the exhaust hood design based on the definition of a simplified CFD model. A parametric model has been developed as a function of key geometrical parameters of both the exhaust hood and the diffuser, taking into account the strong fluid-dynamic coupling between these components. A periodic approximation has been introduced to model the exhaust hood domain, thus allowing to augment the number of the geometrical parameters of the DOE, while keeping the computational effort low. A response surface has been achieved as a function of the key geometrical parameters, therefore an optimization method has allowed identifying the best performing configuration. A 3D model of the optimized periodic geometry has been then generated to assess the effectiveness of the procedure here presented. Finally, the presented procedure has been applied in several off-design operating conditions, in order to find out an optimal geometry for each operating point, evaluating how much they differ from that one got for the design point.