Evaluation of Loss Models Predictions in One-Dimensional Analysis of Centrifugal Compressors

Compressed Air Energy Storage (CAES) systems are emerging as a promising large-scale energy storage technology, where the efficiency of the compression process has a direct impact on the overall cycle performance. Centrifugal compressors represent a critical component, and the availability of reliable prediction tools is essential to support their design, optimization, and integration into such systems. While three-dimensional CFD simulations can provide highly accurate results, they are computationally intensive and not always suitable for preliminary design phases, where faster evaluation methods are required. This thesis focuses on the study of the performance of a one-dimensional model for centrifugal compressor performance prediction, implemented in MATLAB. The model computes key parameters such as pressure ratio and isentropic efficiency directly from the compressor geometry and boundary conditions. A central feature of the work is the implementation of multiple sets of loss correlations from the literature, each describing different mechanisms including incidence, skin friction, clearance, slip, mixing, and diffuser-related effects. A detailed discussion of these correlations is provided, emphasizing their assumptions, validity ranges, and impact on the predicted performance. To assess the predictive capability of the model, the code was applied to centrifugal compressor geometries and datasets reported in the literature, which include experimental measurements of pressure ratio and efficiency. The comparison revealed that the model is able to reproduce performance trends, while also highlighting the discrepancies associated with different sets of correlations. In particular, certain models provided good agreement in terms of pressure ratio prediction, whereas others were more effective in estimating efficiency trends. These results underscore the importance of selecting appropriate loss models depending on the specific performance feature of interest. Overall, the thesis provides a critical review of existing loss modeling approaches. Within the broader framework of CAES applications, the study demonstrates how one-dimensional modelling can effectively balance computational efficiency and predictive accuracy, making it suitable for preliminary design and comparative studies. The modular structure of the code also opens the way for future extensions, including the integration of new correlations and its application to alternative compressor configurations or other turbomachinery components.