Boiler feed pump turbines (BFPTs) are in use at a number of Eskom power stations. They utilise bled steam extracted from the main turbine in order to drive multistage centrifugal pumps which supply the boilers with feedwater. With an increase of renewables in the energy mix, the need for Eskom’s coal-fired power stations to run for extended periods at very low loads has arguably never been this great. Various systems affect the ability of these generation units to run economically at low loads. One such system is the boiler feed pump turbine and its associated pumps. A station was selected from Eskom’s fleet based on access to information and the station being a relatively typical plant. The Unit (a boiler and turbogenerator set) selected for study was one with the most thorough instrumentation available for remote monitoring. The BFPT system of this Unit was modelled in Flownex, a one-dimensional thermofluid process modelling package. The model included individual pump stages, steam admission valves and a stage-by-stage turbine model utilising custom stage components. These turbine stage components represent each stage with nozzles and other standard Flownex components. The boundary conditions of the system were set as functions of generator load in order to represent typical values for use in case studies. The relationships between load and boundary conditions were based on large samples of data from the station’s data capture system (DCS). A corresponding standby electric feed pump system was also modelled in Flownex for a comparative case study. After model validation, a number of case studies were performed, demonstrating the functionality of the model and also providing specific results of value to the station in question. These results include the minimum generator load possible with different steam supplies; maximum condenser back pressure before plant availability is affected; the viability of changing the pump leak-off philosophy; and the effect of electric feed pump use on power consumption. The main recommendations from the case studies were as follows: i. to stroke the steam admission valves as per the design charts, ii. to test the operation of the BFPT down to 40 % generator load, iii. to keep the pump leak-off philosophy unchanged, iv. to maintain the cooling water system and condensers sufficiently to avoid poor condenser vacuum, v. to reconsider the decommissioning of the “cold reheat” steam supply, vi. and, to favour use of the BFPT over the electric feed pumps at all generator loads.

A previously developed turbine modelling methodology, requiring minimal blade passage information, produced a customizable turbine stage component. This stage-by-stage turbine nozzlemodel component was derived from the synthesis of classical turbine theory and classical nozzle theory enabling the component to accurately model a turbine stage. Utilizing Flownex, a thermohydraulic network solver, the turbine stage component can be expanded to accurately model any arrangement and category of turbine. This project focused on incorporating turbine blade passage geometrical information, as it relates to the turbine specific loss coefficients, into the turbine stage component to allow for the development of turbine models capable of predicting turbine performance for various structural changes, anomalies and operating conditions. The development of turbine loss coefficient algorithms as they relate to specific blade geometry data clusters required the investigation of several turbine loss calculation methodologies. A stage-by-stage turbine nozzle-model incorporating turbine loss coefficient algorithms was developed and validated against real turbine test cases obtained from literature. Several turbine models were developed using the loss coefficient governed turbine stage component illustrating its array of capabilities. The incorporation of the turbine loss coefficient algorithms clearly illustrates the correlation between turbine performance deviations and changes in specific blade geometry data clusters.