Part Load Strategy Definition and Annual Simulation for Small Size sCO2 Based Pulverized Coal Power Plant

Abstract

In the near future, due to the growing share of variable renewable energy (VRE) in the electricity mix and the lack of large-scale electricity storage, coal power plants will have to gradually shift their role from base-load operation to providing fluctuating back-up power to meet unpredictable and short-noticed load variations, in order to improve the stability of the electrical grid. However, current coal power plants have been designed to operate at base-load and are not optimized for flexible part-load operation, resulting in an intrinsic inadequacy for fast load variations. The founding idea of the H2020 sCO2-Flex project is to improve the flexibility of pulverized coal power plants by adopting supercritical carbon dioxide (sCO2) Brayton power cycles instead of common steam Rankine cycles. Despite the extensive available literature about the design of sCO2 power systems for different applications (concentrating solar power, waste heat recovery, 4th generation nuclear), there is still limited knowledge about part-load strategies that should be implemented in order to maximize system efficiency during real plant operation. This paper aims to provide a deeper insight about the potential of sCO2 power plants for small modular coal power plants (25–100 MWel) highlighting the difficulties that must be faced during part-load operation in order to ensure high system performances still guaranteeing a safe operation of the cycle components. The selected configuration is a recompressed cycle with high temperature recuperator bypass which is modelled in a MATLAB+REFPROP numerical tool specifically developed to optimize the plant nominal performance, to provide a preliminary sizing of each component and to evaluate and compare different off-design operating strategies. In particular, the off-design behavior and the operational constraints of each component will be implemented based on referenced numerical models, adopting reliable correlations and exploiting ad hoc codes for the performance evaluation.