Surrogate based optimisation of a pump mode startup sequence for a contra-rotating pump-turbine using a genetic algorithm and computational fluid dynamics
Jonathan Fahlbeck, Håkan Nilsson, Saeed Salehi
Abstract
Contra-rotating pump-turbines (CRPT) have been proposed as a new solution for pumped hydro storage (PHS) at low-head sites. The two individually operated runners give both flexibility and complexity during transient operation, such as startup. This paper presents an optimisation approach for the pump mode startup sequence of a CRPT. The aim is to minimise the loads during fast startups and thus extend the lifetime and increase the flexibility of the CRPT. The startup sequence is decomposed into five design variables which determine a valve opening and the speedup of each runner rotational speed. The optimisation procedure includes transient computational fluid dynamics (CFD) simulations and global surrogate-based optimisation (SBO). An objective function, based on an integration of the time derivative of the axial force, is derived and transformed into a Gaussian-process surrogate model. The results show that the SBO can predict an optimal combination of the five variables that limits unfavourable load variations during the pump mode startup sequence. For an optimal startup sequence, the valve should be opened up during 73% of the sequence, the upstream runner should speed up during most of the sequence, and the downstream runner should speed up during the final third of the sequence. The outcome of this work is beneficial for a safe startup operation of possible future PHS with CRPTs.