A Comparison of Synthetic Daily Streamflow Models Applied to Flood Control Planning

Código

I-EBHE0055

Autores

FELIPE TREISTMAN, DÉBORA DIAS JARDIM PENNA, Angela de Oliveira Ghirardi, Daniela de Souza Kyrillos, IGOR PINHEIRO RAUPP, Priscilla Dafne Shu Chan, FERNANDA DA SERRA COSTA, Jorge Machado Damazio

Tema

WG 1.10: Hydrologic Design - Solutions & Communication

Resumo

Accurate streamflow scenario generation is critical for optimizing flood control strategies and ensuring efficient reservoir management in hydropower operations. This study aims to present a comparative analysis of two synthetic daily streamflow models applied in flood control planning for hydropower plants. The performance of these models was evaluated using a comprehensive set of statistical tests against historical streamflow records, and their efficacy was further demonstrated by calculating of reservoir space requirements for flood control in a small Brazilian system. The first model, DIANA, is used by the National Operator of the Electrical System (ONS) in the Annual Flood Control Planning for the Interconnected National System. It is a stochastic multivariate time series model, that considers daily streamflow as a sum of two components, the first one representing the external factors (precipitation) and the second representing the persistent emptying outflow of the watershed. The second model, GHCen - Hybrid Generator of Synthetic Streamflow Scenarios, combines stochastic multivariate time series modeling of daily rainfall with a conceptual rainfall-runoff model. It employs a stochastic approach based on the Hilbert-Huang transform to generate synthetic multisite daily precipitation scenarios, which are latter applied to a rainfall-runoff model called SMAP/ONS. To assess both models, several statistical tests were applied to evaluate the reproduction of key characteristics of the historical streamflow record, with a focus on flood-related statistics. Additionally, the application of both models in generating reservoir volume requirements for flood control was also analyzed and compared with the distribution of required volumes obtained from the historical record. Results from the streamflow statistical tests showed that both models have a great capacity to mimic the key characteristics of historical record and the probabilistic behavior of flood occurrence. GHCen was able to reproduce slightly better daily streamflow high-order moments such as skewness, kurtosis, and the autocorrelation function for lags greater than one day. In terms of mean and standard deviation of annual maximum required daily flood control volume, both models have similar good results. In the assessment regarding the flood control operation rule curve the form of the curves obtained by the two models are very similar, but the volumes of the curve obtained by GHCen were higher than the DIANA?s curve in the middle of the flood season and lower during the remainder of the wet period. Further investigations to clarify this difference are to be carried out. In conclusion, through statistical testing and practical application in reservoir flood control planning, GHCen demonstrated similar performance compared to DIANA, indicating its potential to be used as another option in flood control planning of hydropower plant operation. Other exercises should be done in other Brazilian watershed which have hydropower plants with flood control constraints.