Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Assessing reservoir reliability using classical and long-memory statistics
Considering that most water supply reservoirs were designed and are operated with limited streamflow records, and that persistence (long-memory) is often present in reservoir inflows, it not surprising that reservoirs designed and operated with 100% reliability have failed, due to unaccounted uncertainties. The Descoberto reservoir (Central Brazil) failed to supply the water demand in 2017, although mass balance analyses indicated a failure-free condition during its previous 31 years of operation. The objective of the present study was to identify the presence of long-memory processes in the Descoberto reservoir inflows and to reassess its reliability with alternative methods and statistics. After persistently low inflows were detected in the years preceding reservoir failure, autocorrelation and rescaled-range analyses were performed in the inflow series, confirming the presence of long-term process-LTP (Hurst coefficient = 0.655). Reservoir reliability was then reassessed for the pre-failure period (1986–2016), using three storage-yield-reliability (S-Y-R) methods and LTP statistics. The reassessed pre-failure reservoir reliability was 96.05%, very close to the observed (post failure) reliability (96.87%), estimated by mass balance analysis. The 3% difference observed between the reliabilities using classical and LTP statistics resulted from a higher standard deviation in the reservoir inflows, capturing the long-term variability not evident in the classical analysis. If persistence occurs in inflows, reservoir reliability could be assessed using a similar approach, improving reservoir behavior forecast and management. Keywords: Reliability, Reservoir, Inflows, Hurst phenomenon
Assessing reservoir reliability using classical and long-memory statistics
Considering that most water supply reservoirs were designed and are operated with limited streamflow records, and that persistence (long-memory) is often present in reservoir inflows, it not surprising that reservoirs designed and operated with 100% reliability have failed, due to unaccounted uncertainties. The Descoberto reservoir (Central Brazil) failed to supply the water demand in 2017, although mass balance analyses indicated a failure-free condition during its previous 31 years of operation. The objective of the present study was to identify the presence of long-memory processes in the Descoberto reservoir inflows and to reassess its reliability with alternative methods and statistics. After persistently low inflows were detected in the years preceding reservoir failure, autocorrelation and rescaled-range analyses were performed in the inflow series, confirming the presence of long-term process-LTP (Hurst coefficient = 0.655). Reservoir reliability was then reassessed for the pre-failure period (1986–2016), using three storage-yield-reliability (S-Y-R) methods and LTP statistics. The reassessed pre-failure reservoir reliability was 96.05%, very close to the observed (post failure) reliability (96.87%), estimated by mass balance analysis. The 3% difference observed between the reliabilities using classical and LTP statistics resulted from a higher standard deviation in the reservoir inflows, capturing the long-term variability not evident in the classical analysis. If persistence occurs in inflows, reservoir reliability could be assessed using a similar approach, improving reservoir behavior forecast and management. Keywords: Reliability, Reservoir, Inflows, Hurst phenomenon
Assessing reservoir reliability using classical and long-memory statistics
Henrique M.L. Chaves (Autor:in) / Douglas R. Lorena (Autor:in)
2019
Aufsatz (Zeitschrift)
Elektronische Ressource
Unbekannt
Metadata by DOAJ is licensed under CC BY-SA 1.0
Assessing the Reliability of Urban Reservoir Supplies
| Wiley | 1991
Use of statistics in reservoir operations
| Engineering Index Backfile | 1956
Reservoir Water Quality: Assessing Salinity
Online Contents | 1995
Reservoir capacity yield reliability analysis
| British Library Conference Proceedings | 2001
The Use of Statistics in Reservoir Operations
| ASCE | 2021