For Better Performance Please Use Chrome or Firefox Web Browser

Mohammad Kiani Felavarjani

Grade: 
Master

Hybrid Modeling of Sequential Reservoir Operation using Large Scale Climate Signals

 

Development and capacity expansion of water infrastructures, such as reservoirs, is part of challenges authorities face in meeting the regional water needs.   Another part of this challenge, and the essential one, is the management and operation of water resources systems. After a serious depletion of alternative water resources such as groundwater,   the most important sources of water   in most countries (including Iran) is the water which is supplied from   dam reservoirs. Operational management of these water reservoirs, as a complex system, is very critical and needs careful attention in order to guarantee the sustainability of water supply in severe conditions.

In this research, a hybrid method has been presentedwhich enables us to introduce alternative optimized scenario for future planning and also to have a real-time time operation ofreservoir systems.Combination of several soft computational methods has been employed to demonstrate the competency of these techniques. HarmonySearch algorithm (HS) and SupportVector Machines (SVM) are the two main approaches for optimization of reservoir releases and forecasting of the inflow to reservoir and ultimately determining a real-time strategy for operation of a reservoir. HS algorithm is one of the new methods of optimization which is based on its successful and remarkable performance in recent years, compare to other meta-heuristic optimization techniques. SVM method based on structural risk minimization principlehas been introduced as a powerful method for learning processes and forecasting phenomenon. In addition, the mutual information (MI) has been used to select the efficient and reliable parameters within immense amount of large-scale regional signals to forecast the annual inflows. Regional signals can be considered as parameters and indicators when their changes in terms of time in different stations show the changes in rainfall and temperature which will happen in another time or another place.

The proposed methodology is applied to Zayandehrud reservoir and the application explored the model’s accuracy and its potential for real time prediction of reservoir monthly releases for adaptive planning and management.Besides the runoff from the Zayandeh-rud basin, diverted flow from neighboring basins contribute to the reservoir inflows. This reservoir is the main water structure to reduce the shortages and help to balance the monthly water needs based on limited supply in the region. Historical data indicate that the region is susceptible to hydrological drought periods where consecutive years of below-average annual streamflow occur. Fifty six years (1954-2009) of monthly inflow data were used for this study.