Volume 8 • Issue 2 • 1000299
J Civil Environ Eng, an open access journal
ISSN: 2165-784X
Nayab and Faisal, J Civil Environ Eng 2018, 8:2
DOI: 10.4172/2165-784X.1000299
Research Article Open Access
Jo
u
r
n
a
l
o
f
C
i
v
i
l
&
E
n
v
iro
n
m
e
n
t
a
l
E
n
g
i
n
e
e
ri
n
g
ISSN: 2165-784X
Journal of Civil & Environmental
Engineering
Water Management in Tarbela Dam By using Bayesian Stochastic
Dynamic Programming in Extreme Inflow Season
Ayesha Nayab* and Muhammad Faisal
Department of Statistics, Quaid I Azam University, Islamabad, Pakistan
Abstract
Existing method of forecasting inlows at Tarbela have some limitations, also system needs an adequate operating
policy model to deal with highly volatile inlow of summer months of June, July, August and September. In this paper,
historical data of inlows from 1986 to 2014 have been used to forecast upcoming inlows at dam. Bayesian predictive
distribution is used to predict future inlows. These forecasted inlows were further incorporated into operating policy
model to determine the optimal release during the prescribed months. Weather volatility is a major factor causing
unstable inlows. High temperature during summer period cause high inlows at dam. Considering weather volatility,
this policy model is proposed for the lood season (15th June to 30th September), in which inlows and outlows are
higher than rest of the year. This model maximizes the expected proit from hydro power production, minimizes the
expected loss from lood damage and updates the proper estimate of current stage of reservoir storage.
*Corresponding author: Ayesha Nayab, Department of Statistics, Quaid I
Azam University, Islamabad, 45320, Pakistan, Tel: +92519064000; E-mail:
ayesha.nayab@mail.au.edu.pk
Received November 07, 2017; Accepted February 02, 2018; Published February
07, 2018
Citation: Nayab A, Faisal M (2018) Water Management in Tarbela Dam By using
Bayesian Stochastic Dynamic Programming in Extreme Inlow Season. J Civil
Environ Eng 8: 299. doi: 10.4172/2165-784X.1000299
Copyright: © 2018 Nayab A, et al. This is an open-access article distributed under
the terms of the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original author and
source are credited.
Keywords: Temperature volatility and inlow forecast; Release policy;
Bayesian stochastic dynamic programming
Introduction
Water inlow is the uncertain element which directly afects the
release policy of the dam. Inlow into the reservoir is usually high in
summer due to snowmelt from high temperature. Also rain in the
upstream catchment area is the most critical lood-producing factor
of the basin. Monsoon caused heavy rain during summer and water
elevation at dam surface oten rose to the peak. To manage available
water to generate maximum hydropower and prevent from lood and
damage in downstream area is a big challenge in summer because of
high volatility of temperature.
Tarbela is a multipurpose reservoir, irstly it provides water for
irrigation and almost 50% of the country irrigation system is run by
Tarbela dam. Secondly, it stores water during lood season to escape
from high damage at downstream areas. hirdly, it ofers generation
of hydropower to satisfy 30% of the energy needs. To achieve these
objectives at their best, system needs a reliable forecast for the inlows
and operating policy model. Table 1 presents the water discharge limits
at diferent levels of loods.
At minimum lood limit release estimate is 7000 m
3
/second and
at maximum this release estimate goes to 23000 m
3
/second. In lood
season, when the water rises beyond the capacity of the dam, we need
to appropriately utilize the substantial amount of available water in
such a way that we can generate maximum hydropower with minimum
damage at downstream area. herefore, we need to account uncertainty
of the inlow in optimal operating policy.
he concept of addressing forecast uncertainty along with inlow
uncertainty was presented by many authors in the past. Bayesian
Stochastic Dynamic Programming model incorporates Bayesian
approach within classical Stochastic Dynamic Programming (SDP)
formulation [1]. herefore we used Inlow data as prior knowledge
and forecasted data as sample information. Posterior distribution was
a conditional distribution of forecast over actual inlow to the reservoir.
Use of Bayesian Decision heory (BDT) and Bayesian inference for
predicting upcoming lows of water along with classical forecasting
methods take into account relatively complex problems of natural
resource management [2]. Bayesian Stochastic Dynamic Programming
(BSDP) is also used as stochastic optimization tool for the development
of operating policy model, which incorporates a Bayesian approach
within the classical Stochastic Dynamic Programming (SDP)
formulation (Table 2) [3].
he use of linear autoregressive (AR) model and linear autoregressive-
moving-average (ARMA) model in reservoir optimization is available
in literature. he multi lag autocorrelation model by a single hydrologic
variable, the value of which changes from day to day and is equal to
the conditional mean of the daily inlow [4]. Algorithm which can take
into account the inlow stochasticity for calculating optimal operating
strategies in a multi-reservoir hydroelectric system does not require
discretization of the state space [5].
Case studies with the Brazilian systems give the comparison of
Flood Stage Discharge (m
3
/second)
Low 7000
Medium 10500
Average 14000
High 18500
Very High 23000
(Source: WAPDA Pakistan, Flood Management Manual 2011)
Table 1: Flood limits at Tarbela.
Notation Description
I
t
. Inlow at time t
nh
t
. Net head available at time t
Ƞ Plant eficiency
C
t
Cost at time t due to excess outlow in downstream area
S
t
Storage at time t
R
t
Rain at time t
E
t
Evaporation at time t
O
t
Outlow at time t
λ Estimated threshold level for damage
Table 2: Notations used in BSDP model with description.