Control Engineering Practice 12 (2004) 377–385
Improvements to the water management ofa run-of-river HPP
reservoir: methodology and case study
Dejan Paravan
a,
*, Tomaz Stokelj
b
, Robert Golob
a
a
Faculty of Electrical Engineering, University of Ljubljana, Trzaska 25, 1000 Ljubljana, Slovenia
b
HSE, Cesta v Mestni Log 88A, 1000 Ljubljana, Slovenia
Received 25 January 2001; received in revised form 31 January 2003; accepted 14 April 2003
Abstract
The operational planning for run-of-river hydroelectric power plants with small storage capacities is directly linked to the natural
inflow ofthe reservoir water. An ability to accurately forecast the natural inflow can result in the increased production ofelectric
energy as a result ofan enhanced flexibility in the management ofthe stored water. In this article a novel methodology for building
an efficient short-term water-inflow forecaster using neural network techniques is presented. The emphasis is given to a precise
definition ofthe problem and the tailoring ofthe problem-solving requirements to specific case conditions. The proposed
methodology was applied to the River SoWa cascade hydroelectric system. The efficiency of the algorithm was evaluated in terms of
the reduction in the amount ofwater spillage and the enhanced production-planning accuracy. Finally, some results that justify the
development ofthe short-term water-inflow forecasting algorithm are presented.
r 2003 Elsevier Science Ltd. All rights reserved.
Keywords: Neural networks; Forecasts; Efficiency enhancement; Hydroelectric systems; Power system control; Data reduction; Management
systems
1. Introduction
With the recent changes in the energy sector,
electricity has become a commodity that is subject to
market mechanisms. One feature that distinguishes
electricity from other goods is that electric energy
cannot, in general, be stored, which results in high daily
and seasonal price fluctuations. There are, however,
some exceptions to this rule: hydroelectric power plants
(HPPs) have the ability to accumulate water in
reservoirs and so can adapt their production to meet
demands. This means that efficient water management
can positively affect the income of a HPP company.
The main consideration to take into account when
dealing with the issue ofwater management is the
capacity ofthe HPP’s reservoir. In the case oflarge
accumulations, the water level ofthe reservoir is mainly
affected by seasonal hydrological changes, which
involves long-term planning. On the other hand, the
operation ofa run-of-river HPP is closely related to the
current natural water inflow. Therefore, computational
tools for forecasting the water inflow can be of great
help when it comes to optimising the exploitation ofthe
HPP’s reservoir water.
The research presented in this paper is focused on
improving the water management for a run-of-river
HPP using short-term water-inflow forecasts (STWIFs).
A novel methodology for building an efficient STWIF
algorithm is presented. STWIF algorithms may differ
from case to case, depending on the characteristics of
the river and the HPP’s reservoir capacity. As a
result, the emphasis is on an accurate problem definition
and the problem-solving requirements. The model-
building process was performed according to the
purposes set for the model and the available system
information.
There are several parameters that must be taken into
consideration when predicting water inflow. The most
important ofthese are the precipitation data for the
river basin. The mathematical relationship between the
natural inflow and the amount ofprecipitation at
ARTICLE IN PRESS
*Corresponding author. Tel.: +386-1-4768222; fax: +386-1-
4264651.
E-mail addresses: dejan@strela.fe.uni-lj.is (D. Paravan),
tomaz.stokelj@hse.si (T. Stokelj), robert@strela.fe.uni-lj.is (R. Golob).
0967-0661/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0967-0661(03)00106-0