INTEGRATED MANAGEMENT OF LANDBASED
ACTIVITIES IN THE SÃO FRANCISCO BASIN PROJECT
ANA/GEF/UNEP/OAS

Subproject 4.4 Subsidies for the Formulation of Operational Policies for the
Great Reservoirs in the São Francisco River Basin

Executive Summary of the Final Report

SUBSIDIES FOR THE FORMULATION OF OPERATIONAL
POLICIES FOR THE GREAT RESERVOIRS
IN THE SÃO FRANCISCO RIVER BASIN

Subproject Coordination
Marcos Airton de Sousa Freitas
Agência Nacional de Águas

Consultants
João Eduardo Gonçalves Lopes
Mario Thadeu Leme de Barros
João Luiz Boccia Brandão

Contract CPR/OAS no 31631 L1

November 2002

SUBSIDIES FOR THE FORMULATION OF OPERATIONAL
POLICIES FOR THE GREAT RESERVOIRS
IN THE SÃO FRANCISCO RIVER BASIN

EXECUTIVE SUMMARY

INTRODUCTION
The aim of this Activity 4.4 (Subsidies for the Formulation of Operational Policies for the
Great Reservoirs of the São Francisco River Basin) is to present alternatives for operational
models of the reservoirs in the São Francisco River, taking into account the multiple uses of its
water resources, as part of the Integrated Management of Land Based Activities in the São
Francisco River Basin Project (GEF/ANA/OAS/UNEP).
Strategically, this will provide support to the Basin Committees in the allocation of water for the
various uses, thus establishing a new approach to the integrated management of water resources
and contributing to the sustainable development of the São Francisco River Basin, as well as to
that of the Northeastern Region, as a whole.
The utilization of the existing multiple uses reservoirs in the São Francisco River Basin will bring
benefits to the water users in the region. The economic optimization and the sustainable
exploitation of water resources will increase the gains to the users, contributing to the integrated
management of the Basin.
Sobradinho and Três Marias are the two large capacity reservoirs in the Basin, with pluriannual
regulation cycles. All other reservoirs have lower capacities and an annual operational cycle,
having little control over the multiple uses of the water.
The main objectives of this Activity are:
· Detailing the methodology used in the ONS models for the medium-term planning of
the operation of the Northeastern Subsystem and verifying the possibility of
incorporating multiple uses into them;
· testing the use of other models for simulation of hydropower generation in the São
Francisco River (chapters 3 and 4);
· estimating losses of power resulting from water diversion, for both consumptive and
non-consumptive uses, in each stretch of the river, based on different scenarios (chapter
5.3); and
· evaluating alternatives for the compensation of the losses of power due to the multiple
uses.
i

1. THE MAIN RESERVOIRS IN THE SÃO FRANCISCO BASIN AND BASIC DATA
The São Francisco River Basin, with approximately 640,000 km2, has six power generating
reservoirs, whose basic information is shown in Chart 1.

Chart 1. Hydroelectric Reservoirs in the São Francisco River Basin
Distance
Incremental
Distance
Drainage
Effective Installed
to
drainage
Plant Company
between
Area.
Volume.
Power
mouth
area
plants (km)
(km²)
(hm³)
(MW)
(km)
(km²)
Três Marias
CEMIG
2220 50560
50560
15278
396
Sobradinho CHESF
800 1420
498425
447865
28669
1050
Itaparica CHESF
310 490
587000
88575
3548
1500
Moxotó
CHESF
270 40
599200
12200
226
400
Paulo Afonso 1/3
CHESF
270 0
599200
0 90
1423
Paulo Afonso 4
CHESF
270 0
599200
0 30
2460
Xingó CHESF

210 60
608700
9500 5
3000

Two of these reservoirs have a large capacity: Três Marias and Sobradinho. Both of them have a
pluriannual regulation capacity, meaning that their filling-emptying cycles are longer than one
year. They were built primarily for the production of electrical power, but, given their strategic
importance to the river regulation, other uses may be considered in their operational procedures.
Figure 1 presents the topological scheme of the system.

Figure 1. Scheme of the São Francisco River Basin
ii

The ONS RE 3/199/2001 Report ("Average Monthly Flows in Hydroelectric Reservoirs: 1931-
1988 Period", October 2001), presents the series of natural incremental flows used in the power
generation planning. These incremental series indicate the contribution of the intermediate
drainage area between the two consecutive plants.
Chart 2 shows, for each month of the year, the average values of the incremental series for the
São Francisco plants.

Chart 2. Incremental Average Monthly Flows (m3/s) - 1931 to 1998

Paulo Afonso/
Três Marias (*) Sobradinho
Itaparica
Xingó
Moxotó
Jan
1,535
3,288
42 0 10
Feb
1,435
3,684
81 0 10
Mar
1,160 3,789 195
0
10
Apr
784 3,128 156 0
10
May
480
1,951
42 0 10
June
361 1,312 7
0
10
July
288 1,089 2
0
10
Aug
234 944 1 0 10
Sept
219 822 1 0 10
Oct
290 875 1 0 10
Nov
582 1,317 4
0
10
Dec
1,100
2,338
31 0 10
aver. flow
706
2,045
47 0 10
Specific
14.0
4.6 0.5 0 1.1
discharge
(l/s/ sq. km)
Natural MLT
706 2,750 2,797 2,797 2,807
(*) Not incremental values, as the reservoir is situated further upstream

Incoherencies are verified in the series of incremental flows. Between the Paulo Afonso/Moxotó
system and Itaparica, zero flows are indicated throughout the period, despite the drainage area of
12,200 km2. The incremental flows at Xingó are constant in the entire period and equivalent to 10
m3/s, resulting from a 9,500 km2 drainage area.
The ONS RE 3/159/2001 Report ("Liquid Evaporation in Hydroelectric Plants", August 1st,
2001) presents the studies for estimating the evaporation to be used in power generation planning.
The recommended values for the liquid evaporation are shown in Chart 3 and may be compared
to values used in the past years. The latter, which can be obtained in the Eletrobrás' archives, are
tabulated in Chart 4.
With the exception of Três Marias, there is a considerable difference in the evaporation values in
the other reservoirs. Apparently, there are some inconsistencies in the data collection in Chart 3,
as the liquid evaporation in the semi-arid region is indicated lower than that in the wetter
iii

Southeastern Region. Chart 3 shows apparently low values, which are a third of the values which
have been until now (Chart 4).
Chart 3. Liquid Evaporation in mm/month (SisEvapo - Normal 61-90)

Paulo Afonso/
Três Marias
Sobradinho
Itaparica
Xingó
Moxotó
Jan 3 41 39 38 32
Feb 0 29 30 35 26
Mar 27 23 25 37 22
Apr 47 36 24 51 22
May 63 48 32 61 30
June 64 58 39 64 36
July 60 56 44 48 38
Aug 54 47 48 22 40
Sept 55 51 50 17 39
Oct 42 46 47 9 37
Nov 24 47 46 17 34
Dec 24 56 44 40 34
TOTAL
463 538 468 439 390

Chart 4. Liquid Evaporation in mm/month (SIPOT)

Paulo Afonso/
Três Marias
Sobradinho
Itaparica
Xingó
Moxotó
Jan 2 118 140 140 140
Feb 0 106 109 109 109
Mar 22 81 81 81 81
Apr 40 132 105 105 105
May 51 153 109 109 109
June 55 142 101 101 101
July 50 158 123 123 123
Aug 42 181 158 158 158
Sept 57 197 180 180 180
Oct 48 189 195 195 195
Nov 23 114 158 158 158
Dec 29 98 152 152 152
TOTAL 419 1669 1611 1611 1611

2. MULTIPLE USE OF WATER RESOURCES IN THE SÃO FRANCISCO BASIN
Data for diverse non-consumptive and consumptive uses of water were collected:
Non-consumptive uses -
· Power generation Physical data of the power plants
· Flood Control Maximum flows which cause severe flood damages downstream from
the power plant.
iv

· Navigation Minimum flows to allow a minimal navigable depth, in the navigable
stretches of the river.
Consumptive uses -
· Irrigation Irrigation demands and transposition needs (based on transposition studies
made by institutions working in the Basin).
There are other uses for which the insufficiency of information will not let them be considered in
the operation of the reservoir. Among those there are: domestic water supply, recreation and
tourism, requirements for the dilution of pollutants, fishing activities, etc.
The utilization of hydroelectric plant's reservoirs for multiple purposes might affect its operation.
It might impact its firm power production, as a result of restrictions to maximum and minimum
water levels (which will affect the generating capacity, the performance of the generating units
and the available volume of water to be turbined at the most appropriate moments) and
limitations on maximum spills or minimum downstream flow.
Water diversions for other uses may also reduce the inflow to the hydroelectric reservoirs, thus
reducing their power production. The need to maintain a minimum flow in the river stretch
between the dam and the powerhouse, in the case of plants where the latter is not near the dam,
will also result in a reduction of the turbined flow.
According to Law nr. 9,433 (January 8th, 1997), reservoir water management should always give
priority to multiple uses. In this manner, as hydropower generation is one of those uses, it should
be optimized, jointly with other uses, such as irrigation, navigation, flood control, recreation and
tourism, water supply (domestic, industrial and wildlife) and preservation of aquatic flora and
fauna.
The second revision of the ONS RE 3/092/2001 Report (August7th, 2001) presents the
"Inventory of Restrictions on the Hydraulic Operation of Hydroelectric Power Plants". Chart 5
summarizes the restrictions relative to the São Francisco Basin.

Chart 5. Summary of the existing hydraulic restrictions in the São Francisco Basin
variation in
Operational Restrictions
minimum flow
Company power
plant
outflow rate
(m3/s)
Upstream Downstream
(m3/ s.day)
(m)
(m3/s)
CEMIG
Três Marias
500
400 - 700
-
2,500
CHESF Sobradinho
1,300
1000
-
8,000
CHESF
Itaparica
1,300
-
max =304 m
8,000
CHESF Moxotó
-
-
-
8,000
CHESF
Paulo Afonso 1/3
-
-
-
8,000
CHESF
Paulo Afonso 4
-
-
-
8,000
CHESF
Xingó
1,300
-
-
-
v

The "Annual Flood Prevention Plan - 2000/2001", prepared by the ONS, presents the studies
which have been made for all the Brazilian hydropower system. According to the document,
flood control in the São Francisco Basin is accomplished through the operation of the Três
Marias, Sobradinho and Itaparica reservoirs, with the allocation of waiting volumes, for flood
control.
Recently, ANA, taking into consideration requests, from the Electrical Sector, about present and
future irrigation requirements in the São Francisco River Basin, issued the Resolution 145, in
July 22nd, 2002. These estimates for irrigated areas and the respective water requirements, as
published in the Resolution, can be seen in Charts 6 and 7.

Chart 6. Estimated irrigated area (in hectares) in the São Francisco River Basin
Year
1999 2000 2001 2002 2003 2004 2005 2006 2007
Upstream from
162,407 166,305 170,203 174,101 177,999 181,897 185,795 189,693 193,591
Sobradinho
Between Sobradinho 149,619 153,210 156,801 160,392 163,983 167,575 171,166 174,757 178,348
and Itaparica
Between Itaparica
6,885 7,050 7,216 7,381 7,546 7,711 7,877 8,042 8,207
and Xingó
Downstream from
14,399 14,745 15,090 15,436 15,781 16,127 16,473 16,818 17,164
Xingó
TOTAL
333,310 341,310 349,310 357,310 365,310 373,310 381,310 389,310 397,310

Chart 7. Estimated mean annual discharges (in m3/s) for irrigation in the São Francisco
River Basin
Year
1999 2000 2001 2002 2003 2004 2005 2006 2007
Upstream from
94.2 96.5 98.7 101.0 103.2 105.5 107.8 110.0 112.3
Sobradinho
Between Sobradinho 86.8 88.9 90.9 93.0 95.1 97.2 99.3 101.4 103.4
Itaparica
Between Itaparica
4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8
and Xingó
Downstream from
8.4 8.6 8.8 9.0 9.2 9.4 9.6 9.8 10.0
Xingó
TOTAL
193.3 198.0 202.6 207.2 211.9 216.5 221.2 225.8 230.4

The estimates of irrigated area in the Basin were based on a survey made by CODEVASF. A
growth rate of 8,000 ha/year was projected, maintaining the proportionality. A mean specific
discharge of 0.58 l/s.ha was estimated.
The mean monthly releases to meet the irrigation demands of the São Francisco Basin are
estimated based on the average seasonality factors, provided by ANA and shown in Chart 8.
vi

Chart 8. Average seasonality factors (ANA) for the irrigation demands in the São
Francisco Basin
JAN FEB MAR APR MAY JUNE JULY AUG SEPT OCT NOV DEC
1.144 0.877 0.839 0.815 0.912 0.839 0.884 1.001 1.113 1.286 1.189 1.101

3. ANALYSIS AND DIAGNOSTIC OF THE ONS MODELS
At the present, the operation of the Brazilian power generation system is made by ONS (National
System's Operator). According to existing regulations, the companies which own the reservoirs
must observe the decisions made by ONS, regarding releases to be made.
The ONS has a clearly defined mission and must ratify its proceedings in the ANEEL, the
National Agency of Electrical Power. The ONS aims the minimization of the operational costs of
entire interconnected Brazilian system, in order to provide a firm supply, to meet power demands.
As hydroelectric power generation is just one of the uses of water, the operation should be
optimized together with other uses, such as irrigation, navigation, flood control, recreation and
tourism, while in harmony with the preservation of the environment and of the water quality.
The ONS uses a chain of models and computer programs for defining the operational rules of the
National Interconnected System, to whom the São Francisco power plants belong. Among the
most important models used for defining the operational rules, two of them are primordial:
NEWAVE Optimizes the operation of the Brazilian system by considering four equivalent sub-
systems. It adopts a future horizon of five years, at monthly intervals. It uses dynamic dual
stochastic programming for decisions regarding thermal generations, assessment of warranty of
supply and the estimates of the energetic value of the water. Its main outputs are the water future
value curves, which are used in the DECOMP model to estimate the power produced by each of
the reservoirs in the system.
DECOMP Optimizes the system considering individual plants with a one year horizon, based
on weekly intervals, for the elaboration of the monthly operating plan (PMO). The curves of
future value of the water, determined by the NEWAVE, are used by DECOMP to define the
optimal rules for each of the reservoirs in the system.
The Dynamic Dual Stochastic Programming (PDDE) approach, used by the NEWAVE, avoids
the dimensionality problems associated with discretizing of states' spaces, allowing the
establishment of optimal operating strategies for the interconnected hydrothermal systems. The
main advantages of this technique are the explicit representation of the interchanges among
subsystems and the use of a monthly auto-regressive power inflow model, of pth order, which
may be used in the stages of calculation and/or simulation of the operation. Its disadvantage is
that, given the size of the system, the reservoirs in each subsystem need to be grouped into a
single equivalent plant.
vii

The hydroelectric generation system is represented by an equivalent power model. In this model,
the hydroelectric generating facilities of each region are represented by an equivalent power
reservoir, whose main parameters are the variables in the reservoirs' mass balances. These
variables are the natural inflows, controllable and uncontrollable spills, maximum and minimum
storages, evaporation losses, minimum flows and water diversions for other uses.
The issue of the multiple uses of water is taken into account through the minimum flows
constraints on and the deviations to meet consumptive demands. In these cases, those elements
are considered in the composition of the equivalent reservoirs for each subsystem and are treated
as power losses, by the model.
The non-consumptive uses are represented by the minimum flows and by the establishment
definitions of the maximum and minimum water levels. Usually, considering these types of
restrictions may properly represent uses associated with navigation, recreation and tourism,
assurance of water quality, environmental preservation and fishing activities.
Nevertheless, the unavoidable flaws verified when representing aggregated systems should be
emphasized. The irregular rainfall distribution in a large basin, aggravated by the differences in
reservoir regulating capacities, due to their spatial distribution, are liable to causes distortions in
the calculations.
Another problem related to the aggregation is that it will sacrifice the analysis of multiple uses.
As water uses are specified by reservoir, the aggregation process compromises the possibility of a
distributed analysis of the operation for multiple uses. The NEWAVE may aggregate reservoirs
for a power analysis, but cannot consider the spatial variability of the demand for other uses in
the Basin. Important questions, such as reservoir operation in the dry seasons, with the setting of
priorities for uses other than the power generation, are not well handled with this type of
approach.
The DECOMP (Determination of the Coordination of the Short-Term Operation) is used in the
planning of short-term operations of the hydrothermal generation system. It is part of the chain of
models used in the planning of the generation system and attempts to show with greater details
the elements of this system.
In this manner, the model aims at setting the power generation goals for each plant, in order to
meet the load, while minimizing the expected operating costs of the system, throughout the
planning horizon. This cost is made up of the expenditures with fuel for the thermal plants, added
to eventual fines associated with the failure to meet the load (cost of the deficit) and with
reservoirs' spillage. The planning horizon is up to one year, discretized in monthly stages.
The NEWAVE and DECOMP models, used by the ONS, consider the multiple uses as
operational constraints and as water diversions from the modeled systems.
The consideration of these elements in the modeling may impact the reservoirs' operating
policies, as the increase in the number of restrictions reduces the range of operative possibilities
viii

and degrees of freedom, thus creating many situations in which feasible solutions are not
practicable.
In 2002, the impact of these releases and restrictions has not been significant on the São
Francisco. Nevertheless, with the growing demand, the number of conflicts tends to increase,
requiring that, in the future, the optimization process of hydroelectric systems no longer treat the
multiple uses in the form of restriction equations. The problem should be addressed in an
alternate way, by incorporating multi-objective functions, which can simultaneously represent
benefits from the various uses.
From this analysis, it can be deduced that the NEWAVE and the DECOMP models do not
comply with multiple use operations in a satisfactory way. Serious restrictions should be imposed
to the conceptual approach of the NEWAVE/DECOMP set of programs, once that the obtained
rules are always optimal, from the power generation point of view, but are obtained through a
methodology which is inappropriate for multiple uses. As previously mentioned, the aggregation
and subsequent disaggregation processes bring serious problems to the establishment of rules for
multiple use. Important issues such as the multiple use operation under water rationing criteria or
rules to account for the spatial and temporal distribution of water (irrigation, for example) cannot
be adequately solved by these models.

4. AN ALTERNATIVE METHODOLOGY FOR THE OPERATION OF MULTIPLE
USE RESERVOIRS
This section deals with the methodology of the model to be used in the operation of the reservoirs
in the São Francisco River Basin. A general approach to the mathematical modeling in this field
is made, followed by a number of suggestions for treating the focused case, the São Francisco
River system.
As mentioned in the previous chapter, the models being used by the ONS, especially the
NEWAVE/DECOMP chains, were made for main-frame computers. These are models developed
years ago, when software was not as sophisticated as they are now. Besides, a few years ago, only
large computers were capable of processing routines such as the used in the present models.
Today, there are software and micro-computers to solve the same problems, employing the DSS
(Decision Support Systems) philosophy.
In the present case, the operation of the reservoirs in the São Francisco can be treated both
through simulation and optimization. Several studies were made using these methodologies.
However, a bibliographical review of these studies is not part of the scope of this particular
document. It is the purpose of this work to discuss a methodological proposal of an analysis
which will consider the operation of the system according to a multi-objective concept, more
specifically to the multiple uses of the water.
Considering the modeling aspects, according to the approach previously described, it is
recommendable the use of an optimization model with newly developed algorithms, recently
placed on the market, with high degree of precision and processing speed.
ix

The formulation of the problem for establishing operational rules for a system of reservoirs, with
the use of mathematical optimization, must comply with the following characteristics:
Objective function: definition of a function which maximizes or minimizes certain objectives
specified by the decision maker; the objectives are represented by a function, which is related to
the decision variables, usually flows allowed by the reservoirs. The definition of these flows sets
the operational rules.
Restriction Equations: the variables involved in the optimization should comply with a series of
restriction equations of the following types:
· Physical characteristics of the reservoirs;
· Physical characteristics of the hydraulic and mechanic structures;
· Operational characteristics;
· Governing water balance equations for the reservoirs.
In the case of the São Francisco River, there are basically two conflicting uses involved:
generation of power and irrigation. The water supply demands for domestic and industrial uses
can be lumped together and be deducted from the irrigation total or from the actual inflows to the
reservoirs. However, this is not critical. Therefore, the Restrictions Method is suitable for the São
Francisco, considering the two basic uses (power and irrigation). The supply amounts can be
grouped together into one volume, which can be called consumptive use, and inserted in the
water balance equations. Another important parcel is the evaporation from the basin. This should
be included in the balance equation. Other uses, such as navigation, recreation and conservation,
among others, can be included in the restriction equations.
Another fundamental issue in the optimization is the type of equations involved, linear or not.
Linear optimizations are easier to be solved than the more complex non-linear ones. The
optimization of the operation of reservoirs involves both kinds of equations, and is therefore a
problem to be solved with non-linear methods.

5. SUBSIDIES FOR THE OPERATIONAL PROCEDURES OF THE MAIN
RESERVOIRS IN THE SÃO FRANCISCO BASIN

This section will consolidate the various points focused throughout the report.
The system optimization's tests used demand and consumptive use data provided by ANA. The
methodology described in Section 4 was executed with an application developed with the
EXCEL (97 version) spreadsheet, with the VBA macros and the use of the SOLVER tool.
x

Chart 8 displays the model's output for the three scenarios, varying the irrigation requirements
according to the 2002 demand factor. Irrigation return flow was not considered, as this is a semi-
arid region.
In the case of the dry scenario, it was not possible to attend the minimum flow of 500 m3/s, in
Três Marias, to make navigation possible. The optimization model does not converge to a
practicable solution. Therefore, it was necessary to reduce this restriction from 500 m3/s to 400
m3/s. It should be noticed that this flow is not sufficient to allow navigation downstream from
Três Marias.
Chart 8. Outcome of the optimization

In the same scenario, considering a factor equal to two for the consumptive use, it was not
possible to meet the minimum flow of 1,300 m3/s at Xingó, required to permit water supply to
downstream areas. Here, the flow was reduced to 1,200 m3/s.
In general lines, the results in Chart 8 indicate that each cubic meter of water allocated to
irrigation, in the median the average and dry scenarios, corresponds to a loss of approximately
2.5 MW. In the wet scenario, this loss is approximately 1 MW. The values in Chart 8 were used
to trace the trade-off curves, shown in Figure 2.
The above curves enable the decision makers to evaluate the trade-offs between the different
uses, according to the methodology described in Section 4. It is evident that power generation is
susceptible to irrigation use. Notice that the point corresponding to the flow of 198 m3/s
represents the estimated consumption for 2002.
A preliminary sensitivity analysis was carried out in order to evaluate various components of the
problem. Chart 9 summarizes the evaluated cases.

xi

Figure 2. Trade-off curves between power generation and consumptive uses.

· Case 1 (actual 2002 situation): all the current restrictions, waiting volumes and
minimum flows (presented in sec. 2.2), in addition to consumptive use demands
(presented in 2.) and evaporation data used by the ONS (presented in 1) are included.
· Case 2: does not consider flood control volumes or minimum flow restrictions.
· Case 3: does not consider minimum flow restrictions.
· Case 4: does not consider flood control volumes.
· Case 5: uses reduced evaporation data, estimated with SISEVAPO (presented in sec.1).
· Case 6: neglects the effects of evaporation in the reservoirs.
· Case 7: does not consider restrictions on consumptive uses.
All the cases were analyzed for the three hydrological scenarios: wet, median and dry.

Chart 10 Comparison of cases
Wet Median Dry
Cases
Power
Difference
Power
Difference
Power
Difference
Av.MW
Av MW
Av MW
Av.MW
Av.MW
Av.MW
1
7,172 5,921
4,173(*)

2
7,187 15 5,929 8 4,205 32
3
7,172 0 5,926 5 4,188 15
4
7,187 15 5,924 3 4,183(*)
10
5
7,338 166 6,269 348 4,582 409
6
7,426 254 6,402 481 4,809 636
7
7,368 196 6,327 406 4,669 496
(*) minimum flow reduced to 400 m3/s at Três Marias
xii

The results in Chart 10 show the impact of minimum flows, flood control volumes, evaporation
losses and consumptive use restrictions on the average power generation. These impacts are
summarized as follows:

· Minimum flows: the minimum flow restriction does not impact the generation in the
wet scenario. However, for the median and dry cases, there are small differences in the
order of 0.08% and 0.36%, respectively. It is worth emphasizing that discharging the
minimum flow impedes navigation for the median and dry scenarios, and might not
meet the consumptive uses.

· Waiting volumes: small differences can be noticed in power productions: 0.21%.
0.05% and 0.24% for wet, median and dry periods, respectively. When minimum flow
and flood control restrictions are eliminated, the following differences are seen in the
wet, median and dry scenarios, respectively: 0.21%, 0.14% and 0.77%. It may be
deduced that these two restrictions have little impact on power production.

· Evaporation: evaporation effect is significant in the Basin. When evaporation losses
are not included, power production increases greatly for the three scenarios. For the wet,
median and dry scenarios, the percentages reach 3.5%, 8.1% and 15% respectively.
When considering evaporations estimates made by the SISEVAPO model, these
percentages reach 2.3%, 5.9% and 9.8%, respectively. Thus, it is seen that substituting
the evaporation losses, as proposed by the ONS, may greatly raise the values of the firm
supply from these plants.

· Consumptive uses (irrigation): when consumptive uses are not considered, the
percentual power gains are 2.7%, 6.9% and 12%, for the wet, median and dry scenarios,
respectively. These values are compatible with the evaporation effect.

For illustrative purposes, a few graphs in Figure 2 show the a number of illustrative graphs with
the variation in reservoir storages, in power generations (per plant and by total production) and
the capacity factor for each plant, for case 2 (waiting volume and the minimum flow restrictions
not considered), for the median scenario. The capacity factor is defined as the mean monthly
power produced in the month, divided by the plant's installed capacity. This parameter enables
the identification of the degree of use of the plant.
unrestricted under no restrictions, while maintaining the irrigation demands. This case reflects a
situation in which the generation is not conditioned by flood control or by the maintenance of a
minimum flows. The storage curves show the best situation for an operational policy favoring
power production. It is noticed that water levels in the Sobradinho reservoir are kept at lower
levels for longer periods than in the others. This is probably due to the fact that the optimizer
attempts to reduce the lake's water surface, in order to decrease evaporation losses. The Itaparica
reservoir, the furthest downstream, stays practically full during the period, while the Três Marias,
located at headwaters, is less used, in terms of storage percentage.

xiii

6. CONCLUSION
Throughout his study, several aspects of the operation of the São Francisco's reservoir system
were discussed, viewing the inclusion of multiple uses of water resources in the current
operational rules. The main conclusions derived from the analyses, as well as considerations
made throughout the work, are hereby presented:

Figure 2. No flood control or minimum flow restrictions (Case 2, median scenario)

The Case 2, shown above, was used to check the performance of the hydroelectric system

THE MULTIPLE USE OF WATER RESOURCES IN THE SÃO FRANCISCO BASIN
· The multiple uses of water in the São Francisco River are primarily the hydroelectric
power generation, irrigation, navigation, and, in some stretches, flood control.
· Among these uses, the production of electrical power and irrigation (and the other
consumptive uses) compete most intensely for the water resources in the Basin. This
question will result in important consequences as this competition must intensify in the
future. The Work Group for the Evaluation of the Methodology for Hydroelectric
Plants' Firm Power Estimates has proposed a review of the current methodology. This
should result in the definition of criteria for reallocating the firm power, eventually lost
in consequence of other uses of the water.
xiv

· In this context, it is fundamental to have accurate and sound assessment of the demands,
for irrigation and the other uses. This data will support the studies for determining the
fair allocation of the water resources in the São Francisco Basin.
· This study took irrigation into account, including it in the water balance equation, as a
consumptive use, with no return flow to the Basin itself.
· Another important issue in the Basin is the evaporation loss in the reservoirs. The
climate in the Region is extremely dry, with high evaporation rates. Regarding the
operation of these reservoirs, these losses should be considered, as the operational rules
are affected by these lost volumes.
· It should be emphasized that if the evaporation rates recently reviewed by the Electrical
Sector are used in the modeling, the generation of the São Francisco, based on
preliminary estimates, will increase approximately 300 MW, in average, which
corresponds to 5% of the generated power.
ANALYSIS AND DIAGNOSTIC OF THE ONS MODELS
· The models which are used in the operation of the hydropower system in the São
Francisco River Basin are integrated into the Brazilian system and are processed by the
National Operator of the Electrical System (ONS). The models evaluated in the present
study are the NEWAVE and the DECOMP, which are used in the planning and
programming of the Brazilian electric system.
· Examining these models' documentation it becomes evident that multiple uses could be
accounted for through restriction equations. This is a suitable approach for this kind of
problem. However, the methodology used in these models may not be appropriate to
efficiently deal with the problem of multiple use.
· The NEWAVE model employs aggregation techniques in order to make possible the
optimization of the Brazilian reservoir system possible with Dynamic Programming.
This technique transforms a hundred plants into only four subsystems, resulting in loss
of representability for the multiple uses. The operation of the DECOMP model is
conditioned to results from the NEWAVE. The DECOMP is run for a number of
months ahead, using flow forecasting models, to establish operational rules for each
reservoir in the system. As multiple use is strongly linked to hydrological seasonality,
and this may result in long-term problems (durations longer than a year), both the
DECOMP and the NEWAVE may present flaws when dealing with multipurpose
operation. Therefore, even if multiple use of water is considered, the
NEWAVE/DECOMP package has limitations for this kind of approach.
· Additionally, the NEWAVE/DECOMP is structured for main-frame computers. These
models were developed years ago, when software programs were not as sophisticated as
they are now. A few years ago, only large computers were capable of dealing with huge
routines, such as the ones used by these models. Today, there are software and
xv

microcomputers to solve the same problems, employing the DSS (Decision Support
Systems) philosophy

ALTERNATE METHODOLOGY FOR THE OPERATION OF MULTIPLE USE
RESERVOIRS
· The development of new optimization routines, in parallel with the appearance of faster
computers, with high processing and storage capacities, has resulted in more efficient
and user friendly models. The selection of optimization models for multiple objective
operation of reservoirs should consider recently developed algorithms, with high degree
of accuracy, and assure compatibility with models currently in use by the ONS.
· Adequate approach to the optimization of the multipurpose operation will permit the
simulation of the system.
· In this manner, this study proposes a methodology for the analysis and the establishment
of operational rules for the reservoirs, with the appropriate consideration of the multiple
uses of water in a hydropower system. The proposed method consists of using objective
functions focused on power generation, subject to restrictions imposed by other uses,
while maintaining the individuality of the and using time scales compatible with
multiple uses. With this kind of approach, it is possible to estimate the trade-off curves
for the considered conflicting objectives, especially power generation, irrigation and
flood control. A preliminary version of the model was developed in EXCEL and
Windows Visual Basic. This version can optimize the three regulating reservoirs in the
São Francisco, on a monthly basis, for a 12-month horizon.
SUBSIDIES FOR OPERATIONAL PROCEDURES OF THE MAIN RESERVOIRS IN
THE SÃO FRANCISCO.
· Preliminary processing attested the technical feasibility of this approaching technique to
the problem. The obtained results anticipate a tendency towards an increase in the
competition for the two main uses in the basin, irrigation and power generation. These
results show that, for both the dry, the median and the wet scenarios, for each cubic
meter per second of water diverted to irrigation purposes there is an average power loss
of around 2.5 MW, corresponding approximately 22 GWh/year.
· With the growing demand for irrigation water, including multiple uses in the
optimization rules will become mandatory. Preliminary results from this study showed
that the proper consideration of multiple uses may significantly affect the operation of
certain reservoirs.
· Additionally, when evaporation rates are too high, as in the case of the São Francisco
Basin, the optimization technique employed in the used methodology may alter
significantly the form of operation of some reservoirs such as that of Sobradinho.
xvi

· The expansion in the electrical interconnection capacity of the Northeastern Region,
with the Northern and the Southeastern Regions, will constitute the best alternative for
compensating the power losses imposed by the multiple uses of the water. The need to
consider power generation no longer a priority is justified by the fact that other uses,
particularly irrigation, present today a higher social and economic value.
7. RECOMMENDATIONS
7.1 GENERAL RECOMMENDATIONS
According to the analyses and comments presented in this report, it is recommended that:
· Studies should be contracted, viewing the development and implementation of planning
models for the operation of the reservoir system of the São Francisco. These models
should consider the multiple uses of the water, irrigation in particular, as it is the use
with the greatest impact on power generation.
· Studies should be made in order to review the hydrometeorologic series used in the
planning of the operation of the São Francisco's hydropower system. The verified
inconsistencies in the data have a direct effect on the outcomes of models and analyses.
· A survey to update data irrigation demands and flood control, as well as in other uses,
should be carried out to allow better prospective analyses. If possible, this task should
be carried out together with the Hydrographic Plan for the São Francisco Basin, as
established in the Water Resources Management National System.
· Studies should be made on the development and implantation of models for the analysis
and support of decisions based on a large number of factors, additional tools to the
planning model of the proposed operation. Bearing in mind the collegiate style of the
management of water resources in Brazil, this tool could help the discussions which
ANA, ANEEL, the Committee of the São Francisco River Basin, ONS and other entities
will certainly have when they have to decide on problems and solutions which involve
different competitive uses for the water in the basin;
· A study should be carried out to assess the expansion of the Basin's hydrometeorologic
monitoring network, in order to meet the requirements for planning, operation and
control of the water uses in the São Francisco.
· A precise scientific study on the evaporation processes in the basin, given the enormous
inconsistencies verified among existing historical and statistical data sets. It was
noticed, in the early stages of the present work, that evaporation losses are too high,
particularly in the Sobradinho reservoir, directly affecting the operational rules and
remarkably the electrical sector.
· ANA should discuss more closely with the National Operator of the Electrical System
(ONS) the multiple use issue, as the currently used planning methodology, regarding
power generation, is ineffective for treating conflicting uses in a Basin.
xvii

· The electric interconnecting capability among the Northeastern, Southeastern, Western-
Central and Northern Regions, in order to allow the use of water from São Francisco,
for multiple uses, without compromising the energy supply in the Basin.

7.2 SPECIFIC RECOMMENDATIONS: Development and Implantation of an
optimization model for the multipurpose operation of the São Francisco's reservoirs
For solving operational problems of reservoirs in the São Francisco River, mainly used for power
generation and irrigation, it is recommended the use of a models based on the method of
restrictions. The objective function should focus on the optimization of power production, subject
to various restrictions, including the guarantee of a firm water supply for irrigation. This
recommendation may be justified by the following considerations:

· the method allows clear identification of the trade-offs among conflicting uses (power
generation and irrigation);
· it does not require the prior establishment of relative weights for each of the elements
composing the objective function, thus avoiding subjective parameters ;
· huge mathematical and computational work are not required, in dealing with the
problem, as the method with a general and pondered objective function involves a
greater number of decision variables.
The methodological bases for developing this model, presented in the Final Report, deal with the
general formulation of the problem as one optimization of operation of reservoir systems. In
general terms, the model should comply with the following requirements:

· individualized representation of the reservoirs and capability of modelling at least
twenty 20 of them;
· performing iterations on a monthly basis, with a possibility of continuous processing for
at least a 5-year (60-month) horizon;
· capability of using non-linear optimization routines;
· dealing with the optimization problem using the restriction method, with a power
generation related objective function, subject to meeting the demands for other uses of
water;
· global optimization of the objective function, considering all calculation steps;
· the implementation of graphic and user-friendly interfaces for both the input and output
of data;
xviii

· processing on a Windows platform;
· improved processing time, without compromising the representability precision of the
results;
· capable of future expansion, such as consideration of other restrictions (electric ones,
for example), stochastic treatment of incoming flows and interchange among sub-
systems.
The model should be developed by a team belonging to an institution or research and technologic
development group with vast experience in the mathematical and computational treatment of the
problem, as well as with the operation of multipurpose reservoirs. The work should be done
under ANA's coordination, with participation of ONS, CHESF, CEMIG and ELETROBRÁS.

The development of the model should take 18 months, costing approximately US$ 300,000.00.
The task would require the following experts: one project manager (18 months, US$ 44,000.00),
one international consultant (4 months, US$ 16,000.00), three senior engineers (18 months, US$
180,000.00) and one senior system analyst (18 months US$ 60,000.00).

The final product will be a software to be used by institutions involved in the planning activities
for the operation of the São Francisco system, programming the releases based on trade-offs
among conflicting uses. Additionally, this software may be used in the evaluation and planning of
new developments, or for subsidizing studies in the area of water resources management
xix