Vous êtes sur la page 1sur 6

1

Profit-Based Short-Term Hydro Scheduling considering Head-Dependent Power Generation


S. J. P. S. Mariano, J. P. S. Catalo, Member, IEEE, V. M. F. Mendes, and L. A. F. M. Ferreira

AbstractThis paper is on the problem of short-term hydro scheduling, particularly concerning head-dependent reservoirs under competitive environment. We propose a method, based on nonlinear programming, for optimizing power generation efficiency. The proposed method considers not only that the hydroelectric power generation is a function of the water discharge and of the head, but also that the maximum power generation is head-dependent. Numerical results based on a realistic cascaded hydro system illustrate the proficiency of the proposed method. Index TermsHydroelectric power generation, nonlinear programming, power generation scheduling.

A
b, b x, x

Constraint matrix. Upper and lower bound vectors on constraints. Upper and lower bound vectors on variables.

i , i Power efficiency limits of plant i.


h i , h i Head limits of plant i. l i , l i Water level limits of reservoir i. q ihx q ihn

Maximum water discharge by reservoir i at h i . Maximum water discharge by reservoir i at h i . II. INTRODUCTION

I. NOMENCLATURE The notation used throughout the paper is stated as follows:


I, i Set and index of reservoirs. Forecasted energy price in hour k. Power generation of plant i in hour k. Future value of the water stored in reservoir i. Water storage of reservoir i at end of hour k. Inflow to reservoir i in hour k. Set of upstream reservoirs to reservoir i. Water discharge by reservoir i in hour k. Water spillage by reservoir i in hour k. Power efficiency of plant i in hour k. Head of plant i in hour k. Water level in reservoir i in hour k.
K , k Set and index of hours in the time horizon.

k
p ik

i
v ik
a ik

Mi
q ik s ik

ik
h ik l ik

v i , v i Water storage limits of reservoir i.

q ik
qi

Maximum water discharge by reservoir i in hour k. Minimum water discharge by reservoir i. Vector of decision variables.

F (.) Nonlinear function of variables.

S. J. P. S. Mariano and J. P. S. Catalo are with the Department of Electromechanical Engineering, University of Beira Interior, R. Fonte do Lameiro, 6201-001 Covilha, Portugal (e-mail: sm@ubi.pt; catalao@ubi.pt). V. M. F. Mendes is with the Department of Electrical Engineering and Automation, Instituto Superior de Engenharia de Lisboa, R. Conselheiro Emdio Navarro, 1950-062 Lisbon, Portugal (e-mail: vfmendes@isel.pt). L. A. F. M. Ferreira is with the Department of Electrical Engineering and Computers, Instituto Superior Tcnico, Technical University of Lisbon, Av. Rovisco Pais, 1049-001 Lisbon, Portugal (e-mail: lmf@ist.utl.pt).

N this paper, the short-term hydro scheduling (STHS) problem of a head-dependent cascaded hydro system is considered. Typically, hydro plants with a small storage capacity, known as run-of-the-river hydro plants, are considered to operate under stationary conditions with constant head and at the maximum water level in the reservoirs, corresponding by design to the optimum efficiency operating point. However, it is often required to change this policy, thus incurring into head changes. The operating efficiency is sensitive to the head head change effect [1]. In the STHS problem a time horizon of one to seven days is considered, usually discretized into hourly periods. The STHS problem is treated as a deterministic one. Where the STHS problem includes stochastic quantities such as inflows to reservoirs or energy prices, the corresponding forecasts are used [2]. In a competitive environment, the goal is to maximize the value of total hydroelectric generation throughout the time horizon considered, satisfying all hydraulic constraints, and consequently to maximize the profit of the hydro generating company (H-GENCO) from selling energy [3]. Dynamic programming (DP) is among the earliest methods applied to the STHS problem [4]-[6]. Although DP can handle the nonlinear characteristics present in the hydro model, for systems with cascaded reservoirs the computational burden increases exponentially with problem dimension, causing DP to suffer from the curse of dimensionality, more difficult to avoid in short-term than in long-term optimization without losing the accuracy needed in the model [7]. Artificial intelligence techniques have also been applied to the STHS problem, namely, neural networks [8], [9], and genetic algorithms [10], [11]. However, due to the heuristics used in the search process only sub-optimal solutions can be reached.

A natural approach to STHS is to model the system as a network flow model [12]-[14], because of the underlying network structure subjacent in cascaded hydro systems. The network flow model is often simplified as a linear or piecewise linear one. Linear programming (LP) is a widely used method for STHS [15]-[17]. LP algorithms lead to extremely efficient codes, implementations of which can be found commercially. Mixed-integer linear programming (MILP) is becoming frequently used for STHS [18]-[22], where binary variables allow modeling of start-up costs, which are mainly caused by the increased maintenance of windings and mechanical equipment, and by malfunctions of the control equipment. However, LP typically considers that power generation is linearly dependent on water discharge, thus ignoring head dependence to avoid nonlinearities, leading to inaccuracy. Also, the discretization of the nonlinear dependence between power generation, water discharge and head, used in MILP to model head variations, augment the computational burden required to solve the STHS problem. A nonlinear model has many advantages compared with a linear one. A nonlinear model expresses hydro generation characteristics more accurately and the head change effect can be taken into account [1], [23], [24]. In this paper, we propose a nonlinear programming (NLP) method to solve the STHS problem. We report our experience with the proposed nonlinear model on a realistic hydro system with seven cascaded reservoirs and considering a time horizon of one week. This paper is organized as follows. In Section 3, the mathematical formulation of the STHS problem is provided. Section 4 presents the proposed NLP method to solve the STHS problem. In Section 5, the proposed NLP method is applied on a realistic cascaded hydro system to demonstrate its effectiveness over classical optimization methods that ignore head dependence. Finally, concluding remarks are given in Section 6. III. PROBLEM FORMULATION The STHS problem can be stated as to find out the water discharges, the water storages, and the water spillages, for each reservoir i at all scheduling time periods k that maximizes (or minimizes) the performance criterion subject to all hydraulic constraints. Also, the reservoir storage at the end of the scheduling horizon must be decided according with future operation.

In (1), the first term is related to the revenues of each plant in the hydro system during the time horizon considered and the last term expresses the future value of the water stored in the reservoirs in the last hour K. B. Hydro Constraints The optimal value of the objective function is computed subject to constraints of two kinds: equality constraints and inequality constraints or simple bounds on the variables. The decisions are discretized into one hour periods. 1) Water Balance: The water balance equation for each reservoir is formulated as
v i k = v i , k 1 + a i k +
mM i

(q m k + s m k ) q i k s i k
i I, k K
(2)

assuming that the time required for water to travel from a reservoir to a reservoir directly downstream is less than the one hour period. 2) Power Generation: Power generation is considered a function of water discharge and hydro power efficiency
p i k = q i k i k (h i k ) i I , k K

(3)

We consider hydro power efficiency expressed as the outputinput ratio [25], which in turn depends on the head. The hydro generation characteristic can be graphically represented by a family of nonlinear curves, also known as unit performance curves, each curve for a specific value of the head (see Fig. 1).
p ik

q ik

hi

hi

qi
Fig. 1. Unit performance curves.

q ihx q ihn

q ik

A. Objective Function
In the profit-based STHS problem under consideration, the goal is to maximize the profit of the H-GENCO from selling energy in an electricity market. Thus, the objective function to be maximized can be expressed as

3) Head: The head is considered a function of the water levels in the upstream reservoir, denoted by f (i ) in subscript, and downstream reservoir, denoted by t (i ) in subscript, depending respectively on the water storages in the reservoirs
h ik = l
f (i ) k

k
i =1 k =1

p i k + i (v i K )
i =1

(1)

(v f (i ) k ) l t (i ) k (v t (i ) k ) i I , k K (4)

4) Water Storage: Water storage has lower and upper bounds v i v ik v i

i = (l i l i ) / (v i v i ) i I
l i0 = l i i v i
iI

(15) (16)

i I, k K

(5)

5) Water Discharge: We consider a null lower bound for water discharge, but the upper bound for water discharge may be different for each period k according to the value of the head

Substituting (11) into (3) we have


p i k = q i k ( i h i k + i0 ) i I , k K

(17)

0 q i k q i k (h i k ) i I , k K

(6)

Therefore, substituting (4) and (14) into (17), power generation becomes a nonlinear function of water discharge and water storage, given by
p i k = i f (i ) q i k v f (i ) k i t (i ) q i k v t (i ) k + i q i k

6) Water Spillage: We consider only a null lower bound for water spillage
s ik 0 i I , k K

i I, k K

(18)

(7)

with
0 0 i = i (l 0 f (i ) l t (i ) ) + i

Water spillage can occur when without it the water storage exceeds its upper bound, so spilling is necessary due to safety considerations. The initial water storages and inflows to reservoirs are assumed as known input data. IV. SOLUTION METHODOLOGY NLP can be stated as to maximize
F ( x)

iI

(19)

(8)

subject to
b Axb

The parameters given by the product of ' s by ' s are of crucial importance for the behavior of head-dependent reservoirs in a hydro system, setting optimal reservoirs storage trajectories in accordance to their relative position in the cascade [1]. In our model, the maximum water discharge, thus giving the maximum power generation, is considered head-dependent [26], given by
q i k = i h i k + q i0

(9) (10)

i I, k K

(20)

xxx

where the parameters i and q i0 are given by


i = (q ihn q ihx ) / (h i h i ) i I

In (8), the function F (.) is a nonlinear function of variables. Equation (9) corresponds to the equality constraints in (2), with b = b . Equation (10) corresponds to the inequality constraints or simple bounds on the variables in (5), (6) and (7). The upper bound for water discharge implies a new inequality constraint that will be rewritten into (9). In (3) and (4), hydro power efficiency and water level depend respectively on head and water storage. We consider a linearization of hydro power efficiency of plants, expressed as the output-input ratio. Hence, we consider the hydro power efficiency given by
i k = i h i k + i0 i I, k K
i0

(21) (22)

q i0 = q ihn + i h i

iI

Substituting (4) and (14) into (20), the maximum water discharge becomes a function of water storage, given by
q i k = i f (i ) v f (i ) k t (i ) v t ( i ) k + i

i I, k K

(23) (24)

with
0 i = q i0 i (l 0 f (i ) l t (i ) ) i I

(11)

where the parameters i and

are given by (12) (13)

Hence, the new inequality constraint to be rewritten into (9) is given by


q i k + i f (i ) v

i = ( i i ) / ( h i h i ) i I
i0 = i i h i iI

f (i ) k

t (i ) v t (i ) k i

iI

(25)

V. CASE STUDY The proposed NLP method for the profit-based STHS problem has been applied on a realistic cascaded hydro system. Our model has been implemented in MATLAB and solved using the optimization solver package Xpress-MP. The numerical testing has been performed on a 600-MHz-based processor with 256 MB of RAM.

Also, we consider a linearization of the water level function given by


l i k = i v i k + l i0

i I, k K

(14)

where the parameters i and l i0 are given by

A. Input Data

The realistic cascaded hydro system has seven reservoirs and is illustrated in Fig. 2. The hydro plants numbered in Fig. 2 as 1, 2, 4, 5 and 7 are run-of-the-river hydro plants. The hydro plants numbered as 3 and 6 are storage hydro plants, Hence, for these storage hydro plants the head effect is neglected, due to the small change in the head during the short-term time horizon.
a 1k

The energy prices are considered as deterministic input data in this paper. These energy prices are forecasted using techniques mainly based on time series and ARIMA models [27], [28], or on neural networks [29], [30].
B. Result Analysis

The storage trajectories of the run-of-the-river reservoirs are shown in Fig. 4. The solid lines denote NLP results while the dashed lines denote LP results.
100

1 v 1k

Storage of R1 (%)

75 50 25 0 100

s 1k a 2k
2

q 1k a 3k

v 2k q 2k

3 v 3k

q 3k

Storage of R2 (%) Storage of R4 (%) Storage of R5 (%)


Storage of R7 (%)

s 2k

s 3k
a 4k
4

75 50 25 0 100

v 4k q 4k a 5k

s 4k

a 6k
6

75 50 25 0
100

v 6k q 6k

v 5k q 5k

s 6k

s 5k

v 7k

75 50 25 0

s 7k

q 7k

Fig. 2. Cascaded hydro system.

100 75 50 25 0 0 24 48 72 Time (h) 96 120 144 168

The final water storage in the reservoirs was considered equal to the initial water storage, chosen as 80% of maximum water storage. Hence, the future value of the water stored in the reservoirs is not required in (1). The time horizon chosen is one week divided into 168 hourly periods. The energy prices are illustrated in Fig. 3, where $ is a symbolic quantity.
Energy price ($/MWh)

Fig. 4. Storage trajectories of the reservoirs 1, 2, 4, 5 and 7.

120 90 60 30 0 0 24 48 72 Time (h) 96 120 144 168

The comparison of LP with NLP results, shown in Fig. 4, reveals the influence of considering the head change effect in the behavior of the reservoirs. The upstream reservoir should operate at a suitable high storage level in order to benefit the power generation efficiency of its associated plant, due to the head change effect. The storage level in the last downstream reservoir is lower with NLP than with LP, thereby improving the head for the immediately upstream reservoirs.

Fig. 3. Energy price profile.

Hence, a higher efficiency of the last downstream plant is not important for the overall profit in this hydro system. The discharge profiles for the run-of-the-river reservoirs are shown in Fig. 5. Again, the solid lines denote NLP results while the dashed lines denote LP results.
Discharge for R1 (%)
100

TABLE I COMPARISON OF LP WITH NLP RESULTS Average Storage (%) 23.73 23.73 Average Discharge (%) 70.32 71.97 Total Profit ($ 103) 4567.32 4753.50 CPU (s) 2.25 5.90

Method LP NLP

75 50 25 0

Discharge for R2 (%)

100 75 50 25 0

Although the average water discharge is as expected the same for both optimization methods, the average storage is superior with the NLP method, due to the consideration of the head change effect. Thus, with NLP we have a higher total profit for the H-GENCO, about 4%. Moreover, the extra CPU time required is negligible. Hence, the proposed NLP method provides better results for head-dependent cascaded hydro systems. VI. CONCLUSIONS A NLP method is proposed for the STHS problem. The proposed method considers not only the nonlinear dependence between the power generation, the water discharge and the head, but also that the maximum water discharge, giving the maximum power generation, is a function of the head. Numerical testing results show that the method is computationally adequate to the STHS problem for cascaded hydro systems with run-of-the-river hydro plants, because the head effect has to be considered in order to improve power generation efficiency. The proposed NLP method provides a higher profit for the H-GENCO, due to the consideration of the head change effect, in comparison with an LP method that ignores head-dependency. VII. REFERENCES
[1] J. P. S. Catalo, S. J. P. S. Mariano, V. M. F. Mendes, and L. A. F. M. Ferreira, "Parameterisation effect on the behaviour of a head-dependent hydro chain using a nonlinear model," Electr. Power Syst. Res., vol. 76, pp. 404-412, Apr. 2006. L. A. F. M. Ferreira, T. Andersson, C. F. Imparato, T. E. Miller, C. K. Pang, A. Svoboda, and A. F. Vojdani, "Short-term resource scheduling in multi-area hydrothermal power systems," Int. J. Electr. Power Energy Syst., vol. 11, pp. 200-212, Jul. 1989. J. P. S. Catalo, S. J. P. S. Mariano, V. M. F. Mendes, and L. A. F. M. Ferreira, "Scheduling of head-sensitive cascaded hydro systems: a comparison based on numerical simulation results," Int. J. Power Energy Syst., to be published. S. M. Amado and C. C. Ribeiro, "Short-term generation scheduling of hydraulic multi-reservoir multi-area interconnected systems," IEEE Trans. Power Syst., vol. PWRS-2, pp. 758-763, Aug. 1987. S.-C. Chang, C.-H. Chen, I.-K. Fong, and P. B. Luh, "Hydroelectric generation scheduling with an effective differential dynamic programming algorithm," IEEE Trans. Power Syst., vol. 5, pp. 737-743, Aug. 1990. C. Lyra and L. R. M. Ferreira, "A multiobjective approach to the shortterm scheduling of a hydroelectric power system," IEEE Trans. Power Syst., vol. 10, pp. 1750-1755, Nov. 1995. J. M. Pursimo, H. K. Antila, M. K. Vilkko, and P. A. Lautala, "A shortterm scheduling for a hydropower plant chain," Int. J. Electr. Power Energy Syst., vol. 20, pp. 525-532, Nov. 1998. R.-H. Liang and Y.-Y. Hsu, "Short-term hydro-scheduling using Hopfield neural network," IEE Proc.-Gener. Transm. Distrib., vol. 143, pp. 269-275, May 1996.

Discharge for R4 (%) Discharge for R5 (%)


Discharge for R7 (%)

100 75 50 25 0
100 75 50 25 0

100 75 50

[2]
25 0 0 24 48 72 Time (h) 96 120 144 168

[3]

Fig. 5. Discharge profiles for the reservoirs 1, 2, 4, 5 and 7.

The comparison of LP with NLP results, shown in Fig. 5, reveals that the water discharge changes more quickly from the minimum value to the upper value in the LP results than in the NLP results, due to the head change effect. As a new contribution to earlier studies, some shape adaptation is imposed due to the consideration of the maximum power generation as head-dependent. This implies that there is a slope shape at the most favorable price hours of each day, instead of the normal flat shape when the maximum water discharge was considered constant. The main numerical results for the hydro system are summarized in Table I.

[4] [5]

[6] [7] [8]

6 [9] [10] [11] [12] [13] [14] R. Naresh and J. Sharma, "Hydro system scheduling using ANN approach," IEEE Trans. Power Syst., vol. 15, pp. 388-395, Feb. 2000. P.-H. Chen and H.-C. Chang, "Genetic aided scheduling of hydraulically coupled plants in hydro-thermal coordination," IEEE Trans. Power Syst., vol. 11, pp. 975-981, May 1996. P. T. Leite, A. A. F. M. Carneiro, and A. C. P. L. F. Carvalho, "Energetic operation planning using genetic algorithms", IEEE Trans. Power Syst., vol. 17, pp. 173-179, Feb. 2002. D. Sjelvgren, S. Andersson, T. Andersson, U. Nyberg, and T.S. Dillon, "Optimal operations planning in a large hydro-thermal power system," IEEE Trans. Power App. Syst., vol. PAS-102, pp. 3644-3651, Nov. 1983. P. E. C. Franco, M. F. Carvalho, and S. Soares, "A network flow model for short-term hydro-dominated hydrothermal scheduling problems," IEEE Trans. Power Syst., vol. 9, pp. 1016-1022, May 1994. A. R. L. Oliveira, S. Soares, and L. Nepomuceno, "Short term hydroelectric scheduling combining network flow and interior point approaches," Int. J. Electr. Power Energy Syst., vol. 27, pp. 91-99, Feb. 2005. E. B. Hreinsson, "Optimal short term operation of a purely hydroelectric system," IEEE Trans. Power Syst., vol. 3, pp. 1072-1077, Aug. 1988. M. R. Piekutowski, T. Litwinowicz, and R. J. Frowd, "Optimal shortterm scheduling for a large-scale cascaded hydro system," IEEE Trans. Power Syst., vol. 9, pp. 805-811, May 1994. A. J. Wood and B. F. Wollenberg, Power Generation, Operation and Control. New York: Wiley, 1996. O. Nilsson and D. Sjelvgren, "Mixed-integer programming applied to short-term planning of a hydro-thermal system," IEEE Trans. Power Syst., vol. 11, pp. 281-286, Feb. 1996. X. Guan, A. Svoboda, and C.-A. Li, "Scheduling hydro power systems with restricted operating zones and discharge ramping constraints," IEEE Trans. Power Syst., vol. 14, pp. 126-131, Feb. 1999. G. W. Chang, M. Aganagic, J. G. Waight, J. Medina, T. Burton, S. Reeves, and M. Christoforidis, "Experiences with mixed integer linear programming based approaches on short-term hydro scheduling," IEEE Trans. Power Syst., vol. 16, pp. 743-749, Nov. 2001. A. J. Conejo, J. M. Arroyo, J. Contreras, and F. A. Villamor, "Selfscheduling of a hydro producer in a pool-based electricity market," IEEE Trans. Power Syst., vol. 17, pp. 1265-1272, Nov. 2002. E. Parrilla, J. Garca-Gonzlez, "Improving the B&B search for largescale hydrothermal weekly scheduling problems," Int. J. Electr. Power Energy Syst., vol. 28, pp. 339-348, Jun. 2006. E. Ni, X. Guan, and R. Li, "Scheduling hydrothermal power systems with cascaded and head-dependent reservoirs," IEEE Trans. Power Syst., vol. 14, pp. 1127-1132, Aug. 1999. J. P. S. Catalo, S. J. P. S. Mariano, V. M. F. Mendes, and L. A. F. M. Ferreira, "Nonlinear approach for short-term scheduling of a headsensitive hydro chain," presented at the IEEE Power Tech Conf., St. Petersburg, Russia, 2005. [25] A. Arce, T. Ohishi, and S. Soares, "Optimal dispatch of generating units of the Itaip hydroelectric plant," IEEE Trans. Power Syst., vol. 17, pp. 154-158, Feb. 2002. [26] S. Mariano, J. Catalo, V. Mendes, and L. Ferreira, "Power generation efficiency improvement in cascaded and head-dependent reservoirs," presented at the 15th PSCC, Liege, Belgium, 2005. [27] F. J. Nogales, J. Contreras, A. J. Conejo, and R. Espnola, "Forecasting next-day electricity prices by time series models," IEEE Trans. Power Syst., vol. 17, pp. 342-348, May 2002. [28] J. Contreras, R. Espnola, F. J. Nogales, A. J. Conejo, "ARIMA models to predict next-day electricity prices," IEEE Trans. Power Syst., vol. 18, pp. 1014-1020, Aug. 2003. [29] B. R. Szkuta, L. A. Sanabria, and T. S. Dillon, "Electricity price shortterm forecasting using artificial neural networks," IEEE Trans. Power Syst., vol. 14, pp. 851-857, Aug. 1999. [30] J. P. S. Catalo, S. J. P. S. Mariano, V. M. F. Mendes, and L. A. F. M. Ferreira, "Short-term electricity prices forecasting in a competitive market: a neural network approach," Electr. Power Syst. Res., to be published.

[15] [16] [17] [18] [19] [20]

VIII. BIOGRAPHIES
S. J. P. S. Mariano received the M.Sc. degree from the Instituto Superior Tcnico (IST), Lisbon, Portugal, in 1994 and the Ph.D. degree from the University of Beira Interior (UBI), Covilha, Portugal, in 2002. Since 1992, he has been with the Department of Electromechanical Engineering, UBI, where he is currently an Assistant Professor. J. P. S. Catalo (M04) received the M.Sc. degree from the Instituto Superior Tcnico (IST), Lisbon, Portugal, in 2003 and the Ph.D. degree from the University of Beira Interior (UBI), Covilha, Portugal, in 2007. Since 1999, he has been with the Department of Electromechanical Engineering, UBI, where he is currently an Assistant Professor. V. M. F. Mendes received the M.Sc. and Ph.D. degrees from the Instituto Superior Tcnico (IST), Lisbon, Portugal, in 1987 and 1994, respectively. Since 1988, he has been with the Department of Electrical Engineering and Automation, Instituto Superior de Engenharia de Lisboa (ISEL), Lisbon, Portugal, where he is currently a Coordinator Professor. L. A. F. M. Ferreira received the M.S.E.E. and Ph.D. degrees from the Georgia Institute of Technology, Atlanta, in 1983 and 1986, respectively. From 1986 to 1989, he was with the Pacific Gas and Electric Company, San Francisco, CA, where he was a major developer of the Hydro-Thermal Optimization program. Since 1989, he has been with the Department of Electrical Engineering and Computers, Instituto Superior Tcnico (IST), Lisbon, Portugal, where he is currently an Associate Professor.

[21] [22] [23] [24]