Reactive optimal power ow incorporating margin enhancement constraints with nonlinear interior point method

H. Song, B. Lee and Y.-H. Moon Abstract: A reactive optimal power ow incorporating margin enhancement constraints is described. Margin sensitivity at a steady-state voltage instability point is calculated using invariant space parametric sensitivity, and can provide valuable information for selection of effective control parameters. However, weakest buses in neighbouring regions have high margin sensitivities in a certain range, so control determination using only the sensitivity information might cause violation of operational limits of the base operating point, at which the control is applied to enhance voltage stability margin in the direction of parameter increase. A nonlinear interior point method (NIPM) is applied to solve the optimal power ow formulation with margin enhancement constraints, and shunt capacitances are mainly considered as control variables. In the case study, an example applying the proposed method to the IEEE 118-bus test system is shown, and experiences with margin enhancement constraints are also included.

Introduction

Voltage stability is one of the main factors limiting the secure region in todays power system operation and planning [13]. In the deregulated environment, power systems have been experiencing an increase in uncertainty in terms of voltage stability resulting from diverse power transactions and benet-based operational schemes. In addition, this trend is expected to continue and become more profound. For secure operation of systems, therefore, an effective control strategy against voltage instability needs be established. To select effective control parameters, a margin boundary point, which can be a steady-state voltage instability point or a saddle-node bifurcation point, needs rst to be detected using direct or indirect methods. Then, margin sensitivity at the point is calculated with the concept of invariant space parametric sensitivity [4], which was originally proposed in [5]. With margin sensitivity information, in [6], an optimisation formulation of linear programming is presented for control determination against voltage collapse. In the experience of sensitivity analysis, however, the weakest buses in a voltage control area usually have high margin sensitivities in a certain range, so in cases of reactive injection-type controls, applying the determined control using only margin sensitivity information to some of the weakest buses might cause violation of operational limits,
r IEE, 2005 IEE Proceedings online no. 20050027 doi:10.1049/ip-gtd:20050027 Paper rst received 29th January and in nal revised form 27th May 2005 H. Song is with the School of Electronic and Information Engineering, Kunsan National University, Kunsan, Jeonbuk, 573-701, Republic of Korea. He was with the APSRC, Korea University B. Lee is with the Advanced Power System Research Centre, Korea University, Seoul, 136-701, Republic of Korea Y.-H. Moon is with the Testing & Evaluation Division, Korea Electrotechnology Research Institute, Uiwang, Kyounggi-do, 437-808, Republic of Korea E-mail: leeb@korea.ac.kr IEE Proc.-Gener. Transm. Distrib., Vol. 152, No. 6, November 2005

especially voltage limits, of the base point. Therefore, it should be veried whether the operating point after the control is within the given range of acceptable limits. This paper mainly discusses optimal power ow formulations that can cope with voltage stability margin enhancement as well as operational constraints. In the literature, there are various formulations of voltage stability constrained optimal power ow (VSCOPF), and they can be divided into two classes. Class I [710] is to maintain steady-state voltage stability of the given state. This can be available for corrective control strategies when the system is in severe contingent states and hence does not satisfy the given voltage stability criteria. Class II [1116] is to determine preventive control strategies in the normal state considering voltage stability of multi-contingencies. This paper focuses on VSCOPF in Class I. Class I is also divided into two parts according to whether to consideration of the solution of the base case. In [7], a direct interior point method is applied to solve both preventive and corrective control problems for satisfying a certain level of voltage stability margin, but the condition at the base solution after applying the controls is not taken into account. In [8], the network equations at the two points (the base and maximum points) are included in the formulation, and maximisation of active load distance between the two points is incorporated into the objective function. Thus, the dimension of the problem might be twice that of a conventional optimal power ow (OPF). In [9], instead of using active power margin index, L index constraints are placed as voltage stability constraints at the base case point. For real applications, the threshold value of L index for maintaining voltage security of the given system is determined in advance. In [10], a formulation for incorporating an MW margin constraint is presented explicitly and a sequential quadratic programming method is adopted to solve the formulation. In this method, however, optimisation is performed at the stressed load level by the desired margin.
961

This paper presents a methodology to bring about control of reactive power related variables meeting the given voltage stability margin criteria. The algorithm utilises information of margin sensitivity that can be obtained from voltage stability analysis. In addition, it employs a nonlinear interior point method (NIPM) to solve the reactive optimal power ow problem containing margin enhancement constraints (MEC), constructed with the margin sensitivity. Since the formulation includes the network equations of only one operating point where control will be applied, the dimension of the problem to be solved is the same as with the conventional OPF. Utilising NIPM, in addition, makes it possible to apply the proposed method to the maximum point as well as to the base point. 2 Voltage stability constraints

This Section discusses voltage stability constraints and presents their incorporation of into OPF formulations. Because the main equality constraints in OPF are the power ow equations, OPF can be a tool to maintain and/or improve system security in a steady-state manner. VSCOPF can be applied to enhance steady-state voltage stability. Thus, the thing that should be considered rst is how to measure system voltage stability or equivalently proximity to the static voltage instability point in the current state. To obtain this proximity information, diverse voltage stability indices have been proposed [17, 18]. Voltage stability indices can be mainly classied into two types. Type I is the given state-based voltage stability index, which is evaluated by comparing a physical parameter of the current point and the theoretically calculated critical value, so it is not necessary to obtain the real critical point. L indicator and Z index are included in Type I. Type II is the large deviation-based index, which can be determined by tracking a certain parameter from the base point to the voltage instability point. Active power and reactive power margins are typical indices in Type II, and they are widely used in voltage stability analysis. Formulations for VSCOPF apply one of these voltage stability indices, and the applied index can be incorporated implicitly or explicitly in the objective function or constraints [19]. Assuming that the incorporated index can be explicitly represented in the constraints, VSCOPF can be expressed in a compact form as follows: min s:t: f x g x 0 hmin  hx  hmax vsmin  vsx 1

VSCOPF to a real system, therefore, the threshold value should be adequately decided in advance, from simulation results. In the case that one of the Type II indices is employed for VSCOPF implementation, more complicated formulations are necessary. When the active power margin index is directly applied for the voltage stability constraint, at least the two solutions of the base case and the maximum point should be taken into account, as described in [8]. However, in order to incorporate the network equations of the two points, the problem dimension becomes twice that of the conventional OPF. To maintain the problem size, a simpler constraint needs to be included. In this paper, it is assumed that, before using VSCOPF for control strategies as remedial actions, voltage stability assessment (VSA) is performed, including sensitivity analysis. Thus, a VSCOPF has been developed, which fully utilises the information from VSA such as active power margin and margin sensitivity. In the formulation of the VSCOPF, a margin enhancement constraint (MEC), constructed with the margin sensitivity, is incorporated to obtain the required margin increase from the current margin. The problem size of the VSCOPF is the same as that used in conventional OPFs. This is one approximate method for enhancement of active power margins; therefore, nonlinearity of margin enhancement needs to be reected in the process of nding solutions. 3 Formulations and solution procedure

This paper uses a simplied voltage stability constraint as mentioned above, and it is not difcult incorporating the constraint in the conventional OPF formulation. This Section describes the main formulations in the proposed VSCOPF and the solution procedure.

3.1

Objective function

The objective function in this paper is minimisation of reactive generation control from the base case as follows: 2 X  k 0 Wui ui ui 2 f x
i2Su

where f( ), g( ), h( ) and vs( ) represent the objective function, the network equation functions, the inequality constraint functions for operational limits, and the voltage stability constraint function, respectively. In (1), hmin and hmax correspond to lower and upper limits of h( ), and vsmin is the lower limit of the margin constraint. For the voltage stability constraint function, vs( ), one of the voltage stability indices is adopted, and its lower limit, vsmin, is a threshold value of the index to maintain system voltage stability. If one of the indices in Type I is chosen to be used for the voltage stability constraint, it is not that complicated to implement the corresponding VSCOPF. The difcult part of the implementation is to calculate the monitored parameter at the theoretically critical point, and deciding the threshold value of the voltage stability index as the lower limit of the constraint. For application of this type of
962

where Su represents a set of the selected control variables, Wui represents the weighting factor for control variable ui, and the superscripts (0) and (k) indicate the base case and the kth iteration in the procedure of the VSCOPF. In this paper, shunt capacitance and tap ratio are regarded as control variables for reactive power dispatch to enhance voltage stability margin. Changing generator terminal voltages is excluded in control variables because it is not usually accepted in real applications.

3.2

Margin enhancement constraints (MEC)

The proposed VSCOPF formulation incorporates MEC to provide control strategies for improving system voltage stability. An MEC is constructed with margin sensitivity, proposed in [6], which is obtained at the margin boundary. The boundary can be determined with respect to a specied direction of stress, and margin sensitivity is calculated based on the left eigenvector corresponding to the minimum eigenvalue of the power ow Jacobian at the boundary. Using MCPF [20], in the paper, a path of power ow solutions is traced in the direction of stressing the system, to capture the margin boundary point. At this point, the left eigenvector of the minimum eigenvalue is calculated with the power method of the power ow Jacobian, and margin sensitivity with respect to each control parameter is
IEE Proc.-Gener. Transm. Distrib., Vol. 152, No. 6, November 2005

obtained. A description of MCPF and margin sensitivity calculation is included in the next Subsections. Using margin sensitivity information, MEC at the given state is constructed as follows:   X k 0 SM ;ui ui ui 3 DM req 
i2Su

repeated until MW-difference of the current solution from the previous one is within the given tolerance.

3.4

Margin sensitivity

where DM denotes the required margin enhancement from the current margin, and SM,ui stands for margin k 0 sensitivity of control variable ui. In (3), ui and ui have the same denitions used in (2). In VSCOPF, the MEC is placed on the position of voltage stability constraint in (1).

req

For a system with n buses, the power ow equations including the generation shifting parameter, m, can be expressed compactly as follows: gx; m; p 0 8 where x is the vector containing all the state variables and p is a scalar parameter. At an operating point, the tangent vector, dx d m dpT can be determined from the following formula: 9 gx dx gm d m gp dp 0 where gx, gm, and gp denote Jacobian matrices of g( ) with respect to x, m, and p, respectively. When the operating point is on the margin boundary, one eigenvalue of the power ow Jacobian, gx, is almost zero, so its singularity causes numerical difculty in power ow calculation at and near the margin boundary. Let vT o be the left eigenvector of the zero eigenvalue, and pre-multiply it to the left and right hand sides of (9). Then the rst element can be eliminated due to the denition of the zero left eigenvector. From the remainder, sensitivity with respect to change of an independent parameter p can be obtained as follows: v T gp dm o dp vT o gm 10

3.3 Modied continuation power ow (MCPF)

MCPF is for determining voltage stability margin of active power ow on a set of main transmission lines connecting the study region and others. It is also applicable for obtaining sensitivity information that is used for construction of MEC. A set of active and reactive power ow equations at bus i of an n-bus system can be expressed as follows: 4 PTi d; V PGi PLi0 0 QTi d; V QGi QLi0 0 5

where vectors d and V correspond to bus voltage angle and bus voltage magnitude, respectively, and the subscripts T, G, and L denote injection, generation and load, respectively. To implement the scenario for increasing ow, a parameter, m, representing generation shift is incorporated into (4). Here, it is assumed that region B is the study region, and that active power is injected from region A to B through the interface lines. The PGi, active power generation in region A and region B, can be determined as follows. For region A in which generation increases, 6 PGi PGio kGAi DPGB; total i 2 SA For region B in which generation increases, PGi PGio mkGBi PGBo;total X mkGBi PGBo;total DPGB; total
i2SB

Finally, margin sensitivity, which is indeed sensitivity of the generation shifting margin with respect to p, is calculated by multiplying PGBo,total to (10) as follows: SM ; p PGBo; total vT o gp vT o gm 11

3.5 Application of nonlinear interior point method (NIPM)

For solution of preventive and corrective OPF, an NIPM is used. NIPM is one of the Newton-type methods for obtaining a solution satisfying the KKT (Karush-KuhnTucker) rst-order necessary optimality condition. NIPM rst introduces slack variables (Z0) to transform inequality constraints into equality ones; then, it establishes the Lagrangian function including the objective function, the functions corresponding to the total equality constraints, and a scalar log barrier function with the slack variables. When applying the optimisation formulation of (1), the Lagrangian function can be described as follows: Lx; shL ; shU ; svsL ; l; phL ; phU ; pvsL ; Z f x lT gx phL hx shL hmin phU hx shU hmax pvsL hx svsL vsmin 12 ! X X X lnshLi lnshUi lnsvsLj Z
i i j

i 2 SB 7

where the following notation is used: original active power generation at bus i PGio original total generation in region B PGBo,total total generation decrease in region B DPGB,total fraction of generation increase at bus i in kGai region A fraction of generation decrease at bus i in kGbi region B SA set of generators in region A SB set of generators in region B. To trace the path of solutions starting from the base case solution with respect to change of the generator shifting parameter, m, MCPF applies the locally parameterised continuation method, so it consists of predictor and corrector. The predictor determines the initial guess of the next solution with the tangent vector obtained from the current solution. The corrector calculates the next solution from the initial guess using the Newton-Raphson method. In this paper, the purposes of MCPF are to identify system margins and to obtain margin sensitivities. For the latter purpose, capturing the precise boundary point is important. During the continuation, MCPF checks whether the nose point has been passed; if so, the solution path is traced back with a small step length; then, this procedure is
IEE Proc.-Gener. Transm. Distrib., Vol. 152, No. 6, November 2005

where shL shU svsL l pHl phU the lower slack variable vectors for hminrh(x) the upper slack variable vectors for h(x)rhmax the lower slack variable vector for vsminrvs(x) Lagrangian multiplier vector for g(x) 0 Lagrangian multiplier vector for h(x)shL hmin 0 Lagrangian multiplier vector for h(x)+shU hmax 0
963

pvsL Z

Lagrangian multiplier vector for vs(x)svsL vsmin 0 the barrier parameter.

j1

select a state from the contingency set (Cj : j = 1,...,m) jj+1 evaluate voltage stability margin (MCPF)

Next, the Newton method is applied from an initial guess to solve the KKT rst-order necessary condition P of (12). Because of the log barrier function, Z lnshLi P P lnshUi lnsvsLj , the solution is determined inside the inequality constraints, and the starting point should also be inside the inequality constraints: Step 1: Initialise the primal and dual variables. Step 2: Compute the complementary gap, GC, and power ow residuals. If the gap and the residuals are less than the given tolerance, then stop:
T T GC pT hL shL phU shU pvsL svsL

select the next contingency

no margin satisfied ? no construct MEC (margin sensitivity) yes end of contingency ? yes end

Step 3: Set the barrier parameter, Z, with the following equation: GC Zs 2nC where nC is the number of constraints and s is centering parameter. Step 4: Calculate the correction vectors of the primal and dual variables, solving the KKT condition with the Newton method. Step 5: Determine primal and dual step length without linearly violating the constraints. Step 6: Update the primal and dual variables and go to Step 2. Incorporating the MEC constraints into the formulation of optimisation changes the structure of Hessian matrix from that of the conventional OPF, and makes the solution matrix less sparse. In the MEC of (3), there is linkage between the selected control variables. During the calculation of Newton-type optimisation, this linkage implicates the existence of the virtual lines connecting the buses of the selected control variables, regarding ll-in elements. When a large number of control variables are linked by MEC, to reduce the number of ll-in elements, a bus-ordering scheme needs to be applied considering the network topology as well as the virtual lines.
Fig. 1

perform VSCOPF for C j

Flowchart of solution procedure

Step 3: If the margin does not satisfy the given margin criterion, go to Step 5; otherwise, go to Step 4. Step 4: Select the next contingency and go to Step 2. If the program examines all the contingencies, stop. Step 5: Obtain sensitivity information with (11) and construct the MEC with the sensitivity and margin requirement. Step 6: Execute VSCOPF with the MEC for the control strategy. Step 7: Perform MCPF for verication of the control strategy. If the margin improvement is not sufcient, go to Step 5; otherwise, go to Step 4.

Case study

3.6

Solution procedure

This Subsection describes the overall solution procedure of the proposed algorithm. In this procedure, it is assumed that a set of credible contingencies is provided on a base case with a specic load and generation pattern. From the contingency list, each contingency is examined one by one in terms of voltage stability, performing MCPF. For contingencies with insufcient margins, control strategies are proposed with VSCOPF including the MEC of (3). Since the MEC uses linear margin sensitivities, the real margin improvement with the given control strategy might be less than expected. Thus, MCPF needs to be executed again for verication of the control strategy. If the real margin is not enough, the MEC is reconstructed with the changed lower limit, DMreq in (3), and the VSCOPF is reapplied. In the algorithm, MCPF and VSCOPF are repeatedly performed, until the required margin enhancement is reached. The solution procedure can be described as follows, as in Fig. 1: Step 1: Select one from the given set of contingencies. Step 2: Apply the current contingency to the base case and perform MCPF for obtaining the interface ow margin.
964

This Section provides the numerical results with the modied IEEE 118-bus test system to illustrate the method proposed in this paper. Figure 2 shows a one-line diagram of the system divided into two regions. In Fig. 2, region B is the study region, so when applying MCPF to obtain generation shifting margin or interface ow margin, generation in region B is decreased, while that in region A is increased by the same amount of generation decrease in region B. This direction of generation shift deteriorates system voltage instability, and there exists a limitation of generation shift according to the given state. Table 1 gives the interface lines in the system and their active power ows in the normal state. For this purpose, demand and generation of each region have been modied. This study applies line outages and performs MCPF to identify severe contingencies in terms of interface ow margin, which is by denition active power ow difference from the base point to the maximum point in an fV curve constructed by MCPF. MCPF mainly changes interface ow on a set of the selected lines by generation shift from region B to region A. Table 2 gives the two worst contingencies that have least interface ow margins, and
IEE Proc.-Gener. Transm. Distrib., Vol. 152, No. 6, November 2005

1 3 G 4

2 region B 12 G 11 14 G 5 6 16 17 G 8 9 10 G 27 G 29 G 28 114 G 32 115 G 26 G 25 23 21 22 31 G 113 30 20 24 G G 74 75 G 72 G 7 15 13 G G 19 18 G 73 71 35 G 34 G 117 33 39 40

G 41

G 42

53

G 54

56

G 55

59 G

52 37 43 36 38 G 46 45 44

57

58 63 51 60 64

50 48 61

G 47 G 69 G 68 66 G 62 G 65 G 79 G 118 76 78 80 97 G 96 95 93 94 G 100 108 G 102 91 G 101 103 G 111 110 109 G G 112 G 104 G G 105 107 81 G 98 77 82 G 99 106 region A 49 67

70 G

G 116

84 85 G 87 86 G

83 G 92 G 90

88

89

Fig. 2

One-line diagram of IEEE 118-bus system

Table 1: Interface lines in this study and their initial ows (Case I)
Number From (region A) 33 34 38 24 To (region B) 15 19 30 23 Initial power ow, MW 13.7 25.1 32.4 28.4
V1, p.u.

0.98 0.96 0.94 0.92

1 2 3 4

0.90 0.88 0.86 0.84 0.82 normal contingency 1 contingency 2 0 200 400 600 interface flow, MW 800 1000

Table 2: Interface ow margins of two worst contingencies (Case I)

Contingency 1 2 Outage line 8-5 30-38 Interface ow margin 206.9 MW 563.7 MW

Fig. 3

fV curves of normal and two worst contingencies (Case I)

Fig. 3 shows fV curves of bus 1 in the normal and two worst cases. Note that other contingencies have almost the same interface ow margins as the normal case has. Assume that in the given voltage criteria, interface ow margin in each contingent state should be greater than 300 MW to maintain system voltage security. For contingency 1, then, a control strategy should be established. In [15], a preventive control strategy is proposed with dispatch of reactive power generation in the normal state. However,
IEE Proc.-Gener. Transm. Distrib., Vol. 152, No. 6, November 2005

control in the normal state considering multi-contingencies might be limited by operational limits in some cases; in addition, corrective control is usually considered as a more economical measure in system operation. The proposed algorithm with the VSCOPF is designed for determination of control after a severe contingency to enhance the interface ow margin. After identication of severe contingencies that require corrective control, the solution framework calculates margin sensitivity at the maximum point of each fV curve. In this study, bus shunt capacitance at each bus is mainly regarded as a control parameter, so margin sensitivity with respect to the parameter is calculated using (11). Table 3 shows the
965

margin sensitivity information at contingency 1. From this Table, it can be seen that most of the top 20 sensitive buses are in the upper part of Region B. To improve the interface ow margin at contingency 1, in this study, we selected the shunt capacitances at the top ten sensitive buses as the control variables in the VSCOPF with an MEC constructed as shown in (3). As the required interface ow margin is 300 MW, the required margin enhancement, DMreq, needs to be set to 93.1 ( 300206.9). Here, the margin sensitivities in Table 3 are not exactly sensitivities of interface ow margin, but those of generation shift, as described in (11). However, the system parameter to change active power ow on the selected interface lines is generation shift in MCPF; in addition, if increase in transmission loss can be neglected due to generation shift, sensitivity of generation shift can be applicable for construction of the MEC. In the objective function of (2), the weighting factors of all the control variables are set to 1000 for this case study. Since this VSCOPF is a reactive power dispatch problem, only voltage limits need to be incorporated in the formulation of (1) as the operational constraints. The upper limits of the constraints are set to 1.05 p.u. and the lower limits are set to 0.90 p.u. Then, the proposed VSCOPF is executed to determine the corrective control strategy for bus shunt capacitances. The MEC in the VSCOPF is a linear constraint using linear sensitivity; thus, after execution of the VSCOPF, real margin enhancement needs to be re-evaluated. Figure 4 shows the fV curves of the worst case at each solution step
Table 3: Margin sensitivity with respect to bus shunt capacitance at contingency 1 (Case I)
Rank Bus number 16 14 13 117 12 2 11 1 3 7 Sensitivity Rank Bus number 6 5 4 15 19 33 18 20 17 21 Sensitivity

of re-evaluating interface ow margin with MCPF until the required margin is reached. Table 4 presents interface ow margin and total amount of shunt capacitance to be committed at each solution step. After the rst execution of VSCOPF with DMreq of 93.1 as the lower limit of the MEC, the real margin enhancement is increased by 46.8 MW. Thus, in the second solution step, the lower limit of the MEC is set as follows:  2  req2 req1 DM 13 DM real DM where the superscripts (1) and (2) corresponding to the rst and second solution steps, respectively, and DMreal stands for the real margin enhancement re-evaluated using MCPF. In (13), DMreq(2) is decided considering the ratio of DMreal to DMreq(1). In this case, DMreq(2) is set to 185.2. After the second execution of VSCOPF, the interface ow margin is 299.3 MW, which is very close to the margin requirement. In the third solution step, DMreq is set to 194.4 using the following formula:   14 DM req3 k DM req2 DM req1 DM req1 where k is the margin enhancement correction multiplier. In the third solution step, it is set to 1.1. In the MEC formulation, margin sensitivity obtained at the initial condition in Table 2 is applied in order that a reference for selection of DMreq is required. In this simulation, signicant change in ranking of sensitive buses is not experienced. Table 5 shows margin sensitivities of the top ten buses at each solution step. In Table 5, there is a minor ranking change between bus 3 and 7. However, as the total amount of reactive compensation increases, margin sensitivities of the ten buses tend to decrease. This indicates that effectiveness of control with shunt capacitance is

1 2 3 4 5 6 7 8 9 10

5.897 5.720 5.711 5.313 5.245 5.220 5.102 5.030 4.935 4.935

11 12 13 14 15 16 17 18 19 20

4.753 4.459 4.311 3.533 2.911 2.702 2.517 2.381 2.113 1.968 Interface ow margin Total shunt capacitance Solution step Initial 206.9 MW 0 First 253.7 MW Second 299.3 MW Third 303.7 MW

Table 4: Interface ow margin and total shunt capacitance committed at each solution step (Case I)

17.46 MVAr 34.75 MVAr 36.46 MVAr

0.94 0.92 0.90 V1 , p.u. 0.88 0.86 0.84 0.82 150 200 250 300 350 400 450 500 550 interface flow, MW initial first solution step second solution step third solution step

Table 5: Top ten sensitive buses and their sensitivities at each step (Case I)
Bus number 16 14 13 117 12 2 11 1 3 7 Margin sensitivity Initial 5.897 5.720 5.711 5.313 5.245 5.220 5.102 5.030 4.935 4.935 First 4.469 4.335 4.328 4.034 3.976 3.958 3.869 3.815 3.744 3.745 Second 4.108 3.983 3.976 3.709 3.652 3.635 3.554 3.505 3.439 3.440 Third 4.074 3.950 3.943 3.679 3.621 3.604 3.524 3.476 3.411 3.411

Fig. 4 fV curves before and after VSCOPF execution (Case I, shunt capacitor)
966

IEE Proc.-Gener. Transm. Distrib., Vol. 152, No. 6, November 2005

reduced with a certain shape of a saturated function. In some severe cases, the margin is rather decreased even though shunt capacitance is more applied within a certain range. Therefore, using too large a value for k in (13), when control variables are shunt capacitances, should be carefully considered. In the authors experience, a reasonable range for k is 1.11.3. In this study, the system with less interface ow margin (Case II) is investigated and a control strategy is suggested using the proposed algorithm. With the change of generation pattern, the system is more stressed so that the interface ow level is changed to 242.8 MW. In Case II, the worst contingency is also outage of line 8-5 (contingency 1) and the interface ow margin of the contingency is 56.8 MW. Since the 300 MW criterion is not satised with this margin, a control strategy is needed. Margin sensitivity with respect to shunt capacitance for contingency 1 is given in Table 6, and shunt capacitances at the top ten sensitive buses in Table 6 are chosen as control variables. Then an MEC is constructed, and VSCOPF and MCPF are performed until the required margin enhancement is achieved. In Table 7, interface ow margin and total amount of shunt capacitance in each step are given. Figure 5 shows the fV curves of bus 1 of the solution steps. In the proposed method, tap ratios can be also considered as control variables, and they can be included for the MEC formulation. For Case II, margin sensitivity with respect to each tap ratio at the margin boundary is given in Table 8. The absolute value of the sensitivity with respect to the tap ratio of branch 30-17 is much larger than those of others, so the control of the tap ratio is very effective for improving the margin. In this simulation, the top ve sensitive taps are chosen for the control strategy, and the upper and lower limits of each tap ratio are assumed to be 1.1 and 0.9, respectively. The procedure with MCPF and VSCOPF is applied, after construction of the MEC. Figure 6 shows the fV curves of the solution steps. At the third step, all tap ratios reached their limits and the program stopped. The interface ow margin at the third step is 212.8 MW. For this case, the required margin enhancement cannot be accomplished. As seen in Fig. 6,
Table 6: Margin sensitivity with respect to bus shunt capacitance at contingency 1 (Case II)
Rank Bus number 1 2 3 4 5 16 14 13 117 12 Sensitivity Rank Bus number 6 7 8 9 10 2 11 1 3 7 Sensitivity

0.96 0.94 0.92 V1, p.u. 0.90 0.88 0.86 0.84 0.82 300 initial first solution step second solution step third solution step 350 400 450 500 550 interface flow, MW 600 650 700

Fig. 5 fV curves before and after VSCOPF execution (Case II, shunt capacitor)

Table 8: Margin sensitivities with respect to tap ratios (Case II)

Branch From 30 26 38 63 64 65 68 81 To 17 25 37 59 61 66 69 80 29.7037 4.4695 3.9751 2.1712 1.8071 0.9798 0.5487 0.5105 Margin sensitivity

0.94 0.92 0.90 V1, p.u. 0.88 0.86 0.84 0.82 0.80 300 initial first solution step second solution step third solution step

5.754 5.581 5.572 5.182 5.114

5.091 4.975 4.906 4.813 4.812

350

400 450 500 interface flow, MW

550

600

Fig. 6 fV curves before and after VSCOPF execution (Case II, tap ratio) Table 7: Interface ow margin and total shunt capacitance committed at each solution step (Case II)
Solution step Initial Interface ow margin Total shunt capacitance 56.8 MW 0 First 190.4 MW Second 278.5 MW Third 302.7 MW

49.34 MVAr 78.43 MVAr 94.47 MVAr

there is little margin improvement between the second step and third step. The reason is that the limits of the top three sensitive taps are binding at the second step and only low sensitive taps can move at the third solution step. Controlling only tap ratios cannot provide enough margin enhancement in many cases, because control quantity is limited and nonlinearity to margin is more serious. It is desirable utilising only high-sensitivity taps for control strategies.
967

IEE Proc.-Gener. Transm. Distrib., Vol. 152, No. 6, November 2005

Conclusions

This paper has presented a reactive optimal power ow incorporating a linear margin enhancement constraint, which is constructed with margin sensitivity. Using the linear margin enhancement constraint enables selective control. Selective control is desirable for corrective control determination, and is effective in reducing the number of linkages among control variables causing ll-in elements during factorisation of the Hessian matrix. A nonlinear interior point method is adopted to solve the optimisation problem, so the proposed OPF can cope with the case in which control is applied to the maximum point. Also, a practical method of effectively adjusting the lower limit of the margin enhancement constraint is explained. 6 Acknowledgment

This work was partly supported by MOCIE through the EIRC program with APSRC at Korea University, and partly supported by the Ministry of Science and Technology and the Korea Science and Engineering Foundation. 7 References

1 Taylor, C.W.: Power system voltage stability (McGraw-Hill, New York, USA, 1994) 2 Kundur, P.: Power system stability and control (McGraw-Hill, New York, USA, 1994) 3 Van Cutsem, T., and Vournas, C.: Voltage stability of electric power systems (Kluwer Academic Publishers, Boston, MA, USA, 1998) 4 Lee, B., and Ajjarapu, V.: Invariant subspace parametric sensitivity (ISPS) of structure preserving power models, IEEE Trans. Power Syst., 1996, 11, (2), pp. 845850 5 Dobson, I., and Liu, L.: Computing an optimum direction in control space to avoid saddle node bifurcation and voltage collapse in electrical power systems, IEEE Trans. Autom. Control, 1992, 37, (10), pp. 16161620 6 Feng, Z., Ajjarapu, V., and Maratukulam, D.J.: Comprehensive approach for preventive and corrective control to mitigate voltage collapse, IEEE Trans. Power Syst., 2000, 15, (2), pp. 791797

7 Wang, X., Ejebe, G.C., Tong, J., and Waight, J.G.: Preventive/ corrective control for voltage stability using direct interior point method, IEEE Trans. Power Syst., 1998, 13, (3), pp. 416423 8 Rosehart, W., Canizares, C., and Quintana, V.: Optimal power ow incorporating voltage collapse constraints. Proc. IEEE 1999 PES Summer Meeting, Edmonton, Canada, July 1999, Vol. 2, pp. 820825 9 Kim, S., Song, T.Y., Jeong, M.H., Lee, B., Moon, Y.H., Namkung, J.Y., and Jang, G.: Development of voltage stability constrained optimal power ow (VSCOPF). Proc. IEEE PES Summer Meeting, Vancouver, Canada, July 2001, Vol. 3, pp. 16641669 10 Lin, X., David, A.K., and Yu, C.W.: Reactive power optimization with voltage stability consideration in power market systems, IEE Proc., Gener. Transm. Distrib., 2003, 150, (3), pp. 305310 11 Kubokawa, J., Inoue, R., and Sasaki, H.: A solution of optimal power ow with voltage stability constraints. Proc. Int. Conf. Power System Technology (PowerCon), Perth, Australia, December 2000, Vol. 2, pp. 625630 12 Vaahedi, E., Mansour, Y., Fuchs, C., Granville, S., Latore, M.D.L., and Hamadanizadeh, H.: Dynamic security constrained optimal power ow/VAr planning, IEEE Trans. Power Syst., 2001, 16, (1), pp. 3843 13 Alemadi, N.A., and Schlueter, R.A.: Dispatching for preventive and corrective control of voltage collapse in a deregulated power system. Proc. Int. Symp. on Bulk Power Systems Dynamics and Control-V, Onomichi, Japan, August 2001 14 Capitanescu, F., and Van Cutsem, T.: Preventive control of voltage security margins: A multicontingency sensitivity-based approach, IEEE Trans. Power Syst., 2002, 17, (2), pp. 358364 15 Song, H., Lee, B., Kwon, S.-H., and Ajjarapu, V.: Reactive reservebased contingency constrained optimal power ow (RCCOPF) for enhancement of voltage stability margins, IEEE Trans. Power Syst., 2003, 18, (4), pp. 15381546 16 Menezes, T.V., da Silva, L.C.P., Affonso, C.M., and da Costa, V.F.: MVAR management on the pre-dispatch problem for improving voltage stability margin, IEE Proc., Gener. Transm. Distrib., 2004, 151, (6), pp. 665672 ! Task Force 38-02-11: Indices predicting voltage collapse 17 CIGRE ! Brochure, 1994) including dynamic phenomena (CIGRE 18 IEEE PES Power system stability subcommittee: Voltage stability assessment: concepts, practices and tools (IEEE Press, 2002) 19 Wang, Q., and Ajjarapu, V.: A critical review on preventive and corrective control against voltage collapse, Elec., Power Compon. Syst., 2001, 29, (12), pp. 11331144 20 Sang, H., Kim, S., Lee, B., Kwon, S.H., and Ajjarapu, P.: Determination of interface ow margin using the modied continuation power ow in voltage stability analysis, IEE Proc. Gener. Transm. Distrib., 2001, 148, (2), pp. 128132

968

IEE Proc.-Gener. Transm. Distrib., Vol. 152, No. 6, November 2005