Вы находитесь на странице: 1из 8

Paper accepted for presentation at the 2011 IEEE Trondheim PowerTech

Smart Grid Emergency Control Strategy for Load Tap Changers


Johan Fredrik Baalbergen, Student Member, IEEE, Madeleine Gibescu, Member, IEEE, and Lou van der Sluis, Senior Member, IEEE

AbstractIn this paper a strategy is introduced for emergency control of Load Tap Changers (LTCs) to prevent the system from voltage collapse. The method is based on the fact that the power consumed by constant current and constant impedance loads can be lowered by decreasing the secondary voltage of the LTC. This idea is used to develop a method of indirect load shedding. This method calculates, based on a model of the connected load, the voltage set-point of the secondary side of the LTC for a given amount of load that needs to be shed. With simulations it is demonstrated that a LTC can be applied for load relief using the proposed method. A sensitivity analysis is done to study the effect of an error in the load model parameters. Furthermore it is proven that the method can be used to prevent the power system from a typical voltage instability. Finally a proposal for the implementation in a smart grid multi-agent system is presented. Index TermsSmart grids, power system stability, load management.

I. I NTRODUCTION OAD Tap Changers (LTCs) control the voltage at the load side of the power system [1]. They aim to keep the load voltages near nominal value, and consequently, for constant impedance and constant current loads, LTCs will restore the load to nominal. This behavior is benecial under normal operating conditions. When the system is, however, close to voltage instability, this load restoring property might contribute to a voltage collapse [2], [3]. To prevent the load restoring behavior of the LTC during these situations, the following emergency controls exist: tap blocking, tap locking, tap reversing, and distribution side voltage set-point reduction. A comparison of these methods is given in [4], [5]. With LTC tap blocking, as soon as voltage instability is detected, the tap is blocked [6], [7]. In the method of [6] as soon as voltage instability is detected, all LTCs are blocked. To be able to restore the load to some extent, the LTCs with the smallest impact are released. In [7] a method is proposed where the voltage instability detection method determines the appropriate LTCs which should be blocked. Tap locking is a special form of tap blocking. As soon as voltage instability is detected, the LTC brings its tap to a specic tap position and then blocks [4], [5]. With the tap-reversing method, as soon as voltage instability is detected, the LTC starts to control the primary voltage
This work was nancially supported by SenterNovem under Grant EOSLT06017. J.F. Baalbergen, M. Gibescu, and L. van der Sluis are with the Faculty of Electrical Engineering, Mathematics and Computer Science, Delft University of Technology, the Netherlands (phone: +31 15 27 86236; e-mail: j.f.baalbergen@tudelft.nl.)

instead of the secondary voltage [8]. The logic of the LTC is reversed in this case. If voltage set-point reduction is applied, the reference voltage of the LTC is lowered. For voltage sensitive loads, this subsequently lowers the load. In [9] a method is introduced where the voltage set-point is lowered to a xed minimum value. In [10] a closed-loop method is outlined where, upon detection of a low secondary voltage, the voltage set-point is reduced by 5 %. In addition to this, the tap is increased by two steps at a time to speed up the restoration. The emergency controls of the LTC can be combined with other countermeasures for voltage instability as generator voltage control and load shedding. Examples can be found in [11][17]. The focus in these papers is the coordination of the different control actions. The method proposed in this paper is to use the inverse of the load restoring property of the LTC: where for voltage sensitive loads a higher voltage leads to a higher load, a lower voltage would lead to a lower load. Consequently, loads can be shed indirectly by lowering the voltage at the connection point. The method introduced in this paper is to adapt the voltage set-point of the LTC to obtain a certain amount of virtual load shedding. The benet of this method compared to the existing voltage set-point reduction methods [9], [10] is the fact that with this method the effectiveness of a given control action can be evaluated ahead of time and subsequently can be applied in a smart grid multi-agent system. In this paper the proposed indirect load shedding method will be outlined and a proof of concept will be given. The method is based on an assumed load model. A sensitivity study will be given to determine the sensitivity of the method to an error in the parameters of the load model. It will also be proven that with the proposed method a typical voltage instability scenario can be prevented. Although the method presented in this paper can also be implemented stand-alone, it is the objective of future steps in this research to implement it in an integrated multi-agent smart grid environment. II. V IRTUAL LOAD SHEDDING METHOD The description of the virtual load shedding method will be based on the simple radial power system of Fig. 1. This power system consists of a load area which is fed by a generator via a LTC and a transmission line. A voltage stability problem occurs when the load draws more power from the grid as can be transferred through the transmission line [2], [3]. The idea of the virtual load shedding method is to reduce the

978-1-4244-8417-1/11/$26.00 2011

(G)

(1)

(2) Load

principle be applied periodically to update the parameters as the composition of the load changes. The set-point for the load power can now be calculated with: Pref = Pload (0) PLT C (4)

Grid
Fig. 1. Simple power system.

In this equation Pload (0) is the initial load power. For obtaining the set-point for the voltage Uref , Pref is lled in for Pload in equation (3). The equation can be rewritten as: aP Uref U0
2

power consumption by lowering the secondary voltage of the transformer when such a situation occurs. To determine the amount of load that should be virtually shed the Maximum Loadability Index M LI as proposed in [18] and extended in [19] will be used. The M LI gives for a given transmission line a measure of the relative distance of the current operating point to the point of maximum loadability, which is often considered as stability limit. If the M LI is larger than one, the system is voltage stable; if the M LI equals one, the system is at the stability limit and when the M LI is smaller than one the system is voltage unstable. As suggested in [18], when the M LI is smaller than one, the amount of load that needs to be shed in a certain area to restore voltage stability can be calculated with: PM LI = Pline (M LI 1) 0 M LI < 1 M LI 1 (1)

+ bP

Uref U0

+ cP

Pref =0 P0

(5)

Solving for Uref gives two solutions: bP b2 P 4aP cP 2aP


Pref P0

Uref,1 , Uref,2 = U0

(6)

From these two solutions the highest voltage value should be taken as set-point voltage: Uref = max {Uref,1 , Uref,2 } (7)

In this equation PM LI is the amount of load that needs to be shed to obtain stability and Pline is the MW loading at the receiving end for the current operating point . To prevent the system from an off-set error and to make sure that the emergency control signal is xed from the time that the emergency controller restored the system to a stable operating point, a PI-action is added in this work. The amount of load that should be indirectly shed by the LTC becomes: PLT C = k P + kI 1 s PM LI (2)

If the secondary voltage of the LTC has to stay within certain limits, this can be taken into account. If the voltage is only allowed to deviate Umax from the nominal voltage, the new reference value for the voltage becomes: Uref U0 Umax U0 Uref Umax U0 Uref > Umax

Uref,lim =

(8)

It should be noted that the value of such a limitation is doubtful for the described emergency control of the LTC. The indirect load shedding is performed to avoid a collapse. When, due to the limitation in the minimal secondary voltage, the emergency control is not able to implement the right control action, the system might still collapse. III. P ROOF OF CONCEPT VIRTUAL LOAD SHEDDING
METHOD

In this equation kP is the amount of proportional action, kI the amount of integral action and 1 s the integral operator. With the proposed virtual load shedding the load is shed by lowering the secondary voltage of the LTC. For the load a static ZIP model (load build up from a constant impedance Z, constant current I and a constant power P part) is assumed: Pload = P0 aP U U0
2

+ bP

U U0

+ cP

(3)

In this equation Pload is the real power consumed by the load; P0 the real power consumed at nominal voltage U0 ; aP , bP and cP the fractions of real power for respectively the constant impedance, constant current and constant power parts of the load model; and U the actual voltage. For the parameters it holds that: aP + bP + cP = 1. To estimate these parameters from measurements several methods are proposed in literature. For instance in [20] a method is outlined based on least squares optimization and in [21] another method is introduced based on a hybrid learning algorithm. These methods can in

In this section a proof of concept of the virtual load shedding method will be given. It will be determined whether, when a certain amount of load needs to be virtually shed, the load is indeed reduced by that amount. Note that in this section the part of the method given by equations (4) up and until (8) is tested. The proof of concept will be based on simulations done with the SimPowerSystems toolbox of Matlab/Simulink on the circuit of Fig. 1. The generator and line are modeled as a Th evenin equivalent with a voltage of 115 kV and an impedance Z = 44.1 + j 132.3 m. For the LTC and the load the parameters of the voltage stability model of [22] are used. This load model is a ZIP load and has a power factor of 0.96. For the real power load part the parameters are aP = 0.5, bP = 0, cP = 0.5, P0 = 3384 MW, and U0 = 1.00152 pu. The LTC is controlled with the proposed indirect load shedding control. It is assumed that the parameters of the load model used for determining the voltage set-point (equation (6))

1 0.98 0.96 0 2 0 -2 -4 -6 0 50 50

Primary Voltage (1) Secondary Voltage (2) Reference Secondary Voltage

Relative deviation [%]

Voltage [pu]

-50

100 Time [s]

150

200

Load shedded relative to load to be shed Theoretical maximum deviation Limits LTC

-100 0 0.02 0.04 0.06 Load to be shed [pu] 0.08 0.1

Tap position

Fig. 3. Relative error in the amount of load that is shed with the new LTC control as function of the amount of load that needs to be shed.

100 Time [s]

150

200

The relative error/deviation (solid line) is calculated with: Pshedded (Pto shed ) = Pshedded Pto shed 100% Pto shed (9)

Load power [pu]

1.02 1 0.98 0.96 0.94 0.92 0

50

100 Time [s]

150

200

Fig. 2. Voltages (upper graph), tap setting (middle graph) and load power (lower graph) if at t = 50 seconds 4 % of the load needs to be shed.

are equal to the actual load parameters. At t = 50 s the controller will be assigned to virtually shed 4 % of the load. The result is shown in Fig. 2. In the upper graph the LTCs primary voltage (solid line), secondary voltage (dashed line) and the reference for the secondary voltage (dotted line) are given. Before t = 50 s no load needs to be shed. The primary voltage is lower than one per unit because of line drop losses. The reference for the secondary voltage equals one. The actual secondary voltage is somewhat higher due to the dead-band introduced by the discrete control of the tap positions. After t = 50 s the indirect load shedding is started. The reference of the secondary voltage drops immediately. The actual secondary voltage is lowered in steps every 5 s until at t = 85 s the LTC has its tap at a position where the voltage is within the LTCs dead-band (0.0125 pu). The nal secondary voltage is slightly too high (by 0.03 %). During the time period 50 t 85 s the primary voltage increases a litle bit due to the fact that with lowering the secondary voltage, the current is also lowered and subsequently the voltage drop across the line. In the middle graph of Fig. 2 the tap position of the LTC is shown. As expected to lower the secondary voltage the tap position increases. In the lower graph of Fig. 2 the power consumption of the load is shown. From this graph the delayed load shedding behavior of the LTC is visible. The load that is nally shed is 0.04 pu. This equals the amount that should be shed. In Fig. 3 the error of the actual load that is shed is shown as function of the amount of load that needs to be shed (solid line) together with the theoretical maximum error (dashed line) and the LTCs limits (vertical dotted lines). This gure gives a detailed picture of the capability and limitations of the method.

The discontinuities in Pshedded (Pto shed ) are due to the discontinues nature of the tap position. The theoretical maximum error (dashed line) is given by the dead-band in the LTCs voltage control. This dead-band in the voltage will appear as a dead-band in the power to be shed. The voltage magnitude deviates due to this fact at most: 1 2 UDB . Differentiating the load model equation (equation (3)) and substituting the maximum voltage deviation gives the theoretical maximum error in the power: Pshedded (UDB ) = 1 UDB P0 2aP 2 U U0 + bP 100% (10)

As can be seen from Fig. 3 the error in the amount of load that is shed is constricted by the theoretical maximum error whenever the LTC operates within its limits. As can be expected the error is most of the time lower that the maximum error. This is because of the fact that the deviation in the secondary voltage does not always reach its maximum value 1 UDB ). (2 The minimum and maximum limits of the LTC (vertical dotted lines) limit the region for which the LTC can shed load. The minimum limit in the power that can be shed is related to the amount of power that can be shed by one tap change. The maximum limit is related to the maximum tap position. LTC load shedding control is only feasible within this region. For the simulations under consideration the error is negative due to the choice of the dead-band and the voltage step of the LTC. The LTC will change its tap to obtain the set-point voltage based on the dead-band: the tap is changed if the difference between the measured voltage Um and the reference voltage is larger than half the dead-band (|Um Uref | 1 2 UDB ). For these simulations the dead-band UDB is chosen to be twice the voltage step per tap Utap , which is according to [2] a typical value. In the virtual load shedding procedure the tap is changed from a lower tap position to a higher one. Because UDB = 2 Utap this implies that the voltage will never end in the lower part of the dead-band. This is illustrated for starting points in the higher and the lower part of the dead-band in Fig. 4. The voltage will be too high and the load that is shed is too low.

Utap

Uref UDB
Fig. 4. Illustration for changing the LTCs secondary voltage from a high voltage to the reference voltage (with UDB = 2 Utap ).

The fact that with the proposed method less power is shed than required will be counteracted by the integral action of the PI-controller in the M LI control loop. This integral action will counteract the offset error. Because the switching action of equation (1) ensures that when the system is stable PM LI does not become below one, the controller will settle itself to a point where more power than required is shed. The relative error is large when small amounts of power have to be shed. This can be explained by the fact that when the amount of load that needs to be shed is in the order of magnitude of the equivalent power dead-band of the LTC, this dead-band will have a relatively large inuence. IV. S ENSITIVITY ANALYSIS In this section the sensitivity of the load shedding method to the different load model parameters aP , bP , cP , U0 and P0 used for control is studied. Simulations are done with the circuit of Fig. 1. Furthermore simulations are carried out with varying values for Umax . In the previous section it was shown that in the ideal case the dead-band of the LTC is of major importance for the error between the load that needs to be shed and the load that is actually shed. To take into account the error for the ideal case, three different cases of load to be shed are compared (see Fig. 3):

0.0400 pu to be shed, which gives in the ideal case an error close to zero; 0.0425 pu to be shed, which gives in the ideal case an error halfway between zero and the theoretical maximum error; 0.0450 pu to be shed, which gives in the ideal case an error close to the theoretical maximum error.

The actual amount of load that is shed is determined and the error of the actual shedded load relative to the amount of load to be shed (equation (9)) is determined. The result for the sensitivity analysis is given in Fig. 5. The left hand side graphs give the sensitivity of the method to the load model parameters aP , bP and cP . During the sensitivity analysis the requirement that: aP + bP + cP = 1 is fullled. This means that one parameter is replaced by another one. The cases investigated are aP replaced by cP (upper left hand graph), aP replaced by bP (middle left hand graph) and cP replaced by bP (lower left hand graph). The relation between these parameters are respectively: cP = 1 aP , bP = 0.5 aP and bP = 0.5 cP . The lower x-axis of the graph gives the

parameter that is replaced and the upper x-axis gives the new parameter by which the old parameter is replaced. For the load model parameter aP which is replaced by cP when the deviation in aP is positive this means that the emergency control thinks that the load has a larger share of constant impedance part than it actually has. cP is on the contrary larger than expected implying that the constant power part of the load is larger than expected. This means that the voltage set-point that is calculated is too high and subsequently not enough load is indirectly shed. The opposite is also true: when the deviation in aP is negative the LTC emergency control expects a load with a smaller constant impedance part and the voltage set-point that is calculated is too low. In this case more load is indirectly shed than necessary. When the load model parameter aP is replaced by bP and the load model parameter cP is replaced by bP a similar effect can be seen. In these cases because of the fact that bP is involved, the deviation in aP respectively cP can only be negative because bP is zero for the actual load model. The increase in the relative error in the amount of load that is actually shed is smaller for these two cases than it was for the case that aP is replaced by cP . This is because the voltage dependency of a constant impedance load and a constant current load or a constant current load and a constant power load are more comparable than the voltage dependency of a constant impedance load and a constant power load. The right hand graphs give the sensitivity of the method to the load model parameters P0 (upper right hand graph) and U0 (middle right hand graph); and the constraint Umax (lower right hand graph). It can be seen that the indirect load shedding method is rather sensitive for deviations in both parameters P0 and U0 . These parameters should be determined carefully. It can be seen that as long as Umax > 0.04 pu, the fact that the maximum voltage deviation is limited does not introduce an extra error. The maximum allowable voltage deviation of 10 % given by the Dutch grid code (see [23]) will thus not introduce an extra error in the method. In Fig. 5 the discrete behavior of the LTC can be noticed for all cases: the deviation in the parameter does not inuence the relative error in the amount of load to be shed as long as it does not introduce a step in the tap setting. Furthermore, due to the fact that in the ideal case not enough power is shed, a deviation in a parameter which results in less power shed, is more detrimental than when the opposite occurs. A deviation in a parameter which results in more power shed can even be advantageous. This last statement is, of course, only true for the case when the deviation introduces a deviation in the tap position. V. P REVENTION OF VOLTAGE COLLAPSE The system used for demonstrating the voltage collapse prevention with the proposed indirect load shedding method is given in Fig. 6 ( [3], [22]). Note that in this section the full method (equations (1) up and until (8)) is tested. The system consists of a generating area and a mainly load area with a local generator. The two areas are connected to each other via a transmission corridor of ve parallel lines.

Deviation in cP 0.1 30 Relative deviation [%] 20 10 0 -10 -20 -30 -0.1 -0.05 0 0.05 Deviation in aP Deviation in bP 0.4 30 Relative deviation [%] 20 10 0 -10 -20 -30 -0.2 -0.15 -0.1 -0.05 Deviation in aP Deviation in bP 0.2 0 Relative deviation [%] -5 -10 -15 -20 -25 -100 -30 -0.2 -0.15 -0.1 -0.05 Deviation in cP 0 0 0.02 0.04 0.06 Umax [pu] 0.08 0.1 0.15 0.1 0.05 0 0 Relative deviation [%] -20 -40 -60 -80
0.0400 pu to be shed 0.0425 pu to be shed 0.0450 pu to be shed

0.05

-0.05

-0.1

100 50 0 -50 -100 -150 -0.1

0.1

Relative deviation [%]

-0.05 0 0.05 Deviation in P0 [pu]

0.1

0.3

0.2

0.1

100 50 0 -50 -100 -150 -0.1

Relative deviation [%]

-0.05 0 0.05 Deviation in U0 [pu]

0.1

Fig. 5. Sensitivity analysis of the indirect load shedding method for the different parameters used by the method. Three levels of load to be shed are used: 0.0400 pu (solid line), 0.0425 pu (dashed line) and 0.0450 pu (dotted line).

Bus 9 voltage [pu]

Tap position [pu]

1 0.8 0.6 0 20 40 60 80 100 120 t [s]

8 6 4 2 0 -2 0 1.2 20 40 60 80 100 120 t [s]

Load power [pu]

1 0.8 0.6 0
With Emergency Control Without Emergency Control

GEN3 eld current [pu]

Bus 6 voltage [pu]

4.5 4 3.5 3 2.5 0

20 40 60 80 100 120 t [s]

M LI

1 0.8 0.6 0 20 40 60 80 100 120 t [s]

1 0.8 0 20 40 60 80 100 120 t [s]

20 40 60 80 100 120 t [s]

Fig. 7.

Simulation of typical voltage instability with (solid line) and without (dashed line) the proposed emergency control.

GEN1

GEN3

1400 MVA 10

GEN2

2200 MVA

Fig. 6.

Simulation model for voltage instability scenario.

In the load area two load buses are present: bus 9 and bus 10. The load at bus 9 is a 50 % constant power and 50 % constant impedance load of 3384 MW and 971 MVar. This load is fed via a LTC with a tap size of 0.015625 pu and a dead-band of 0.03125 pu. The load at bus 10 is a constant power load of 3271 MW and 1015 MVar. GEN1 is an innite bus and GEN2 and GEN3 are classical synchronous generators equiped with governors, voltage regulators and power system stabilizers. The model of GEN3 also contains an Over-Excitation Limiter (OXL) for which the model with integral control of eld current is used [2]. For the remaining power system components standard models are used. At t = 5 s one of the transmission lines trips. Simulations are done for both cases: without emergency control and with the proposed emergency control. For the indirect load shedding method it is assumed that there is no error in the load model parameters and Umax is neglected. Furthermore the settings for the PI control loop are kP = 0.6 and kI = 0.4. The M LI is determined for the parallel equivalent of the transmission corridor and this M LI is used for the emergency control. The results for the simulations are given in Fig. 7. The upper left hand graph shows the bus 9 voltage, the upper middle graph the LTC tap position and the upper right hand graph

the load power consumed at bus 9. The lower left hand graph shows the bus 6 voltage, the lower middle graph the M LI of the transmission corridor and the lower right hand graph the eld current of generator GEN3. The result of the simulation without emergency control is given by the dashed lines of Fig. 7. At t = 5 s the bus 9 voltage decreases due to the trip of the line. Almost immediately the voltage increases again due to the action of the exciter of GEN3. The bus 9 voltage, however, does not restore to the reference value of the LTC and subsequently the LTC will decrease its tap in order to increase the secondary voltage. At rst the action of the LTC indeed leads to an increase in the bus 9 voltage. The maximum allowed eld current for the long term (3.06 pu) is, however, exceeded. After 30 s the OXL of GEN3 starts to limit the eld current. A decrease of the tap setting does not increase the bus 9 voltage anymore and even decreases it. This is because the reactive power injection of GEN3 is limited. Finally the voltage collapses. The M LI indicates the instability immediately after the trip of the line. By contrast, the results for the simulation for the case with the proposed emergency controller are given by the solid lines of Fig. 7. The rst conclusion is that the voltages do not collapse. The indirect load shedding starts immediately after the trip of the line. The LTCs tap position increases and the bus 9 voltage decreases. Subsequently the load power is also decreased. Note that the M LI quickly reaches a value above one. This is because of the switching action in equation (1) combined with the integral action of equation (2) which does not allow a decrease in PM LI . It can be seen that the indirect load shedding is benecial for the eld current of GEN3. At each decrease in the load power, the eld current decreases. The value for the eld current when at t 40 s the total required amount of load has been shed is, however, still too high. The OXL of GEN3 limits the eld current at t 60 s. The M LI decreases and the LTC emergency controller will decrease the reference for the secondary voltage again. More load is virtually shed by increasing the tap position and the system is kept stable.

VI. LTC LOAD SHEDDING IMPLEMENTATION IN SMART


GRID CONTROL

only allowed to deviate Umax from the nominal voltage, the total voltage step becomes: Ulim = U Umax U Umax U > Umax (12)

The proposed LTC emergency control with indirect load shedding is an integral part of the set of control actions available for the Hierarchical Agent Based Voltage Instability Prevention (HABVIP) system [24]. In this system each substation is controlled by a substation agent and each actuator is controlled by an actuator agent. Between the agents there is a hierarchical structure: actor agents are supervised by substation agents and between the substation agents the normal power system hierarchy based on voltage levels is followed. So an agent controlling a 50 kV substation is higher in rank than the agent of a underlying 10 kV substation. A substation can only communicate with other substations which are electrically connected to the substation and with actor agents of actuators which are directly connected at the substation. Voltage instability detection is done by the substation agents based on the M LI computation. As soon as a voltage instability is detected, all substation agents which detect this instability will cooperate and determine on the best control action to counteract the instability. This process is supervised by the highest substation agent in hierarchy that senses the problem. This agent will determine based on equation (1) the amount of load relief necessary in the area to restore voltage stability. Subsequently this agent will ask all agents lower in hierarchy what amount of load relief they can provide. The substations will aggregate the amount of load relief they can provide and communicate with substations higher in the hierarchy. This aggregation is based on the actuators directly connected to the substation and the aggregated amount communicated from substation agents lower in hierarchy. Based on the signals of the aggregated load relief all lower level substations can give, the highest substation in hierarchy will determine which amount the substations lower in hierarchy should give. The substation agents one level lower in hierarchy will determine for their actuators and the substations one level lower in hierarchy which entity has to provide which amount of load relief. In this way the higher level agents only have to determine an aggregated control action which is decomposed at the appropriate level. The indirect load shedding method described by equations (4) - (8) can be implemented in actor agents which control the LTCs. Based on the load relief ordered by the substation agent that supervises the LTC, the actuator agent will calculate the required voltage set-point. This new set-point will be used by the conventional control of the LTC. The actor agent controlling the LTC should also determine which amount of power it has available for load relief. This is dependent on the number of tap changes from the initial tap position r(0) to the maximum tap position rmax and the voltage step per tap (Utap ). From this, the maximum step in the secondary voltage can be calculated: U = (rmax r(0)) Utap (11)

From this, the minimum secondary voltage of the LTC can be found: Umin = U (0) Ulim (13)

In this equation U (0) is the initial load voltage. The value found for Umin with equation (13) is substituted for U in equation (3) and gives the power consumed by the load at minimum voltage Pmin . From this the available power for indirect load shedding by means of the LTC can be calculated: Pmax = Pload (0) Pmin (14)

This amount of power will be communicated to the supervising substation agent. This paper only focused on the new control strategy for the LTC. The total HABVIP controller should take more control options into account. For instance in [25] possibilities for intelligent control of decentralized generators to prevent the system from a voltage collapse are dened. These should be part of the set of possible control actions. Coordination of the agents and optimization of the control actions will be the task of the highest agent in hierarchy affected by the problem. Actuators will be grouped by impact level. Actuators with low negative impact (such as the increase in local generation) will get a higher priority than actuators with a high negative impact (such as load shedding). Coordination should also take place within these impact groups. For larger systems the time needed for communication might be a limiting factor: when the control is delayed too much, a collapse might already be unavoidable. In [19], however, it is proven that detection of voltage instability with the M LI gives enough time for implementing control actions. The exact optimization and coordination of the control actions is a subject for future research. VII. C ONCLUSION In this paper a new voltage instability emergency control method for LTCs is proposed. The method is based on the fact that the power consumed by constant current and constant impedance loads can be lowered by decreasing the secondary voltage of the LTC. This idea is used to develop a method of indirect load shedding. This method calculates the voltage set-point for the LTC based on the amount of load that needs to be indirectly shed. The amount of necessary load shedding is determined based on the Maximum Loadability Index. Based on simulations it is demonstrated that the method proposed can indeed be used to give load relief. Due to the dead-band of the LTC, with the proposed method less power is shed than required. The theoretical maximum value of this error can be calculated. Furthermore it is shown that with the proposed method a voltage instability in a typical test system can be prevented. With the help of a PI feedback loop the error

When the secondary voltage of the LTC has to stay within certain limits, this can be taken into account. If the voltage is

due to the dead-band in the voltage controller is counteracted. The proposed method uses a model for the load to determine the voltage set-point. The method is insensitive to an error in the model parameters as long as the error does not imply another tap position for the LTC. The proposed method can be implemented in an LTC actor agent as part of the Hierarchical Agent Based Voltage Instability Prevention system. The coordination between the different control actions and the exact multi-agent system is subject for further research. ACKNOWLEDGMENT This research is part of the Dynamic State-Estimation and Voltage Stability of Transmission and Distribution Grids with a large share of Distributed Generation Capacity-project. This is a joint project of TU Delft, ECN, Alliander, KEMA, Stedin and Delta N.V. In this project the consequences on voltage stability of DG are being investigated and appropriate control strategies are designed. Information about the project can be found at: www.devs-project.nl. The authors would like to thank the members of this consortium for the valuable discussions. R EFERENCES
[1] P. H. Schavemaker and L. v. d. Sluis, Electrical Power System Essentials. Chichester: John Wiley & Sons, 2008. [2] T. van Cutsem and C. Vournas, Voltage Stability of Electric Power Systems. Boston: Kluwer, 1998. [3] C. W. Taylor, Power System Voltage Stability. New York: McGraw-Hill, 1994. [4] C. Vournas, A. Metsiou, M. Kotlida, V. Nikolaidis, and M. Karystianos, Comparison and combination of emergency control methods for voltage stability, in Proc. IEEE Power Eng. Soc. General Meeting, vol. 2, 2004, pp. 1799 1804. [5] T. Van Cutsem and C. Vournas, Emergency voltage stability controls: an overview, in Proc. IEEE Power Eng. Soc. General Meeting, June 2007. [6] T. Van Cutsem, An approach to corrective control of voltage instability using simulation and sensitivity, IEEE Trans. Power Syst., vol. 10, no. 2, pp. 616 622, May 1995. [7] C. Vournas and G. Manos, Emergency tap-blocking to prevent voltage collapse, in Proc. PowerTech, vol. 2, September 2001. [8] B. Otomega, V. Sermanson, and T. Van Cutsem, Reverse-logic control of load tap changers in emergency voltage conditions, in Proc. IEEE PowerTech, vol. 1, June 2003. [9] R. Nielsen and N. Markushevich, Dynamic system load control through use of optimal voltage and var control, in Proc. IEEE Ind. App. Dynamic Modeling Control Applications for Industry Workshop, 1998, pp. 29 32. [10] C. Vournas and N. Sakellaridis, Problems and solutions for local identication of voltage instability and emergency control, in Proc. IEEE Power Eng. Soc. General Meeting, Jul. 2008. [11] T. Overbye, Computation of a practical method to restore power ow solvability, IEEE Trans. Power Syst., vol. 10, no. 1, pp. 280 287, feb. 1995. [12] S. Granville, J. Mello, and A. Melo, Application of interior point methods to power ow unsolvability, IEEE Trans. Power Syst., vol. 11, no. 2, pp. 1096 1103, may 1996. [13] D. Popovic, V. Levi, and Z. Gorecan, Co-ordination of emergency secondary-voltage control and load shedding to prevent voltage instability, Generation, Transmission and Distribution, IEE Proceedings-, vol. 144, no. 3, pp. 293 300, may 1997. [14] A. Conceicao and C. Castro, A new approach to dening corrective control actions in case of infeasible operating situations, in Proc. IEEE PowerTech, vol. 3, 2001. [15] M. Larsson and D. Karlsson, Coordinated system protection scheme against voltage collapse using heuristic search and predictive control, IEEE Trans. Power Syst., vol. 18, no. 3, pp. 1001 1006, Aug. 2003. [16] A. Lerm and A. Silva, Avoiding Hopf bifurcations in power systems via set-points tuning, IEEE Trans. Power Syst., vol. 19, no. 2, pp. 1076 1084, may 2004.

[17] L. Ran, F. Spinato, P. Taylor, R. Wilson, and D. Jackman, Coordinated AVR and tap changing control for an autonomous industrial power system, Generation, Transmission and Distribution, IEE Proceedings-, vol. 153, no. 6, pp. 617 623, november 2006. [18] B. Venkatesh, R. Ranjan, and H. B. Gooi, Optimal reconguration of radial distribution systems to maximize loadability, IEEE Trans. Power Syst., vol. 19, no. 1, pp. 260266, Feb. 2004. [19] F. Baalbergen, M. Gibescu, and L. van der Sluis, Agent-based real-time voltage instability detection with maximum loadability index, in Proc. 45th Int. Univ. Power Eng. Conf., Sep. 2010. [20] M. Sadeghi and G. Sarvi, Determination of ZIP parameters with least squares optimization method, in Electrical Power Energy Conference (EPEC), 2009 IEEE, 2009, pp. 1 6. [21] H. Bai, P. Zhang, and V. Ajjarapu, A novel parameter identication approach via hybrid learning for aggregate load modeling, IEEE Trans. Power Syst., vol. 24, no. 3, pp. 1145 1154, 2009. [22] P. Kundur, Power System Stability and Control. New-York: McGrawHill, 1994. [23] Energiekamer, Netcode elektriciteit - voorwaarden als bedoeld in artikel 31, lid 1, sub a van de elektriciteitswet 1998 (in dutch), 2009. [Online]. Available: http://www.energiekamer.nl [24] F. Baalbergen, M. Gibescu, and L. van der Sluis, Outline of a new hierarchical agent-based voltage instability protection system, in Proc. IEEE Power Eng. Soc. Transmission and Distribution Conf., Apr. 2010. [25] , Voltage stability consequences of decentralized generation and possibilities for intelligent control, in Proc. IEEE ISGTE Europe Conf., Gothenburg, Sweden, Oct. 2010.

Johan Fredrik Baalbergen was born in Heemstede, the Netherlands, on August 11, 1983. He received his B.Sc. degree in Electrical Engineering from Delft University of Technology, the Netherlands, in 2005 and his M.Sc. degree in Electrical Engineering from the same University, in 2007. He is currently working on a PhD in the Electrical Power Systems group of the same university. M.Sc. Baalbergens main topics of interest are the implementation of renewable energy sources in power systems, smart grids and voltage stability.

Madeleine Gibescu received her Dipl.Eng. in Power Engineering from the University Politehnica, Bucharest, Romania in 1993 and her MSEE and Ph.D. degrees from the University of Washington in 1995 and 2003, respectively. She has worked as a Research Engineer for ClearSight Systems, and as a Power Systems Engineer for the AREVA T&D Corp. of Bellevue, Washington. She is currently an Assistant Professor with the Electrical Power Systems group at Delft University of Technology, the Netherlands.

Lou van der Sluis obtained his M.Sc. in electrical engineering from Delft University of Technology in 1974. He joined the KEMA High Power Laboratory in 1977. In 1990 he became a part-time professor and since 1992 he has been employed as a full-time professor at Delft University of Technology in the Power Systems Department.

Вам также может понравиться