Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

1

Validation of a Two-Time Step HVDC Transient Stability Simulation Model including Detailed HVDC Controls and DC Line L/R Dynamics
Rebecca M. Brandt, Udaya D. Annakkage, Senior Member, IEEE, Dennis P. Brandt, Senior Member, IEEE, and Niraj Kshatriya, Student Member, IEEE
DC voltage and current controllers; it assumes that the controllers are fast enough to allow the DC voltage and current to instantaneously follow their respective setpoints. It uses steady state DC equations to back-calculate the corresponding firing angles, alpha and gamma, based on the DC voltage and current and the AC commutating bus voltages. A real HVDC link will not recover instantaneously from a fault, and because the DC controllers are not included in the response model, the user is required to input many model parameters, such as dc voltage and current ramp rates and ac voltage blocking and bypassing levels, in order to preprogram the HVDC model to respond to a certain fault in a certain manner. This is a drawback to a response-type model; instead of running the simulation to see how the HVDC will respond in the system, the user specifies how it will respond. Other drawbacks of this type of model include: --The model response is not very accurate because the L/R dynamics of the dc link are not modeled and therefore the dc current transients are not represented, which can have a significant impact on the response. --It can be time-consuming and difficult for users to select appropriate model input parameters, especially without having a benchmark with which to compare the model response. --Depending on the strength of the ac system, different sets of input parameters may be required to represent the HVDC during various system powerflow conditions and for various disturbances. --It is not possible to represent DC line faults. Most of the important dynamics of synchronous machines can be captured in a transient stability simulation with a cycle time step of integration. Therefore it is an unnecessary burden to run the computer simulation with a much smaller time step solely for the purpose of capturing the effects of fast HVDC controls. Instead a method is proposed to combine the two features: the ability to include the detailed fast HVDC controls while still running the transient stability simulation of the large network at a cycle time step. In order to demonstrate the method, the commercially available transient stability program PSS/E was used to develop a user-written HVDC model that included the fast DC controls and the DC line L/R dynamics. A two-time step approach [4] was used to run the HVDC model at an internal time step smaller than the large network time step in order to minimize the impact on

AbstractA two-time step approach to transient stability modeling of detailed HVDC controls and L/R line dynamics is validated against a detailed electromagnetic transients model, and compared to the response-type HVDC model typically used in transient stability studies. The two-time step method combines the ability to include the detailed fast HVDC controls while still running the transient stability simulation of the large network at a cycle time step, minimizing the impact to the overall computation time. To reduce complexity of the model, the detailed HVDC model assumes the AC commutating bus voltages to be unchanging during the internal fast time step loop of the HVDC link. Validation results show improved accuracy and numerical stability of the detailed HVDC transient stability model when compared to the typical response-type model, both in weak and strong test systems with Effective Short Circuit Ratios (ESCR) ranging from 2.0 to 5.0. Index TermsHVDC transmission, Power system modeling, Power system simulation, Power system transient stability

I. INTRODUCTION

RANSIENT stability system studies for large networks are typically performed at a cycle time step. This can create a problem for networks containing High Voltage Direct Current (HVDC) transmission lines and Flexible AC Transmission Systems (FACTS) devices, whose controller bandwidths are often much wider than what could be accommodated with a cycle time step. Due to the computation time required to run a transient stability simulation, it is not practical to use a time step that would be sufficiently small to facilitate detailed HVDC control modeling in large networks. Instead, a response-type model is often used to represent HVDC in stability programs [4,5,8]. The HVDC response-type model does not represent the

This work was supported by TransGrid Solutions, Inc. of Winnipeg, Manitoba, Canada. www.transgridsolutions.com. R. M. Brandt, TransGrid Solutions Inc., Winnipeg, MB, CANADA. (phone: 204-480-4044; fax: 204-989-4858 e-mail: rbrandt@transgridsolutions.com). U. D. Annakkage, University of Manitoba, Winnipeg, MB CANADA (email: annakkag@ee.umanitoba.ca). D. P. Brandt, TransGrid Solutions Inc., Winnipeg, MB, CANADA. (e-mail: dbrandt@transgridsolutions.com). N. Kshatriya, University of Manitoba., Winnipeg, MB, CANADA. (e-mail: nkshatri@ee.umanitoba.ca).

1-4244-0493-2/06/$20.00 2006 IEEE.

computation time, while still allowing the details of the HVDC to be accurately represented. To keep the model as simple as possible, the AC commutating bus voltages were assumed to remain unchanged during the internal time step loop, being aware that this assumption could become critical for severe nearby disturbances and in weaker AC systems. The objectives are to: 1) Determine the potential impacts of the unchanging AC voltage assumption: How large of a disturbance can be applied while retaining accuracy of the model response? How weak of a system can such a model be used in? 2) Compare the response of the detailed two-time step model with the response-type model and an electromagnetic transient model. Determine if the accuracy of the detailed model response significantly improved, along with the ease of use of the model compared to time-consuming parameter-setting of the response-type model. II. HVDC CONVERTER MODEL A. DC Line Representations 1) Overhead Line and Back-to-Back Models. The overhead line and back-to-back model shown in Figure 1 ignores the dc line capacitance. The DC line is modeled using one differential equation as given in (1) to represent the entire inductance of the dc line conductor, smoothing reactors and converter transformer reactances.
dI dc Vdcr Vdci Rdc Idc . = dt 2 .0 (Lsr + Lline + Lxfmr )

dV cap dt

Icap . Ccap

(5)

Fig. 2. Cable representation of DC line.

3) Algebraic Equations: All Models. The algebraic equations are the same for both dc line representations. The steady state DC equations, which include the converter transformers, are used to calculate the rectifier and inverter DC voltages as well as the real and reactive powers at both converters. These steady state dc equations require as inputs the DC current(s) as calculated from the differential equations described in (1), (2), (3) and (5), as well as the firing angles, alpha and gamma, which are outputs of the HVDC controls described in section II B. The set of algebraic equations are given in (6) (13) [6].
Vdcr = 1 .35 Vlr cos Vdci = 1 .35 Vli cos 3 Xcr Idc

(6) (7) (8) (9) (10) (11) (12) (13)

r = cos 1 (cos rec i = cos 1 (cos inv


Pacr = Idc Vdcr Paci = Idc Vdci Qacr = Pacr tan r
Qaci = Paci tan i

Xcr Idc 2 Vlr Xci Idc 2 Vli

Xci Idc
)

(1)

Fig. 1. Overhead line representation of DC line.

2) Cable Model. The cable model shown in Figure 2 represents the dc line capacitance as a lumped capacitor in the middle of the cable to form a T line model. The dc line is modeled using three differential equations: one to represent the capacitance (2), and the remaining two, (3) and (5), to represent inductances on either side of the capacitance [7].
Rdc Idcr 2 .0 . (Lsr + Lline + Lxfmr ) Rdc Vdci + Vcap Idci 2.0 . = (Lsr + Lline + Lxfmr ) Vdcr Vcap

B. Basic HVDC Controls DC current or power and DC voltage or gamma are two typical operator setpoints for an HVDC link [9]. Under normal operation, the rectifier is usually responsible for controlling the DC current, and the inverter is responsible for controlling the DC voltage. The PI-controllers that control DC voltage and current at each converter produce a firing angle to maintain the DC voltage and current at their setpoints, or references, according to the steady state DC equations described above. There are additional control functions that come into operation during abnormal system conditions. These include: 1) Voltage Dependent Current Limits (VDCL). This function reduces the DC current reference if the DC voltage becomes too low according to a DC voltage-current characteristic. The rectifier and inverter have separate VDCL characteristics. 2) Inverter Control Mode Selection. The inverter is also

dI dcr = dt dI dci dt

(2)

(3) (4)

Icap = Idcr Idci

equipped with a current controller that is operating in the background and only becomes active if the rectifier cannot maintain its scheduled DC current and is sitting at minimum alpha. The inverter current order is less than the rectifier current order by a current margin, usually 10%. When the dc current drops below the inverter current order, the inverter takes over current control. 3) Current Error Control (CEC). This function provides for a smooth transition between operating points when the inverter is changing from voltage to current control or vice versa.

Fig. 3. Basic HVDC Control Concpet.

dynamics in the approximate frequency range of 0.1 Hz to 10 Hz, using the general assumption that fast electromagnetic transients (less than 0.1 Hz) die out within milliseconds (with the exception of subsynchronous resonance), and do not affect the stability simulation which is often run to 5 or 10 seconds [1]. However, the fast controls of HVDC converters do have a direct impact on the transient stability behavior of power systems. During the course of one simulation time step (after initialization is complete) as shown in Figure 4, the PSS/E transient stability program performs the following steps [2]: 1) Calls model subroutines to update the state variables based on the calculation of time derivative equations from the previous time step, and calculates new source currents for all current-injection models. 2) Calls model subroutines to retrieve current injections, and performs the network solution. This is an iterative process that calculates the network powerflow solution AC voltages and angles based on current injections at each bus in order to satisfy the P and Q requirements at load buses and the P and V requirements at generator buses until the solution tolerance is achieved. 3) Calls model subroutines to calculate the new state variable time derivatives. 4) Performs numerical integration, and updates the time step.

C. Internal Time Step for the HVDC Model The integration time constants associated with the HVDC controllers and measurement circuits are in the range of 2 ms to 20 ms. Considering a simple first-order integration process, the maximum time step of a transient stability simulation must be no more than of the smallest integration time constant present in the models in order to avoid numerical instability. PSS/E uses a second-order integration technique, however the PSS/E Application Guide suggests using a time step to 1/5 of the shortest integration time step being modeled to ensure numerical stability [2]. The same theory applies to the differential equation associated with the line capacitance in the DC cable model. It is required to run the HVDC model at a smaller internal time step because the typical cycle, 8.33 ms or 10 ms, time step used in most stability simulations is too large for the HVDC controls model and dc line differential equations to be numerically stable.
Fig. 4. Solution Steps to One PSS/E Simulation Time Step.

III. TRANSIENT STABILITY SIMULATION The traditional Transient Stability Simulation method is briefly reviewed in section III A using PSS/E as the example. The proposed hybrid simulation concept is then introduced in section III B. A. Traditional Transient Stability Program Simulation Concepts Transient stability programs are used to study power system

B. Hybrid Simulator 1) Method: The two-time step, or hybrid simulator, approach to modeling is achieved by wrapping a piece of user-written code around the user-written HVDC model. As depicted in Figure 4, when the main PSS/E transient stability program performs the call to the HVDC subroutine, this piece of code effectively takes over the role of the transient stability program by performing steps 1) to 4) (from Transient Stability

Simulation section III part A) on the HVDC model alone at an internal time step of 50 sec to ensure numerical stability. Once the HVDC model has been processed internally the required number of times to equate to the network solution time step (e.g. 10 ms), control is given back to the real PSS/E program to perform its regular operation. 2) Potential Problem: Even though the HVDC model is being executed at a small time step, it is still only interfacing with the large network solution every PSS/E time step. The assumption during this internal process is that the AC commutating bus voltages remain unchanged. The network AC voltages are only updated when the PSS/E transient stability program performs step 2) every PSS/E time step. This means that during the internal time step process the HVDC controls are producing firing angles and calculating updates to DC voltage and current but the AC voltage is not being affected, which is incorrect. This will result in some amount of error in where the HVDC controls end up at the end of each internal process because the AC voltage affects the error signal inputs to these controllers. The amount of error will depend on how weak the network is at the interface point with the HVDC link, and how severe a disturbance is being simulated, i.e. how fast the AC commutating bus voltages are changing. Methods, such as estimating updates to the AC commutating bus voltages every 50 sec using known Thevenin equivalents and updated converter P and Q injections, could be used to reduce the internal AC voltage error, however this would add a further degree of complexity to the internal process and would require more inputs from the user, such as system strengths at the rectifier and inverter buses. The question is whether or not the HVDC model still produces an accurate response in spite of this AC voltage assumption. The HVDC Converter Model Validation in section IV will address this question.

model every detail of the HVDC controls in PSS/E, especially since some of the controls operate based on single-phase quantities in PSCAD/EMTDC, which are not available in a positive-sequence transient stability program such as PSS/E. The purpose of the PSCAD/EMTDC model was to provide benchmark test results. The idea was to compare the response between the PSS/E response-type model CDC6 and the PSS/E detailed HVDC two-time step model against the much more detailed three-phase PSCAD/EMTDC benchmark model. The model was validated using three test systems by simulating solid and remote three-phase faults applied at the rectifier and inverter commutating buses. The main waveforms of interest were the DC current, which was expected to be significantly improved due to its representation as a state variable in the L/R dynamics, as well as the AC commutating bus voltages because these quantities demonstrate the impacts of the HVDC model at the interface point to the rest of the network. Other DC and AC quantities were also compared. A. Test System 1 A simple Thevenin equivalent 2-bus test system with an Effective Short Circuit Ratio (ESCR) of 2.0 at both the rectifier and inverter was used to test the accuracy of the HVDC models in a weak system. The HVDC was a bipolar link with a rating of 500 kV and 1800 MW.
Rectifier AC Bus Voltage (pu) 1.4 1.2 1 0.8 0.6 0.4 1.4 1.2 PSCAD PSSE CDC6 PSCAD Two time-step

IV. HVDC CONVERTER MODEL VALIDATION As previously mentioned, the commercial transient stability program PSS/E was used to demonstrate the model validation. The idea could also be implemented using any other transient stability program in which the user has the capability to develop user-written models. Three HVDC converter models were set up for performing the model validation; one in PSS/E using the library responsetype model CDC6, one in PSS/E using the user-written twotime step detailed HVDC model described in this paper, and one detailed electromagnetic transients HVDC model in PSCAD/EMTDC. PSCAD/EMTDC is an electromagnetic transients program implementing a three-phase time domain switching solution that is typically run using a 50 sec time step. The PSS/E detailed HVDC model implemented a simplified version of the same HVDC controls used in the PSCAD/EMTDC model. It was not practical or necessary to

1 0.8 0.6 0.4 0 0.2 0.4 Time (s)


Fig. 5. Solid three-phase fault at inverter commutating bus. Comparison of rectifier AC bus voltages. Top: PSCAD and PSSE two-time step HVDC model; Bottom: PSCAD and PSSE CDC6 library model.

0.6

0.8

Figure 5 shows the rectifier AC voltage response to a solid three-phase inverter fault. In the weak system, the two-time step model performed significantly better than the responsetype model, both in accuracy and numerical stability. Similar accuracy of results was also seen for the various test cases and for other AC and DC quantities in this test system.

B. Test System 2 A simple Thevenin equivalent 2-bus test system with an ESCR of 5.0 at both the rectifier and inverter was used to test the accuracy of the HVDC models in a strong system. The HVDC was a bipolar link with a rating of 500 kV and 1800 MW.
DC current (kA) 3 2 1 0 PSCAD Two time-step

Rectifier Bus AC Voltage (pu) 1.2 1.1 1 0.9 0.8 1.2 1.1 1 0.9 0.8 PSCAD PSSE CDC6 PSCAD Two time-step

3 2

PSCAD PSSE CDC6


1.015

2 Time (s)

Generator 11speed (pu) PSCAD Two time-step

1 0 0 0.2 0.4 Time (s) 0.6 0.8 1

1.01 1.005 1 0.995 1.015 1.01 1.005

Fig. 6. Solid three-phase fault at rectifier commutating bus. Comparison of DC current. Top: PSCAD and PSSE two-time step HVDC model; Bottom: PSCAD and PSSE CDC6 library model.

PSCAD PSSE CDC6

Figure 6 shows the DC current response to a solid threephase rectifier fault. The two-time step model performed significantly better than the response-type model in terms of accuracy. Similar accuracy of results was also seen for the various test cases and for other AC and DC quantities in this test system. C. Test System 3 The third test system is a 12-bus system with a 250 kV, 350 MW HVDC monopole link placed between bus 1 (rectifier) and bus 3 (inverter) [3]. The HVDC data and controls were based on the CIGRE HVDC benchmark model. The purpose of performing tests using this test system was to validate the HVDC model in a typical AC network that included machine dynamics. The test system contained four machines. Figure 7 shows the rectifier AC voltage, DC current and generator speed responses to a solid three-phase inverter fault. The two-time step model performed significantly better than the response-type model in terms of accuracy. The damping and phase angle of rotor angle oscillations produced by the two-time step model very closely resembles those of the detailed PSCAD model, whereas the response-type model produced a significant phase shift and much larger damping. Similar accuracy of results was also seen for the various test cases and for other AC and DC quantities in this test system.

1 0.995

2 Time (s)

DC current (pu) 2.5 2 1.5 1 0.5 0 2.5 2 1.5 1 0.5 0 0 1 2 Time (s) 3 4 5 PSCAD PSSE CDC6 PSCAD Two time-step

Fig. 7. Solid three-phase fault at inverter commutating bus. Tops: PSCAD and PSSE two-time step HVDC model; Bottoms: PSCAD and PSSE CDC6 library model.

V. CONCLUSION The two-time step approach to modeling HVDC controls and L/R line dynamics in a transient stability program was seen to produce accurate results, especially when considering the fundamental differences between the solution algorithms of an electromagnetic transient program and a transient stability program. The benefit of the two-time step method is that the impact to the simulation time is minimized by allowing the transient stability network solution to be run at the large cycle time step, while still accurately representing the details of the fast HVDC controls and L/R line dynamics. The commercially available transient stability program PSS/E was used to develop a user-written HVDC model that included the fast DC controls and the DC line L/R dynamics, which was then compared to results of an electromagnetic transients program PSCAD/EMTDC as a benchmark. To keep the HVDC model as simple as possible, the AC commutating bus voltages were assumed to remain unchanged during the internal time step loop. This assumption produces a slight error in the output of HVDC controls because the firing angles are changing but not affecting the AC commutating bus voltages. As expected, the effects of the assumption were slightly more evident, i.e. less accuracy, in the weakest test system, however the results of the two-time step HVDC model were still relatively accurate, especially when compared to the generic response-type model. The results of the two-time step model were also more numerically stable than the generic response-type model in a weak AC system. It was observed that a solid three-phase fault could be applied to either AC commutating bus in a weak system of ESCR 2.0, while still retaining a fairly accurate and stable model response. The AC voltage assumption was not seen to be a significant issue. In comparing the response of the detailed two-time step model with the response-type model against an electromagnetic transient model, the accuracy of the detailed model response was significantly improved. In particular, the DC current transients were accurately represented, which improved the accuracy of the AC commutating bus voltages and therefore resulted in an overall more accurate AC network solution. The ease of use of the detailed HVDC controls model compared to time-consuming parameter setting of the response-type model is also a benefit. The HVDC control parameters of the two-time step model only need to be entered once based either on known values from the actual HVDC system or typical values. They do not need alteration for different disturbances, as is sometimes the case for parameters such as voltage and current ramp rates of the response-type model. A further benefit of the two-time step HVDC model is that it could be programmed to go a step further and allow the user the ability to simulate HVDC line faults and re-start sequences, thereby providing the user with a more accurate picture of the potential impacts of such DC disturbances on

the transient stability of the AC system, which is not possible with the generic response-type model. VI. ACKNOWLEDGMENT The authors thank Mojtaba Mohaddes and Pete Kuffel of TransGrid Solutions Inc. for the knowledge and expertise they provided on HVDC controls and HVDC modeling aspects.

VII. REFERENCES
[1] [2] [3] P. Kundur, Power System Stability and Control. McGraw Hill, 1993, ch. 10, ch. 13. PSS/E 29, Volume II: Program Application Guide, Power Technologies Inc., Schenectady, NY, October 2002, ch. 11. Jiang S. ; Annakkage U. D. ; Gole A. M., A Platform for Validation of FACTS Models, IEEE Transactions on Power Delivery: TPWRD.2005.852301. S. Arabi, P. Kundur, J.H. Sawada, Appropriate HVDC Transmission Simulation Models for Various Power System Stability Studies, IEEE Trans. on Power Systems, vol. 13, no. 4, pp. 12921297, Nov. 1998. Bradley K. Johnson, HVDC Models used in Stability Studies, IEEE Trans. on Power Delivery, vol. 4, no. 2, pp. 11531163, Apr. 1989. HVDC Handbook, First Edition, GE Industrial and Power Systems, Schenectady, NY, 1994, ch. 3. PSS/E 29, Volume II: Program Application Guide, Power Technologies Inc., Schenectady, NY, October 2002, ch. 17, p. 59. HVDC Handbook, First Edition, GE Industrial and Power Systems, Schenectady, NY, 1994, ch. 4. HVDC Handbook, First Edition, GE Industrial and Power Systems, Schenectady, NY, 1994, ch. 5.

[4]

[5] [6] [7] [8] [9]

VIII. BIOGRAPHIES
Rebecca M. Brandt received her B.Sc. degree in Electrical Engineering at the University of Manitoba in 2001. She is currently a System Studies Engineer with TransGrid Solutions Inc., an engineering consulting company located in Winnipeg. She has been a member of the Association of Professional Engineers and Geoscientists of Manitoba since 2004. Udaya D. Annakkage (M'95-SM'04) received the B.Sc. (Eng) degree from the University of Moratuwa, Sri Lanka, in 1981 and the M.Sc. and Ph.D. degrees from the University of Manchester, UK, in 1984 and 1987 respectively. He is currently a Professor at the University of Manitoba (since 2001), Winnipeg, MB, Canada. Previously, he was with the University of Moratuwa (1987-1990) and the University of Auckland, New Zealand (1991-2000). Dennis P. Brandt (M 85- SM 04) received his B.S. degree in Electrical Engineering at the University of Manitoba in 1975. He is currently Vice President and a Director of TransGrid Solutions Inc. (since 2002) in Winnipeg, Manitoba and is a senior consultant on HVDC and FACTS systems. He has previously worked for Manitoba Hydro (17 yrs.) on the Nelson River HVDC system. He also worked for Brown Boveri Switzerland (1986-87) as a senior HVDC System controls engineer responsible for the control & protection concept in the Pacific Intertie Expansion Project. Since 1992 he has been President and Technical director of Brandt Consultants Inc. in Winnipeg, Manitoba. Niraj Kshatriya (S'04) received the B.E.E. degree from the Sardar Patel University, Vallabh Vidhya Nagar, India in 1992 and the M.Sc. degree in 2004 from the University of Manitoba, Winnipeg, MB, Canada, where he is currently pursuing the Ph.D. degree. He worked for Gujarat Narmada Valley Fertilizers Co. (GNFC) Ltd. from 1992 to 2001.

You might also like