1 Thevenin Equivalents, Adjoint Equivalents and Power Flow Equations Daniel F. C. M. S. Dias Abstract — This work starts by presenting a theoretical study about power flow, voltage stability, Thevenin’s theorem and Tellegen’s theorem applied to adjoint networks where some existing models are also included. ext follows the approach of the methods of analysis used in this dissertation, with special attention to the two models of Thevenin equivalent, where each one represents an operating situation (situation of short circuit and normal operating conditions of an electrical power system), and corresponding calculation of voltage at its terminals. In addition to these, it is also shown how to calculate the power flow (software MATPOWER) and the method chosen to indicate the voltage collapse ( ). As the simulation part, two types of disturbances are carried out: at a load bus and at a branch. The first is tested in three electrical power systems (3, 6 and 39 buses), in order to compare the two operation situations and their respective equivalent Thevenin models with the solution of the chosen power flow and obtain conclusions about them. The second is analyzed only in the 39 bus system, because it is the most complex one. These disturbances occur on the critical buses and branches of each system, in order to obtain more accurate results. Index Terms — Electrical Power Systems, Power Flow, Tellegen’s Theorem, Thevenin’s Theorem, Voltage Stability. I. INTRODUCTION T HE electrical power systems (EPS) are normally studied using the power flow equations, knowing that their interest is not usually in the solution of all currents and voltages, but the current and voltage of a small part of the network. So, given the need to improve the speed of calculations of these systems, it is convenient to be able to replace much of the network by a simple equivalent circuit. It is in this context that, while looking for alternatives to the solution of the power flow equations for a small part of the system, the Thevenin’s theorem and the Tellegen’s Theorem together with the adjoint networks appear as possible valid equivalent models. Thevenin’s theorem plays an important role in the study of linear systems. Knowing that the EPS are essentially nonlinear, the question lies in the proper function of Thevenin’s theorem in such systems beyond the usual calculation of short circuits (already used with guarantees). Among the circuit analysis theorems, Tellegen’s theorem is unusual because it depends solely on the Kirchhoff’s laws and the network topology. This theorem can therefore be applied to all electrical systems that comply with Kirchhoff’s laws, be they linear/nonlinear and variants/invariants in time. Tellegen’s theorem applied to the adjoint network technology is already considered as a valid and fast solution to this kind of problems. Its implementation performance in these systems has been extremely accurate after different types of disturbances (variations in voltage, impedance, power, changes in the network structure, etc.). With increasing demand on the quality, safety and operating conditions of the electrical power grids, as well as economic benefits and environmental constraints, the EPS are operated closer to their limits. Knowing this, it is understandable that the voltage stability is an issue with significant importance in the electrical industry of the last years, which makes it convenient to study it in this work. Thus, it is quite interesting to study and develop models of Thevenin equivalents and adjoint equivalents (using Tellegen’s theorem) in electrical power systems, testing them over a number of perturbations, particularly on the system’s voltage stability limits. In this work a theoretical study is presented of its important principles, including some already existing models. Next follows the approach of the analysis models used, including the chosen method of power flow (Newton’s method), the voltage collapse proximity indicator and two models of Thevenin’s equivalent for the two different operating situations (situation of short circuit and normal operating conditions of an electrical power system), called respectively Model 1 and 2. Both these models (especially Model 2, because it works with the normal operation situation of an EPS), along with their corresponding calculation of the voltage at its terminals are the original contributions of this work. In the end, there are some case studies conducted with three different power systems (3, 6 and 39 bust test systems), in order to compare those two Thevenin equivalents for both different situations with the power flow equations. For this, some disturbances in the load buses of all the three systems and disturbances in a branch of the 39 bus system are considered, starting always from an initial case also studied. II. THEORETICAL PRINCIPLES A. Power Flow In electrical power systems, power flow (1) is the steady state solution including the network, generators and loads. It is a fundamental problem in these kinds of systems. The power flow and its solution consist of the following steps: - Formulation of the mathematical model representative of the system; - Specification of the bus type and quantities for each one; - Numerical solution of the power flow equations, obtaining the magnitude and argument of the voltages of each bus; - Solution of the power transfer in each branch. 2 B. Voltage Stability Voltage stability (2) refers to the ability of a power system to maintain steady voltages at all buses in the system after being subjected to a disturbance from a given initial operating condition. It depends on the ability to maintain/restore equilibrium between load demand and load supply from the power system. Studies have been performed to predict voltage collapse (lower limit of the voltage from which the system becomes unstable) with both static and dynamic approaches. Since the objective of this work does not require the precision of dynamic analysis methods, mostly static analysis methods are referred in the following paragraphs. The P-V and Q-V curve methods (3) are classical techniques to predict voltage collapse. These curves are produced by running a serious of load flow cases. The critical point is where the curve makes an inversion from the upper to the lower level. Sterling et. al. (4) studied voltage collapse at load buses of the network using the concept of maximum power transfer between two buses. Using Thevenin’s equivalent for the relevant bus, the open circuit voltage and Thevenin’s equivalent impedance are calculated and utilized in deriving the stability index. This index is the ratio of Thevenin’s impedance to load impedance and is at a maximum of 1 when both are equal, indicating the system’s maximum stable loading at that bus. If the impedance ratio is lower than 1, the system is said to be stable. The system looses stability when the impedance ratio exceeds 1. Jasmon et. al. (5) formulates a voltage stability index, L, from the voltage equation derived from a two bus network and it is computed using Thevenin equivalent circuit of the power system referred to a load bus. To maintain a stable system at a specific load bus, the index has to be lower than 1. The system starts to be close to its voltage stability limit when the index approaches 1. Gubina et. al. (6) used Tellegen’s theorem associated with Thevenin’s theorem to calculate an index of voltage stability, obtaining satisfactory results. Tellegen’s theorem is used to simplify determination of the Thevenin parameters, which are calculated in a different way than adaptive curve-fitting techniques. To evaluate the system’s voltage stability at a given bus, it requires the voltage and current phasors measurements only. C. Thevenin’s Theorem Thevenin’s theorem (7) states that a linear circuit ended in two points a and b, containing any number of voltage generators, can be replaced by a single voltage generator in series with an impedance connected between those two points, called respectively VTH and ZTH (Fig. 1). The voltage generator VTH is equal to the open circuit voltage measured between a and b; the impedance ZTH is the impedance of the circuit measured between a and b when the voltage generators are short circuited. Fig. 1 – Thevenin’s equivalent circuit. This theorem is widely used for the calculation of short circuits, used with the superposition theorem. Short circuit (1) means a low impedance path, resulting from a defect, through which the current (generally very high) is closed. This is an abnormal situation in EPS that requires immediate action due to the damage it may do. Thus, the calculation of the short circuit currents is a fundamental task in this type of systems. Like said in the section before, Thevenin’s theorem is recently being used to predict the voltage collapse. Besides the already referred (3), (4), (5) and (6), others like Wang et. al. (8) and Yu et. al. (9) proposed Thevenin based models to observe and assess voltage stability. The first converts the whole system into an equivalent two bus system by multiple measured synchronized phasors, using those phasors of the target bus and the buses connected to it which are measured in WAMS (wide area measurement system). The relationship of the load impedance magnitude and the Thevenin’s equivalent impedance magnitude is analyzed when voltage stability problem occurs. Yu et. al. (9) monitors the weak voltage load bus groups with Thevenin equivalent parameters, presenting a power flow data based simulation method to track those parameters. Rao et. al. (10) investigates the validity of his Thevenin model in obtaining Q-V characteristics of P-Q loads, the main consideration for analysis of the problem of voltage instability and reactive power management. The model works with a very reasonable accuracy in non-linear systems, and cannot be the same as the one that is determined for fault analysis. To find the Thevenin model, the load buses were approximated to constant voltage sources, as well as the generator buses. The Thevenin impedance is computed at the load bus. The error increases with the system size and also with the wide variations in Q. Bahadornejad et. al. (11) develops a method to estimate power system Thevenin impedance, that is based on signal processing on the measured data in the load bus. It is shown that the cross-correlations of the changes in the load voltage and current with respect to the changes in the load admittance can be used to estimate the system Thevenin impedance. The required steps are clearly shown, being the method validated by simulation, also on real data. D. Tellegen’s Theorem and Adjoint !etworks Tellegen’s theorem (12) expresses a relationship between magnitudes of the branches of two adjoint networks, i.e. networks with the same graph. In its strong form, and as being adjoint networks, expresses the considering and 3 orthogonality between the voltages of a network and the currents of the other. This theorem and the concept of adjoint networks have been applied to a variety of functions and in several fields, but it’s their application to power systems that interests to this work. Examples of some of those applications to power systems are the voltage stability index calculated by Gubina et.al. (6), some works from Ferreira ( (13) and (14)) and from both Ferreira and Jesus ( (15) and (16)). Both of them confirmed Tellegen’s theorem and adjoint networks as a solid alternative to the normal power flow equations, using them to calculate voltages and currents of EPS after disturbances in that system. III. ANALYSIS MODELS A. Power Flow The computer program chosen to solve power flows is the MATPOWER (17), which is a package of MatLab files developed by PSERC (Power Systems Engineering Research Center) researchers. In this work, the power flow solver is the one based on a standard Newton’s method (18) using a full Jacobian, updated at each iteration. In order to return the solution of the power flows from MATPOWER in output arguments and to be able to use them, a MatLab file was created. This file also calculates the currents injected at each bus and transferred in each branch. B. Voltage Collapse Proximity Indicator Despite all the methods to predict the phenomenon of voltage collapse referred before, the one chosen to use in this work was the one from Moghavvemi et. al. (19), due to its simplicity and computational mathematics (compared to other methods, the solution is faster to find) combined with its proven capabilities in predicting the voltage collapse. Considering a typical transmission line of a power system network, Fig. 2 shows its representation through an equivalent Thevenin network, modeled with its parameters. Fig. 2 – Transmission line modeled with its parameters. In this method, the load of the line is treated as the power that is transferred at the receiving end through that particular line only, instead of the total load at the bus. Therefore, ∠ is the line impedance, ∠ф is the corresponding load impedance and ф = ⁄ . It is considered the most frequent case, where only the modulus of the load impedance is varied while ф remains constant. With the increase of demand in load, decreases and current I increases, which leads to a voltage drop at the receiving end. The maximum real power that can be transferred to the receiving end , and the maximum real power loss in the line , can then be obtained using the boundary condition ! ⁄! = 0, which leads to ⁄# = 1. Base on these maximum permissible quantities, the following Voltage Collapse Proximity Indicators were proposed by the author (note that the values of and are obtained from conventional power flow calculations): %&' ()* = = . % 012ф [1] +,- # ∙ # − ф . 4 ∙ 012 4 7 2 %&'8(## = = . 012 [2] +,- %# ∙ # − ф . 4 ∙ 012 4 7 2 As the loading condition moves closer to the critical operation point, both %&' ()* and %&'8(## approach the value 1. Close to the voltage collapse point, %&' ()* is less sensitive to further loading, while %&'8(## is found to be highly sensitive. Therefore, values of %&' ()* are adequate to indicate for indicating the voltage collapse in the line, but evaluation of %&'8(## helps to locate the voltage collapse point exactly. Because MATPOWER does not save the values of the power loss in the line from the power flow’s solution (needed to calculate the indicator %&'8(## ), and knowing that exact precision is not needed, in this work only the %&' ()* is used as a voltage collapse proximity indicator. Thus, a MatLab file was created to calculate the value of this indicator for all branches of a system, indicating the critical branch and critical bus, which is the one in the receiving end. C. Model 1 – Thevenin’s Equivalent for the short circuit situation Under the assumption of the short circuit theory, and to verify the proper function of Thevenin’s equivalent in short circuits, a model was developed to calculate its equivalent for this situation that does not depend on computer programs like MatLab but in the software OrCAD PSpice together with PSpice Schematics (very important in an electrical engineer student’s life). Thus, the following steps are used in the construction of the electrical circuits for the calculation of ZTH: - Replace the impedances between the system’s branches by resistors and capacitors/coils (in their respective metric system); - Replace the load buses by their equivalent load impedance, through: ||. [3] 9: = ∗ <: - Place the existing generators in short circuit (instead of using its transient reactance, like in a normal short circuit calculation); 4 - Place the AC voltage generator (VAC device) on the bus where the short circuit happens, with the same value as the pre-fault voltage on that bus; @ A:* @ A:* - Find the values of %>? and 'B through two separate simulations (20), and thus the value of ZTH with [4]. Note that @ A:* the value of %>? is measured by using VPRINT1 (from PSpice Schematics), changing the AC, MAG and PHASE cells (they have to be selected and entered OK). After @ A:* removing the VPRINT1 device, the value of 'B is measured by the insertion of an IPRINT device (from PSpice Schematics) between the terminals (between the bus and the ground or between the two buses), that also has to have the AC, MAG and PHASE cells selected and entered OK. The pertinent data from both simulations (set the analysis for AC Sweep/Noise with a linear sweep beginning and ending at a frequency of 50 Hz) are found in the output files of PSpice. @ A:* % [4] 9 >? = >? @ A:* 'B The value VTH of the equivalent circuit is the electrical potential difference between the specific bus and the ground or between buses i and j in the pre-fault situation, depending on the case. D. Model 2 – Thevenin’s Equivalent for the normal operating conditions Ohm’s law (21) states that the voltage across conducting materials is directly proportional to the current flowing through the material. Thus, the impedance is defined as the ratio of the electrical potential difference between two points and the current through these two points: [5] 9= Model 2 was then created based on these assumptions and, unlike Model 1, is applied in normal operating conditions of an EPS. With the aim of injecting power on the bus where the equivalent should be found, a unitary current source is inserted between the bus and the ground or between the buses i and j, as shown in Fig. 3. Thus, the value of 1+j0 A is added to the previous injected current in bus i. That same value is removed from the previous injected current in bus j, if there is such a bus, i.e., -1+j0 A added. Solving the power flows before and after the current injection, obtaining both voltage values, it is possible to find Thevenin’s impedance for both cases: DAE,8 − AEA:A,8 [6] A 9 >? = A ∆ DAE,8 DAE,8 GA − H I − GAEA:A,8 − HAEA:A,8 I [7] A 9 >? = ∆ The value VTH of the equivalent circuit is, like for Model 1, the electrical potential difference between the specific bus and the ground or between buses i and j in the pre-fault situation, depending on the case. Two MatLab files were created for this model, in order to be able to find the correspondent power injected in the buses for both cases, after the insertion of the current source in the circuit. These files run numerous iterations, during each are held several cases of small changes (increase/decrease) in the values of the load power involved. At the end of each iteration, the best option to achieve the unitary value of injected current is chosen. E. Calculating the Voltage between the Thevenin’s Equivalent Terminals After calculating Thevenin’s equivalent parameters VTH and ZTH (via the two described models), it is necessary to calculate the voltage Vab (voltage between the terminals a and b of Fig. 1). Depending from the case, these terminals can be the bus i and the ground (Vab applied to a bus) or the buses i and j (Vab applied to a branch). In order to find the impedance between terminals a and b (Zab) when looking to find Vab applied to a bus, one has to pay attention to the load before and after the disturbance. Considering Zo as the initial load impedance and Zc as the final load impedance, it is concluded that Zc equals to the parallel between Zo and Zab, like in Fig. 4. Thus, Zab is calculated by: 9K 9,J = [8] 9 1 − L KN 9M Fig. 4 – The parallel between the impedances Zo and Zab equals to Zc. Fig. 3 – Current injection of 1 A: (a) in bus i; (b) between buses i and j. Knowing all parameters, the voltage Vab between the bus j and the ground is found by: 9,J [9] ,J = >? ∙ 9 >? + 9,J 5 When it comes to Thevenin’s equivalent circuit between buses i and j (branch), the objective is not to calculate the voltage at its terminals depending on the load disturbances but to make changes in its network. Therefore, in order to disturb the system, Zab should lean to zero (short circuit) or to infinity M (open circuit). Considering ,J as the initial impedance of the branch, the way to put Zab as a short circuit is given by [10] and to put it as open circuit is given by [11]. 9M,J [10] 9E,J = [11] 9E,J = ∗ 9M,J For this section, two MatLab files were also created, that have the similarity of calculating Thevenin’s equivalent parameters for the two previously described models. Besides that, one of them calculates Vab between the bus i and the ground after load disturbances in that bus and the other one calculates the effects of the branch disturbances, which are observed by graphs from MatLab itself. IV. DISTURBANCE CASE STUDIES A. Initial Analysis In order to compare both models of Thevenin’s equivalent for the two different operating situations with the power flow by Newton’s method, three test systems (3-bus, 6-bus and 39bus) are used. Both initial data from the 3-bus and 6-bus systems are given in the Appendix, and were taken from (10) and (15), respectively. The initial data from the 39-bus system was taken from (17). In this section, MATPOWER was used to make an initial analysis of each of those three systems, indicating all the power flow results including all the values of the currents (injected at each bus and transferred in each branch). B. Critical buses and branches In order to truly evaluate the two models, the critical branch and corresponding critical bus of each system is found, using the voltage collapse proximity indicator studied before (%&' ()* ). In the 3-bus system, the critical branch is the one between the buses 1 and 3 (branch number 2), being the bus number 3 the critical one. The 6-bus system showed that the critical branch is the one between the buses 1 and 5 (branch number 3), being the bus number 5 the critical one. For the 39 New England bus system, because the calculated critical bus was a generator bus (the bus 31 maintains a constant voltage with power variations), it was necessary to resort to some of the references ((22) and (23)) to find the critical bus of this well known system. Therefore, the critical branch used was the one between the buses 12 and 11 (branch number 35) and the critical bus of this system is the bus number 12. C. Disturbances on a load bus Since the power load is complex, the voltage depends on the combination of active and reactive power of the concerned bus. The point of voltage collapse will then be affected by how these two powers vary independently and/or the combination of the two, i.e., the apparent power. Thus, for each test system there are three variations of power load (only active power, only reactive power and both powers) only on the critical bus, so that it could tend to voltage instability. Through these disturbances, both the magnitude and the argument of the voltage are recorded by the two Thevenin’s equivalents models and by the power flow equations from MATPOWER (Newton’s method). In the 3-bus system, Model 1 presented excellent results, having a maximum voltage magnitude error of 1,50% near voltage collapse when changing only the active power. Model 2 produced satisfactory results near voltage collapse, having a maximum voltage magnitude error of 14,99% for the reactive power variations. When studying the 6-bus system, Model 1 again showed excellent results. The maximum voltage magnitude error was with the active power disturbance, with a percentage error of 1,69%. On the other hand, Model 2 produced the worst results of all the three systems. Near voltage collapse, this method had voltage magnitude errors of 23,11%, 33,81% and 29,80% when variations of active, reactive and apparent powers occurred, respectively. In the 39-bus system, Model 1 presented the worst maximum voltage magnitude error (active power variations only), with 2,78%. Still, with an error so low, it is possible to say that the results were also excellent. Model 2 had again satisfactory results (although worst than in the 3-bus system), with maximum voltage magnitude errors near voltage collapse of 16,61%, 15,41% and 15,53% when variations of active, reactive and apparent powers occurred, respectively. Through the obtained results of this study, it is possible to conclude that Model 1 for the short circuit situation, like expected, presents good results for the voltage’s magnitude calculation. On the other hand, Model 2 presents considerable errors when the voltage in the critical bus is near its collapse. Note that, in both cases, the errors in the models tend to grow according to the proximity of the voltage collapse point, as expected. D. Disturbances on a branch Because the possible obtained results for all systems had the same conclusions for this kind of disturbances, only the 39-bus system was selected to be studied (since it is the biggest one from all the three test systems). After the calculation of Thevenin’s equivalent parameters, with equations [4] and [7], and in order to make the critical branch tend to short circuit or open circuit, various branches (n=1000) are placed in parallel with the terminals a and b (critical branch of the system) or the load is increased progressively, respectively. The curves of voltage and current between these terminals are then analyzed and verified that the branch do tend to those situations. In the short circuit situation the voltage Vab tends to zero and the current Iab tends to its P maximum value (',J = QR ). On the other hand, in the open SQR circuit situation the voltage Vab tends to its maximum value (Vab=VTH) and the current Iab to zero. 6 V. CONCLUSION This work was initiated with a study on the fundamental theoretical principles. Thus, a literature review was performed on the themes outlined in the work, including some of its existing methods. After it, an approach of the analysis models used was made, including the two models of Thevenin’s equivalent for the two operating situations (situation of short circuit and normal operating conditions of an electrical power system) and their corresponding calculation of the voltage at its terminals. Model 1 uses the software OrCAD PSpice together with PSpice Schematics to create electrical circuits based on the short circuits theory and correspondent Thevenin’s equivalent. This actually proved to be impractical, due to the need to build a new circuit with every new system or network reconfiguration. Model 2 is based on Ohm’s law and on the definition of impedance, and works on the normal conditions of nonlinear systems, like the EPS. For this model, because MATPOWER (program used to calculate power flows) does not let you insert a unitary current source directly in a bus of the system, some MatLab algorithms had to be created. These algorithms make it possible to inject unitary current sources in one or 2 buses of any kind of system, an essential requirement in this Thevenin’s equivalent model. In the simulation part of this work two different kinds of disturbances were made (after the calculation of the critical buses and branches of all testing systems, starting from an initial case study). Disturbances in the critical load bus were tested in three different systems (3-bus, 6-bus and 39-bus), while disturbances on a critical branch was only made in the 39-bus system. Model 1, because it uses the situation of short circuits, was expected to have good results on the load bus disturbance studies, and it did. On the other hand, because Model 2 works with the nonlinearity of the electrical power systems, it was expected to not have good results. That assumption was correct, since this model usual presented considerable errors when the bus was near voltage collapse (maximum error of 33,81%, for the 6-bus system). In both models, its errors tended to grow according to the proximity of the voltage collapse point. When concerning the disturbances on the critical branch, both Thevenin’s equivalent models had similar results. When the branch tended for short circuit and open circuit, both Vab and Iab presented expected results. As future work, and since in this work it was concluded that Thevenin’s equivalent models for nonlinear electrical power systems have problems, it seems interesting to find other kind of equivalent models to use in those nonlinear systems. Thinking already in that objective, a theoretical study about Tellegen’s theorem applied to adjoint networks was made, given its validity in this type of problems. It is expected that these adjoint equivalents will obtain accurate results when compared to the Thevenin’s equivalents developed in this work and even with the power flow equations solution. APPENDIX Fig. 5 – Representation of the 3 bus test system. TABLE I BRANCH DATA FOR THE 3 BUS TEST SYSTEM (p.u.) From To Bus Bus 1 1 2 Resistance Reactance Susceptance 2 0,02 0,08 0,02 3 0,02 0,08 0,02 3 0,02 0,08 0,02 TABLE II BUS DATA FOR THE 3 BUS TEST SYSTEM (p.u.) Bus Type Magnitude of Voltage Angle Pg Qg Pc Qc 1 Slack 1,02 0,00 - - - - 2 Gen 1,01 - 0,50 - - - 3 Load - - - - 1,40 0,50 No. Fig. 6 – Representation of the 6 bus test system. 7 TABLE III BRANCH DATA FOR THE 6 BUS TEST SYSTEM (p.u.) From To Bus Bus 1 Resistance Reactance Susceptance 2 0,10 0,20 0,00 1 4 0,05 0,20 0,00 1 5 0,08 0,30 0,00 2 3 0,05 0,25 0,00 2 4 0,05 0,10 0,00 2 5 0,10 0,30 0,00 2 6 0,07 0,20 0,00 3 5 0,12 0,26 0,00 3 6 0,02 0,10 0,00 4 5 0,20 0,40 0,00 5 6 0,10 0,30 0,00 TABLE IV BUS DATA FOR THE 6 BUS TEST SYSTEM (p.u.) Bus Type Magnitude of Voltage Angle Pg Qg Pc Qc 1 Slack 1,05 0,00 - - - - 2 Gen 1,05 - 0,50 - - - 3 Gen 1,07 - 0,60 - - - 4 Load - - - - 0,70 0,70 5 Load - - - - 0,70 0,70 6 Load - - - - 0,70 0,70 No. Fig. 7 – Representation of the 39 bus test system. REFERENCES 1. Paiva, J. P. Sucena. Redes de Energia Eléctrica – Uma Análise Sistémica. Instituto Superior Técnico : IST Press, 2005. 2. Definition and Classification of Power System Stability. IEEE/CIGRE Joint Task Force. Maio 2004, IEEE Trans.Power Systems, Vols. 19, no.3, pp. 1387 - 1401. 3. An analysis of voltage collapse in electric power systems. Bhat, R. and Gross, C.A. Outubro 2005, Proceedings of the 37th Annual North American, pp. 325 - 331. 4. Voltage collapse proximity indicator: behaviour and implications. Chebbo, A.M.; Irving, M.R.; Sterling, M.J.H. Maio 1992, IEE Proceedings C, Vols. 139, no.3, pp. 241 - 252. 5. A new technique for voltage stability analysis in a power system and improved loadflow algorithm for distribution network. Rahman, T.K.A.; Jasmon, G.B. Novembro 1995, Proceedings of EMPD'95, Vols. 2, 21 - 23, pp. 714 - 719. 6. Local Voltage-Stability Index Using Tellegen's Theorem. Smon, I., Verbic, G. and Gubina, F. Agosto 2006, IEEE Transactions on Power Systems, Vols. 21, no.3, pp. 1267 - 1275. 7. Stevenson, William D. Elementos de Análise de Sistemas de Potência. s.l. : McGraw-Hill. 8. A Method for Voltage Stability Assessment Based on Wide Area Measurement System. Su, Yongchun and Wang, Xiaoming. Março 2009, APPEEC 2009, pp. 1 - 4. 9. Tracking of Thevenin Equivalent Parameters on Weak Voltage Load Bus Groups. An, Tianyu, et al. Outubro/Novembro 2006, IEEE PES PSCE '06, pp. 1570 - 1576. 10. Model Validation Studies in Obtaining Q-V Characteristics of P-Q Loads in Respect of Reactive Power Management and Voltage Stability. Govinda Rao, G. and Ramachandra Murthy, K.V.S. Dezembro 2006, International Conference on PEDES '06, pp. 1 - 5. 11. System thevenin impedance estimation using signal processing on load bus data. Bahadornejad, M. and Ledwich, G. Novembro 2003, Sixth International Conference on ASDCOM, Vol. 1, pp. 274 - 279. 12. Jesus, C.M.S.C. Aplicação do Teorema de Tellegen e de Redes Adjuntas ao Cálculo de Tensões e Perdas em Sistemas de Energia Eléctrica. Instituto Superior Técnico. Dezembro, 2004. Dissertação de Doutoramento. 13. Tellegen’s Theorem and Power Systems - !ew Load Flow Equations, !ew Solution Methods. Ferreira, L.A.F.M. Abril 1990, IEEE Transactions on Circuits and Systems, Vols. 37, no. 4, pp. 519 - 526. 14. Accuracy and Complexity in the Modelling of Power Systems Subject to Large !etwork Disturbances in a Framework given by Tellegen’s Theorem and Adjoint !etworks. Ferreira, L.A.F.M. Singapura : s.n., Junho 1991, IEEE International Symposium on Circuits and Systems, Vol. 2, pp. 1228 1232. 15. Power Flow by Adjoint !etworks. Jesus, C.M.S.C. and Ferreira, L.A.F.M. Setembro 2003, IEEE PES Transmission and Distribution Conference and Exposition, Vol. 1, pp. 196 - 200. 16. Local !etwork Power Flow Analysis: an Accuracy Level Comparison for Two Sets of Equations. Ferreira, L.A.F.M., Jesus, C.M.S.C. Novembro 2006, IEEE Transactions on Power Systems, Vols. 21, no. 4, pp. 1616 - 1623. 17. Zimmerman, R.D., Murillo-Sánchez, C.E. MATPOWER: A MATLAB Power System Simulation Package. [Online] Setembro 2007. http://www.pserc.cornell.edu/matpower/. 18. Power Flow Solution by !ewton's Method. Tinney, W.F. and Hart, C.E. Novembro 1967, IEEE Transactions on Power Apparatus and Systems, Vols. PAS-86, no. 11, pp. 1449 - 1460 . 19. Real-time contingency evaluation and ranking technique. Moghavvemi, M. and Faruque, O. Setembro 1998, IEE Proceedings, Vols. 145, no.5, pp. 517 - 524. 20. Robbins, Allan and Miller, Wilhelm C. Circuit Analysis with Devices: Theory and Practice. 2. s.l. : Cengage Learning, 2003. pp. 722 - 725. 21. William H. Hayt, Jr., Kemmerly, Jack E. and Durbin, Steven M. Engineering Circuit Analysis. sexta edição. s.l. : McGraw-Hill, 2002. 22. Calculation of the Extreme Loading Condition of a Power System for the Assessment of Voltage Stability. Semlyen, A., Gao, B. and Janischewskyj, W. Fevereiro 1991, IEEE Transactions on Power Systems, Vols. 6, no.1, pp. 307 - 315. 23. Method of Combined Static and Dynamic Analysis of Voltage Collapse in Voltage Stability Assessment. Hasani, M. and Parniani, M. 2005, IEEE/PES Transmission and Distribution Conference and Exhibition: Asia and Pacific, pp. 1 - 6.