You are on page 1of 7

Neurocomputing 73 (2010) 27372743

Contents lists available at ScienceDirect

Neurocomputing
journal homepage: www.elsevier.com/locate/neucom

Adaptive neuro-fuzzy inference system based representative quality power factor for power quality assessment
N. Rathina Prabha a,n, N.S. Marimuthu b, C.K. Babulal c
a

EEE Department, Mepco Schlenk Engineering College, Sivakasi 626005, India National Engineering College, Kovilpatti 628503, India c EEE Department, Thiagarajar College of Engineering, Madurai 625015, India
b

a r t i c l e i n f o
Article history: Received 30 September 2009 Received in revised form 18 January 2010 Accepted 25 April 2010 Communicated by A. Zobaa Available online 6 May 2010 Keywords: ANFIS Power quality Representative quality power factor Sinusoidal and non-sinusoidal situations

abstract
Under sinusoidal operating conditions of electric power system, the classical denitions of apparent power and power factor work well as long as the loads are linear and the source voltage waveform is sinusoidal. Increase in use of power electronic devices, adjustable speed drives and other nonlinear loads cause the voltage and current waveforms to become non-sinusoidal and highly distorted. A new adaptive neuro-fuzzy inference system based representative quality power factor (ANFIS RQPF) is proposed in this paper to represent the existing different power factorsdisplacement power factor, transmission efciency power factor and oscillation power factor. The ANFIS RQPF can represent an essential module for evaluating and amalgamating the three power factors. The ANFIS RQPF was applied to different caseslinear, nonlinear, sinusoidal and non-sinusoidal considering lagging and leading power factors. It is shown that the ANFIS RQPF is expressive and accurately represents the existing power factors in all cases and in all situations. Taking into consideration the advantages of the ANFIS such as simplicity, ease of application, exibility, speed and ability to deal with imprecision and uncertainties, this factor can be useful for power quality assessment, cost-effective analysis of power quality mitigation techniques, as well as billing purposes, in these situations. & 2010 Elsevier B.V. All rights reserved.

1. Introduction Apparent power and power factor are not physical quantities by themselves, but they characterize physical phenomena, however in a way that may depend on the situation considered. The caution is that in some special situations, the same quantity may characterize more than a single phenomenon, but this is not necessarily the case in other or more general situations. Hence, it may be necessary, to generalize a single concept, introduced for particular situations, by more than one concept in more general situations. In sinusoidal situations, power factor denition is unique and expressive. However in non-sinusoidal situations and/ or nonlinear load, different power factors are proposed to express these situations. New denitions of electrical quantities in nonstandard situations are needed because of the changes in the situation in power systems. Denitions of the power factor related to various quality aspects may be useful to compare and optimize the effectiveness of techniques for compensation of the loads with respect to the various quality aspects. Therefore, there is a need

Corresponding author. Tel.: + 91 4562 235310, + 91 9443415805. E-mail addresses: nrpeee@mepcoeng.ac.in, nrp_me@yahoo.co.in (N. Rathina Prabha), ckbeee@tce.edu (C.K. Babulal). 0925-2312/$ - see front matter & 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.neucom.2010.04.012

for appropriate representation of apparent power and power factor under these conditions. Denitions for apparent power and power factor under these conditions are proposed in [112]. For instance, the IEEE Working Group on Non-sinusoidal Situations [5] has recommended separating the fundamental current and voltage components from the harmonic components, in order to calculate the fundamental apparent power and subsequently the displacement power factor (DPF) to facilitate applying engineering economic techniques, such as power factor correction. This also allows monitoring the fundamental power content separately from harmonic power content. In [10], Willems proposed separating the apparent power and the power factor denitions to characterize power transmission efciency and power oscillations in distorted current and voltage waveforms. As a result, two alternative power factors were dened: the transmission efciency power factor (TEPF) and the oscillation power factor (OSCPF). The analysis in [10] has contributed to dening characteristics that are useful to describe different aspect of power quality, as well with respect to the supplier as the load, and may lead to an application in the denition of quality requirements as well as in the tariff structure. Note that the maximum value of the oscillation power factor is 0.8165, and occurs in case of pure resistive load which explains the unavoidable oscillation even in the sinusoidal situation, while

2738

N. Rathina Prabha et al. / Neurocomputing 73 (2010) 27372743

the minimum value is zero which occurs in case of the pure reactive element, where there is continuous oscillation with zero average or active power. Another important power factor that is useful for separately monitoring the fundamental power from the harmonic power as well as easily applying in many engineering economic techniques, such as power factor correction was recommended by the IEEE Working Group in [5]. They recommended the separation of the fundamental power components from the non-fundamental components, and hence calculating the fundamental active and fundamental apparent power. Power factor is fundamentally an index of the quality of power that allows a user in a deregulated market to select an electricity provider on the basis of level of quality of the delivered power [13,14]. As a result, there is a need to evaluate the quality of the power delivered through evaluating a power factor index. Fuzzy system methodology has been demonstrated to allow solving uncertain and vague problems [15,16]. Example applications for power quality and power systems are included in [1722]. In [24], a fuzzy-based representative quality power factor (RQPF) is introduced. The RQPF evaluates a power factor index as a single value that represents the three power factors identied earlier, namely the displacement, transmission efciency and oscillation power factors. The fuzzy-logic-based approach calculates the RQPF using the Mamdanis fuzzy inference system (FIS). An ANFIS based total demand distortion factor for power quality evaluation is proposed in [26]. An ANFIS based method for determination of available transfer capability is proposed in [27] to determine the step length of Homotophy continuation power ow method by considering the values of load bus voltage and change in load bus voltage. In this paper, an ANFIS based RQPF is introduced. The RQPF evaluates a power factor index as a single value that represents the three power factors identied earlier, namely the displacement, transmission efciency and oscillation power factors. An ANFIS based approach is proposed to calculate the RQPF. The advantages of using the ANFIS are simplicity, ease of application, exibility, speed and ability to deal with imprecision and uncertainties. The proposed approach is tested for linear and nonlinear loads supplied from sinusoidal and non-sinusoidal sources, while considering lagging and leading power factors. The proposed ANFIS based RQPF can be successfully applied for evaluating the power quality, while considering distorted waveforms. The proposed RQPF can represent an essential ANFIS module in that application for evaluating the power factors, while aggregating it with other modules outputs such as total voltage harmonic distortion and total demand distortion. The organization of the paper is as follows: some basic concepts of an ANFIS are introduced in Section 2. The ANFIS based module used for calculating the ANFIS RQPF is explained in Section 3. The simulation results for applying the ANFIS based module to different cases are presented in Section 4. Effectiveness of an ANFIS RQPF in precise billing of customers is discussed in Section 5. Conclusions and suggestions for future work are given in Section 6.

controller is to learn and achieve good performance in the presence of disturbances and uncertainties. The design of membership functions is done by the ANFIS batch learning technique, which amounts to tuning an FIS with back propagation algorithm based on a collection of inputoutput data pairs. 2.1. ANFIS architecture Generally, an ANFIS is a multilayer feed forward network in which each node performs a particular function (node function) on incoming signals. For simplicity, we consider two inputs x and y and one output z. Suppose that the rule base contains two fuzzy if-then rules of Takagi and Sugeno type [25]. Rule 1 : IF x is A1 and y is B1 THEN f1 P1 x Q1 y R1 Rule 2 : IF x is A2 and y is B2 THEN f2 P2 x Q2 y R2 1

The ANFIS architecture is a ve layer feed forward network as shown in Fig. 1. An adaptive network [25] is a multilayer feed forward network, in which each node performs a particular function (node function) on incoming signals as well as a set of parameters pertaining to this node. The formulas for the node functions may vary from node to node, and the choice of each node function depends on the overall inputoutput function which the adaptive network is required to carry out. Note that the links in an adaptive network only indicate the ow direction of signals between nodes; no weights are associated with the links. To reect different adaptive capabilities, we use both circle and square nodes in an adaptive network. A square node (adaptive node) has parameters, while a circle node (xed node) has none. The parameter set of an adaptive network is the union of the parameter sets of each adaptive node. In order to achieve a desired inputoutput mapping, these parameters are updated according to the given training data and a gradient-based learning procedure is used. Layer 1: every node in this layer is a square node with a node function (the membership value of the premise part) O1 mAi x i 2

where x is the input to the node i, and Ai is the linguistic label associated with this node function. Layer 2: every node in this layer is a circle node labeled P, which multiplies the incoming signals. Each node output represents the ring strength of a rule. O2 mAi xmBi y, i where i 1 : 2 3

Layer 3: every node in this layer is a circle node labeled N (normalization). The ith node calculates the ratio of the ith rules ring strength to the sum of all ring strengths. O3 W i i Wi , W1 W2 where i 1 : 2 4

2. Adaptive neuro-fuzzy inference system The objective of an ANFIS is to integrate the best features of fuzzy systems and neural networks. The ANFIS utilizes Representation of prior knowledge into a set of constraints (network topology) to reduce the optimization search space, from fuzzy systems and adaptation of back propagation to structured network to automate FC parametric tuning, from neural networks, to improve performance. The design objective of the fuzzy

Fig. 1. ANFIS architecture.

N. Rathina Prabha et al. / Neurocomputing 73 (2010) 27372743

2739

Layer 4: every node in this layer is a square node with a node function O4 W i fi W i Pi x Qi y Ri , i where i 1 : 2 5 Layer 5: the single node in this layer is a circle node labeled S that computes the overall output as the summation of all incoming signals O5 system output, i where i 1 : 2 6

Eq. (6) represents the overall output of the ANFIS, which is functionally equivalent to the fuzzy system in [4]. 2.2. ANFIS learning algorithm In this subsection, the hybrid learning algorithm is explained briey. The ANFIS learning algorithm uses a two-pass learning cycle. In the forward pass, S1 is unmodied and S2 is computed using a least squared error (LSE) algorithm (off-line learning). In the backward pass, S2 is unmodied and S1 is computed using a gradient descent algorithm (usually back propagation). Thus, the hybrid learning algorithm is a combination of both the back propagation and the least square algorithms. The back propagation is used to identify the nonlinear parameters (premise parameters) and the least square is used for the linear parameters in the consequent parts.

From the ANFIS structure shown in Fig. 2, it has been observed that when the values of the premise parameters are xed, the overall output can be expressed as a linear combination of the consequent parameters. The hybrid learning algorithm is a combination of both back propagation and the least square algorithms. Each epoch of the hybrid learning algorithm consists of two passes, namely forward pass and backward pass. In the forward pass of the hybrid learning algorithm, functional signals go forward up to layer 4 and the consequent parameters are identied by the least squares estimate.

3. ANFIS based RQPF determination This section explains the ANFIS based approach used to calculate the ANFIS RQPF, which is a single value index that represents an amalgamation of the existing power factors: DPF, TEPF and OSCPF. Fig. 3 shows a schematic diagram of the ANFIS RQPF module proposed. This module was built using the ANFIS tool box available in MATLAB 7.5. The design procedure is as follows. 3.1. Input and output fuzzication The inputs to the ANFIS RQPF module are DPF, TEPF and OSCPF. The values of input variables and output variables for an ideal case and non-ideal case are shown in Table 1. While the RQPF ranges0 1, a value of 1 indicates no harmonics (distortion), while a value near 0 indicates high distortion. The triangular forms of memberships were used here to represent the three input variables, namely DPF, TEPF and OSCPF and are shown in Fig. 4. Three membership functions were used to identify the range of all
Table 1 Rule viewer output for ideal and non-ideal cases. Case Input DPF Ideal case Non ideal case 1.0 0.633 TEPF 1.0 0.741 Output OSCPF 0.8165 0.585 RQPF [24] 0.947 0.666 ANFIS RQPF 1 0.424

Fig. 2. ANFIS structure.

Fig. 3. Schematic diagram of ANFIS based RQPF module.

2740

N. Rathina Prabha et al. / Neurocomputing 73 (2010) 27372743

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2

INPUT VARIABLES 1 & 2


Low Medium High

4. The training data yields the membership functions for the input variables which are shown in Fig. 4. 5. The number of membership functions and their range is changed in order to t the training data output for the given values of DPF, TEPF and OSCPF, the ANFIS module calculates the ANFIS RQPF. Table 1 shows the rule viewer input and output for the two cases (ideal and non-ideal). Comparison of the values of RQPF and ANFIS RQPF for the same input shows that better performance is achievable using the proposed ANFIS based approach, compared to the FIS based approach in [24]. 4. Applications and results The ANFIS is applied to different test cases that include linear and nonlinear loads supplied from sinusoidal and non-sinusoidal sources. For given values of the DPF, TEPF and OSCPF, the ANFIS module will calculate the RQPF. Table 1 lists the output of the rule viewer for two cases, an ideal and a non-ideal case. An ideal case corresponds to the sinusoidal linear load, while the non-ideal case corresponds to any other case than the sinusoidal linear load. Fig. 5 shows the rule viewer diagram for the ideal case of Table 1. 4.1. Linear load supplied from sinusoidal source Fig. 6 shows a circuit consisting of a linear load supplied from a sinusoidal source (f1 60 Hz) through a line having an impedance of (5+ j5) O. The source voltage has the following time domain equation: vt 100 sin 377t: 7 The load voltage Vbn and the load current are used to calculate the DPF, TEPF and OSCPF. Using the proposed ANFIS RQPF module, the value of the ANFIS RQPF can be obtained. Seven cases are considered here as listed in Table 2. The values for these cases are chosen to represent different values of power factors to help explain and evaluate the ANFIS approach. Cases 26 represent lagging power factor, while Case 7 represents leading power factor. Once the approach fullls all of the requirements previously determined, it can be applied to all other complex cases. The circuit chosen is intended to be simple to help explain and evaluate the proposed approach. The simulation results for these cases are shown in Table 3. From Table 3, Case 1, involving a pure resistive load results in coinciding values of the DPF and TEPF, but the OSCPF is different because its maximum value is not one, but 0.8165. Also the ANFIS RQPF has its maximum value of 1 in this case, since all of the power factors that it represents are at their maximum value. For the remaining loading conditions, it is clear that the ANFIS RQPF tracks the changes in the three power factors. 4.2. Linear load supplied from non-sinusoidal source Referring to Fig. 6, when the source voltage is non-sinusoidal and contains in addition to the fundamental component, the third, fth, seventh, eleventh and fteenth, the time domain equation for this source voltage [23] could be written as vt 100 sin 377t 81 sin 1131t 60:6 sin 1885t 37 sin 2639t 15:7 sin 3393t 2:4 sin 4147t 6:3 sin 4900t 7:9 sin 5654t: 8 Table 3 represents the results of applying the new ANFIS approach to the seven cases while considering the non-sinusoidal source also. Comparing the sinusoidal source and non-sinusoidal

0.3

0.4 0.5 0.6 DPF&TEPF INPUT VARIABLE 3

0.7

0.8

0.9

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.1 0.2

Low Medium High

0.3

0.4 OSCPF

0.5

0.6

0.7

0.8

Fig. 4. Membership functions of input variables: (a) displacement factor (DPF) and transmission efciency factor (TEPF) and (b) oscillation power factor (OSCPF).

the three input variables. The ANFIS RQPF is required to identify the limits of the RQPF. The ANFIS RQPF output is represented using the triangular membership function due to its simplicity. 3.2. ANFIS if-then rules Twenty seven rules were developed in the ANFIS RQPF module. The development of these rules is based on different combinations of the input linguistic variables using the AND operator through the IF part of the statement. While the output linguistic variables help in determining the correct value for ANFIS RQPF through the THEN part of the ANFIS If-Then rules. The design steps are as follows: 1. The computation of an RQPF depends on the choice of three power factors. Ideal values should be selected for the input variables. 2. The performance of the ANFIS RQPF module depends on the size of the training data set. Thus, the training data for the three input variables should cover a wide range of all possibilities. 3. The training data for the ANFIS RQPF module is selected from the three input variables as described in Section 2.

N. Rathina Prabha et al. / Neurocomputing 73 (2010) 27372743

2741

Fig. 5. Rule viewer diagram for ideal case of Table 1.

Table 3 Comparison of an RQPF [24] and ANFIS RQPF. Case RLoad + jXLoad
n

Source and load type

RQPF

ANFIS RQPF

j20

j5

4 Fig. 6. Linear loads supplied from sinusoidal or non-sinusoidal source. 5 Table 2 Seven cases considered for ANFIS RQPF. Case 1 2 3 4 5 6 7 RLoad in O 1 0 0 20 20 20 20 XLoad in O 0 20 5 20 113 5 20 6

20+ j20

20+ j113

20+ j5

20 j20

SL SNL NSL NSNL SL SNL NSL NSNL SL SNL NSL NSNL SL SNL NSL NSNL SL SNL NSL NSNL SL SNL NSL NSNL SL SNL NSL NSNL

0.947 0.788 0.947 0.789 0.0521 0.0521 0.0521 0.0521 0.0521 0.0521 0.0521 0.0521 0.703 0.677 0.653 0.682 0.29 0.408 0.253 0.328 0.882 0.771 0.877 0.769 0.703 0.674 0.764 0.703

1.0 0.932 1.0 0.787 0 0 0 0 0.1 0 0 0 0.703 0.663 0.642 0.69 0.304 0.312 0.207 0.338 0.906 0.868 0.92 0.768 0.703 0.72 0.595 0.702

source conditions, under linear load, it can be inferred that their results look similar except that in some cases the TEPF and the OSCPF give different results as expected, when the source is the nonsinusoidal, and the ANFIS RQPF as a result try to track these changes.

n Source and load type: SLsinusoidal source, linear load; SNLsinusoidal source, non-linear load; NSLnon-sinusoidal source, linear load; NSNLnon-sinusoidal source, non-linear load.

4.3. Nonlinear load supplied from sinusoidal source Fig. 7 shows a circuit consisting of nonlinear load represented by a diode with linear load, supplied from a sinusoidal source that

can be expressed as in Eq. (7). The diode has a snubber resistance of 500 O and snubber capacitance of 250 nF. Note that now the load voltage Vbn is the voltage across the whole nonlinear load. Table 3 shows the simulation results for these cases, also. In this case, the situation is different. For example, considering Case 1 of the pure resistive load with diode, the DPF gives 1, while the TEPF gives a value less than one. This means that the power

2742

N. Rathina Prabha et al. / Neurocomputing 73 (2010) 27372743

transmitted to the load is less than that generated, although the load is purely resistive and there is no reactive element. Also, the OSCPF is less than 0.8165 which means that there is an oscillating power between the source and the load with zero average power, and this could explain why the transmission efciency is not equal to one in this case. The ANFIS RQPF, in this case, will have a value less than one which proves that it is very expressive and accurately represents the three power factor values. 4.4. Nonlinear load supplied from non-sinusoidal source The circuit shown in Fig. 7, is supplied by a non-sinusoidal voltage source as expressed in Eq. (8) with load representing different cases as listed in Table 2. Table 3 shows the simulation results of applying the new ANFIS approach for these cases also. This is the case most likely to be encountered in practice. The load is nonlinear and the source is non-sinusoidal. So, as shown in Table 3, there is no coincidence between the DPF values and the other power factor values, which indicates the presence of oscillating power, and that the useful power transmitted to the load is less than the generated one in all loading conditions.

5. Customer billing Charging the customer precisely, based on the power factor maintained by them is very crucial in the point of view of all stakeholders. Neither the Utility nor the customer must suffer, due to an incorrect bill. A solid proof for satisfying the customer with the reason for which he/she has been penalized is also crucial. Fig. 5 shows the rule viewer diagram with three inputs and output of the ANFIS based RQPF determination method for the ideal case in Table 1. The proposed ANFIS RQPF expresses quantitatively and successfully discriminates between these situations, since its value does not exceed unity. The following section explains the usefulness of ANFIS RQPF in two critical cases of penalty and benet to the customers.

5.1. Penalty to the customer Table 4 shows the selected cases of DPF and ANFIS RQPF values less than TEPF. In these situations, though the fundamental power factor is low, TEPF is high which penalize the utility. Hence, the customers pay less than the actual penalty. The numerical value of an ANFIS RQPF indicates the proportionality of an additional penalty to the customer. Thus, the ANFIS RQPF helps the utility to charge more penalties to the customer.

5.2. Benet to the customer Table 5 shows the selected cases of DPF and ANFIS RQPF values greater than TEPF. In these situations, though the fundamental power factor is high, TEPF is low that penalize the customer. Hence, the customers pay more than the actual penalty. The numerical value of an ANFIS RQPF indicates the proportionality of benet to the customer. Thus, the ANFIS RQPF helps the customer to pay fewer penalties to the utility. Hence, the ANFIS RQPF provides the tradeoff amalgamation of the three power factor values and helps the utility to charge the correct amount and penalty for the customers based on the power factor maintained by them.

Fig. 7. Nonlinear loads supplied from sinusoidal or non-sinusoidal source. Table 4 Critical cases of TEPF higher than ANFIS RQPF and DPF. Case RLoad + jXLoad
n Source and load type

DPF

TEPF

OSCPF

RQPF

ANFIS RQPF

5 7
n

20 + j113 20 j20

SNL NSL

0.3467 0.7071

0.5369 0.864

0.318 0.774

0.408 0.764

0.312 0.595

Source and load type: SNLsinusoidal source, non-linear load; NSLnon-sinusoidal source, linear load.

Table 5 Critical cases of TEPF less than ANFIS RQPF and DPF. Case RLoad +jXLoad
n Source and load type

DPF

TEPF

OSCPF

RQPF

ANFIS RQPF

1 4

1 20+ j20

5 6 7

20+ j113 20+ j5 20 j20

SNL NSNL SNL NSL NSNL NSL NSNL SNL NSNL SNL NSNL

1 1 0.8032 0.706 0.8768 0.1626 0.4604 0.9747 0.9855 0.9902 0.999

0.6797 0.684 0.5369 0.522 0.5369 0.0925 0.1808 0.6658 0.6389 0.5 0.4534

0.6929 0.695 0.6047 0.594 0.5537 0.1296 0.2477 0.6855 0.6704 0.6 0.5397

0.788 0.789 0.677 0.653 0.682 0.253 0.328 0.771 0.769 0.674 0.703

0.932 0.787 0.663 0.642 0.69 0.207 0.338 0.868 0.768 0.72 0.702

Source and load type: SNLsinusoidal source, non-linear load; NSLnon-sinusoidal source, linear load; NSNLnon-sinusoidal source, non-linear load.

N. Rathina Prabha et al. / Neurocomputing 73 (2010) 27372743

2743

6. Conclusion A new adaptive neuro-fuzzy inference system based representative quality power factor is proposed in this paper to represent the existing different power factorsdisplacement power factor, transmission efciency power factor and oscillation power factor. The ANFIS RQPF can represent an essential module for evaluating and amalgamating the three power factors. From the application of the proposed ANFIS based approach for calculating the representative quality power factor in linear or nonlinear loading conditions and supplied from sinusoidal or nonsinusoidal sources, considering lagging and leading power factor cases, it is found that this factor is very expressive and accurately represents these power factors in all cases. The power of the ANFIS RQPF is to characterize qualitatively and, at the same time, quantitatively the degree of electric power system utilizations by a single index, while carrying all of the characteristics of the three power factors that it represents. Also, the use of an ANFIS to calculate this factor has the advantage of being simple, easy to implement, exible, can be easily altered, adjusted, and it contains its knowledge base; so there is no need for an expert. The ANFIS RQPF can be effective in making a cost-effective analysis for applying the power factor correction devices and power quality mitigation techniques. Since the ANFIS RQPF can completely characterize the system power factors performance, this might help facilitate the comparative economic analysis which requires an estimation of the costs associated with the variations of the power factors due to disturbances and the costs of different PQ mitigation alternatives. The ANFIS RQPF will be useful for billing purposes, since it has proven to be very sensitive to any changes in power factors especially under non-sinusoidal operating conditions; therefore, customers will be charged the correct amount based on the power factor maintained by them. References
[1] P.S. Filipski, Polyphase apparent power and power factor under distorted waveform conditions, IEEE Trans. Power Del. 6 (3) (1991) 11611165. [2] A.E. Emanuel, On the denition of power factor and apparent power in unbalanced polyphase circuits with sinusoidal voltage and currents, IEEE Trans. Power Del. 8 (3) (1993) 841852. [3] P.S. Filipski, Y. Baghzouz, M.D. Cox, Discussion of power denitions contained in the IEEE dictionary, IEEE Trans. Power Del. 9 (3) (1994) 12371244. [4] L.S. Czarnecki, Power related phenomena in three-phase unbalanced systems, IEEE Trans. Power Del. 10 (3) (1995) 11681176. [5] Practical denitions for power systems with non-sinusoidal waveforms and unbalanced loads: a discussion, IEEE Trans. Power Del., vol. 11, no. 1, pp. 7987, Jan. 1996. [6] D. Sharon, Power factor denitions and power transfer quality in nonsinusoidal situations, IEEE Trans. Instrum. Meas. 45 (3) (1996) 728733. [7] Denitions for the measurement of electric quantities under sinusoidal, nonsinusoidal, balanced, or unbalanced conditions, IEEE Stand. 14592000, 2000. [8] F. Ghassemi, New concept in ac power theory, Proc. Inst. Electr. Eng., Gen. Transm. Distrib. 147 (6) (2000) 417424. [9] C. Sankaran, Power Quality, CRC Press, 2002. [10] J.L. Willems, Reections on apparent power and power factor in nonsinusoidal and polyphase situations, IEEE Trans. Power Del. 19 (2) (2004) 835840. [11] J.L. Willems, J.A. Ghijselen, A.E. Emanuel, The apparent power concept and the IEEE standard 14592000, IEEE Trans. Power Del. 20 (2, pt. 1) (2005) 876884. [12] H. Lev-Ari, A.M. Stankovic, A decomposition of apparent power in polyphase unbalanced networks in non-sinusoidal operation, IEEE Trans. Power Syst. 21 (1) (2006) 438440. [13] The modeling and simulation of propagation of harmonics in electric power networks Part I: concepts, models and simulation techniques, IEEE Trans. Power Del., vol. 11, no. 1, pp. 452465, January 1996. [14] M.H. Shwehdi and M.R. Sultan, Power correction capacitors; essentials and cautions, presented at the IEEE Power Eng. Soc. Summer Meet., Seattle, July 1620, 2000.

[15] L.A. Zadeh, Fuzzy sets, Inf. Control 8 (1965) 338353. [16] H.-J. Zimmermann, Fuzzy Set Theory and Its Applications, third ed., Kluwer, Norwell, MA, 1996. [17] A. Elmitwally, M.S. Kandil, M. Elkateb, A fuzzy-controlled versatile system for harmonics, unbalance and voltage sag compensation, in: Proc. IEEE Power Eng. Soc. Summer Meet., 2000, pp. 14391444. [18] S.A. Farghal, M.S. Kandil, A. Elmitwally, Quantifying electric power quality via fuzzy modeling and analytic hierarchy processing, in: Proc. Inst. Electr. Eng., Gen. Transm. Distrib 149 (1) (January 2002) 4449. [19] W.R.A. Ibrahim, M.M. Morcos, Articial intelligence and advanced mathematical tools for power quality applications: a survey, IEEE Trans. Power Del. 17 (2) (2002) 668673. [20] A.D. Grey, Power factor improvement using fuzzy logic control of AC synchronous motor, in: Proc. IEEE Southeast Conf., April 2005, pp. 193199. [21] W. Zhang and Y. Liu, Fuzzy logic controlled particle swarm for reactive power timization considering voltage stability, in: Proc. 7th Int. Power Eng. Conf., November 2005, pp. 15. [22] H.-S. Ko, J. Jatskevich, Power quality control of hybrid wind power generation system using FuzzyLqr controller, IEEE Trans. Energy Convers. 22 (2) (2007) 516527. [23] IEEE recommended practice and requirements for harmonic control in electric power systems, IEEE Stand. 5191992, 1993. [24] Walid G. Morsi, M.E. El-Hawary, A new fuzzy-based representative quality power factor for non-sinusoidal situations, IEEE Trans. Power Delivery 23 (2) (2008) 930936. [25] J.-S. Jang, ANFIS: adaptive network based fuzzy inference system, IEEE Trans. Syst., Man Cybern. 23 (3) (1993) 665685. [26] N. Rathina Prabha, N.S. Marimuthu, C.K. Babulal, An adaptive neuro-fuzzy inference system based total demand distortion factor for power quality evaluation, Elsevier J. Neurocomput. 73 (13) (2009) 315323. [27] C.K. Babulal, P.S. Kannan, A novel approach for atc computation in deregulated environment, J. Electr. Syst. 2 (3) (2006) 146161.

N. Rathina Prabha received her B.E. (EEE) and M.E. (Power System Engg.) degrees from Thiagarajar College of Engineering, Madurai, India. She is presently working as an Assistant Professor at Mepco Schlenk Engineering College, Sivakasi, India. Her areas of interest include Power Quality, Power System Control and Electrical Machines.

N. S. Marimuthu received his B.E. and M.Sc. (Engg.) degrees from Annamalai University, Chidambaram, India. He received his Ph.D. degree from I.I.T. Kharagpur, India, in 1988. He is presently Principal, National Engineering College, Kovilpatti, India. His areas of interest include Power System Dynamics and Power Electronics.

C. K. Babulal received his B.E. (EEE) and M.E. (Power System Engg.) degrees from Thiagarajar College of Engineering, Madurai, India and Ph.D. degree from Madurai Kamaraj University, Madurai, India in 1991, 1993 and 2007, respectively. Presently, he is an Assistant Professor in the Department of Electrical and Electronics Engineering at Thiagarajar College of Engineering, Madurai, India. His areas of interest include Power System Voltage Stability, Power System Restructuring and Intelligent System Applications in Power Systems.

You might also like