Power Quality Improvement With Shunt Hybrid
Active Power Filter Using ANN-Based Predictive and
Adaptive Controllers
Mr. CH. VNV HariKrishna Smt. CH. Sujatha Sri. P. Bala KoteswaraRao M.Tech Scholar Associate professor Assistant professor
Department of EEE Department of EEE Department of EEE
Gudlavalleru Engineering College Gudlavalleru Engineering College Gudlavalleru Engineering College Gudlavalleru, AP, India. Gudlavalleru, AP, India. Gudlavalleru, AP, India.
---***---
Abstract—This paper ensures that improve the dynamic performance of a shunt type Hybrid active power filter how to compensates the harmonic contents in the network. This paper is combination of adaptive and predictive control techniques for fast convergence and reduced calculations. The adaptive and predictive properties of artificial neural networks are used for fast estimate of the compensating current. Here the dc-link voltage is used in a predictive controller to generate the first estimate the compensating current then after by convergence of the algorithm by an adaptive controller ANN i.e. adaline based network. Weights in adaline are automatically adjusted to minimize the total harmonic distortion of the source current. This paper MATLAB Simulations confirms the validity of the proposed scheme for all kinds of loads i.e. balanced and unbalanced for a three-phase three-wire system.
Index Terms—Adaptive, current control, nonlinear loads, shunt active power filter (SAPF), Shunt Hybrid active power filter (HAPF), total harmonic distortion (THD), voltage source inverter(VSI).
I. INTRODUCTION
In recent years power quality distortion becomes serious problem in electrical power systems due increasing of nonlinear loads drawing non sinusoidal currents. When source is connected to linear load there is no effect on source then system will be stable. Also when source is connected to non linear load, load will produces harmonics on load side, due to these harmonics source will become non linear. Control of ac power using thyristors and other semiconductor switches are mostly employs to fed controlled electrical power to electrical loads, such as adjustable speed drives, furnaces etc. these controllers are also used in HVDC systems and renewable electrical power generation. Harmonics in the network are not only increase the losses but also produce unwanted disturbance to the communication network, more voltage and/or current stress, etc.. Different harmonic elimination techniques are proposed, e.g., passive filter, active power line conditioner, and also hybrid active filter, have been proposed and used [1]–[8]. Recent technological advanced switching devices and availability of low cost controlling devices, e.g. DSP-/field-programmable-gate-array-based system, makes active power line conditioner a natural choice to compensate for harmonics. Shunt-type Hybrid active power filter (HAPF) is used to eliminate the current harmonics.
The dynamic performance of an HAPF is mainly dependent on how quickly and accurately the harmonic contents are extracted from the load current. Many harmonic elimination methods are available, and their performance have been explored. Proposed techniques include traditional d−q [2] and p−q theory [3]–[5] based approaches, Shunt active power filter and application of adaptive filters [6], genetic algorithm, artificial neural network (ANN) etc., for quick estimation of the compensating current [8]. Hybrid active power filter is the combination of passive filter and active power filter. Different types of combinations are there for hybrid active power filter. Here in this paper shunt type series connected hybrid active filer is proposed.
Recently, Artificial Neural Networks have attracted much different applications, including the APF. A full “neuromimetic” strategy involving several adalines has been reported by Abdslam et al. [15].Active power filters have been shown to be very interesting alternative to compensate power distribution systems. Shunt hybrid active power filter are more suitable to compensate current harmonic components and displacement power factor, while series topologies present better characteristics to compensate voltage distortions. Hybrid topology significantly improves the compensation characteristics of simple passive filters, making active power filter available for different high power applications, at relative lower cost. Better performance has been observes compared to discrete Fourier transform and fast Fourier transform, or Kalman filtering-based approaches. Tey et al. [16] reported a modified version. An additional PI controller is used to regulate the dc-link voltage. The adaptive controller can adapts for unbalance and change in working conditions.. An adaline-based harmonic compensation is reported by Singh et al. [17]. Weights are computed online by the LMS algorithm. Recently we have to seen lot of applications for artificial neural networks. Here integration of Ann based predictive and adaptive control techniques are used to generate gating pulses for voltage source inverter.
Fig. 1. (a) HAPF to compensate for a nonlinear load. (b) Single phase shunt HAPF
techniques are reported for power controllers. These are also applied to Active Power filter. The implementation of APF using power balancing at the dc-link is reported by Singh et al.[10]. The dc link voltage is used to find the magnitude of the supply current for self supporting dc bus. However, no detail analysis of the dynamics of the dc-link voltage is available. This paper is an combination of predictive and adaptive control techniques for fast convergence and reduced caluclations. Two ANN based controllers are used for such purpose.
The predictive controller generates immediately the first estimate of the compensating current quickly after the change in load is identified. The changes in voltage across the capacitor is used for this technique. This is followed by an Adeline based controller to fast converge to the steady value. This paper is presented in eight sections. Section II is presents the over view of hybrid active filter. Section III deals with the formulation of the problem. Section IV investigates the dynamics of the dc link voltage calculation . Predicting algorithm is covered in Section V. Adeline-based algorithm is presented in Section VI. Section VII presents the simulation and experimental results. Section VIII concludes the work.
II SHUNT HYBRID ACTIVE POWER FILTER
Hybrid active power filter is a combination of active power filter and passive filter. Different types combinations of hybrid active power filter(HAPF) is there i.e. Parallel and series combinations. Here in this paper series combination
shunt connected Hybrid active power filter is presented Passive filters have been extensively to compensate current
harmonics and the displacement power factor in power distribution systems. However, the poor flexibility of passive filters to adapt to variable load compensation requirements constitutes a major disadvantage. This results in power factor overcompensation in the case of low load operation. Fig .(2). Show the hybrid active filter topology since passive filter is
[image:2.594.65.520.53.320.2]connected in series to the active power filter through a coupled transformer, it imposes a voltage signal at its primary terminals that forces the circulating of current harmonics through the passive filter, improving its compensation characteristics.
Fig. 2.Hybrid Active Power Filter Configuration
[image:2.594.311.534.508.669.2]topology.
III.ESTIMATION OF CURRENT REFERENCE
Fig. 1(a) shows the HAPF compensating a nonlinear load. Fig. 1(b) shows the single phase shunt HAPF schematic diagram. A general expression for the load current (Fig. 1(b) is
iL(t) = iα1(t) + iβ1(t) + ih(t). (1)
The in phase and quadrature components of the phase current at fundamental frequency are iα1 and iβ1, respectively. All
other harmonics are included in ih. The perphase source
voltage and the corresponding in-phase component of the load current may be expressed as
vs(t) = Vm cos ωt (2)
iα1(t) = Iα1 cos ωt. (3) Assuming that the APF will compensate for harmonic and reactive power the compensating current becomes Ic(t) = iL(t) − iα1(t) = iL(t) − Iα1 cos ωt (4)
Where 𝐼𝛼1 is the peak magnitude of in phase current that the mains supply and hence needs to estimated. Once 𝐼𝛼1 is estimated over, the reference current for the HAPF may easily be set as per (4). iL(t) may measured using current sensors. In our proposed scheme, estimation of Iα1 is calculated by two stages. A single layer ANN based algorithm first predicts the value of Iα1 following which an adaline based ANN is used for fast convergence. Note that the inverter also draws a small current isα(t) to maintain the dc-link voltage. IV. CONTROL OF DC-LINK VOLTAGE IN APF The dynamics of the dc link voltage is an indirect measure of the performance of the APF. Whenever there is a change in the load, the voltage across the dc link capacitor also undergoes a corresponding change. A controller is used to keep the voltage regulated at a desire value. In this following section, a simple process of the dynamics of the dc-link voltage is first carried out. Parameters that preside over the dynamics are identified, following which an algorithm is developed it estimate the compensating current of the APF. To maintain the dc bus voltage to a desired magnitude, the capacitor draws in phase i.e., in phase with the source voltage current isα. This is in addition to the compensating current ic. From the power balance equation pdc = cdcvdc dvdc dt (5)
Where pdc is the power required to maintain the voltage Vdc across the dc link.
v
si(t)i
sαi(t) i = a,b,c − i=a,b,cRf(i2sαi(t)+ i2ci(t)) − 1/
2 i = a,b,cLf d/dt(i2sαi(t )+ i2 ci(t)) =i
dc(t)v
dc(t) =p
dc (6)Where Rf and Lf are the resistance and inductance of the inductor that is connected in between the common coupling point and the voltage source inverter. We remembers that Isα supplies the system loss at the steady state and Charges, discharges the capacitor during transient to maintain the dc link voltage. Considering that “power” is a scalar quantity, (6) for a balancedthree-phase system may be expressed as
3vs t isα t − 3Rf i2sα t + i2c t −3
2Lf d dt i 2 sα t + i2c t = idc t vdc t = pdc (7)
Applying small perturbations in is, isα, vdc and vs, the following new set of variables may be obtained: is(t) = Ic+ ∆ic (8)
isα(t) = Isα+ ∆isα (9)
vdc(t) = Vdc + ∆vdc (10)
vs(t) = Vs+ ∆vs (11)
where Ic, Isα and Vs are rms and Vdc is the dc value of the corresponding values at the operating point. After at steady state 3𝑉𝑠Isα− 3𝑅𝑓(𝐼2sα+ 𝐼2c) = 0 (12)
Substituting (8)–(12) in (7), the following equation is obtained: 3 ∆𝑣𝑠𝐼𝑠𝛼+ ∆𝑣𝑠𝑖𝑠𝛼 − 6𝑅𝑓 𝐼𝑠𝛼∆𝑖𝑠𝛼+ 𝐼𝑐∆𝑖𝑐 − 3𝐿𝑓(𝐼𝑠𝛼 𝑑∆𝑖𝑠𝛼 𝑑𝑡 +𝐼𝑐 𝑑 ∆𝑖𝑐 𝑑𝑡 ) = CdcVdc dΔVdc dt (13)
Converting the variables to s-domain and after rearranging, (13) may be expressed as ΔVdc(s )= KGs (s)G1(s)G2(s) 1+KGs (s)G1(s)G2(s) ΔV∗dc(s) − G2(s)G3(s) 1 + KGs(s)G1(s)G2(s)ΔIc(s) + 1+KGs (s)G2(s)G4(s)G2(s)G4(s) ΔVs(s) (14)
where
[image:3.594.323.540.516.684.2]K is the small signal gain. Detail derivation of (14)from(13)is available in the Appendix.
Fig. 3. Block diagram of ANN-based peak value predictor
distortion in source voltage and reference dc-bus voltage is not considered. Therefore, (14) further modifies to
ΔVdc(s) = − 1+KGs (s)G1(s)G2(s)−G2(s)G3(s) ΔIc(s). (15)
This explores the possibility of extracting an estimate of the compensating current from the change in Vdc . The gains of the PI controller used to regulate the dc-link voltage are preside over by the following two inequalities(26):
Is < cdcv∗dc
3Kp Lf (16)
Is ≤ 2R KpVs
fKp+Lf KI (17)
All the ac quantities in (16) and (17) are expressed as rms value. Is, 𝑣∗dc and Vs are the source current, reference dc-link voltage and source voltage, respectively. Equations (16) and (17) are used to generate an initial guess of Kp and
KI and also to settheir limits.
V. ANN-BASED FAST ESTIMATION OF COMPENSATING CURRENT
An ANN-based PI controller plays a important dual role. It ensures faster reference generation and is also accountable for better regulating of dc-bus voltage. The block daigram of the system (i.e. ANN-tuned adaptive PI controller) is shown in Fig. 4. To reduce computational burden, a single-layer ANN structure issued. The input vector as expressed in (18) is gives to the state exchanger. In our scheme, error voltage and its gradient are chosen as the state of the system to ensure faster corrective action
u = [𝑉∗
𝑑𝑐 𝑣𝑑𝑐]𝑇 (18)
The task of the state generator block is to generate states
x1 andx2 as follows:
𝑥1= 𝑣𝑒 𝑘 𝑥2= 𝛿𝑥1
𝛿𝑘 (19)
Where 𝑣𝑒 𝑘 = 𝑉∗𝑑𝑐− 𝑣𝑑𝑐(𝑘).The output error z(k) is represented as
z(k)= 𝑣0 (k)− 𝑣0 (k−1). (20 ) The output vo(k) is fed to output state to estimateIα1.
Neuron cell generates controlling signal through interrelated gathering as
u(k) = u(k−1) +Σ 2𝑖=1𝑤𝑖 𝑘 𝑥𝑖(𝑘) (21) where wi is the weight of the system.
Here, a neuron is trained by Hebb’s rule [27], [28]. Therefore, the change of weight of the neuron cell at kth instant may be represented as
𝑤𝑖 𝑘 + 1 = (1 − 𝑐) 𝑤𝑖(k) + η𝑟𝑖(𝑘) (22)
𝑟𝑖 𝑘 = z(k)u(k) 𝑥𝑖(𝑘 ) (23)
where ri is the progressive signal, η is Hebb’s studying ratio (learning rate), and “c” is a constant. Substituting (22) and (23) in (21), the following equation may be obtained:
Δ 𝑤𝑖(k) i(k) = 𝑤𝑖 𝑘 + 1 − 𝑤𝑖(k)
= − c[𝑤𝑖(k) –η z(k) u(k) 𝑥𝑖(𝑘)/c] (24)
Δwi(k)is the change of weight at kth step. Weights of the neuron are tuned according to Hebb’s assumption. Hebb’s assumption is popularly known as the covariance algorithm. Finally, for the PI controller, the weights are represented by
𝑤1 (k+1) = 𝑤1 (k) + η 𝐼𝑧(𝑘)𝑥1 (k) (25) 𝑤2 (k+1) = 𝑤2 (k) + η𝑃𝑧 (k) 𝑥2 (k) (26)
Whenever the ANN is initiated, it starts with a set of controller gains to generate the first estimate of the compensating current. These initial values of controller parameters are set by offline training of the ANN. The controller parameters are then adjusted following (28) and (29) to regulate the dc-link voltage.
VI . ADAPTIVE CURRENT DETECTION TECHNIQUE
The ANN in Section IV provides an initial guess for any change in system dynamics. To generate more accurate reference for APF, load current samples are fed to the adaline based network shown in Fig. 3. Adaline is designed to minimize the total harmonic distortion (THD) of source current. Uncompensated source current sample s(k) may be represented as
𝑠 𝑘 = 𝐼𝛼1cos(𝜔𝑘𝑡𝑠) +
𝐼𝛽1 sin (𝜔𝑘𝑡𝑠) + Rn=2In cos(𝑛𝜔𝑘𝑡𝑠+𝜙𝑛) (27)
Where 𝑡𝑠 is the step size in discrete domain. The square of error terms for kth sample may be expressed as
ε2(k) =
[
[s2 k − 2s k a(k)]a2(k) + 1
]
(28) Wherea2(k)=I2
α1(k)cos2 (𝜔𝑘𝑡𝑠) + I2β1(k)sin2 (𝜔𝑘𝑡𝑠) (29)
Equation (31) may also be represented as
ε2 (k) = {s
2(k)−2s k [XT k α k α−T k X(k)]}
XT k α k α−T k X(k) + 1 (30)
where the vector
α(k) = [𝐼𝛼1 𝑘 , 𝐼𝛽1 𝑘 ] (31) and the input vector
X(k) = [cos 𝜔𝑘𝑡𝑠 , sin 𝜔𝑘𝑡𝑠]𝑇 (32)
VI. SIMULATION RESULTS
Simulations have been conducted for balanced and unbalanced loads using SIMULINK for different controller configurations. Switching frequency of the inverter is set at 10 kHz, and the dead time of the inverter is set at 1μs.The whole system is built in SIMULINK where the ANN routine is called whenever necessary. A 110-V 50 HZ mains supplying a load of 3 KVA is considered. First, simulation study is made for the case with only predictive algorithm. Fig.(4) shows the waveform for balanced and nonlinear load. A diode bridge feeding a highly inductive network is treated as the nonlinear load. The figure shows the source, load, and compensating currents in top-to-bottom order. Load change has occurred at 5 ms. The initial estimate of the source current is extracted from the dip in capacitor voltage (according to the algorithm explained in Section IV). Although a quick estimate helped, the waveform quality is poor due to the lack of any corrective mechanism in the system
.
Fig.4. (a)
Fig. 4. (b)
Fig. 4. (c) Fig.4. (a) Source current of phase A (b) Load current of phase A
(c) Compensating current of phase A.
Balanced three-phase nonlinear load is considered similar to the case with predictive algorithm. Fig. (4) show the
Performance of the HAPF with predictive ANN (simulation results). Fig.4.(a) shows Source current of phase A. Fig.4.(b) shows Load current of phase A. Fig.4.(c) shows Compensating current of phase A.
Next, the adaptive algorithm is tried. Simulation is conducted to check the performance of the system for a step change in load.
Fig. 5. (a)
Fig. 5. (b)
Fig. 5. (c) Fig.5. (a) Source current of phase A (b) Load current of phase A
(c) Compensating current of phase A.
Fig. (5) show the Performance of the HAPF with adaptive
controller (simulation results). Fig.5.(a) shows Source current of phase A. Fig.5.(b) shows Load current of phase A. Fig.5.(c) shows Compensating current of phase A.
Now, to have the advantage of predictive and adaptive controllers, the system is run with both the algorithms. Fig. (6) show the simulation with both the predictive and adaptive controllers in operation. The results have confirmed very satisfactory performance in terms of waveform quality and response time. The THD value for the simulation result is 1.49%. Fig. (6) show the simulation daigram for the Performance of the HAPF with predictive and adaptive controllers Fig.6. (a) shows Source current of phase A.
[image:5.594.300.558.102.528.2] [image:5.594.38.285.281.644.2]Fig.6. (a)
Fig.6. (b)
Fig.6. (c) Fig.6. (a) Source current of phase A (b) Load current of phase A
(c) Compensating current of phase A.
Fig. (7). THD value of HAPF with predictive and adaptive controllers (simulation results).
Fig.(7). Shows the percentage of THD value of HAPF with predictive and adaptive controller. The result of THD have confirmed very satisfactory performance in terms of waveform quality and response time Fig. (8). Shows the comparison of THD values of shunt active power filter and hybrid active power filter. This Comparision shows the performance of HAPF is more better than shunt active power filter
.
SAPF
%THD OF SOURSE CURRENT
HAPF %THD OF SOURSE
CURRENT predictive
controller 6.02
predictive
controller 4.97
adaptive
controller 3.73
adaptive
controller 2.95
predictive and adaptive controllers
2.25
predictive and adaptive controllers
1.49
Fig. (8). Comparision of THD values for SAPF and HAPF
VIII. CONCLUSION
The combination of adaptive and predictive ANN
based controller for a shunt type HAPF has been presented in this paper to improve the dynamic performance, convergence and reduce the computational requirement. The predictive algorithm is derived from an ANN-based PI controller used to regulate the dc-link voltage in the HAPF. This is followed by
an adaline based THD minimization method. Adaline is
trained by CG method to minimize THD. Use of only two weights and two input vectors makes the convergence very fast and simple. The system is extensively simulated in MATLAB/SIMULINK model.
APPENDIX
Taking Laplace transformation of (13), the following equation may be obtained:
3ΔVs(s)Isα +ΔIsα(s)(3Vs − 6Rf Isα − 3sLf Isα)
−ΔIc(s)(6Rf Ic + 3sLf Ic) = sΔVdc(s)CdcVdc. (A1)
Consider
G1(s) = − 3(Vs − 2Rf Isα + sLf Isα) (A2)
G2(s) = 1
[image:6.594.37.289.34.484.2] [image:6.594.312.549.153.336.2]G3(s) =3(2Rf Ic + sLf Ic) (A4) G4(s) =3Isα. (A5) Substituting (A2)–(A5) in (A1), the following equation may be derived:
ΔVs(s)G4(s) + ΔIsα(s)G1(s) − ΔIc(s)G3(s)
= ΔVdc(s)G2(s)1 (A6)
The relation between isα and vdc may be expressed as
isα = Kdv dcdt (A7) where K is the small-signal gain and isα is the part of the source current used to stabilize the dc-bus voltage. Instantaneous dc link voltage is compared with the reference dc voltage, and the error voltage in Laplace domain may be expressed as
Ve(s) = V∗dc
s − Vdc(s). (A8)
A PI controller may be used to maintain the capacitor voltage. Therefore, the output of the PI controller may be expressed as
Vo(s) = (KP + KIs )Ve(s). (A9)
Thus, (A7) may be modified as ΔIsα(s) = KGs(s)(ΔV ∗dc
s − ΔVdc(s)) (A10)
where
(KP + KIs) = Gs(s). (A11)
Substituting values of ΔIsα(s) in (A1), (14) is obtained (in Section IV).
REFERENCES
[1] B. Singh, K. Al-Haddad, and A. Chandra, “A review of active power filters for power quality improvements,” IEEE Trans. Ind. Electron., vol. 46, no. 5, pp. 960–971, Oct. 1999.
[2] S. Bhattacharya, T. M. Frank, D. M. Divan, and B. Banerjee, “Active filter system implementation,” IEEE Ind. Appl. Mag., vol. 4, no. 5, pp. 47–63, Sep./Oct. 1998.
[3] H. Akagi, Y. Kanazawa, and A. Nabae, “Instantaneous reactive power compensators comprising switching devices without energy storage com-ponents,” IEEE Trans. Ind. Appl., vol. IA-20, no. 3, pp. 625–630, May 1984.
[4] F. Z. Peng, G. W. Ott, Jr., and D. J. Adams, “Harmonic and reactive power compensation based on the generalized instantaneous reactive power the-ory for three phase four wire system,” IEEE Trans. Power
Electron., vol. 13, no. 6, pp. 1174–1181, Nov. 1998.
[5] R. S. Herrera and P. Salmeron, “Instantaneous reactive power theory: A reference in the nonlinear loads compensation,” IEEE Trans. Ind. Elec-tron., vol. 56, no. 6, pp. 2015–2022, Jun. 2009.
[6] T.Mahalekshmi “Current Harmonic Compensation and Power Factor Improvement by Hybrid Shunt Active Power Filter” International Journal of Computer Applications (0975 – 8887) Volume 4 – No.3, July 2010.
[7] Elsa Susan Daniel and G.Abirami “Selective Harmonic Elimination Using Shunt Hybrid Active Power Filters Operating At Different Switching Frequencies” International Journal of Innovative Research In Electrical & Electronics, Instrumentation And Control Engineering Vol. 1, Issue 1, April 2013.
[8] L. Chen and A. Jouanne, “A Comparison and Assessment of Hybrid Filter Topologies and Control Algorithms”, Proceedings of The IEEE Power Electronics Specialists Conference (PESC), June 17-21, 2001, Vancouver, Canada: IEEE Trans. 2001, Pp. 565-570.
[9] D.Rivas, L.Moran, J. Dixon and J.Espinoza, “A simple control scheme
for hybrid active power filter”, IEE proc-Gener. Transm. Distrib, Vol. 149, No 4, July 2002.
[10]B. N. Singh, B. Singh, A. Chandra, and K. Al-Haddad, “Design and
digital implementation of active filter with power balance theory,” Proc
Inst. Elect. Eng.—Electr. Power Appl., vol. 152, no. 5, pp. 1149–1160,
Sep. 2005.
[11]A. Hamadi, S. Rahmani, and K. Al-Haddad, “A hybrid passive filter configuration for VAR control and harmonic compensation,” IEEE Trans.Ind. Electron., vol. 57, no. 7, pp. 2419–2434, Jul. 2010.
[12] B. Widrow and M. A. Lehr, “30 years of adaptive neural networks: Perceptron, madaline and back propagation,” Proc. IEEE, vol. 78, no. 9, pp. 1415–1442, Sep. 1990.
[13] L. H. Tey, P. L. So, and Y. C. Chu, “Improvement of power quality using adaptive shunt filter,” IEEE Trans. Power Del., vol. 20, no. 2, pp. 1558– 1568, Apr. 2005
[14] power filters,” IEEE Trans. Ind. Electron., vol. 54, no. 1, pp. 61–76,
Feb. 2007 D. O. Abdeslam, P. Wira, J. Merckle, D. Flieller, and Y. A. Chapuis, “A unified artificial neural network architecture for active.
[15] L. H. Tey, P. L. So, and Y. C. Chu, “Improvement of power quality using adaptive shunt filter,” IEEE Trans. Power Del., vol. 20, no. 2, pp. 1558– 1568, Apr. 2005.
[16] B. Singh, V. Verma, and J. Solanki, “Neural network-based selective
compensation of current quality problems in distribution system,” IEEE
Trans. Ind. Electron., vol. 54, no. 1, pp. 53–60, Feb. 2007.
