Professional Documents
Culture Documents
Abstract: As a promising renewable alternative, a significant contribution to a photovoltaic power generation system in the future,
but also brought new problems related to the integration of power quality, including voltage control and reactive power compensation.
This study proposes a new adaptive neural network DC / AC converter-based control. In this study, the use of neural network control
to increase the DC / AC converter transient response. Aiming at these problems, explore the effect of reliability and power quality
and reactive power compensation system optimization, and propose coherent active and reactive power distribution in order to
alleviate this problem. In an area of photovoltaic power plant simulation. Simulation results verify the effectiveness of the method.
The results show that the proposed method helps to achieve this adaptive control.
Key words: Renewable energy, reactive power compensation, Harmonic Detection, Neural Networks.
Modeling of photovoltaic generators in a different The company can be changed to, dP / dVPV 0.
way. These models differ in some aspects, such as the
3690
Simulation is based on the data sheet (single crystal, is obtained by inverse coordinate transformation by using a
pp1237, 37wp) parameters of PV modules and solar phase locked loop (PLL). Knowledge of the values of
modules are listed in table (1). ^
ec
eb
Harmonic currents generated in the supply system and ea
installation. Voltage distortion caused by harmonic R
voltage, and over zero noise network [25-27]. Harmonics is PLL
THDV
f
n 2 Vn
2
4
network faults use of the area. For this reason, the artificial
neural network is generally in sensor verification, testing
V1 and intelligent monitoring applications. Figure 1 shows the
basic topology used artificial neural networks.
THDI
f 2
n 2 In
5
Auto-associative neural network topology has 19 inputs,
I1 five hidden layer, 19 outputs. In FIG. 3, 19 represents a
valid line power system harmonics.
where V1 and I1 Rms value of the fundamental
component of voltage and current, while Vn and I n the Inputs Hidden Layer
RMS harmonic component of voltage and current.
C. PV inverter Outputs
The configuration of the grid-connected inverter is
shown in Fig. 2. The superscript ^ refers to the estimated
variables and quantities, and the superscript * denotes Harmonic
the reference variables and quantities. The current
( ia , ib , ic ) and voltage ( ea , eb , ec ) variables transform
Fig. 3 Representative topology of the ANN
from the three phase stationary coordinate system to the Provided neural network input signal:
two-phase rotating coordinate system. The grid angle ( T )
3691
1, cos( wn), sin(wn), Use a neuron by neuron training algorithm
cos( 2wn), sin(2wn), convergence has significantly improved compared to the
conventional error backpropagation training fcc. Also,
X ( n) (6)
# before Lauenburg Marquardt training and NBN
cos( Mwn), sin(Mwn) comparison can be drawn fed neural network better
Network right value: training program fully connected.
1 l 1 l m (7) IV. HARMONIC DETECTION
E [dk f (netk )]2
2k1
2k1
[d k f ( w jk y j )]2
j 0
Cage herein derived average switching model and
Calculation output: synchronous rotating coordinate system control three-phase
y ( n) X ( n ) W ( n ) inverter with LCL type. With these models, the grid
M
(8) voltage grid current impact of analyzed and it was
i0 [ I pm cos( nwn ) I qm sin(nwn )] discovered that not only directly affected by the grid
m 1 current grid voltage, but also by injection grid voltage grid
I p1 (n N 1) I p1 (n N 2) " I p1 (n) current reference error. This work provides the basis for
I q1 (n N 1) I q1 (n N 2) " I q1 (n) suppression of harmonics and unbalance the inverter grid
connection in the subsequent chapters of the injection gate
I p 2 (n N 1) I p 2 (n N 2) " I p 2 (n)
current. Instantaneous active power can be determined as:
(9)
C I q 2 (n N 1) I q 2 (n N 2) " I q 2 (n)
p V i Va ia Vbib Vc ic 12
# # #
I
pM ( n N 1 ) I pM ( n N 2 ) " I pM ( n ) Then, the instantaneous active current vector i p is
I
qM ( n N 1) I qM ( n N 2) " I qM ( n ) defined as
Provided data window of length N, enter a set length iap
of historical data during adaptive prediction. In a neural p
network, the latest set of network power N value is ip ibp V 13
obtained: t n N 1, t N 2 , , Right Calculated i V V
values of n timeas shown in the following matrix cp
Since the three-phase current vector is always equal to
Where I p and i q subscript m represents the first the sum of i p and iq ,the instantaneous compensating
harmonic; in parentheses represents the n-th time. *
Therefore, a harmonic order of each row represents a current reference ic can be written as
calculated value of each time period; each column p
represents the calculated value at a time for each order ic* iq i p i V i 14
harmonics. In order to predict the harmonics of the V V
Thus, for any given three-phase current, instantaneous
n 1 time, just find the matrix continuation to the harmonic current compensation command shown in Figure
first n 1 . 2. In addition, non-linear loads and unbalanced loads can
Historical error before the N 1 time and n time cause the power section, around the average active power,
error is calculated as follows: P oscillation power section without any contribution, active
and should be removed. This is part of a low-pass filter can
'w jk KG ko y j K (d k ok )ok (1 ok ) y j be used to extract the active power flow (LPF).
l
'vij KG j xi K (G k w jk ) y j (1 y j ) xi
y o A. Detection Structure
k 1 Figure 4 shows the harmonic detection using adaptive
(10) neural network implementation. Constituted by the PLL
Where k 0,1, " , N 1 .Total error: frequency voltage, which is the basic frequency, adaptive
l neural network prediction algorithm, which can give more
G jy (G ko w jk ) y j (1 y j ) (11) harmonic component and the fundamental active or
k 1 reactive currents.
Neural Networks
B. Adaptive neural network training Weight adjustment Harmonic
ANN components
Phase
Each output artificial neural network is possible the Voltage Phase-Locked
Shift
Adaptive Negative sequence
Loop Prediction
range [-1,1]. Output trained artificial neural network, if the Array
Algorithm
Fundamental
Reactive
difference exceeds a threshold between PV output and
Current
reduce output over - take 1, take training and value + 1 if
the difference is less than the threshold. Threshold can be + -
Fig. 4 Harmonic Detection Chart
set to any value between the minimum and maximum 1 to
19. If, on the other hand, is set to a higher value, and then, V. SIMULATION RESULTS
the result is an ideal value calculated at the intensity level
In order to verify the performance of the detection
is correct.
algorithm, through the use of simulation software for the
3692
photovoltaic power station psacd. Single-phase 220V This work was supported by the Qinghai Province key
power supply voltage rms, and use the PV power plant and Laboratory of Photovoltaic Power Generation and
grid-connected inverter as a harmonic source. Grid-connected Technology (Grant No. 2014-Z-Y34A).
In a phase current, Figure 6 is a reactive current References
waveform, Figure 7is a current waveform, containing a
[1] NOROOZIAN R, GHAREHPETIAN G B. An investigation on
reactive power. This can be seen from the two figures, this combined operation of active power filter with photovoltaic arrays
detection method can quickly detect the harmonic currents [J]. Electrical Power and Energy Systems, 2013, 46: 392399.
within 5 ms from the figure this detection method can be [2] REISI A R, MORADI M H, SHOWKATI H. Combined
used for active power filter harmonic current detection, has photovoltaic and unified power quality controller to improve power
quality [J]. Solar Energy, 2013, 88: 154162.
a good detection accuracy and tracking speed. [3] PENG Fang-zheng, OTT G W, ADAMS D J. Harmonic and
reactive power compensation based on the generalized
instantaneous reactive power theory for three-phase four-wire
systems [J]. IEEE Transactions on Power Electronics, 1998, 13(6):
11741181.
[4] PENG Fang-zheng, LAI J S. Generalized instantaneous reactive
power theory for three-phase power systems [J]. IEEE Transactions
on Instrumentation and Measurement, 1996, 45(1): 293297.
[5] Aten M, Werner H (2003) Robust multivariable control design for
(a)A phase Grid-connected voltage HVDC back to back schemes. IEEE Xplore-Gener Transm Distrib
150(6):761767
[6] Skogestad S, Postlethwaite I (2005) Multivariable feedback control:
analysis and design. Wiley, Chichester
[7] Teodorescu R, Blaabjerg F, Liserre M et al (2006)
Proportionalresonant controllers and filters for grid-connected
voltage-source converters. IEEE Xplore-Electr Power Appl
153(5):50762
[8] Y. Xu, L. M. Tolbert, J. N. Chiasson, F. Z. Peng, and J. B.
bA phase Harmonic currents Campbell, Generalized instantaneous non-active power theory for
Fig.6 Waveforms of harmonic current without reactive power STATCOM, IET Electr. Power Appl., vol. 1, no. 6, pp. 853861,
Nov. 2007.
[9] M. Rezaei, H. Iman-Eini, and S. Farhangi, Grid-connected
photovoltaic system based on a cascaded H-Bridge inverter, J.
Power Electron., vol. 12, no. 4, pp. 57886, Jul. 2002.
[10] Y. Shi, L. Liu, H. Li, and Y. Xue, A single-phase grid-connected
PV converter with minimal dc-link capacitor and low-frequency
ripple-free maximum power point tracking, in Proc. 5th IEEE
Energy Convers.Congr. Expo., Denver, CO, Sep. 1519, 2013, pp.
aA phase Fundamental active current 23852390.
[11] L. Liu, H. Li, Y. Xue, and W. Liu, Decoupled active and reactive
power control for large scale grid-connected photovoltaic systems
using cascaded modular multilevel converter, to be published
IEEE Trans. Power Electron. 2014.
[12] D. Aquila, M. Liserre, V. G. Monopoli, and P. Rotondoi,
Overview of PI-based solutions for the control of dc buses of a
single-phase Hbridge multilevel active rectifier, IEEE Trans. Ind.
Appl., vol. 44, no. 3, pp. 857866, May/Jun. 2008.
bA phase Harmonic currents [13] H.T.Wu and Y. H. Liu, Novel STATCOM Control Strategy for
Fig.7 Waveforms of harmonic current with reactive power Wind Farm Reactive Power Compensation, IEEE Transactions on
Energy Conversion 2011 IEEE.
[14] Chong Han, Alex Q. Huang, Mesut E. Baran, STATCOM Impact
VI. CONCLUSIONS Study on the Integration of a Large Wind Farm into a Weak Loop
This paper discusses the influence of reactive power Power System, IEEE Transactions on Energy Conversion, vol. 23,
Issue 1, Mar. 2008,pp.226233.
compensation grid interactive photovoltaic system cascade
[15] Zhen Xiang, Da Xie, Jinxia Gong, YanChi Zhang, Dynamic
operational performance. Analysis of system stability and Charactcristics Analysis of STATC0M for Reactive ComPensation
reliability problems caused by asymmetric active. Reactive in Wind Farm, Automation of Electric Power systems, vol. 32,
power compensation and allocation, to alleviate this Issue 9, May10,2008,pp.9295.
[16] S.Kahrobaee, S. Afshania, V. Salehipoor (University of Tehran),
problem. The output voltage of each module is verified to
Reasonable Reactive Power Control and Voltage Compensation
determine the direct power distribution. Distribution of the forWind Farms Using FACTS Devices, Nordic Wind Power
relationship between the voltage and power distribution, Conference 22 23 May 2006, ESPOO, Finland.
shows a wide range of power. [17] Ahmed Maria, Mauro Facca & John Diaz De Leon IESO
Philosophy On Reactive Power Compensation North American
ACKNOWLEDGEMENTS WINDPOWER June 2007 issue.
3693