PATTERN RECOGNITION FOR FAULT DIAGNOSIS OF

SOLAR POWER INVERTER BY TRAJECTORY IMAGE

UNDERSTANDING

Jaeho Hwang

, Nanhwa Kim

, Neajoung Kwak

and Wonpyo Hong

Dept. of Electronic Eng., Hanbat National University, Daejeon 305-719, Korea

Dept. of Building Services Eng., Hanbat National University, Daejeon 305-719, Korea

Keywords: Pattern Recognition, Image Understanding, Fault Diagnosis, Decision Tree.

Abstract: This paper presents an approach based on pattern recognition to detect and diagnose faults of solar power

inverter by its fault trajectory image understanding. The drive system for simulation is modeled using

Matlab Simulink toolboxes. Solar power device uses control/filter structure to connect the pulse width

modulation (PWM) inverter. Multistage diagnosis factors are calculated from faults patterning procedure. It

is based on the analysis of the vector trajectory and of the space syntax in faulty image mode.

1 INTRODUCTION

A Solar power inverter is a type of power electronics

inverter that is made to convert the DC electricity

from photovoltaic solar cells into AC sinusoidal one

under various kinds of load, building appliances and

a utility grid. In case of medium voltage solar

inverter, switching mode PWM converters have

been widely used because of its high efficiency and

output power. However, electrical faults may exist in

any component of the drive system. Once abrupt

faults such as the breakdown of switching devices,

the failure of the capacitor or inductor in the low-

pass filter occur, the whole system will lose its

operation and even propagate a series of troubles to

whole power system. In addition to electronic

troubles, the solar plant output voltage varies in a

wide range. It has to be converged within a specified

range in use of controller and a big input capacitor

which is connected in parallel to the solar cells in

order to fit the input voltage of DC-to-AC power

inverter. The troubles in this device and input

voltage ripple are the additional faults of solar power

inverter.

The recent researches on fault diagnosis of

power inverters have been focused related to three-

phase induction motor derive system, inverter faults

in variable speed AC drives, load short/open circuit

and mechanical/insulation failure of the induction

motor(Guan et al., 2007, Son et al., 2004, Ye and

Wu, 2001). The method for fault detection and

diagnosis is mainly based on the current trajectory

and its instantaneous frequency at the output side of

the inverter. However, with the increasing concern

about natural energy source and environmental

demand, the need to produce the green energy such

as solar energy to replace fossil fuels has

significantly increased. In an effort to utilize the

solar energy, photovoltaic(PV) generation with

power inverter is an effective method to supply the

flexible power in a grid, not only AC motors,

connected small and large power generation plants.

This paper develops a real-time faulty diagnosis

method for photovoltaic inverter system based on

the pattern recognition of current vector calculation.

The performance and characteristic of the system

faults are evaluated as the type of source image

parameters for patternization.

2 SYSTEM AND FAULT

ANALYSIS

2.1 System Schemes

The scheme of the proposed PV inverter system can

be built in different ways, depending on the size of

the system and on the desired energy management. It

is composed of photovoltaic module, electrical,

electronic device for control, DC/DC converter,

477

Hwang J., Kim N., Kwak N. and Hong W. (2010).

PATTERN RECOGNITION FOR FAULT DIAGNOSIS OF SOLAR POWER INVERTER BY TRAJECTORY IMAGE UNDERSTANDING.

In Proceedings of the International Conference on Computer Vision Theory and Applications, pages 477-480

 SciTePress

inverter and grid. The output power produced by

photovoltaic modules is significantly affected by the

weather conditions. The PV sources depend on solar

radiation and cell temperature during sun hour

period. In order to extract maximum output power

from PV source, solar power controller has to be

incorporated to regulate the voltage to the Maximum

Power Point Tracking(MPPT)(Bellini et al., 2008).

This is achieved through controller of PV inverter

system. Due to vary solar output, night or

intermittent sun condition, battery is used to work as

a standalone power source charged from PV source.

It can be also connected to Grid sources as a backup

source.

The single-stage for PV inverter system,

illustrated in Fig.1, is employed to supply AC power

to an available load. The output voltage is quite low

and the MPPT is controlled by the inverter.

Figure 1: Single-stage PV inverter system.

On the utility inverter, a high frequency Pulse

width modulated(PWM) inverter is designed and

applied in order to maintain a power factor and low

harmonic current, where the stored DC power in the

battery is digitized to produce a sequence of PWM

pulses at the output of inverter.

The general scheme for simple PV inverter is

shown in Fig.2. Here the controller or controller

with charge regulator is incorporated to regulate the

output power of PV source.

Figure 2: Grid tie PV inverter system.

The PV output voltage, which can vary in a wide

range, has to be isolated in order to fit the input

voltage of inverter. In case of Fig.2, it is realized by

controller or charge-regulator with controller.

The schematic block diagram, where the grid

system and PV system works as primary and back-

up source, in shown Fig.3.

The system switch connects the inverter during

sun hour and PV supplies electric power to the load

through inverter. During night or intermittent sun

condition, the grid source alone charges the battery

and the load gets its power from the grid. The

system switches over to the grid during inverter cut

off.

Figure 3: Grid tie PV inverter system.

2.2 Fault Analysis

Various types of faults might occur on the different

parts of the PV inverter system, which include the

following:

ⅰ) PV source;

ⅱ) Controller, charge regulator or rectifier:

breakdown of device, faults in the circuit ;

ⅲ) Battery: capacitor breakdown;

ⅳ) Load: one phase-missing, close to ground, short,

open;

ⅴ ) PWM inverter: faults on turning on-off or

thyristor devices;

ⅵ) Grid: switching fault, short;

In this paper, only faults occurred in the invert,

controller, load and grid are considered. The other

faults will be studied in another paper. Switching

faults of PWM inverter, open or short circuit faults

of control devices and defects on load and grid are

analyzed in diagnosis model.

The grid system and PWM inverter are three-

phase devices. The signals used for diagnosis model

are inverter output three-phase currents. The faults

in the inverter drive system due to electrical or

switching causes are reflected in the current wave

form at the output side of the inverter. The three-

phase signals are transformed into a two-phase

rector space by Concordia transform. It transforms a

three-phase system into a further simplified two line

currents based on

abc

i i i  





(1)

i i i





(2)

The trajectory of vector

( , )i i i





periodically preserved image on



plane because

the trajectory for each fault mode is unique.

VISAPP 2010 - International Conference on Computer Vision Theory and Applications

478

3 DIAGNOSIS MODEL AND

PATTERN RECOGNITION

The diagnostics of PV inverter which is designed to

draw photovoltaic energy from a battery can be

implemented by checking the trajectory image of the

current vector in



plane(Peuget et al, 1998). It

has its own trajectory mode, normal or faulty one.

They can be plotted for the current of the inverter

and different patterns are easily classified when

switching device faults occur in the inverter. Each

trajectory image represents full or half circle under

ideal operation condition. In case of one switching

fault, the phase currents of the load are no longer

sinusoidal. The current of that phase can flow in one

direction. The phase voltage during the half period

does not appear because the phase current does not

flow in positive or negative state. The phase current

of fault device is null during half of the current

period. The relation between



and



3ii





by equation (1) and (2). Therefore the

corresponding trajectory is a half circle in



plane.

3.1 Diagnosis Parameters

But both switching devices are faulty, the trajectory

becomes a narrow line or a sector within a right

angle. If both devices belong to one phase, the

trajectory moves the



axis for phase A and

becomes one line with 120 degree delay for phase B

and C. If both devices belong to different phase, the

trajectory becomes a sector within a right angle. The

trajectory refers to a fault mode. In order to easily

identify a fault mode of the PV inverter, four

parameters related to the trajectory image are to be

proposed within a normalized unit circle; shape,

region, distributed angle and typical vector angle.

 Shape: line, fanwise sector

 Region: six regions (Fig. 4)

Figure 4: Trajectory regions.



Distributed angle



(Fig. 5(a)):

60 ( ),mm

1, 2,



Typical vector angle



(Fig. 5(b)):

30 ( ),nn

2,… ,11

Typical vector is decided to halve the distributed

angle.

(a)



(b)



Figure 5: Distributed/typical vector angle (examples).

3.2 Tree for Pattern Classification

In fault analysis, a tree diagram is used as a pattern

recognition model which maps observation about

faults to pattern. The goal is to create a model that

classifies the faults based on the proposed diagnosis

parameters. Each interior node corresponds to one of

the parameters, that is, input variables. This tree can

be learned to refer to the outcome of the fault pattern

(Table1, Fig. 6).

Table 1: Tree configuration.

node 1

node 2

()

pattern

line

b11,90

b12, 150

b13, 210

sector

node 3,

()





node 4,

()





b21, 60

b211~ b216,

2nk

D4~D9

b22, 120

b221~ b226,

21nk

D10~D1

b23, 180

b231~ b236,

2nk

D16~D2

0,1,2, ,5, :n b branch    

PATTERN RECOGNITION FOR FAULT DIAGNOSIS OFSOLAR POWER INVERTER BY TRAJECTORY IMAGE

UNDERSTANDING

479

Figure 6: Tree and pattern recognition.

3.3 Simulation for Faulty Pattern

In general the PV inverter is a combination of

electric and electronic devices. The proposed

simulation model for PV inverter is grouped into

several modules provided in Matlab Simulink as

shown in Fig.2. The load current is simulated using

three phase inverter module in SimPowerSystem

library. In table2, the patterns for different faults

modes are listed. The fault refers to the open-

circuited of the relevant power-electronic device.

The devices E1 and E6 are respectively upper and

lower device of phase A. E3 and E2 are related to

phase B, and E5 and E4 refer to upper/lower device

of phase C.

Table 2: Fault patterns.

pattern

open-

circuited

devices

pattern

open-

circuited

devices

pattern

open-

circuited

devices



D15



D16



D10



D17



D11



D18



D12



D19



D13



D20



D14



D21

: and

4 CONCLUSIONS

This paper proposed a method for patternization and

its graphic recognition based on the analysis of

trajectory modes image understanding when three

phase PV inverter faults occur. System schemes and

diverse fault mode are introduced. After parameters

for diagnosis are identified, a decision tree is

composed. The fault pattern can easily diagnose the

each switching fault. This knowledge-based method

has been tested in simulation using Matlab Simulink

toolboxes. The proposed method can apply to an

experimental system.

REFERENCES

Guan, Y., Sun, D., He, Y.(2007).Mean Current Vector

Based Online Real-Time Fault Diagnosis for Voltage

Source Inverter fed Induction Motor Drives, IEEE

International Electric Machines & Drives

Conference,2,1114 – 1118.

Son, H.I., Kim, T.J., Kang, D.W., Hyun, D.S.(2004). Fault

Diagnosis And Neutral Point Voltage Control When

The 3-Level Inverter Fault Occur, IEEE Power

Electronics Specialists Conference

Ye, Z. and Wu, B.(2001). Simulation of Electrical Fault of

Three Phase Induction Motor Drive System, IEEE

Power Electronics Specialists Conference ,1,75 – 80.

Bellini, A., Bifaretti, S., Iacovone, V.(2008). Resonant

DC-DC Converters for Photovoltaic Energy

Generation Systems, SPEEDAM 2008 International

Symposium on Power Electronics, Electrical Drives,

Automation and Motion, 815-820.

Peuget, R., Courtine, S., Rognon, J.(1998). Fault Detection

and Isolation on a PWM Inverter by Knowledge-

Based Model, IEEE Transactions on Industry

Applications, 34(6), 1318-1326.

VISAPP 2010 - International Conference on Computer Vision Theory and Applications

480