Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
CA 02823546 2013-07-02
SOUND ABNORMALITY DETECTION SYSTEM FOR POWER DEVICE AND
METHOD THEREOF
FIELD OF THE INVENTION
[0001] The invention relates to the field of electric power device
detection, and
particularly to a sound abnormality detection system and method for a
substation device.
BACKGROUND OF THE INVENTION
[0002] At present, normal running of the substation still relies mainly on
manual on-site
patrol inspection and maintenance. This way has two problems as follows: 1. a
safety problem
of a patrol inspection staff, the substation is a dangerous environment which
is full of high
voltage and discharge, which is prone to cause injuries to the body of the
patrol inspection
staff; 2. an effect problem of the manual patrol inspection, the staff is
prone to get tired and
slack when he/she performs repeated mechanical works, which may lower the
patrol
inspection effect and quality of the patrol inspection, moreover, the manual
patrol inspection
can not meet the suddenness and timeliness of a fault that occurs to the
substation device.
Therefore, the effect of the manual patrol inspection is instable, and the
reliability thereof can
not be improved.
[0003] Some substations begin to try to use an automatic, intelligent system
to assist or
replace the manual patrol inspection, and demand for the unattended substation
is increasing.
In the aspect of sound detection for a substation device, researches have been
made mainly on
substation device monitoring method and an ultrasonic detection method. The
substation
device monitoring method relies mainly on an audio sensor and a central
processor mounted
on the device body, which has two disadvantages as follows: 1. it requires a
fixed point and
closely attaching to the body, therefore the arrangement position of the audio
sensor must be
fixed, which is inflexible; 2. each substation device needs to be provided
with a separate
audio sensor, too many sensor devices will lead to arrangement difficulties
and higher cost.
The ultrasonic detection method has the disadvantages of high cost and being
unable to
achieve real-time continuous detection.
- -
CA 02823546 2013-07-02
SUMMARY OF THE INVENTION
[0004] The technical problem to be solved by the invention is to provide a
sound
abnormality detection system and method with flexible detection and low cost,
which mainly
uses a technique of sound signal processing and pattern recognition to perform
sound
detection to a substation device, to determine whether there are abnormalities
or internal
faults in the device, thereby maintaining normal operation of the substation
and meanwhile
ensuring the safety of a staff.
To solve the technical problem described above, the technical solution used by
the
invention is as follows.
1. First technical solution
[0005] A sound abnormality detection system for an electric power device, the
system
includes an audio detection subsystem and a computer subsystem wirelessly
communicating
with the audio detection subsystem.
[0006] The audio detection subsystem includes an audio collection module, a
video service
module, an audio training and identifying module, a storage module, and a
first
communication module; an output of the audio collection module is connected
with an input
of the video service module, an output of the video service module is
connected with a
corresponding input of the audio training and identifying module, a
corresponding output of
the audio training and identifying module is connected with an input of the
first
communication module, and the storage module is connected with the audio
training and
identifying module bidirectionally.
[0007] The computer subsystem includes a host computer installed with a base
station
backstage software system, an audio data display module, an alarm module and a
second
communication module; an output of the second communication module is
connected with a
corresponding input of the host computer; corresponding outputs of the
computer host are
respectively connected with an input of the audio data display module and an
input of the
alarm module.
[0008] Preferably, the system further includes a patrol inspection platform,
the patrol
inspection platform includes a walking mechanism and a control module for
controlling the
walking mechanism and the audio detection subsystem is provided on the patrol
inspection
- 2 -
CA 02823546 2013-07-02
platform.
[0009] The audio training and identifying module includes an ARM9
microprocessor and a
DSP microprocessor.
[0010] The patrol inspection platform uses an electric power patrol inspection
robot with a
model of LER-3000.
2. Second technical solution
[0011] A detection method applied in the sound abnormality detection
system for an
electric power device, which includes the following steps:
(1) starting a electric power patrol inspection robot;
(2) moving the electric power patrol inspection robot to an electric power
device
patrol inspection point according to a designated route, triggering an
initiator of the audio
collection module, so that the audio collection module collects audio data of
the electric
power device;
(3) receiving, by a video service module, the audio data collected by the
audio
collection module, performing digital signal compression and transmission;
(4) receiving, by an audio training and identifying module, the audio data
transmitted by the video service module, and performing training or
identification to the audio
data, which includes:
[0012] firstly determining whether to perform training on the audio data, and
if a training is
required, entering a training program to train the operating sound of a normal
substation
device so as to obtain an audio feature library, and storing the audio feature
library in the
storage module; and
[0013] if no training is required, entering an identifying program, in which
the DSP
microprocessor firstly extracts a sub-band feature of the currently collected
audio data, and
then compares the sub-band feature with the audio feature library stored in
the storage
module;
[0014] (5) transmitting, by the audio training and identifying module, a
comparison result to
the host computer by a real-time communication between the first communication
module and
the second communication module, wherein if the comparison result is abnormal,
a alarm
- 3 -
CA 02823546 2013-07-02
module sounds an alarm sound, and an audio data display module displays the
waveform data
and the sub-band feature of the audio; otherwise, only the waveform data and
the sub-band
feature of the audio are output; and
[0015] (6) determining whether an patrol inspection route has been inspected
completely,
and if the patrol inspection route have been inspected completely, ending;
otherwise,
performing step (2).
[0016] The audio detection subsystem in the invention gives the detection
result mainly for
abnormalities or internal faults that may occur in the substation and can be
determined from
the sound. Substation device faults capable of being reflected by a sound
mainly include:
device internal discharge, insulator discharge, electromagnetic vibration
sound abnormality,
fan or oil pump electric motor rotation abnormalities, and capacitor sound
abnormality, etc.
The workflow of the audio detection subsystem is basically as follows: an
patrol inspection
robot repeatedly performs patrol inspection in the substation according to a
designated route
or task, and the patrol inspection robot may perform patrol inspection in one
day for several
times, and the patrol inspection covers most of geographical range of the
substation. The
audio collection module in the audio detection subsystem may collect in real-
time the
operating sound of the substation device at a location where the patrol
inspection robot arrives.
The audio training and identifying module identifies the sound, and thereby
determines
whether there are abnormalities or faults in the device. If it is found that
there is a fault, the
audio detection subsystem sends an alarm message to the base station backstage
software
system to notify the staff on duty in the substation, thereby the staff on
duty may quickly take
settlement measures to eliminate the fault or hazard, thereby ensuring safe
operation of the
substation.
[0017] The beneficial effects of using the technical solution described above
are as follows:
1) the working way in which the manual patrol inspection is assisted or
replaced by the patrol
inspection platform may effectively eliminate a working risk of a staff and
protect personal
security; 2) the audio detection subsystem may effectively detect abnormal
sound of the
substation device and improve patrol inspection reliability and stability to
maintain normal
operation of the substation; 3) the patrol inspection platform repeatedly
inspects all the
devices in the substation, which is flexible, convenient, freely controlled,
and only one sound
pickup device is required to be assembled, thereby the number of the audio
sensors and
- 4 -
CA 02823546 2013-07-02
configuration difficulty in other solutions are reduced, the advantages of low
cost, simplicity
of devices and high reliability are obtained; 4) the audio detection subsystem
communicates
with the computer subsystem in real time, the computer subsystem can receive,
analyze,
display all kinds of audio data in real time, and sound an alarm for an
abnormal result in time,
thereby assists the staff to quickly eliminate an operation hazard or fault of
the device, and
ensures patrol inspection quality.
BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 shows a functional block diagram of an embodiment of the
invention;
[0019] Figure 2 is a flow chart of a detection method of an embodiment of the
invention;
and
[0020] Figure 3 is a software flow chart of an audio training and identifying
module of the
invention.
DETAILED DESCRIPTION OF THE INVENTION
[0021] Referring to Figure 1, the embodiments include an audio detection
subsystem and a
computer subsystem wirelessly communicating with the audio detection
subsystem;
[0022] The audio detection subsystem includes an audio collection module, a
video service
module, an audio training and identifying module, a storage module, and a
first
communication module; an output of the audio collection module is connected
with an input
of the video service module, an output of the video service module is
connected with a
corresponding input of the audio training and identifying module, a
corresponding output of
the audio training and identifying module is connected with an input of the
first
communication module, and the storage module is connected with the audio
training and
identifying module bidirectionally.
[0023] The computer subsystem includes a host computer installed with a base
station
backstage software system, an audio data display module, an alarm module and a
second
communication module; a corresponding output of the computer host is
respectively
connected with an input of the audio data display module and an input of the
alarm module;
and an output of the second communication module is connected with a
corresponding input
- 5 -
CA 02823546 2013-07-02
of the host computer.
[0024] The embodiment further includes a patrol inspection platform, and the
patrol
inspection platform uses an electric power patrol inspection robot
(hereinafter referred to as
robot) with a model of LER-3000. The electric power patrol inspection robot is
a carrier of the
audio detection subsystem and is responsible for carrying the audio detection
subsystem and
performing the patrol inspection task. A corresponding output of the robot is
connected with
an initiator of the audio collection module, and when the robot moves to a set
device point
according to a designated route, the corresponding output of the robot will
trigger the initiator
of the audio collection module, and the audio collection module begins to
collect the
operating sound of the device.
[0025] The audio collection module adopts a sound pickup device which can
achieve a
high-fidelity sound recording operation in an outdoor and long-distance
environment. The
sound pick-up device is mounted in the front part of the inspect robot,
collects the sound of
the substation device at a location where the robot arrives. The sound pickup
device uses a
firm, durable and beautiful metal housing, has an electroplated surface, and
has the
characteristic of waterproof, moisture proof, high-fidelity and low noise,
thereby providing
reliable audio data for the subsequent audio signal processing.
[0026] The video service module is directly connected with the sound pickup
device, for
receiving audio data of the sound pickup device, performing the function of
digital signal
compression and transmission. In which, the compression format adopts an A-law
PCM
(Pulse Code Modulation) format, and the transmission way adopts the TCP/IP
protocol
transmission standard.
[0027] The audio training and identifying module is mounted in the robot, and
consists of
an ARM9 microprocessor and a DSP microprocessor; the ARM9 microprocessor
adopts an
S3C2440 chip, for receiving the audio data transmitted from the video service
module so as to
achieve the function of data communication; the DSP microprocessor adopts an
TMS320X2812 chip, for performing the function of audio data training and
identifying. The
audio training and identifying includes a training program and an identifying
program. The
training program trains operating sound of a normal substation device by a VQ
(Vector
Quantization) algorithm to obtain a normal sound feature library, and stores
the feature library
in the storage module; the training work is required to be performed only
once, and so that all
- 6 -
CA 02823546 2013-07-02
the audio identification works can be performed by using the feature library.
The identifying
program is that, the DSP microprocessor firstly extracts a sub-band feature of
the currently
collected audio data, and then compares the sub-band feature with the audio
feature library
stored in the storage module in advance. If the comparison result is displayed
as FALSE, it
indicates that there is an abnormal sound in the operation of a certain device
in the substation,
meanwhile the comparison result is sent to the base station backstage software
system by the
ARM9.
[0028] The VQ (vector quantization) algorithm is an important model
classification method.
In the audio identifying system, the vector quantization can be used to create
model for each
of the audio data to be identified. That is, in the training phase, performing
unsupervised
model clustering to the audio feature with an LBG algorithm to form respective
codebooks; in
the identifying phase, searching for the average minimum quantization
distortion distance
between the testing sound and each codebook, and the one with a minimum
distance is output
as the identifying result. Finally, whether the audio data is abnormal can be
determined.
[0029] The storage module is a FLASH storage device connected with the ARM9
microprocessor. The storage module is adapted to store the audio feature
library formed by the
training, the audio identifying result, an operation system kernel and a file
system, etc..
[0030] The computer subsystem is provided in the main control room of the
substation, and
can communicate with the audio detection subsystem provided on the patrol
inspection
platform via the first communication module and the second communication
module, the first
communication module and the second communication module meet the 802.11n
protocol;
the computer subsystem receives the abnormal result of audio detection in time
and sound an
alarm via the alarm module; meanwhile the audio data, such as waveform data
and sub-band
feature data, can be showed to a staff on duty or a relevant staff, and this
task is implemented
by the audio data display module.
[0031] The sound abnormality detection method for an electric power device
using the
above technical solution includes the following steps, referring to Figure 2
to Figure 3:
(1) starting an electric power patrol inspection robot;
(2) moving the electric power patrol inspection robot to an electric power
device
patrol inspection point according to a designated route, triggering an
initiator of the audio
- 7 -
CA 02823546 2013-07-02
collection module, so that the audio collection module collects audio data of
the electric
power device;
(3) receiving, by the video service module, the audio data collected by the
audio
collection module, performing digital signal compression and transmission;
(4) receiving, by the audio training and identifying module, the audio data
transmitted from the video service module, and performing training or
identification to the
audio data, which includes:
[0032] determining firstly whether there is audio feature data of the device
in the audio
feature library, if there is audio feature data of the device in the audio
feature library, a
training program is not required to be performed, and the process proceeds to
an identifying
grogram directly, and otherwise if there is no audio feature data of the
device in the audio
feature library, the process proceeds with the training program; and
[0033] if the training is required, entering the training program to train the
operating sound
of a normal substation device so as to obtain an normal audio feature library,
and storing the
normal audio feature library in the storage module;
[0034] if the training is not required, entering an identifying program, in
which the DSP
microprocessor firstly extracts a sub-band feature of the currently collected
audio data, and
then compares the sub-band feature with the audio feature library stored in
the storage
module;
[0035] (5) transmitting, by the audio training and identifying module, the
comparison result
to the host computer by the real-time communication between the first
communication
module and the second communication module, wherein if the comparison result
is abnormal,
the alarm module sounds an alarm, and the audio data display module displays
the waveform
data and the sub-band feature of the audio; otherwise, if the comparison
result is not abnormal,
only the waveform data and the sub-band feature of the audio are output and;
[0036] (6) determining whether the patrol inspection route has been completely
inspected,
and if the patrol inspection route has been inspected completely, ending;
otherwise,
performing step (2).
[0037] In summary, when the robot moves to the vicinity of the substation
device, the robot
starts the audio collection module, and the operating sound of the substation
device is
- 8 -
CA 02823546 2013-07-02
=
recorded by the sound pickup device at the sampling rate of 8k, the recorded
sound is
transmitted to the video service module, and is encoded and compressed by the
video service
module, to form an audio signal of the A-law PCM format. The digital signal
will enter the
audio training and identifying module.
[00381 The audio training and identifying module has two branches, i.e., the
training
function and the identifying function. Before the robot performs the patrol
inspection task for
the first time, the robot starts the training function. The training function
refers to: receiving
the sound transmitted from the video service module, extracting a feature of
the sound,
generating an audio feature library, and storing the feature library and the
transmitted sound in
the storage module. The training function is required to be performed only
once to generate
the audio feature library, thereby making preparations for the identifying
function. The
identifying function refers to: after receiving the sound transmitted from the
video service
module, extracting the feature of the sound to be identified, taking out a
sound feature
library having been trained in advance from the storage module, and comparing
the sound
feature library with the feature of the sound to be identified, thereby
obtaining the comparison
result. If the comparison result is FALSE, it indicates that the sound to be
identified is
abnormal, so the audio training and identifying module wirelessly communicates
with the
computer subsystem, and then the alarm module sounds an alarm, meanwhile the
audio data
display module can display some data and features of the sound.
- 9 -