CN114285038A - APF resonance prediction method based on voting perceptron - Google Patents

APF resonance prediction method based on voting perceptron Download PDF

Info

Publication number
CN114285038A
CN114285038A CN202111399708.5A CN202111399708A CN114285038A CN 114285038 A CN114285038 A CN 114285038A CN 202111399708 A CN202111399708 A CN 202111399708A CN 114285038 A CN114285038 A CN 114285038A
Authority
CN
China
Prior art keywords
apf
resonance
current
power grid
maximum value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202111399708.5A
Other languages
Chinese (zh)
Other versions
CN114285038B (en
Inventor
张宁
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Xixing Technology Co ltd
Original Assignee
Shanghai Xixing Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shanghai Xixing Technology Co ltd filed Critical Shanghai Xixing Technology Co ltd
Priority to CN202111399708.5A priority Critical patent/CN114285038B/en
Publication of CN114285038A publication Critical patent/CN114285038A/en
Application granted granted Critical
Publication of CN114285038B publication Critical patent/CN114285038B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
    • Y02E40/20—Active power filtering [APF]

Landscapes

  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention provides an APF resonance prediction method based on a voting sensor, which comprises the following steps: collecting waveform data of the power grid voltage and the load current before and after APF resonance; determining characteristic values of training samples for APF resonance prediction; constructing a classification criterion of the feature vector and the value sensor; selecting a loss function and a parameter learning method of the perceptron; iteratively calculating a weight vector of the perceptron; and using the learnt sensor classification criterion in the resonance judgment of the APF. Compared with the prior art, the method can accurately predict the APF to generate resonance according to some characteristics of the voltage and the load current in the power grid, and carry out shutdown operation on the APF in advance, thereby avoiding the adverse effect of the APF and the load resonance on the power grid.

Description

APF resonance prediction method based on voting perceptron
Technical Field
The invention relates to the technical field of power electronics and neural networks, in particular to an APF (adaptive Power Filter) resonance prediction method based on a voting sensor.
Background
The active filter (APF) is connected in parallel in the power grid and is equivalent to a controllable reactive and harmonic current source, the reactive current and the harmonic current of the active filter can quickly change along with the change of the load reactive current and the harmonic current, the reactive power and the harmonic needed by the power grid system are automatically compensated, dynamic reactive compensation is realized on the reactive power and the harmonic of the power grid, and the active filter belongs to an important component part of a Flexible Alternating Current Transmission System (FACTS).
Due to the complex characteristics of the grid load, the APF may resonate with a special load in practical application. The APF can have serious consequences once it resonates in the grid. The light results in the load protection of the factory, and the heavy results in the burning of the transformer or the load.
Patent document CN112994004A (application number: CN 202011580570.4) discloses a hybrid active filter resonance suppression strategy considering control delay, which is a novel resonance suppression method with a hybrid active filter applied to high-voltage direct-current power transmission as a research object. However, the impedance characteristic of the hybrid active filter is complex, resonance risk is easily generated, and in addition, due to calculation delay and sampling delay, the hybrid active filter may present a negative impedance characteristic, which greatly endangers the stability of an alternating current/direct current system. Therefore, the invention analyzes the influence of digital delay on the system stability by establishing an impedance model, designs a self-adaptive resonance suppression strategy based on the least mean square algorithm, adopts the design idea of feedforward compensation, and utilizes the waveform of the next moment to be predicted to eliminate the phase lag caused by control delay, thereby not only improving the impedance characteristic of the filter, but also adaptively adjusting the prediction matrix to be applied to the conditions of working conditions and circuit parameter change.
Patent document CN106253279A (application number: CN 201610712875.3) discloses an algorithm for quickly extracting and suppressing the harmonic component of the current on the grid side, and effectively suppresses the resonance between the active filter and the grid by performing detection comparison and judgment. A harmonic-resistant control algorithm applied to an active filter detects load current in real time, extracts harmonic components in the load current through a sliding window DFT algorithm, provides current reference for a control module, and achieves the harmonic suppression function of the active filter. The invention can protect equipment under the condition of weak power grid damping, prevent the equipment from being damaged by resonance, simultaneously can quickly extract harmonic components of resonance current, and quickly responds to an internal module, thereby achieving the aim of resonance inhibition.
Patent document CN103378595A (application number: cn201210105400. x) discloses a method for performing optimal configuration on parameters of a parallel hybrid active filter (SHAPF) by using a relationship between capacitance reactive compensation power and series-parallel resonant frequency as a constraint condition, wherein an Improved Particle Swarm Optimization (IPSO) is adopted, and a time-varying nonlinear trigonometric function method is proposed to control the parameters according to a relationship between a parameter speed of the particle swarm optimization and an inertia factor, so that a convergence speed of the algorithm is accelerated, and local optimization is prevented from being trapped. And simulation verification is carried out through Matlab, so that the parameter design of SHAPF is optimized and configured, and a good filtering effect is achieved. In the example application, the resonance is effectively avoided, and the method has certain engineering application value.
The prior art proposes an algorithm which can suppress resonance under the condition of aiming at certain specific loads or specific power grids, but the APF faces more situations in practical application, and once resonance occurs, serious results can occur. The invention aims to provide an APF resonance prediction method based on a voting sensor, which can accurately predict APF resonance according to some characteristics of voltage and load current in a power grid, carry out shutdown operation on the APF in advance and avoid adverse effects of APF and load resonance on the power grid.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide an APF resonance prediction method and system based on a voting sensor. The method can conveniently and accurately predict the resonance risk of the APF in the operation of the power grid, and shut down the APF before resonance occurs, thereby ensuring safe and reliable operation of the power grid.
In order to achieve the purpose, the invention adopts the technical scheme that:
the invention adopts a neural network of a perceptron algorithm to predict APF resonance, and comprises the following steps:
step 1: collecting waveform data of grid voltage and load current before and after APF current suddenly increases
The occurrence of resonance in the APF generally occurs in the case of a sudden increase in output current, but the increase in APF output current does not necessarily cause resonance. A data acquisition function is designed in the APF equipment, once the output current of the APF is suddenly increased, the waveform data of the grid voltage and the load current before and after the output current is increased are stored in a flash memory in the APF equipment, and the data are uploaded to a server through GPRS.
Step 2: the eigenvalues of the training samples used for APF resonance prediction are determined. The resonance of the APF in the power grid is usually caused by the operation of a special load in the power grid, and the special load operation which can cause the APF resonance is usually caused by the change of the current and the phase of the voltage in the power grid, or the sudden change of the voltage or the current in the power grid, and the possibility of generating a peak voltage or a surge current in the voltage or the current in the power grid. So the eigenvalues of the selected training samples are:
1. APF resonance or current sudden increase maximum value of three-phase grid voltage phase change in previous mains supply period
2. Maximum value of three-phase grid current phase change in APF resonance or sudden current increase in previous mains supply period
3. Maximum value of three-phase grid voltage slope in previous mains supply period due to APF resonance or sudden current increase
4. Maximum value of three-phase grid current slope in previous mains supply period due to APF resonance or sudden current increase
5. Three-phase network voltage maximum value in previous mains supply period caused by APF resonance or sudden current increase
6. Three-phase network current maximum value in previous mains supply period due to APF resonance or sudden current increase
And selecting N groups of resonant data from the data of the server, extracting the characteristic values, selecting N groups of data with suddenly increased current but no resonance, extracting the characteristic values, and obtaining a training set of 2N samples. phva = { phva = } { (phva)(1),phva(2),phva(3),…,phva(2N)Wherein phva(1)For APF resonance or sudden increase of current extracted from the first set of data, the maximum value of the phase change of the A-phase grid voltage in the previous mains supply period, phva(2N)And the maximum value of the voltage phase change of the A-phase power grid in the previous mains supply period is suddenly increased for APF resonance or current extracted from the 2N group of data. phvb and phvc are analogically analogized and are respectively the maximum value of the voltage phase change of the B, C-phase power grid.
phia=｛phia(1),phia(2),phia(3),…,phia(2N)Therein phia(1)For APF resonance or sudden current increase extracted from the first group of data, the maximum value of the phase change of the A-phase power grid current in the previous mains supply period，phia(2N)And (3) suddenly increasing the APF resonance or current extracted from the 2N group of data to the maximum value of the phase change of the A-phase power grid current in the previous mains supply period. And the parity and the phic are analogized, and the values are respectively the maximum values of the current phase change of the B, C-phase power grid.
sva=｛sva(1),sva(2),sva(3),…,sva(2N)Therein sva(1)Sva maximum value of voltage slope of A-phase power grid in previous commercial power cycle for APF resonance or sudden current increase extracted from first set of data(2N)And (3) the maximum value of the voltage slope of the A-phase power grid in the previous mains supply period is suddenly increased for APF resonance or current extracted from the 2N group of data. svb and svc, and so on, are B, C phase grid voltage slope maxima, respectively.
sia=｛sia(1),sia(2),sia(3),…,sia(2N)Where sia(1)For APF resonance or sudden current increase extracted from the first set of data, the maximum value of the current slope of the A-phase power grid in the previous mains supply period, sia(2N)And (3) suddenly increasing the APF resonance or current extracted from the 2N group of data to the maximum value of the current slope of the A-phase power grid in the previous mains supply period. And sib and sic, and so on, are B, C phase grid current slope maximum values respectively.
vpa=｛vpa(1),vpa(2),vpa(3),…,vpa(2N)Therein vpa(1)Vpa maximum value of A-phase network voltage in previous mains cycle for APF resonance or current extracted from first set of data(2N)And (3) suddenly increasing the maximum value of the A-phase grid voltage in the previous mains supply period for APF resonance or current extracted from the 2N group of data. vpb and vpc, and so on, are the maximum values of B, C phase grid voltages respectively.
ipa=｛ipa(1),ipa(2),ipa(3),…,ipa(2N)Therein ipa(1)Vpa maximum value of A-phase grid current in previous mains cycle for APF resonance or sudden increase of current extracted from first set of data(2N)And (3) the APF resonance or current extracted from the 2N group data is suddenly increased by the maximum value of the A-phase grid current in the previous mains supply period. vpb and vpc, and so on, BAnd C-phase grid current maximum value.
y=｛y(1),y(2),y(3),…,y(2N)Store the class label in y (define resonance as 1 and no resonance as-1). y is(1)Class label, y, for whether the first set of data is resonant(2N)A class label of whether the 2N group data is resonant.
And step 3: constructing vector X(n)=[phva(n),phvb(n),phvc(n),phia(n),phib(n),phic(n),sva(n),svb(n),svc(n),sia(n),sib(n),sic(n),vpa(n),vpb(n),vpc(n),ipa(n),ipb(n),ipc(n)]T
If there are 2N sets of data in step 2, then 2N vectors X can be constructed from the 2N sets of data(n)（n=1，2，3，…,2N）。
Constructing classification criteria of the value sensing device:
Figure 100002_RE-DEST_PATH_IMAGE001
(1)
wherein
Figure DEST_PATH_IMAGE002
Is a weight vector for the perceptron and,
Figure DEST_PATH_IMAGE003
is a feature value extracted from a set of data needed to predict whether the APF will resonate,
Figure DEST_PATH_IMAGE004
is the result of prediction (
Figure 557446DEST_PATH_IMAGE004
The frequency of the signal that is not 1 will resonate,
Figure 756477DEST_PATH_IMAGE004
= -1 non-resonance)
And 4, step 4: initializing weight vectors
Figure DEST_PATH_IMAGE005
0，
Figure 598531DEST_PATH_IMAGE005
0Is composed of
Figure 69439DEST_PATH_IMAGE005
The initial value of (c).
And 5: handle
Figure 414970DEST_PATH_IMAGE005
k-1And in step 2X (k) Substituting the k-th prediction result into the formula (1)
Figure DEST_PATH_IMAGE006
。
Step 6: calculating the difference between the model prediction and the true tag if
Figure DEST_PATH_IMAGE007
The sample is predicted incorrectly if
Figure DEST_PATH_IMAGE008
The sample prediction is correct. The following formula was chosen as the loss function:
Figure DEST_PATH_IMAGE009
if it is not
Figure DEST_PATH_IMAGE010
I.e. sample prediction error, then a loss function
Figure DEST_PATH_IMAGE011
. If it is not
Figure DEST_PATH_IMAGE012
Then the sample prediction is correct, then the function is lostNumber of
Figure DEST_PATH_IMAGE013
。
And 7: updating weight vectors
Figure RE-481417DEST_PATH_IMAGE014
Using a random gradient descent with a gradient of each update
Figure RE-373281DEST_PATH_IMAGE015
Figure RE-933575DEST_PATH_IMAGE016
And 8: repeating the iteration for 2N times, and calculating
Figure RE-705222DEST_PATH_IMAGE014
(k=1，2，3，…,2N)
And step 9: computing
Figure RE-468910DEST_PATH_IMAGE014
Average value of (2)
Figure RE-413732DEST_PATH_IMAGE017
。
Figure RE-766216DEST_PATH_IMAGE018
Use of
Figure 277304DEST_PATH_IMAGE017
As a weight vector used in APF products, the weight vector with formula (1) is used to predict whether APF will generate resonance.
Figure DEST_PATH_IMAGE019
（2）
Step 10: when APF is in operation, the sliding window is extracted by 20mX = [ phva, phvb, phvc, phia, phib, phic, sva, svb, svc, sia, sib, sic, vpa, vpb, vpc, ipa, ipb, ipc within s]TCalculating by substituting X into formula (2)
Figure DEST_PATH_IMAGE020
If, if
Figure DEST_PATH_IMAGE021
The APF is immediately turned off to avoid the resonance condition.
Drawings
Other features, objects and advantages of the invention will become more apparent upon reading of the detailed description of non-limiting embodiments with reference to the following drawings:
FIG. 1 is a three-phase grid voltage waveform before and after resonance occurs;
fig. 2 shows three-phase grid current waveforms before and after resonance occurs.
Fig. 3 is a flow chart of an APF resonance prediction method.
Detailed Description
The present invention will be described in detail with reference to specific examples. The following examples will assist those skilled in the art in further understanding the invention, but are not intended to limit the invention in any way. It should be noted that it would be obvious to those skilled in the art that various changes and modifications can be made without departing from the spirit of the invention. All falling within the scope of the present invention.
According to the APF resonance prediction method based on the voting sensor, as shown in FIG. 1-FIG. 2, the method comprises the following steps:
step 1: collecting waveform data of grid voltage and load current before and after APF current suddenly increases
The occurrence of resonance in the APF generally occurs in the case of a sudden increase in output current, but the increase in APF output current does not necessarily cause resonance. A data acquisition function is designed in the APF equipment, once the output current of the APF is suddenly increased, the waveform data of the grid voltage and the load current before and after the output current is increased are stored in a flash memory in the APF equipment, and the data are uploaded to a server through GPRS.
Step 2: the eigenvalues of the training samples used for APF resonance prediction are determined. The resonance of the APF in the power grid is usually caused by the operation of a special load in the power grid, and the special load operation which can cause the APF resonance is usually caused by the change of the current and the phase of the voltage in the power grid, or the sudden change of the voltage or the current in the power grid, and the possibility of generating a peak voltage or a surge current in the voltage or the current in the power grid. So the eigenvalues of the selected training samples are:
1. APF resonance or current sudden increase maximum value of three-phase grid voltage phase change in previous mains supply period
2. Maximum value of three-phase grid current phase change in APF resonance or sudden current increase in previous mains supply period
3. Maximum value of three-phase grid voltage slope in previous mains supply period due to APF resonance or sudden current increase
4. Maximum value of three-phase grid current slope in previous mains supply period due to APF resonance or sudden current increase
5. Three-phase network voltage maximum value in previous mains supply period caused by APF resonance or sudden current increase
6. Three-phase network current maximum value in previous mains supply period due to APF resonance or sudden current increase
And selecting 1000 groups of resonant data from the data of the server, extracting the characteristic value, selecting 1000 groups of data with suddenly increased current but no resonance, extracting the characteristic value, and obtaining a training set of 2000 samples. phva = { phva = } { (phva)(1),phva(2),phva(3),…,phva(2000)Wherein phva(1)For APF resonance or sudden increase of current extracted from the first set of data, the maximum value of the phase change of the A-phase grid voltage in the previous mains supply period, phva(2000)And the APF resonance or current extracted from the 2000 th group of data is suddenly increased by the maximum value of the voltage phase change of the A-phase power grid in the previous power supply period. phvb and phvc are analogically analogized and are respectively the maximum value of the voltage phase change of the B, C-phase power grid.
phia=｛phia(1),phia(2),phia(3),…,phia(2000)Therein phia(1)Extracting for the first set of dataThe APF resonance or the sudden increase of the current of the A-phase power grid in the previous mains cycle has the maximum value of the phase change of the current of the A-phase power grid, phia(2000)The phase change of the a-phase grid current in the previous mains cycle is increased suddenly for the APF resonance or current extracted in group 2000. And the parity and the phic are analogized, and the values are respectively the maximum values of the current phase change of the B, C-phase power grid.
sva=｛sva(1),sva(2),sva(3),…,sva(2000)Therein sva(1)Sva maximum value of voltage slope of A-phase power grid in previous commercial power cycle for APF resonance or sudden current increase extracted from first set of data(2000)The maximum value of the voltage slope of the A-phase power grid in the previous mains supply period is suddenly increased for APF resonance or current extracted from the 2000 th group of data. svb and svc, and so on, are B, C phase grid voltage slope maxima, respectively.
sia=｛sia(1),sia(2),sia(3),…,sia(2000)Where sia(1)For APF resonance or sudden current increase extracted from the first set of data, the maximum value of the current slope of the A-phase power grid in the previous mains supply period, sia(2000)The slope of the a-phase grid current in the previous mains cycle is suddenly increased for the APF resonance or current extracted in group 2000. And sib and sic, and so on, are B, C phase grid current slope maximum values respectively.
vpa=｛vpa(1),vpa(2),vpa(3),…,vpa(2000)Therein vpa(1)Vpa maximum value of A-phase network voltage in previous mains cycle for APF resonance or current extracted from first set of data(2000)The APF resonance or current extracted for group 2000 data suddenly increases the maximum value of the a-phase grid voltage in the previous mains cycle. vpb and vpc, and so on, are the maximum values of B, C phase grid voltages respectively.
ipa=｛ipa(1),ipa(2),ipa(3),…,ipa(2000)Therein ipa(1)Vpa maximum value of A-phase grid current in previous mains cycle for APF resonance or sudden increase of current extracted from first set of data(2000)Extracting for the 2000 th group of dataThe APF of (a) resonates or current suddenly increases the maximum value of the a-phase grid current in the previous mains cycle. vpb and vpc, and so on, are the maximum values of B, C phase grid currents respectively.
y=｛y(1),y(2),y(3),…,y(2000)Store the class label in y (define resonance as 1 and no resonance as-1). y is(1)Class label, y, for whether the first set of data is resonant(2000)The class label is whether group 2000 data is resonant.
And step 3: constructing vector X(n)=[phva(n),phvb(n),phvc(n),phia(n),phib(n),phic(n),sva(n),svb(n),svc(n),sia(n),sib(n),sic(n),vpa(n),vpb(n),vpc(n),ipa(n),ipb(n),ipc(n)]T
With 2000 sets of data in step 2, then 2N vectors X can be constructed from the 2N sets of data(n)（n=1，2，3，…,2000）。
Constructing classification criteria of the value sensing device:
Figure 614876DEST_PATH_IMAGE001
(3)
wherein
Figure 431522DEST_PATH_IMAGE002
Is a weight vector for the perceptron and,
Figure 168534DEST_PATH_IMAGE003
is a feature value extracted from a set of data needed to predict whether the APF will resonate,
Figure 504969DEST_PATH_IMAGE004
is the result of prediction (
Figure 934813DEST_PATH_IMAGE004
The frequency of the signal that is not 1 will resonate,
Figure 238755DEST_PATH_IMAGE004
=1 does not resonate).
And 4, step 4: initializing weight vectors
Figure 513879DEST_PATH_IMAGE005
0Either ← 0 (all-zero vector),
Figure 701890DEST_PATH_IMAGE005
0is composed of
Figure 568215DEST_PATH_IMAGE005
Is started.
And 5: handle
Figure 828295DEST_PATH_IMAGE005
k-1And in step 2X (k) Substituting the k-th prediction result into the formula (1)
Figure DEST_PATH_IMAGE022
。
Step 6: calculating the difference between the model prediction and the true tag if
Figure DEST_PATH_IMAGE023
The sample is predicted incorrectly if
Figure DEST_PATH_IMAGE024
The sample prediction is correct. The following formula was chosen as the loss function:
Figure 329946DEST_PATH_IMAGE009
if it is not
Figure 827923DEST_PATH_IMAGE010
I.e. sample prediction error, then a loss function
Figure 661887DEST_PATH_IMAGE011
. If it is not
Figure 612526DEST_PATH_IMAGE012
Then the sample prediction is correct, then the penalty function
Figure 308080DEST_PATH_IMAGE013
。
And 7: updating weight vectors
Figure 722881DEST_PATH_IMAGE014
Using a random gradient descent with a gradient of each update
Figure 665429DEST_PATH_IMAGE015
Figure 103364DEST_PATH_IMAGE016
And 8: repeating the steps 5-7, iterating for 2000 times, and calculating
Figure 865259DEST_PATH_IMAGE014
(k=1，2，3，…,2000)
And step 9: computing
Figure 868987DEST_PATH_IMAGE014
Average value of (2)
Figure 513595DEST_PATH_IMAGE017
。
Figure DEST_PATH_IMAGE025
Use of
Figure 986296DEST_PATH_IMAGE017
As a weight vector used in APF products, the weight vector with formula (1) is used to predict whether APF will generate resonance.
Figure 538500DEST_PATH_IMAGE019
（2）
Step 10: when the APF is in operation, the sliding window extracts X = [ phva, phvb, phvc, phia, phib, phic, sva, svb, svc, sia, sib, sic, vpa, vpb, vpc, ipa, ipb, ipc within 20ms]TCalculating by substituting X into formula (2)
Figure 599997DEST_PATH_IMAGE020
If, if
Figure 697397DEST_PATH_IMAGE021
The APF is immediately turned off to avoid the resonance condition.
The foregoing description of specific embodiments of the present invention has been presented. It is to be understood that the present invention is not limited to the specific embodiments described above, and that various changes or modifications may be made by one skilled in the art within the scope of the appended claims without departing from the spirit of the invention.

Claims (4)

1. An APF resonance prediction method based on a voting sensor is characterized by comprising the following steps:
step 1: collecting waveform data of the power grid voltage and the load current before and after the APF current is suddenly increased;
step 2: determining characteristic values of training samples for APF resonance prediction;
and step 3: constructing a vector X and a classification criterion of a value sensor;
and 4, step 4: initializing a weight vector;
and 5: calculating a prediction result;
step 6: calculating the difference between the model prediction and the real label;
and 7: updating the weight vector;
and 8: repeating the step 5-7;
and step 9: calculating the average value of the weight vectors to obtain a resonance classification criterion required by the APF;
step 10: when the APF runs, the sliding window extracts the characteristic vector within 20ms, and the characteristic vector is substituted into the classification criterion obtained by the previous calculation to judge whether resonance occurs.
2. The voting sensor-based APF resonance prediction method of claim 1, wherein in the step 1:
a data acquisition function is designed in the APF equipment, once the output current of the APF is suddenly increased, the waveform data of the grid voltage and the load current before and after the output current is increased are stored in a flash memory in the APF equipment, and the data are uploaded to a server through GPRS.
3. The voting sensor-based APF resonance prediction method of claim 1, wherein in the step 2:
the eigenvalues of the selected training samples are: the method comprises the steps that APF resonance or current suddenly increases the maximum value of phase change of a three-phase power grid in a previous mains supply period, APF resonance or current suddenly increases the maximum value of phase change of a three-phase power grid current in the previous mains supply period, APF resonance or current suddenly increases the maximum value of voltage slope of the three-phase power grid in the previous mains supply period, APF resonance or current suddenly increases the maximum value of current slope of the three-phase power grid in the previous mains supply period, APF resonance or current suddenly increases the maximum value of the three-phase power grid voltage in the previous mains supply period, and APF resonance or current suddenly increases the maximum value of the three-phase power grid current in the previous mains supply period.
4. The voting sensor-based APF resonance prediction method of claim 1, wherein in the step 3:
constructing classification criteria of the value sensing device:
Figure RE-DEST_PATH_IMAGE001
(1)
wherein
Figure RE-976710DEST_PATH_IMAGE002
Is a weight vector for the perceptron and,
Figure RE-109752DEST_PATH_IMAGE003
is a feature value extracted from a set of data needed to predict whether the APF will resonate,
Figure RE-163158DEST_PATH_IMAGE004
is the result of prediction (
Figure RE-810128DEST_PATH_IMAGE004
The frequency of the signal that is not 1 will resonate,
Figure RE-353105DEST_PATH_IMAGE004
=1 does not resonate).
CN202111399708.5A 2021-11-24 2021-11-24 APF resonance prediction method based on voting perceptron Active CN114285038B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111399708.5A CN114285038B (en) 2021-11-24 2021-11-24 APF resonance prediction method based on voting perceptron

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111399708.5A CN114285038B (en) 2021-11-24 2021-11-24 APF resonance prediction method based on voting perceptron

Publications (2)

Publication Number Publication Date
CN114285038A true CN114285038A (en) 2022-04-05
CN114285038B CN114285038B (en) 2023-10-31

Family

ID=80869862

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111399708.5A Active CN114285038B (en) 2021-11-24 2021-11-24 APF resonance prediction method based on voting perceptron

Country Status (1)

Country Link
CN (1) CN114285038B (en)

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106253279A (en) * 2016-08-23 2016-12-21 南京亚派科技股份有限公司 A kind of anti-resonance control algolithm being applied to active filter
CN112217213A (en) * 2020-10-24 2021-01-12 青岛鼎信通讯股份有限公司 Resonance point real-time calculation method based on APF device
CN112448392A (en) * 2020-11-09 2021-03-05 广西电网有限责任公司电力科学研究院 Regional power grid harmonic treatment method and device based on complex-valued neural network

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106253279A (en) * 2016-08-23 2016-12-21 南京亚派科技股份有限公司 A kind of anti-resonance control algolithm being applied to active filter
CN112217213A (en) * 2020-10-24 2021-01-12 青岛鼎信通讯股份有限公司 Resonance point real-time calculation method based on APF device
CN112448392A (en) * 2020-11-09 2021-03-05 广西电网有限责任公司电力科学研究院 Regional power grid harmonic treatment method and device based on complex-valued neural network

Also Published As

Publication number Publication date
CN114285038B (en) 2023-10-31

Similar Documents

Publication Publication Date Title
Gupta et al. Fuzzy logic controlled shunt active power filter for reactive power compensation and harmonic elimination
CN111740575B (en) An Adaptive Identification Method of Inverter Model Parameters Based on Steepest Descent Method
Wang et al. DC fault detection and classification approach of MMC-HVDC based on convolutional neural network
Ni et al. Multi-machine power system control based on dual heuristic dynamic programming
CN113298124B (en) Parameter matching-based multi-level direct current fault arc detection method
CN115632394A (en) Photovoltaic power station transient model construction and parameter identification method based on PPO algorithm
CN115224698A (en) Reactive power-voltage optimization control method of new energy power system based on digital twin
Priyadarshi et al. An experimental performance verification of continuous mixed P‐norm based adaptive asymmetrical fuzzy logic controller for single stage photovoltaic grid integration
CN114285038A (en) APF resonance prediction method based on voting perceptron
CN107123991B (en) Dynamic voltage restorer control method and device based on fuzzy particle swarm algorithm
Shah et al. Power quality improvement using ultra capacitor based dynamic voltage restorer with real twisting sliding mode control
He et al. Electrical signature analysis for open-circuit faults detection of inverter with various disturbances in distribution grid
Sujith et al. Implementation of psoann optimized PI control algorithm for shunt active filter
Joshi et al. Backpropagation and particle swarm optimization based neural network for power quality improvement in grid connected systems
Wu et al. Fault diagnosis of recessive weakness in superbuck converter based on kpca-ipnn
CN105846431A (en) Power harmonics current signal tracking control method
Kumar et al. Impementation of Self-Learning ANN-ANFIS Based Dynamic Voltage Restorer for Voltage Power Quality Improvement
Meena et al. Control of DSTATCOM in distribution systems by artificial intelligence based controller
CN108110761B (en) Fuzzy high-order sliding mode active power filter control method based on linearization feedback
Fei et al. Adaptive Fuzzy-Neural-Network Control of Active Power Filter Using Fuzzy Backstepping Approach
Kumar et al. A Novel Optimized Golden Eagle Based Self-Evolving Intelligent Fuzzy Controller to Enhance Power System Performance
Wang et al. Fault through technique of synchronverter based on voltage feedforward compensation
Janggi et al. Artificial neural network controller for single-phase active power filter
Jianfeng et al. Parameter optimization design method of LCL filter based on harmonic stability of VSC system
Siregar et al. Dynamic voltage restorer quality improvement analysis using particle swarm optimization and artificial neural networks for voltage sag mitigation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant