CN111541227B - Wide area protection system of AT full parallel power supply network based on artificial intelligence - Google Patents

Wide area protection system of AT full parallel power supply network based on artificial intelligence Download PDF

Info

Publication number
CN111541227B
CN111541227B CN202010193509.8A CN202010193509A CN111541227B CN 111541227 B CN111541227 B CN 111541227B CN 202010193509 A CN202010193509 A CN 202010193509A CN 111541227 B CN111541227 B CN 111541227B
Authority
CN
China
Prior art keywords
fault
layer
wide area
area protection
network
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.)
Active
Application number
CN202010193509.8A
Other languages
Chinese (zh)
Other versions
CN111541227A (en
Inventor
尹建斌
闫雪松
王洪友
邢志杰
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.)
Tianjin Keyvia Electric Co ltd
Original Assignee
Tianjin Keyvia Electric 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 Tianjin Keyvia Electric Co ltd filed Critical Tianjin Keyvia Electric Co ltd
Priority to CN202010193509.8A priority Critical patent/CN111541227B/en
Publication of CN111541227A publication Critical patent/CN111541227A/en
Application granted granted Critical
Publication of CN111541227B publication Critical patent/CN111541227B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02H—EMERGENCY PROTECTIVE CIRCUIT ARRANGEMENTS
    • H02H7/00—Emergency protective circuit arrangements specially adapted for specific types of electric machines or apparatus or for sectionalised protection of cable or line systems, and effecting automatic switching in the event of an undesired change from normal working conditions
    • H02H7/26—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured
    • H02H7/261—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured involving signal transmission between at least two stations
    • H02H7/262—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured involving signal transmission between at least two stations involving transmissions of switching or blocking orders
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Energy or water supply
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02H—EMERGENCY PROTECTIVE CIRCUIT ARRANGEMENTS
    • H02H7/00—Emergency protective circuit arrangements specially adapted for specific types of electric machines or apparatus or for sectionalised protection of cable or line systems, and effecting automatic switching in the event of an undesired change from normal working conditions
    • H02H7/26—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured
    • H02H7/261—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured involving signal transmission between at least two stations
    • H02H7/263—Sectionalised protection of cable or line systems, e.g. for disconnecting a section on which a short-circuit, earth fault, or arc discharge has occured involving signal transmission between at least two stations involving transmissions of measured values

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Tourism & Hospitality (AREA)
  • Marketing (AREA)
  • Biomedical Technology (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Public Health (AREA)
  • Water Supply & Treatment (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Primary Health Care (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Molecular Biology (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Locating Faults (AREA)
  • Supply And Distribution Of Alternating Current (AREA)

Abstract

The invention provides a wide area protection system of an AT full parallel power supply network based on artificial intelligence, which comprises: s1, arranging wide area protection devices in power supply intervals of the AT full-parallel power supply network, wherein each wide area protection device in the power supply intervals is provided with one wide area protection device, and all the wide area protection devices are connected with each other; s2, dividing each power supply interval into a main station and a plurality of sub stations, and uploading data of the sub station wide area protection devices to the main station wide area protection devices in real time; and S3, embedding an intelligent learning network in the wide area protection device of the main place. The invention provides a wide area protection system for an AT full parallel power supply network, the system not only can selectively remove fault lines, but also can position different types of faults, and has certain learning capability, different fault data can be learned, relevant parameters of the system can be corrected, and the accuracy of the system is improved.

Description

Wide area protection system of AT full parallel power supply network based on artificial intelligence
Technical Field
The invention belongs to the field of electrified traction railways, and particularly relates to a wide-area protection system of an AT full-parallel power supply network based on artificial intelligence.
Background
The high-speed railway in China generally adopts an AT power supply mode, so that the voltage of a traction network can be improved, and the energy consumption can be reduced; the full parallel power supply mode is that the uplink and the downlink are connected through an AT transformer on the basis of the power supply of a complex line AT; when faults occur at different positions randomly, the fault positions can be quickly positioned, fault lines can be cut off, and meanwhile, the fault lines can stably run; this is a necessary condition for ensuring stable operation of the traction power supply system;
however, AT present, a relay protection scheme cannot effectively solve the problem of protection selectivity aiming AT a full parallel AT power supply mode; when a point of fault occurs on a traction network, both an upper circuit breaker and a lower circuit breaker of a feeder line of a substation need to trip, then the reclosing of the feeder line protection of the substation is matched with the voltage loss tripping and the detected voltage reclosing of an AT station and a subarea to realize the cutting isolation of a power supply arm AT a fault side and recover the power supply of a power supply arm AT a non-fault side, so that the power failure range is expanded, a fault line and a non-fault line are powered off AT the same time, and the recovery power supply time of the non-fault line is about 5 seconds, which is very unfavorable for a high-speed heavy haul railway; the AT full-parallel traction power supply network fault positioning method adopts the ratio of uplink current to downlink current, ideal equivalence exists when a model is constructed, and a measured result may have deviation; secondly, with the increase of short circuit data, scientific analysis can not be carried out only by using a single formula, and the short circuit data are qualified within the deviation of the engineering allowable range, so that a plurality of deviations can not be corrected in time.
Disclosure of Invention
In view of the above, the present invention provides a wide area protection system based on an artificial intelligence AT all parallel power supply network, which aims to overcome the above-mentioned defects in the prior art.
In order to achieve the purpose, the technical scheme of the invention is realized as follows:
wide area protection system based on AT full parallel power supply network of artificial intelligence includes:
s1, arranging wide area protection devices in power supply intervals of the AT full-parallel power supply network, wherein each wide area protection device in the power supply intervals is provided with one wide area protection device, and all the wide area protection devices are connected with each other;
s2, dividing each power supply interval into a main station and a plurality of sub stations, and uploading data of the sub station wide area protection devices to the main station wide area protection devices in real time;
s3, embedding an intelligent learning network in the wide area protection device of the main station, wherein the intelligent learning network consists of a fault analysis network and a learning network;
s4, when a fault occurs, the wide area protection device of the main station brings data in the fault time into a fault analysis network according to the fault time scale, and the fault analysis network calculates and calculates the fault position;
s5, after the fault analysis network measurement and calculation are completed, the main station wide area protection device locks the tripping of the non-fault line according to the measurement and calculation result and reports the fault position;
and S6, inputting the actual fault position information of the patrol inspection into a main station wide area protection device, training the learning network, and guiding fault analysis network fault positioning by the trained learning network.
Further, in step S1, the power supply section is a section of normal power supply range composed of a traction substation, an AT station, a sub-station, a power feeder, a power supply line, and a contact network; each wide area protection device is connected with the corresponding satellite clock, and the wide area protection devices are connected through beacon channels.
Further, in step S2, the main substation wide area protection device receives real-time data of the sub-substation wide area protection device once every 10ms through the beacon channel, so as to monitor the state of the sub-substation.
Further, the specific working manner of the fault analysis network in step S4 is as follows:
s41, the wide area protection device in the traction substation reads that the uplink T line current, the uplink F line current, the downlink T line current, the downlink F line current and the suck-up current are respectively as follows:
Figure GDA0003536398570000021
reading the voltage of an uplink T line, the voltage of an uplink F line, the voltage of a downlink T line and the voltage of a downlink F line as follows:
Figure GDA0003536398570000031
the wide area protection device in the AT station reads the same data, and the parameter names are sequentially
Figure GDA0003536398570000032
The wide area protection device in the subarea station reads the same data, and the parameter names are sequentially
Figure GDA0003536398570000033
S42, judging the power supply line, if so
Figure GDA0003536398570000034
Judging the fault type to be a TR/FR fault, otherwise, judging the fault type to be a TF fault;
s43, when the TF fault is judged to be a fault,
Figure GDA0003536398570000035
continuously judging the current direction in the place with larger current, wherein the current direction is a downlink fault if the current direction is from the uplink to the downlink, and the current direction is an uplink fault if the current direction is from the downlink to the uplink;
s44, if it is judged that TR/FR fault occurs
Figure GDA0003536398570000036
Judging as a downlink TR fault;
s45, under the condition that the fault type and the fault line are both determined, the fault analysis network adopts the uplink-downlink current ratio to carry out fault location, and the fault location formula is as follows:
Figure GDA0003536398570000037
wherein,
Figure GDA0003536398570000038
L1for up-link arm length, L2For descending power supply arm length, L is fault distance, Lc: and (5) correcting the distance.
Further, the specific operation of learning the network in step S4 is as follows;
s46, building a learning network based on an artificial neural network, wherein the learning network consists of an input layer, a hidden layer and an output layer;
an input layer: the relevant measurement information collected by the wide area protection device is selected as input, and the following formula is shown:
Figure GDA0003536398570000041
hidden layer: the number of the hidden layers is two, and the number of nerves in each layer is 27; the excitation function of each layer of the hidden layer is logsig; the learning rate is 0.1; the maximum training times is 10000; the error goals achieved by training are: 1 e-30; the performance function adopts a sumsqr function;
an output layer: outputting three quantities of a fault type Y, a fault row G and a fault distance D as output;
s47, defining the working principle of the learning network: [ SS AT SP]T→[Y G D]TEstablishing a mapping relation between an input layer and an output layer;
s48, adjusting the weight matrix and the threshold matrix of the artificial neural network, and reducing the output error value of the learning network.
Further, in step S48, a specific method for reducing the output error value of the learning network is as follows:
s481, defining an Input layer as Input, a hidden layer first layer as output hidden1, a hidden layer second layer as output hidden2 and an output layer as Layout; carrying out normalization processing on the matrix of the input layer and the matrix of the output layer by means of premmx;
s482, initializing weights and thresholds among the input layer, the hidden layer and the output layer; defining the weights among the four layers as: [ W1]27×27,[W2]27×27,[W3]3×27(ii) a The thresholds between the four layers are defined as: [ B1]27×1,[B2]27×1,[B3]27×1；
S483, according to the artificial neural network principle, defining the transmission relation among the input matrix, the weight matrix and the threshold matrix:
hidden1＝logsig(W1•Input+B1)
hidden2＝logsig(W2•hidden1+B2)
layout＝logsig(W3•hidden2+B3)；
s484, setting the actual output as Right, defining the output error of the network as E, and setting the network performance function as sumsqr:
Figure GDA0003536398570000051
s485, adjusting a weight matrix and a threshold matrix of the artificial neural network;
suppose that the amount of change of the corresponding weight value of the i-th layer neuron to the j-th layer neuron is Δ wij；yiIs the output of the ith layer, xjFor the input of the j-th layer, in order to reduce the error the fastest, the weight change should be proportional to the reduction of the error E of each layer, i.e. the negative derivative of the error function to the weight, and the learning rate is alpha, delta wijThe calculated relationship of (a) is as follows:
Figure GDA0003536398570000052
s486, above formula djFor the operator at the j-th layer, the derivation of the weight value by the input at the j-th layer is as follows:
Figure GDA0003536398570000053
let m be the output layer, when j equals m,
Figure GDA0003536398570000054
wherein,
Figure GDA0003536398570000055
is the output of the j-th layer,
Figure GDA0003536398570000056
when j is less than m and is not an output layer, then
Figure GDA0003536398570000057
Wherein, f (x) is an excitation function between layers; each layer derives the excitation function by f' (x) ═ f (x) (1-f (x)), the above equation can be simplified:
dj＝dj+1Wjj+1f(x)(1-f(x))
from the above recursion formula, the operator of the previous layer can be sequentially deduced from the error of the output layer, and after the operator is obtained, the variation of the weight and the threshold of each layer is calculated as follows:
Δwij＝dj·yi
Δdj＝dj
s487, adjusting the weights and the threshold values of the neurons in each layer, wherein the relationship between the connection weights and the unadjusted neurons in each layer is shown as the following equation:
wij(t+1)＝wij(t)+Δwij(t)+μΔwij(t-1)
dij(t+1)＝dj(t)+Δdj(t)
wherein t is the corrected times, mu is the inertia coefficient, and the training result is combined with the empirical value to automatically adjust.
Further, the specific steps of step S5 are as follows:
s51, after the measurement and calculation are completed, the main station wide area protection device determines a fault interval and a fault navigation category and sends a locking signal to the circuit breaker in the non-fault row;
and S52, calculating the fault position according to the measurement information of the related quantity of the wide area protection system and the uplink and downlink current ratio principle by the intelligent learning network, and reporting the fault position.
Further, in step S6, a specific method for guiding the learning network to the failure analysis network is as follows:
s61, training the actual fault position and the corresponding data as input by the learning network, and perfecting and modifying the weight value and the threshold value of the learning network;
and S62, after the fault analysis network finishes the fault position measurement and calculation according to the fault data, the learning network analyzes and judges the fault data, and if the fault data is similar to the training data in the learning network, the learning network corrects the fault position measurement and calculation result.
Compared with the prior art, the invention has the following advantages:
the invention provides a set of wide area protection system for an AT full parallel power supply network, the system not only can selectively remove fault lines, but also can position different types of faults, and has certain learning capability, so that different fault data can be learned, relevant parameters of the system can be corrected, and the accuracy of the system is improved; meanwhile, the method provides an idea for processing big data, and a large amount of short-circuit data can be more scientifically applied.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and, together with the description, serve to explain the invention without limitation. In the drawings:
fig. 1 is a schematic composition diagram of a wide area protection system of an AT all parallel power supply network based on artificial intelligence according to an embodiment of the present invention;
fig. 2 is a flowchart of the wide area protection system of the AT all parallel power supply network based on artificial intelligence according to the embodiment of the present invention;
fig. 3 is a flowchart illustrating the operation of a fault analysis network in a wide-area protection system based on an artificial intelligence AT all-parallel power supply network according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a learning network in a wide area protection system of an AT all parallel power supply network based on artificial intelligence according to an embodiment of the present invention;
fig. 5 is a flowchart illustrating the operation of learning network in the wide area protection system of the AT full parallel power supply network based on artificial intelligence according to the embodiment of the present invention;
fig. 6 is a schematic diagram of a failure analysis of a second uplink interval of an AT all parallel power supply system based on artificial intelligence according to an embodiment of the present invention.
Detailed Description
It should be noted that the embodiments and features of the embodiments of the present invention may be combined with each other without conflict.
In the description of the present invention, it is to be understood that the terms "central," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like are used in the orientation or positional relationship indicated in the drawings, which are merely for convenience in describing the invention and to simplify the description, and are not intended to indicate or imply that the referenced device or element must have a particular orientation, be constructed and operated in a particular orientation, and are therefore not to be construed as limiting the invention. Furthermore, the terms "first", "second", etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first," "second," etc. may explicitly or implicitly include one or more of that feature. In the description of the invention, the meaning of "a plurality" is two or more unless otherwise specified.
In the description of the invention, it is to be noted that, unless otherwise explicitly specified or limited, the terms "mounted", "connected" and "connected" are to be construed broadly, e.g. as being fixed or detachable or integrally connected; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meaning of the above terms in the creation of the present invention can be understood by those of ordinary skill in the art through specific situations.
The invention will be described in detail with reference to the following embodiments with reference to the attached drawings.
The wide area protection system based on artificial intelligence AT full parallel power supply network comprises
S1, in a power supply interval, the AT full parallel power supply system is provided with a wide area protection device in each power supply interval;
s2, uploading data of the subsytems to the main station in real time in the wide area protection system in the operation process;
s3, embedding a set of intelligent learning network in the wide area protection device in the main station, which is a core part for completing calculation; the intelligent learning network consists of a fault analysis network and a learning network.
S4, when a fault occurs, the wide area protection device in the main station brings data in the three stations at the fault time into a fault analysis network according to the fault time scale;
s5, after the calculation is completed, the main station can accurately lock the non-fault line trip, and the fault position is reported;
s6, after the patrol of the staff is completed, no matter whether the distance measurement result is correct or not, the actual fault position can be input into the wide area protection device, the intelligent learning network in the wide area protection device can automatically complete one-time learning, the trained learning network can provide guidance for fault analysis network fault positioning, and the accuracy of the next calculation is guaranteed.
Specifically, in step S1, the power supply section includes a section of normal power supply range formed by the traction substation, the AT station, the sub-station, the electric railway feeder, the power supply line, the contact network, and the like; a wide area protection device is respectively arranged in the traction substation, the AT station and the subarea station; in the station, all the devices are connected with the satellite clock in the station, so that the time is ensured to be completely synchronous; the devices are connected through beacon channels to form a wide area protection system as shown in fig. 1.
Specifically, in step S2, the main station receives the real-time data of the small station once every 10ms through the beacon channel, and monitors the state of the small station.
Specifically, in step S3, when a fault occurs, the master records a time scale of the fault occurrence and sends a fault signal to the slave station; after receiving the fault signal, the small place actively uploads the fault information starting from the time scale, and the fault information is uploaded once every 10 ms. After the master station receives the fault data, if the uploaded fault data are stable and complete, the data are brought into an intelligent network, and a data transmission stopping signal is sent to the slave station.
Specifically, in step S4, the intelligent learning network is composed of a failure analysis network and a learning network, wherein the working principle of the failure analysis network is as follows:
the wide area protection device in the traction substation reads that the uplink T line current, the uplink F line current, the downlink T line current, the downlink F line current and the suck-up current are respectively as follows:
Figure GDA0003536398570000091
reading the voltage of an uplink T line, the voltage of an uplink F line, the voltage of a downlink T line and the voltage of a downlink F line as follows:
Figure GDA0003536398570000092
the wide area protection device in the AT station reads the same data, and the parameter names are sequentially
Figure GDA0003536398570000093
The wide area protection device in the subarea station reads the same data, and the parameter names are sequentially
Figure GDA0003536398570000094
In the power supply line, if present
Figure GDA0003536398570000095
And judging the fault type to be TR/FR fault, otherwise, judging the fault type to be TF fault.
If the TF fault is judged to be a TF fault,
Figure GDA0003536398570000101
and judging the current direction in which the current is larger, and if the current flows from the uplink to the downlink, determining that the current is a downlink fault, and vice versa.
If it is determined as TR/FR fault, if it is
Figure GDA0003536398570000102
It may be judged as a downlink TR failure. The other fault types are the same as the fault row judgment principle, and the fault judgment flow is shown in fig. 3.
Under the condition that the fault type and the fault line are both determined, the fault analysis network carries out fault location by adopting the current ratio of uplink and downlink, specifically, taking the uplink fault as an example, the uplink and downlink equations can be written by kirchhoff's law as follows:
Figure GDA0003536398570000103
in the formula:
Figure GDA0003536398570000104
L1for up-link arm length, L2For descending power supply arm length, L is fault distance, Lc: and (5) correcting the distance. Z1Is an uplink impedance, Z2Is the downstream impedance, Z12Is the uplink and downlink mutual impedance.
Formula 1, taken together, can be solved:
Figure GDA0003536398570000105
since the uplink and downlink parameters are equivalent, Z can be assumed1-Z12＝Z2-Z12And formula 2 can be simplified as
Figure GDA0003536398570000106
LC: and correcting the distance and setting according to field parameters.
In step S4, the learning network part in the intelligent learning network is built based on an artificial neural network, and the working principle is as follows:
the learning network consists of three parts, namely an input layer, a hidden layer and an output layer.
An input layer: the relevant measurement information collected by the wide area protection device is selected as input, as shown in the following formula 2:
Figure GDA0003536398570000107
hidden layer: the number of the hidden layers is two, and the number of nerves in each layer is 27;
an output layer: the three quantities of the fault type Y, the fault row G and the fault distance D are output as the output. The excitation function of each layer of the network hidden layer is logsig; the learning rate is 0.1; the maximum training times is 10000; the error goals achieved by training are: 1 e-30; the performance function adopts a sumsqr function;
learning the network working principle: [ SS AT SP]T→[Y G D]TThat is, a mapping relationship is established between the input layer and the output layer, and a schematic diagram is shown in fig. 4;
(1) the Input layer is Input, the hidden layer first layer outputs hidden1, the hidden layer second layer outputs hidden2, and the output layer is Layout. Normalizing the matrix of the input layer and the output layer by means of premnx
(2) And initializing weights and thresholds among the input layer, the hidden layer and the output layer. The weights between the four layers are respectively: [ W1]27×27,[W2]27×27,[W3]3×27. The thresholds between the four layers are: [ B1]27×1,[B2]27×1,[B3]27×1
(3) According to the artificial neural network principle, the transfer relationship among the input matrix, the weight matrix and the threshold matrix can be written as follows:
hidden1＝logsig(W1·Input+B1)
hidden2＝logsig(W2·hidden1+B2)
layout＝logsig(W3·hidden2+B3)
(4) assuming that the actual output is Right, the output error of the network is defined as E, and the network performance function is sumsqr, then:
Figure GDA0003536398570000111
(5) adjusting a weight matrix and a threshold matrix of the artificial neural network, and reducing an output error E of the network;
suppose that the amount of change of the corresponding weight value of the i-th layer neuron to the j-th layer neuron is Δ wij；yiIs the output of the ith layer, xjFor the input of the j-th layer, in order to reduce the error the fastest, the weight change should be proportional to the reduction of the error E of each layer, i.e. the negative derivative of the error function to the weight, and the learning rate is alpha, delta wijThe calculated relationship of (a) is as follows:
Figure GDA0003536398570000112
djan operator at the j-th layer; the derivation of the input to the weight value at the j-th layer in the above equation is as follows:
Figure GDA0003536398570000121
when the output layer is m, and when j is m, the operator of the j-th layer can be known by a full derivative formula
Figure GDA0003536398570000122
Wherein,
Figure GDA0003536398570000123
is the output of the j-th layer,
Figure GDA0003536398570000124
when j is less than m and is not an output layer, then
Figure GDA0003536398570000125
Where f (x) is the excitation function between layers. Each layer derives the excitation function by f' (x) ═ f (x) (1-f (x)), the above equation can be simplified:
dj＝dj+1Wjj+1f(x)(1-f(x))
from the above recursion formula, the operator of the previous layer can be sequentially deduced from the error of the output layer, and after the operator is obtained, the variation of the weight and the threshold of each layer is calculated as follows:
Δwij＝dj·yi
Δdj＝dj
adjusting the weight and the threshold value of each layer of neurons, wherein the connection weight and the relationship before the adjustment between the neurons in each layer are shown as the following formula:
wij(t+1)＝wij(t)+Δwij(t)+μΔwij(t-1)
dij(t+1)＝dj(t)+Δdj(t)
wherein t is the corrected times, mu is the inertia coefficient, and the training result is combined with the empirical value to automatically adjust.
Specifically, in step S5, after the wide area protection device can determine the fault section and the fault line after completing the calculation, the wide area protection device in the main station sends a blocking signal to the non-fault line circuit breaker, and cooperates with the conventional protection to implement quick, reliable and accurate fault removal; and meanwhile, the network calculates the position of the fault according to the measurement information of the related quantity of the wide area system and an uplink and downlink current ratio principle and reports the position.
Specifically, in step S6, the specific method for guiding the learning network to the failure analysis network is as follows:
s61, training by using the actual fault position and the corresponding data as input by the learning network, and perfecting and modifying the weight value and the threshold value of the learning network;
and S62, after the fault analysis network finishes the fault position measurement and calculation according to the fault data, the learning network analyzes and judges the fault data, and if the fault data is similar to the training data in the learning network, the learning network corrects the fault position measurement and calculation result.
When the patrol personnel arrives at the site to determine the real fault type, fault line and fault position, and deviation within a certain error allowable range may exist, the patrol personnel can inform the worker in the site to record the real data into the protection device, so that the protection device can train and learn the data of the time, and can automatically correct the calculation result when similar faults occur next time; along with the accumulation of fault data, the generalization capability of the network of the wide area protection system is stronger, the analysis of the fault is more accurate, and the operation of the line is guaranteed.
The invention provides a set of wide area protection system for an AT full parallel power supply network, the system not only can selectively remove fault lines, but also can position different types of faults, and has certain learning capability, so that different fault data can be learned, relevant parameters of the system can be corrected, and the accuracy of the system is improved; meanwhile, the method provides an idea for processing big data, and a large amount of short-circuit data can be more scientifically applied.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and should not be taken as limiting the invention, so that any modifications, equivalents, improvements and the like, which are within the spirit and principle of the present invention, should be included in the scope of the present invention.

Claims (6)

1. AT full parallel power supply network's wide area protection system based on artificial intelligence, its characterized in that includes:
s1, arranging wide area protection devices in power supply intervals of the AT full-parallel power supply network, wherein each wide area protection device in the power supply intervals is provided with one wide area protection device, and all the wide area protection devices are connected with each other;
s2, dividing each power supply interval into a main station and a plurality of sub stations, and uploading data of the sub station wide area protection devices to the main station wide area protection devices in real time;
s3, embedding an intelligent learning network in the wide area protection device of the main station, wherein the intelligent learning network consists of a fault analysis network and a learning network;
s4, when a fault occurs, the wide area protection device of the main station brings data in the fault time into a fault analysis network according to the fault time scale, and the fault analysis network calculates and calculates the fault position;
s5, after the fault analysis network measurement and calculation are completed, the main station wide area protection device locks the tripping of the non-fault line according to the measurement and calculation result and reports the fault position;
s6, inputting the actual fault position information of the patrol inspection into a main station wide area protection device, training a learning network, and guiding fault analysis network fault positioning by the trained learning network;
the specific working mode of the fault analysis network in the step S4 is as follows:
s41, the wide area protection device in the traction substation reads that the uplink T line current, the uplink F line current, the downlink T line current, the downlink F line current and the suck-up current are respectively as follows:
Figure FDA0003536398560000011
reading the voltage of an uplink T line, the voltage of an uplink F line, the voltage of a downlink T line and the voltage of a downlink F line as follows:
Figure FDA0003536398560000012
the wide area protection device in the AT station reads the same data, and the parameter names are sequentially
Figure FDA0003536398560000013
The wide area protection device in the subarea station reads the same data, and the parameter names are sequentially
Figure FDA0003536398560000014
S42, judging the power supply line, if so
Figure FDA0003536398560000015
Judging the fault type to be a TR/FR fault, otherwise, judging the fault type to be a TF fault;
s43, when the TF fault is judged to be a fault,
Figure FDA0003536398560000021
continuously judging the current direction in the place with larger current, wherein the current direction is a downlink fault if the current direction is from the uplink to the downlink, and the current direction is an uplink fault if the current direction is from the downlink to the uplink;
s44, if it is judged that TR/FR fault occurs
Figure FDA0003536398560000022
Judging as a downlink TR fault;
s45, under the condition that the fault type and the fault line are both determined, the fault analysis network adopts the uplink-downlink current ratio to carry out fault location, and the fault location formula is as follows:
Figure FDA0003536398560000023
wherein,
Figure FDA0003536398560000024
L1for up-link arm length, L2For descending power supply arm length, L is fault distance, Lc: correcting the distance;
in step S1, the power supply section is a normal power supply range including a traction substation, an AT station, a substation, an electric railway feeder line, a power supply line, and a catenary; each wide area protection device is connected with the corresponding satellite clock, and the wide area protection devices are connected through beacon channels.
2. The wide area protection system for an AT all-parallel power supply network based on artificial intelligence of claim 1, wherein: in step S2, the main substation wide area protection device receives real-time data of the sub substation wide area protection device once every 10ms through the beacon channel, so as to monitor the state of the sub substation.
3. The wide area protection system for an AT all-parallel power supply network based on artificial intelligence of claim 1, wherein the learning network in step S3 works as follows;
s46, building a learning network based on an artificial neural network, wherein the learning network consists of an input layer, a hidden layer and an output layer;
an input layer: the relevant measurement information collected by the wide area protection device is selected as input, and the following formula is shown:
Figure FDA0003536398560000031
hidden layer: the number of the hidden layers is two, and the number of nerves in each layer is 27; the excitation function of each layer of the hidden layer is logsig; the learning rate is 0.1; the maximum training times is 10000; the error goals achieved by training are: 1 e-30; the performance function adopts a sumsqr function;
an output layer: outputting three quantities of a fault type Y, a fault row G and a fault distance D as output;
s47, defining the working principle of the learning network: [ SS AT SP]T→[Y G D]TEstablishing a mapping relation between an input layer and an output layer;
s48, adjusting the weight matrix and the threshold matrix of the artificial neural network, and reducing the output error value of the learning network.
4. The wide area protection system for an AT all-parallel power supply network based on artificial intelligence of claim 3, wherein in the step S48, the specific method for reducing the output error value of the learning network is as follows:
s481, defining an Input layer as Input, a hidden layer first layer as output hidden1, a hidden layer second layer as output hidden2 and an output layer as Layout; carrying out normalization processing on the matrix of the input layer and the matrix of the output layer by means of premmx;
s482, initializing weights and thresholds among the input layer, the hidden layer and the output layer; defining the weights among the four layers as: [ W ]1]27×27,[W2]27×27,[W3]3×27(ii) a The thresholds between the four layers are defined as: [ B ]1]27×1,[B2]27×1,[B3]27×1；
S483, according to the artificial neural network principle, defining the transmission relation among the input matrix, the weight matrix and the threshold matrix:
hidden1＝logsig(W1·Input+B1)
hidden2＝logsig(W2•hidden1+B2)
layout＝logsig(W3•hidden2+B3)；
s484, setting the actual output as Right, defining the output error of the network as E, and setting the network performance function as sumsqr:
Figure FDA0003536398560000041
s485, adjusting a weight matrix and a threshold matrix of the artificial neural network, and reducing an output error E of the network;
suppose that the amount of change of the corresponding weight value of the i-th layer neuron to the j-th layer neuron is Δ wij；yiIs the output of the ith layer, xjFor the input of the j-th layer, in order to reduce the error the fastest, the weight change should be proportional to the reduction of the error E of each layer, i.e. the negative derivative of the error function to the weight, and the learning rate is alpha, delta wijThe calculated relationship of (a) is as follows:
Figure FDA0003536398560000042
s486, formula djFor the operator at the j-th layer, the derivation of the weight value by the input at the j-th layer is as follows:
Figure FDA0003536398560000043
when the output layer is m, and when j is m, the operator of the j-th layer can be known by a full derivative formula
Figure FDA0003536398560000044
Wherein,
Figure FDA0003536398560000045
is the output of the j-th layer,
Figure FDA0003536398560000046
when j is less than m and is not an output layer, then
Figure FDA0003536398560000047
Wherein, f (x) is an excitation function between layers; each layer derives the excitation function by f' (x) ═ f (x) (1-f (x)), the above equation is reduced to:
dj＝dj+1Wj j+1f(x)(1-f(x))
from the above recursion formula, the operator of the previous layer can be sequentially deduced from the error of the output layer, and after the operator is obtained, the variation of the weight and the threshold of each layer is calculated as follows:
Δwij＝dj•yi
Δdj＝dj
s487, adjusting the weights and the threshold values of the neurons in each layer, wherein the relationship between the connection weights and the unadjusted neurons in each layer is shown as the following equation:
wij(t+1)＝wij(t)+Δwij(t)+μΔwij(t-1)
dij(t+1)＝dj(t)+Δdj(t)
wherein t is the corrected times, mu is the inertia coefficient, and the training result is combined with the empirical value to automatically adjust.
5. The wide area protection system for the AT all-parallel power supply network based on artificial intelligence of claim 1, wherein the specific steps of the step S5 are as follows:
s51, after the measurement and calculation are completed, the main station wide area protection device determines a fault interval and a fault navigation category and sends a locking signal to the circuit breaker in the non-fault row;
and S52, calculating the fault position according to the measurement information of the related quantity of the wide area protection system and the uplink and downlink current ratio principle by the intelligent learning network, and reporting the fault position.
6. The wide-area protection system for an AT all-parallel power supply network based on artificial intelligence of claim 1, wherein in the step S6, the specific method for guiding the fault analysis network by the learning network is as follows:
s61, training by using the actual fault position and the corresponding data as input by the learning network, and perfecting and modifying the weight value and the threshold value of the learning network;
and S62, after the fault analysis network finishes the fault position measurement and calculation according to the fault data, the learning network analyzes and judges the fault data, and if the fault data is similar to the training data in the learning network, the learning network corrects the fault position measurement and calculation result.
CN202010193509.8A 2020-03-18 2020-03-18 Wide area protection system of AT full parallel power supply network based on artificial intelligence Active CN111541227B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010193509.8A CN111541227B (en) 2020-03-18 2020-03-18 Wide area protection system of AT full parallel power supply network based on artificial intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010193509.8A CN111541227B (en) 2020-03-18 2020-03-18 Wide area protection system of AT full parallel power supply network based on artificial intelligence

Publications (2)

Publication Number Publication Date
CN111541227A CN111541227A (en) 2020-08-14
CN111541227B true CN111541227B (en) 2022-04-26

Family

ID=71969174

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010193509.8A Active CN111541227B (en) 2020-03-18 2020-03-18 Wide area protection system of AT full parallel power supply network based on artificial intelligence

Country Status (1)

Country Link
CN (1) CN111541227B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4063166A (en) * 1975-06-05 1977-12-13 Bbc Brown Boveri & Company Limited Method for locating a fault on a line near to a measuring location with the aid of substitute signals
CN103983896A (en) * 2014-04-24 2014-08-13 云南电力试验研究院（集团）有限公司电力研究院 Distribution network line single end distance measurement result calibration method based on distribution network power quality sampled data
CN107315131A (en) * 2017-07-05 2017-11-03 成都交大许继电气有限责任公司 AT power supply trouble ranging data acquisition methods based on IEC61850 and fault waveform
CN108090658A (en) * 2017-12-06 2018-05-29 河北工业大学 Arc fault diagnostic method based on time domain charactreristic parameter fusion
CN110416979A (en) * 2019-07-22 2019-11-05 成都运达润泰信息科技有限公司 AT subregion institute bus bar protecting method in one's power under all-parallel AT traction system mode

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4063166A (en) * 1975-06-05 1977-12-13 Bbc Brown Boveri & Company Limited Method for locating a fault on a line near to a measuring location with the aid of substitute signals
CN103983896A (en) * 2014-04-24 2014-08-13 云南电力试验研究院（集团）有限公司电力研究院 Distribution network line single end distance measurement result calibration method based on distribution network power quality sampled data
CN107315131A (en) * 2017-07-05 2017-11-03 成都交大许继电气有限责任公司 AT power supply trouble ranging data acquisition methods based on IEC61850 and fault waveform
CN108090658A (en) * 2017-12-06 2018-05-29 河北工业大学 Arc fault diagnostic method based on time domain charactreristic parameter fusion
CN110416979A (en) * 2019-07-22 2019-11-05 成都运达润泰信息科技有限公司 AT subregion institute bus bar protecting method in one's power under all-parallel AT traction system mode

Also Published As

Publication number Publication date
CN111541227A (en) 2020-08-14

Similar Documents

Publication Publication Date Title
EP4030365A1 (en) Multi-mode multi-service rail transit analog simulation method and system
CN111768074B (en) Novel intelligent train operation adjusting method
CN100470252C (en) Electric network functional failure travelling wave positioning method
CN103324992B (en) Transformer risk prediction method based on markov and entropy weight fuzzy comprehensive evaluation
CN111724589A (en) Multi-source data-based highway section flow estimation method
CN105139268A (en) Power grid safety risk assessment method
CN109902948A (en) A kind of power transmission line monitoring system and method based on big data
CN103364693A (en) Electric transmission line traveling wave fault distance measurement method based on regional data
CN107870287A (en) A kind of localization method of distribution network failure
CN107167709A (en) A kind of electric network fault localization method and alignment system
CN116050940B (en) Labor dispatch winner and winner management system of mechanic platform
CN104698338A (en) Overhead power transmission line fault traveling wave locating method
CN116228466B (en) Big data analysis system of smart power grids
CN111242463A (en) Fault location method of AT single-wire power supply system based on BP neural network
CN111541227B (en) Wide area protection system of AT full parallel power supply network based on artificial intelligence
CN113376476A (en) PHM-based operation and maintenance system and method for medium and low voltage power distribution network
Zhang et al. An AI based high-speed railway automatic train operation system analysis and design
CN107067126A (en) It is a kind of based on power flow transfer than thermally-stabilised key transmission channel recognition method
CN105226666B (en) To the check method of reactive apparatus running status in transformer station's automatism voltage control
CN109389520A (en) A kind of electric power system fault method for pushing and system
US20240092407A1 (en) Method and device for multi-train operation trend deduction
CN1232416C (en) Controlling means and device for electric locomotive to pass the misphasing space automatically
CN114696467B (en) Method and system for analyzing and processing fault tripping event of high-voltage transmission long line
CN110458405A (en) It is a kind of based on electric power-Characteristics of micrometeorology data power system security method for early warning
CN109377044A (en) A kind of trouble hunting Cost Analysis Method of intelligent substation protection equipment

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