CN104318485A - Power transmission line fault identification method based on nerve network and fuzzy logic - Google Patents

Abstract

The invention relates to a power transmission line fault identification method based on a nerve network and fuzzy logic. The method comprises the following steps: 1), continuously acquiring a lead temperature, a lead inclination angle and lead tension data, and calculating a corresponding lead temperature deviation value, a corresponding lead inclination angle deviation value and a corresponding lead tension deviation value; 2), according to the three deviation values, calculating corresponding membership values, and if at least one of the three membership values is 1, executing step 3); 3), performing data fusion on the three membership values, and respectively calculating a line fault-free probability, a line potential fault probability and a line fault probability; and 4), according to the line potential fault probability, the line fault probability and fault duration, outputting an evaluation result. Compared to the prior art, the method provided by the invention has the advantages of growth performance, accurate prediction and the like.

Description

A kind of transmission line malfunction method of discrimination based on neural network and fuzzy logic

Technical field

The present invention relates to a kind of overhead transmission line state on_line monitoring technology, especially relate to a kind of transmission line malfunction method of discrimination based on neural network and fuzzy logic.

Background technology

Electric system by sending out in a large number, give, defeated, join, the equipment connection such as electricity consumption forms, the reliability of these equipment and ruuning situation directly decide the stable of whole system and safety, also determine the q&r of power supply.High pressure electric line is the important step that electric energy realizes long-distance sand transport, is the lifeblood of whole electrical network, and the long-term safety stable operation of electrical network in the factor serious threat such as meteorological disaster, artificial destruction, circuit long-time running are aging.In January, 2008, motherland tens provinces, south and area suffer from serious ice and snow weather, and most circuit presents large area icing, and transmission pressure load and transmission tower load cause broken string and shaft tower collapse accident because of severe overweight, electric power facility destroys heavy, causes unprecedented loss.Therefore, we should strengthen the monitoring to transmission line status, give warning in advance the safe and stable operation ensureing electrical network.Present stage monitors line status mainly through the artificial way of patrol to circuit, and human resources are limited, far can not meet actual requirement, let alone reaches the object to the monitoring completely in real time of transmission line of electricity situation.Along with the progress of computer communication network technology and sensor network technique, we have the ability to realize remote monitoring to power transmission state monitoring.The monitoring system of present stage can photograph transmission line information and circuit surrounding enviroment information by various kinds of sensors information and video camera, by corresponding communication network, information is sent to the Surveillance center of setting, thus staff judges by the information collected the situation that electric line is instant, give warning in advance and discovery situation solve fault in time, guarantee the safe and stable operation of circuit.

Transmission pressure on-line monitoring system plays an important role in whole operation states of electric power system monitoring system.The running status of circuit can be drawn by the information such as temperature, inclination angle, sag, tension force, load of wire.When line load is overweight, the temperature of wire raises, inclination angle increases, thus causes damage and the quick aging of wire and gold utensil, and the sag of wire is crossed conference and made short circuit and flashover easily become possibility.And when wire icing is excessive, the inclination angle of wire, sag, tension force and load all can become large, broken string and shaft tower correspondingly can be caused to be out of shape and to collapse.Power line conductive line monitoring system can draw the status information of circuit timely, early warning in time before fault occurs, goes to process circuit question to staff, effectively can reduce the incidence of line fault the more time, the utilization factor of raising equipment, is worth further investigation.

Summary of the invention

Object of the present invention is exactly the transmission line malfunction method of discrimination based on neural network and fuzzy logic providing a kind of accuracy of judgement in order to overcome defect that above-mentioned prior art exists.

Object of the present invention can be achieved through the following technical solutions:

Based on a transmission line malfunction method of discrimination for neural network and fuzzy logic, the method comprises the following steps:

1) continuous acquisition conductor temperature, wire inclination angle and wire tension data, and calculate corresponding conductor temperature deviate v ₁(k), wire inclination deviation value v ₂(l) and wire tension deviate v ₃(m);

2) according to conductor temperature deviate v ₁(k), wire inclination deviation value v ₂(l) and wire tension deviate v ₃m () calculates corresponding fault and is subordinate to angle value, and judge that three faults are subordinate to angle value and whether are 0, if NO, then performs step 3), if yes, then return step 1);

3) adopt BP neural network to be subordinate to angle value to three faults and carry out data fusion, calculate circuit probability of nonfailure, circuit incipient fault probability and line fault probability respectively;

4) evaluation result is exported according to circuit incipient fault probability, line fault probability and trouble duration.

The fault of described conductor temperature is subordinate to angle value and is:

μ(V ₁)＝u[v ₁(k)-D ₁]

Wherein: u (x) is unit step function, D ₁for the decision threshold of conductor temperature;

The fault at described wire inclination angle is subordinate to angle value and is:

μ(V ₂)＝u[v ₂(l)-D ₂]

Wherein: D ₂for the decision threshold at wire inclination angle;

The fault of described wire tension is subordinate to angle value and is:

μ(V ₃)＝u[v ₃(m)-D ₃]

Wherein: D ₃for the decision threshold of wire tension.

Described BP neural network is three input three output modes, comprises an input layer, two hidden layers and an output layer, transport function Sigmoid type function:

$O=1/[1+\exp (-\mathrm{\Σ}{x}_p{w_{\mathrm{ij}}}^n-\mathrm{\θ})]$

Wherein: O represents that neuron exports, x _pfor input; w _ij ⁿfor the connection weights of n-th layer i-th node and (n+1) layer jth node; θ is the initial value of supposition, is set to 0 as just started.

The nodes of described two hidden layers is 7.

Described step 4) specifically comprise step:

401) circuit incipient fault probability y is imported ₂with line fault probability y ₃if, y ₂and y ₃all be greater than 0.7, then outlet line breaks down, if y ₂and y ₃all be less than 0.2, then outlet line non-fault, otherwise, perform step 402);

402) the fuzzy membership angle value of computational scheme incipient fault probability, line fault probability and trouble duration;

403) utilize the fuzzy membership angle value of incipient fault probability, line fault probability and trouble duration to carry out fuzzy reasoning in conjunction with fuzzy rule base and export fuzzy reasoning result U, and export evaluation result.

The fuzzy membership angle value of described circuit incipient fault probability is:

$\mathrm{\μ}\left(y_2\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(y_2-{\mathrm{\μ}}_1)}^2/2{\mathrm{\σ}}^2]$

Wherein: y ₂for circuit incipient fault probability, σ is the standard deviation of circuit incipient fault probability, and value is 0.4, μ ₁for the mean value of circuit incipient fault probability, value is 0,0.5 and 1;

The fuzzy membership angle value of described line fault probability is:

$\mathrm{\μ}\left(y_3\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(y_3-{\mathrm{\μ}}_2)}^2/2{\mathrm{\σ}}^2]$

Wherein: y ₃for line fault probability, σ is the standard deviation of line fault probability, and value is 0.4, μ ₂for the mean value of line fault probability, value is 0,0.5 and 1;

The fuzzy membership angle value of described trouble duration is:

$\mathrm{\μ}\left(T\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(T-{\mathrm{\μ}}_3)}^2/2{\mathrm{\σ}}^2]$

Wherein: T is trouble duration, σ is the standard deviation of trouble duration, and value is 0.4, μ ₃for the threshold value of trouble duration, value is T ₁and T ₂.

Compared with prior art, the present invention has the following advantages:

1) when transmission line of electricity breaks down, often the inclination angle of wire, sag, tension force and load all can become large, therefore can be drawn the running status of circuit by the information such as temperature, inclination angle, sag, tension force, load of wire, never avoid broken string and shaft tower distortion to collapse.

2) accuracy of prediction can be improved based on the transmission line status appraisal procedure of neural network and fuzzy logic decision.

Accompanying drawing explanation

Fig. 1 is the transmission line status assessment models block diagram that the present invention is based on neural network and fuzzy logic decision;

Fig. 2 is that the present invention assesses design cycle;

Fig. 3 is BP neural network design frame chart of the present invention;

Fig. 4 is fuzzy logic decision system block diagram of the present invention.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.The present embodiment is implemented premised on technical solution of the present invention, give detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.

As shown in Figure 1, transmission line status assessment models adopts BP neural network fusion circuit local fault to be subordinate to angle value, and fuzzy logic decision infers the end-state of line fault.Through data prediction, Monitoring Data first show that local line fault is subordinate to angle value before entering neural network, line fault is subordinate to angle value and merges rear outlet line non-fault, incipient fault as the input of neural network and have the shape probability of states such as fault, do according to a preliminary estimate according to probability of malfunction to line status, to still judging that the probable value of line fault conditions is input to fuzzy logic decision system, according to expertise and historical data design fuzzy rule base, decision making package calculates final assessment result.

Based on a transmission line malfunction method of discrimination for neural network and fuzzy logic, as shown in Figure 2, method comprises the following steps:

1) continuous acquisition conductor temperature, wire inclination angle and wire tension data, and calculate corresponding conductor temperature deviate v ₁(k), wire inclination deviation value v ₂(l) and wire tension deviate v ₃(m), deviate is the difference of detected value and standard value;

2) according to conductor temperature deviate v ₁(k), wire inclination deviation value v ₂(l) and wire tension deviate v ₃m () calculates corresponding fault and is subordinate to angle value, and judge that three faults are subordinate to angle value and whether are 0, if NO, then performs step 3), if yes, then return step 1);

The fault of conductor temperature is subordinate to angle value and is:

μ(V ₁)＝u[v ₁(k)-D ₁]

Wherein: u (x) is unit step function, D ₁for the decision threshold of conductor temperature;

The fault at wire inclination angle is subordinate to angle value and is:

μ(V ₂)＝u[v ₂(l)-D ₂]

Wherein: D ₂for the decision threshold at wire inclination angle;

The fault of wire tension is subordinate to angle value and is:

μ(V ₃)＝u[v ₃(m)-D ₃]

Wherein: D ₃for the decision threshold of wire tension.

Above-mentioned 3 decision threshold are determined according to the specification of wire.

3) adopt BP neural network to be subordinate to angle value to three faults and carry out data fusion, calculate circuit probability of nonfailure, circuit incipient fault probability and line fault probability respectively;

As shown in Figure 3, neural network model design adopts BP neural network fusion line fault to be subordinate to angle value, calculates and outlet line probability of malfunction value.Input signal x ₁, x ₂, x ₃be respectively the fault that fault is subordinate to angle value, inclination angle causes that the temperature after data prediction causes to be subordinate to the fault that angle value and pulling force cause and to be subordinate to angle value, the output signal R after BP neural network fusion ₁, R ₂r ₃represent the states such as circuit non-fault, incipient fault, line fault respectively, y ₁, y ₂, y ₃represent that it occurs this shape probability of state respectively.

BP neural network is supervised learning neural network, generally first carry out off-line learning (data training and performance verification), then application on site carries out fault diagnosis (neural network trained drops into practical application), its learning process and use procedure separate, and fault diagnosis result accuracy is high.Owing to being three input three output modes, the number of hidden nodes that in literary composition, selection one is larger is trained, and generally chooses by following experimental formula:

$n_H=\sqrt{n_i+n_O}+l$

In formula, n _hfor hidden node number; n _ifor input node number; n _ofor output node number; L is the integer between 1-10.According to practical problems analytical test, neural network hidden layer is finally defined as 2 layers, and hidden node number is 7.

BP neural network model design ground floor is the input layer of neural network, and the second layer and third layer are hidden layer, and the 4th layer is output layer.

Transport function Sigmoid type function.

O＝1/[1+exp(-∑x _pw _ij ⁿ-θ)]

Wherein: O represents output, x _pfor input; w _ij ⁿfor the connection weights of n-th layer i-th node and (n+1) layer jth node; θ is threshold value.

Select some input data values according to reality detection and expertise, and they are divided into training dataset and verification msg collection.Training dataset is used for training BP neural network.Here 14 data points are selected, desired output R ₁, R ₂r ₃.Utilize the data point selected to screw up discipline neural network weight, design the BP neural network being determined malfunction by arbitrary data point.Table 1 represents the corresponding output valve of malfunction, and 100 represent unfaulty conditions, and 010 represents circuit incipient fault state, and 001 represents that circuit has malfunction.The coordinate figure x of each data point ₁, x ₂, x ₃as input value, output valve is R ₁, R ₂r ₃, according to output valve computational analysis circuit unfaulty conditions probability, incipient fault probability, have probability of malfunction, result is with y ₁, y ₂, y ₃export.Network training uses Sigmoid function as transport function.

4) be subordinate to according to any one fault in circuit incipient fault probability, line fault probability and trouble duration (trouble duration is with step 1) and start when angle value is 1 to calculate, until this time interval judged) export evaluation result, specifically comprise step:

Table 1

401) circuit incipient fault probability y is imported ₂with line fault probability y ₃if, y ₂and y ₃all be greater than 0.7, then outlet line breaks down, if y ₂and y ₃all be less than 0.2, then outlet line non-fault, otherwise, perform step 402);

402) the fuzzy membership angle value of computational scheme incipient fault probability, line fault probability and trouble duration;

The anti fuzzy method of fuzzy reasoning result adopts gravity model appoach computational scheme probability of malfunction, and be greater than 0.6 if analyze according to calculating gained probability U, decision-making is line fault conditions (PB), and being less than or equal to 0.6 decision-making is unfaulty conditions (PS).

Fuzzy logic decision system block diagram as shown in Figure 4, y ₂(incipient fault probability), y ₃the input quantity of fuzzy introduction in (probability of malfunction), T (trouble duration) corresponding fuzzy logic system, the output quantity of the corresponding fuzzy logic control system of U.Given y ₂domain be A, [0,1]; y ₃domain B, [0,1]; The domain of T is C, [0,1]; The domain of U is D, [0,1], and the fuzzy language value of division is { having fault, non-fault }.By experience and can by y to the historical experience probability of the order of severity of line fault ₂, y ₃fuzzyly change into 3 grades, honest (PB), center (PM), just little (PS), change into 2 grades by fuzzy for T and U, honest (PB) and just little (PS).The membership function of these fuzzy sets is represented with normal distyribution function μ (x).

The fuzzy membership angle value of circuit incipient fault probability is:

$\mathrm{\μ}\left(y_2\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(y_2-{\mathrm{\μ}}_1)}^2/2{\mathrm{\σ}}^2]$

Wherein: y ₂for circuit incipient fault probability, σ is the standard deviation of circuit incipient fault probability, and value is 0.4, μ ₁for the mean value of circuit incipient fault probability, value is 0,0.5 and 1;

The fuzzy membership angle value of line fault probability is:

$\mathrm{\μ}\left(y_3\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(y_3-{\mathrm{\μ}}_2)}^2/2{\mathrm{\σ}}^2]$

Wherein: y ₃for line fault probability, σ is the standard deviation of line fault probability, and value is 0.4, μ ₂for the mean value of line fault probability, value is 0,0.5 and 1;

The fuzzy membership angle value of trouble duration is:

$\mathrm{\μ}\left(T\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(T-{\mathrm{\μ}}_3)}^2/2{\mathrm{\σ}}^2]$

Wherein: T is trouble duration, σ is the standard deviation of trouble duration, and value is 0.4, μ ₃for the threshold value of trouble duration, value is T ₁and T ₂, T ₁and T ₂depend on the circumstances, 0 and 1 can be respectively.

Circuit incipient fault probability, line fault probability are changed into 3 grades by fuzzy: honest (PB), center (PM), just little (PS);

Trouble duration and fuzzy reasoning result is fuzzy changes into 2 grades: honest (PB) and just little (PS).

Such as, by y ₂substitute into following formula:

$\mathrm{\μ}\left(y_2\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(y_2-{\mathrm{\μ}}_1)}^2/2{\mathrm{\σ}}^2]$

If work as μ ₁when being 0.5, μ (y ₂) maximum, then circuit incipient fault probability is turned to PS by fuzzy, if μ ₁when being 0.5, μ (y ₂) maximum, then fuzzyly turn to PM, if μ ₁when being 1, μ (y ₂) maximum, then fuzzyly turn to PB.

By y ₃substitute into following formula:

$\mathrm{\μ}\left(y_3\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(y_3-{\mathrm{\μ}}_2)}^2/2{\mathrm{\σ}}^2]$

If work as μ ₂when being 0.5, μ (y ₃) maximum, then line fault probability is turned to PS by fuzzy, if μ ₂when being 0.5, μ (y ₃) maximum, then fuzzyly turn to PM, if μ ₂when being 1, μ (y ₃) maximum, then fuzzyly turn to PB.

T is substituted into following formula:

$\mathrm{\μ}\left(T\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(T-{\mathrm{\μ}}_3)}^2/2{\mathrm{\σ}}^2]$

If work as μ ₃when being 0.5, μ (T) is maximum, then trouble duration is turned to PS by fuzzy, if μ ₃when being 1, μ (T) is maximum, then fuzzyly turn to PB.

Table 2

y 2	y 3	T	U
				PS	PS	PS	PS
PS	PM	PS	PS
				PS	PB	PS	PS
PS	PS	PB	PS
				PS	PM	PB	PB
PS	PB	PB	PB
				PM	PS	PS	PS
PM	PM	PS	PB
				PM	PS	PB	PB
PM	PB	PS	PB
				PB	PS	PS	PS
PB	PS	PB	PB

Utilize " If-Then " rule (R rule), in conjunction with actual electric line on-line monitoring system, get rid of the fuzzy rule of contradiction, totally 12 rules, as shown in table 2.The minimum operation rule of Mamdani is adopted to calculate fuzzy relation based on fuzzy rule base, the anti fuzzy method of fuzzy reasoning result adopts gravity model appoach computational scheme probability of malfunction, if be greater than 0.6 according to the probability analysis of calculating gained, decision-making is line fault conditions, being less than or equal to 0.6 decision-making is unfaulty conditions, namely as in table 2, table look-up obtain U be PB then outlet line there is fault, if U is PS, outlet line does not have fault.

Because line fault is by various factors, be difficult to set up specific analytic model between influence factor and fault.Single factors assessment accuracy rate is not high, in order to improve the accuracy rate of Guangdong power system state estimation, the present invention is based on transmission line online monitoring system, consider the multi-sensor monitoring parameters such as temperature signal, dip angle signal, pulling force signal, propose the transmission line status assessment models based on neural network and fuzzy logic decision.First carry out pre-service to Monitoring Data to calculate local fault and be subordinate to angle value, then the probable value that BP neural computing that angle value input trains draws line fault conditions is subordinate to, the probable value of last fuzzy uncertain is as the input variable of inference system, set up fuzzy rule base according to expertise, complex reasoning goes out the accurate malfunction of circuit.This method can realize the assessment of intelligent transmission line status, and assessment result accuracy rate improves greatly.

Claims (6)

1., based on a transmission line malfunction method of discrimination for neural network and fuzzy logic, it is characterized in that, the method comprises the following steps:

1) continuous acquisition conductor temperature, wire inclination angle and wire tension data, and calculate corresponding conductor temperature deviate v ₁(k), wire inclination deviation value v ₂(l) and wire tension deviate v ₃(m);

2) according to conductor temperature deviate v ₁(k), wire inclination deviation value v ₂(l) and wire tension deviate v ₃m () calculates corresponding fault and is subordinate to angle value, and judge that three faults are subordinate to angle value and whether are 0, if NO, then performs step 3), if yes, then return step 1);

3) adopt BP neural network to be subordinate to angle value to three faults and carry out data fusion, calculate circuit probability of nonfailure, circuit incipient fault probability and line fault probability respectively;

4) evaluation result is exported according to circuit incipient fault probability, line fault probability and trouble duration.

2. a kind of transmission line malfunction method of discrimination based on neural network and fuzzy logic according to claim 1, it is characterized in that, the fault of described conductor temperature is subordinate to angle value and is:

μ(V ₁)＝u[v ₁(k)-D ₁]

Wherein: u (x) is unit step function, D ₁for the decision threshold of conductor temperature;

The fault at described wire inclination angle is subordinate to angle value and is:

μ(V ₂)＝u[v ₂(l)-D ₂]

Wherein: D ₂for the decision threshold at wire inclination angle;

The fault of described wire tension is subordinate to angle value and is:

μ(V ₃)＝u[v ₃(m)-D ₃]

Wherein: D ₃for the decision threshold of wire tension.

3. a kind of transmission line malfunction method of discrimination based on neural network and fuzzy logic according to claim 1, it is characterized in that, described BP neural network is three input three output modes, comprises an input layer, two hidden layers and an output layer, transport function Sigmoid type function:

O＝1/[1+exp(-∑x _pw _ij ⁿ-θ)]

Wherein: O represents that neuron exports, x _pfor input; w _ij ⁿfor the connection weights of n-th layer i-th node and (n+1) layer jth node; θ is the initial value of supposition, is set to 0 as just started.

4. a kind of transmission line malfunction method of discrimination based on neural network and fuzzy logic according to claim 3, it is characterized in that, the nodes of described two hidden layers is 7.

5. a kind of transmission line malfunction method of discrimination based on neural network and fuzzy logic according to claim 1, is characterized in that, described step 4) specifically comprise step:

401) circuit incipient fault probability y is imported ₂with line fault probability y ₃if, y ₂and y ₃all be greater than 0.7, then outlet line breaks down, if y ₂and y ₃all be less than 0.2, then outlet line non-fault, then, step 402 don't be performed);

402) the fuzzy membership angle value of computational scheme incipient fault probability, line fault probability and trouble duration;

403) utilize the fuzzy membership angle value of incipient fault probability, line fault probability and trouble duration to carry out fuzzy reasoning in conjunction with fuzzy rule base and export fuzzy reasoning result U, and export evaluation result.

6. a kind of transmission line malfunction method of discrimination based on neural network and fuzzy logic according to claim 5, it is characterized in that, the fuzzy membership angle value of described circuit incipient fault probability is:

$\mathrm{\μ}\left(y_2\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(y_2-{\mathrm{\μ}}_1)}^2/{2\mathrm{\σ}}^2]$

Wherein: y ₂for circuit incipient fault probability, σ is the standard deviation of circuit incipient fault probability, and value is 0.4, μ ₁for the mean value of circuit incipient fault probability, value is 0,0.5 and 1;

The fuzzy membership angle value of described line fault probability is:

$\mathrm{\μ}\left(y_3\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(y_3-{\mathrm{\μ}}_2)}^2/{2\mathrm{\σ}}^2]$

Wherein: y ₃for line fault probability, σ is the standard deviation of line fault probability, and value is 0.4, μ ₂for the mean value of line fault probability, value is 0,0.5 and 1;

The fuzzy membership angle value of described trouble duration is:

$\mathrm{\μ}\left(T\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\exp [-{(T-{\mathrm{\μ}}_3)}^3/{2\mathrm{\σ}}^2]$

Wherein: T is trouble duration, σ is the standard deviation of trouble duration, and value is 0.4, μ ₃for the threshold value of trouble duration, value is T ₁and T ₂.