CN106951687A - Transformer insulated Stress calculation and evaluation method based on fuzzy logic and evidential reasoning - Google Patents

Abstract

The invention discloses a kind of transformer insulated stress appraisal method based on fuzzy logic and evidential reasoning, comprise the following steps：Field data is normalized first, will describe the dimensionless variable that the input datas such as transformer normal value, limiting value are transformed to [0,1] region；Secondly obfuscation input data, that is, map that to fuzzy membership functions, and trapezoidal membership function, generalized bell membership function or Gauss member function can be used；Then transformer stress appraisal tree-model is built, that is, builds each node and its connected mode from transformer input data to transformer insulated stress, node layer is associated with Fuzzy Correlation matrix up and down；Then the weights between node are determined, i.e., the objective weight-values analyzed using the subjective weights and characteristic vector of level of analysis processing method obtain synthesis weights；Transformer insulated stress is finally obtained, i.e., using inference mechanism and combining evidences rule, determines the stress levels overall of transformer.This method has advantage in terms of the ability of accuracy, flexibility and processing measuring uncertainty.

Description

Transformer insulation stress calculation and evaluation method based on fuzzy logic and evidence reasoning

[ technical field ] A method for producing a semiconductor device

The invention relates to the technical field of power systems, in particular to a transformer insulation stress calculation and evaluation method based on fuzzy logic and evidence reasoning.

[ background of the invention ]

One of the most critical devices in an electrical power system is the transformer, whose availability and reliability are not allowed to suffer. Load optimization of existing transformers is a subject of much utility priority, which necessitates that the equipment operate under the stress level limitations of its design. As the MVA and kV ratings of transformers continue to increase, which inevitably leads to premature failure of the original specifications, the stress levels of the transformers need to be evaluated and managed as planned. Transformers typically age over time and thus have a progressively lower ability to withstand operational stresses.

The core, bushings and windings of a transformer are generally considered to have infinite life due to their very slow aging rate. The oil paper insulation system is therefore a system with a limited lifetime, determining the lifetime of the transformer. Since the oil can be reconditioned, recycled or replaced, unlike paper, the paper will determine the ultimate end of life of the transformer. The insulation life of a transformer is mainly affected by water, oxygen, pollution, heat and faults. The first three can be reduced by current insulation handling and preservation techniques, but the last two must be maintained by appropriate cooling and protection systems. The expected insulation life can be seen as a function of the transformer design, material quality and process, the operational stresses imposed on the transformer and its corresponding maintenance. Events such as short circuits, over-voltages and moisture ingress can overload the transformer and accelerate its insulation degradation. Most transformers have a design life of up to 40 years. Transformer failure is a complex event involving mainly oil paper electrolyte, oxidation, hydrolysis and pyrolytic degradation processes.

To determine the approximate stress level of the insulation, a number of tests may be performed. These test data, when combined, can be used to determine when the transformer enters an area of increased unreliability. This would allow managers to plan what actions should be taken next in the maintenance, loading, relocation or decommissioning of the transformer. There are two major sources of transformer stress-failure and aging. Fault stresses are the result of exceeding electrical and adiabatic withstand limits in the form of partial discharges, arcing, local overheating, and the like. They are routinely diagnosed by Dissolved Gas Analysis (DGA). Aging stress is the result of degradation of insulation over time. They are generally characterized by oxidation, hydrolysis and pyrolysis by-products of the oilpaper insulation system. These stresses are evaluated by measuring insulation variables such as moisture content, interfacial tension, dielectric strength, acidity, furan, tension, dielectric strength, degree of polymerization, Tan, aldehydes, alcohols, and ketones. The primary goal of the asset manager is to maximize the utilization of the transformer over its design life. A number of methods have been used for transformer state estimation and fault to be used for transformer state estimation and fault diagnosis, including but not limited to fuzzy logic, neural networks, bayesian networks, neural fuzzy and evidence reasoning. In some utilities, a scoring method is typically used that compares the test results to industry-accepted thresholds or previous test results on the device, and sometimes device test data similar to its design, structure, age, and operating environment. Due to the increased demand for reliable power supplies, reduced maintenance budgets and strict minimum downtime, accurate stress assessment models are required which have to include as much status data about the transformer as possible.

The transformer stress level assessment can be viewed as a tree with two main branches, fault and aging, each main branch in turn having a sub-branch and a sub-branch. The subtrees are the inputs to the evaluation and represent various parameters of the transformer measurement, such as the size and rate of gas produced by the fault, dielectric loss, moisture and furans, as shown in fig. 1.

[ summary of the invention ]

The invention provides a transformer insulation stress calculation and evaluation method based on fuzzy logic and evidence reasoning.

The invention adopts the following technical scheme:

the transformer insulation stress calculation method based on fuzzy logic and evidence reasoning comprises the following steps:

(1) normalizing field data, namely converting dissolved gases of hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, carbon dioxide and furan in the transformer oil, acidity, moisture content, interfacial tension, dielectric strength and Tan into dimensionless variables of a [0,1] area;

(2) constructing a transformer stress evaluation tree model by taking the data normalized in the step (1) as input, namely constructing each node from transformer input data to transformer insulation stress and a connection mode thereof, and using a fuzzy incidence matrix R for upper and lower nodes_m×nAssociating;

(3) fuzzification processing is carried out on the data normalized in the step (1), and finally a central defuzzification method is used for obtaining a total stress value of the transformer;

(4) determining weights among nodes of transformer stress evaluation tree model, namely weights W using analytical hierarchy processing method_type1And weight W of eigenvector analysis_type2To obtain a composite weight W_final；

(5) Adopting the synthesis weight W obtained in the step (4)_finalAnd (4) adjusting the total stress value of the transformer obtained in the step (3), namely determining the total stress level of the transformer by using an inference mechanism and an evidence synthesis rule.

Further, if the normalized variable is 1, it indicates that the equipment is operated in the area with high fault risk, if the normalized variable is 0, it indicates that the insulation condition is in the best state, and for the normalized variable between 0 and 1, the following steps are further divided from low to high in sequence: the first level, the second level, the third level, the fourth level and the fifth level.

Further, according to different situations, different fuzzy membership functions are adopted for fuzzification processing, which specifically comprises the following steps:

trapezoidal membership function: the trapezoid membership function is determined by four parameters a, b, c and d, wherein the parameters a and b determine the angle of the trapezoid, the parameters c and d determine the shoulder of the trapezoid, and the expression is shown as formula (1):

wherein, x is a normalized variable, the parameter a is a rising starting point of the trapezoid membership function, and the trapezoid membership function value is increased from 0 to 1 from the point b; the parameter b is the rising end point of the trapezoidal membership function, and the trapezoidal membership function value reaches the maximum value 1 from the point b; the parameter c is the starting point of the decline of the trapezoidal membership function, and the trapezoidal membership function value declines from 1 to 0 from the point c; the parameter d is the end of the fall of the trapezoidal membership function, which remains 0 from point d.

Generalized bell membership function: the generalized bell-shaped membership function is determined by three parameters a ', b ' and c ', and is expressed as formula (2):

where x is the normalized variable, a 'is the scaling ratio to the normalized variable x, and b' represents a factorB ' and a ' together establish the shape of the curve, c ' being the abscissa of the center point of the curve.

The gaussian membership function is determined by two parameters c' and sigma, respectively defining the center and shape of the curve, and is expressed by the following formula (3):

wherein, x is a normalized variable, c' determines the center of the Gaussian distribution curve, and sigma determines the dispersion degree of the Gaussian distribution curve.

Further, the specific method for obtaining the total stress value of the transformer in the step (3) comprises the following steps: an optimal stress evaluation level of the transformer is established based on the respective basic confidence assignments p (λ) for each evaluation level, applying an evidence reasoning criterion, and based on the following conditions.

For M rules, the sum L of the outputs of all rules_outThe expression of (a) is:

where M is the number of rules for the input trigger, L^jIndicating that in the current input, the row vector inference engine unites all triggered rules;

after the output of each node of the transformer stress evaluation tree model is obtained, a central defuzzification method is used for determining the expression of clear output y (x) of a fuzzy logic system as follows:

where u is the number of quantization levels output, y^lIs the quantity output at the first quantization level, μ_B(y^l) Is the true membership value of the domain point.

Further, the fuzzy logic rule at each node is:

wherein A is_iAnd B_iInput representing a node of the hierarchy, C_iRepresenting the output of the node of the current level, also as part of the input of the node of the next level, f_iIs a numerical value mapping rule of two layers of nodes,andis a specific value, i represents a node label.

Further, the weight W is synthesized_finalCalculated according to the following formula:

W_final＝e^θW_type1+(1-e^θ)W_type2(11)

wherein, theta represents weight type W_type1The specific value of the contribution rate is flexibly determined according to the historical data of the transformer.

The synthesis mode of the evidences of other nodes in the transformer stress evaluation tree model except the nodes where the fault stress and the aging stress are located is as follows:

wherein, P is the node number where the variables are merged, and P is 1,2, 3. N is the number of input variables at each node, W_PThe node weighting matrix is used for carrying out weighted summation on fuzzy membership values of all levels so as to obtain variable values after node combination; matrix Ψ_PElement mu of_1j，μ_2j...μ_5jFuzzy membership values representing five levels associated with the jth input variable;

stress at failureAnd aging stressThe multiple evidence combination rule of (2) is defined as:

wherein,give thisThe degree of conflict of these sources of evidence,andnot zero, their intersection creates an empty set.

A transformer insulation stress evaluation method based on fuzzy logic and evidence reasoning is disclosed, which is based on the respective basic confidence distribution (P (lambda))₁，(P(λ))₂，…，(P(λ))₅The optimum stress level is established according to the following conditions:

InsulDeg＝subscript{max{(P(λ))₁,(P(λ))₂,...,(P(λ))₅}} (10)

wherein InsulDeg is the insulation stress level of the transformer, (P (lambda))_rFor the basic confidence allocation of the r-th rank, r ∈ {1, 2,3, 4, 5}, the script is a subscript function, i.e., it satisfies the script { (P (λ))_rR and requires that the best rank must have a basic confidence allocation that exceeds the next best rank by some threshold, which is not less than 5%.

Compared with the prior art, the invention has at least the following beneficial effects: the existing transformer insulation performance evaluation is mainly based on a grade standard given by the industry in practice, and the insulation performance of the transformer is determined according to the grade value after manual grading. The grading standard file is relatively stable, and the transformer performance and other aspects are continuously updated, so the transformer insulation performance evaluation method based on manual grading has certain limitations. The invention adopts a fuzzy logic and evidence reasoning method to evaluate the insulation stress of the transformer, and the result shows that the application of evidence reasoning improves the flexibility and the capability of processing the measurement uncertainty, and the application of fuzzy logic has advantages in the aspect of a large number of expert databases.

[ description of the drawings ]

FIG. 1 is a schematic diagram of a stress evaluation tree model of one of the transformers according to the present invention;

FIG. 2 is a circuit diagram of fuzzy logic stress evaluation of a transformer according to the present invention;

fig. 3 is a numerical mapping rule chart of the insulation stress evaluation of the transformer according to the present invention.

[ detailed description ] embodiments

The technical solution of the present invention will be described in detail with reference to the accompanying drawings and preferred embodiments.

The invention provides a transformer insulation stress calculation method based on fuzzy logic and evidence reasoning, which comprises the following steps:

s100: normalizing field data, namely using amplitudes and change rates of variables such as dissolved gas hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, carbon dioxide, furan, acidity, moisture content, interfacial tension, dielectric strength, tan and the like in the transformer oil as input data, converting the input data into dimensionless variables in a [0,1] area, carrying out fuzzy processing on the normalized data, namely mapping the normalized data to a fuzzy membership function, wherein a trapezoidal membership function, a generalized bell-shaped membership function or a Gaussian membership function can be used;

s200: constructing a transformer stress evaluation tree model, namely constructing each node from transformer input data to transformer insulation stress and a connection mode thereof, and using a fuzzy incidence matrix R for upper and lower nodes_m×nAssociating;

s300: determining weights among nodes of transformer stress evaluation tree model, namely weights W using analytical hierarchy processing method_type1And weight W of eigenvector analysis_type2To obtain a composite weight W_final；

S400: and obtaining the insulation stress of the transformer, namely determining the total stress level of the transformer by using an inference mechanism and an evidence synthesis rule.

S500: the total stress level of the transformer was evaluated.

Referring to fig. 1, the method of constructing the transformer stress evaluation tree model according to the present invention is as follows:

transformer insulation stress is mainly due to two main stress sources, fault stress and aging stress. The failure stress is divided into two main aspects of thermal failure stress and electrical failure stress, wherein the thermal failure stress is mainly caused by over-heated oil and over-heated paper; the electrical fault stress is divided into two specific aspects of partial discharge and electric arc, and the faults are generally measured by the content of carbon monoxide, carbon dioxide, acetylene, ethylene, methane, hydrogen, acetylene and other gases. The aging stress is conceptually divided into three types, namely chemical aging stress, physical aging stress and electrical aging stress, the chemical aging stress is generally calculated by furan and acidity, the physical aging stress can be evaluated by moisture content and interfacial tension, and the electrical aging stress is specifically expressed by dielectric strength and Tan. Therefore, in the transformer insulation stress evaluation, the base layer input is a numerical value of such factors as carbon monoxide, carbon dioxide, ethylene, acetylene, methane, hydrogen, acetylene, furan, acidity, moisture content, interfacial tension, dielectric strength, and Tan. The amplitude and the change rate of the factors such as carbon monoxide, carbon dioxide, ethylene, acetylene, methane, hydrogen and acetylene are used for evaluating the degrees of overheating oil, overheating paper, partial discharge and electric arc, wherein the data of the overheating oil and the overheating paper can further evaluate the thermal fault stress, and the data of the partial discharge and the electric arc are used for evaluating the electrical fault stress, and the thermal fault stress and the electrical fault stress jointly determine the fault stress of the transformer. Chemical aging stress, physical aging stress and electrical aging stress can be evaluated respectively by using the amplitudes and the change rates of the factors of furan, acidity, moisture content, interfacial tension, dielectric strength and Tan, and the stresses of the three factors jointly determine the aging stress of the transformer. The failure stress and aging stress of the transformer can be evaluated for the stress level of the transformer as a whole.

In step S100, the input data is normalized, i.e. transformed into a dimensionless variable between 0 and 1, by using values describing the normal and limit values of the transformer. Because all inputs are in the range of 0 to 1, normalization of the variables allows for uniformity of the created data and minimizes errors during evaluation.

Normalized variable is represented by y_iIndicate if y is_iA 1 means that the useful properties of the insulation have been lost and the device is operating in an area of high risk of failure. If yi is zero, it means that the stress level is minimal, and therefore the insulation conditions are optimal. For further division, the present invention divides five levels between 0 and 1: from 0 to 1, in order: the first level, the second level, the third level, the fourth level and the fifth level.

FIG. 2 shows a fuzzy model for transformer stress level calculation, with the output being used for transformer stress level calculation. In this model, each input has a corresponding output node, where the output of node 25 is the overall transformer stress level. All are formed by_iThe elements are all located in the region R [0,1]]And (4) the following steps.

After normalization, fuzzification processing needs to be performed on the normalized data, that is, different fuzzy membership functions are adopted according to different situations, specifically:

firstly, a fuzzy logic unit is created, and the specific operation comprises a plurality of steps of fuzzification, defuzzification and the like. From the measurement data, a fuzzy logic stress assessment can be developed.

Trapezoidal membership function:

the trapezoid membership function is determined by four parameters a, b, c and d, wherein the parameters a and b determine the angle of the trapezoid, the parameters c and d determine the shoulder of the trapezoid, and the expression is shown as formula (1):

wherein, x is a normalized variable, the parameter a is a rising starting point of the trapezoid membership function, and the trapezoid membership function value is increased from 0 to 1 from the point b; the parameter b is the rising end point of the trapezoidal membership function, and the trapezoidal membership function value reaches the maximum value 1 from the point b; the parameter c is the starting point of the decline of the trapezoidal membership function, and the trapezoidal membership function value declines from 1 to 0 from the point c; the parameter d is the end of the fall of the trapezoidal membership function, which remains 0 from point d.

Generalized bell membership function:

the generalized bell-shaped membership function is determined by three parameters a ', b ' and c ', and is expressed as formula (2):

where x is the normalized variable, a 'is the scaling ratio to the normalized variable x, and b' represents a factorB ' and a ' together establish the shape of the curve, c ' being the abscissa of the center point of the curve.

The Gaussian membership function is determined by two parameters c' and sigma, and the expression is shown as formula (3):

wherein, x is a normalized variable, c' determines the center of the Gaussian distribution curve, and sigma determines the dispersion degree of the Gaussian distribution curve.

Here, an expression is given for a gaussian membership function that transforms the normalized input into a fuzzy variable.

The following example illustrates the fuzzification process with gaussian membership functions:

y_ifuzzy set of (1) and fuzzy membership mu of five trapezoidal membership functions thereof_n(y_i) As shown in formulas (4) to (8):

in the evaluation tree model of fig. 1, a fuzzy logic rule is formed at each node such as nodes 1 to 25.

The logic rules of the neighboring nodes are described by the following logic expressions:

wherein A is_iAnd B_iInput representing a node of the hierarchy, C_iRepresenting the output of the node of the current level, also as part of the input of the node of the next level, f_iIs a numerical value mapping rule of two layers of nodes,andis a specific value, i represents a node label.

Numerical mapping rule f_iFor different nodes, the mapping rules are different, and the numerical mapping rules of the nodes are described layer by layer as follows:

for the bottom layer, the value mapping rule is to synthesize the amplitudes and the change rates of the factors of carbon monoxide, carbon dioxide, ethylene, acetylene, methane, hydrogen, acetylene, furan, acidity, moisture content, interfacial tension, dielectric strength and Tan, taking methane as an example, the change rate of the methane gas content is related to the existence and severity of the transformer fault, the amplitude of the methane gas content also reflects the insulation state of the transformer to a certain extent, but the importance of the amplitude is not as high as the change rate, and the mode of mapping the amplitude and the change rate of the methane gas to the methane gas factor is shown in fig. 3 (a).

Secondly, the numerical values of carbon monoxide and carbon dioxide determine the evaluation factor of the overheated paper, ethylene and ethane determine the evaluation factor of the overheated oil, methane and hydrogen determine the evaluation factor of partial discharge, hydrogen and acetylene determine the evaluation factor of the electric arc, and the overheated oil, the overheated paper, the partial discharge and the electric arc together determine the thermal failure stress and the electrical failure stress; furan and acidity determine the chemical aging stress, moisture content and interfacial tension determine the physical aging stress, and dielectric strength and Tan determine the chemical aging stress. These logic rules for evaluation may be referred to in the relevant literature and are not described in detail.

At the top level, the rule of mapping the values of the fault stress determined by the thermal fault stress and the electrical fault stress, and the transformer insulation stress determined by the chemical aging stress, the physical aging stress, and the electrical aging stress together is shown in fig. 3 (b).

For M rules, the sum L of the outputs of all rules_outThe expression of (a) is:

where M is the number of rules for the input trigger, L^jIndicating that the row vector inference engine joins all triggered rules in the current input.

And finally, after the output of each node of the transformer stress evaluation tree model is obtained, determining the clear output y (x) of the fuzzy logic system by using a central defuzzification method as follows:

where u is the number of quantization levels output, y^lIs the quantity output at the first quantization level, μ_B(y^l) Is the true membership value of the domain point.

The model has a total of 13 input variables, each of which is characterized by two parameters: amplitude and rate of change thereof. Thus, as can be seen from the fuzzy logic analog circuit diagram of fig. 2, there are 26 input ports.

The method for obtaining the synthesis weight in step S300 is as follows:

after normalizing the live data and mapping them to fuzzy membership functions, to combine them at any node. Firstly, they are weighted properly to reflect their strength in stress evaluation, that is, subjective weight can be obtained by analyzing the hierarchical processing method AHP, and the weight can be objectively calculated from the obtained matrix by applying eigenvector analysis. For the dissolved gases hydrogen, methane, ethylene, acetylene, carbon monoxide and carbon dioxide, a comprehensive weighting method is applied, i.e. a combination of subjective AHP weights in a certain proportion and objective weights formed based on the standards of these gases in a certain proportion is used.

Weight W using analytic hierarchy process_type1And weight W of eigenvector analysis_type2To obtain a composite weight W_finalComprises the following steps:

W_final＝e^θW_type1+(1-e^θ)W_type2(11)

wherein, theta represents weight type W_type1The specific value of the contribution rate can be flexibly determined according to historical data of the transformer.

And (4) calculating the resultant weight of the dissolved gas at each node by using the formula (11).

And adjusting the output result by adopting the synthesis weight:

the synthesis mode of the evidences of other nodes in the transformer stress evaluation tree model except the nodes where the fault stress and the aging stress are located is as follows:

and is provided with

Wherein, P is the node number where the variables are merged, and P is 1,2, 3. N is the number of input variables at each node, W_PThe node weighting matrix is used for carrying out weighted summation on fuzzy membership values of all levels so as to obtain variable values after node combination. Matrix Ψ_PElement mu of_1j，μ_2j...μ_5jRepresenting five levels of fuzzy membership values associated with the jth input variable.

Finally, the outputs of nodes 23 and 24 give evidence of supporting fault and aging stresses, resulting in a total stress level of the transformer at node 25.

Evidence Reasoning (ER) analysis is applied to the nodes 25, evidence reasoning being a tool for dealing with ambiguity and uncertainty problems. The basic confidence distribution of each node subset in the transformer stress evaluation tree model meets the following conditions:

where p (λ) ═ 0,1, denotes the basic confidence allocation for the subset λ of Ω, and p (Φ) denotes the basic confidence allocation for the empty subset, so p (Φ) ═ 0.

Two sources of evidence fault stress at node 25And aging stressThe multiple evidence combination rule of (2) is defined as:

wherein,given the degree of conflict of these sources of evidence,andnot zero, their intersection creates an empty set.

In the above method, the "evidence" refers to the output result, for example, the node evidence, i.e., the output result of the node.

The invention provides a transformer insulation stress evaluation method based on fuzzy logic and evidence reasoning, which comprises the following steps:

a respective basic confidence allocation (P (lambda))₁，(P(λ))₂，…，(P(λ))₅The optimum stress level is established according to the following conditions:

InsulDeg＝subscript{max{(P(λ))₁,(P(λ))₂,...,(P(λ))₅}} (16)

wherein InsulDeg is the insulation stress level of the transformer, (P (lambda))_rFor the basic confidence allocation of the r-th rank, r ∈ {1, 2,3, 4, 5}, the script is a subscript function, i.e., it satisfies the script { (P (λ))_rR and requires that the best rank must have a basic confidence allocation that exceeds the next best rank by some threshold, which is not less than 5%.

The evaluation criteria of the insulation stress of the transformer are as follows:

a first grade: the transformer has stable insulation condition and can normally run; a second stage: the insulation condition of the transformer has a deterioration trend, but the transformer can still normally operate; third level: the insulation state of the transformer is degraded, and monitoring in operation should be enhanced; fourth level: the deterioration degree of the insulation performance of the transformer is large, and the maintenance is reasonably arranged; and a fifth grade: the insulation performance of the transformer is seriously deteriorated, and the maintenance is required to be arranged as soon as possible.

Only three transformer case studies are provided herein, the transformers being named a, B and C.

The total stress level of transformer a using the measurement data is as follows: fuzzy logic with trapezoidal membership functions 28.74%, generalized bell membership functions 28.07%, gaussian membership functions 51.25% and evidence reasoning 13.53%. The total stress level of variable a is consistent with actual test data.

The second transformer B is a generator transformer. According to the relevant information of the company, the transformer is in a good state and has slight paper aging. Its stress level is based on Evidence Reasoning (ER) 11.56%, trapezoidal membership function (GBMF) 9.66% and Gaussian Membership Function (GMF) 26.53%.

The transformer C of the third example is a power station unit transformer, in good condition. Stress level based on ER 17.8%, fuzzy logic; TMF 53.76%, GBMF 53.77% and GMF 39.66%.

The technical features of the embodiments described above may be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the embodiments described above are not described, but should be considered as being within the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims (8)

1. The transformer insulation stress calculation method based on fuzzy logic and evidence reasoning is characterized by comprising the following steps of:

(1) normalizing field data, namely converting dissolved gases of hydrogen, methane, ethane, ethylene, acetylene, carbon monoxide, carbon dioxide and furan in the transformer oil, acidity, moisture content, interfacial tension, dielectric strength and Tan into dimensionless variables of a [0,1] area;

(2) constructing a transformer stress evaluation tree model by taking the data normalized in the step (1) as input, namely constructing input data from the transformerEach node to the insulation stress of the transformer and the connection mode thereof, the fuzzy correlation matrix R for the upper and lower nodes_m×nAssociating;

(3) fuzzification processing is carried out on the data normalized in the step (1), and finally a central defuzzification method is used for obtaining a total stress value of the transformer;

(4) determining weights among nodes of transformer stress evaluation tree model, namely weights W using analytical hierarchy processing method_type1And weight W of eigenvector analysis_type2To obtain a composite weight W_final；

(5) Adopting the synthesis weight W obtained in the step (4)_finalAnd (4) adjusting the total stress value of the transformer obtained in the step (3) to obtain the insulation stress of the transformer, namely determining the total stress level of the transformer by using a reasoning mechanism and an evidence synthesis rule.

2. The transformer insulation stress calculation method based on fuzzy logic and evidence reasoning is characterized in that if the normalized variable is 1, the device is indicated to operate in a region with high fault risk, if the normalized variable is 0, the insulation condition is indicated to be in an optimal state, and the normalized variable is between 0 and 1, and the steps are further divided into the following steps from low to high: the first level, the second level, the third level, the fourth level and the fifth level.

3. The transformer insulation stress calculation method based on fuzzy logic and evidence reasoning according to claim 1, characterized in that different fuzzy membership functions are adopted for fuzzification processing according to different situations, and the method specifically comprises the following steps:

trapezoidal membership function: the trapezoid membership function is determined by four parameters a, b, c and d, wherein the parameters a and b determine the angle of the trapezoid, the parameters c and d determine the shoulder of the trapezoid, and the expression is shown as formula (1):

$f\left(x;a,b,c,d\right)=\left\{\begin{array}{c}0,x\≤a\\ \frac{x-a}{b-a},a\≤x\≤b\\ 1,b\≤x\≤c\\ \frac{d-x}{d-c},c\≤x\≤d\\ 0,x\≥d\end{array}\right\}---\left(1\right)$

wherein, x is a normalized variable, the parameter a is a rising starting point of the trapezoid membership function, and the trapezoid membership function value is increased from 0 to 1 from the point b; the parameter b is the rising end point of the trapezoidal membership function, and the trapezoidal membership function value reaches the maximum value 1 from the point b; the parameter c is the starting point of the decline of the trapezoidal membership function, and the trapezoidal membership function value declines from 1 to 0 from the point c; the parameter d is the descending terminal point of the trapezoid membership function, and the trapezoid membership function value is kept to be 0 from the point d;

generalized bell membership function: the generalized bell-shaped membership function is determined by three parameters a ', b ' and c ', and is expressed as formula (2):

$f(x;a^{\′},b^{\′},c^{\′})=\frac{1}{1+|\frac{x-c^{\′}}{a^{\′}}{|}^{2b^{\′}}}---\left(2\right)$

where x is the normalized variable, a 'is the scaling ratio to the normalized variable x, and b' represents a factorB ' and a ' together define the shape of the curve, c ' is the abscissa of the center point of the curve;

the gaussian membership function is determined by two parameters c' and sigma, respectively defining the center and shape of the curve, and is expressed by the following formula (3):

$f(x;\mathrm{\σ},c^{\′\′})=e^{\frac{-{(x-c^{\′\′})}^2}{2{\mathrm{\σ}}^2}}---\left(3\right)$

wherein, x is a normalized variable, c' determines the center of the Gaussian distribution curve, and sigma determines the dispersion degree of the Gaussian distribution curve.

4. The transformer insulation stress evaluation method based on fuzzy logic and evidence reasoning according to claim 1 or 3, characterized in that the specific method for obtaining the total stress value of the transformer in the step (3) is as follows:

for M rules, the sum L of the outputs of all rules_outThe expression of (a) is:

$L_{out}={\∪}_{j=1}^M{L}^j---\left(9\right)$

where M is the number of rules for the input trigger, L^jIndicating that in the current input, the row vector inference engine unites all triggered rules;

after the output of each node of the transformer stress evaluation tree model is obtained, a central defuzzification method is used for determining the expression of clear output y (x) of a fuzzy logic system as follows:

$y\left(x\right)=\frac{\underset{i=1}{\overset{u}{\Σ}}{y}^l{\mathrm{\μ}}_B\left(y^l\right)}{\underset{i=1}{\overset{u}{\Σ}}{\mathrm{\μ}}_B\left(y^l\right)}---\left(10\right)$

where u is the number of quantization levels output, y^lIs the quantity output at the first quantization level, μ_B(y^l) Is the true membership value of the domain point.

5. The transformer insulation stress evaluation method based on fuzzy logic and evidence reasoning according to claim 4, wherein the fuzzy logic rule at each node is as follows:

$IF(A_i=\tilde{a})AND(B_i=\tilde{b})THEN(C_i=f_i(A_i,B_i\left)\right)$

wherein A is_iAnd B_iInput representing a node of the hierarchy, C_iRepresenting the output of the node of the current level, also as part of the input of the node of the next level, f_iIs a numerical value mapping rule of two layers of nodes,andis a specific value, i represents a node label.

6. The transformer insulation stress evaluation method based on fuzzy logic and evidence reasoning according to claim 1, characterized in that the weight W is synthesized_finalCalculated according to the following formula:

W_final＝e^θW_type1+(1-e^θ)W_type2(11)

wherein, theta represents weight type W_type1The specific value of the contribution rate is flexibly determined according to the historical data of the transformer.

7. The transformer insulation stress evaluation method based on fuzzy logic and evidence reasoning according to claim 1, characterized in that the specific method of step (5) is as follows:

the synthesis mode of the evidences of other nodes in the transformer stress evaluation tree model except the nodes where the fault stress and the aging stress are located is as follows:

$F_P=\underset{i=1}{\overset{N}{\Σ}}{W}_P{\mathrm{\Ψ}}_P^T---\left(12\right)$

${\mathrm{\Ψ}}_P=\left[\begin{array}{cccc}{\mathrm{\μ}}_{1j}& {\mathrm{\μ}}_{1j+1}& ...& {\mathrm{\μ}}_{1N}\\ {\mathrm{\μ}}_{2j}& {\mathrm{\μ}}_{2j+1}& ...& {\mathrm{\μ}}_{2N}\\ {\mathrm{\μ}}_{3j}& {\mathrm{\μ}}_{3j+1}& ...& {\mathrm{\μ}}_{3N}\\ {\mathrm{\μ}}_{4j}& {\mathrm{\μ}}_{4j+1}& ...& {\mathrm{\μ}}_{4N}\\ {\mathrm{\μ}}_{5j}& {\mathrm{\μ}}_{5j+1}& ...& {\mathrm{\μ}}_{5N}\end{array}\right]---\left(13\right)$

wherein, P is the node number where the variables are merged, and P is 1,2, 3. N is the number of input variables at each node, W_PThe node weighting matrix is used for carrying out weighted summation on fuzzy membership values of all levels so as to obtain variable values after node combination; matrix Ψ_PElement mu of_1j，μ_2j...μ_5jFuzzy membership values representing five levels associated with the jth input variable;

stress at failureAnd aging stressThe multiple evidence combination rule of (2) is defined as:

$p\left(\mathrm{\λ}\right)={\mathrm{\Σ}}_{{\tilde{y}}_1\∩{\tilde{y}}_2=\mathrm{\λ}}\frac{p_1\left({\tilde{y}}_1\right)p_2\left({\tilde{y}}_2\right)}{1-{\mathrm{\Σ}}_{{\tilde{y}}_1\∩{\tilde{y}}_2=\mathrm{\Φ}}{p}_1\left({\tilde{y}}_1\right)p_2\left({\tilde{y}}_2\right)}---\left(15\right)$

wherein,given the degree of conflict of these sources of evidence,andnot zero, their intersection creates an empty set.

8. A method for evaluating the insulation stress of a transformer based on fuzzy logic and evidence reasoning according to any of claims 1 to 7, characterized in that the respective basic confidence score (P (λ))₁，(P(λ))₂，…，(P(λ))₅The optimum stress level is established according to the following conditions:

InsulDeg＝subscript{max{(P(λ))₁,(P(λ))₂,...,(P(λ))₅}} (16)

wherein InsulDeg is the insulation stress level of the transformer, (P (lambda))_rFor the basic confidence allocation of the r-th rank, r ∈ {1, 2,3, 4, 5}, the script is a subscript function, i.e., it satisfies the script { (P (λ))_rR and requires that the best rank must have a basic confidence allocation that exceeds the next best rank by some threshold, which is not less than 5%.