A method for optimal operation life decision of power equipment based on technical economy

Abstract

The utility model provides a kind of based on technical economy's electric power equipment optimal operation life decision method, belong to renewable energy technology field, to solve the problem that existing calculation electric power equipment operation life is not combined with economy and technology, it includes: constructing the full life cycle cost estimation module of electric power equipment;Using the output result of full cycle cost estimation module combined with the uniform speed deterioration trend of operation and maintenance cost and the fault probability of overhaul cost, constructs the static economic life estimation model of electric power equipment;Using the output result of static economic life estimation model combined with the time value of fund, constructs the dynamic economic life estimation model of electric power equipment;Output dynamic economic life estimation result;Construct the technical life prediction model of electric power equipment based on HGWO-LS-SVM algorithm, output technical life prediction result;According to dynamic economic life estimation result and technical life prediction result, obtain the optimal operation life of electric power equipment.It is used to calculate electric power equipment operation life.

Description

TECHNICAL FIELD

[0001] The present application relates to a kind of optimal operation life decision-making method of power equipment based on technical economy, belong to renewable energy technical field. BACKGROUND

[0002] Wind power, photovoltaic and other renewable energy development has important role in reducing carbon emissions, but, wind power, photovoltaic and other renewable energy and electric vehicle and other flexible load have randomness, cause influence to the stable operation of entire power system, bring new challenges to the investment and management of power grid.Therefore, scientifically assesses the operation life of power equipment, reasonably determines the equipment replacement opportunity, is the solid foundation and security guarantee of building new type power system.

[0003] In the quality research of power equipment, equipment performance evolution law has been widely concerned.Because of the wear and tear of parts, equipment will continue to deteriorate in the process of operation, affect the normal production activities.The main factors causing distribution network equipment failure include: material aging, external damage, environmental impact, etc.

[0004] In the quality evaluation method of power equipment of prior art, there are:

[0005] A Gaussian cloud distribution network equipment power supply quality evaluation model is proposed in the prior art, and Technique for Order Preference by Similarity to an Ideal Solution is used for sorting.A classic bath curve is used to describe the equipment failure probability, and an electrical equipment reliability optimization method is designed, which can enhance the reliability of equipment.In addition, a power transformer deterioration and maintenance model is established, and a bath curve is also used to describe the equipment failure probability, and the wear tail of the curve is analyzed to study the accelerated deterioration process.However, the problem of this kind of method is that in the study of bath curve, the failure rate function is not continuous, not smooth, and the distribution shape parameter is not fixed.

[0006] Equipment economic life is related to the economic benefit and operation efficiency of power grid, and economic life is the time when the annual average cost of equipment is lowest.In calculating the economic life of equipment, full life cycle theory needs to be applied.The total life cycle cost and total ownership cost of transformer are calculated in the prior art, which improves the economy of equipment.An equipment economic life model is established, and the year when the total discount cost is lowest is found.However, these researches do not consider the influence of time value of money when calculating economic life, and ignore the influence of equipment deterioration on operation and maintenance cost.

[0007] Device quality evaluation and technical life prediction are of great significance for the safe and reliable operation of power grid. In the prior art, a comprehensive evaluation index system of distribution network equipment is designed from five aspects of equipment economy, performance, intelligence, coordination and standardization, the index weight is determined by subjective and objective weighting methods, and finally the equipment quality level can be evaluated. In addition, the health state of power grid equipment is evaluated by using fuzzy expert estimation and fuzzy set theory. In addition, support vector regression and grey wolf algorithm are integrated to predict the future operation state of distribution transformer. In addition, on the basis of analyzing the deterioration reason of equipment, an improved time-varying prediction function is proposed to evaluate the future fault state of equipment. In addition, the transformer failure probability and expected life are studied from the aspects of economy and technology. When predicting the quality of equipment, these methods do not consider the influence of external environment such as temperature and humidity.

[0008] Therefore, the existing research has the following problems when calculating the economic life:

[0009] 1. The prior art does not comprehensively consider the time value of funds, equipment deterioration speed and equipment failure rate, which may cause the calculated result to be inconsistent with the actual result.

[0010] 2. The prior art ignores the influence of external environment such as temperature and humidity on the quality of equipment when predicting the technical life.

[0011] 3. The selection of equipment replacement time in the prior art often does not consider both economy and safety. SUMMARY

[0012] The purpose of the present application is to solve the problem that the economic and technical properties are not combined when calculating the operation life of power equipment, and to provide a power equipment optimal operation life decision method based on technical economy.

[0013] The power equipment optimal operation life decision method based on technical economy comprises the following steps:

[0014] S1, a full life cycle cost calculation module of power equipment is constructed; the output result of the full cycle cost calculation module is combined with the uniform deterioration trend of operation and maintenance cost and the failure probability of maintenance cost to construct a static economic life calculation model of power equipment; the output result of the static economic life calculation model is combined with the time value of funds to construct a dynamic economic life calculation model of power equipment; and the dynamic economic life calculation result is outputted;

[0015] S2, a technical life prediction model of power equipment based on HGWO-LS-SVM algorithm is constructed, and the technical life prediction result is outputted;

[0016] S3, the optimal operation life of power equipment is obtained according to the dynamic economic life calculation result and the technical life prediction result.

[0017] Preferably, the specific method for constructing the technical life prediction model of power equipment based on the HGWO-LS-SVM algorithm and outputting the technical life prediction results as described in S2 includes:

[0018] S2-1. Conduct a quality assessment on the scrapped equipment, and use the kernel density estimation KDE method to fit the quality score with the equipment life to obtain the quality score corresponding to the critical scrapping point, and obtain the kernel function.

[0019] S2-2. Construct a technical life prediction model for power equipment based on least squares support vector machine (LS-SVM), and input the differential evolution algorithm (DE) into the gray wolf optimization algorithm (GWO) to obtain the improved gray wolf algorithm (HGWO).

[0020] S2-3. The improved Grey Wolf Algorithm (HGWO) is used to optimize the parameters of the penalty factor in LS-SVM and the kernel function obtained in S2-1 to obtain the quality scores of power equipment at different times.

[0021] S2-4. Compare the quality scores obtained in S2-3 for different periods with the quality scores corresponding to the critical scrapping point obtained in S2-1. When the quality scores for different periods are lower than the quality scores corresponding to the critical scrapping point, the corresponding time is the technical life prediction result.

[0022] Preferably, the specific method for constructing the full life cycle cost calculation module for power equipment as described in S1 includes:

[0023] Total life cycle cost C LCC for:

[0024]

[0025] Among them, C I C represents the initial investment cost. O,n C represents the operating and maintenance cost in year n. F,n L represents the failure cost in year n. R Let j represent the residual value of the equipment, where j = 1, 2, ..., n.

[0026] Preferably, the specific method for constructing the static economic life calculation model of power equipment as described in S1 includes:

[0027] The average one-time cost C in year n I-R,n yes:

[0028]

[0029] Assuming the equipment deteriorates at a uniform rate and the maintenance cost increases by a constant amount each year, then the average annual maintenance cost C in year n is... O,n yes:

[0030]

[0031] wherein C O,1 represents the initial value of the operation and maintenance cost, and a represents the increment of the operation and maintenance cost caused by the deterioration;

[0032] the average annual maintenance cost C F,n in the nth year is:

[0033]

[0034] wherein, represents the average annual periodic maintenance cost in the nth year, represents the average annual failure maintenance cost in the nth year;

[0035]

[0036] wherein, represents the initial value of the periodic maintenance cost, and β represents the increment of the periodic maintenance cost caused by the deterioration;

[0037]

[0038] wherein, represents the cost of each failure maintenance, and h(n) represents the failure rate in the "bathtub curve";

[0039] The static economic life measurement model is:

[0040] C a (n) = C I-R,n + C O,n + C F,n ;

[0041] N0 = [min C a (n)] -1 ;

[0042] wherein N0 represents the static economic life, and C a (n) represents the static average annual cost in the nth year.

[0043] Preferably, the specific method for constructing the dynamic economic life measurement model of the power equipment according to S1 comprises:

[0044] The dynamic economic life N0' is:

[0045] N0' = [min C a '(n)] -1 ;

[0046] wherein C a '(n) represents the dynamic average annual cost in the nth year;

[0047] C a′(n)=C I ′ -R,n +C′ O,n +C′ F,n ;

[0048] Among them, C I ′ -R,n Let C′ represent the dynamic investment cost in year n. O,n Let C′ represent the dynamic operating and maintenance cost in year n. F,n This represents the dynamic cost of failure in year n.

[0049] C I ′ -R,n =C I ·(A / P,i,n)-L R ·(A / F,i,n);

[0050] Where (A / P,i,n) represents the capital recovery factor for equal share payments, (A / F,i,n) represents the sinking fund factor for equal share payments, and i represents the basic internal rate of return;

[0051]

[0052] Among them, C O,1 This represents the operating and maintenance cost in the first year, and (P / F,i,n) represents the present value factor for a single payment.

[0053] in:

[0054]

[0055] Preferably, the specific method for fitting the quality score to the equipment life using the kernel density estimation KDE method described in S2-1 to obtain the quality score corresponding to the critical scrapping point and to obtain the kernel function includes:

[0056] Kernel density estimation (KDE) is used to fit the quality score to the equipment lifespan to obtain the quality score corresponding to the critical scrap point. The form of kernel density estimation is as follows:

[0057]

[0058] in, This indicates the KDE fit with respect to The probability density function, Let N represent a random variable used to represent the quality score, h represent the sample size, and K represent the bandwidth. h Represents the kernel function. R represents the sample observations. d Let represent a d-dimensional set of real numbers, and k represent the k-th sample;

[0059] The kernel function adopts a Gaussian kernel function, a probability density function is expressed as:

[0060]

[0061] Preferably, the specific method of S2-2 for constructing a technical life prediction model of power equipment based on a least squares support vector machine (LS-SVM) comprises:

[0062] An LS-SVM model extended from a support vector machine (SVM) is proposed:

[0063] First, samples are mapped from an original space to a feature space:

[0064]

[0065] y represents an output vector, x represents an un-input vector, represents a mapping function, ω and b both represent parameters to be determined;

[0066] The determination process of ω and b is equivalent to the minimization of an intermediate variable R:

[0067]

[0068] wherein, ||ω||2 2 represents the complexity of a control model, c represents an adjustment parameter, and controls the punishment degree of an error exceeding, and R emp represents an error control function;

[0069] The minimization problem is further expressed as:

[0070]

[0071] J(ω,ξ) represents a loss function, l represents the total number of samples, j(x p ) represents a mapping function of the pth input vector, w represents a complexity parameter, x p represents the pth input vector, ξ p represents a relaxation factor;

[0072] A Lagrange function L(·) is constructed:

[0073]

[0074] wherein, α p represents a Lagrange multiplier;

[0075] According to a KKT optimization condition, the following can be obtained:

[0076]

[0077] After ω and ξ are eliminated, the following can be obtained:

[0078]

[0079] where I = [1,..., 1], with kernel function instead of x m represents the mth vector, x n represents the nth vector.

[0080] The regression function f(x) is:

[0081]

[0082] where K(x, x p ) represents the kernel function of x and x p .

[0083] The radial basis function RBF is represented as:

[0084]

[0085] where x h represents the hth vector, x g represents the gth vector; and σ is the kernel width.

[0086] Preferably, S2-2 said differential evolution algorithm DE into the grey wolf optimization algorithm GWO, the specific method for obtaining improved grey wolf algorithm HGWO includes:

[0087] First, DE is used to realize population mutation, and then the optimal three individual positions are obtained according to the fitness, and the distance between the remaining grey wolves and the optimal grey wolf is calculated:

[0088] D a = |C1X a (t) - X(t)|

[0089] D b = |C2X b (t) - X(t)|

[0090] D c = |C3X c (t) - X(t)|

[0091] where t represents the iteration number, D a , D b and D c represent the distance respectively; C1, C2 and C3 represent random vectors, X a , X b and X c represent the three optimal positions closest to the prey, and X represents the current solution position.

[0092] The rest of the wolves are repositioned according to the three optimal positions above:

[0093] X1(t+1) = X a (t) - A1D a

[0094] X2(t+1) = X b (t) - A2D b

[0095] X3(t+1) = X c (t) - A3D c

[0096] A1, A2 and A3 all represent convergence factors;

[0097] The final positions of the rest of the individuals are:

[0098]

[0099] C = 2r1;

[0100] A = 2ar2-a;

[0101] Wherein, C represents a disturbance parameter, A represents a convergence factor, r1 and r2 represent random numbers in [0, 1], and a represents a variable related to the number of iterations, which decreases from 2 to 0;

[0102] The better individuals are retained, and A and C are updated until the condition is met.

[0103] Preferably, the specific method for obtaining the quality score of the power equipment at different periods as described in S2-3 and the specific method for determining the technical life prediction result when the quality score at different periods is lower than the quality score corresponding to the critical scrap point as described in S2-4 include:

[0104] Mapping the quality-related data to the hidden layer space through LS-SVM kernel mapping;

[0105] Obtaining the optimal values of the adjustment parameters c and the kernel width σ by applying the HGWO algorithm;

[0106] Obtaining the predicted value of the quality score of the equipment at different periods after inputting the LS-SVM;

[0107] When the quality score at different periods is lower than the quality score corresponding to the critical scrap point, the corresponding time is the technical life prediction result:

[0108] M0 = (if S(n) < S) -1 ;

[0109] Wherein, M0 represents the technical life prediction result, S(n) represents the predicted value of the equipment quality score in the nth year, and S represents the quality score corresponding to the critical scrap point.

[0110] Preferably, the specific method for obtaining the optimal operation life of the power equipment according to the dynamic economic life calculation result and the technical life prediction result comprises the following steps of:

[0111] The optimal operation life R of the power equipment is:

[0112] R = min (N0', M0) ;

[0113] Wherein, N0' represents the dynamic economic life, and M0 represents the technical life prediction result.

[0114] Advantages of the present application: the present application combines the economy and technology of power equipment operation, and proposes a power equipment optimal operation life decision model. The dynamic economic life calculation model is constructed by combining the uniform speed degradation trend of operation and maintenance cost and the failure trend of maintenance cost, and taking into account the time value of money. The technical life prediction model is constructed based on the least square support vector machine (LS-SVM) by combining internal factors and external factors, and the parameters are determined by using the improved grey wolf optimization algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0115] Figure 1 is a flow chart of the power equipment optimal operation life decision method based on technical economy according to the present application;

[0116] Figure 2 is a flow chart of the dynamic economic life calculation model of the power equipment according to the present application;

[0117] Figure 3 is a curve diagram of the static economic life;

[0118] Figure 4 is a device quality score probability density function;

[0119] Figure 5 is a flow chart of the technical life prediction model of the power equipment based on the HGWO-LS-SVM algorithm according to the present application;

[0120] Figure 6 is a device quality score diagram from 2010 to 2020;

[0121] Figure 7 is a reliability curve diagram;

[0122] Figure 8 is a failure rate curve diagram;

[0123] Figure 9 is a device reliability curve diagram from 2010 to 2020;

[0124] Figure 10 is a curve graph of the cumulative failure rate of the equipment in each quarter from 2010 to 2020;

[0125] Figure 11 is a curve graph of the daily maximum temperature in each quarter from 2010 to 2020;

[0126] Figure 12 is a curve graph of the precipitation in each quarter from 2010 to 2020. DETAILED DESCRIPTION

[0127] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0128] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0129] The present application will be further described below with reference to the drawings and specific embodiments, but is not limited by the embodiments.

[0130] Embodiment 1:

[0131] The present embodiment will be described below. Figures 1-5 The present embodiment describes a power equipment optimal operation life decision-making method based on technical economy, which comprises:

[0132] S1, constructing a full life cycle cost estimation module of the power equipment; using the output result of the full cycle cost estimation module, combining the uniform deterioration trend of the operation and maintenance cost and the failure probability of the maintenance cost, constructing a static economic life estimation model of the power equipment; using the output result of the static economic life estimation model, combining the time value of the fund, constructing a dynamic economic life estimation model of the power equipment; outputting the dynamic economic life estimation result;

[0133] S2, constructing a technical life prediction model of the power equipment based on the HGWO-LS-SVM algorithm, outputting the technical life prediction result;

[0134] S3, obtaining the optimal operation life of the power equipment according to the dynamic economic life estimation result and the technical life prediction result.

[0135] Further, the specific method of S2 constructing the technical life prediction model of the power equipment based on the HGWO-LS-SVM algorithm and outputting the technical life prediction result comprises:

[0136] S2-1. Conduct a quality assessment on the scrapped equipment, and use the kernel density estimation KDE method to fit the quality score with the equipment life to obtain the quality score corresponding to the critical scrapping point, and obtain the kernel function.

[0137] S2-2. Construct a technical life prediction model for power equipment based on least squares support vector machine (LS-SVM), and input the differential evolution algorithm (DE) into the gray wolf optimization algorithm (GWO) to obtain the improved gray wolf algorithm (HGWO).

[0138] S2-3. The improved Grey Wolf Algorithm (HGWO) is used to optimize the parameters of the penalty factor in LS-SVM and the kernel function obtained in S2-1 to obtain the quality scores of power equipment at different times.

[0139] S2-4. Compare the quality scores obtained in S2-3 for different periods with the quality scores corresponding to the critical scrapping point obtained in S2-1. When the quality scores for different periods are lower than the quality scores corresponding to the critical scrapping point, the corresponding time is the technical life prediction result.

[0140] Furthermore, the specific method for constructing the life-cycle cost calculation module for power equipment as described in S1 includes:

[0141] Total life cycle cost C LCC for:

[0142]

[0143] Among them, C I C represents the initial investment cost. O,n C represents the operating and maintenance cost in year n. F,n L represents the failure cost in year n. R Let j represent the residual value of the equipment, where j = 1, 2, ..., n.

[0144] Furthermore, the specific method for constructing the static economic life calculation model for power equipment as described in S1 includes:

[0145] The average one-time cost C in year n I-R,n yes:

[0146]

[0147] Assuming the equipment deteriorates at a uniform rate and the maintenance cost increases by a constant amount each year, then the average annual maintenance cost C in year n is... O,n yes:

[0148]

[0149] Among them, C O,1 α represents the initial value of operation and maintenance costs, and α represents the increase in operation and maintenance costs caused by degradation.

[0150] the average annual maintenance cost C of the nth year F,n is:

[0151]

[0152] wherein, represents the average annual periodic maintenance cost of the nth year, represents the average annual breakdown maintenance cost of the nth year;

[0153]

[0154] wherein, represents the initial value of the periodic maintenance cost, and β represents the increment of the periodic maintenance cost caused by the deterioration;

[0155]

[0156] wherein, represents the breakdown maintenance cost each time, and h(n) represents the failure rate in the "bathtub curve";

[0157] The static economic life measurement model is:

[0158] C a (n) = C I-R,n + C O,n + C F,n ;

[0159] N0 = [minC a (n)] -1 ;

[0160] wherein, N0 represents the static economic life, and C a (n) represents the static average annual cost of the nth year.

[0161] Further, the specific method for constructing the dynamic economic life measurement model of the power equipment by S1 comprises:

[0162] The dynamic economic life N0' is:

[0163] N0' = [minC a '(n)] -1 ;

[0164] wherein, C a '(n) represents the dynamic average annual cost of the nth year;

[0165] C a '(n) = C I '(n-1) + C -R,n '(n-1) + C O,n '(n-1) + C F,n ;

[0166] Among them, C I ′ -R,n Let C′ represent the dynamic investment cost in year n. O,n Let C′ represent the dynamic operating and maintenance cost in year n. F,n This represents the dynamic cost of failure in year n.

[0167] C I ′ -R,n =C I ·(A / P,i,n)-L R ·(A / F,i,n);

[0168] Where (A / P,i,n) represents the capital recovery factor for equal share payments, (A / F,i,n) represents the sinking fund factor for equal share payments, and i represents the basic internal rate of return;

[0169]

[0170] Among them, C O,1 This represents the operating and maintenance cost in the first year, and (P / F,i,n) represents the present value factor for a single payment.

[0171] in:

[0172]

[0173] Furthermore, the specific method described in S2-1 for using the kernel density estimation KDE method to fit the quality score to the equipment lifespan, obtain the quality score corresponding to the critical scrapping point, and obtain the kernel function includes:

[0174] Kernel density estimation (KDE) is used to fit the quality score to the equipment lifespan to obtain the quality score corresponding to the critical scrap point. The form of kernel density estimation is as follows:

[0175]

[0176] in, This indicates the KDE fit with respect to The probability density function, Let N represent a random variable used to represent the quality score, h represent the sample size, and K represent the bandwidth. h Represents the kernel function. R represents the sample observations. d Let represent a d-dimensional set of real numbers, and k represent the k-th sample;

[0177] The kernel function used is a Gaussian kernel function, and the probability density function is... Represented as:

[0178]

[0179] Further, the specific method for constructing the technical life prediction model of the power equipment based on the least squares support vector machine LS-SVM in S2-2 comprises the following steps:

[0180] An LS-SVM model extended from the support vector machine SVM is proposed:

[0181] First, the sample is mapped from the original space to the feature space:

[0182]

[0183] y represents an output vector, x represents an un-input vector, represents a mapping function, ω and b both represent to-be-determined parameters;

[0184] The determination process of ω and b is equivalent to the minimization of an intermediate variable R:

[0185]

[0186] wherein, ||ω||2 2 represents the complexity of the control model, c represents an adjustment parameter, and the penalty degree of error control, R emp represents an error control function;

[0187] The minimization problem is further represented as:

[0188]

[0189] J(ω,ξ) represents a loss function, l represents the total number of samples, j(x p ) represents a mapping function of the pth input vector, w represents a complexity parameter, x p represents the pth input vector, and ξ p represents a relaxation factor;

[0190] A Lagrange function L(·) is constructed:

[0191]

[0192] wherein, α p represents a Lagrange multiplier;

[0193] According to the KKT optimization condition, the following can be obtained:

[0194]

[0195] After ω and ξ are eliminated, the following can be obtained:

[0196]

[0197] wherein, I=[1,…,1], is replaced by a kernel function; xm represents the mth vector, x n represents the nth vector;

[0198] The regression function f(x) is:

[0199]

[0200] where K(x, x p ) represents the kernel function of x and x p ;

[0201] The radial basis function RBF is represented as:

[0202]

[0203] where x h represents the hth vector, x g represents the gth vector; and σ is the kernel width.

[0204] Further, the method for introducing the differential evolution algorithm DE into the grey wolf optimization algorithm GWO to obtain the improved grey wolf optimization algorithm HGWO includes:

[0205] First, DE is used to realize population mutation, and then the optimal three individual positions are obtained according to the fitness, and the distances of the remaining grey wolves and the optimal grey wolf are calculated:

[0206] D a = |C1X a (t)-X(t)|

[0207] D b = |C2X b (t)-X(t)|

[0208] D c = |C3X c (t)-X(t)|

[0209] where t represents the iteration number, D a , D b and D c represent the distances respectively; C1, C2 and C3 represent random vectors, X a , X b and X c represent the three optimal positions closest to the prey, and X represents the current solution position;

[0210] The remaining grey wolves are repositioned according to the above three optimal positions:

[0211] X1(t+1) = X a (t)-A1D a

[0212] X2(t+1) = X b (t) - A2D b

[0213] X3(t+1) = X c (t) - A3D c

[0214] A1, A2 and A3 all represent convergence factors;

[0215] The final position of the remaining individual is:

[0216]

[0217] C = 2r1;

[0218] A = 2ar2-a;

[0219] Wherein, C represents a disturbance parameter, A represents a convergence factor, r1 and r2 represent random numbers of [0, 1], and a represents a variable related to the number of iterations, which decreases from 2 to 0;

[0220] The better individual is reserved, A and C are updated, and the condition is met.

[0221] Further, the method for obtaining the quality score of the power equipment in different periods according to S2-3 and the method for obtaining the technical life prediction result when the quality score in different periods is lower than the quality score corresponding to the critical scrap point according to S2-4 include:

[0222] Map the quality-related data to the hidden layer space through LS-SVM kernel mapping;

[0223] Obtain the optimal values of the adjustment parameters c and the kernel width σ by applying the HGWO algorithm;

[0224] After inputting the LS-SVM, the predicted value of the quality score of the equipment in different periods is obtained;

[0225] When the quality score in different periods is lower than the quality score corresponding to the critical scrap point, the corresponding time is the technical life prediction result:

[0226] M0 = (if S(n) < S) -1 ;

[0227] Wherein, M0 represents the technical life prediction result, S(n) represents the predicted value of the quality score of the equipment in the nth year, and S represents the quality score corresponding to the critical scrap point.

[0228] Further, the method for obtaining the optimal operation life of the power equipment according to the dynamic economic life calculation result and the technical life prediction result according to S3 includes:

[0229] The optimal operation life R of the power equipment is:

[0230] R = min(N0', M0);

[0231] Wherein, N0' represents a dynamic economic life, and M0 represents a technical life prediction result.

[0232] In the application, the power equipment has characteristics of many points and wide surfaces, multi-source heterogeneity and the like, and the application mainly researches the distribution network equipment. When economic life calculation and technical life prediction are performed, three key problems need to be solved: (1) acquisition of annual cost, state, performance and the like of the equipment; (2) determination of performance evolution law of the equipment over time; and (3) evaluation of annual quality of the equipment. The specific solving approaches of the above three key problems are as follows:

[0233] The data fields contained in the multi-service system are as shown in Table 1:

[0234] Table 1

[0235] Business system Data field Production management system Fields such as account, retirement, defect, failure, maintenance, operation, etc. Enterprise resource planning system Fields such as account, asset, cost, material, material, etc. Safety monitoring system Fields such as reliable operation event and statistics, uninterrupted power supply event and statistics, etc. Lightning positioning system Fields such as lightning information, lightning trip, etc. Environmental meteorological system Fields such as temperature, humidity, rainfall, altitude, etc.

[0236] According to the data fields contained in the multi-service system, relevant information for economic life calculation and technical life prediction is extracted.

[0237] There are three kinds of evolution laws of the power equipment deterioration over time, which are acceleration, uniform speed and deceleration. When internal reasons or external reasons change, the equipment quality state will rise. The internal reasons mainly refer to maintenance on the equipment, such as replacement of severely worn parts. The external reasons mainly refer to the influence of the environment on the equipment quality, such as passing through the high-temperature and high-humidity season in the coastal area.

[0238] There are three periods of the power equipment failure rate evolution law over time, which are a running-in period, a stable period and a wear-out period, as shown in a "bathtub curve". Figure 3 After the equipment is put into operation, the running-in period needs to go through multiple debugging, and small failures often occur; when the debugging is completed, the failure rate is low to enter the stable period; with the overall aging of the equipment, the failure rate is high to enter the wear-out period. In addition to the "bathtub curve", a reliability function can also reflect the equipment failure rate.

[0239] When the reliability function and the "bathtub curve" are fitted, it is necessary to ensure that the curves are greater than zero, continuous and first decreasing and then increasing. Existing researches show that at least 3 variables need to be introduced. There are dozens of fitting results at present, after comparing the accuracy of various fitting models, the applied reliability function and "bathtub curve" are as follows:

[0240] Reliability function:

[0241]

[0242] Failure rate function:

[0243]

[0244] wherein R(t) is a function of reliability with respect to life; h(t) is a function of failure rate with respect to life; a and b are shape parameters, a e (0, 1), b > 0; l is a failure rate function adjustment factor; T is the design life of the equipment; t is the service year of the equipment, t e (0, T);

[0245] The quality evaluation process of power equipment is as follows:

[0246] Following the principles of comprehensiveness, systematicness, independence, quantifiability, etc., a power equipment quality evaluation index system is constructed, including the operation condition, availability factor, failure frequency, downtime, defect level, defect elimination time, etc. of the equipment;

[0247] The weights of each index are determined by means of analytic hierarchy process, entropy weight method, etc. In order to improve objectivity and scientificalness, a combined weighting method can be used.

[0248] A comprehensive evaluation model is constructed, commonly used are fuzzy comprehensive evaluation, TOPSIS, matter-element extension, etc.

[0249] The quality evaluation results of the equipment in each period can be obtained by extracting the data of each index of the equipment in different periods from the multi-service system and performing quality evaluation according to the above process, and can be converted into quality scores.

[0250] Dynamic economic life measurement model based on time value of money:

[0251] A power equipment life cycle cost measurement model is constructed, considering the uniform degradation trend of operation and maintenance cost and the failure trend of repair cost, a static economic life measurement model of the equipment is proposed, and finally considering the time value of money, a dynamic economic life measurement model of the equipment is proposed, and the calculation process is as shown in Figure 2

[0252] Equipment life cycle cost model:

[0253] The life cycle cost of power equipment includes initial investment cost, operation and maintenance cost, repair cost, and residual value, which can be expressed as:

[0254]

[0255] In the formula, C LCC represents the life cycle cost of power equipment; C I represents the initial investment cost; C O,t represents the operation and maintenance cost in the tth year; C F,t represents the failure cost in the tth year; L R represents the equipment residual value. ​

[0256] Static economic life calculation model considering degradation failure:

[0257] In the total life cycle cost, maintenance costs increase year by year due to equipment deterioration, and repair costs also increase year by year due to increased equipment failure rates. The calculation methods for each annual average cost are as follows:

[0258] Annual one-time cost: Annual one-time cost refers to the annual amortized cost after deducting the residual value from the investment cost.

[0259]

[0260] In the formula, C I-R,n This represents the average annual one-time cost in year n.

[0261] Average annual operation and maintenance cost: Assuming the equipment deteriorates at a uniform rate, the operation and maintenance cost will increase by a certain amount each year.

[0262]

[0263] In the formula, C O,n C represents the average annual maintenance cost in year n. O,1 α represents the initial value of operation and maintenance costs; α represents the increase in operation and maintenance costs caused by degradation.

[0264] Average annual maintenance cost: Maintenance costs include scheduled maintenance costs and fault repair costs. Scheduled maintenance is carried out according to the maintenance guidelines for various types of equipment. Considering the uniform deterioration of equipment, the cost of scheduled maintenance increases by a certain amount each year. Fault repair is carried out based on the fault situation, and the annual cost is related to the probability of failure.

[0265]

[0266] In the formula, C F,n This represents the average annual maintenance cost in year n. Let this be the average annual cost of scheduled maintenance in year n. Let $\frac{ ...

[0267]

[0268] In the formula, β represents the initial value of the periodic maintenance cost; β represents the increase in periodic maintenance cost caused by deterioration.

[0269]

[0270] In the formula, is the cost of repairing each fault; h is the failure rate in the "bathtub curve".

[0271] Static economic life: the average annual cost of power equipment in the whole life cycle decreases first and then increases, showing a "U" type curve. The time (N0) when the average annual cost is the lowest is the economic life, as shown in Figure 3

[0272] The calculation method of static economic life is:

[0273] C a (n) = C I-R,n + C O,n + C F,n

[0274] N0 = [min C a (n)] -1

[0275] In the formula, C a (n) is the static average annual cost in the nth year; N0 is the static economic life.

[0276] The dynamic economic life calculation model considering the time value of money is:

[0277] The running life of power equipment is generally long, and the time value of money has a greater impact on economic life. The dynamic economic life calculation model is:

[0278] C′ a (n) = C′ I-R,n + C′ O,n + C′ F,n

[0279] N′0 = [min C′ a (n)] -1

[0280] In the formula, C′ a (n) is the dynamic average annual cost in the nth year; N′0 is the dynamic economic life; the variables with a prime on the upper right corner represent dynamic cost.

[0281] After considering the time value of money, the cost calculation method is revised as:

[0282] C′ I-R,n = C I ·(A / P,i,n)-L R ·(A / F,i,n)

[0283]

[0284] The corresponding equivalent formula of funds is:

[0285]

[0286] ​In the formula, i represents the basic internal rate of return; (A / P, i, n) represents the equal installment capital recovery coefficient; (A / F, i, n) represents the equal installment sinking fund coefficient; (P / F, i, n) represents the present value coefficient of one-time payment.

[0287] HGWO-LS-SVM-based technical life prediction model: Considering internal factors (reliability, failure rate) and external factors (temperature, humidity), a power equipment technical life prediction model is constructed based on the least squares support vector machine (LS-SVM). The differential evolution algorithm (DE) is introduced into the grey wolf optimization algorithm (GWO), and the improved grey wolf algorithm (HGWO) is applied to optimize the penalty factor and kernel function of LS-SVM, and the quality score of the equipment at different periods is predicted. When the quality score is lower than the corresponding time of the given standard, the technical life of the equipment is obtained.

[0288] Kernel density estimation: In order to study the relationship between the quality of power equipment and the life, the quality of the scrapped equipment is evaluated, and the kernel density estimation (KDE) method is used to fit the quality score and the life of the equipment, and the quality score corresponding to the critical scrap point is found. Kernel density estimation is used to estimate unknown density function, which belongs to one of the non-parametric test methods. Compared with the participation estimation, the result is closer to the true distribution, and the general form is:

[0289]

[0290] In the formula, is a random variable, which represents the quality score; is the probability density function of the KDE fitting about N is the sample size; h is the bandwidth; K h is the kernel function, which has several types such as Gaussian, Triangle, Epanechnikov, etc.

[0291] The present application adopts Gaussian kernel function, which can be represented as:

[0292]

[0293] The quality score of the transformer equipment reaching the technical life is mainly distributed on both sides of 60 points, as shown in Figure 4 .

[0294] LS-SVM model: LS-SVM is an extension of support vector machine (SVM), which can effectively improve the solving speed and convergence accuracy. First, the sample is mapped from the original space to the feature space

[0295]

[0296] where ω, b are parameters to be determined; is a mapping function; x is an input vector; y is an output vector.

[0297] The determination of ω, b can be equivalently transformed into the minimization of the following problem, i.e.,

[0298]

[0299] where ||ω|| is a norm of ω. 2 controls the complexity of the model; c is a tuning parameter that controls the degree of penalty for exceeding the error; R emp is an error control function.

[0300] The minimization optimization problem can be further expressed as:

[0301]

[0302] where ξ i is a relaxation factor.

[0303] A Lagrange function is constructed as:

[0304]

[0305] where α i is a Lagrange multiplier.

[0306] According to the KKT optimization condition, we have:

[0307]

[0308] Eliminating ω and ξ, we have:

[0309]

[0310] where I = [1, …, 1]; which can be replaced by a kernel function K(x, x i ).

[0311] Finally, the regression function is:

[0312]

[0313] The key problem is to determine a suitable kernel function. The radial basis function (RBF) kernel function has strong adaptability and good performance. The RBF is expressed as follows:

[0314]

[0315] where σ is the kernel width.

[0316] Parameter optimization based on HGWO:

[0317] GWO has the advantages of few parameters, fast speed, high precision, etc., but it is easy to fall into local optimum. DE has the advantages of strong robustness and simple structure, which can improve the shortcomings of GWO. The HGWO process is as follows: first, use DE to realize population mutation, then get the optimal three individual positions according to the fitness, and calculate the distance between the remaining gray wolves and the optimal gray wolf.

[0318] D a = |C1X a (t) - X(t)|

[0319] D b = |C2X b (t) - X(t)|

[0320] D c = |C3X c (t) - X(t)|

[0321] In the formula, t represents the number of iterations; D a , D b and D c represent the distance; C1, C2 and C3 represent random vectors; X a , X b , X c are the positions of the three optimal positions closest to the prey; X represents the current solution position;

[0322] The remaining gray wolves reposition according to the above three optimal positions:

[0323] X1(t+1) = X a (t) - A1D a

[0324] X2(t+1) = X b (t) - A2D b

[0325] X3(t+1) = X c (t) - A3D c

[0326] In the formula, A is the convergence factor. The final position of the remaining individual is:

[0327]

[0328] C = 2r1

[0329] A = 2ar2 - a

[0330] In the formula: C is the disturbance parameter; r1 and r2 are random numbers in [0, 1]; a is related to the number of iterations, decreasing from 2 to 0.

[0331] Then the better individual is reserved, A and C are updated, whether the condition is met is judged, if met, the result is output, if not met, recalculate.

[0332] The prediction model precision test:

[0333] The residual sum of squares e SSE , the mean absolute percentage error e MAPE , the mean square error e MSE are selected to measure the prediction precision, and the expressions of various indexes are as follows:

[0334]

[0335] In the formula, Y i and respectively represent the i-th actual value and predicted value.

[0336] HGWO-LS-SVM prediction process:

[0337] The power equipment technical life prediction model based on the HGWO-LS-SVM algorithm is proposed, first, the quality related data is mapped to the hidden layer space through LS-SVM kernel, then the optimal value of c and sigma is obtained by applying the HGWO algorithm, the prediction value of the equipment quality score in different periods is obtained after inputting LS-SVM, and the time when the quality score is first lower than the given standard is found, that is, the equipment predicted life.

[0338] M0=(if S(n)<S -1

[0339] In the formula, M0 is the predicted technical life of the equipment, S(n) is the predicted quality score in the n-th year, and S is the given equipment operation quality standard. The process is shown in Figure 5 .

[0340] The optimal operation life decision process under the joint of the two lives:

[0341] According to the dynamic economic life calculation result and the technical life prediction result, the optimal operation life of the power equipment is determined, and the specific process is as Figure 1 .

[0342] R=min(N0',M0)

[0343] In the formula, R is the optimal operation life.

[0344] In the application, a 10kV oil-immersed distribution transformer of a certain power company is taken as an example, and the cost and quality related data information is extracted from the ERP and PMS systems. The distribution transformer asset information is shown in Table 2.

[0345] Table 2

[0346] Device name 94840 distribution transformer Device code 15M00000013266481 Unit details State Grid Some City Power Supply Company PMS device type code 302 Asset sub-class code 160200001 Rated capacity 500kVA Voltage level AC 10kV Operation status In operation Health status Normal state Device model SCB9-D-500 / 10 Date of manufacture 2010 / 4 / 1 Date of commissioning 2010 / 6 / 1

[0347] 94840 distribution transformer cost data as shown in Table 3.

[0348] Table 3

[0349] Device name 94840 distribution transformer Upper limit of life 18 years Service life until 2020 10 years Asset residual value rate 5％ Original asset value 57593.7 yuan Net asset residual value 2879.69 yuan Initial operation and maintenance cost 1439.84 yuan Operation and maintenance cost growth value 400 yuan Initial regular maintenance 863.91 yuan Regular maintenance cost growth value 300 yuan Average cost of each failure maintenance 30000 yuan

[0350] Due to the large difference between temperature and precipitation in different seasons in the province, the quality of the equipment will be affected to a certain extent, therefore, the time granularity of all quality-related data is quarterly, a total of 5 groups and 44 dimensions. In order to simplify the analysis, each quarter of each year is represented in a numerical sequence format, as shown in Table 4.

[0351] Table 4

[0352] Date 2010 / 3 / 1 2010 / 6 / 1 2010 / 9 / 1 2010 / 12 / 1 2014 / 3 / 1 … 2020 / 9 / 1 2020 / 12 / 1 Sequence 1 2 3 4 5 … 43 44

[0353] First, by using the quality comprehensive evaluation method, the quality score of 94840 distribution transformer in each quarter can be obtained, such as Figure 6 .

[0354] Then, by using the fitting method, according to the historical failure law of distribution network equipment, the parameter values are: α = 0.4, β = 0.1, λ = 0.01, η = 1. The reliability curve and failure rate curve are drawn as Figure 7 and Figure 8 .

[0355] The first 44 dimensions of Figure 7 and Figure 8 are intercepted to obtain the reliability and cumulative failure rate of 94840 distribution transformer in each quarter from 2010 to 2020, as shown in Figure 9 and Figure 10 .

[0356] Finally, according to the environmental meteorological system, the daily average maximum temperature and precipitation in this area are obtained, as shown in Figure 11 and Figure 12 .

[0357] Among the above 2010-2020 quarterly data, 1:36 dimensions are sample sets, which are used to train the model; 37:44 dimensions are test sets, which are used to verify the accuracy of the model prediction; 44:79 dimensions are to be predicted sets, which are used to judge the technical life of the equipment.

[0358] Static economic life measurement:

[0359] According to the asset information and cost data of distribution transformer, the static economic life is calculated, and the calculation results are as follows:

[0360] Table 5

[0361]

[0362] It can be seen that the static economic life of distribution transformer is 12 years, which is lower than the expected service life of the equipment. This occurs mainly because the impact of the time value of money is not considered.

[0363] Dynamic economic life measurement: After considering the time value of money, the dynamic economic life measurement results are as follows:

[0364] Table 6

[0365]

[0366] It can be seen that the annual average cost of distribution transformer is the lowest, which is 13899 yuan, and the dynamic economic life is 14 years, which is close to the actual service life of the equipment. Because the equipment in the distribution network has been in use for a long time, the time value of money cannot be ignored. After considering the time value of money, the calculated economic life is basically consistent with the actual expected situation of the power supply bureau.

[0367] With the increase of the service life of the equipment, the annual average investment cost and the annual average net residual value of the distribution transformer decrease year by year, and the speed of decrease gradually slows down. Because the equipment is considered to be uniformly deteriorated, the annual average operation and maintenance cost and the annual average periodic maintenance cost increase uniformly year by year. The annual average fault maintenance cost first decreases and then increases, which is consistent with the "bathtub curve" law representing the fault.

[0368] The economic life calculation method proposed in this application can reduce the difficulty of obtaining the annual cost of the equipment, and only needs to determine the deterioration trend and the failure probability of the equipment. For the deterioration trend and the failure probability, this project has carried out sufficient qualitative analysis and scientific quantitative modeling, found out the rules of the equipment operation and described them with mathematical models. The calculation method realizes the optimization of the equipment scrapping and replacement time, and provides decision reference for technical improvement and overhaul.

[0369] The previously calculated theoretical economic life often does not match the actual economic life, and the theoretical economic life is usually ahead of time. This problem mainly occurs because the time value of money is not considered. According to the economic life calculation results of the transformer, the dynamic economic life lags behind the static economic life. Power grid companies in various provinces can apply this method to improve the accuracy of economic life measurement in the physical asset analysis and evaluation table.

[0370] The power equipment technology model realizes the change from "passive" replacement after the equipment fails to "active" prevention before the equipment fails. The reliable operation of the power system depends largely on the reliable operation of the equipment. Once the electrical equipment fails seriously, if there is no timely maintenance and repair, it will cause an accident, and the power operation and maintenance management is in a "passive" position. By applying the technical life prediction method of this project, the future operation and maintenance activities of the equipment can be accurately grasped, and the power operation and maintenance management is in a "active" position, which improves the management level of the operation and maintenance of the equipment.

[0371] The provincial company can formulate maintenance, repair and replacement plan according to the quality score change trend of power equipment. The current power equipment quality management only focuses on the shortcomings of the current state and does not consider the influence of the past state on the future. The application of the HGWO-LS-SVM technology life prediction model proposed in this paper can judge the development trend, so as to truly master the operation condition of the equipment and formulate targeted operation and maintenance plan.

[0372] In summary, the optimal operation life decision result of combining the economic life and the technical life of power equipment is 14 years. The economic life of the equipment reflects the operation efficiency of the distribution network, and reasonable application of the economic life estimation method can realize accurate investment and lean operation; the technical life of the equipment reflects the fault frequency of the distribution network, and reasonable application of the technical life prediction method can realize safe operation and reliable service. The joint application of economic life and technical life can take into account the benefits of reliable operation and economic efficiency.

[0373] Improving the operation and maintenance management level of power equipment is the safety guarantee and effective support for building a new power system with new energy as the main body. In order to avoid replacing power equipment that has not reached the life limit and can still be safely operated, and to prevent delaying the replacement of power equipment that has reached the life limit and cannot be safely operated, this paper proposes an optimal operation life decision method for power equipment under the joint application of economic life and technical life. The empirical results show that: 1) the economic life estimation method not only considers the equipment deterioration trend and failure probability, but also takes into account the time value of funds, and finally calculates the economic life as 14 years with an annual average minimum cost of 13899 yuan, solving the problem that the static economic life does not match the actual life. 2) The technical life prediction method considers internal and external factors, and HGWO improves the convergence speed and solution accuracy of the process of obtaining LS-SVM optimization parameters, and finally predicts the technical life as 16 years, which can provide decision reference for technical improvement and overhaul. 3) Finally, by combining the decision results of the two kinds of life, the optimal operation life of power equipment is determined as 14 years, which not only ensures safe operation but also saves cost.

[0374] Although the present application is described herein with reference to particular embodiments, it is to be understood that these examples are merely hypothetical illustrations of the principles and applications of the present application. It should therefore be understood that numerous modifications can be made to the exemplary embodiments and that other arrangements can be devised without departing from the spirit and scope of the present application as defined by the appended claims. It should be understood that the features described in connection with separate embodiments can be used in combination with features described in connection with other embodiments. It should also be understood that features described in connection with individual embodiments can be used in other described embodiments.

Claims

1. A method for optimal operation life decision of power equipment based on techno-economic efficiency, characterized in that, It comprises: S1, constructing a full life cycle cost estimation module of power equipment; adopting the output results of the full life cycle cost estimation module combined with the uniform deterioration trend of operation and maintenance cost and the failure probability of maintenance cost, a static economic life estimation model of power equipment is constructed; the output results of the static economic life estimation model are combined with the time value of funds to construct a dynamic economic life estimation model of power equipment; and the dynamic economic life estimation results are outputted; S2, constructing a technical life prediction model of power equipment based on HGWO-LS-SVM algorithm, and outputting the technical life prediction results; S3, obtaining the optimal operation life of the power equipment according to the dynamic economic life estimation results and the technical life prediction results; The specific method of S2 comprises: S2-1, quality assessment is performed on the scrapped equipment, the quality score is fitted with the equipment life by using the kernel density estimation KDE method, the quality score corresponding to the critical scrap point is obtained, and the kernel function is obtained; S2-2, a technical life prediction model of power equipment is constructed based on least squares support vector machine LS-SVM, differential evolution algorithm DE is introduced into grey wolf optimization algorithm GWO to obtain improved grey wolf algorithm HGWO; S2-3, the penalty factor in LS-SVM and the kernel function obtained in S2-1 are optimized by using the improved grey wolf algorithm HGWO to obtain the quality scores of power equipment in different periods; S2-4, the quality scores in different periods obtained in S2-3 are compared with the quality score corresponding to the critical scrap point obtained in S2-1, when the quality score in different periods is lower than the quality score corresponding to the critical scrap point, the corresponding time is the technical life prediction result.

Images