Power grid material sampling inspection strategy making method and system based on multi-dimensional analysis

Abstract

The invention relates to the field of power grid material quality management and control, in particular to the technical field of quality evaluation, and more particularly relates to a power grid material sampling inspection strategy making method and system based on multiple dimensions, aiming at a special application scene of continuous small-batch arrival acceptance inspection of power grid materials by multiple suppliers, and aiming at the target of'quality-quality few inspection and quality-quality multiple inspection '. By introducing the AQL theory and integrating multiple factors such as the historical supply quality level of suppliers and the bearing capacity of detection resources, a total cost minimization model under the condition of maximum supplier sampling inspection strength and quality level evaluation sorting consistency is constructed, so that the optimal sampling inspection strategy of different power grid materials and different suppliers can be intelligently obtained, and the sampling inspection efficiency is improved. And a support is provided for an electric power company to make a sampling inspection scheme more accurately and realize reasonable allocation and utilization of resources.

Description

Technical Field

[0001] The present invention relates to the field of power grid material quality control, specifically to the field of quality evaluation technology, and more specifically to a method and system for formulating a multi-dimensional power grid material sampling inspection strategy. Background Art

[0002] As an important part of national infrastructure, the quality of power grid materials is directly related to the safe and stable operation of the power grid and the reliability of power supply. With the acceleration of the construction of new power systems with new energy as the main body, new changes have emerged, such as the rapid development of distributed new energy and the continuous emergence of new energy forms.

[0003] At present, the scale of new energy grid connection continues to expand, and the "double high" characteristics of a high proportion of new energy and a high proportion of power electronic equipment access to the power grid are becoming increasingly prominent. The traditional routine sampling strategy can no longer meet the sampling requirements under the new situation.

[0004] On the one hand, the current sampling inspection strategy is basically formulated based on the material supply volume and the material sampling inspection pass rate. The application of historical business data related to quality is limited, resulting in some parameter settings still relying on the experience and judgment of experts and lacking a scientific basis; on the other hand, the current sampling inspection strategy only relies on the sampling inspection pass rate to evaluate "high quality" or "poor quality". There is a lack of more refined sampling inspection strategies for suppliers with different supply quality levels and the same type of materials arriving in different warehouses in the same time period.

[0005] Therefore, it is necessary to optimize the power grid material sampling strategy in response to new forms to solve the pain points and difficulties in the current sampling work, further improve the scientificity and rationality of material sampling work, improve the accuracy and representativeness of sampling results, and provide an accurate data basis for material quality evaluation and supplier evaluation and profiling. Summary of the invention

[0006] In view of the fact that the existing sampling inspection methods are facing the new challenges of the continuous expansion of the scale of new energy grid connection and the increasingly prominent "double high" characteristics of a high proportion of new energy and a high proportion of electronic power equipment access, there is a problem that the risk of "sick" materials entering the grid is not completely avoided, and quality incidents still occur after the materials are put into operation. This paper provides a method and system for formulating power grid material sampling inspection strategies based on multiple dimensions to improve the sampling inspection efficiency, reduce the sampling inspection cost, and improve the representativeness of the sampling inspection results.

[0007] In a first aspect, the present invention provides a method for formulating a power grid material sampling inspection strategy based on multi-dimensional analysis, which is characterized by comprising the following steps:

[0008] S1: Collect the actual sampling population and sample size of multiple suppliers, and calculate the original qualified rate p of each supplier based on the total qualified and unqualified number of sampling tasks i, use Wilson algorithm to correct the original qualified rate and obtain the corrected original qualified rate Pr i ;

[0009] S2: Establish quality level evaluation model Pr i * , inputting each supplier's basic qualified rate, unqualified degree and quality improvement index into the quality level evaluation model, and the quality level evaluation model analyzes them and outputs a quality level evaluation value;

[0010] S3: Use the quality level evaluation value to establish a model that maximizes the consistency of quality evaluation ranking to determine the sampling ratio of each supplier:

[0011] S4: Establish a model to minimize the total annual inspection cost based on the inspection ratio;

[0012] S5: Genetic algorithm and multi-objective Beluga optimization algorithm are used to build an intelligent solver. The intelligent solver solves the model of minimizing the total annual inspection cost and obtains the optimal inspection strategy.

[0013] Preferably, the corrected original qualified rate Pr is obtained i The methods include:

[0014] Take the Wilson confidence interval The average of the upper and lower limits is used to correct the original qualified rate to obtain the corrected qualified rate Pr i :

[0015]

[0016] Where: k i represents the batches sampled from the i-th supplier within a certain historical period; u i represents the total number of qualified materials sampled from the i-th supplier; v i Represents the total number of unqualified materials sampled from the i-th supplier; m i Represents the total number of samples inspected in the historical arrivals of the i-th supplier;

[0017] For the jth batch of sampled materials (j=1, 2, ..., k i ), when the sampling result is "qualified", let u ij =m ij , v ij =0, recorded as (m ij , 0); when the sampling result is "unqualified", let u ij =0, v ij =m ij , where u ij represents the number of qualified products of the jxth sampled materials of the i-th supplier, v ijrepresents the number of unqualified products in the jth batch of sampled materials from the i-th supplier; p i represents the original qualified rate of the ith supplier; Pr i represents the corrected qualified rate of the i-th supplier after considering the confidence level; z α Represents the quantile of the normal distribution, z α Represents the quantile of the normal distribution, with a value of 1.96, corresponding to a 95% confidence level.

[0018] Preferably, the supplier quality level evaluation value Pr i * Obtained according to the following formula:

[0019]

[0020] Where: Pr i represents the comprehensive basic qualification rate of the ith supplier, Pr i ∈[0, 1], Pr i =APr i 1 +BPr i 2 +CPr i 3 , Pr i 1 , Px i 2 , Pr i 3 represents the sampling qualified rate of the i-th supplier in the past three years, the past year, and the past month; A, B, and C are Pr after considering the difference in long-term and short-term impacts. i 1 , Pr i 2 , Pr i 3 The weights of A, B, and C are 0.1, 0.2, and 0.7 respectively;

[0021] P i represents the most recent cumulative score in the bad behavior record of the i-th supplier, P i ∈[0, 10];

[0022] P i ′ is P i After normalization, the data

[0023] P i ′∈[0,1];I i represents the quality improvement index of the i-th supplier, I i =R 2 -R 1(Formula 4), R 1 , R 2 are the sampling pass rates of suppliers in the last two six months. When there is no data in any of the time periods, let I i =0,I i ∈[-0.5, 0.7].

[0024] Preferably, in step S3, constructing a maximum quality evaluation sorting consistency model specifically includes the following steps:

[0025] S3-1: Setting sequence

[0026] Set up supplier quality level evaluation sequence Sort the supplier quality level evaluation values ​​from high to low to meet

[0027] set up represents a set of sampling strategies for supplier i, and the actual sampling ratio of the supplier corresponding to the strategy Under the premise of "less inspection for high quality and more inspection for low quality", there exists a sampling ratio sequence Sr I = {Sr 1 , ..., Sr i , ..., Sr I}, satisfying Sr 1 ≤...≤Sr i ≤…≤Sr I ;

[0028] S3-2: Determine the benchmark sampling ratio

[0029] Assume sr i ′ is the benchmark sampling ratio for each supplier, Sr i ′ According to the quality level evaluation value of each supplier and the minimum and maximum thresholds of the sampling ratio, the interval mapping function is used to determine whether Sr′ is satisfied. 1 ≤…≤Sr′ i ≤…≤Sr′ I ;

[0030] For the quality level evaluation value, the quality level evaluation interval can be obtained When the minimum sampling ratio of all suppliers is a and the maximum sampling ratio is b, the sampling ratio benchmark range [a, b] can be obtained. A mapping relationship is established based on the sampling ratio benchmark range interval and the quality level evaluation interval:

[0031]

[0032] According to (Formula 6), (Formula 7) is derived:

[0033]

[0034] S3-3: Establishing the objective function of minimizing the deviation distance

[0035] Calculate Sr i The actual value and the given benchmark sampling ratio Sr i The distance d between i , d i =Sr i -Sr i ′, d is the total deviation distance of all suppliers, then there is an objective function that satisfies the consistency of quality evaluation ranking:

[0036]

[0037] Among them, the ratio of the difference between the actual sampling ratio of adjacent ranking suppliers and the interpolation of the benchmark ratio is close to

[0038] [0.9, 1], that is

[0039] Preferably, the model for minimizing the total annual inspection cost is obtained by the following formula:

[0040]

[0041] Among them, W 1i Costs incurred for receiving materials, W 2i Costs incurred for rejecting materials, Pr i is the sampling pass rate, N i，总 The supplier's total annual arrival volume and sampling base and the corresponding sample size

[0042] Preferably, the intelligent solver solves the model for minimizing the total annual sampling cost by the following steps:

[0043] S5-1: Determine the sample size using the GB / T2828.1-2012 normal inspection single sampling plan retrieval table The batch size is determined based on the sample size and the cumulative arrival volume of each supplier throughout the year. The value range of The combination value retrieval table of

[0044] S5-2: Based on The value range and correspondence of Generate multiple sampling strategies using intelligent solvers And calculate the sampling ratio Sr under each sampling strategy i ,

[0045] S5-3: Set the sampling ratio Sr under each sampling strategyi Construct a decreasing sequence of quality evaluation in descending order from high to low, and construct an increasing sequence of sampling ratio according to the increasing order of sampling ratio. Determine the benchmark sampling ratio by presetting the sampling ratio range [a, b]. Generate multiple sets of sampling ratio sequences Sr I , Sr. l = {Sr 1 , Sr. 2 , ..., Sr i , ..., S r I}, based on the consistency rule of quality level evaluation ranking, the deviation distance d between the actual sampling ratio and the benchmark sampling ratio is calculated by the following formula:

[0046] d i =Sr i -Sr′ i

[0047]

[0048] MIN(d)

[0049] Find the minimum deviation distance d and select the sampling ratio sequence that meets the inspection conditions as the optimal sequence solution S5-4: Calculation Corresponding sampling strategy sequence SIS I The sampling cost of the sampling strategy is selected as the optimal sampling strategy sequence with the minimum total cost.

[0050] In a second aspect, the present invention provides a system for formulating a power grid material sampling inspection strategy based on multi-dimensional analysis, comprising the following modules:

[0051] A data acquisition module, which is used to acquire data of all suppliers of power grid materials for which sampling inspection strategies are to be formulated, so as to be used in formulating sampling inspection strategies;

[0052] The data analysis module, after the data module obtains the data, analyzes the supplier quality level evaluation value Pr based on the historical data. i * , and the benchmark sampling ratio Sr′ of supplier i determined based on the quality level evaluation results i ;

[0053] Strategy output module, the strategy output module includes an optimal sequence solution calculation module, a sampling strategy sequence generation module, and an optimal sampling strategy generation module; the optimal sequence solution calculation module uses the quality level value to establish a maximum quality evaluation ranking consistency model and obtain the optimal sequence solution Sr I *The sampling strategy sequence generation module automatically generates the optimal sequence solution Sr I * Multi-group sampling strategy SIS i ; The optimal sampling strategy generation module is based on the sampling strategy SIS i The sampling ratio in the model is used to establish a model that minimizes the total annual sampling cost;

[0054] An intelligent solver is constructed using genetic algorithms and multi-objective Beluga optimization algorithms. The intelligent solver solves the model for minimizing the total annual inspection cost and obtains the optimal inspection strategy.

[0055] Preferably, the strategy formulation module uses the quality level value to establish a model for maximizing the consistency of quality evaluation ranking, and the process includes:

[0056] Get the supplier quality level evaluation sequence from high to low in

[0057] Based on the supplier quality level evaluation value and the minimum and maximum thresholds of the sampling ratio, the supplier benchmark sampling ratio sequence is determined through the interval mapping function, and further based on the quality level evaluation interval The sampling ratio benchmark range is [a, b], where a is the supplier's minimum sampling ratio and b is the supplier's maximum sampling ratio. A mapping relationship is established:

[0058]

[0059] A mapping relationship is established based on the sampling ratio benchmark range interval and the quality level evaluation interval to obtain and output the maximum quality evaluation ranking consistency model:

[0060]

[0061] Costs incurred based on receiving materials W 1i , the cost of rejecting materials 2i , sampling inspection qualified rate Pr i 、The total amount of goods received by suppliers throughout the year N i总 , sampling base and the corresponding sample size Output minimizes the total annual sampling cost;

[0062]

[0063] Based on the sampling ratio sequence as the optimal sequence solution and total cost Minimum, intelligent solver solves the model of minimizing the total annual sampling cost and outputs the optimal sampling strategy sequence Beneficial Effects

[0064] (1) The present invention collects actual sampling data from suppliers and uses the Wilson algorithm to correct the pass rate. This scientific method reduces the interference of human factors and makes the sampling results more accurate. This method takes into account the confidence level, so that the corrected pass rate can better represent the actual quality level of the supplier, and improves the accuracy and representativeness of the sampling results.

[0065] (2) The model for minimizing the total annual inspection cost proposed by the present invention takes into account the costs of receiving and rejecting materials, as well as factors such as the inspection pass rate and the supplier's arrival volume. Through this model, the total cost under different inspection strategies can be accurately calculated, and the strategy with the lowest cost can be found. This method not only reduces unnecessary inspections, but also optimizes resource allocation, thereby effectively reducing the overall inspection cost.

[0066] (3) Based on the sampling ratio, a model for minimizing the total annual sampling cost was established. In this process, Bayesian statistical theory was introduced to infer the non-conforming rate. This method estimates the non-conforming rate of materials more scientifically through the transformation of prior distribution and posterior distribution. Combined with the AQL theory, a total cost minimization model was constructed to maximize the consistency of supplier sampling intensity and quality level evaluation ranking, making the formulation of sampling strategy more scientific and reasonable.

[0067] (4) Using intelligent algorithms to solve the optimal sampling strategy, an intelligent solver is constructed by introducing genetic algorithms and multi-objective White Whale optimization algorithms. The intelligent solver solves the model of minimizing the total annual sampling cost, automatically processes a large amount of data, quickly generates the optimal sampling strategy, and obtains the optimal sampling strategy. This significantly improves the efficiency of sampling work. This method not only improves the solving speed, but also improves the accuracy of strategy formulation, making the sampling strategy more intelligent and automated. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 A flow chart of a method for formulating a power grid material sampling inspection strategy based on multi-dimensional analysis provided in Example 1 of the present invention;

[0069] Figure 2 A schematic diagram of a calculation flow of a supplier's revised qualified rate after considering the confidence level provided in Example 1 of the present invention;

[0070] Figure 3 This is a logical diagram for solving the model for minimizing the total annual sampling cost provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0071] The present invention is further described below in conjunction with specific embodiments, but the protection scope of the present invention is not limited thereto.

[0072] Example 1

[0073] This embodiment uses the following method to obtain a power grid material sampling strategy, which includes the following steps:

[0074] S1: Collect the actual sampling population and sample size of multiple suppliers, and calculate the original qualified rate p of each supplier based on the total qualified and unqualified number of sampling tasks i , use Wilson algorithm to correct the original qualified rate and obtain the corrected original qualified rate Pr i ;

[0075] Get the corrected original pass rate Pr i The methods include:

[0076] Take the Wilson confidence interval The average of the upper and lower limits is used to correct the original pass rate to obtain the corrected pass rate Pri:

[0077]

[0078] Where: k i represents the batches sampled from the i-th supplier within a certain historical period; u i represents the total number of qualified materials sampled from the i-th supplier; v i Represents the total number of unqualified materials sampled from the i-th supplier; m i Represents the total number of samples inspected in the historical arrivals of the i-th supplier;

[0079] For the jth batch of sampled materials (j=1, 2, ..., k i ), when the sampling result is "qualified", let u ij =m ij , v ij =0, recorded as (m ij , 0); when the sampling result is "unqualified", let u ij =0, v ij =m ij , where u ij represents the number of qualified products of the jth batch of sampled materials from the i-th supplier, v ij represents the number of unqualified products in the jth batch of sampled materials from the i-th supplier; p i represents the original qualified rate of the ith supplier; Pr i represents the corrected qualified rate of the i-th supplier after considering the confidence level; z α Represents the quantile of the normal distribution, z α Represents the quantile of the normal distribution, with a value of 1.96, corresponding to a 95% confidence level.

[0080] Preferably, the supplier quality level evaluation value Pr i* Obtained according to the following formula:

[0081]

[0082] Where: Pr i represents the comprehensive basic qualification rate of the ith supplier, Pr i ∈[0, 1], Pr i =APr i 1 +BPr i 2 +CPr i 3 , Pr i 1 , Pr i 2 , Pr i 3 represents the sampling qualified rate of the i-th supplier in the past three years, the past year, and the past month; A, B, and C are Pr after considering the difference in long-term and short-term impacts. i 1 , Pr i 2 , Pr i 3 The weights of A, B, and C are 0.1, 0.2, and 0.7 respectively;

[0083] P i represents the most recent cumulative score in the bad behavior record of the i-th supplier, P i ∈[0, 10];

[0084] P i ′ is P i After normalization, the data

[0085] P i ′∈[ 0 , 1 ];I i represents the quality improvement index of the i-th supplier, I i =R 2 -R 1 (Formula 4), R 1 , R 2 are the sampling pass rates of suppliers in the last two six months. When there is no data in any of the time periods, let I i =0,I i ∈[-0.5,0.7].

[0086] S2: Establish quality level evaluation model Pr i *, inputting each supplier's basic qualified rate, unqualified degree and quality improvement index into the quality level evaluation model, and the quality level evaluation model analyzes them and outputs a quality level evaluation value;

[0087] The supplier quality level evaluation value Pr i * Obtained by the following formula:

[0088]

[0089] Where: Pr i represents the comprehensive basic qualification rate of the ith supplier, Pr i ∈[0, 1], Pr i =APr i 1 +SPr i 2 +CPr i 3 , Pr i 1 , Pr i 2 , Pr i 3 is the representative sampling qualified rate of the i-th supplier, A, B, and C are Pr after considering the difference in long-term and short-term impacts. i 1 , Pr i 2 , Pr i 3 The weight of i 1 , Pr i 2 , Pr i 3 represents the sampling qualified rate of the i-th supplier in the past three years, the past year, and the past month; A, B, and C are Pr after considering the difference in long-term and short-term impacts. i 1 , Pr i 2 , Pr i 3 The weights of A, B, and C are 0.1, 0.2, and 0.7 respectively;

[0090] In the present invention, an exponential function with base e is introduced to i ′、I i Nonlinear transformation is performed to make the quality level evaluation model more sensitive. Specifically, when P i ' is higher (that is, the original unqualified degree Pi is lower), The supplier quality level evaluation value Pr will be significantly increased. i * will also increase accordingly; when P i ' is lower (i.e. the original unqualified degree P i Higher), The supplier quality level evaluation value Pr will be significantly reduced. i * Then it becomes smaller. When I i When it is high, it indicates that the supplier performs well in quality improvement, and the dynamic adjustment factor Will significantly increase, thereby improving the supplier quality level evaluation value Pr i * On the contrary, when I i When it is low, it means that the supplier's supply quality has not improved, and the dynamic adjustment factor Will significantly reduce and make the supplier quality level evaluation value Pr i * reduce.

[0091] P i Represents the most recent cumulative score in the bad behavior record of the i-th supplier. In the field of power grid material management, a point system management system is implemented, which is also an important basis for grading suppliers after quality problems are found in random inspections. Combined with the analysis of the cumulative score data of supplier bad behavior, the supplier unqualified degree P is introduced i As a "penalty" term (the penalty is P i >0, appropriately increase sampling), P i is the cumulative score of all unqualified products of supplier i in the past six months. Considering the lag of data validity and the announcement of penalty results, P i Take the most recent cumulative score in the supplier's bad behavior record. i The higher the value, the lower the quality level evaluation value.

[0092] P i ′ is P i After normalization, the data , P i i′∈[0,1];

[0093] I i represents the quality improvement index of the i-th supplier, I i =R 2 -R 1 (Formula 4), R 1 , R 2 are the sampling qualified rates of suppliers in the last two consecutive time periods. When there is no data in any of the time periods, let I i =0; preferably, R1 , R 2 They are respectively the supplier’s sampling inspection pass rates in the last two to six months.

[0094] The embodiment of the present invention uses the change of the supplier's sampling pass rate in the past two and a half years to obtain the quality improvement index I i Preferably, when R 1 , R 2 When the sampling inspection pass rate of the supplier in the last two six months is respectively R, the sampling inspection pass rate of the supplier in the previous six months is recorded as R 1 The qualified rate of random inspection in the last six months is recorded as R 2 If the sampling pass rate in the most recent six months is greater than the sampling pass rate in the previous six months, then I i It can be used as a "reward" item (reward means appropriate reduction of random inspections), otherwise it can be used as a "penalty" item. i It only works if the supplier has sampling data in the past two and a half years and the difference between the sampling results in the two time periods is not 0. Otherwise, let I i =0, does not affect the final supplier quality level evaluation, preferably, I i ∈[-0.5, 0.7];

[0095] S3: The method of using the quality level value to establish a model to maximize the consistency of quality evaluation ranking to determine the sampling ratio of each supplier includes:

[0096] S3-1: Setting sequence

[0097] Set up supplier quality level evaluation sequence Sort the supplier quality level evaluation values ​​from high to low to meet

[0098] set up represents a set of sampling strategies for supplier i, and the actual sampling ratio of the supplier corresponding to the strategy Under the premise of "less inspection for high quality and more inspection for low quality", there exists a sampling ratio sequence Sr I = {Sr 1 , ..., Sr i , ..., Sr I}, satisfying Sr 1 ≤...≤Sr i ≤…≤Sr I .

[0099] S3-2: Determine the benchmark sampling ratio

[0100] Assume Sr i ′ is the benchmark sampling ratio for each supplier, Sr i′ According to the quality level evaluation value of each supplier and the minimum and maximum thresholds of the sampling ratio, the interval mapping function is used to determine whether Sr′ is satisfied. 1 ≤…≤Sr′ i ≤…≤Sr′ I .

[0101] For the quality level evaluation value, the quality level evaluation interval can be obtained When the minimum sampling ratio of all suppliers is a and the maximum sampling ratio is b, the sampling ratio benchmark range [a, b] can be obtained. A mapping relationship is established based on the sampling ratio benchmark range interval and the quality level evaluation interval:

[0102]

[0103] According to (Formula 6), (Formula 7) is derived:

[0104]

[0105] S3-3: Establishing the objective function of minimizing the deviation distance

[0106] Calculate Sr i The actual value and the given benchmark sampling ratio Sr i The distance d between i , d i =Sr i -Sr i ′, d is the total deviation distance of all suppliers, then there is an objective function that satisfies the consistency of quality evaluation ranking:

[0107]

[0108] Among them, the ratio of the difference between the actual sampling ratio of adjacent ranking suppliers and the interpolation of the benchmark ratio is close to [0.9, 1], that is,

[0109] S4: A model for minimizing the total annual inspection cost is established based on the inspection ratio. The model for minimizing the total annual inspection cost is obtained by the following formula:

[0110]

[0111] Among them, W 1i Costs incurred for receiving materials, W 2i Costs incurred for rejecting materials, Pr i is the sampling pass rate, N i总 The supplier's total annual arrival volume and sampling base and the corresponding sample size

[0112] Due to the importance and particularity of power grid materials for the safe and stable operation of the power grid system, power grid companies generally adopt the "0 in, 1 out" principle when accepting and sampling. If all the samples of the batch of samples are qualified, the batch of materials will be accepted. If there is only one unqualified product, the whole batch of materials will be directly rejected.

[0113] In the present invention, the Bayesian statistical theory is introduced to find the expected failure rate. Let π(1-Pr i ) is the overall parameter 1-Pr i A probability density function, in acceptance sampling, the parameter 1-Pr i is the failure rate. Assuming that the failure rate of materials in the historical sampling process follows a certain distribution, the distribution is the failure rate 1-Pr i The prior distribution of π(1-Pr i ). When a sample size is When randomly selecting and checking whether the samples are qualified, the number of defective products X usually follows a binomial distribution and (1-Pr i )∈(0,1), according to the properties of the estimated parameters and the transformation conditions between the prior distribution and the posterior distribution, the posterior distribution π(1-Pr i |x) is based on the distribution of the sample X and 1-Pr i The prior distribution of π(1-Pr i ) is a distribution calculated by the conditional probability distribution method in probability theory. Assume that the probability density function of the random variable X is p(x; 1-Pr i ), in acceptance sampling, this random variable is used as the distribution of the sampling population. Based on the distribution of the population, the sample x and the parameter 1-Pr i The joint probability density function is h(x|1-Pr i )=p(x|1-Pr i )π(1-Pr i ). When x is given, the parameter 1-Pr i The posterior probability density function π(1-Pr i |x) is:

[0114]

[0115] Where m(x) is the marginal distribution of the sample,

[0116] Considering the failure rate (1-Pr i )∈(0,1), with the failure rate 1-Pr i As a random variable of beta distribution, the probability density function of failure rate is: Where p>0, q>0, B(p,q) is the Beta function, Γ(x) is the gamma function. The mean μ and variance σ of this distribution are 2 :

[0117]

[0118] Move the item to get:

[0119]

[0120] At the same time, from the perspective of the supply chain, the main cost of random inspection for arrival acceptance is composed of three parts: random inspection fee (A), user loss caused by missed inspection (B), and loss to suppliers caused by rejection of batches of materials (R), thereby obtaining the costs under acceptance and rejection conditions, as follows:

[0121] (1) Total cost composition in the case of acceptance

[0122] When all samples pass the inspection, the cost of receiving the batch of materials is equal to the sum of the sampling inspection costs of all samples and the user's losses caused by the possible omission of inspection of the remaining uninspected materials. The total acceptance cost W is obtained. 1i :

[0123]

[0124] Among them, μ(1-Pr i |x=0) is the pass rate 1-Pr i The expectation of the posterior distribution is calculated using the Bayesian statistical method described above.

[0125] (2) Total cost composition in case of rejection

[0126] When defective products are found in the sample inspection, the cost of rejecting the batch of materials is equal to the sum of the sampling inspection costs of all samples and the loss to supplier i caused by rejecting the batch of materials. The total rejection cost W is 2i :

[0127]

[0128] Considering both acceptance and rejection, there is a certain probability of acceptance in a single sampling inspection, denoted by L(1-Pr i ) is the failure rate 1-Pr i When the power grid material sampling inspection Ac=0, Then the expected cost W for a random inspection of each supplier can be calculated: i :

[0129]

[0130] Based on this, the annual sampling cost for each supplier can be obtained.

[0131] Thus, the annual sampling cost for all suppliers is obtained:

[0132] S5: Use genetic algorithm and multi-objective White Whale optimization algorithm to build an intelligent solver, which solves the model of minimizing the total annual inspection cost, such as Figure 2 As shown in Figure 2, the method for obtaining the optimal sampling strategy includes:

[0133] S5-1: Determine the sample size using the GB / T2828.1-2012 normal inspection single sampling plan retrieval table (Table 1) The batch size is determined based on the sample size and the cumulative arrival volume of each supplier throughout the year. The value range of

[0134] Table 1 AQL retrieval table for random inspection of power grid materials upon arrival

[0135]

[0136] S5-2: Based on The value range and correspondence of Generate all possible sampling strategies using intelligent solvers And calculate the sampling ratio Sr under each strategy i ,

[0137] S5-3: For all possible Sr i , construct a decreasing sequence of quality evaluation in descending order from high to low, and construct an increasing sequence of sampling ratio according to the increasing order of sampling ratio, and determine the benchmark sampling ratio by presetting the sampling ratio range [a, b]. Generate multiple sets of sampling ratio sequences Sr I , Sr. I = {Sr 1 , Sr. 2 , ..., Sr i , ..., Sr I}, based on the consistency rule of quality level evaluation ranking, calculate the deviation distance d between the actual sampling ratio and the benchmark sampling ratio;

[0138] d i =Sr i -Sr′ i

[0139]

[0140] MIN(d)

[0141] Find the minimum deviation distance d and select the sampling ratio sequence that meets the inspection conditions as the optimal sequence solution

[0142] S5-4: Due to Corresponding to the only Sr i , but the same Sr i May correspond to multiple groups of sampling strategies SIS i , so further calculation All possible sampling strategy sequences SIS I The sampling cost of the sampling strategy is selected as the optimal sampling strategy sequence with the minimum total cost.

[0143]

[0144] Example 2

[0145] On the basis of Example 1, the data of 10 typical suppliers of a certain product of a certain company is taken as an example to further illustrate the sampling inspection strategy formulation method based on multi-dimensional analysis described in the present invention. First, the historical arrival data, historical sampling inspection qualification, bad behavior processing and other data of 10 typical suppliers are collected. For the convenience of calculation, according to the survey of the management requirements of the power grid material sampling inspection business, some parameters of the model are assigned values, Y 0 =200, a=0.05, b=0.27, N i总 Take the total annual arrival volume of supplier i in 2023, and the initial cost value is A = 5000 yuan, B = 12000 yuan, and R = 8000 yuan. In addition, considering the differences in the historical supply quantities of different suppliers, in order to ensure that the evaluation of their sampling pass rate can be at the same level and to better test the effectiveness of the model, the Wilson algorithm is introduced to transform the sampling pass rate, and the model trial uses the adjusted sampling pass rate data. The historical arrival volume, sampling pass rate and quality level evaluation results of 10 typical suppliers are shown in Table 2.

[0146] Table 2 Historical arrival quantity, qualified rate and quality level evaluation values ​​of 10 typical suppliers

[0147]

[0148] According to the GB / T 2828.1-2012 normal inspection single sampling plan table, combined with the arrival data of 10 typical suppliers, the sample size can be obtained. and batch The value range of is shown in Table 3.

[0149] Table 3 Correspondence between sample size and batch range

[0150]

[0151] (1) Option 1: Different sample sizes for each supplier

[0152] Assuming that each supplier implements a different sample size plan, the solution procedure is used to calculate each one and generate all possible sampling strategy combinations for 10 typical suppliers. And the corresponding sampling ratio Some calculation results are shown in Table 4.

[0153] Table 4 Sampling strategy combination calculation table (only part of the data is listed for illustration)

[0154]

[0155]

[0156] Screening the optimal strategy combination that meets the consistency requirements of sampling intensity and quality evaluation ranking The deviations between these sequences and the supplier quality rating sequences are calculated, and the calculation results are shown in Table 5.

[0157] Table 5 Cost calculation results under the consistent optimal strategy combination (only part of the data is listed for illustration)

[0158]

[0159] Select the optimal strategy set based on maximum consistency and minimum total cost This determines the sampling strategy for each supplier As shown in Table 6.

[0160] Table 6 Optimal sampling strategies for 10 suppliers

[0161]

[0162]

[0163] Thus, the sampling strategy formulation method disclosed in the present invention obtains the sampling strategy of the historical data of 10 typical suppliers of a certain product of a certain company in 2023. The sampling strategy provides the optimal sampling base and samples for different suppliers.

[0164] Example 3

[0165] On the basis of Example 1, the feasibility of the sampling inspection strategy of the present invention is further verified and illustrated.

[0166] Combining the sample size code table of GB / T 2828.2-2012 and the normal inspection single sampling plan (Table 1 and Table 2), according to the rule of Ac=0, Re=1, the inspection level and acceptance quality limit (AQL) that can be implemented by the above optimal sampling strategy can be approximately determined, as shown in Table 7. It can be seen that the inspection level requirements for suppliers with poor quality are stricter.

[0167] Table 7 Executable inspection level and AQL query results under the optimal sampling strategy

[0168]

[0169] To further verify the feasibility of the above sampling strategy, the arrival data of 10 suppliers in 2023 are used to simulate the generation of sampling tasks. By comparing the sampling tasks generated by the new strategy with the differences in historical sampling task items and sample sizes, it can be found that under the new strategy, the number of sampling tasks generated is less than that of the original strategy due to the accumulation of suppliers' arrival volume in a short period of time, but the sample size of a single generated task increases, as shown in Table 8.

[0170] Table 8 Comparison of task generation between two sampling strategies

[0171]

[0172] Assuming that each supplier implements the same sample size plan, the calculations are performed one by one according to the solution procedure and the solutions are solved one by one. The optimal sampling strategy for 10 typical suppliers And the corresponding sampling ratio Some calculation results are shown in Tables 9 to 11. It can be found that when the sample size is larger, a higher inspection level and AQL are performed.

[0173] Table 9 Optimal sampling strategy calculation results table ( Total cost: RMB 8.9814 million)

[0174]

[0175]

[0176] Table 10 Optimal sampling strategy calculation results table ( Total cost: RMB 8,286,600)

[0177]

[0178] Table 11 Optimal sampling strategy calculation results table ( Total cost: RMB 7,410,600)

[0179]

[0180]

[0181] The above sampling strategy is also used to simulate the generation of sampling tasks for the 2023 arrival data of 10 suppliers. By comparing the sampling tasks generated by the new strategy with the historical sampling task items and sample sizes (as shown in Table 12), it can be seen that the new strategy generates fewer tasks but more samples, which can better reflect the overall quality of batch materials through sampling.

[0182] Table 12 Comparison of sampling tasks generated by new and old strategies

[0183]

[0184] To further test the effectiveness of the model, we take A∈[2500,10000], B∈[7000,15000], R∈[500,10000], and observe the detection cost, the loss caused by missed detection, and the supplier loss adjustment. When the sampling ratio and sample size of Scheme 2 are determined, the batch There are no multiple choices, so the cost does not affect the generation of the optimal sampling strategy; the following only explains the changes in the optimal sampling strategy of Scheme 1.

[0185] Table 13 The impact of detection cost A on the optimal sampling strategy

[0186]

[0187]

[0188]

[0189] Table 14 The impact of loss B caused by missed inspection on the optimal sampling strategy

[0190]

[0191]

[0192] Table 15 The impact of supplier loss R on the optimal sampling strategy

[0193]

[0194]

[0195] It can be seen from Tables 13-15 that the unit inspection cost and supplier losses have no effect on the sampling strategy; the losses caused by missed inspections do have an impact on the sampling strategy. When B ≤ 8,500 yuan, a new strategy will be generated. Comparing the two strategies, it is found that B is more reasonable when it is the initial value.

[0196] Example 4

[0197] The embodiment of the present invention provides a system for formulating a power grid material sampling inspection strategy based on multi-dimensional analysis, including the following modules:

[0198] A data acquisition module, which is used to acquire data of all suppliers of power grid materials for which sampling inspection strategies are to be formulated, so as to be used in formulating sampling inspection strategies;

[0199] The data analysis module, after the data module obtains the data, analyzes the supplier quality level evaluation value Pr based on the historical data. i * , and the benchmark sampling ratio Sr′ of supplier i determined based on the quality level evaluation results i ;

[0200] A strategy formulation module, wherein the strategy formulation module uses the quality level value to establish a model that maximizes the consistency of quality evaluation ranking to determine the sampling ratio of each supplier, and establishes a model that minimizes the total annual sampling cost based on the sampling ratio;

[0201] The output module uses a genetic algorithm and a multi-objective Beluga optimization algorithm to construct an intelligent solver, and the intelligent solver solves a model for minimizing the total annual sampling cost to obtain an optimal sampling strategy.

[0202] The strategy formulation module uses the quality level value to establish a model that maximizes the consistency of quality evaluation ranking, and the process includes:

[0203] Get the supplier quality level evaluation sequence from high to low in

[0204] Based on the supplier quality level evaluation value and the minimum and maximum thresholds of the sampling ratio, the supplier benchmark sampling ratio sequence is determined through the interval mapping function, and further based on the quality level evaluation interval The sampling ratio benchmark range is [a, b], where a is the supplier's minimum sampling ratio and b is the supplier's maximum sampling ratio. A mapping relationship is established:

[0205]

[0206] A mapping relationship is established based on the sampling ratio benchmark range interval and the quality level evaluation interval to obtain and output the maximum quality evaluation ranking consistency model:

[0207]

[0208] Costs incurred based on receiving materials W 1i , the cost of rejecting materials 2i , sampling inspection qualified rate Pr i、The total amount of goods received by suppliers throughout the year N i总 , sampling base and the corresponding sample size Output minimizes the total annual sampling cost;

[0209]

[0210] Based on the sampling ratio sequence as the optimal sequence solution and total cost Minimum, intelligent solver solves the model of minimizing the total annual sampling cost and outputs the optimal sampling strategy sequence Furthermore, the present invention is based on the model solving intelligent algorithm, uses programming languages ​​such as Python to build an intelligent solver, configures multiple sets of solution algorithm packages, and supports the model to flexibly select adaptive algorithm packages and efficiently solve.

[0211] The embodiments are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention.

Claims

1. A method for formulating a power grid material sampling inspection strategy based on multi-dimensional analysis, characterized by: The steps include: S1: Collect the actual sampling population and sample size of multiple suppliers, and calculate the original qualified rate p of each supplier based on the total qualified and unqualified number of sampling tasks i , use Wilson algorithm to correct the original qualified rate and obtain the corrected original qualified rate Pr i ; S2: Establish quality level evaluation model Pr i * , inputting each supplier's basic qualified rate, unqualified degree and quality improvement index into the quality level evaluation model, and the quality level evaluation model analyzes them and outputs a quality level evaluation value; S3: Use the quality level evaluation value to establish a maximum quality evaluation ranking consistency model to determine the optimal solution, and obtain the sampling ratio sequence of each supplier that meets the optimal solution; S4: Establish a model to minimize the total annual inspection cost based on the inspection ratio sequence; S5: Genetic algorithm and multi-objective Beluga optimization algorithm are used to build an intelligent solver. The intelligent solver solves the model of minimizing the total annual inspection cost and obtains the optimal inspection strategy.

2. The method for formulating a power grid material sampling inspection strategy based on multi-dimensional analysis according to claim 1 is characterized in that: Get the corrected original pass rate Pr i The methods include: Take the Wilson confidence interval The average of the upper and lower limits is used to correct the original qualified rate to obtain the corrected qualified rate Pr i : Where: k i represents the batches sampled from the i-th supplier within a certain historical period; u i represents the total number of qualified materials sampled from the i-th supplier; v i Represents the total number of unqualified materials sampled from the i-th supplier; m i Represents the total number of samples inspected in the historical arrivals of the i-th supplier; For the jth batch of sampled materials (j=1, 2, ..., k i ), when the sampling result is "qualified", let u ij =m ij , v ij =0, recorded as (m ij , 0); when the sampling result is "unqualified", let u ij =0, v ij =m ij , where u ij represents the number of qualified products of the jth batch of sampled materials from the i-th supplier, v ij represents the number of unqualified products in the jth batch of sampled materials from the i-th supplier; p i represents the original qualified rate of the i-th supplier; Pr i represents the corrected qualified rate of the i-th supplier after considering the confidence level; 2 α Represents the quantile of the normal distribution.

3. The method for formulating a power grid material sampling strategy based on multi-dimensional analysis according to claim 1 is characterized in that: Supplier quality level evaluation value Pr i * Obtained according to the following formula: Where: Pr i represents the comprehensive basic qualification rate of the ith supplier, Pr i ∈[0, 1], Pr i =APr i 1 +BPr i 2 +CPr i 3 , Pr i 1 , Pr i 2 , Pr i 3 The i-th supplier sets the sampling pass rate within the cycle, A, B, and C are Pr after considering the difference in long-term and short-term impact. i 1 , Pr i 2 , Pr i 3 The weight of P i represents the most recent cumulative score in the bad behavior record of the i-th supplier, P i ∈[0, 10]; P i ′ is P i After normalization, the data P i ′∈[0,1];I i represents the quality improvement index of the i-th supplier, I i =R2-R1 (Formula 4), R1, R2 are respectively When there is no data in any time period, let I i =0.

4. The method for formulating a power grid material sampling inspection strategy based on multi-dimensional analysis according to claim 1 is characterized in that In step S3, constructing a maximum quality evaluation ranking consistency model specifically includes the following steps: S3-1: Setting sequence Set up supplier quality level evaluation sequence Sort the supplier quality level evaluation values ​​from high to low to meet set up represents a set of sampling strategies for supplier i, and the actual sampling ratio of the supplier corresponding to the strategy (Sr i ∈(0,1]), under the premise of "less inspection for high quality and more inspection for low quality", there exists a sampling ratio sequence Sr I = {Sr1, ..., Sr i ,.....Sr I }, satisfying Sr1≤...≤Sr i ≤…≤Sr I ; S3-2: Determine the benchmark sampling ratio Assume Sr i ′ is the benchmark sampling ratio for each supplier, Sr i ′ is determined by the interval mapping function according to the quality level evaluation value of each supplier and the minimum and maximum thresholds of the sampling ratio, satisfying Sr′1≤…≤Sr′ i ≤…≤Sr′ I ; For the quality level evaluation value, the quality level evaluation interval can be obtained When the minimum sampling ratio of all suppliers is a and the maximum sampling ratio is b, the sampling ratio benchmark range [a, b] can be obtained. A mapping relationship is established based on the sampling ratio benchmark range interval and the quality level evaluation interval: According to (Formula 6), (Formula 7) is derived: S3-3: Establishing the objective function of minimizing the deviation distance Calculate Sr i The actual value and the given benchmark sampling ratio Sr i The distance d between i , d i =Sr i -Sr i ′, d is the total deviation distance of all suppliers, then there is an objective function that satisfies the consistency of quality evaluation ranking: Among them, the ratio of the difference between the actual sampling ratio of adjacent ranking suppliers and the interpolation of the benchmark ratio is close to [0.9, 1], that is, 5. The method for formulating a power grid material sampling inspection strategy based on multi-dimensional analysis according to claim 1 is characterized in that: The model for minimizing the total annual inspection cost is obtained by the following formula: Among them, W 1i Costs incurred for receiving materials, W 2i Costs incurred for rejecting materials, Pr i is the sampling pass rate, N i总 The supplier's total annual arrival volume and sampling base and the corresponding sample size 6. The method for formulating a power grid material sampling strategy based on multi-dimensional analysis according to claim 1 is characterized in that: The intelligent solver solves the model for minimizing the total annual inspection cost by the following steps: S5-1: Determine the sample size using the GB / T2828.1-2012 normal inspection single sampling plan retrieval table The batch size is determined based on the sample size and the cumulative arrival volume of each supplier throughout the year. The value range of The combination value retrieval table of S5-2: Based on The value range and correspondence of Generate multiple sampling strategies using intelligent solvers And calculate the sampling ratio Sr under each sampling strategy i , S5-3: Set the sampling ratio Sr under each sampling strategy i Construct a decreasing sequence of quality evaluation in descending order from high to low, and construct an increasing sequence of sampling ratio according to the increasing order of sampling ratio. Determine the benchmark sampling ratio by presetting the sampling ratio range [a, b]. Generate multiple sets of sampling ratio sequences Sr I , Sr. I = {Sr1, Sr2, ..., Sr i , ..., Sr I }, based on the consistency rule of quality level evaluation ranking, the deviation distance d between the actual sampling ratio and the benchmark sampling ratio is calculated by the following formula: d i =Sr i -Sr′ i MIN(d) Find the minimum deviation distance d and select the sampling ratio sequence that meets the inspection conditions as the optimal sequence solution S5-4: Calculation Corresponding sampling strategy sequence SIS I The sampling cost of the sampling strategy is selected as the optimal sampling strategy sequence with the minimum total cost.

7. A power grid material sampling strategy formulation system based on multi-dimensional analysis, characterized by: Includes the following modules: A data acquisition module, which is used to acquire data of all suppliers of power grid materials for which sampling inspection strategies are to be formulated, so as to be used in formulating sampling inspection strategies; The data analysis module, after the data module obtains the data, analyzes the supplier quality level evaluation value Pr based on the historical data. i * , and the benchmark sampling ratio Sr of supplier i determined based on the quality level evaluation results i ′; Strategy output module, the strategy output module includes an optimal sequence solution calculation module, a sampling strategy sequence generation module, and an optimal sampling strategy generation module; the optimal sequence solution calculation module uses the quality level value to establish a maximum quality evaluation ranking consistency model and obtain the optimal sequence solution The sampling strategy sequence generation module automatically generates a solution that satisfies the optimal sequence Multi-group sampling strategy SIS i ; The optimal sampling strategy generation module is based on the sampling strategy SIS i The sampling ratio in the algorithm is used to establish a model that minimizes the total annual sampling cost; and the genetic algorithm and the multi-objective White Whale optimization algorithm are used to build an intelligent solver. The intelligent solver solves the model that minimizes the total annual sampling cost and obtains the optimal sampling strategy.

8. The multi-dimensional grid material sampling strategy formulation system according to claim 7 is characterized in that: The strategy formulation module uses the quality level value to establish a model that maximizes the consistency of quality evaluation ranking, and the process includes: Get the supplier quality level evaluation sequence from high to low in Based on the supplier quality level evaluation value and the minimum and maximum thresholds of the sampling ratio, the supplier benchmark sampling ratio sequence is determined through the interval mapping function, and further based on the quality level evaluation interval The sampling ratio benchmark range is [a, b], where a is the supplier's minimum sampling ratio and b is the supplier's maximum sampling ratio. A mapping relationship is established: A mapping relationship is established based on the sampling ratio benchmark range interval and the quality level evaluation interval to obtain and output the maximum quality evaluation ranking consistency model: Costs incurred based on receiving materials W 1i , the cost of rejecting materials 2i , sampling inspection qualified rate Pr i 、The total amount of goods received by suppliers throughout the year N i总 , sampling base and the corresponding sample size Output minimizes the total annual sampling cost; Based on the sampling ratio sequence as the optimal sequence solution and total cost Minimum, intelligent solver solves the model of minimizing the total annual sampling cost and outputs the optimal sampling strategy sequence

Images