Method for classifying homogeneous geographic areas of electromechanical system component failure influence factors

Abstract

The invention discloses an electromechanical system component failure influence factor homogenized geographic region classification method, which comprises the following steps of: identifying failure influence factors of electromechanical system components; determining space granularity and subspace granularity which need to be subjected to geographic region classification; building a failure influence factor panel data model considering a data generation time weight; applying the proposed method to clustering analysis of the failure influence factors of the photovoltaic module; and the effectiveness and practicability of the method are verified by comparing and analyzing clustering results. By carrying out failure influence factor geographic area homogenization classification on the components of the electromechanical system, the requirement of precise evaluation of the field reliability is met, and a basis is provided for analysis and design of the electromechanical system facing the field reliability and the components of the electromechanical system.

Description

Technical Field

[0001] The invention relates to cluster analysis of time series data of failure influencing factors, which is applicable to on-site reliability assessment of electromechanical system components, and in particular to a classification method for homogeneous geographical areas of failure influencing factors of electromechanical system components. Background Art

[0002] In the actual application of complex electromechanical systems such as photovoltaic air-conditioning systems, photovoltaic power generation systems and air-conditioning systems, the field reliability of their components is affected by multiple factors, namely failure influencing factors. These failure influencing factors are not only coupled with each other, but also change with the change of service geographical area, resulting in significant geographical regional differences in the field reliability of components and systems.

[0003] Traditional component field reliability analysis and evaluation methods are usually based on test results under standard experimental conditions or data obtained from on-site operation and maintenance in a specific geographical area, ignoring the geographical differences in failure influencing factors. This causes a considerable difference between the evaluation results and the on-site reliability level of the components, greatly affecting the effectiveness of component and system design and operation and maintenance decisions based on on-site reliability.

[0004] Therefore, to accurately analyze and evaluate the field reliability of components, the geographical differences in failure influencing factors must be considered. By dividing geographical regions with consistent failure influencing factors into homogeneous categories and conducting field reliability analysis and evaluation according to the geographical region categories with homogeneous failure influencing factors, the impact of geographical differences in failure influencing factors on field reliability analysis and evaluation results can be effectively controlled. Summary of the invention

[0005] In view of the problems existing in the on-site reliability analysis and evaluation of the above-mentioned complex electromechanical system components, the present invention proposes a classification method for homogeneous geographical areas of electromechanical system component failure influencing factors, obtains the coarse classification results of provincial administrative divisions in geographical areas with homogeneous component failure influencing factors and the detailed classification results of prefecture-level administrative divisions based on the coarse classification results, thereby laying a foundation for the accurate analysis and evaluation of the on-site reliability of components and systems.

[0006] The purpose of the present invention is achieved by adopting the following technical solutions: A classification method for homogeneous geographical areas of factors affecting failure of electromechanical system components, the steps of which are as follows:

[0007] 1) Identify the failure factors of electromechanical system components:

[0008] 1.1 Identify the various failure modes (FM) of electromechanical system components from the three dimensions of function, performance and structure. τ (τ=1,2,…,p);

[0009] 1.2 Analyze all the influencing factors that cause each failure mode to occur, and obtain the set of influencing factors of the failure mode

[0010] 1.3 Comprehensively consider the set of influencing factors of each failure mode and obtain the set of influencing factors of failure of electromechanical system components {z1 z2 … z ω … z s}(ω=1,2,…,s);

[0011] 2) Determine the spatial granularity and its sub-spatial granularity for geographical area classification, and quantitatively describe the failure influencing factors of each spatial granularity and its sub-spatial granularity:

[0012] 2.1 The provincial administrative division unit is used as the spatial granularity for geographical area classification, and the prefecture-level administrative division unit included in the spatial granularity is used as the sub-spatial granularity;

[0013] 2.2 The time granularity of failure influencing factors is measured in months. The average value of the ωth failure influencing factor at the tth time granularity of the vth subspace granularity in the jth year under the i-th space granularity is

[0014] 2.3 The average value of the ωth failure influencing factor at the tth time granularity in the jth year of the i-th spatial granularity is The calculation formula is as follows:

[0015]

[0016] Where: N i represents the number of subspace granularity v under the i-th space granularity;

[0017] 3) Construct a panel data model of failure influencing factors that takes into account the weight of data generation time:

[0018] 3.1 The average values ​​of failure influencing factors at the subspace granularity and space granularity constitute time series data according to the order of the time granularity t of data generation. The importance of the later generated data is greater than that of the earlier generated data;

[0019] 3.2 The importance of the failure influencing factor data generated at the t-th time granularity in the j-th year within a total period T of φ years is its weight, and the calculation formula is as follows:

[0020]

[0021] Where: θ jis the importance of the failure influencing factor data generated in the jth year, and its calculation formula is as follows:

[0022]

[0023] 3.3 Comprehensively consider the time series data of failure influencing factors and their weights, and construct a spatial granular failure influencing factor panel data model with homogeneous failure influencing factors:

[0024]

[0025] Where: N is the number of spatial granularity;

[0026] 3.4 Panel data model of failure influencing factors at subspace granularity within the spatial granularity of homogeneous failure influencing factors:

[0027]

[0028] Where: N k is the number of subspace granularities within the space granularity k;

[0029] 4) Spatial granularity of failure influencing factors and geographical region cluster analysis:

[0030] 4.1 Time series data of the panel data model for factors affecting spatial granularity failure After normalization, we can get the vector composed of the elements in the failure influencing factors:

[0031] Z i =[z i1 z i2 … z iω … z is ] (i=1,2,…,N);

[0032] 4.2 After clustering, the i-th spatial granularity becomes the i-th spatial granularity in category k, denoted as:

[0033] Z ki =[z ki1 z ki2 … z kiω …z kis ] (i=1,2,…,N);

[0034] The centroid of each category k and the centroid of the whole composed of all N spatial granularities for:

[0035]

[0036] Where: Θ kis the number of spatial granularities divided into categories k (k = 1, 2, ..., C), and

[0037] 4.3 Comprehensive consideration of time series data The geometric structure of the membership information and the validity function index XBI(C) and KI(C) values ​​determine the optimal number of categories C. * , the calculation formula is as follows:

[0038]

[0039] Where: is the membership factor, μ ki is the spatial granularity Z i Membership to category k:

[0040]

[0041] 4.5 Using the fuzzy C-means clustering algorithm for iterative calculation, the spatial granularity Z i (i=1,2,…,N) to determine the specific category k (k=1,2,…,C * );

[0042] 4.6 For isolated spatial granularity Z α (α∈[1,2,…,N]), then it is classified into the adjacent category β(β=1,2,…,C β )center The Mahalanobis distance D αβ The smallest category, Mahalanobis distance D αβ The calculation formula is as follows:

[0043]

[0044] Where: Z α and The covariance of

[0045] 4.7 The geographical area composed of all spatial granularities of the same category is the coarse geographical area with homogeneous failure influencing factors;

[0046] 5) Cluster analysis of failure influencing factors at the subspace granularity by geographical region:

[0047] 5.1 Statistical Spatial Granularity Failure Factors Specific Categories of Results from the Cluster Analysis of Geographical Regions k (k = 1, 2, ..., C * ) contains the number of subspace granularities N k ;

[0048] 5.2 For each spatial granularity category, the subspace granularity N it contains k Repeat the cluster analysis from 4.1 to 4.6 to obtain the cluster analysis results of the subspace granularity of each spatial granularity category;

[0049] 5.3 The geographical area formed by all subspace granularities of the same category is the subdivided geographical area with homogeneous failure influencing factors;

[0050] 6) Application and verification of the method:

[0051] 6.1 Apply the proposed method to cluster analysis of factors affecting photovoltaic module failure;

[0052] 6.2 By comparing and analyzing the clustering results, the effectiveness and practicality of the method are verified.

[0053] The present invention solves the need for accurate on-site reliability assessment by classifying the influencing factors of component failure of electromechanical systems into homogeneous geographical regions, provides a basis for the analysis and design of electromechanical systems and their components for on-site reliability, and can also provide accurate geographical region classification results for the analysis and assessment prediction of on-site reliability of electromechanical system components. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Weight distribution of time series data of failure influencing factors;

[0055] Figure 2 Fuzzy C-means clustering calculation process. DETAILED DESCRIPTION

[0056] The present invention is further described below with reference to the accompanying drawings and examples. Figure 1 to Figure 2 ,The classification method of homogeneous geographical areas of factors affecting the failure of electromechanical system components, and its specific implementation steps are as follows:

[0057] Step 1: Identify the factors affecting the failure of photovoltaic modules:

[0058] The various failure forms, i.e., failure modes, of photovoltaic modules are identified dimensionally from the three dimensions of function, performance, and structure. The results are shown in Table 1.

[0059] Table 1 Field failure modes of photovoltaic modules

[0060]

[0061] All influencing factors that lead to the failure mode are analyzed one by one to obtain the set of influencing factors of the failure mode. The set of influencing factors of each failure mode is combined to obtain the set of failure influencing factors (temperature, temperature difference, humidity, sunshine, solar radiation, wind speed, precipitation, dust and SO2 / NO2 concentration) of photovoltaic modules (as shown in Table 2).

[0062] Table 2 Failure modes and their corresponding failure influencing factors

[0063] Failure influencing factors Failure Mode temperature Hot spots and oxidation of battery cells, yellowing, browning and delamination of EVA packaging materials Temperature difference The battery cell is cracked, the surface glass is broken, the back panel is cracked, and the EVA packaging material is delaminated. humidity Cell corrosion, EVA packaging material hydrolysis, electrical connector electrode corrosion sunshine Hot spots on the cell, output power below the threshold; yellowing and browning of EVA packaging materials Solar Radiation EVA packaging materials turn yellow and brown Wind speed Back panel cracking, surface glass cracking, battery cell corrosion, electrical connector electrode corrosion precipitation EVA packaging material hydrolysis, battery cell corrosion, electrical connector electrode corrosion dust Electrical connector contact failure; battery cell hot spot, output power below the threshold <![CDATA[SO2 / NO2 concentration]]> Cell corrosion, electrical connector electrode corrosion, backplane thinning

[0064] Step 2: Determine the spatial granularity and its sub-spatial granularity for geographical area classification, and quantitatively describe the failure influencing factors of each spatial granularity and its sub-spatial granularity:

[0065] Conduct homogeneous geographical classification of factors affecting PV module failure. The provincial administrative division unit is the spatial granularity for geographical classification, and the prefecture-level administrative division unit included in the spatial granularity is the sub-spatial granularity.

[0066] Taking 9 years and 108 months from 2014 to 2022 as the time span, the time granularity of failure influencing factors is measured in months. The monthly average values ​​of each failure influence at each time granularity of each sub-space granularity under each space granularity can be collected as follows:

[0067] (1) The monthly average data of failure influencing factors such as temperature, temperature difference, humidity, sunshine, precipitation and wind speed can be analyzed and extracted from the China Surface Climate Data Monthly Value Dataset released by the China Meteorological Science Data Sharing Service Network;

[0068] (2) The monthly average data of dust failure influencing factors can be analyzed and extracted from the dust monthly value data set released by the China National Environmental Monitoring Center. The monthly average data of SO2 / NO2 failure influencing factors can be obtained from the monthly value data set released by the China Air Quality Online Monitoring and Analysis Platform.

[0069] The monthly average value of the failure influencing factor data of all sub-space granularities at each time granularity under each space granularity is the data of the failure influencing factor at each time granularity of each space granularity, as shown in Table 3.

[0070] Table 3 Time series data of factors affecting photovoltaic module failure at spatial granularity

[0071]

[0072]

[0073] Step 3: Construct a panel data model of failure influencing factors that takes into account the weight of data generation time:

[0074] The average values ​​of failure influencing factors at the subspace granularity and spatial granularity constitute the time series data of the subspace granularity and spatial granularity respectively according to the order of the time granularity of data generation. As an example, Table 2 gives the time series data of photovoltaic module failure influencing factors at the spatial granularity of some provincial administrative units.

[0075] Table 4 and Figure 1 The weights of the failure influencing factor data generated at each time granularity are given, and the failure influencing factor data generated at each time granularity in the same year have the same weight.

[0076] Table 4 Data weights of failure influencing factors generated at each time granularity

[0077]

[0078] By comprehensively considering the time series data of failure influencing factors and their weights, we can obtain the panel data model of failure influencing factors at a spatial granularity and the panel data model of failure influencing factors at a sub-space granularity within a geographical area category with homogeneous failure influencing factors composed of several spatial granularities.

[0079] Step 4: Spatial granularity cluster analysis of failure influencing factors in geographical areas:

[0080] Firstly, the time series data of the panel data model of spatial granularity failure influencing factors are normalized to obtain the vector composed of each element in the failure influencing factor set; then the centroid of each category and the centroid of the whole composed of all spatial granularities are calculated respectively; then the optimal number of categories is determined to be VIII by comprehensively considering the effectiveness function indicators XBI(C) and KI(C); finally, the Figure 2 The fuzzy C-means clustering algorithm shown in FIG. 1 classifies the spatial granularity. The steps of the fuzzy C-means clustering algorithm are as follows:

[0081] Step 1: Set the iterative calculation termination accuracy ε and membership factor and the number of clusters C*, t=0;

[0082] Step 2: Initialize the membership matrix and cluster centroid

[0083] Step 3: Update the membership matrix and cluster centroid

[0084] Step 4: If the iterative calculation termination accuracy requirement is met, the iterative calculation ends; otherwise, t = t + 1, go to step 4 to continue the iteration;

[0085] Step 5: Spatial Granularity Z i (i=1,2,…N) to determine the specific cluster k (k=1,2,…C* );

[0086] Step 6: Isolate spatial granularity Z one by one α (α=1,2,…N) calculate the Mahalanobis distance between it and the center of the adjacent category;

[0087] Step 7: Determine the isolated space granularity Z one by one α The category to which (α=1,2,…N) belongs.

[0088] The homogeneous and roughly classified geographical regions of failure influencing factors are obtained as shown in Table 5.

[0089] Table 5 Homogeneity of factors affecting PV module failure by geographic region (Taiwan, Hong Kong and Macau are not included in this statistics)

[0090] Rough classification Classification results I Beijing, Tianjin, Hebei, Shanxi II Inner Mongolia, Gansu, Ningxia, Xinjiang III Heilongjiang, Jilin, Liaoning IV Shandong, Henan, Shaanxi V Shanghai, Jiangsu, Anhui, Zhejiang, Hubei VI Qinghai, Tibet VII Sichuan, Chongqing, Guizhou, Hunan VIII Guangdong, Guangxi, Hainan, Fujian, Jiangxi, Yunnan

[0091] Step 5: Cluster analysis of failure influencing factors at the subspace granularity by geographic region:

[0092] Taking Category V, which is composed of the spatial granularity of five provincial administrative units including Shanghai, Jiangsu, Anhui, Zhejiang and Hubei in the coarse classification of spatial granularity, as an example, a geographical regional clustering analysis of failure influencing factors at the sub-spatial granularity was conducted.

[0093] Taking into account the land area of ​​Shanghai, it is considered as a subspace granularity, so the number of subspace granularities included in this category is 58 in total.

[0094] By repeating the cluster analysis process at the spatial granularity, the homogeneous subdivision of the geographical area categories of the failure influencing factors of Category V can be obtained as shown in Table 6.

[0095] Table 6. Homogeneous breakdown of factors affecting failure of PV modules in Category V regions by geographical region category

[0096] Segment Categories Clustering results I Chuzhou, Hefei, Huainan, Lu'an, Ma'anshan, Nanjing, Taizhou, Yancheng, Yangzhou, Zhenjiang II Jingmen, Jingzhou, Qianjiang, Suizhou, Tianmen, Xiantao, Xiangyang, Yichang III Changzhou, Huzhou, Jiaxing, Nantong, Shanghai, Suzhou, Wuxi IV Anqing, Chizhou, Hangzhou, Tongling, Wuhu, Xuancheng V Ezhou, Huanggang, Huangshi, Wuhan, Xianning, Xiaogan VI Huangshan, Jinhua, Lishui, Quzhou, Shaoxing VII Haozhou, Fuyang, Huaibei, Suzhou, Xuzhou VIII Ningbo, Taizhou, Wenzhou, Zhoushan IX Bengbu, Huai'an, Lianyungang, Suqian X Enshi, Shennongjia Forest Area, Shiyan

[0097] Step 6: Application and verification of the method:

[0098] According to the application example of the classification method of homogeneous geographical regions of failure influencing factors of electromechanical system components in photovoltaic power generation system photovoltaic components, the proposed method can systematically identify and quantitatively describe the failure influencing factors of components, and obtain the classification results of geographical regions with consistent component failure influencing factors through the coarse geographical region clustering of homogeneous spatial granularity failure influencing factors and the sub-geographical region clustering of geographical homogeneity of sub-space granularity failure influencing factors based on the coarse clustering results. The method is general and can provide technical support for the analysis, design and operation and maintenance decision-making of field reliability components and systems.

Claims

1. A classification method for homogeneous geographical areas of factors affecting failure of electromechanical system components, characterized by the following steps: 1) Identify the failure factors of electromechanical system components: 1.1 Identify the various failure modes (FM) of electromechanical system components from the three dimensions of function, performance and structure. τ (τ=1,2,…,p); 1.2 Analyze all the influencing factors that cause each failure mode to occur, and obtain the set of influencing factors of the failure mode 1.3 Comprehensively consider the set of influencing factors of each failure mode and obtain the set of influencing factors of failure of electromechanical system components {z1 z2… z ω … z s }(ω=1,2,…,s); 2) Determine the spatial granularity and its sub-spatial granularity for geographical area classification, and quantitatively describe the failure influencing factors of each spatial granularity and its sub-spatial granularity: 2.1 The provincial administrative division unit is used as the spatial granularity for geographical area classification, and the prefecture-level administrative division unit included in the spatial granularity is used as the sub-spatial granularity; 2.2 The time granularity of failure influencing factors is measured in months. The average value of the ωth failure influencing factor at the tth time granularity of the vth subspace granularity in the jth year under the i-th space granularity is 2.3 The average value of the ωth failure influencing factor at the tth time granularity in the jth year of the i-th spatial granularity is The calculation formula is as follows: Where: N i represents the number of subspace granularity v under the i-th space granularity; 3) Construct a panel data model of failure influencing factors that takes into account the weight of data generation time: 3.1 The average values ​​of failure influencing factors at the subspace granularity and space granularity constitute time series data according to the order of the time granularity t of data generation. The importance of the later generated data is greater than that of the earlier generated data; 3.2 The importance of the failure influencing factor data generated at the t-th time granularity in the j-th year within a total period T of φ years is its weight, and the calculation formula is as follows: Where: θ j is the importance of the failure influencing factor data generated in the jth year, and its calculation formula is as follows: 3.3 Comprehensively consider the time series data of failure influencing factors and their weights, and construct a spatial granular failure influencing factor panel data model with homogeneous failure influencing factors: Where: N is the number of spatial granularity; 3.4 Panel data model of failure influencing factors at subspace granularity within the spatial granularity of homogeneous failure influencing factors: Where: N k is the number of subspace granularities within the space granularity k; 4) Spatial granularity of failure influencing factors and geographical region cluster analysis: 4.1 Time series data of the panel data model for factors affecting spatial granularity failure After normalization, we can get the vector composed of the elements in the failure influencing factors: WITH i =[of i1 With i2 … With iω … With is ]; 4.2 After clustering, the i-th spatial granularity becomes the i-th spatial granularity in category k, denoted as: WITH ki =[of ki1 With ki2 … With kiω … With kis ] (i=1,2,…,N); The centroid of each category k and the centroid of the whole composed of all N spatial granularities for: Where: Θ k is the number of spatial granularities divided into categories k (k = 1, 2, ..., C), and 4.3 Comprehensive consideration of time series data The geometric structure of the membership information and the validity function index XBI(C) and KI(C) values ​​determine the optimal number of categories C. * , the calculation formula is as follows: Where: is the membership factor, μ ki is the spatial granularity Z i Membership to category k: 4.5 Using the fuzzy C-means clustering algorithm for iterative calculation, the spatial granularity Z i (i=1,2,…,N) to determine the specific category k (k=1,2,…,C * ); 4.6 For isolated spatial granularity Z α (α∈[1,2,…,N]), then it is classified into the adjacent category β(β=1,2,…,C β )center The Mahalanobis distance D αβ The smallest category, Mahalanobis distance D αβ The calculation formula is as follows: Where: Z α and The covariance of 4.7 The geographical area composed of all spatial granularities of the same category is the coarse geographical area with homogeneous failure influencing factors; 5) Cluster analysis of failure influencing factors at the subspace granularity by geographical region: 5.1 Statistical Spatial Granularity Failure Factors Specific Categories of Results from the Cluster Analysis of Geographical Regions k (k = 1, 2, ..., C * ) contains the number of subspace granularities N k ; 5.2 For each spatial granularity category, the subspace granularity N it contains k Repeat the cluster analysis from 4.1 to 4.6 to obtain the cluster analysis results of the subspace granularity of each spatial granularity category; 5.3 The geographical area formed by all subspace granularities of the same category is the subdivided geographical area with homogeneous failure influencing factors; 6) Application and verification of the method: 6.1 Apply the proposed method to cluster analysis of factors affecting photovoltaic module failure; 6.2 By comparing and analyzing the clustering results, the effectiveness and practicality of the method are verified.

Images