CN109840541A - A kind of network transformer Fault Classification based on XGBoost - Google Patents
A kind of network transformer Fault Classification based on XGBoost Download PDFInfo
- Publication number
- CN109840541A CN109840541A CN201811482922.5A CN201811482922A CN109840541A CN 109840541 A CN109840541 A CN 109840541A CN 201811482922 A CN201811482922 A CN 201811482922A CN 109840541 A CN109840541 A CN 109840541A
- Authority
- CN
- China
- Prior art keywords
- xgboost
- data set
- data
- transformer fault
- dga
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The present invention provides a kind of network transformer Fault Classification based on XGBoost.The network transformer Fault Classification based on XGBoost includes the following steps: step 1: obtaining and integrate the DGA data set of multiple transformers；Step 2: the data after being normalized by the data in DGA data set that step 1 obtains and will be pretreated are given XGBoost and are trained, a certain number of post-class processings are constructed to be fitted the preceding residual error once learnt, and optimal XGBoost parameter combination is found by grid search, to improve the diagnosis accuracy to transformer fault.
Description
Technical field
The invention belongs to transformer fault diagnosis technical fields, more particularly to a kind of power grid transformation based on XGBoost
Device Fault Classification.
Background technique
Power transformers (power transformer) be used to by electrical power conversion that power plant generates be transferred to the whole world
User (customer) in factory/company.Power transformer is one of key equipment of electric system, its operating status is very
Determine that can power grid good work in big degree.Thereby it is ensured that the good operating status for becoming power transformer is very heavy
It wants, such transformer could provide reliable and lasting electric power, this is necessary in real world.Currently, being permitted
More Utilities Electric Co.s implement various status assessments and maintenance measure, dissolved gas analysis to the state of transformer
(DGA) be exactly it is one such, DGA is the key that a kind of concentration based on dissolved gas in transformer insulation oil and gas generate
The method that rate is detected and predicted to the failure of transformer, such as key gas, IEC ratio, Rogers ratio and
Dornenburg ratio method.
DGA method can make some estimations to the operating status of transformer, unfortunately, although they operate letter
It is single, but exist and encode the problems such as incomplete, boundary is excessively absolute, these methods often provide different prediction results, say
These bright methods be it is highly inaccurate, will lead to many failures correctly can not timely be found, also can cause to be stranded to tester
It is difficult.These problems have driven many researchers to go to study the method based on machine learning to diagnose to transformer fault.
With the development of intelligent algorithm and machine learning algorithm, support vector machines, neural network, post-class processing, principal component point
The technologies such as analysis are all gradually applied to the fault diagnosis of transformer and achieve certain achievement.But since transformer data are past
Toward be very difficult to obtain and it is very rare, although these intelligent algorithms can obtain fairly good classification results, often
Over-fitting can be fallen into.For example, although support vector machines is good at handling Small Sample Database, but the essence of two classifiers due to it,
Keep its efficiency in the problem of processing more classification (classification of transformer is more classification problems, includes various faults type) lower.
And neural network is when handling Small Sample Database, although thering is very strong learning ability to be easily trapped into local optimum and leading
Cause over-fitting.Post-class processing is although high-efficient, but the learning ability of single tree is too weak and is also easily trapped into over-fitting and asks
Topic.
Summary of the invention
It is an object of the invention in view of the drawbacks of the prior art or problem, provide a kind of power grid based on XGBoost to become
Depressor Fault Classification.
Technical scheme is as follows: a kind of network transformer Fault Classification based on XGBoost includes as follows
Step:
Step 1: obtaining and integrate the DGA data set of multiple transformers；
Step 2: the number after being normalized by the data in DGA data set that step 1 obtains and will be pretreated
It is trained according to XGBoost is given, constructs a certain number of post-class processings and the preceding residual error once learnt is fitted, and lead to
It crosses grid search and finds optimal XGBoost parameter combination, to improve the diagnosis accuracy to transformer fault.
Preferably, in step 2, XGBoost model is specifically included:
Input: input can be expressed as D by the transformer data acquisition system after preliminary screening and normalization, data set
={ X1,X2,X3,…,Xd, the number of sample, X in d data seti={ x1,x2,…,xn, y } and indicate each sample data, xi
Indicate the feature of each dimension, 8 kinds of fault types of y ∈ { 0,1,2,3,4,5,6,7 } indication transformer；
Output: setting XGBoost selects softmax as target, returns to the classification of prediction；
Loss function: XGBoost requires loss function to be can dimpling loss function.
Technical solution provided by the invention has the following beneficial effects:
The network transformer Fault Classification based on XGBoost incorporates DGA gas data from different sources,
And the ratio for combining gas with various forms the relatively large DGA data set of quantity；
Moreover, the fault diagnosis and classification of integrated study (XGBoost) for transformer can permit transformer
There are a small amount of missing values (at this moment very common, due to the severe running environment of transformer) in DGA data, pass through
The integrated study ability building of XGBoost largely can handle high inclination and the classification of polymorphic continuous data type data returns
Tree, and optimal model is obtained by grid search；
It is trained in addition, giving the DGA data set obtained after pretreatment to XGBoost, constructs a large amount of classification and return
Tree, is constantly fitted previous prediction result to improve the diagnosis accuracy to transformer fault.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to embodiments, to the present invention
It is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to
Limit the present invention.
The description of specific distinct unless the context otherwise, the present invention in element and component, the shape that quantity both can be single
Formula exists, and form that can also be multiple exists, and the present invention is defined not to this.Although step in the present invention with label into
It has gone arrangement, but is not used to limit the precedence of step, unless expressly stated the order of step or holding for certain step
Based on row needs other steps, otherwise the relative rank of step is adjustable.It is appreciated that used herein
Term "and/or" one of is related to and covers associated listed item or one or more of any and all possible groups
It closes.
A kind of network transformer Fault Classification based on XGBoost provided by the invention, includes the following steps:
Step 1: obtaining and integrate the DGA data set of multiple transformers；
Step 2: the number after being normalized by the data in DGA data set that step 1 obtains and will be pretreated
It is trained according to XGBoost is given, constructs a certain number of post-class processings and the preceding residual error once learnt is fitted, and lead to
It crosses grid search and finds optimal XGBoost parameter combination, to improve the diagnosis accuracy to transformer fault.
Specifically, extreme gradient promotes (Extreme Gradient Boosting, XGBoost), is expanding for tree promotion
Machine learning system is opened up, is a kind of novel classification device based on post-class processing (CART) set.
It is parallel in XGBoost supported feature granularity, XGBoost before training, can to being ranked up in advance to data,
Then block structure is saved as, iteration later is recursive to use this structure, and greatly reduces calculation amount.
There is the data set of n sample m dimensional feature for one
XGBoost is a kind of tree aggregation model, and the output of target is predicted using K function superposition.
It is the set of post-class processing (CART) composition,
Q is sample data xiTo the mapping of the leaf node of CART tree, for indicating the structure of one tree, T indicates the leaf in one tree
The number of child node.Each fkIt is equivalent to the weight (score) of q a mapping and its leaf node, this weight is a company
Continuous value and help to realize efficient optimization algorithm.wiIndicate the weight of i-th of leaf node.
Final XGBoost model in order to obtain needs to train a series of post-class processing, this integrated study model
Objective function definition as shown in (2).L indicate can dimpling loss function, this loss function is for measuring true tag y and pre-
Survey label valueBetween gap.Ω is a regular terms, and the regular terms set by K is superimposed to obtain, and is arrived for smooth last study
Weight, punish the complexity of model, prevent over-fitting.Such objective function meeting final choice goes out one by a series of
The model of relatively simple anticipation function composition, has stronger generalization ability.If regular terms is set as 0, (2) in fact
It is traditional gradient boosted tree.
Integrated model in formula (2) includes function as parameter, therefore it is excellent not to be available traditional optimization method progress
Change, this model is trained in an iterative manner.IfIt is the prediction result of i-th of sample of t iteration, objective function
As shown in (3), ftRepresent the t times iteration creation new tree, selected by formula (3) most can lift scheme ft, pass through ft
To be fitted the prediction result of last iteration and the residual error of true value.
During grad enhancement, XGBoost carrys out optimization object function using the second Taylor series, reaches simplest shape
Shown in formula such as formula (4), whereinandThey are loss respectively
The single order and second dervative of function.Ij=i | q (xi)=j } indicate leaf node j sample number.
When the structure q (x) of one tree is given, the optimal weights of leaf node calculate by formula (5)
It arrives, the quality of tree construction q (x) can be calculated by formula (6).
It can not usually go to enumerate all possible tree construction, the method for use is to pass through iteration since root node
Mode branch is added to tree, it is assumed that ILAnd IRIt is the sample set of left subtree and right subtree after dividing, I=I respectivelyL∪IR,
Formula (7) be used to assess candidate split vertexes.
In embodiments of the present invention, the transformer fault prediction model based on XGBoost is by the change after feature selecting
Depressor data are input into XGBoost, and XGBoost passes through the approximation in split point finding algorithm according to the data after pretreated
Algorithm establishes first CART tree, is predicted according to this tree sample, predicted value and true value are compared, obtained residual
Residual error is simultaneously constructed next CART regression tree as new label information and sample data together and carrys out regression criterion by difference.Therefore,
One tree is added every time just can be such that the value of loss function constantly reduces.
Specifically, XGBoost model specifically includes:
Input: input can be expressed as D by the transformer data acquisition system after preliminary screening and normalization, data set
={ X1,X2,X3,…,Xd, the number of sample, X in d data seti={ x1,x2,…,xn, y } and indicate each sample data, xi
Indicate the feature of each dimension, 8 kinds of fault types of y ∈ { 0,1,2,3,4,5,6,7 } indication transformer；
Output: setting XGBoost selects softmax as target, returns to the classification of prediction；
Loss function: XGBoost requires loss function to be can dimpling loss function；Using merror, (more classification are wrong
Accidentally rate) and mlogloss (the negative log-likelihood for being defined as the true tag of given probability classification prediction).
It is obvious to a person skilled in the art that invention is not limited to the details of the above exemplary embodiments, Er Qie
In the case where without departing substantially from spirit or essential attributes of the invention, the present invention can be realized in other specific forms.Therefore, no matter
From the point of view of which point, the present embodiments are to be considered as illustrative and not restrictive, and the scope of the present invention is by appended power
Benefit requires rather than above description limits, it is intended that all by what is fallen within the meaning and scope of the equivalent elements of the claims
Variation is included within the present invention.
In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped
Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should
It considers the specification as a whole, the technical solutions in the various embodiments may also be suitably combined, forms those skilled in the art
The other embodiments being understood that.
Claims (3)
1. a kind of network transformer Fault Classification based on XGBoost, characterized by the following steps:
Step 1: obtaining and integrate the DGA data set of multiple transformers；
Step 2: the data after being normalized by the data in DGA data set that step 1 obtains and will be pretreated are handed over
It is trained to XGBoost, constructs a certain number of post-class processings and the preceding residual error once learnt is fitted, and pass through net
Optimal XGBoost parameter combination is found in lattice search, to improve the diagnosis accuracy to transformer fault.
2. a kind of network transformer Fault Classification based on XGBoost according to claim 1, which is characterized in that
XGBoost model specifically includes:
Input: input can be expressed as D=by the transformer data acquisition system after preliminary screening and normalization, data set
{X1,X2,X3,…,Xd, the number of sample, X in d data seti={ x1,x2,…,xn, y } and indicate each sample data, xiTable
Show the feature of each dimension, 8 kinds of fault types of y ∈ { 0,1,2,3,4,5,6,7 } indication transformer；
Output: setting XGBoost selects softmax as target, returns to the classification of prediction；
Loss function: XGBoost requires loss function to be can dimpling loss function.
3. a kind of network transformer Fault Classification based on XGBoost according to claim 1, which is characterized in that
In step 1, the ratio that joined gas with various in DGA data set enriches the feature quantity in DGA data set.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811482922.5A CN109840541A (en) | 2018-12-05 | 2018-12-05 | A kind of network transformer Fault Classification based on XGBoost |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811482922.5A CN109840541A (en) | 2018-12-05 | 2018-12-05 | A kind of network transformer Fault Classification based on XGBoost |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109840541A true CN109840541A (en) | 2019-06-04 |
Family
ID=66883169
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811482922.5A Pending CN109840541A (en) | 2018-12-05 | 2018-12-05 | A kind of network transformer Fault Classification based on XGBoost |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109840541A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298085A (en) * | 2019-06-11 | 2019-10-01 | 东南大学 | Analog-circuit fault diagnosis method based on XGBoost and random forests algorithm |
| CN110458360A (en) * | 2019-08-13 | 2019-11-15 | 腾讯科技（深圳）有限公司 | Prediction technique, device, equipment and the storage medium of hot resource |
| CN110781206A (en) * | 2019-12-02 | 2020-02-11 | 国网河北省电力有限公司电力科学研究院 | Method for predicting whether electric energy meter in operation fails or not by learning meter-dismantling and returning failure characteristic rule |
| CN110849617A (en) * | 2019-11-22 | 2020-02-28 | 深圳市通用互联科技有限责任公司 | Conveyor belt fault detection method and device, computer equipment and storage medium |
| CN110866366A (en) * | 2019-11-26 | 2020-03-06 | 南京工程学院 | XGboost algorithm-based island detection method for photovoltaic microgrid containing PHEV |
| CN110956010A (en) * | 2019-11-01 | 2020-04-03 | 国网辽宁省电力有限公司阜新供电公司 | Large-scale new energy access power grid stability identification method based on gradient lifting tree |
| CN110987439A (en) * | 2019-12-05 | 2020-04-10 | 山东超越数控电子股份有限公司 | Aeroengine fault prediction method based on Logitics regression and Xgboost model |
| CN111060755A (en) * | 2019-11-28 | 2020-04-24 | 北京济松科技有限公司 | Electromagnetic interference diagnosis method and device |
| CN111177375A (en) * | 2019-12-16 | 2020-05-19 | 医渡云（北京）技术有限公司 | Electronic document classification method and device |
| CN111612036A (en) * | 2020-04-20 | 2020-09-01 | 国网浙江省电力有限公司嘉兴供电公司 | Oil-immersed transformer fault diagnosis method based on particle swarm optimization XGboost |
| CN112183590A (en) * | 2020-09-14 | 2021-01-05 | 浙江大学 | Transformer fault diagnosis method based on Oneclass SVM algorithm |
| CN112507790A (en) * | 2020-11-03 | 2021-03-16 | 北方工业大学 | Fault diagnosis method and system of complementary classification regression tree based on difference evolution |
| CN112595918A (en) * | 2020-12-26 | 2021-04-02 | 广东电网有限责任公司广州供电局 | Low-voltage meter reading fault detection method and device |
| CN112782589A (en) * | 2021-01-26 | 2021-05-11 | 武汉理工大学 | Vehicle-mounted fuel cell remote fault classification diagnosis method and device and storage medium |
| CN112966425A (en) * | 2021-04-06 | 2021-06-15 | 昆明理工大学 | Slope stability prediction and evaluation method |
| CN113702728A (en) * | 2021-07-12 | 2021-11-26 | 广东工业大学 | Transformer fault diagnosis method and system based on combined sampling and LightGBM |
| CN114638384A (en) * | 2022-05-17 | 2022-06-17 | 四川观想科技股份有限公司 | Fault diagnosis method and system based on machine learning |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106067066A (en) * | 2016-05-31 | 2016-11-02 | 西安工程大学 | Method for diagnosing fault of power transformer based on genetic algorithm optimization pack algorithm |
| CN108551167A (en) * | 2018-04-25 | 2018-09-18 | 浙江大学 | A kind of electric power system transient stability method of discrimination based on XGBoost algorithms |
-
2018
- 2018-12-05 CN CN201811482922.5A patent/CN109840541A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106067066A (en) * | 2016-05-31 | 2016-11-02 | 西安工程大学 | Method for diagnosing fault of power transformer based on genetic algorithm optimization pack algorithm |
| CN108551167A (en) * | 2018-04-25 | 2018-09-18 | 浙江大学 | A kind of electric power system transient stability method of discrimination based on XGBoost algorithms |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298085A (en) * | 2019-06-11 | 2019-10-01 | 东南大学 | Analog-circuit fault diagnosis method based on XGBoost and random forests algorithm |
| CN110458360A (en) * | 2019-08-13 | 2019-11-15 | 腾讯科技（深圳）有限公司 | Prediction technique, device, equipment and the storage medium of hot resource |
| CN110458360B (en) * | 2019-08-13 | 2023-07-18 | 腾讯科技（深圳）有限公司 | Method, device, equipment and storage medium for predicting hot resources |
| CN110956010A (en) * | 2019-11-01 | 2020-04-03 | 国网辽宁省电力有限公司阜新供电公司 | Large-scale new energy access power grid stability identification method based on gradient lifting tree |
| CN110956010B (en) * | 2019-11-01 | 2023-04-18 | 国网辽宁省电力有限公司阜新供电公司 | Large-scale new energy access power grid stability identification method based on gradient lifting tree |
| CN110849617A (en) * | 2019-11-22 | 2020-02-28 | 深圳市通用互联科技有限责任公司 | Conveyor belt fault detection method and device, computer equipment and storage medium |
| CN110866366A (en) * | 2019-11-26 | 2020-03-06 | 南京工程学院 | XGboost algorithm-based island detection method for photovoltaic microgrid containing PHEV |
| CN111060755A (en) * | 2019-11-28 | 2020-04-24 | 北京济松科技有限公司 | Electromagnetic interference diagnosis method and device |
| CN110781206A (en) * | 2019-12-02 | 2020-02-11 | 国网河北省电力有限公司电力科学研究院 | Method for predicting whether electric energy meter in operation fails or not by learning meter-dismantling and returning failure characteristic rule |
| CN110987439A (en) * | 2019-12-05 | 2020-04-10 | 山东超越数控电子股份有限公司 | Aeroengine fault prediction method based on Logitics regression and Xgboost model |
| CN110987439B (en) * | 2019-12-05 | 2022-03-22 | 超越科技股份有限公司 | Aeroengine fault prediction method based on Logitics regression and Xgboost model |
| CN111177375A (en) * | 2019-12-16 | 2020-05-19 | 医渡云（北京）技术有限公司 | Electronic document classification method and device |
| CN111612036A (en) * | 2020-04-20 | 2020-09-01 | 国网浙江省电力有限公司嘉兴供电公司 | Oil-immersed transformer fault diagnosis method based on particle swarm optimization XGboost |
| CN112183590A (en) * | 2020-09-14 | 2021-01-05 | 浙江大学 | Transformer fault diagnosis method based on Oneclass SVM algorithm |
| CN112507790A (en) * | 2020-11-03 | 2021-03-16 | 北方工业大学 | Fault diagnosis method and system of complementary classification regression tree based on difference evolution |
| CN112507790B (en) * | 2020-11-03 | 2023-09-29 | 北方工业大学 | Fault diagnosis method and system of complementary classification regression tree based on differential evolution |
| CN112595918A (en) * | 2020-12-26 | 2021-04-02 | 广东电网有限责任公司广州供电局 | Low-voltage meter reading fault detection method and device |
| CN112782589A (en) * | 2021-01-26 | 2021-05-11 | 武汉理工大学 | Vehicle-mounted fuel cell remote fault classification diagnosis method and device and storage medium |
| CN112966425A (en) * | 2021-04-06 | 2021-06-15 | 昆明理工大学 | Slope stability prediction and evaluation method |
| CN112966425B (en) * | 2021-04-06 | 2022-06-07 | 昆明理工大学 | Slope stability prediction and evaluation method |
| CN113702728A (en) * | 2021-07-12 | 2021-11-26 | 广东工业大学 | Transformer fault diagnosis method and system based on combined sampling and LightGBM |
| CN114638384A (en) * | 2022-05-17 | 2022-06-17 | 四川观想科技股份有限公司 | Fault diagnosis method and system based on machine learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109840541A (en) | A kind of network transformer Fault Classification based on XGBoost | |
| CN110929847A (en) | Converter transformer fault diagnosis method based on deep convolutional neural network | |
| CN102521656A (en) | Integrated transfer learning method for classification of unbalance samples | |
| CN108229732A (en) | ExtremeLearningMachine wind speed ultra-short term prediction method based on error correction | |
| CN105701509B (en) | A kind of image classification method based on across classification migration Active Learning | |
| CN106022954A (en) | Multiple BP neural network load prediction method based on grey correlation degree | |
| CN113468817B (en) | Ultra-short-term wind power prediction method based on IGOA (insulated gate bipolar transistor) optimized ELM (ELM) | |
| CN109861211A (en) | A kind of power distribution network dynamic reconfiguration method based on data-driven | |
| CN107067033A (en) | The local route repair method of machine learning model | |
| CN110837915A (en) | Low-voltage load point prediction and probability prediction method for power system based on hybrid integrated deep learning | |
| CN104102951A (en) | Short-term wind power prediction method based on EMD (Empirical Mode Decomposition) historical data preprocessing | |
| CN106651199A (en) | Steam pipe network scheduling rule system based on decision-making tree method | |
| CN105894212A (en) | Comprehensive evaluation method for coupling and decoupling ring of electromagnetic ring network | |
| CN107818328A (en) | With reference to the deficiency of data similitude depicting method of local message | |
| CN114021433A (en) | Construction method and application of dominant instability mode recognition model of power system | |
| CN107808245A (en) | Based on the network scheduler system for improving traditional decision-tree | |
| CN109615246A (en) | A kind of active distribution network economical operation state determines method | |
| CN112508254A (en) | Method for determining investment prediction data of transformer substation engineering project | |
| Chi et al. | Application of BP neural network based on genetic algorithms optimization in prediction of postgraduate entrance examination | |
| CN110991723A (en) | Application method of artificial intelligence in seasonal load prediction | |
| CN106611181A (en) | Method for constructing cost-sensitive two-dimensional decision tree | |
| CN113496255B (en) | Power distribution network mixed observation point distribution method based on deep learning and decision tree driving | |
| CN110059737A (en) | Distribution transformer connection relationship discrimination method based on integrated deep neural network | |
| WO2021243930A1 (en) | Method for identifying composition of bus load, and machine-readable storage medium | |
| CN109886288A (en) | A kind of method for evaluating state and device for power transformer |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20190606 Address after: 11 004 No. 18 Ningbo Road, Shenyang City, Liaoning Province Applicant after: INFORMATION COMMUNICATION BRANCH, STATE GRID LIAONING ELECTRIC POWER Co.,Ltd. Applicant after: Nanjing University of Aeronautics and Astronautics Applicant after: STATE GRID CORPORATION OF CHINA Address before: 11 004 No. 18 Ningbo Road, Shenyang City, Liaoning Province Applicant before: INFORMATION COMMUNICATION BRANCH, STATE GRID LIAONING ELECTRIC POWER Co.,Ltd. Applicant before: Nanjing University of Aeronautics and Astronautics |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190604 |