Home Deep learning technology of Internet of Things Blockchain in distribution network faults
Article Open Access

Deep learning technology of Internet of Things Blockchain in distribution network faults

  • Hong Zhang EMAIL logo , Rui Li and Chuncheng Shi
Published/Copyright: August 11, 2022
Download Article (PDF)
Become an author with De Gruyter Brill
Submit Manuscript Author Information
Journal of Intelligent Systems
From the journal Journal of Intelligent Systems Volume 31 Issue 1

Abstract

Nowadays, the development of human society and daily life are inseparable from the power supply. Therefore, people also put forward higher requirements for the reliability of distribution network, but power companies can only passively deal with distribution network failures, which is a bottleneck for the improvement of distribution network reliability. The Internet of Things (IoT) is the best solution for online equipment status monitoring and basic data sharing for large, widely distributed, relatively fixed, and large numbers of equipment. The construction of the IoT for power distribution equipment faces many important problems, including the selection of networking, equipment selection, and interaction standards. When researching the implementation plan, research on the distribution of IoT market was carried out. Based on the grid, the idea of optimizing the investment selection plan of the power distribution using IoT was discussed, and a result verification model was established. After the completion of the theoretical part, a case study of medium-voltage distribution grid equipment management and medium-voltage distribution network equipment management based on the grid was carried out by taking a real enterprise application situation as an example. Realizing fault diagnosis of distribution network will not only provide decision support for operation and maintenance of distribution network for power companies, but also have great economic and social benefits. Aiming at the shortcomings of single data mining method in distribution network fault diagnosis, hybrid data mining method is proposed. First, rough set theory is used to reduce the original fault data and form a simplified rule set. Because of the non-linearity of distribution network fault and the strong learning ability, adaptability, and robustness of Bayesian network, Bayesian network can be used to classify distribution network faults. Therefore, a simplified fault diagnosis system is established in this paper, and its correctness is confirmed. Then, the learning and training are carried out by using Bayesian network to call the simplest rule set, which has the characteristics of short learning and training time and high diagnostic accuracy.

Keywords: distribution network fault diagnosis; data mining; information entropy; Bayesian network

1 Introduction

When the distribution network fails, especially when there are many uncertain factors such as information loss due to channel interference, protection, switch malfunction, rejection, and so on, the response of power system will be complicated, which will cause more difficulties [1,2]. Therefore, developing a fault diagnosis method of distribution network [3,4,5], which is sensitive to error information, strong fault-tolerant, fast, and accurate is important. Interaction and integration are the ability to liberate the power grid’s human management mode and realize the ultimate intelligent management. Interaction contains two connotations. On the one hand, it is the interaction between grid operators and grid equipment. Through wearable smart devices, grid personnel can “communicate” with the grid to handle affairs without leaving home. On the other hand, it is the interaction between the power users and the grid. Users use their own smart appliances or distributed power sources to “communicate” with the grid on transaction and service issues. Integration is to optimize and integrate the management process and classified business of power grid enterprises to realize all-round intelligence in production, marketing, and scheduling [6,7]. This is the ultimate goal of the power distribution Internet of Things (IoT).

At present, the distribution network is closely related to human life, and every household cannot do without the power system. Therefore, once the distribution network fails, it will directly affect social production and people’s lives, and bring huge losses to social production. Fault diagnosis of distribution network with high automation level [8,9,10] puts forward a correct and effective regional blackout recovery strategy to help dispatchers to accurately locate the fault location, isolate the fault area, and quickly restore the power supply in the fault area. Some faults will inevitably occur in power system due to long operation period, heavy maintenance workload, and aging of line insulation [11,12]. It is necessary to quickly monitor and eliminate faults, by realizing fault location and quick fault recovery of distribution network [13,14,15,16], and guarantee the normal life of human beings. In addition, due to the shortcomings of some existing fault diagnosis methods, it is difficult to achieve satisfactory results, so the improvement of distribution network fault diagnosis methods is imminent. Although the current domestic power distribution IoT technology research has been relatively mature, the management problems faced by this field are also very prominent. Since most of the power distribution IoT construction in various places is in the planning stage, it is still difficult to have a model to follow for the investment selection and supplier selection. From a technical point of view, there are a wide range of smart devices for power distribution IoT in the market, and the prices are relatively expensive, but there are few practical cases. So how to complete the equipment selection, how to choose the installation point, and what installation measures to take at different installation points are the issues to be discussed in management [17,18,19,20].

Data mining technology is a combination of database technology and artificial intelligence [21,22,23,24], and the research field is more active in the international community [25]. Data mining involves a large number of disciplines and methods. The commonly used data mining methods include neural network method [26], genetic algorithm, rough set method, and decision tree method. In data mining, decision tree classification method represents a classification [27,28,29].

It is necessary to determine the complex relationship between fault and its influencing factors, and then build a model to diagnose the distribution network fault. By using data mining technology to diagnose faults at feeder level [30,31,32], it can fine manage the patrol work of feeder, optimize the emergency repair station of distribution network, and provide guidance for the operation and maintenance work of feeder [33].

Aiming at eliminating the shortcomings, such as poor fault tolerance, low diagnosis speed and the uncertainty of the processed data, of single data mining method, new breakthroughs have been made in diagnosis, to ensure power supply, save human resources, and make rational use of various resources of distribution network, so as to maximize the economic and social benefits of distribution network.

2 Proposed method

2.1 Fault diagnosis of distribution network

The work of the distribution network includes four links: “generating, transforming, transmitting and distributing”. The working condition of distribution network has a great influence on the quality of power to the consumers, and its importance is getting higher and higher. Therefore, it is very important to improve the speed and accuracy of finding faults in a relatively short time, and to find out the fault area accurately, so as to maximize and guarantee the normal power supply to users. At the same time, the rapid and accurate treatment of the fault can also reduce the damage of the equipment and narrow the scope of the fault.

2.1.1 Fault location based on recloser and segmenter

The fault location method based on recloser and segmenter is realized by making full use of the action relationship between recloser and segmenter. By setting the action time and time of recloser and segmenter, the fault location can be accurately located. Therefore, the whole process is simple and easy to implement. However, different types of power grids have different configuration requirements for reclosers, and although the application of reclosers and sectional switches still has advantages in a certain environment, it has its own shortcomings, and the fault location processing time is too long; therefore, it is difficult for the network structure to achieve the best state.

2.1.2 Fault location based on feeder terminal unit (FTU)

In normal operation of distribution network, FTU can collect operation information corresponding to line switches to realize online monitoring of distribution network. In case of failure, FTU can upload relevant information to the control system, and finally find out the fault location through corresponding analysis, and determine the optimal recovery scheme.

According to the current information of each node in the fault zone, the network matrix of the fault zone can be judged and searched out. With the deepening of the research of matrix, the speed of matrix algorithm has been greatly improved, and its application scope is more and more extensive.

2.1.3 Fault current discrimination method

Based on graph theory, this method realizes fault diagnosis according to the expansion model of distribution network. This method can finally get a fault judgment matrix. The concrete steps are to first write the network description matrix according to the expansion model of distribution network. Then, according to the information matrix and the network description matrix, the rough description matrix is normalized, and finally the rough description is obtained. This method is based on the change in power flow. The structure and parameters of the system will change when faults occur. This change makes the calculation and analysis of power flow take more time, which will affect the speed of diagnosis and the quick recovery of faults. In addition, in normal operation, the power flow value of some lines is close to zero, such as online operation, so even if the power flow is used to judge the fault, the accuracy of diagnosis cannot be ensured.

2.2 Data mining

Database technology, artificial intelligence, and mathematical statistics are three powerful technical pillars of data mining research.

2.2.1 Introduction to data mining

Noisy and incomplete data. The essence of data mining is to filter the intuitive information in mass data and discover the inherent unknown information. It often requires constant adjustments to obtain these unknowns from the data. At the same time, data mining is developed on the basis of multi-disciplinary technology integration, including database, machine learning, statistics, high-performance computing, and visualization.

2.2.2 Data mining process

2.2.2.1 Data preparation

Data mining will mainly extract knowledge from some data, that is, it extracts data related to data mining technology from the database according to user’s requirements. In this process, in order to form a real database, some database operations will be used to process the data.

  1. In this stage, the actual mining operations are carried out, including the key points of deciding how to generate hypothetical knowledge by selecting appropriate tools to discover knowledge and verify the discovered knowledge.

  2. The results are analyzed, explained, and evaluated. Generally, according to the data mining operation, we decide which analysis method to use.

2.2.3 Application of data mining technology in distribution network faults

More and more power supply equipment are diversified, and the detection of various equipment by power supply companies is becoming more and more perfect. With more and more diagnostic information, the application of data mining technology also has a solid foundation.

Because of channel failure, monitoring equipment failure, and other reasons, some parameters of distribution network equipment are lost or even wrong. Traditional expert system cannot use historical data rules to analyze, process, and give the final decision. It is very difficult to find the model that is consistent with the current fault in the complicated historical data. The following processing is needed to determine the power network fault. Cluster analysis is clustering analysis of faults, wherein each fault satisfying the similarity between any two data samples (similarity can be calculated by specific values of sample description attributes, usually expressed by sample distance) is relatively high. Based on cluster analysis, the similarity of distribution network faults is processed and analyzed, and the faults are classified to lay the foundation for the next correlation analysis. Association rule analysis of faults refers to the analysis of the influence and association relations among different data in historical data. By integrating fault areas with certain association rules, a concise expression can be obtained. By comparing the support degree and reliability index under association rules, we can judge the usefulness of association conclusion between faults.

Diagnosis process is a decision-making process. The basis of knowledge will affect the correctness of decision-making. The richer and more reliable the knowledge is, the more correct the decision-making will be. It is difficult for humans to discover the hidden knowledge and rules in front of them, and data mining can make up for the shortcomings of humans in this regard. Therefore, data mining technology is needed in fault diagnosis of distribution network. It is precisely because of this demand that data mining technology is developing continuously. Data mining theory can integrate these methods effectively and compare and analyze them. Thus, the system, function, and process of information mining are defined, which lay a good foundation for the application of various advanced theories and technologies in condition monitoring and fault diagnosis, and makes the whole process clear and easy to implement.

2.3 Application of data mining technology based on information entropy in distribution network fault diagnosis

2.3.1 Information entropy

Shannon refers to the concept of thermodynamics and the average amount of information excluded from redundancy as “information entropy,” and gives the mathematical expression for calculating information entropy. There is a measure of the value of information, and more inferences can be made about the flow of knowledge.

Essentially, information entropy reflects the information content represented by events. Therefore, the greater the uncertainty of variables, the more information they can represent, and the greater the value of information entropy:

(1) H ( x ) = − ∑ i + 1 n p ( x i ) ln ( p ( x i ) ) .

In formula (1) p(x i ) is the probability of x i occurrence. At present, information entropy has been widely used in various fields. Entropy is an important choice for constructing evaluation convergence function in path search. When more and more information is known, the probability of the remaining unknown events is lower, and once these unknown events occur, they represent a new search direction. In the aspect of fault diagnosis, different faults are expressed differently at different measuring points. When a fault is expressed differently from other faults at one measuring point, that is, the change caused by the fault at that measuring point is not found at other measuring points, then it shows that the fault can be distinguished accurately by the measuring point, and the occurrence of the fault can be identified when the specific value is detected by the measuring point.

  1. Non-negative: H(X) ≥ 0

  2. Certainty: H(1,0) = H(1,0,0) = H(1,0,0,0) = … H(1,0,…,0) = 0

  3. H(XY) = H(X) + H(Y)

  4. Extremum: H ( p 1 , p 2 , … , p n ) ≤ H 1 n , 1 n , … , 1 n = log n

The calculation of information entropy is very complicated. The information with multiple preconditions can hardly be calculated.

2.3.2 Information entropy introduction

A large number of alarm signals are generated, some of which are related and some are independent. Independent features can provide complementary information, so they should be retained. Relevant features can generate redundant information and increase the computational workload, so they need to be eliminated. Due to the complexity of rule extraction, in order to find the optimal reduction from the reduction combination, the information entropy reduction algorithm is used.

However, outlier detection algorithms based on distance and bias require parameters to be determined in advance, which can lead to wrong conclusions if the parameters are not selected properly.

Since information entropy only depends on the probability of each attribute in the record, the value of the attribute can be either numeric or non-numeric (such as character). That is to say, information entropy can integrate numerical attributes and nominal attributes in a unified framework for processing.

2.4 Bayesian network

Bayesian network technology and Bayesian network have overcome many conceptual and computational difficulties of rule-based systems.

Edge is used to represent the dependency of each variable. Each attribute node in the structure diagram networks represented by directed acyclic graphs can be viewed from both quantitative and qualitative perspectives, which can be more intuitive qualitative analysis of the impact of those variables on its final results. Quantitatively, Bayesian networks can express the dependence of different variables on their parent nodes by using conditional probability distribution, and the relationship of each node in Bayesian networks can be expressed by using joint probability distribution. For example, if X 1, X 2, …, X n is used to represent variable nodes in a network, the joint probability distribution of each variable can be obtained by multiplying the conditional probability of each variable using formula (2):

(2) P ( X 1 , X 2 , … , X n ) = ∏ i = 1 n P ( X i | π ( X i ) ) .

3 Experiments

When power system fails, it will produce a series of effects, such as abnormal drop in voltage, instantaneous increase in current, and change in public power. After the fault information is transmitted to the fault diagnosis system, various data are transmitted and the transmitted data are analyzed and processed, and finally the location where the fault occurs is diagnosed. The general steps of applying data mining technology based on information entropy to distribution network fault diagnosis are as follows:

  1. Original decision table is established.

  2. Based on the original decision table of fault diagnosis, an identifiable matrix is established according to formula (3).

    (3) c i j = { a ∈ A : a ( x i ) = a ( x j ) } , D ( x i ) ≠ D ( x j ) , 0 , D ( x i ) = D ( x j ) , − 1 , a ( x i ) = a ( x j ) , D ( x i ) ≠ D ( x j ) ,

  3. Find out the core attributes and get the reduction results.

  4. The average mutual information in attribute reduction results is obtained by using the method of solving average.

    (4) I ( X ; Y ) = ∑ x , y p ( x , y ) log b p ( x , y ) p ( x ) p ( y ) .

  5. According to the actual situation, the joint probability of conditional attributes in different fault areas is obtained by using formulas (5) and (6). According to the calculation results, the fault area is determined and compared with the predetermined fault area.

(5) P ( X | C k ) = ∏ i = 1 n P ( x k | C k ) ,

(6) P ( X | C k ) = 1 N N c k + N x i N .

4 Discussion

4.1 Fault diagnosis model of distribution network

Figure 1 is a simple distribution network model with three areas, Sec 1, Sec 2, and Sec 3.

Figure 1 Simple distribution network model.
Figure 1

Simple distribution network model.

4.2 Action principle of relay protection

The basic principles of relay protection: 1. When each electrical component has internal and external faults (including normal operation), the difference in current phase or power direction between two sides can constitute the protection of various differential principles, such as longitudinal differential protection, phase difference high frequency protection, directional high frequency protection, and so on. 2. The protection of differential principle can only operate when the internal fault of the protected component occurs, but does not respond to the external fault. So, it is considered to be absolutely selective. When various relay protection devices are constructed according to the above principles, their parameters can be reflected by the current and voltage in each phase, or only by the current and voltage of one of the symmetrical components (negative, zero, or positive sequence).

4.3 Fault diagnosis model of distribution network

Formula (3) is used to form the rules for the establishment of recognizable matrices, i, j = 1,2, …, n. Conditional attributes are represented by A, while a(x i ) and a(x j ) represent elements of row I and column J in the set of conditional attributes A, respectively. According to the formation rules of the discernible matrix, the c ij = c ji in the above definition matrix is a symmetric matrix, so the upper triangle or the lower triangle is needed. According to the definition, when i = j, c ij = 0, the diagonal elements of the identifiable matrix are all 0.

Appeal rules can give an identifiable matrix, which is based on the original decision table of fault diagnosis, and the corresponding identifiable matrix can be derived from the rules of the formation of identifiable matrix. According to the rules of core attribute extraction, the elements containing carbon dioxide (CO2) and carbon trioxide (CO3) in the discernible matrix are removed, and the retained elements are carbon monoxide (CO1), risk ratio (RR1); central cannabinoid (CB1), RR1; CB1, CO1. Combining the above three sets of elements with CO2 and CO3, respectively, the attribute reduction result of the element decision table is achieved.

In general, the number of attribute reduction combinations is not always certain. When there is only one attribute reduction combination extracted, this is the best attribute reduction combination. But when there are multiple combinations in the attribute reduction results, it needs to be operated to find the simplest attribute reduction combination. In this paper, the results of attribute reduction include the values of CO1, CO2, and CO3, which can also be calculated by formula (3). Because the association degree of these three attribute reduction combinations is different, if we need to use Bayesian network model reasoning, we must assume that each conditional attribute is independent of each other. In order to minimize the influence of independence hypothesis on the results of probabilistic reasoning, it is necessary to find the attribute reduction combination with the least dependency among attributes. Results can be obtained by using the method of solving the average mutual information [formula (4)]. The combination of the simplest attribute reduction is the one with the smallest average mutual information.

According to the formula: (1) CO1, RR1, CO2, and CO3 are 0.0869; (2) CB1, RR1, CO2, and CO3 are 0.1192; and (3) CB1, CO1, CO2, and CO3 are 0.072. It is concluded that the combination of the simplest attribute reduction is CB1, CO1, CO2, and CO3. A new decision table is generated after the combination of the simplest attribute reduction is obtained, as shown in Table 1.

Table 1

Decision table after attribute reduction

Sample number CB1 CO1 CO2 CO3 Fault area
1 1 1 0 0 Sec 1
2 0 0 1 0 Sec 2
3 0 0 0 1 Sec 3
4 1 0 1 0 Sec 2
5 1 0 0 1 Sec 3
6 1 0 0 0 Sec 2/Sec 3
7 0 0 0 0 NO

4.4 Bayesian network model

From Figure 1, a Bayesian network model with five nodes can be built, as shown in Figure 2.

Figure 2 Bayesian network structure diagram.
Figure 2

Bayesian network structure diagram.

Next the prior probability of each fault area is calculated based on the existing fault information, as shown in Table 2.

Table 2

Prior probability of each fault region

Fault area Sec 1 Sec 2 Sec 3 Sec 2/Sec 3 NO
Prior probability 1/7 2/7 2/7 1/7 1/7

According to the Bayesian theorem, in the Bayesian network model, every child node can find the parent node which has a direct relationship with it. In order to simplify the operation process, this article regards each child node and parent node as independent. The joint probability of each conditional attribute taking a specific value under a certain fault condition can be expressed as the product of the conditional probability of each attribute taking a corresponding value under such fault condition, such as formula (5). In the formula, the fault type is C k , and the condition attribute is x k . P(X|C k ) can be obtained from the sample of historical fault information record. If there is no fault type C k and condition attribute x k in the record, formula (6) is used. The conditional probabilities of CB1, CO1, CO2, and CO3 in different fault areas can be calculated from the formulas (5) and (6), as shown in Table 3.

Table 3

Conditional probability of CB1, CO1, CO2, and CO3 in different fault areas

Sec 1 Sec 2 Sec 3 Sec 2/Sec 3 NO
CB1 = 0 1/10 1/2 1/2 1/10 1
CB1 = 1 1 1/2 1/2 1 1/11
CO1 = 0 1/13 1 1 1 1
CO1 = 1 1 1/15 1/15 1/8 1/8
CO2 = 0 1 1/19 1 1 1
CO2 = 1 1/9 1 1/16 1/9 1/9
CO3 = 0 1 1 1/19 1 1
CO3 = 1 1/9 1/16 1 1/9 1/9

According to the probability of occurrence of conditional attributes under different fault types, Sec 1 is set as fault area, and the conditional attributes of short circuit protection are CB1 = 0, CO1 = 1, CO2 = 0, CO3 = 0. Condition attributes of the fault are assumed to be CB1 = 0, CO1 = 1, CO2 = 0, CO3 = 0, and then calculated according to the method, the most likely area is Sec 1. According to the calculation results, it can be judged that the fault area is Sec 1, which is exactly the same as the predetermined fault area. The diagnostic error curve of the Bayesian network model is shown in Figure 3.

Figure 3 Diagnostic error curve of Bayesian network model.
Figure 3

Diagnostic error curve of Bayesian network model.

First, this study compares the shortcomings of various existing methods in practical application. Aiming at the problem of diversification of results, a method of combining rough set theory based on information entropy and Bayesian network is proposed. Table 3 is used as the main tool. Then, the average mutual information of information entropy is used to reduce the feature parameters of samples, solve the problems of too large scale of network, and too slow speed of classification and recognition, so that the structure of Bayesian network is simplified, the training time of network is shortened, and the accuracy of fault identification is improved. The distribution network model is established, and the experimental results are summarized and analyzed. The characteristics of this method are more intuitively displayed. In this study, according to the characteristics of distribution network, Bayesian network is selected as the classifier. The design and training process of the network are introduced in detail, which was verified by an example and good results were achieved. The sampling frequency is 4 kHz. The number of measurement nodes of the PMU device is related to the compressed sensing algorithm. In order to recover the original signal more accurately, the voltage amplitude changes before and after the fault are calculated, combined with the node impedance, and the reconstruction result is shown in Figure 4.

Figure 4 The reconstruction results.
Figure 4

The reconstruction results.

When a fault occurs, the protection and circuit breaker action information is shown in Table 4.

Table 4

Protection and circuit breaker action information

Serial number Time Action information
1 16:00:09:307 L19-20 left end current I stage protection start
2 16:00:09:343 L19-20 right-side current stage I protection starts
3 16:00:09:393 CBz0-19 trip
4 16:00:10:089 L2-19 left end current stage II protection start
5 16:00:10:124 CB-21 trip

According to the method of fault location based on the multi-sensor data fusion algorithm based on the enhanced belief divergence measure, the fusion results are shown in Table 5.

Table 5

Fusion results

Element A1 A2 A3
B3 0.015 0.957 0.028
B4 0.082 0.915 0.003
B5 0.283 0.632 0.085
B6 0.986 0.013 0.001
B7 0.074 0.854 0.072

In the calculation example, the interruption time of all ILs is mainly concentrated in the time period of 7:45–13:15; L#13 was interrupted in the time period of 19:30–24:00; L#23 was interrupted in the time period of 17:00–22:30; L#29 was interrupted at 16:45–23:30. The time of failure is shown in Figure 5.

Figure 5 The time of failure.
Figure 5

The time of failure.

Reconstruction amplitude of each bus and line within the fault range is shown in Table 6.

Table 6

Reconstruction amplitude of each bus and line within the fault range data

Element Reconstruction amplitude
BI 83
B2 192
B3 120
B19 482
B20 501
B21 254
L12 141
L23 189

Compared with the original fault conditions, the fault recovery control strategy generated by the method in this study can greatly restore the system load. The control strategy that takes into account the interaction of supply and demand further reduces the load loss on the basis of only considering the network reconfiguration, which proves the generation strategy of this study. The effectiveness of active distribution network resilience under extreme disasters is improved. The total loss of system load before and after the fault recovery control is shown in Figure 6.

Figure 6 The total loss of system load before and after the fault recovery control.
Figure 6

The total loss of system load before and after the fault recovery control.

5 Conclusion

When the power system fails, it will cause a blackout of large area. However, the faults of the power system, especially the power grid, are inevitable. In order to quickly monitor and eliminate the faults, and quickly realize the fault location and fault type identification of the transmission network, so as to facilitate the quick recovery after the overhaul and accident, a high-quality fault diagnosis system is needed.

After the application of the IoT has been popularized in recent years, the number of interface devices used in the sensors of animals, plants, and machinery will far exceed the number of mobile phones. This will further promote the economic development and progress, and create another industry with unlimited potential to promote economic benefits for the society. According to the current demand for the IoT in various fields of the society, in recent years, a large number of sensors and electronic meters used in the IoT will need to be produced, and the production of these requires a lot of manpower and can provide a lot of jobs for the people of the society. The production of these meters at the same time can speed up the production of information technology components. To truly build an efficient and useful IoT, it must have the necessary components. This study tried to apply the methods of information entropy, Bayesian network, and decision tree commonly used in data mining to the above methods through simulation examples. Through the application of the above methods, fault diagnosis becomes more accurate and effective, and the following conclusions are drawn.

By installing more precise and optimized engine equipment in the smart grid, power providers can break the limitations of “traditional” networks and build smart grids that can more actively manage power failures. Power suppliers can use these engine equipment to achieve more optimized management of power failures, manage the complex topology and resource constraints in the grid, and at the same time, through optimized engine equipment and IoT, power generation equipment can be identified more quickly and accurately. In this way, the power supplier can more efficiently arrange the sequence of power failure detection and power equipment restoration. Thus, at least 30% of the power outage time and frequency of power outages can be saved, and the economic losses caused by power outages can be reduced accordingly, and reducing the time and frequency of power outages increases the reliability of the power grid and the satisfaction of customers. Information entropy is a high mathematical method. Through the application of this method, the time of diagnosis process is reduced, and good results are achieved. Bayesian network is used to achieve better fault classification. By training the Bayesian network with samples, choosing an optimal network to realize classification, and by calling the toolbox method, the output of fault report is realized. The application of multi-variable decision tree in distribution network fault diagnosis can effectively simplify and describe knowledge, reduce the redundancy of diagnosis information, greatly improve the diagnosis efficiency, and the results are easy to understand.

Blockchain technology has unique advantages in terms of security. Its high redundancy and data are not easy to be tampered with. It coincides with the strict requirements of the distribution network protection and control system for information security. Each node in the blockchain network saves complete data, and through a carefully designed consensus mechanism, ensures the consistency of each node’s data. The newly generated data are associated with existing data blocks in the form of blocks, forming a chain structure, so that all historical data can be traced and cannot be tampered with. Blockchain technology was first applied in the field of digital encryption currency, and its high security based on cryptographic principles has been widely recognized. With the improvement of power supply quality requirements of users, more scholars and experts will pay attention to the fault diagnosis of distribution network. With the rise of data mining methods in artificial intelligence, more attention will be paid to the application of data mining in distribution network. In this study, although the information entropy is used to carry out profound research on the fault troubleshooting of the distribution network, there are still many shortcomings. The depth and breadth of the research in this article are not enough. In the process of this research, the selection and acquisition of experimental data are absolutely ideal. The completeness and effectiveness are not enough. Our academic level research is also limited. This research is still in the preliminary stage. In the future work, we will study suitable methods from more perspectives based on the existing technology and level, and means to continuously improve the effectiveness of the method.

  1. Funding information: This work was supported by Jilin Province Foundation Department research projects (20210101485JC), Jilin Provincial Department of Education research projects (No. JJKH20210673KJ).

  2. Conflict of interest: Authors state no conflict of interest.

References

[1] Akhavan-Rezai E, Haghifam MR, Fereidunian A. Data-driven reliability modeling, based on data mining in distribution network fault statistics Powertech. IEEE Buchar. 2009. 10.1109/PTC.2009.5281796.Search in Google Scholar

[2] Shu H, Cao L, Xu Z, Liu K. The research of multidimensional information decision mining based on information entropy. Int Forum Inf Technol Appl. 2009. 10.1109/IFITA.2009.559.Search in Google Scholar

[3] Cui S, Shang Z, Wang Q, et al. Research of fault indication system based on WIA industrial wireless technology for distribution network. Control & Decision Conference; 2013.Search in Google Scholar

[4] Adil M, Ali J, Attique M, Jadoon MM, Abbas S, Alotaibi SR, et al. Three byte-based mutual authentication scheme for autonomous Internet of Vehicles. IEEE Trans Intell Transport Syst. 2021;23(7):9358–69.10.1109/TITS.2021.3114507Search in Google Scholar

[5] Adil M, Khan MK, Jamjoom M, Farouk A. MHADBOR: AI-enabled administrative distance based opportunistic load balancing scheme for an agriculture internet of things network. IEEE Micro. 2021;42(1):41–50.10.1109/MM.2021.3112264Search in Google Scholar

[6] Adil M, Jan MA, Mastorakis S, Song H, Jadoon MM, Abbas S, et al. Hash-MAC-DSDV: mutual authentication for intelligent IoT-based cyber-physical systems. IEEE Internet Things J. 2021. 10.1109/JIOT.2021.3083731.Search in Google Scholar

[7] Farouk A, Alahmadi A, Ghose S, Mashatan A. Blockchain platform for industrial healthcare: vision and future opportunities. Computer Commun. 2020;154:223–35.10.1016/j.comcom.2020.02.058Search in Google Scholar

[8] Liao H, Yang L, Geng X. Application research of new distributed data mining model based on intelligent agent and web service in digital gas fields. International Conference On Computational & Information Sciences; 201110.1109/ICCIS.2011.98Search in Google Scholar

[9] Chen S, Jia HE. Research on cost management system of distribution network construction projects based on data mining china. International Conference On Electricity Distribution; 2013.10.1109/CICED.2012.6508454Search in Google Scholar

[10] Li XM, Liu H, Wang WX, Ye Z, Lv HB, Lv ZH. Big data analysis of the internet of things in the digital twins of smart city based on deep learning. Future Gener Computer Syst. 2021;128:167–77.10.1016/j.future.2021.10.006Search in Google Scholar

[11] Li Y, Zuo Y, Song H, Lv Z. Deep learning in security of internet of things. IEEE Internet Things J. 2021;99:1–1142.10.1109/JIOT.2021.3106898Search in Google Scholar

[12] Zhang Z. Application and research of data mining technology in communication network environment. Adv Mater Res. 2014;989–994:3814–7.10.4028/www.scientific.net/AMR.989-994.3814Search in Google Scholar

[13] Zheng CF, Jiang L, Jiang LQ, Wu ZJ. Application and research of Bayesian network in data mining. Adv Mater Res. 2012;532–533:738–42.10.4028/www.scientific.net/AMR.532-533.738Search in Google Scholar

[14] Liu M, Qu M, Zhao B. Research and citation analysis of data mining technology based on Bayesian algorithm. Mob Netw Appl. 2016;22(3):1–9.10.1007/s11036-016-0797-2Search in Google Scholar

[15] Liang GY, Gao Y, Zhang L, Xu AR. Research of state estimation in power network based on data mining technology. Appl Mech Mater. 2013;448–453:2594–8.10.4028/www.scientific.net/AMM.448-453.2594Search in Google Scholar

[16] Yu DR, He HX, He HX, Zhang XX. The research of visualization data mining in industrial data. Appl Mech Mater. 2013;274:5.10.4028/www.scientific.net/AMM.274.545Search in Google Scholar

[17] Zhao Q, Wu S, Ying C, Jia X. The research of information filtering technology based on Bayesian network. Procedia Environ Sci. 2011;11(Part B):545–51.10.1016/j.proenv.2011.12.086Search in Google Scholar

[18] Sui JY, Wang YN, Ren Z. The distribution network reconfiguration based on spatial data mining. Appl Mech & Mater. 2013;427–429:2748–51.10.4028/www.scientific.net/AMM.427-429.2748Search in Google Scholar

[19] Zhou J, Zhang X, Jia C. Application of fault location method based on multi-source fault information in repair of distribution network. North China Electr Power. 2015;1070–1072:6.10.4028/www.scientific.net/AMR.1070-1072.678Search in Google Scholar

[20] Gao Z, Peng Z, Gao N, Chen B. A distribution network fault data analysis method based on association rule mining. Power & Energy Engineering Conference; 2015.Search in Google Scholar

[21] Osamh. IK, Ghaida MA. Energy efficient routing and reliable data transmission protocol in WSN. Int J Adv Soft Comput Appl. 2020;12(3):45–53.Search in Google Scholar

[22] Khalaf OI, Abdulsahib GM, Sabbar BM. Optimization of wireless sensor network coverage using the bee algorithm. J Inf Sci Eng. 2020;36(2):377–86.Search in Google Scholar

[23] Rajendran S, Khalaf OI, Alotaibi Y, Alghamdi S. MapReduce-based big data classification model using feature subset selection and hyperparameter tuned deep belief network. Sci Rep. 2021;11:24138. 10.1038/s41598-021-03019-y.Search in Google Scholar PubMed PubMed Central

[24] Khalaf OI, Abdulsahib GM. Optimized dynamic storage of data (ODSD) in IoT based on blockchain for wireless sensor networks. Peer-to-Peer Netw Appl. 2021;14:2858–73. 10.1007/s12083-021-01115-4.Search in Google Scholar

[25] Zhang HP, He HY, Gu MX. Research of fiber-optical fault diagnosis based on support vector machine (SVM) mining. Fifth International Conference on Intelligent Systems Design & Engineering Applications; 2014Search in Google Scholar

[26] Tao Z, Pan YM, Guo KY, Wang ZP. Fault location of distribution network based on immune algorithm. Power Syst Prot Control. 2014;42(1):77–83.Search in Google Scholar

[27] Shu HC, Sun XF, Da-Jun SI. A rough set approach to distribution network fault location based on fault complain call information. Power Syst Technol. 2004;28(1):64–8.Search in Google Scholar

[28] Sheng S, Ma Z, Wu J, Gao N. Distribution network fault restoration based on improved adaptive genetic algorithm. Algorithm Second International Conference on Intelligent Computation Technology & Automation; 2009.Search in Google Scholar

[29] Guo K. Research on location selection model of distribution network with constrained line constraints based on genetic algorithm. Neural Comput Appl. 2020;32(6):1679–89.10.1007/s00521-019-04257-ySearch in Google Scholar

[30] Yan W. Application research of data mining technology about teaching quality assessment in colleges and universities. Proc Eng. 2011;15:4241–5.10.1016/j.proeng.2011.08.795Search in Google Scholar

[31] Wang L, Tong YF. Application research of data mining technology in power dispatching management system. International Conference on Smart Grid & Electrical Automation; 2016.Search in Google Scholar

[32] Zhang WQ. Application research of data mining technology on growth management of forestry. Adv Mater Res. 2014;846–847:995–8.10.4028/www.scientific.net/AMR.846-847.995Search in Google Scholar

[33] Zhu Q, Wang Y, Li J. Public transport IC card data analysis and operation strategy research Based on data mining technology. Int Forum Computer Sci Tech Appl. 2009;1422–5; 10.1109/IFCSTA.2009.352.Search in Google Scholar
Received: 2022-01-24
Revised: 2022-03-31
Accepted: 2022-05-22
Published Online: 2022-08-11

© 2022 Hong Zhang et al., published by De Gruyter

This work is licensed under the Creative Commons Attribution 4.0 International License.

Articles in the same Issue

  1. Research Articles
  2. Construction of 3D model of knee joint motion based on MRI image registration
  3. Evaluation of several initialization methods on arithmetic optimization algorithm performance
  4. Application of visual elements in product paper packaging design: An example of the “squirrel” pattern
  5. Deep learning approach to text analysis for human emotion detection from big data
  6. Cognitive prediction of obstacle's movement for reinforcement learning pedestrian interacting model
  7. The application of neural network algorithm and embedded system in computer distance teach system
  8. Machine translation of English speech: Comparison of multiple algorithms
  9. Automatic control of computer application data processing system based on artificial intelligence
  10. A secure framework for IoT-based smart climate agriculture system: Toward blockchain and edge computing
  11. Application of mining algorithm in personalized Internet marketing strategy in massive data environment
  12. On the correction of errors in English grammar by deep learning
  13. Research on intelligent interactive music information based on visualization technology
  14. Extractive summarization of Malayalam documents using latent Dirichlet allocation: An experience
  15. Conception and realization of an IoT-enabled deep CNN decision support system for automated arrhythmia classification
  16. Masking and noise reduction processing of music signals in reverberant music
  17. Cat swarm optimization algorithm based on the information interaction of subgroup and the top-N learning strategy
  18. State feedback based on grey wolf optimizer controller for two-wheeled self-balancing robot
  19. Research on an English translation method based on an improved transformer model
  20. Short-term prediction of parking availability in an open parking lot
  21. PUC: parallel mining of high-utility itemsets with load balancing on spark
  22. Image retrieval based on weighted nearest neighbor tag prediction
  23. A comparative study of different neural networks in predicting gross domestic product
  24. A study of an intelligent algorithm combining semantic environments for the translation of complex English sentences
  25. IoT-enabled edge computing model for smart irrigation system
  26. A study on automatic correction of English grammar errors based on deep learning
  27. A novel fingerprint recognition method based on a Siamese neural network
  28. A hidden Markov optimization model for processing and recognition of English speech feature signals
  29. Crime reporting and police controlling: Mobile and web-based approach for information-sharing in Iraq
  30. Convex optimization for additive noise reduction in quantitative complex object wave retrieval using compressive off-axis digital holographic imaging
  31. CRNet: Context feature and refined network for multi-person pose estimation
  32. Improving the efficiency of intrusion detection in information systems
  33. Research on reform and breakthrough of news, film, and television media based on artificial intelligence
  34. An optimized solution to the course scheduling problem in universities under an improved genetic algorithm
  35. An adaptive RNN algorithm to detect shilling attacks for online products in hybrid recommender system
  36. Computing the inverse of cardinal direction relations between regions
  37. Human-centered artificial intelligence-based ice hockey sports classification system with web 4.0
  38. Construction of an IoT customer operation analysis system based on big data analysis and human-centered artificial intelligence for web 4.0
  39. An improved Jaya optimization algorithm with ring topology and population size reduction
  40. Review Articles
  41. A review on voice pathology: Taxonomy, diagnosis, medical procedures and detection techniques, open challenges, limitations, and recommendations for future directions
  42. An extensive review of state-of-the-art transfer learning techniques used in medical imaging: Open issues and challenges
  43. Special Issue: Explainable Artificial Intelligence and Intelligent Systems in Analysis For Complex Problems and Systems
  44. Tree-based machine learning algorithms in the Internet of Things environment for multivariate flood status prediction
  45. Evaluating OADM network simulation and an overview based metropolitan application
  46. Radiography image analysis using cat swarm optimized deep belief networks
  47. Comparative analysis of blockchain technology to support digital transformation in ports and shipping
  48. IoT network security using autoencoder deep neural network and channel access algorithm
  49. Large-scale timetabling problems with adaptive tabu search
  50. Eurasian oystercatcher optimiser: New meta-heuristic algorithm
  51. Trip generation modeling for a selected sector in Baghdad city using the artificial neural network
  52. Trainable watershed-based model for cornea endothelial cell segmentation
  53. Hessenberg factorization and firework algorithms for optimized data hiding in digital images
  54. The application of an artificial neural network for 2D coordinate transformation
  55. A novel method to find the best path in SDN using firefly algorithm
  56. Systematic review for lung cancer detection and lung nodule classification: Taxonomy, challenges, and recommendation future works
  57. Special Issue on International Conference on Computing Communication & Informatics
  58. Edge detail enhancement algorithm for high-dynamic range images
  59. Suitability evaluation method of urban and rural spatial planning based on artificial intelligence
  60. Writing assistant scoring system for English second language learners based on machine learning
  61. Dynamic evaluation of college English writing ability based on AI technology
  62. Image denoising algorithm of social network based on multifeature fusion
  63. Automatic recognition method of installation errors of metallurgical machinery parts based on neural network
  64. An FCM clustering algorithm based on the identification of accounting statement whitewashing behavior in universities
  65. Emotional information transmission of color in image oil painting
  66. College music teaching and ideological and political education integration mode based on deep learning
  67. Behavior feature extraction method of college students’ social network in sports field based on clustering algorithm
  68. Evaluation model of multimedia-aided teaching effect of physical education course based on random forest algorithm
  69. Venture financing risk assessment and risk control algorithm for small and medium-sized enterprises in the era of big data
  70. Interactive 3D reconstruction method of fuzzy static images in social media
  71. The impact of public health emergency governance based on artificial intelligence
  72. Optimal loading method of multi type railway flatcars based on improved genetic algorithm
  73. Special Issue: Evolution of Smart Cities and Societies using Emerging Technologies
  74. Data mining applications in university information management system development
  75. Implementation of network information security monitoring system based on adaptive deep detection
  76. Face recognition algorithm based on stack denoising and self-encoding LBP
  77. Research on data mining method of network security situation awareness based on cloud computing
  78. Topology optimization of computer communication network based on improved genetic algorithm
  79. Implementation of the Spark technique in a matrix distributed computing algorithm
  80. Construction of a financial default risk prediction model based on the LightGBM algorithm
  81. Application of embedded Linux in the design of Internet of Things gateway
  82. Research on computer static software defect detection system based on big data technology
  83. Study on data mining method of network security situation perception based on cloud computing
  84. Modeling and PID control of quadrotor UAV based on machine learning
  85. Simulation design of automobile automatic clutch based on mechatronics
  86. Research on the application of search algorithm in computer communication network
  87. Special Issue: Artificial Intelligence based Techniques and Applications for Intelligent IoT Systems
  88. Personalized recommendation system based on social tags in the era of Internet of Things
  89. Supervision method of indoor construction engineering quality acceptance based on cloud computing
  90. Intelligent terminal security technology of power grid sensing layer based upon information entropy data mining
  91. Deep learning technology of Internet of Things Blockchain in distribution network faults
  92. Optimization of shared bike paths considering faulty vehicle recovery during dispatch
  93. The application of graphic language in animation visual guidance system under intelligent environment
  94. Iot-based power detection equipment management and control system
  95. Estimation and application of matrix eigenvalues based on deep neural network
  96. Brand image innovation design based on the era of 5G internet of things
  97. Special Issue: Cognitive Cyber-Physical System with Artificial Intelligence for Healthcare 4.0.
  98. Auxiliary diagnosis study of integrated electronic medical record text and CT images
  99. A hybrid particle swarm optimization with multi-objective clustering for dermatologic diseases diagnosis
  100. An efficient recurrent neural network with ensemble classifier-based weighted model for disease prediction
  101. Design of metaheuristic rough set-based feature selection and rule-based medical data classification model on MapReduce framework
https://doi.org/10.1515/jisys-2022-0064
Keywords for this article
distribution network fault diagnosis; data mining; information entropy; Bayesian network
Creative Commons
BY 4.0

Articles in the same Issue

  1. Research Articles
  2. Construction of 3D model of knee joint motion based on MRI image registration
  3. Evaluation of several initialization methods on arithmetic optimization algorithm performance
  4. Application of visual elements in product paper packaging design: An example of the “squirrel” pattern
  5. Deep learning approach to text analysis for human emotion detection from big data
  6. Cognitive prediction of obstacle's movement for reinforcement learning pedestrian interacting model
  7. The application of neural network algorithm and embedded system in computer distance teach system
  8. Machine translation of English speech: Comparison of multiple algorithms
  9. Automatic control of computer application data processing system based on artificial intelligence
  10. A secure framework for IoT-based smart climate agriculture system: Toward blockchain and edge computing
  11. Application of mining algorithm in personalized Internet marketing strategy in massive data environment
  12. On the correction of errors in English grammar by deep learning
  13. Research on intelligent interactive music information based on visualization technology
  14. Extractive summarization of Malayalam documents using latent Dirichlet allocation: An experience
  15. Conception and realization of an IoT-enabled deep CNN decision support system for automated arrhythmia classification
  16. Masking and noise reduction processing of music signals in reverberant music
  17. Cat swarm optimization algorithm based on the information interaction of subgroup and the top-N learning strategy
  18. State feedback based on grey wolf optimizer controller for two-wheeled self-balancing robot
  19. Research on an English translation method based on an improved transformer model
  20. Short-term prediction of parking availability in an open parking lot
  21. PUC: parallel mining of high-utility itemsets with load balancing on spark
  22. Image retrieval based on weighted nearest neighbor tag prediction
  23. A comparative study of different neural networks in predicting gross domestic product
  24. A study of an intelligent algorithm combining semantic environments for the translation of complex English sentences
  25. IoT-enabled edge computing model for smart irrigation system
  26. A study on automatic correction of English grammar errors based on deep learning
  27. A novel fingerprint recognition method based on a Siamese neural network
  28. A hidden Markov optimization model for processing and recognition of English speech feature signals
  29. Crime reporting and police controlling: Mobile and web-based approach for information-sharing in Iraq
  30. Convex optimization for additive noise reduction in quantitative complex object wave retrieval using compressive off-axis digital holographic imaging
  31. CRNet: Context feature and refined network for multi-person pose estimation
  32. Improving the efficiency of intrusion detection in information systems
  33. Research on reform and breakthrough of news, film, and television media based on artificial intelligence
  34. An optimized solution to the course scheduling problem in universities under an improved genetic algorithm
  35. An adaptive RNN algorithm to detect shilling attacks for online products in hybrid recommender system
  36. Computing the inverse of cardinal direction relations between regions
  37. Human-centered artificial intelligence-based ice hockey sports classification system with web 4.0
  38. Construction of an IoT customer operation analysis system based on big data analysis and human-centered artificial intelligence for web 4.0
  39. An improved Jaya optimization algorithm with ring topology and population size reduction
  40. Review Articles
  41. A review on voice pathology: Taxonomy, diagnosis, medical procedures and detection techniques, open challenges, limitations, and recommendations for future directions
  42. An extensive review of state-of-the-art transfer learning techniques used in medical imaging: Open issues and challenges
  43. Special Issue: Explainable Artificial Intelligence and Intelligent Systems in Analysis For Complex Problems and Systems
  44. Tree-based machine learning algorithms in the Internet of Things environment for multivariate flood status prediction
  45. Evaluating OADM network simulation and an overview based metropolitan application
  46. Radiography image analysis using cat swarm optimized deep belief networks
  47. Comparative analysis of blockchain technology to support digital transformation in ports and shipping
  48. IoT network security using autoencoder deep neural network and channel access algorithm
  49. Large-scale timetabling problems with adaptive tabu search
  50. Eurasian oystercatcher optimiser: New meta-heuristic algorithm
  51. Trip generation modeling for a selected sector in Baghdad city using the artificial neural network
  52. Trainable watershed-based model for cornea endothelial cell segmentation
  53. Hessenberg factorization and firework algorithms for optimized data hiding in digital images
  54. The application of an artificial neural network for 2D coordinate transformation
  55. A novel method to find the best path in SDN using firefly algorithm
  56. Systematic review for lung cancer detection and lung nodule classification: Taxonomy, challenges, and recommendation future works
  57. Special Issue on International Conference on Computing Communication & Informatics
  58. Edge detail enhancement algorithm for high-dynamic range images
  59. Suitability evaluation method of urban and rural spatial planning based on artificial intelligence
  60. Writing assistant scoring system for English second language learners based on machine learning
  61. Dynamic evaluation of college English writing ability based on AI technology
  62. Image denoising algorithm of social network based on multifeature fusion
  63. Automatic recognition method of installation errors of metallurgical machinery parts based on neural network
  64. An FCM clustering algorithm based on the identification of accounting statement whitewashing behavior in universities
  65. Emotional information transmission of color in image oil painting
  66. College music teaching and ideological and political education integration mode based on deep learning
  67. Behavior feature extraction method of college students’ social network in sports field based on clustering algorithm
  68. Evaluation model of multimedia-aided teaching effect of physical education course based on random forest algorithm
  69. Venture financing risk assessment and risk control algorithm for small and medium-sized enterprises in the era of big data
  70. Interactive 3D reconstruction method of fuzzy static images in social media
  71. The impact of public health emergency governance based on artificial intelligence
  72. Optimal loading method of multi type railway flatcars based on improved genetic algorithm
  73. Special Issue: Evolution of Smart Cities and Societies using Emerging Technologies
  74. Data mining applications in university information management system development
  75. Implementation of network information security monitoring system based on adaptive deep detection
  76. Face recognition algorithm based on stack denoising and self-encoding LBP
  77. Research on data mining method of network security situation awareness based on cloud computing
  78. Topology optimization of computer communication network based on improved genetic algorithm
  79. Implementation of the Spark technique in a matrix distributed computing algorithm
  80. Construction of a financial default risk prediction model based on the LightGBM algorithm
  81. Application of embedded Linux in the design of Internet of Things gateway
  82. Research on computer static software defect detection system based on big data technology
  83. Study on data mining method of network security situation perception based on cloud computing
  84. Modeling and PID control of quadrotor UAV based on machine learning
  85. Simulation design of automobile automatic clutch based on mechatronics
  86. Research on the application of search algorithm in computer communication network
  87. Special Issue: Artificial Intelligence based Techniques and Applications for Intelligent IoT Systems
  88. Personalized recommendation system based on social tags in the era of Internet of Things
  89. Supervision method of indoor construction engineering quality acceptance based on cloud computing
  90. Intelligent terminal security technology of power grid sensing layer based upon information entropy data mining
  91. Deep learning technology of Internet of Things Blockchain in distribution network faults
  92. Optimization of shared bike paths considering faulty vehicle recovery during dispatch
  93. The application of graphic language in animation visual guidance system under intelligent environment
  94. Iot-based power detection equipment management and control system
  95. Estimation and application of matrix eigenvalues based on deep neural network
  96. Brand image innovation design based on the era of 5G internet of things
  97. Special Issue: Cognitive Cyber-Physical System with Artificial Intelligence for Healthcare 4.0.
  98. Auxiliary diagnosis study of integrated electronic medical record text and CT images
  99. A hybrid particle swarm optimization with multi-objective clustering for dermatologic diseases diagnosis
  100. An efficient recurrent neural network with ensemble classifier-based weighted model for disease prediction
  101. Design of metaheuristic rough set-based feature selection and rule-based medical data classification model on MapReduce framework
Sign up now to receive a 20% welcome discount
Institutional Access
How does access work?
Have an idea on how to improve our website?
Please write us.
© 2025 De Gruyter Brill
Downloaded on 7.9.2025 from https://www.degruyterbrill.com/document/doi/10.1515/jisys-2022-0064/html
Scroll to top button