Mengru Ma, Yingjie Chen, Qingbin Yu, Zhongxin Du, Wei Ding
The geo-referenced video consists of space-temporal information such as time, spatial location, camera shooting direction, camera viewing angle, viewable distance, etc. This type of video is widely applicable with sensors loaded on video capture devices. The application of geo-referenced video queries is increasingly popular recently (e.g., travel recommendation, intelligent transportation, road anomaly detection). And each of them needs to realize the query process of geo-referenced video at a specific time or spatial location. However, existing mobile video indexing methods still have room for improvement. There still exist problems with efficiency and accuracy. In this paper, we proposed a novel indexing method named MVideoIndex. MVideoIndex can process point or range queries quickly by utilizing Minimum Bounding Tilted Rectangle (MBTR) in leaf nodes based on the linear change of movement direction in geo-referenced videos. For representing the viewable regions of geo-referenced videos along the trajectory better, we constructed the index with a memory buffer limit to avoid the situation, where the query target falls into a large index and is inconvenient to query. We experimentally analyzed the performance of MVideoIndex and the state-of-art video index method GeoVideoIndex to verify our theory. The performance shows that MVideoIndex is capable of reducing the index construction time and query time, presenting a better performance than other methods. We further compared the impact of the memory buffer threshold size on query efficiency and found that the optimal memory buffer threshold size is about 8-kilometer Byte. We also conducted experiments to explore the effect of MVideoIndex and GeoVideoIndex on different datasets and found a more suitable application scenario for MVideoIndex.
{"title":"MVideoIndex: Querying and Indexing of Geo-referenced Videos","authors":"Mengru Ma, Yingjie Chen, Qingbin Yu, Zhongxin Du, Wei Ding","doi":"10.1145/3546000.3546017","DOIUrl":"https://doi.org/10.1145/3546000.3546017","url":null,"abstract":"The geo-referenced video consists of space-temporal information such as time, spatial location, camera shooting direction, camera viewing angle, viewable distance, etc. This type of video is widely applicable with sensors loaded on video capture devices. The application of geo-referenced video queries is increasingly popular recently (e.g., travel recommendation, intelligent transportation, road anomaly detection). And each of them needs to realize the query process of geo-referenced video at a specific time or spatial location. However, existing mobile video indexing methods still have room for improvement. There still exist problems with efficiency and accuracy. In this paper, we proposed a novel indexing method named MVideoIndex. MVideoIndex can process point or range queries quickly by utilizing Minimum Bounding Tilted Rectangle (MBTR) in leaf nodes based on the linear change of movement direction in geo-referenced videos. For representing the viewable regions of geo-referenced videos along the trajectory better, we constructed the index with a memory buffer limit to avoid the situation, where the query target falls into a large index and is inconvenient to query. We experimentally analyzed the performance of MVideoIndex and the state-of-art video index method GeoVideoIndex to verify our theory. The performance shows that MVideoIndex is capable of reducing the index construction time and query time, presenting a better performance than other methods. We further compared the impact of the memory buffer threshold size on query efficiency and found that the optimal memory buffer threshold size is about 8-kilometer Byte. We also conducted experiments to explore the effect of MVideoIndex and GeoVideoIndex on different datasets and found a more suitable application scenario for MVideoIndex.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123576837","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
We propose a parallel exact diagonalization method for solving the large-scale Hubbard model. The core of this algorithm is the parallelization of the Lanczos algorithm, for which we propose a hierarchical communication model and a fast strategy for finding nonzero elements of large-scale matrix, starting only from the symmetry of Hamiltonian matrix. The effect of our parallel algorithm was tested on the Tianhe-2 supercomputer, where the strong scaling efficiency could reach 53% for 30,000 cores in a 140-billion dimensional matrix, and the weak scaling efficiency remained above 40% for 60,000 cores in a 730-billion dimensional matrix.
{"title":"Large-scale parallel exact diagonalization algorithm of the Hubbard model on Tianhe-2 supercomputer","authors":"Biao Li, Jie Liu","doi":"10.1145/3546000.3546001","DOIUrl":"https://doi.org/10.1145/3546000.3546001","url":null,"abstract":"We propose a parallel exact diagonalization method for solving the large-scale Hubbard model. The core of this algorithm is the parallelization of the Lanczos algorithm, for which we propose a hierarchical communication model and a fast strategy for finding nonzero elements of large-scale matrix, starting only from the symmetry of Hamiltonian matrix. The effect of our parallel algorithm was tested on the Tianhe-2 supercomputer, where the strong scaling efficiency could reach 53% for 30,000 cores in a 140-billion dimensional matrix, and the weak scaling efficiency remained above 40% for 60,000 cores in a 730-billion dimensional matrix.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129088621","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Aiming to strengthen the stability of operation and maintenance of the urban rail transit network cloud platform at this stage, it is emerging to solve the security mechanism of the intelligent urban railway cloud platform. In this paper, we proposed a zero-trust network security solution for the rail transit system network construction. First, we built a zero-trust network construction for smart city rail transit at the architecture level, it can break the phenomenon of information security silo of rail transit line platform and minimize the system security risk based on a zero-trust network. Next, we focus on building a cloud security brain for urban rail transit networks and proposed the self-learning trust algorithm for a zero-trust network. Specifically, we illustrated the modified network model and constructed a dynamic updating user trust profile as the trustworthy access list. The parameters of the self-learning trust algorithm consist of the state, available chain road bandwidth, waiting for queue state of network traffic, linkage actions, and so on. We adopted a dynamic self-learning strategy for adjusting mitigation policy, the learning step predicted the state of the predetermined congestion and selected the rich links for execution. Finally, experiments show the efficiency of our secure mechanism of railway cloud platform based on zero-trust security architecture.
{"title":"Secure Mechanism of Intelligent Urban Railway Cloud Platform Based on Zero-trust Security Architecture","authors":"Y. Qlu","doi":"10.1145/3546000.3546015","DOIUrl":"https://doi.org/10.1145/3546000.3546015","url":null,"abstract":"Aiming to strengthen the stability of operation and maintenance of the urban rail transit network cloud platform at this stage, it is emerging to solve the security mechanism of the intelligent urban railway cloud platform. In this paper, we proposed a zero-trust network security solution for the rail transit system network construction. First, we built a zero-trust network construction for smart city rail transit at the architecture level, it can break the phenomenon of information security silo of rail transit line platform and minimize the system security risk based on a zero-trust network. Next, we focus on building a cloud security brain for urban rail transit networks and proposed the self-learning trust algorithm for a zero-trust network. Specifically, we illustrated the modified network model and constructed a dynamic updating user trust profile as the trustworthy access list. The parameters of the self-learning trust algorithm consist of the state, available chain road bandwidth, waiting for queue state of network traffic, linkage actions, and so on. We adopted a dynamic self-learning strategy for adjusting mitigation policy, the learning step predicted the state of the predetermined congestion and selected the rich links for execution. Finally, experiments show the efficiency of our secure mechanism of railway cloud platform based on zero-trust security architecture.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133552593","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Silicon interposer enables 2.5D stacking of memory chips and processor chips to pursue advanced memory access performance. In interposer-based system, different traffic transfers through network-on-interposer (NoI) lays on the silicon interposer which makes NoI throughput important to transmit the mass of data. However, the performance of the existing topology varies under different traffic patterns. In this paper, we use reinforcement learning (RL) is adapted to further optimize the throughput of NoI in various traffic. We design a dedicated RL framework for NoI enviroment to enable performance improvement. Three algorithms are used to maximize the throughput as well as reward in the RL Model. Simulation results demonstrate that the proposed RL approach provide higher throughput both in memory traffic and coherence traffic.
{"title":"Reinforcement Learning Enabled Throughput Optimization for Interconnection Networks of Interposer-based system","authors":"Shuhao Ling, Huaien Gao, Jiasong Chen, Dawei Liu","doi":"10.1145/3546000.3546002","DOIUrl":"https://doi.org/10.1145/3546000.3546002","url":null,"abstract":"Silicon interposer enables 2.5D stacking of memory chips and processor chips to pursue advanced memory access performance. In interposer-based system, different traffic transfers through network-on-interposer (NoI) lays on the silicon interposer which makes NoI throughput important to transmit the mass of data. However, the performance of the existing topology varies under different traffic patterns. In this paper, we use reinforcement learning (RL) is adapted to further optimize the throughput of NoI in various traffic. We design a dedicated RL framework for NoI enviroment to enable performance improvement. Three algorithms are used to maximize the throughput as well as reward in the RL Model. Simulation results demonstrate that the proposed RL approach provide higher throughput both in memory traffic and coherence traffic.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125667510","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The explosion of massive data poses a severe challenge to the storage cost and data storage reliability of traditional storage systems. As the number of storage nodes increases in a distributed storage system, the node failure probability increases. Erasure code technology as a redundancy strategy can greatly save the cost of storage space while providing security for data compared with multi-copy technology. In real large-scale data centers, the repair costs vary due to differences in storage node attributes and link status changes between nodes. In the repair process based on the erasure code mechanism, it is required to select a certain number of provider nodes from the surviving nodes to transmit data to the newcomer nodes, which leads to the problem of selecting nodes to participate in the repair process. In this paper, a tree-type repair scheme considering node selection (TR-NS) is designed, including an algorithm for selecting two types of nodes according to nodes performance and an algorithm for establishing the optimal repair tree to complete the reconstruction of the failure data. The experimental results show that the repair time of the TR-RS scheme proposed in this paper is significantly reduced.
{"title":"A Node Selection Scheme for Data Repair Using Erasure Code in Distributed Storage System","authors":"Yuejin Huang, Miao Ye, Yue Cai","doi":"10.1145/3546000.3546003","DOIUrl":"https://doi.org/10.1145/3546000.3546003","url":null,"abstract":"The explosion of massive data poses a severe challenge to the storage cost and data storage reliability of traditional storage systems. As the number of storage nodes increases in a distributed storage system, the node failure probability increases. Erasure code technology as a redundancy strategy can greatly save the cost of storage space while providing security for data compared with multi-copy technology. In real large-scale data centers, the repair costs vary due to differences in storage node attributes and link status changes between nodes. In the repair process based on the erasure code mechanism, it is required to select a certain number of provider nodes from the surviving nodes to transmit data to the newcomer nodes, which leads to the problem of selecting nodes to participate in the repair process. In this paper, a tree-type repair scheme considering node selection (TR-NS) is designed, including an algorithm for selecting two types of nodes according to nodes performance and an algorithm for establishing the optimal repair tree to complete the reconstruction of the failure data. The experimental results show that the repair time of the TR-RS scheme proposed in this paper is significantly reduced.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127687722","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Modern x86-64 processors have strong performance due to long vector units. Therefore long vector units are widely used in CNN-like neural network model inference on modern x86-64 processors. However the performance of GNN inference on modern x86-64 processors is poor. Unfortunately, with the development of GNNs and the increase of graph datasets, GNN inference performance meets the serious challenge on x86-64 processors. In this paper, we study the problem of poorly optimized DGL-based GAT models on the x86-64 platform, and analyze the main performance bottlenecks in this case. In order to optimize the performance of DGL on the two main x86-64 platform CPUs of Intel and AMD, we implement a simple and effective task allocator to balance the task load among multiple cores and use vector instructions to optimize the core operators in DGL. In addition, we also propose corresponding optimization ideas for the NUMA architecture. The experimental results show that our optimization method improves the performance of the basic DGL version by up to 3.12x and 2.6x on Intel and AMD platforms.
{"title":"Optimize DGL Operations on x86-64 Multi-Core Processors","authors":"Chao Liu, Huayou Su, Y. Dou, Qinglin Wang","doi":"10.1145/3546000.3546018","DOIUrl":"https://doi.org/10.1145/3546000.3546018","url":null,"abstract":"Modern x86-64 processors have strong performance due to long vector units. Therefore long vector units are widely used in CNN-like neural network model inference on modern x86-64 processors. However the performance of GNN inference on modern x86-64 processors is poor. Unfortunately, with the development of GNNs and the increase of graph datasets, GNN inference performance meets the serious challenge on x86-64 processors. In this paper, we study the problem of poorly optimized DGL-based GAT models on the x86-64 platform, and analyze the main performance bottlenecks in this case. In order to optimize the performance of DGL on the two main x86-64 platform CPUs of Intel and AMD, we implement a simple and effective task allocator to balance the task load among multiple cores and use vector instructions to optimize the core operators in DGL. In addition, we also propose corresponding optimization ideas for the NUMA architecture. The experimental results show that our optimization method improves the performance of the basic DGL version by up to 3.12x and 2.6x on Intel and AMD platforms.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132824271","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Y. Zhu, Haixing Zhao, Jianqiang Huang, Xiaoying Wang
In real life, there are many cases that cannot be described by the network abstracted as the graph, but can be described perfectly by the hypernetwork abstracted as the hypergraph. Different from the network, the hypernetwork structure is more complex and poses a great challenge to the existing network representation learning methods. Therefore, in order to overcome the challenge of the hypernetwork structure, a hypernetwork representation learning method with the transformation strategy is proposed. Firstly, as three types of transformation strategies from the hypergraph to the graph, line graph, incidence graph and 2-section graph are combined into three types of integral graphs with the hyperedge information, namely incidence graph + 2-section graph, line graph + incidence graph and line graph + incidence graph + 2-section graph. Secondly, a shallow neural network algorithm is trained respectively on five types of networks abstracted as incidence graph, 2-section graph, incidence graph + 2-section graph, line graph + incidence graph and line graph +incidence graph + 2-section graph to obtain node representation vectors. Finally, the evaluation experiment is conducted on four different types of hypernetwork datasets. The experimental results demonstrate that the node classification performance of 2-section graph is better than that of other graphs, and the link prediction performance of incidence graph + 2-section graph is better than that of other graphs.
{"title":"Hypernetwork Representation Learning With the Transformation Strategy","authors":"Y. Zhu, Haixing Zhao, Jianqiang Huang, Xiaoying Wang","doi":"10.1145/3546000.3546020","DOIUrl":"https://doi.org/10.1145/3546000.3546020","url":null,"abstract":"In real life, there are many cases that cannot be described by the network abstracted as the graph, but can be described perfectly by the hypernetwork abstracted as the hypergraph. Different from the network, the hypernetwork structure is more complex and poses a great challenge to the existing network representation learning methods. Therefore, in order to overcome the challenge of the hypernetwork structure, a hypernetwork representation learning method with the transformation strategy is proposed. Firstly, as three types of transformation strategies from the hypergraph to the graph, line graph, incidence graph and 2-section graph are combined into three types of integral graphs with the hyperedge information, namely incidence graph + 2-section graph, line graph + incidence graph and line graph + incidence graph + 2-section graph. Secondly, a shallow neural network algorithm is trained respectively on five types of networks abstracted as incidence graph, 2-section graph, incidence graph + 2-section graph, line graph + incidence graph and line graph +incidence graph + 2-section graph to obtain node representation vectors. Finally, the evaluation experiment is conducted on four different types of hypernetwork datasets. The experimental results demonstrate that the node classification performance of 2-section graph is better than that of other graphs, and the link prediction performance of incidence graph + 2-section graph is better than that of other graphs.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128388957","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Sijin Cheng, Y. Zhou, Hao Sun, Junjian Tian, He Zhu, Jinfu Zhu, Hanliang Liao
In order to study the thermal mechanical characteristics of the broken line of the connecting pipe and pipeline joint of the transmission line, the tensile test of the steel cored aluminum strand steel wire, the thermodynamic analysis of the broken line joint of the connecting pipe and the metallographic test of the steel core of the broken line joint were carried out successively. The tensile strength of the steel core of the steel cored aluminum strand, the temperature load curve at different times and the metallographic test results of the steel core of the broken line joint were obtained, Finally, the exposed section of steel core is tested by metallography. The results show that the tensile strength of steel cored aluminum strand meets the standard, the heating of connecting pipe caused by conductor current will affect the calculated breaking force of steel core, and the exposed section of steel core has been running at high temperature for a period of time before being pulled off.
{"title":"Analysis of thermal-mechanical characteristics and metallographic detection of broken connection of transmission line","authors":"Sijin Cheng, Y. Zhou, Hao Sun, Junjian Tian, He Zhu, Jinfu Zhu, Hanliang Liao","doi":"10.1145/3546000.3546030","DOIUrl":"https://doi.org/10.1145/3546000.3546030","url":null,"abstract":"In order to study the thermal mechanical characteristics of the broken line of the connecting pipe and pipeline joint of the transmission line, the tensile test of the steel cored aluminum strand steel wire, the thermodynamic analysis of the broken line joint of the connecting pipe and the metallographic test of the steel core of the broken line joint were carried out successively. The tensile strength of the steel core of the steel cored aluminum strand, the temperature load curve at different times and the metallographic test results of the steel core of the broken line joint were obtained, Finally, the exposed section of steel core is tested by metallography. The results show that the tensile strength of steel cored aluminum strand meets the standard, the heating of connecting pipe caused by conductor current will affect the calculated breaking force of steel core, and the exposed section of steel core has been running at high temperature for a period of time before being pulled off.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130152392","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The proposal and development of a multi-energy collaborative low-carbon energy management system put forward new solutions for gradually increasing the proportion of new energy power generation, optimizing energy utilization efficiency, and reducing greenhouse gas emissions. Because this paper fully considers the differences in dynamic characteristics among various energy sources, and the optimal scheduling is also different according to the load under different time scales and the prediction accuracy results of new energy. In the multi-energy coordinated comprehensive energy park system of cold, heat, electricity, and gas, the multi time scale scheduling strategy reasonably coordinates a variety of equipment to optimize its operation, which is of great significance to the economic and reliable operation of the system, Based on the detailed modelling of the equipment, combined with the time-sharing price and time-sharing natural gas price, and considering the multi-energy complementary operation, this paper dynamically adjusts the equipment operation state to deal with the small-scale uncertain fluctuation of load and new energy output, so as to meet the economic and reliable operation of the system.
{"title":"Control strategy of multi-energy collaborative low carbon energy management system","authors":"L. Zhong, Xi Hu, Jun Lei, Gaopeng Guo, Haojun Yan","doi":"10.1145/3546000.3546034","DOIUrl":"https://doi.org/10.1145/3546000.3546034","url":null,"abstract":"The proposal and development of a multi-energy collaborative low-carbon energy management system put forward new solutions for gradually increasing the proportion of new energy power generation, optimizing energy utilization efficiency, and reducing greenhouse gas emissions. Because this paper fully considers the differences in dynamic characteristics among various energy sources, and the optimal scheduling is also different according to the load under different time scales and the prediction accuracy results of new energy. In the multi-energy coordinated comprehensive energy park system of cold, heat, electricity, and gas, the multi time scale scheduling strategy reasonably coordinates a variety of equipment to optimize its operation, which is of great significance to the economic and reliable operation of the system, Based on the detailed modelling of the equipment, combined with the time-sharing price and time-sharing natural gas price, and considering the multi-energy complementary operation, this paper dynamically adjusts the equipment operation state to deal with the small-scale uncertain fluctuation of load and new energy output, so as to meet the economic and reliable operation of the system.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122444104","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The condition monitoring and fault Early Warning of wind turbine can find its faults early and reduce its failure rate and maintenance cost. This paper presents a fault diagnosis method of wind turbine generator based on residual autoencoder network (RAE). The proposed RAE has an autoencoder network structure. The encoding network is responsible for extracting the feature vector reflecting the distribution law of supervisory control and data acquisition (SCADA) data, the decoding network is responsible for reconstructing SCADA data according to the feature vector, and training the RAE network according to the reconstruction error of input data and reconstructed data. There are several shortcut connections between the corresponding layers of the encoder and decoder of the RAE. Shortcut connections introduce the shallow features in the encoder into the decoder and combines them with the deep semantic features in the decoder. Moreover, the shortcut connections allow the network to get additional supervision during back propagation process, avoiding the problem of gradient disappearance. Through the simulation analysis of the recorded data before and after generator fault, the effectiveness of the proposed RAE network for wind turbine generator fault diagnosis is verified.
{"title":"Fault Early Warning of Wind Turbine Generator based on Residual Autoencoder Network","authors":"Zhaoyang Wang","doi":"10.1145/3546000.3546029","DOIUrl":"https://doi.org/10.1145/3546000.3546029","url":null,"abstract":"The condition monitoring and fault Early Warning of wind turbine can find its faults early and reduce its failure rate and maintenance cost. This paper presents a fault diagnosis method of wind turbine generator based on residual autoencoder network (RAE). The proposed RAE has an autoencoder network structure. The encoding network is responsible for extracting the feature vector reflecting the distribution law of supervisory control and data acquisition (SCADA) data, the decoding network is responsible for reconstructing SCADA data according to the feature vector, and training the RAE network according to the reconstruction error of input data and reconstructed data. There are several shortcut connections between the corresponding layers of the encoder and decoder of the RAE. Shortcut connections introduce the shallow features in the encoder into the decoder and combines them with the deep semantic features in the decoder. Moreover, the shortcut connections allow the network to get additional supervision during back propagation process, avoiding the problem of gradient disappearance. Through the simulation analysis of the recorded data before and after generator fault, the effectiveness of the proposed RAE network for wind turbine generator fault diagnosis is verified.","PeriodicalId":196955,"journal":{"name":"Proceedings of the 6th International Conference on High Performance Compilation, Computing and Communications","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-06-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127563083","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：