检索结果-内蒙古大学图书馆

Machine learning at the Grid Edge: data-driven Impedance Models for Model-Free Inverters

ieee TRANSACTIONS ON POWER ELECTRONICS 2024年第8期39卷 10465-10481页

作者： Li, Yufei Liao, Yicheng Zhao, Liang Chen, Minjie Wang, Xiongfei Nordstrom, Lars Mittal, Prateek Poor, H. Vincent Princeton Univ Dept Elect & Comp Engn Princeton NJ 08544 USA Princeton Univ Andlinger Ctr Energy & Environm Princeton NJ 08544 USA Xi An Jiao Tong Univ Sch Elect Engn State Key Lab Elect Insulat & Power Equipment Xian 710049 Peoples R China Energinet Fredericia 7000 Denmark Aalborg Univ Dept Energy AAU Energy Aalborg 9220 Denmark

It is envisioned that the future electric grid will be underpinned by a vast number of smart inverters linking renewables at the grid edge. These inverters' dynamics are typically characterized as impedances, which are crucial for ensuring grid stability and resiliency. However, the physical implementation of these inverters may vary widely and may be kept confidential. Existing analytical impedance models require a complete and precise understanding of system parameters. They can hardly capture the complete electrical behavior when the inverters are performing complex functions. Online impedance measurements for many inverters across multiple operating points are impractical. To address these issues, we present the InvNet, a machine learning framework capable of characterizing inverter impedance patterns across a wide operation range, even with limited impedance data. Leveraging transfer learning, the InvNet can extrapolate from physics-based models to real-world ones and from one inverter to another with the same control framework but different control parameters with very limited data. This framework demonstrates machine learning as a powerful tool for modeling and analyzing black-box characteristics of grid-tied inverter systems that cannot be accurately described by traditional analytical methods, such as inverters under model-predictive control. Comprehensive evaluations were conducted to verify the effectiveness of the InvNet.

关键词： Grid edge impedance machine learning model-free inverter transfer learning

来源：评论

学校读者我要写书评

暂无评论

Model Reconstruction Based on Physical-Informed Polynomial Neural ODEs 13

Model Reconstruction Based on Physical-Informed Polynomial N...

引用

13th ieee data driven control and learning systems conference, DDCLS 2024

作者： Chen, Zhang Zhu, Wei Li, Jingsui Bian, Hanlin Pei, Chao School of Computer Science and Technology Chongqing University of Posts and Telecommunications Chongqing400065 China Chongqing University of Posts and Telecommunications Key Laboratory of Intelligent Analysis and Decision on Complex Systems Chongqing400065 China

ISBN: (纸本)9798350361674

Neural networks are capable function approximators for data, but their lack of interpretability poses challenges. Physical-informed Polynomial Neural Ordinary Differential Equations (PIPNODEs) are proposed. They combine the constraints of prior knowledge of the data with interpretable Polynomial Neural Ordinary Differential Equations (PolyNODEs) to discover or approximate the governing equations. In this work, Neural Ordinary Differential Equations (NODEs) are categorized into two layers: a physical-informed term, which represents the priori knowledge of the data, and a discrepancy term, which is derived from the approximation of PolyNODEs. Pruning strategies are also used to remove redundant terms with coefficients lower than a certain threshold. We demonstrate the effectiveness and flexibility of this framework in a series of numerical algorithms with varying levels of prior knowledge. We found that having correct but incomplete physical information accelerates training and approaching thresholds compared to having incorrect physical information. © 2024 ieee.

关键词： data modeling discprepancy modeling interpretability Neural Ordinary Differential Equations physics-informed machine learning polynomial neural network

来源：评论

学校读者我要写书评

暂无评论

Automatic Extraction Technology of Ionospheric Vertical data with Handwritten Font 13

Automatic Extraction Technology of Ionospheric Vertical Data...

引用

13th ieee data driven control and learning systems conference, DDCLS 2024

作者： Su, Guichang Zhang, Ruikun Liu, Xiangpeng School of Mathematics and Physics Qingdao University of Science and Technology Qingdao266061 China

ISBN: (纸本)9798350361674

To solve the problems such as large size, dense text area and various handwriting styles, an automatic data extraction technology based on DB and CRNN algorithms is proposed, which mainly includes four modules: image preprocessing, form text detection, text recognition and layout processing. Firstly, contour extraction and tilt correction are used to preprocess the scanned images of handwriting vertical data. Then, Hough transform is used for table detection and DB algorithm is used for text detection. The image segmentation function is added before DB detection algorithm to improve detection accuracy. Finally, CRNN algorithm is used for text recognition, and coordinate fusion algorithm is used to save the recognition results into Excel standardized format to realize automatic data extraction and saving. The experimental results show that the text detection recall rate of the proposed algorithm is 96.77%, and the F-value of the comprehensive evaluation index of text recognition is 95.2%. Compared with other algorithms, its effectiveness is verified. Therefore, the automatic data extraction technology proposed in this paper has high practicability and can meet the actual needs of engineering applications. © 2024 ieee.

关键词： Hough transforms

来源：评论

学校读者我要写书评

暂无评论

Defect Detection of Composite Materials Using Channel Attention and Convolutional Autoencoder 13

Defect Detection of Composite Materials Using Channel Attent...

引用

13th ieee data driven control and learning systems conference, DDCLS 2024

作者： Jiang, Jiahao Jiang, Zhili Wang, Fumin Yao, Yuan Liu, Yi Institute of Process Equipment and Control Engineering Zhejiang University of Technology Hangzhou310023 China National Tsing Hua University Department of Chemical Engineering Hsinchu30013 Taiwan

ISBN: (纸本)9798350361674

As an effective non-destructive technology, infrared thermography is often used to detect inner defects in composites. However, non-homogeneous background in the thermal images reduces image quality inevitably. A squeeze and excitation convolutional autoencoder thermography (SE-CAT) technique is introduced for detecting underlying defects. Specifically, the channel attention mechanism assigns weights to different convolutional channels to highlight certain prominent features. As a result, the defect information can be extracted in a targeted manner using the convolutional autoencoder module. Finally, an experiment of carbon fiber reinforced composite materials shows good defect detection performance of the proposed SE-CAT method. © 2024 ieee.

关键词： Carbon fiber reinforced plastics

来源：评论

学校读者我要写书评

暂无评论

Hybrid Physics-Inspired Machine learning Framework for Predictive Maintenance of Forklift Chains: Leveraging Sensor data Characteristics 11

Hybrid Physics-Inspired Machine Learning Framework for Predi...

引用

11th International conference on control, Mechatronics and Automation (ICCMA)

作者： Akcatepe, Osman Moeckel, Michael Univ Appl Sci Aschaffenburg Lab Hybrid Modelling Aschaffenburg Germany

ISBN: (纸本)9798350315684

Bridging the gap between physics-based modeling and data-driven machine learning promises to reduce the amount of training data required and to improve explainability in predictive maintenance applications. For a small fleet of industrial forklift trucks, we develop a physically inspired framework for predicting remaining useful life (RUL) for selected components by integrating physically motivated feature extraction, degradation modelling and machine learning. The discussed approach is promising for situations of limited data availability or large data heterogeneity, which often occurs in fleets of customized vehicles optimized for particular tasks.

关键词： Physics-inspired machine learning Feature engineering data-driven modeling Forklifts Predictive Maintenance Remaining Useful Life (RUL) Industry 4.0

来源：评论

学校读者我要写书评

暂无评论

Reinforcement learning for Fuzzy Structured Adaptive Optimal control of Discrete-Time Nonlinear Complex Networks

引用

ieee TRANSACTIONS ON FUZZY systems 2024年第11期32卷 6035-6043页

作者： Wu, Tao Cao, Jinde Xiong, Lianglin Park, Ju H. Lam, Hak-Keung Southeast Univ Sch Math Nanjing 210096 Peoples R China Southeast Univ Frontiers Sci Ctr Mobile Informat Commun & Secur Nanjing 210096 Peoples R China Purple Mt Labs Nanjing 211111 Peoples R China Yunnan Open Univ Sch Media & Informat Engn Kunming 650504 Peoples R China Yeungnam Univ Dept Elect Engn Gyongsan 38541 South Korea Kings Coll London Dept Engn London WC2R 2LS England

This article focuses on fuzzy structural adaptive optimal control issue of discrete-time nonlinear complex networks (CNs) via adopting the reinforcement learning (RL) and Takagi-Sugeno fuzzy modeling approaches, where the control gains are subjected to structured constraints. In accordance with the Bellman optimality theory, the modified fuzzy coupled algebraic Riccati equations (CAREs) are constructed for discrete-time fuzzy CNs, while the modified fuzzy CAREs are difficult to solve directly through mathematical approaches. Then, a model-based offline learning iteration algorithm is developed to solve the modified fuzzy CAREs, where the network dynamics information is needed. Moreover, a novel data-driven off-policy RL algorithm is given to compute the modified fuzzy CAREs, and the structural optimal solutions can be obtained directly by using the collected state and input data in the absence of the network dynamics information. Furthermore, the convergence proofs of the presented learning algorithms are provided. In the end, the validity and practicability of the theoretical results are explicated via two numerical simulations.

关键词： Adaptive optimal control discrete-time nonlinear complex networks fuzzy coupled algebraic Riccati equations (CAREs) reinforcement learning (RL) structural learning iteration Adaptive optimal control discrete-time nonlinear complex networks fuzzy coupled algebraic Riccati equations (CAREs) reinforcement learning (RL) structural learning iteration

来源：评论

学校读者我要写书评

暂无评论

A data-driven Prediction Method for Lithium-ion Batteries based on Variational Bayes and Particle Filtering 13

A Data-Driven Prediction Method for Lithium-ion Batteries ba...

引用

13th ieee data driven control and learning systems conference, DDCLS 2024

作者： Wang, Hao Deng, Yi Hou, Ru Xin Liu, Qie School of Automation Chongqing University Chongqing400044 China

ISBN: (纸本)9798350361674

The remaining useful life (RUL) prediction and state of health (SOH) assessment of lithium-ion batteries are the key technologies for achieving equipment prediction and health management (PHM). Accurate RUL prediction results can serve as evaluation indicators for battery management systems(BMS), ensuring their safe and stable operation. This paper proposes a model parameter estimation method based on particle filtering (PF) and variational Bayesian (VB) to address the issues of high degradation noise and small data sample size in lithium-ion batteries, which lead to large prediction model errors and poor real-time performance. Considering the impact of uncertainty in noise, this paper uses a hybrid model to describe the data degradation process of lithium-ion batteries at different time intervals and probabilistically describes the measurement noise and model parameters. On this basis, the variational Bayes method is used to perform posterior estimation on each parameter, and the particle filter optimization method is used to dynamically update the values of the model parameters. This paper conducts experimental verification using publicly available datasets from NASA. The results indicate that this method can achieve good predictive performance. © 2024 ieee.

关键词： Particle Filtering Remaining Useful Life Prediction Variational Bayes

来源：评论

学校读者我要写书评

暂无评论

STDP-driven Development of Attention-Based People Detection in Spiking Neural Networks

引用

ieee TRANSACTIONS ON COGNITIVE AND DEVELOPMENTAL systems 2024年第1期16卷 380-387页

作者： Safa, Ali Ocket, Ilja Bourdoux, Andre Sahli, Hichem Catthoor, Francky Gielen, Georges G. E. IMEC B-3001 Leuven Belgium Katholieke Univ Leuven ESAT B-3000 Leuven Belgium VUB ETRO B-1050 Brussels Belgium

This letter provides, to the best of our knowledge, a first analysis of how biologically plausible spiking neural networks (SNNs) equipped with spike-timing-dependent plasticity (STDP) can learn to detect people on the fly from nonindependent and identically distributed (non-i.i.d) streams of retina-inspired, event camera data. Our system works as follows. First, a short sequence of event data, capturing a walking human from a flying drone, is forwarded in its natural order to an SNN-STDP system, which also receives teacher spiking signals from the neural activity readout block. Then, when the end of the learning sequence is reached, the learned system is assessed on testing sequences. In addition, we also present a new interpretation of anti-Hebbian plasticity as an overfitting control mechanism and provide experimental demonstrations of our findings. This work contributes to the study of attention-based development and perception in bioinspired systems.

关键词： Bioinspired vision continual learning (CL) spike-timing-dependent plasticity (STDP)

来源：评论

学校读者我要写书评

暂无评论

Mechanism Enhanced data-driven Method for Reliability Improvement of Load Margin Estimation

引用

ieee TRANSACTIONS ON POWER systems 2024年第2期39卷 3715-3724页

作者： Gao, Han Wang, Lixin Yang, Deyou Cai, Guowei Northeast Elect Power Univ Sch Elect Engn Jilin 132012 Jilin Peoples R China Harbin Univ Sci & Technol Sch Elect & Elect Engn Harbin 150000 Peoples R China

The data-driven method for security monitoring is promising for power system voltage stability estimation, whereas the reliability of such method is greatly affected due to the generalizability weakened by operational variability. In this article, a mechanism knowledge enhanced machine learning method for load margin estimation is established by integrating power system mechanism into intelligent neural network (NN) model. Firstly, the correlation between the load power factor and load margin explored fromcurve P-V were analyzed, from which the mechanism knowledge of voltage stability is extracted. Then, the mathematical logical representation of mechanism knowledge is given, which was further integrated into the loss function of the NN model. The proposed enhanced model constrains the NN model to provide mechanistically consistent results, as well as helps achieve an excellent generalizability performance. The effectiveness of the proposed method is verified by numerical simulations using the ieee 39-bus test system and the ieee 145-bus test system, the results of which shows that the generalization ability of the enhanced NN is significantly improved with the proposed method, thus the accuracy, reliability and robustness of the load margin estimation are greatly enhanced.

关键词： Intelligent load margin estimation mechanism knowledge neural network generalization ability

来源：评论

学校读者我要写书评

暂无评论

Robust Variational Bayesian Filter for Accurate Flight Altitude Estimation with Uncertain Measurement Noise 13

Robust Variational Bayesian Filter for Accurate Flight Altit...

引用

13th ieee data driven control and learning systems conference, DDCLS 2024

作者： Guo, Caiqin Shen, Kai Huang, Deqing Institude of Systems Science and Technology Southwest Jiaotong University Chengdu611756 China

ISBN: (纸本)9798350361674

Accurate altitude information is essential for achieving safe flight, especially during take-off and landing. The Kalman filter (KF) is an optimal state estimation method that can be utilized for flight altitude estimation. In practical scenario, considering the influence of sensor characteristics and changes in the external environment, uncertain measurement phenomena may arise, which leads to the deviation between the estimated filter state and the true state, leading to filter divergence. In this paper, a robust measurement distribution-based variational Bayesian filtering algorithm is presented as a solution to this issue. The heavy-tailed properties of the measurement noise are modeled using the Student-t distribution, while the variational Bayesian approach is employed to dynamically estimate the statistical characteristics of the measurement noise in real time. Simulation and experiment results demonstrate that the proposed algorithm outperforms both the traditional KF algorithm and the variational Bayesian Adaptive KF (VBAKF) algorithm when dealing with uncertain measurement noise. © 2024 ieee.

关键词： Kalman filters

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：