Article
Statistics & Probability
Eckhard Liebscher, Ostap Okhrin
Summary: This paper explores the semi-parametric estimation of the multivariate elliptical distribution when the dimensionality increases with the sample size. We prove the almost sure convergence of the estimator and derive its convergence rates, which depend on the sample size, dimensionality, and kernel bandwidth. Additionally, we demonstrate the almost sure convergence and convergence rates of the sample covariance matrix under the Frobenius norm. Extensive simulation studies provide support for the theoretical findings.
JOURNAL OF MULTIVARIATE ANALYSIS
(2023)
Article
Statistics & Probability
Yinpu Li, Antonio R. Linero, Jared Murray
Summary: We propose a Bayesian nonparametric model based on BART for conditional distribution estimation. The model is flexible, has a default prior specification, and is computationally convenient. We introduce a method for targeted smoothing of the BART models to address the distinguished role of the response. Theoretical analysis shows that the posterior distribution of the proposed model concentrates at close to the minimax optimal rate adaptively over smoothness classes in the high-dimensional regime. We also propose a data augmentation algorithm for fitting the model, which can be easily extended to existing BART samplers. Simulation studies and an application using medical expenditure panel survey data demonstrate the performance of our methodology.
JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
(2023)
Article
Water Resources
David A. Benson, Diogo Bolster, Stephen Pankavich, Michael J. Schmidt
Summary: Traditional interpolation techniques for particle tracking involve binning and convolutional formulas using pre-determined kernels. Each particle in the cloud is a sample from the Green's function, and the kernel for interpolation should replicate the cloud itself. An iterative method is proposed to discern the form of the kernel during the process of interpolating the Green's function.
ADVANCES IN WATER RESOURCES
(2021)
Article
Mathematical & Computational Biology
Siyun Liu, Tao Yu
Summary: In this article, a method for density estimation of data with a mixture structure is proposed, which nonparametrically estimates component density functions through weighted kernel density estimation. Extensive simulation studies and real data examples demonstrate the superiority of the proposed method over existing methods in most cases.
STATISTICS IN MEDICINE
(2021)
Article
Engineering, Electrical & Electronic
Tingting Zhang, Wen Zhang, Qi Zhao, Jian Chen, Yicheng Zhang
Summary: This article proposes an adaptive particle filter method by using randomized quasi-Monte Carlo sampling to enhance computational efficiency. The method applies low-discrepancy Sobol sequence for uniform sampling and achieves higher convergence rate with fewer particles, thus greatly improving computation efficiency. Experimental results demonstrate the accuracy and scalability of the proposed method.
IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT
(2023)
Article
Computer Science, Artificial Intelligence
Zhiyuan Zhao, Xuelong Li
Summary: Crowd counting, an essential task in crowd analysis with great significance in public safety, has gained increasing attention recently. Existing methods for crowd counting using convolutional neural networks and density maps fail to accurately represent scale changes caused by perspective effects. To overcome this challenge, we propose a scale-sensitive crowd density map estimation framework, which incorporates adaptive density maps, deformable density map decoders, and auxiliary branches. Experimental results demonstrate the effectiveness of the proposed framework, and visualization shows that deformable convolutions capture the scale variation of targets.
IEEE TRANSACTIONS ON IMAGE PROCESSING
(2023)
Article
Engineering, Electrical & Electronic
Yijun Xu, Jaber Valinejad, Mert Korkali, Lamine Mili, Yajun Wang, Xiao Chen, Zongsheng Zheng
Summary: This paper proposes a Bayesian inference framework to simultaneously estimate the topology and state of a three-phase, unbalanced power distribution system using a limited number of available measurements. The framework utilizes an adaptive importance sampling procedure that efficiently recovers the full Bayesian posterior distributions of the system topology, achieving excellent performance.
IEEE TRANSACTIONS ON POWER SYSTEMS
(2022)
Article
Statistics & Probability
Y. Y. Linke, I. S. Borisov, P. S. Ruzankin
Summary: Consistent weighted least square estimators are proposed for nonparametric regression models with random regression functions. The estimators achieve uniform convergence to the random regression function with explicit upper bounds, regardless of the correlation structure of the design points. The new estimators are illustrated to be more accurate than the Nadaraya-Watson estimators through simulation examples and applied to real earthquake data in Japan from 2012 to 2021.
Article
Mathematics
Pierre Lafaye de Micheaux, Frederic Ouimet
Summary: This paper introduces five new asymmetric kernel c.d.f. estimators and compares their performance with traditional methods and other kernel c.d.f. estimators. The LogNormal and Birnbaum-Saunders kernel c.d.f. estimators are found to perform the best overall.
Article
Mathematics
Yanchun Zhao, Mengzhu Zhang, Qian Ni, Xuhui Wang
Summary: Learning density estimation is crucial for probabilistic modeling and reasoning with uncertainty. B-spline basis functions provide advantages in density estimation due to their local support and numerical computation efficiency. However, selecting the bandwidth for uniform B-splines is challenging and computationally expensive. In this study, we propose an adaptive strategy for density estimation using nonuniform B-splines by introducing an error indicator attached to each interval, which is an approximation of the local information entropy. Our numerical experiments demonstrate that the local density estimation with nonuniform B-splines outperforms the uniform B-spline, achieving better estimation results and alleviating the overfitting phenomenon.
Article
Engineering, Civil
Shahid Latif, Slobodan P. Simonovic
Summary: The joint probability modelling of storm surges and rainfall events in low-lying coastal areas is crucial for assessing compound flood risk. This study proposes a nonparametric approach using the Bernstein copula estimator and Beta kernel copula density. The research shows that neglecting the compound effect of storm surge and rainfall in coastal flood risk assessment may lead to underestimated failure probability statistics.
WATER RESOURCES MANAGEMENT
(2022)
Article
Energy & Fuels
Yanhong Luo, Xu Wang, Shijie Yan
Summary: The adaptive kernel density estimation and Cornish-Fisher series combined method is introduced to evaluate the out of limit risk of photovoltaic power distribution network. A probability distribution model and risk indexes are established for the assessment.
Article
Multidisciplinary Sciences
I Laroussi, M. Madi
Summary: The purpose of this work is to present a novel mode of convergence, complete second-order moment convergence with rate, which gives better performances and is easier to obtain than the almost complete convergence for the nonparametric estimators with kernels of the density function, distribution function, and quantile function. The proposed approach imposes less conditions on the kernel function due to the use of the mean squared error expression.
JOURNAL OF TAIBAH UNIVERSITY FOR SCIENCE
(2022)
Article
Automation & Control Systems
Minh-Lien Jeanne Nguyen, Claire Lacour, Vincent Rivoirard
Summary: This paper studies the estimation of the conditional density given observation data and an independent identically distributed sample. It provides an adaptive fully-nonparametric strategy based on kernel rules, and proposes a new fast iterative algorithm to select the bandwidth of the kernel rule. The results show that the pointwise estimator achieves quasi-optimal convergence rate in terms of both regularity and sparsity, and the computational complexity of the method is only O(dn log n).
JOURNAL OF MACHINE LEARNING RESEARCH
(2022)
Article
Mathematics
Tomas Ruzgas, Mantas Lukauskas, Gedmantas Cepkauskas
Summary: This paper explores various non-parametric multivariate density estimation techniques, the inversion formula algorithm, and probability mixture models. Different estimation methods are recommended based on dimensions and sample sizes to improve accuracy.
Article
Computer Science, Information Systems
Seyed Abbas Seifossadat, Mohammad Rastegar, Mohammad Mohammadi, Soroush Senemmar, Jie Zhang
Summary: This study proposes a hierarchical optimization model to optimize the locations and component sizes of micro-energy hubs (mu EHs) in integrated electricity and natural gas distribution systems. The planning process considers seasonal climate changes and uncertainties in demand and renewable energy generation. Simulation results demonstrate that the optimized planning model can reduce line losses, purchased electricity, and imported natural gas.
IEEE SYSTEMS JOURNAL
(2022)
Article
Engineering, Electrical & Electronic
Saeed Nejadfard-jahromi, Mohammad Mohammadi
Summary: This study presents a nonparametric data-driven VSA method for probabilistic voltage stability assessment of distribution systems. It employs Bernstein vine copula to estimate the unknown underlying dependence structure among variables and utilizes global sensitivity analysis to rank the loads and renewable energy sources according to their contribution to voltage stability. The proposed method provides an efficient approach for voltage stability assessment.
IET GENERATION TRANSMISSION & DISTRIBUTION
(2022)
Article
Engineering, Electrical & Electronic
Mohammadhadi Rouhani, Mohammad Mohammadi, Marco Aiello
Summary: This paper proposes a novel hybrid nonparametric algorithm for short-term probabilistic power flow analysis in radial distribution networks. The algorithm takes into account probabilistic nonlinear loads, unit generators, and charging patterns, while considering temporal-spatial correlation and multivariate uncertainties. Simulation results demonstrate the effectiveness of the proposed method in terms of improved prediction accuracy and reduced computation burden.
ELECTRIC POWER SYSTEMS RESEARCH
(2022)
Article
Automation & Control Systems
Shahabodin Afrasiabi, Mousa Afrasiabi, Benyamin Parang, Mohammad Mohammadi, Haidar Samet, Tomislav Dragicevic
Summary: This study proposes a new algorithm for differential protection, using fast gated recurrent neural network (FGRNN) and a loss function based on information theory concept, which can effectively distinguish between internal faults and external disturbances, and has faster and more reliable performance.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2022)
Article
Energy & Fuels
Mohammad Hossein Oboudi, Mohammad Mohammadi, Dimitris N. N. Trakas, Nikos D. D. Hatziargyriou
Summary: This paper proposes a risk-based resilience enhancement framework against earthquakes for power systems. The framework consists of four phases, including earthquake modeling, vulnerability assessment, risk assessment, and resilience enhancement. It supports the decision-making of distribution system operators for retrofitting substation components and underground cables to enhance distribution system resilience. The framework utilizes various metrics such as repair cost, customer interruption cost, and power generation cost to assess seismic risk and determine optimal retrofitting strategies for resilience enhancement.
JOURNAL OF ENERGY ENGINEERING
(2023)
Article
Engineering, Electrical & Electronic
Amir Rostami, Mohammad Mohammadi, Hadis Karimipour
Summary: This paper presents an effective reliability assessment technique for cyber-physical generation systems that incorporates cybersecurity issues and non-normal random variables with nonlinear dependencies. It adopts a Bayesian attack graph for vulnerability analysis and models the probabilistic nature of attack paths. The proposed method also assesses the capacity outage probability table in power systems using a hybrid approximate-analytical method.
IRANIAN JOURNAL OF SCIENCE AND TECHNOLOGY-TRANSACTIONS OF ELECTRICAL ENGINEERING
(2023)
Article
Engineering, Electrical & Electronic
Farshid Nasrfard, Mohammad Mohammadi, Mazaher Karimi
Summary: The degradation of power system components can be reduced by implementing preventative maintenance. Optimizing inspection and preventative maintenance rates is crucial in order to avoid undesirable consequences. Traditional approaches are not suitable for practical and large-scale systems due to complexity and computational burdens.
ELECTRIC POWER SYSTEMS RESEARCH
(2023)
Article
Energy & Fuels
Alireza Pourdaryaei, Amidaddin Shahriari, Mohammad Mohammadi, Mohammad Reza Aghamohammadi, Mazaher Karimi, Kimmo Kauhaniemi
Summary: The purpose of this work was to demonstrate the connection between load flow analysis and long-term voltage stability by introducing a new voltage stability assessment based on a multi-machine dynamic model. The proposed method was validated using the IEEE 118-bus test system.
Article
Engineering, Electrical & Electronic
Amir Rostami, Mohammad Mohammadi, Hadis Karimipour
Summary: This paper presents a reliability assessment technique for cyber-physical power systems (CPPSs) that incorporates cybersecurity issues and considers non-normal random variables with non-linear dependencies. The proposed model uses bidirectional encoder representations from transformers (BERTs) to predict the severity of cyber vulnerabilities through textual analysis. A Bayesian attack graph and a Markov model are used to simulate attack paths and demonstrate the consequences of cyber attacks. The proposed method shows higher accuracy and faster convergence compared to other methods.
INTERNATIONAL JOURNAL OF ELECTRICAL POWER & ENERGY SYSTEMS
(2023)
Review
Engineering, Electrical & Electronic
Ali Reza Abbasi, Mohammad Mohammadi
Summary: The evaluation of system performance uncertainty is crucial due to the increasing presence of distributed energy resources and uncertainties in power networks, necessitating effective load flow analysis technologies. Probabilistic load flow analysis can uncover the stochastic properties of power systems, but nonparametric tools are also needed as uncertainties may not conform to conventional probability density functions. Past studies have explored various uncertainty modeling techniques in probabilistic load flow, but no single best methodology exists, as the load flow problem and type of uncertain input variables determine the optimal strategy. This research provides a comprehensive overview and classification of load flow strategies in distribution networks from different perspectives, as well as evaluates and compares uncertainty modeling techniques in stochastic load flow based on accuracy, complexity, and simulation time, benefiting engineers, scientists, and researchers in this field.
ELECTRIC POWER SYSTEMS RESEARCH
(2023)
Article
Engineering, Electrical & Electronic
Farshid Dehghani, Mohammad Mohammadi, Mazaher Karimi
Summary: This study proposes a transparent methodology and a set of metrics to quantify the resilience of distribution systems subjected to hurricanes. The methodology includes probabilistic modeling of failure probability and restoration time, age-dependent fragility analysis, and assessment of resilience considering different lifetimes. The metrics, vulnerability rate and restoration rate, accurately assess and quantify power system resilience.
INTERNATIONAL JOURNAL OF ELECTRICAL POWER & ENERGY SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Shahabodin Afrasiabi, Mousa Afrasiabi, Mohammad Amin Jarrahi, Mohammad Mohammadi, Jamshid Aghaei, Mohammad Sadegh Javadi, Miadreza Shafie-Khah, Joao P. S. Catalao
Summary: In this article, a WAMS-based load modeling method is proposed, which combines impedance-current-power and induction motor, and utilizes deep learning techniques to understand the time-varying and complex behavior of the load. The method is shown to be effective and robust in numerical experiments, and outperforms other methods significantly.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Alireza Pourdaryaei, Mohammad Mohammadi, Hamza Mubarak, Abdallah Abdellatif, Mazaher Karimi, Elena Gryazina, Vladimir Terzija
Summary: This study proposes a method for electricity price forecasting using multi-head self-attention and Convolutional Neural Networks (CNN), and also develops a feature selection technique using mutual information and neural networks. The simulation results demonstrate the efficiency of the proposed method and its potential to aid effective energy management and decision-making in the electricity industry.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Computer Science, Information Systems
Fatemeh Jahani, Mohammad Mohammadi, Alireza Pourdaryaei, Ebrahim Shayesteh, Mazaher Karimi
Summary: This paper presents a new technique that combines multi-criteria decision analysis and gray system theory to solve stochastic multi-criteria decision-making problems with uncertain weight information. The proposed technique demonstrates relevance and efficacy in asset management optimization in electric power systems.
Article
Computer Science, Information Systems
Solmaz Farahbod, Taher Niknam, Mohammad Mohammadi, Jamshid Aghaei, Sattar Shojaeiyan
Summary: This article presents a probabilistic approach for wind speed prediction, including designing a deep network to understand the temporal and spatial features of wind speed data, and proposing a modified loss function and kernel density estimator to improve prediction accuracy.
