A5101884505- Optimal Power Flow Distribution
- Power System Optimization and Stability
- Semiconductor materials and devices
- Advanced Memory and Neural Computing
- Neural Networks and Applications
- Energy Load and Power Forecasting
- Ferroelectric and Negative Capacitance Devices
- Electric Power System Optimization
- Advancements in Semiconductor Devices and Circuit Design
- Power Quality and Harmonics
- Low-power high-performance VLSI design
- Microgrid Control and Optimization
- Parallel Computing and Optimization Techniques
- Smart Grid Energy Management
- Image and Signal Denoising Methods
- Power System Reliability and Maintenance
- Advanced Neural Network Applications
- Blind Source Separation Techniques
- Grey System Theory Applications
- Advanced Data Storage Technologies
- Machine Learning and ELM
- Brain Tumor Detection and Classification
- Integrated Circuits and Semiconductor Failure Analysis
- Islanding Detection in Power Systems
- Model Reduction and Neural Networks
Taiwan Semiconductor Manufacturing Company (Taiwan)
2021-2024
Shiga University of Medical Science
2022-2024
Chiba Hospital
2024
Meiji University
2002-2023
IBM Research - Tokyo
2007-2021
Kobe University
2015-2018
Kyushu University
1984-2013
NEC (Japan)
1996-2002
Mitsubishi Group (Japan)
2002
Mitsubishi Electric (Japan)
1996
There have been observed several voltage instability phenomena in electric power systems where receiving end voltages oscillate remarkably or get much lower than the nominal values. Those tend to occur heavy loaded conditions and seem be related multiple load flow solution problem for following reasons. It has confirmed analytically by simulations that solutions are likely appear under heavy-loaded conditions, individual of pair different features from each other, standpoint stability,...
From the cloud to edge devices, artificial intelligence (AI) and machine learning (ML) are widely used in many cognitive tasks, such as image classification speech recognition. In recent years, research on hardware accelerators for AI devices has received more attention, mainly due advantages of at edge: including privacy, low latency, reliable effective use network bandwidth. However, traditional computing architectures (such CPUs, GPUs, FPGAs, even existing accelerator ASICs) cannot meet...
Computing-in-memory (CIM) is being widely explored to minimize power consumption in data movement and multiply-and-accumulate (MAC) for edge-AI devices. Although most prior work focuses on analog-based CIM (ACIM) leverage the BL charge/discharge operation, lack of accuracy caused by transistor variation ADC an issue [1]–[3]. In contrast, a digital-based (DCIM) approach realizes enough flexibility various input weight bit widths [4], while also benefiting from technology scaling. This paper...
The computational load, for accurate AI workloads, is moving from large server clusters to edge devices; thus enabling richer and more personalized applications. Compute-in-memory (CIM) beneficial edge-AI specifically ones that are MAC-intensive. However, realizing better power-performance-area (PPA) high accuracy a major challenge practical CIM implementation. Recent work examined tradeoffs between MAC throughput, energy efficiency analog based [1–3]. On the other hand, digital-CIMs (DCIM),...
Hardware acceleration of deep learning using analog non-volatile memory (NVM) requires large arrays with high device yield, accuracy Multiply-ACcumulate (MAC) operations, and routing frameworks for implementing arbitrary neural network (DNN) topologies. In this article, we present a 14-nm test-chip Analog AI inference—it contains multiple phase change (PCM)-devices, each array capable storing 512 <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML"...
Recently SRAM-based digital compute-in memory (D-CIM) [1] has demonstrated excellent energy/area efficiency, with full precision of 4b/8b integer multiply-accumulate operations, it better programmability, hardware reuse and scalability, in addition, can effectively leverage technology scaling for PPA. Nonetheless, several new challenges remain, including huge peak currents resulting from high parallel operation, long delays adder trees, scalable architectures that support various neural...
Compute-in-memory (CIM) is being widely explored to minimize power consumption related data movement and multiply-and-accumulate (MAC) operations for AI edge devices. Compared analog based CIMs, digital-based CIMs (DCIM), which include small, distributed SRAM banks a customized MAC unit, realize massively parallel computation with no accuracy loss better power-performance-area (PPA) scaling advanced technologies. However, balancing operating efficiency per area (TOPS/mm <sup...
A novel method for predicting power system voltage harmonics with an artificial neural network is presented. The based on the backpropagation learning technique feedforward networks. promise of proposed in prediction shown. In order to demonstrate its effectiveness, applied observed through a personal computer measurement and performance compared that conventional methods.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
The present study aimed to determine the clinical value of serum procalcitonin (PCT) level in predicting postoperative infections after hepatectomy. Medical records 301 consecutive patients who underwent hepatectomy were retrospectively reviewed. We divided into infection-positive and infection-negative groups. investigated changes perioperative inflammatory markers such as C-reactive protein (CRP) PCT level. Associations between infectious complications evaluated identify predictive factors...
Objectives: The clinical significance of early administration pancrelipase after distal pancreatectomy (DP) has not been reported, and its effect on new-onset diabetes mellitus (DM) clarified. This study aimed to investigate the DP postoperative nutritional status, skeletal muscle mass, DM. Methods: We retrospectively reviewed medical records 76 patients who underwent DP. Delayed-release high-titer was administered daily, starting day 3 (EP group). Postoperative mass index (SMI), DM were...
This paper presents a clustering method for preprocessing input data of short-term load forecasting in power systems. Clustering the prior to with artificial neural network (ANN) decreases prediction errors observed. In this paper, an MLP ANN is used deal one-step ahead daily maximum forecasting, and deterministic annealing (DA) employed classify into clusters. The DA based on principle entropy statistical mechanics evaluate globally optimal classification. proposed successfully applied real...
Enhancing video quality is critical for achieving a boosted user experience on smart devices including mobiles, televisions, and monitors. Practical hardware designs should deliver high performance with minimal resources under the stringent limitations related to bandwidth, area energy budget. The widespread usage of deep-learning algorithms in image processing tasks, super-resolution (SR) noise-reduction (NR), has further emphasized necessity energy-efficient solutions. Therefore, an...
This paper proposes a new probabilistic method for short-term load forecasting with the Gaussian processes (GP). In recent years, degree of uncertainty increases as power system becomes more deregulated and competitive. The players are concerned maximizing profit while minimizing risk in market. As result, it is important to consider predicted appropriately. proposed aims at extending average point into that posterior distribution handle forecasting. this paper, hyperparameters covariance...
This paper presents a new method for meter placement in power system state estimation. It plays an important role transmission security control. focuses on to enhance topological observability that discusses the relationship between and network configurations. Mathematically, problem results complicated combinatorial optimization. That implies it is hard solve larger systems. To overcome problem, this tabu search-based efficiently. As meta-heuristic approach, search quite effective solving...
This paper proposes a new method for determining capacitor placement in distribution systems. The proposed makes use of parallel tabu search to evaluate better solution terms computational effort and accuracy. is hard solve sense global optimization due the high nonlinear mixed integer problem. To problem efficiently, this focuses on that one efficient meta-heuristics. Tabu than conventional methods adaptive memory allows escape from local minimum cost function. In paper, improve performance...
This paper proposes a hybrid method for short-term load forecasting in power systems. Short-term is one of the most important problems system operation and planning. Therefore, more accurate models are required to handle it appropriately. The proposed based on fuzzy regression tree data mining multi-layer perceptron (MLP) artificial neural networks. works discover rules from actual classify input into some classes. On other hand, MLP used predict one-step ahead loads. aims clarify nonlinear...
Short-term load forecasting plays a key role in power system operation and planning. This paper presents method for data mining short-term systems. makes use of to clarify the nonlinear relationship between input output variables forecasting. Data discovers useful knowledge rules large bases. is more attractive because difficulty understanding The obtained model structure explains importance variables. It may be classified into classification regression trees. handles tree since corresponds...
This paper proposes a new method for voltage stability assessment in power systems. The proposed is based on hybrid of optimal data mining and an artificial neural network (ANN). Voltage main concern system operation planning. In recent years, the deregulated market brings about uncertain events increases degree uncertainty. As result, operators are faced with more complicated To understand conditions appropriately, they need feature extraction index. this paper, to estimate index extract...
In this paper, a new continuation power flow method is proposed to analyze static voltage stability in three-phase unbalanced radial distribution systems. The useful for evaluating the P-V curves that give maximum loading point. improves conventional way arc-length parameterization and nonlinear predictor are used successfully applied IEEE 13-node
Describe an artificial neural net (ANN) based approach to prediction of power system harmonic voltages. The effectiveness recurrent networks is examined. Recurrent have the advantage being able consider dynamics a time series, unlike conventional feedforward ANN. Four are applied fifth voltage. A comparison made four network models from standpoint accuracy and computational efforts.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
This paper proposes an interval-analysis based power flow computation for calculating multiple solutions in systems. The interval technique provides more reliable solution with the upper and lower bounds. makes use of Krawczyk method to evaluate solution. is effective solving nonlinear equation due feature that it can provide all proposed demonstrated sample
This paper proposes an efficient method for capacitor placement in distribution systems. The proposed is based on parallel tabu search that considers the decomposition of neighborhood into subneighborhoods and multiple lengths search. Capacitor important problem to maintain voltage profile objective determine location size shunt capacitors be installed at nodes so transmission loss installation cost are minimized. difficult solve sense it has high nonlinearity handles both discrete...
This paper presents a new method for forecasting of PV generation output. The output systems is significantly affected by the weather conditions. As result, one most difficult time series forecasting. However, power system operators require more accurate prediction model to deal with operation such as economic load dispatching, unit commitment, etc. proposed makes use hybrid intelligent that consists Generalized Radial Basis Function Network (GRBFN), Deterministic Annealing (DA), and...
In this paper, a kernel machine based method is proposed for short-term load forecasting. This paper makes use of Informative Vector Machine (IVM) to provide better prediction results with The Kernel technique an extension Support (SVM) that very useful pattern recognition deal the regression model quantitative variables. IVM has advantage constructed limited number learning data through new information theory. successfully applied real forecasting in Japan.