Duy Thanh Nguyen 교수

Convolutional neural networks (CNNs) require numerous computations and external memory accesses. Frequent accesses to off-chip memory cause slow processing and large power dissipation. For real-time object detection with high throughput and power efficiency, this paper presents a Tera-OPS streaming hardware accelerator implementing a you-only-look-once (YOLO) CNN. The parameters of the YOLO CNN are retrained and quantized with the PASCAL VOC data set using binary weight and flexible low-bit acti

In restructured power systems, there are many independent players who benefit from demand response (DR). These include the transmission system operator (TSO), distributors, retailers, and aggregators. This paper proposes a new concept-demand response eXchange (DRX)-in which DR is treated as a public good to be exchanged between DR buyers and sellers. Buyers need DR to improve the reliability of their own electricity-dependent businesses and systems. Sellers have the capacity to significantly mod

Convolutional neural networks (CNNs) require both intensive computation and frequent memory access, which lead to a low processing speed and large power dissipation. Although the characteristics of the different layers in a CNN are frequently quite different, previous hardware designs have employed common optimization schemes for them. This paper proposes a layer-specific design that employs different organizations that are optimized for the different layers. The proposed design employs two laye

This paper presents the design and evaluation of an effective market-clearing scheme for trading demand response (DR) in a deregulated power system. The proposed scheme is called demand response exchange (DRX), in which DR is treated as a public good to be exchanged between two groups of participating agents, namely DR buyers and DR sellers. While buyers require DR and are willing to pay for it, sellers have the capacity to curtail customer loads to supply DR on request. The DRX market clearing

In this paper we investigate efficient schemes for scheduling demand response (DR) in a deregulated environment. We begin with an analysis of partial schemes which are being implemented in many established markets. These schemes can be considered inefficient due to externalities that occur amongst DR beneficiaries including the transmission network company (Transco), distribution network companies (Discos), and retail electricity companies (Recos). This limitation of existing schemes motivates t

A DRAM device requires periodic refresh operations to preserve data integrity, which incurs significant power consumption. This paper proposes a new memory architecture to reduce the power consumption by refresh operations by slowing down the refresh rate. Slow refresh may cause a loss of data stored in a DRAM cell, which affects the correctness of the computation using the lost data. The proposed memory architecture attempts to avoid the problem caused by lost data by taking advantage of the er

In this paper we describe an analytical method for estimating uncertainty in distribution loads, with applications to network monitoring and optimal meter placement. Our model works by assessing the impact of consumption aggregation from customers of different types, in order to determine the mean and variance of loading profile in each distribution transformer in the network. Such parameters are then adjusted by correlating with real-time information taken from the supervisory control and data

DRAM devices require periodic refresh operations to preserve data integrity. Slowing down the refresh rate can reduce the energy consumption; however, it may cause a loss of data stored in the DRAM cell. This paper proposes a new memory architecture of soft approximation for deep learning applications, which reduces the refresh energy consumption while maintaining accuracy and high performance. Utilizing the error-tolerant property of deep learning applications, the proposed memory architecture

Residual block is a very common component in recent state-of-the art CNNs such as EfficientNet or EfficientDet. Shortcut data accounts for nearly 40% of feature-maps access in ResNet152 [8]. Most of the previous DNN compilers, accelerators ignore the shortcut data optimization. This paper presents ShortcutFusion, an optimization tool for FPGA-based accelerator with a reuse-aware static memory allocation for shortcut data, to maximize on-chip data reuse given resource constraints. From TensorFlow

In this paper we develop a novel mathematical model to describe the localized economic impact of load recovery by electricity consumers following demand response (DR) event participation. In this model-termed “securitization”-cost and benefit associated with an amount of load recovery change exponentially with time-varying stochastic discount rates. These volatile rates, which can be estimated using the principle of Brownian motion, are further examined via case studies on the Roy Billinton test

There is a need to integrate Demand Response (DR) into the management of the electricity market and power system within both planning and operational timescales. This paper reviews some issues of DR for domestic and small business consumers, in both the United States and Australian contexts. The advantages and limitations of two types of DR, namely reliability-based DR and price-based DR are discussed through examining related studies. The paper claims that neither price-based DR nor reliability

This paper introduces a new and separated market for the trading of Demand Response (DR) in a restructured power system. This market is called Demand Response eXchange or shortly DRX, in which DR is treated as a virtual resource to be exchanged between DR buyers and sellers. Buyers - including the Transmission System Operator (TSO), distributors, and retailers - need DR to improve the efficiency and reliability of their own businesses and systems. Sellers - including consumers though an aggregat

This thesis presents the development of a new and separate market for trading Demand Response (DR) in a deregulated power system. This market is termed Demand Response eXchange (DRX), in which DR in the form of hourly load reduction is considered a product to be negotiated between two groups of market participants, namely buyers and sellers. DR buyers, including all transmission companies (Transcos), distribution companies (Discos), and retail companies (Recos) need DR for their risk management

This paper focuses on designing process of a micro-controller based system with remote control interface to be used for home automation and demand-side automatic meter reading (AMR). The system consists of a master node called energy modem (EM) and many slave nodes called energy appliance controller (ECs). The bi-directional and asynchronous power line communication (PLC) links among energy modem and energy load controllers were successfully established. Energy metering, reporting, and control s