Development Of Graphics Processing Units Research Articles

The state-of-charge predication of lithium-ion battery energy storage system using data-driven machine learning

Accurate estimation of state-of-charge (SOC) is critical for guaranteeing the safety and stability of lithium-ion battery energy storage system. However, this task is very challenging due to the coupling dynamics of multiple complex processes inside the lithium-ion battery and the lack of measure to monitor the variations of a battery’s internal properties. Recently, with the continuous development of Graphic Processing Unit (GPU) computing power, there is an increasing interest in applying deep-learning as SOC estimation approaches. In this paper, a novel SOC estimation scheme for lithium-ion energy storage system is proposed based on Convolutional Neural Network and Long Short-Term Memory (CNN-LSTM) neural network. This method is completely driven by the actual operating data from a photovoltaic energy storage system without using any artificial battery models or inference systems. Compared with traditional SOC estimation methods, the CNN-LSTM model can overcome the deviation in estimation caused by voltage jump at the end of charge and discharge, provide satisfied SOC estimation results during stabilized stage and various charging/discharging stages of the assembled lithium-ion batteries in the system. The calculation results indicate that this method enables fast and accurate SOC estimation with an RMSE of less than 0.31% over the entire operating data of the photovoltaic energy storage system for a full day.

A Survey of GPU Multitasking Methods Supported by Hardware Architecture

The ability to support multitasking becomes more and more important in the development of graphic processing unit (GPU). GPU multitasking methods are classified into three types: temporal multitasking, spatial multitasking, and simultaneous multitasking (SMK). This article first introduces the features of some commercial GPU architectures to support multitasking and the common metrics used for evaluating the performance of GPU multitasking methods, and then reviews the GPU multitasking methods supported by hardware architecture (i.e., hardware GPU multitasking methods). The main problems of each type of hardware GPU multitasking methods to be solved are illustrated. Meanwhile, the key idea of each previous hardware GPU multitasking method is introduced. In addition, the characteristics of hardware GPU multitasking methods belonging to the same type are compared. This article also gives some valuable suggestions for the future research. An enhanced GPU simulator is needed to bridge the gap between academia and industry. In addition, it is promising to expand the research space with machine learning technologies, advanced GPU architectural innovations, 3D stacked memory, etc. Because most previous GPU multitasking methods are based on NVIDIA GPUs, this article focuses on NVIDIA GPU architecture, and uses NVIDIA's terminology. To our knowledge, this article is the first survey about hardware GPU multitasking methods. We believe that our survey can help the readers gain insights into the research field of hardware GPU multitasking methods.

Importance of Machine Learning in Making Investment Decision in Stock Market

Executive Summary Predicting stock trends in the financial market is always demanding but satisfying as well. With the growing power of computing and the recent development of graphics processing unit and tensor processing unit, analysts and researchers are applying advanced techniques such as machine learning techniques more and more to predict stock price trends. In recent years, researchers have developed several algorithms to predict stock trends. To assist investors interested in investing in the stock market, preferably for a short period, it has become necessary to review research papers dealing on machine learning and analyse the importance of their findings in the context of how stock price trends generate trading signals. In this article, to achieve the stated task, authors scrutinized more than 50 research papers focusing on various machine learning algorithms with varied levels of input variables and found that though the performance of models measured by root-mean-square error (RMSE) for regression and accuracy score for classification models varied greatly, long short-term memory (LSTM) model displayed higher accuracy amongst the machine and deep learning models reviewed. However, reinforcement learning algorithm performance measured by profitability and Sharpe ratio outperformed all. In general, traders can maximize their profits by using machine learning instead of using technical analysis. Technical analysis is very easy to implement, but the profit based on it can vanish too soon or making a profit using technical analysis is almost difficult because of its simplicity. Hence, studying machine, deep and reinforcement learning algorithms is vital for traders and investors. These findings were based on the literature review consolidated in the result section.

An Improved Bidirectional Gated Recurrent Unit Method for Accurate State-of-Charge Estimation

State-of-Charge (SOC) estimation of lithium-ion batteries have a great significance for ensuring the safety and reliability of battery management systems in electrical vehicle. Deep learning method can hierarchically extract complex feature information from input data by building deep neural networks (DNNs) with multi-layer nonlinear transformations. With the development of graphic processing unit, the training speed of the network is faster than before, and it has been proved to be an effective data-driven method to estimate SOC. In order to further explore the potential of DNNs in SOC estimation, take battery measurements like voltage, current and temperature directly as input and SOC as output, an improved method using the Nesterov Accelerated Gradient (NAG) algorithm based Bidirectional Gated Recurrent Unit (Bi-GRU) network is put forward in this paper. Notably, to address the oscillation problem existing in the traditional gradient descent algorithm, NAG is used to optimize the Bi-GRU. The gradient update direction is corrected by considering the gradient influence of the historical and the current moment, combined with the estimated location of the parameters at the next moment. Compared to state-of-the-art estimation methods, the proposed method enables to capture battery temporal information in both forward and backward directions and get independent context information. Finally, two well-recognized lithium-ion batteries datasets from University of Maryland and McMaster University are applied to verify the validity of the research. Compared with the previous methods, the experimental results demonstrate that the proposed NAG based Bi-GRU method for SOC estimation can improve the precision of the prediction at various ambient temperature.

Optimized Block-Based Algorithms to Label Connected Components on GPUs

Connected Components Labeling (CCL) is a crucial step of several image processing and computer vision pipelines. Many efficient sequential strategies exist, among which one of the most effective is the use of a block-based mask to drastically cut the number of memory accesses. In the last decade, aided by the fast development of Graphics Processing Units (GPUs), a lot of data parallel CCL algorithms have been proposed along with sequential ones. Applications that entirely run in GPU can benefit from parallel implementations of CCL that allow to avoid expensive memory transfers between host and device. In this paper, two new eight-connectivity CCL algorithms are proposed, namely Block-based Union Find (BUF) and Block-based Komura Equivalence (BKE). These algorithms optimize existing GPU solutions introducing a block-based approach. Extensions for three-dimensional datasets are also discussed. In order to produce a fair comparison with previously proposed alternatives, YACCLAB, a public CCL benchmarking framework, has been extended and made suitable for evaluating also GPU algorithms. Moreover, three-dimensional datasets have been added to its collection. Experimental results on real cases and synthetically generated datasets demonstrate the superiority of the new proposals with respect to state-of-the-art, both on 2D and 3D scenarios.

PELLR: A Permutated ELLPACK-R Format for SpMV on GPUs

The sparse matrix vector multiplication (SpMV) is inevitable in almost all kinds of scientific computation, such as iterative methods for solving linear systems and eigenvalue problems. With the emergence and development of Graphics Processing Units (GPUs), high efficient formats for SpMV should be constructed. The performance of SpMV is mainly determinted by the storage format for sparse matrix. Based on the idea of JAD format, this paper improved the ELLPACK-R format, reduced the waiting time between different threads in a warp, and the speed up achieved about 1.5 in our experimental results. Compared with other formats, such as CSR, ELL, BiELL and so on, our format performance of SpMV is optimal over 70 percent of the test matrix. We proposed a method based on parameters to analyze the performance impact on different formats. In addition, a formula was constructed to count the computation and the number of iterations.

Application of artificial intelligence methods to recognize pathologies on medical images

The rapid development of graphics processing units (GPU) allowed to use in practice resource-intensive methods of artificial intelligence such as neural networks. At present, the authors got preliminary results on detection of some forms of leukocytes. The article presents the first results of the work.

A Case Study of the Base Technology for the Smart Grid Security: Focusing on a Performance Improvement of the Basic Algorithm for the DDoS Attacks Detection Using CUDA

Since the development of Graphic Processing Unit (GPU) in 1999, the development speed of GPUs has become much faster than that of CPUs and currently, the computational power of GPUs exceeds CPUs dozens and hundreds times in terms of decimal calculations and costs much less. Owing to recent technological development of hardwares, general-purpose computing and utilization using GPUs are on the rise. Thus, in this paper, we have identified the elements to be considered for the Smart Grid Security. Focusing on a Performance Improvement of the Basic Algorithm for the Stateful Inspection to Detect DDoS Attacks using CUDA. In the program, we compared the search speeds of GPU against CPU while they search for the suffix trees. For the computation, the system constraints and specifications were made identical during the experiment. We were able to understand from the results of the experiment that the problem-solving capability improves when GPU is used. The other finding was that performance of the system had been enhanced when shared memory was used explicitly instead of a global memory as the volume of data became larger.

CUDA-Based SSA Method in Application to Calculating EM Scattering From Large Two-Dimensional Rough Surface

The small slop approximation (SSA) is an accurate method to calculate the electromagnetic (EM) scattering properties of rough surfaces. However, its computational complexity restricts its application to smaller domains and there is always the need for speedup in very large cases using pure central processing units (CPUs) hardware. With the development of graphics processing units (GPUs), more processors are dedicated to perform independent calculations. In addition, NVIDIA introduced a parallel computing platform, compute unified device architecture (CUDA), which provides researchers an easy way to use processors on GPU. To calculate EM scattering properties on GPU, we reformulate the SSA method with CUDA to take advantage of GPU threads. Because each thread executes synchronously and deals with a corresponding point data of rough surface, the CUDA-based SSA method calculates faster than the pure-CPU equivalent. To overcome memory limitations, the data of large rough surface are stored on hard disk. Moreover, a subsidiary thread is used to deal with the process of data transmission between the memory and the hard disk and reduce transmitting time further. The factors, block size, data transfers, and register, are also discussed in the optimization of the CUDA application. Test cases running on a NVIDIA GTX 460 GPU indicate that two orders of magnitude speedup, including file input and output, is obtained with our new formulation.

TOWARDS FFT-BASED DIRECT NUMERICAL SIMULATIONS OF TURBULENT FLOWS ON A GPU

The accurate simulation of turbulence and the implementation of corresponding turbulence models are both critical to the understanding of the complex physics behind turbulent flows in a variety of science and engineering applications. Despite the tremendous increase in the computing power of central processing units (CPUs), direct numerical simulation of highly turbulent flows is still not feasible due to the need for resolving the smallest length scale, and today's CPUs cannot keep pace with demand. The recent development of graphics processing units (GPU) has led to the general improvement in the performance of various algorithms. This study investigates the applicability of GPU technology in the context of fast-Fourier transform (FFT)-based pseudo-spectral methods for DNS of turbulent flows for the Taylor–Green vortex problem. They are implemented on a single GPU and a speedup of unto 31x is obtained in comparison to a single CPU.

Gnss Signal Processing in Gpu

ABSTRACT Signal processing of the global navigation satellite systems (GNSS) is a computationally demanding task due to the wide bandwidth of the signals and their complicated modulation schemes. The classical GNSS receivers therefore utilize tailored digital signal processors (DSP) not being flexible in nature. Fortunately, the up-to-date parallel processors or graphical processing units (GPUs) dispose sufficient computational power for processing of not only relatively narrow band GPS L1 C/A signal but also the modernized GPS, GLONASS, Galileo and COMPASS signals. The performance improvement of the modern processors is based on the constantly increasing number of cores. This trend is evident not only from the development of the central processing units (CPUs), but also from the development of GPUs that are nowadays equipped with up to several hundreds of cores optimized for video signals. GPUs include special vector instructions that support implementation of massive parallelism. The new GPUs, named as general-purpose computation on graphics processing units (GPGPU), are able to process both graphic and general data, thus making the GNSS signal processing possible. Application programming interfaces (APIs) supporting GPU parallel processing have been developed and standardized. The most general one, Open Computing Language (Open CL), is now supported by most of the GPU vendors. Next, Compute Unified Device Architecture (CUDA) language was developed for NVidia graphic cards. The CUDA language features optimized signal processing libraries including efficient implementation of the fast Fourier transform (FFT). In this paper, we study the applicability of the GPU approach in GNSS signal acquisition. Two common parallel DSP methods, parallel code space search (PCSS) and double-block zero padding (DBZP), have been investigated. Implementations in the C language for CPU and the CUDA language for GPU are discussed and compared with respect to the acquisition time. It is shown that for signals with long ranging codes (with 10230 number of chips - Galileo E5, GPS L5 etc.). Paper presented at the "European Navigation Conference 2012", held in Gdansk, Poland

Fast GPU Based Adaptive Filtering of 4D Echocardiography

Time resolved three-dimensional (3D) echocardiography generates four-dimensional (3D+time) data sets that bring new possibilities in clinical practice. Image quality of four-dimensional (4D) echocardiography is however regarded as poorer compared to conventional echocardiography where time-resolved 2D imaging is used. Advanced image processing filtering methods can be used to achieve image improvements but to the cost of heavy data processing. The recent development of graphics processing unit (GPUs) enables highly parallel general purpose computations, that considerably reduces the computational time of advanced image filtering methods. In this study multidimensional adaptive filtering of 4D echocardiography was performed using GPUs. Filtering was done using multiple kernels implemented in OpenCL (open computing language) working on multiple subsets of the data. Our results show a substantial speed increase of up to 74 times, resulting in a total filtering time less than 30 s on a common desktop. This implies that advanced adaptive image processing can be accomplished in conjunction with a clinical examination. Since the presented GPU processor method scales linearly with the number of processing elements, we expect it to continue scaling with the expected future increases in number of processing elements. This should be contrasted with the increases in data set sizes in the near future following the further improvements in ultrasound probes and measuring devices. It is concluded that GPUs facilitate the use of demanding adaptive image filtering techniques that in turn enhance 4D echocardiographic data sets. The presented general methodology of implementing parallelism using GPUs is also applicable for other medical modalities that generate multidimensional data.

Multilevel fast multipole algorithm enhanced by GPU parallel technique for electromagnetic scattering problems

AbstractAlong with the development of graphics processing Units (GPUS) in floating point operations and programmability, GPU has increasingly become an attractive alternative to the central processing unit (CPU) for some of compute‐intensive and parallel tasks.In this article, the multilevel fast multipole algorithm (MLFMA) combined with graphics hardware acceleration technique is applied to analyze electromagnetic scattering from complex target. Although it is possible to perform scattering simulation of electrically large targets on a personal computer (PC) through the MLFMA, a large CPU time is required for the execution of aggregation, translation, and deaggregation operations. Thus GPU computing technique is used for the parallel processing of MLFMA and a significant speedup of matrix vector product (MVP) can be observed. Following the programming model of compute unified device architecture (CUDA), several kernel functions characterized by the single instruction multiple data (SIMD) mode are abstracted from components of the MLFMA and executed by multiple processors of the GPU. Numerical results demonstrate the efficiency of GPU accelerating technique for the MLFMA. © 2010 Wiley Periodicals, Inc. Microwave Opt Technol Lett 52: 502–507, 2010; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/mop.24963

Recent Development of Molecular Simulation Based on GPU in Material Science

Molecular simulation can provide mechanism insights into how material behaviour related to molecular properties and microscopic details of the arrangement of many molecules. With the development of Graphics Processing Unit (GPU), scientists have realized general purpose molecular simulations on GPU and the Common Unified Device Architecture (CUDA) environment. In this paper, we provided a brief overview of molecular simulation and CUDA, introduced the recent achievements in molecular simulation based on GPU in material science, mainly about Monte Carlo method and Molecular Dynamics. The recent research achievements have shown that GPUs can provide unprecedented computational power for scientific applications. With optimized algorithms and program codes, a single GPU can provide a performance equivalent to that of a distributed computer cluster. So, study of molecular simulations based on GPU will accelerate the development of material science in the future.