• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.1
• /
• pp.1-14
• /
• 2024

Time-series forecasting is extensively used in the actual world. Recent research has shown that Transformers with a self-attention mechanism at their core exhibit better performance when dealing with such problems. However, most of the existing Transformer models used for time series prediction use the traditional encoder-decoder architecture, which is complex and leads to low model processing efficiency, thus limiting the ability to mine deep time dependencies by increasing model depth. Secondly, the secondary computational complexity of the self-attention mechanism also increases computational overhead and reduces processing efficiency. To address these issues, the paper designs an efficient multi-layer attention-based time-series forecasting model. This model has the following characteristics: (i) It abandons the traditional encoder-decoder based Transformer architecture and constructs a time series prediction model based on multi-layer attention mechanism, improving the model's ability to mine deep time dependencies. (ii) A cross attention module based on cross attention mechanism was designed to enhance information exchange between historical and predictive sequences. (iii) Applying a recently proposed sparse attention mechanism to our model reduces computational overhead and improves processing efficiency. Experiments on multiple datasets have shown that our model can significantly increase the performance of current advanced Transformer methods in time series forecasting, including LogTrans, Reformer, and Informer.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.7
• /
• pp.2407-2424
• /
• 2022

Recently, Transformer has made great progress in video retrieval tasks due to its high representation capability. For the structure of a Transformer, the cascaded self-attention modules are capable of capturing long-distance feature dependencies. However, the local feature details are likely to have deteriorated. In addition, increasing the depth of the structure is likely to produce learning bias in the learned features. In this paper, an improved Transformer structure named TransDCS (Transformer with Dynamic Convolution and Shortcut) is proposed. A Multi-head Conv-Self-Attention module is introduced to model the local dependencies and improve the efficiency of local features extraction. Meanwhile, the augmented shortcuts module based on a dual identity matrix is applied to enhance the conduction of input features, and mitigate the learning bias. The proposed model is tested on MSRVTT, LSMDC and Activity-Net benchmarks, and it surpasses all previous solutions for the video-text retrieval task. For example, on the LSMDC benchmark, a gain of about 2.3% MdR and 6.1% MnR is obtained over recently proposed multimodal-based methods.

• Journal of the Architectural Institute of Korea Structure & Construction
• /
• v.35 no.8
• /
• pp.177-184
• /
• 2019

Recently, electric power usages and peak loads from buildings are increasing due to higher outdoor air temperatures and/or abnormal climate during the summer period. As one of the eco-friendly measures, a renewable energy system has been received much attention. Particularly, interest on a photovoltaic (PV) system using solar energy has been rapidly increasing in a building sector due to its broad applicability. In using the PV system, one of important factors is the PV efficiency. The normal PV efficiency is determined based on the STC(Standard Test Condition) and the NOCT(Nominal Operating Cell Temperature) performance test. However, the actual PV efficiency is affected by the temperature change at the module surface. Especially, higher module temperatures generally reduce the PV efficiency, and it leads to less power generation from the PV system. Therefore, the analysis of the relation between the module temperature and PV efficiency is required to evaluate the PV performance during the summer period. This study investigates existing algorithms for calculating module surface temperatures and analyzes resultant errors with the algorithms by comparing the measured module temperatures.

• 한국신재생에너지학회:학술대회논문집
• /
• 2007.11a
• /
• pp.246-249
• /
• 2007

The application of photovoltaics into building as integrated building components has been paid more attention worldwide. Photovoltaics or solar electric modules are sol id state devices, directly converting solar radiation into electricity; the process does not require fuel and any moving parts, and produce no pollutants. And the prefab building method is very effective because the pre- manufactured building components is simply assembled to making up buildings in the construction fields especially the sandwich panel. So, this paper describes a design and performance test of the 10kW poly metal pv module(pmpp) system. It is concluded that the prediction of BIPV system's performance should be based on the more accurate PV module installation.

• Proceedings of the KIEE Conference
• /
• 2008.07a
• /
• pp.1110-1111
• /
• 2008

The life time of PV module is semi-permanent. But, because of installation and module fabrication process, its important part can not be finished. In this paper, we analyze 15 years old modules made from different company. Among the PV modules, the maximum power drop ratio was 12.23% minimum and 80.63% maximum. Also the effect of solar cell's short circuit current difference was analyzed. The PV module exposed about 65days, its the maximum power drop ratio was 1.29% minimum and 23.43% maximum. It is for reduction of current value. And the reason for current reduction was due to reduction of parallel resistance of solar cell. To prevent early degradation, it is need to have attention to fabrication, installation and maintenance.

• KIEAE Journal
• /
• v.6 no.4
• /
• pp.17-24
• /
• 2006

The application of photovoltaics into building as integrated building components has been paid more attention worldwide. Photovoltaics or solar electric modules are solid state devices, directly converting solar radiation into electricity; the process does not require fuel and any moving parts, and produce no pollutants. And the prefab building method is very effective because the pre- manufactured building components is simply assembled to making up buildings in the construction fields especially the sandwich panel. Architecture considerations for the integration of PV module to building envelope such as building structure, construction type, safety, regulation, maintenance etc. have been carefully refelected from the early stage of BIPV module design. Trial product of BIPV module are manufactured and sample construction details for demonstration building are purposed. Therefore, this paper intends to advanced its practical use by proposing how to get integrated PV system which can be applied to prefab building material, and how to apply it.

• Journal of the Microelectronics and Packaging Society
• /
• v.26 no.4
• /
• pp.7-13
• /
• 2019

Power module is getting attention from electronic industries such as solar cell, battery and electric vehicles. Transient liquid phase (TLP) boding, sintering with Ag and Cu powders and wire bonding are applied to power module packaging. Sintering is a popular process but it has some disadvantages such as high cost, complex procedures and long bonding time. Meanwhile, TLP bonding has lower bonding temperature, cost effectiveness and less porosity. However, it also needs to improve ductility of the intermetallic compounds (IMCs) at the joint. Wire boding is also an important interconnection process between semiconductor chip and metal lead for direct bonded copper (DBC). In this study, TLP bonding using Sn-based solders and wire bonding process for power electronics packaging are described.

• Proceedings of the KIPE Conference
• /
• 2014.07a
• /
• pp.79-80
• /
• 2014

Voltage source multilevel module converter attracts more and more attention recently. The core component of the voltage source multilevel module converter is the valve based on IGBT. So the test circuit for the valve is very important, reliable test method can guarantee the converter valve design meet the operation requirement. This paper analyzes the valve voltage and current stress during the operation, and according to IEC standard test requirement, object, condition, introduces a kind of test circuit. Finally, through the simulation model, to verify the test circuit can provide the proper test condition for the voltage source multilevel module converter valve.

• Journal of Broadcast Engineering
• /
• v.24 no.6
• /
• pp.1113-1121
• /
• 2019

In this paper, we propose the face anti-spoofing method based on combination of luminance and chrominance with convolutional neural networks. The proposed method extracts luminance and chrominance features independently from live and fake faces by using stacked convolutional neural networks and auxiliary networks. Unlike previous methods, an attention module has been adopted to adaptively combine extracted features instead of simply concatenating them. In addition, we propose a new loss function, called the contrast loss, to learn the classifier more efficiently. Specifically, the contrast loss improves the discriminative power of the features by maximizing the distance of the inter-class features while minimizing that of the intra-class features. Experimental results demonstrate that our method achieves the significant improvement for face anti-spoofing compared to existing methods.

• Journal of IKEEE
• /
• v.24 no.4
• /
• pp.1141-1147
• /
• 2020

The hornet species are so similar in shape that they are difficult for non-experts to classify, and because the size of the objects is small and move fast, it is more difficult to detect and classify the species in real time. In this paper, we developed a system that classifies hornets species in real time based on a deep learning algorithm using a boundary box. In order to minimize the background area included in the bounding box when labeling the training image, we propose a method of selecting only the head and body of the hornet. It also experimentally compares existing boundary box-based object recognition algorithms to find the best algorithms that can detect wasps in real time and classify their species. As a result of the experiment, when the mish function was applied as the activation function of the convolution layer and the hornet images were tested using the YOLOv4 model with the Spatial Attention Module (SAM) applied before the object detection block, the average precision was 97.89% and the average recall was 98.69%.