Recent ICRF coupling experiments on EAST

Recent ion cyclotron resonance frequency(ICRF) coupling experiments for optimizing ICRF heating in high power discharge were performed on EAST. The coupling experiments were focus on antenna phasing and gas puffing, which were performed separately on two ports of the ion cyclotron resonance heating(ICRH) system of EAST. The antenna phasing was performed on the I-port antenna, which consists of four toroidally spaced radiating straps operating in multiple phasing cases; the coupling performance was better under low wave number ∣k_‖∣(ranging from 4.5 to 6.5). By fuelling the plasma from gas injectors, placed as uniformly spaced array from top to bottom at each side limiter of the B-port antenna, which works in dipole phasing, the coupling resistance of the B-port antenna increased obviously.Furthermore, the coupling resistance of the I-port antenna was insensitive to a smaller rate of gas puffing but when the gas injection rate was more than a certain value(1021 s~(-1)), a sharp increase in the coupling resistance of the I-port antenna occurred, which was mainly caused by the toroidal asymmetric boundary density arising from gas puffing. A more specific analysis is given in the paper.     

基于AM-LSTM的风电场内多点位风电功率预测

为了提高大型风电场的风电功率预测的准确性,构建基于多点位的注意力机制(AM)和长短期记忆神经网络(LSTM)混合预测模型.对于风电功率影响因素特征提取模块,利用属性简约方法,在保证不影响原有的分类效果上,去掉多余属性,大幅度提高预测效率.对于风电功率预测模块,首先,利用15分钟数据间隔,不同高度的风速等作为模型输入,然后通过LSTM处理输入的时间序列与风电功率之间的非线性关系,接下来通过AM进一步优化LSTM的权重,最终得到风电功率的预测结果.通过国外某电场数据的验证:将数据集按照季度分开,各个模型预测结果最为优的是第四季度,对于基于多点位的注意力机制和长短期记忆混合预测模型的均方根误差和平均绝对误差分别比前馈神经网络预测模型降低48.8％和17.4％,比单一长短期记忆神经网络预测模型降低37.2％和7.8％.     

色彩对室内设计的影响

色彩在建筑装饰设计中,具有相当重要的作用。与形状相比,色彩更能引起人的视觉反映,而且还直接影响着人们的心理和情绪。因为在人体的各种知觉中,视觉是最主要的感觉,据说人依靠眼睛可获得约87%的外来信息。而眼睛只有通过光的作用在物体上造成色彩才能获得印象。故色彩有唤起人的第一视觉的作用。     

Design and test of voltage and current probes for EAST ICRF antenna impedance measurement

On the experimental advanced superconducting tokamak(EAST), a pair of voltage and current probes(V/I probes) is installed on the ion cyclotron radio frequency transmission lines to measure the antenna input impedance, and supplement the conventional measurement technique based on voltage probe arrays. The coupling coefficients of V/I probes are sensitive to their sizes and installing locations, thus they should be determined properly to match the measurement range of data acquisition card. The V/I probes are tested in a testing platform at low power with various artificial loads. The testing results show that the deviation of coupling resistance is small for loads RL??2.5 Ω, while the resistance deviations appear large for loads RL??1.5 Ω, which implies that the power loss cannot be neglected at high VSWR. As the factors that give rise to the deviation of coupling resistance calculation, the phase measurement error is the more significant factor leads to deleterious results rather than the amplitude measurement error. To exclude the possible ingredients that may lead to phase measurement error, the phase detector can be calibrated in steady L-mode scenario and then use the calibrated data for calculation under H-mode cases in EAST experiments.     

基于mRMR和VMD-AM-LSTM的短期风功率预测

为了提高模型预测风功率的准确率,提出了一种基于最大相关-最小冗余筛选、变分模态分解、注意力机制和长短期记忆神经网络的短期风功率预测方法.首先使用变分模态分解算法将风功率序列分解成几个中心频率不同的分量;再对各个分量结合最大相关-最小冗余筛选出的气象特征分别建立注意力机制和长短期记忆混合预测模型;最后将各个分量的预测结果...     

基于RAGAN的工业过程运行指标前馈-反馈多步校正

针对工业过程运行指标反馈校正存在滞后及一步推理校正模型可解释性差的问题,提出了基于递归注意力生成对抗网络(Recurrent attention generative adversarial networks, RAGAN)的运行指标前馈-反馈多步校正方法.该方法采用基于负相关正则化的集成随机权神经网络,建立综合生产指标预报模型,为校正提供前馈信息补偿反馈校正的滞后性.提出的RAGAN校正,采用多步校正实现一次调整的思想,将当前时刻运行指标映射到低维潜变量空间简化数据复杂度,利用长短期记忆(Long short-term memory, LSTM)模型实现数据的分步输入,提高模型可解释性;采用分布式注意力(Distributed attention, DA)机制构建数据读入网络,使校正环节获取任务相关性更高的数据,降低任务复杂度,减小噪声干扰,利用校正后的运行指标,保证系统的综合指标尽可能地跟随设定值运行.采用中国西部地区最大选矿厂实际数据的仿真实验,验证了所提方法的有效性.