A CNN-Based Single-Stage Occlusion Real-Time Target Detection Method

Aiming at the problem of low accuracy of traditional target detection methods for target detection in endoscopes in substation environments, a CNN-based real-time detection method for masked targets is proposed. The method adopts the overall design of backbone network, detection network and algorithmic parameter optimisation method, completes the model training on the self-constructed occlusion target dataset, and adopts the multi-scale perception method for target detection. The HNM algorithm is used to screen positive and negative samples during the training process, and the NMS algorithm is used to post-process the prediction results during the detection process to improve the detection efficiency. After experimental validation, the obtained model has the multi-class average predicted value (mAP) of the dataset. It has general advantages over traditional target detection methods. The detection time of a single target on FDDB dataset is 39 ms, which can meet the need of real-time target detection. In addition, the project team has successfully deployed the method into substations and put it into use in many places in Beijing, which is important for achieving the anomaly of occlusion target detection.

Gearbox Fault Diagnosis using Adaptive Zero Phase Time-varying Filter Based on Multi-scale Chirplet Sparse Signal Decomposition

When used for separating multi-component non-stationary signals, the adaptive time-varying filter(ATF) based on multi-scale chirplet sparse signal decomposition(MCSSD) generates phase shift and signal distortion. To overcome this drawback, the zero phase filter is introduced to the mentioned filter, and a fault diagnosis method for speed-changing gearbox is proposed. Firstly, the gear meshing frequency of each gearbox is estimated by chirplet path pursuit. Then, according to the estimated gear meshing frequencies, an adaptive zero phase time-varying filter(AZPTF) is designed to filter the original signal. Finally, the basis for fault diagnosis is acquired by the envelope order analysis to the filtered signal. The signal consisting of two time-varying amplitude modulation and frequency modulation(AM-FM) signals is respectively analyzed by ATF and AZPTF based on MCSSD. The simulation results show the variances between the original signals and the filtered signals yielded by AZPTF based on MCSSD are 13.67 and 41.14, which are far less than variances (323.45 and 482.86) between the original signals and the filtered signals obtained by ATF based on MCSSD. The experiment results on the vibration signals of gearboxes indicate that the vibration signals of the two speed-changing gearboxes installed on one foundation bed can be separated by AZPTF effectively. Based on the demodulation information of the vibration signal of each gearbox, the fault diagnosis can be implemented. Both simulation and experiment examples prove that the proposed filter can extract a mono-component time-varying AM-FM signal from the multi-component time-varying AM-FM signal without distortion.

Are diurnal time-budgets and activity patterns density-dependent in the Shelduck (Tadorna tadorna) wintering in Algeria? An analysis across multiple temporal scales

Background: The Shelduck (Tadorna tadorna) is a characteristic waterbird species of inland wetlands in northeastern Algeria. Its wintering behavior in relation to changes of local abundances and foraging group density is poorly known. Objectives: This study aims at monitoring patterns of diurnal activities and the variation of behavioral time-budgets in relation to numbers of wintering Shelducks. We investigate temporal variations of diurnal activities across multipletime scales and consider their interrelationships. Methods: Assessments of local population abundance were weekly surveyed during two wintering seasons (2010– 2012), whereas diurnal activities (feeding, sleeping, swimming, preening, loafing, flying, courtship, and antagonism) were studied three times a month during seven hours (08:00–16:00) using the Scan method. Time budget variations of each behavioral activity were tested using nested ANOVAs following multiple time scales. Generalized linear mixedeffects models (GLMM) tested whether variations in diurnal activities were density-dependent. Results: During the wintering season, Shelduck’s numbers followed a bell-shaped trend, which indicated that the species was typically a wintering migrant in Sabkha Djendli. The first individuals arrived onsite in October–November then numbers reached a peak in January (up to 2400 individuals in 2012) with steady density during December–February, afterward individuals left the site progressively until late April when the site is deserted. During both wintering seasons, diurnal activities were dominated by feeding (60%), followed by sleeping (12%) then swimming and preening with 9% and 8%, respectively. The rest of the activities (loafing, flying, courtship and antagonistic behaviors) had low proportions of time budget. ANOVAs showed that activity time budgets varied significantly following multiple time scales (year, season, month, day, semi-hour). Time budgets of diurnal activities during each wintering season were significantly interrelated. Correlations patterns between the two seasons were similar. GLMMs revealed that the variations of diurnal activities were not density-dependent, except for preening and swimming. Conclusion: During the wintering season, habitats of Sabkha Djendli are important for waterbirds, including the Shelduck that used the lake mainly for food-foraging and resting. The 2400 individuals censused in mid-winter are important locally and at the North African scale. This stresses the need to strengthen the protection status of this wetland and mitigate degradation sources that threaten wintering waterfowl.

Research on multi-time scale doubly-fed wind turbine test system based on FPGA+CPU heterogeneous calculation

As the proportion of renewable energy increases, the interaction between renewable energy devices and the grid continues to enhance. Therefore, the renewable energy dynamic test in a power system has become more and more important. Traditional dynamic simulation systems and digital-analog hybrid simulation systems are difficult to compromise on the economy, flexibility and accuracy. A multi-time scale test system of doubly fed induction generator based on FPGA+ CPU heterogeneous calculation is proposed in this paper. The proposed test system is based on the ADPSS simulation platform. The power circuit part of the test system is setup up using the EMT(electromagnetic transient simulation) simulation, and the control part uses the actual physical devices. In order to realize the close-loop testing for the physical devices, the power circuit must be simulated in real-time. This paper proposes a multi-time scale simulation algorithm, in which the decoupling component divides the power circuit into a large time scale system and a small time scale system in order to reduce computing effort. This paper also proposes the FPGA+CPU heterogeneous computing architecture for implementing this multitime scale simulation. In FPGA, there is a complete small time-scale EMT engine, which support the flexibly circuit modeling with any topology. Finally, the test system is connected to an DFIG controller based on Labview to verify the feasibility of the test system.

An Effective Numerical Calculation Method for Multi-Time-Scale Mathematical Models in Systems Biology

The improvements of high-throughput experimental devices such as microarray and mass spectrometry have allowed an effective acquisition of biological comprehensive data which include genome, transcriptome, proteome, and metabolome (multi-layered omics data). In Systems Biology, we try to elucidate various dynamical characteristics of biological functions with applying the omics data to detailed mathematical model based on the central dogma. However, such mathematical models possess multi-time-scale properties which are often accompanied by time-scale differences seen among biological layers. The differences cause time stiff problem, and have a grave influence on numerical calculation stability. In the present conventional method, the time stiff problem remained because the calculation of all layers was implemented by adaptive time step sizes of the smallest time-scale layer to ensure stability and maintain calculation accuracy. In this paper, we designed and developed an effective numerical calculation method to improve the time stiff problem. This method consisted of ahead, backward, and cumulative algorithms. Both ahead and cumulative algorithms enhanced calculation efficiency of numerical calculations via adjustments of step sizes of each layer, and reduced the number of numerical calculations required for multi-time-scale models with the time stiff problem. Backward algorithm ensured calculation accuracy in the multi-time-scale models. In case studies which were focused on three layers system with 60 times difference in time-scale order in between layers, a proposed method had almost the same calculation accuracy compared with the conventional method in spite of a reduction of the total amount of the number of numerical calculations. Accordingly, the proposed method is useful in a numerical analysis of multi-time-scale models with time stiff problem.

融合多特征信息与GWO-SVM的机械关键设备故障诊断

为了提高机械关键设备故障诊断的精度,建立机械关键设备故障诊断模型。文章提出一种融合机械关键设备故障信号多特征信息与灰狼优化算法(Grey Wolf Optimization Algorithm,GWO)改进支持向量机(Support Vector Machine,SVM)(GWO-SVM)的机械关键设备故障诊断模型。首先,提取机械关键设备故障信号的时域特征、频域特征和多尺度加权排列熵特征,分别对比不同特征的机械关键设备故障诊断结果。其次,为提高SVM模型性能,运用GWO算法对SVM模型的惩罚参数P和核函数参数g进行优化选择,提出一种融合多特征信息与GWO-SVM的机械设备故障诊断模型。与GA-SVM、PSO-SVM和SVM相比,基于GWO-SVM的机械设备故障诊断模型的诊断精度最高。这里算法可以有效提高机械关键设备故障诊断正确率,为机械关键设备故障诊断提供了新的方法。

沿海中小河流域地表-地下水库联合调控洪水资源利用研究:Ⅰ.方法与模型

针对沿海中小河流域地表-地下水库复杂工程系统,因缺少联合调控方法而导致洪水资源利用难的问题,解析地表-地下水库系统结构与功能,定量模拟地表水库调度、闸坝调控下河流和地下水动力过程,并利用智能替代模型提升水动力过程计算速度;在此基础上通过研发时/日/月多时间尺度模型耦合嵌套方法,构建了兼顾计算精度与效率的地表-地下水库联合调控模型。该模型实现了洪水、地表-地下水相互作用和供水等时间尺度差异明显的调节过程的统一调配,可为沿海中小河流域洪水资源利用提供新思路与技术支撑。

基于ResCNN-BiGRU的四川方言语音识别

由于基于深度卷积神经网络的语音识别模型中缺乏对特定方言音素特征的提取能力,造成方言发音底层特征部分信息丢失,进而导致方言识别准确率不高、鲁棒性差等问题。针对上述问题,提出一种结合残差网络(RestNet)和双向门控循环网络(BiGRU)的模型,该模型以GFCC特征图为输入,同时在残差网络中设计多尺度卷积模块,通过不同大小的卷积核提取特征,然后使用双向门控循环网络捕捉序列数据中的长期依赖关系,最后采用连接时序分类算法进行标签软对齐,实现四川方言语音识别模型。在四川方言语料库上的实验结果表明,提出的模型识别性能优于现有基准模型。

SSA-VMD与小波分解结合的GNSS坐标时序降噪方法

利用麻雀搜索算法(sparrow search algorithm, SSA)优化变分模态分解(VMD),然后结合小波分解(WD),提出一种GNSS坐标时间序列降噪方法IVMD-WD。利用仿真信号和10个基准站的实测数据进行GNSS坐标时间序列降噪实验。结果表明,IVMD-WD方法的降噪效果优于经验模态分解(EMD)、集合经验模态分解(EEMD)和WD,能够更加有效地剔除GNSS坐标时间序列中的噪声。

考虑动态频率约束的多能源电力系统日前-日内优化调度

提出考虑系统动态频率约束的含风电-电池储能-抽蓄电力系统协调优化调度策略。首先,建立含风电机组、电池储能和抽水蓄能的电力系统频率响应模型,并对系统调频响应过程进行分段线性化处理,降低计算的复杂度。然后,根据各机组的爬坡速度、启停约束、功率跟踪能力等因素的差异,建立日前-日内协调的多时间尺度优化调度模型。日前阶段制定考虑动态频率约束的火电机组和抽水蓄能的机组启停和出力计划;日内阶段优化电池储能的充放电状态计划,并调整火电和抽蓄的出力计划,促进风电消纳。最后,在含风电、变速抽蓄和电池储能的改进IEEE-39节点系统中进行算例分析,结果证明所提方法能提高系统频率稳定性,减少弃风量,兼顾电力系统经济调度和频率稳定目标。

基于EMD-DELM-LSTM组合模型的湖泊水位多时间尺度预测

针对水位时间序列具有线性与非线性混合、不确定性高等特点带来的预测困难问题,提出了一种基于经验模态分解(EMD)、长短时记忆网络(LSTM)和深度极限学习机(DELM)的EMD-DELM-LSTM组合模型,其中DELM和LSTM采用并联结构预测,并与EMD串联连接。首先使用EMD将原始信号分解为若干个具有单一特征的本征模态函数(IMFs),再将IMFs分类重组为高、中、低频信号后输入DELM-LSTM并联结构中进行预测并重构。以广州某大学重要湖泊为例验证模型的有效性,结果表明,与EMD-LSTM、EMD-DELM、LSTM、DELM和BiLSTM模型相比,本模型在不同时间尺度下的预测性能均有显著提升,其中40 min时间尺度下的预测性能提升效果最为明显,分别较对比模型提升43.08%、22.92%、45.79%、30.92%和47.31%。可见,本模型对于不同时间尺度的水位预测具有良好的可靠性和稳定性。

随机复合材料结构非线性热-力耦合模拟的统计高阶多尺度方法

对于具有复杂随机细观构造的复合材料结构的非线性热-力耦合问题的随机多尺度建模和计算仍是一个具有挑战性的问题。本文发展了一个新的统计高阶多尺度方法,克服了随机多尺度问题直接模拟时巨大的计算量,实现了具有随机复合材料结构非线性热-力耦合问题的数值模拟。借助统计多尺度渐近分析和泰勒级数方法,本文严格推导了可以精确分析随机复合材料结构宏-细观尺度非线性热-力耦合响应的统计高阶多尺度计算模型。然后,通过局部误差分析证明了统计高阶多尺度计算模型中高阶校正项在保持计算模型局部能量和动量守恒的重要意义。进一步,建立了可以高效模拟随机复合材料结构非线性热-力耦合行为的具有离线和在线两阶段的时空多尺度算法。最后,通过数值实验验证了统计高阶多尺度方法的计算高效率和高精度。

单变量时间序列的MHAGRU-MCCE分类方法

在单变量时间序列分类任务中,有效利用时间序列的多尺度特征、时间依赖特征对提高分类准确率至关重要。针对现有模型在综合利用多尺度特征、时间依赖特征方面局限,本文提出一种结合多尺度条件卷积增强模块(MCCE)与基于多头注意力机制的门控循环单元(MHAGRU)新型混合模型MHAGRU-MCCE,MCCE从不同尺度捕捉丰富时序特征,MHAGRU侧重于提取时间序列数据中依赖关系。在UCR的85个公共数据集上,与MACNN、AFFNet、OS-CNN、LITETime、MLP和LSTM-FCN等6种主流基于深度学习时间序列分类模型相比,验证表明MHAGRU-MCCE在平均准确率(MA)上分别提升0.66%、2.04%、3.45%、2.70%、12%和2.89%,并取得最高算术平均排名(AMR)=2.45、几何平均排名(GMR)=1.98,充分证明MHAGRU-MCCE在处理单变量时间序列分类问题上的有效性、优越性。

基于MTS-BiGRU-DMHSA的工业负荷预测方法

工业用电占我国全社会用电量比重大,通过工业负荷预测了解负荷趋势和用电量信息,有助于电网安全稳定运行,为电力部门发电规划提供依据,且有助于工业用户优化生产工艺和降低成本。为了兼顾工业负荷波动的不确定性以及工业用户用电行为的规律性特征,提出一种基于多时间尺度(MTS)特征的工业负荷预测方法MTS-BiGRU-DMHSA,利用MTS特征融合挖掘工业负荷的周期趋势特征和局部波动特征,提升工业负荷表征的可解释性。此外,双层多头自注意力(DMHSA)机制利用注意力权重聚焦重要特征,在挖掘输入特征关联性的同时捕捉时序关联性,强化重要特征变量与关键时间步的信息表达。在中国某工业企业五面受总柜实采数据上完成实验验证,采用2种评价指标对所提方法及5种基于神经网络的预测方法进行对比分析。实验结果表明,所提方法相较于对比方法平均误差降低逾20%,其中4.67%得益于对MTS特征的运用。通过对比各方法计算效率证实了所提方法的综合性能优势,可视化实验结果与对比分析也验证了其在工业负荷预测任务上的可行性。

含分布式电源配网多时间尺度电压-无功控制

针对大规模分布式电源接入配电网导致的网损增加、电压偏移增大和电压越限等问题,研究考虑不同种类和响应速度调压设备的多时间尺度电压-无功协同控制方法。在日前阶段,建立以最小化网损及电压偏差为目标,考虑有载调压分接头、电容器组等传统设备的日前优化模型,采用线性化和二阶锥方法进行高效求解;在日内阶段,考虑不同分布式电源及调压设备的调节成本差异,提出基于电压灵敏度原理的无功-有功高效电压控制模型。仿真算例表明,该方案能降低电压偏移和网络损耗(降低34.73%),当电压越限发生时,能根据电压灵敏度以及各设备调控成本优化控制目标,在调整电压的同时兼顾经济性。

基于滚动优化的电-热-气-冷系统多时间尺度低碳运行

电-热-气-冷多能联供型微网对实现能源可持续发展具有重要的应用价值。针对多能联供系统碳排放量较高和负荷模型预测不准确问题,提出了一种基于滚动优化的电-热-气-冷系统多时间尺度低碳运行策略。首先,建立电-热-气-冷系统设备模型。其次,构建日前与日内两阶段模型,在日前调度阶段引入含赏罚因数的碳交易机制,通过将卷积神经网络(convolutional neural networks,CNN)与双向长短期记忆网络(bi-directional long short term memory,Bi-LSTM)进行结合对风光功率进行预测,并以运行成本最低为目标进行优化。之后,建立日内多时间尺度的优化调度模型,以调度成本最低为目标进行求解。最后,以某市综合能源系统为研究对象进行分析。结果表明,所提出的方法能够有效减少碳排放,提高负荷模型预测的准确度的同时实现多能联供系统的低碳经济运行。

新型电力系统“源-网-荷-储”多元灵活性提升方法与路径

以高比例新能源为主体的新型电力系统在局部时段灵活性严重不足,需要挖掘各类资源的灵活性以提高电力系统多时间尺度可调节能力。探讨了不同时间尺度的灵活性供需平衡特性;总结了面向新型电力系统“源-网-荷-储”各环节的灵活性提升方法,突出各类资源的灵活性互补效果;分析了不同类型电力市场的交易机制相互配合衔接对提高电力系统多时间尺度灵活性的作用,可以联合提高灵活性资源的跨时空配置效率,有效释放电力系统多时间尺度灵活性。

基于Time-Causality模型的供热用气量预测分析

目前关于时间序列预测的特征选择一直是研究的热点,但很少有学者分析多时间尺度下不同特征对预测的差异。提出基于Granger关系的Time-Causality预测模型,利用Granger关系进行特征选择,引入时间维度作为输入维度,并利用LSTM模型进行实验,在多时间尺度下分析预测供热用气量的特征。实验结果表明:Time-Causality模型能筛选到更有助于用气量预测的特征;从不同的时间尺度预测,所选取的特征不同;每个特征的预测作用也可能会随时间尺度的变化而变化。这为长期和短期预测提供理论和实践支持。

考虑微电网群和需求响应的配电网日前-日内优化调度策略

微电网中光伏等新能源的渗透率不断提升,其间歇性与不确定性给配电网系统带来了巨大挑战。考虑配电网负荷预测的不确定性,根据不同时间尺度调控策略调用微电网内分布式资源来降低不确定性对配电网稳定性的影响。首先,针对负荷和光伏预测的不确定性和需求侧负荷的不同时间尺度特性,建立了配电网-微电网群-微电网的申报和决策方案,系统采用日前-日内多时间尺度滚动优化调度策略;然后,日前优化以配电网和微电网群的运行成本最小为目标,日内优化基于日前调度结果以微电网内可控资源调度成本最小为目标,对日前调度量进行修正;最后,采用目标级联法对所建立的模型进行求解,通过算例证明所提方法可以降低系统峰谷差和提高系统经济性。

计及电-气-热-氢需求响应的综合能源系统多时间尺度低碳运行优化策略

为更好地促进综合能源系统(IES)的低碳经济运行,提升多时间尺度运行优化管理水平,提出了一种考虑电-气-热-氢需求响应与阶梯式碳排放费用机制的多时间尺度低碳运行优化策略。综合考虑阶梯式碳排放费用机制,关注电、气、热之外的氢负荷需求,通过优化耦合设备中的燃气轮机热电比的可调性,建立了日前-日内滚动-实时三阶段的多时间尺度优化模型。以IES整体运行成本、碳排放费用成本、弃风弃光成本最小为经济目标,并将原非线性问题转化为混合整数线性问题,调用Gurobi求解器求解。通过仿真对比分析各类情形下的优化结果,验证了所提策略在促进低碳运行、发挥系统设备灵活性与促进清洁能源消纳方面的有效性。