A Two-Layer Encoding Learning Swarm Optimizer Based on Frequent Itemsets for Sparse Large-Scale Multi-Objective Optimization

Traditional large-scale multi-objective optimization algorithms(LSMOEAs)encounter difficulties when dealing with sparse large-scale multi-objective optimization problems(SLM-OPs)where most decision variables are zero.As a result,many algorithms use a two-layer encoding approach to optimize binary variable Mask and real variable Dec separately.Nevertheless,existing optimizers often focus on locating non-zero variable posi-tions to optimize the binary variables Mask.However,approxi-mating the sparse distribution of real Pareto optimal solutions does not necessarily mean that the objective function is optimized.In data mining,it is common to mine frequent itemsets appear-ing together in a dataset to reveal the correlation between data.Inspired by this,we propose a novel two-layer encoding learning swarm optimizer based on frequent itemsets(TELSO)to address these SLMOPs.TELSO mined the frequent terms of multiple particles with better target values to find mask combinations that can obtain better objective values for fast convergence.Experi-mental results on five real-world problems and eight benchmark sets demonstrate that TELSO outperforms existing state-of-the-art sparse large-scale multi-objective evolutionary algorithms(SLMOEAs)in terms of performance and convergence speed.

Transient responses of double-curved sandwich two-layer shells resting on Kerr's foundations with laminated three-phase polymer/GNP/fiber surface and auxetic honeycomb core subjected to the blast load

This work uses refined first-order shear theory to analyze the free vibration and transient responses of double-curved sandwich two-layer shells made of auxetic honeycomb core and laminated three-phase polymer/GNP/fiber surface subjected to the blast load.Each of the two layers that make up the double-curved shell structure is made up of an auxetic honeycomb core and two laminated sheets of three-phase polymer/GNP/fiber.The exterior is supported by a Kerr elastic foundation with three characteristics.The key innovation of the proposed theory is that the transverse shear stresses are zero at two free surfaces of each layer.In contrast to previous first-order shear deformation theories,no shear correction factor is required.Navier's exact solution was used to treat the double-curved shell problem with a single title boundary,while the finite element technique and an eight-node quadrilateral were used to address the other boundary requirements.To ensure the accuracy of these results,a thorough comparison technique is employed in conjunction with credible statements.The problem model's edge cases allow for this kind of analysis.The study's findings may be used in the post-construction evaluation of military and civil works structures for their ability to sustain explosive loads.In addition,this is also an important basis for the calculation and design of shell structures made of smart materials when subjected to shock waves or explosive loads.

Two-Layer Attention Feature Pyramid Network for Small Object Detection

Effective small object detection is crucial in various applications including urban intelligent transportation and pedestrian detection.However,small objects are difficult to detect accurately because they contain less information.Many current methods,particularly those based on Feature Pyramid Network(FPN),address this challenge by leveraging multi-scale feature fusion.However,existing FPN-based methods often suffer from inadequate feature fusion due to varying resolutions across different layers,leading to suboptimal small object detection.To address this problem,we propose the Two-layerAttention Feature Pyramid Network(TA-FPN),featuring two key modules:the Two-layer Attention Module(TAM)and the Small Object Detail Enhancement Module(SODEM).TAM uses the attention module to make the network more focused on the semantic information of the object and fuse it to the lower layer,so that each layer contains similar semantic information,to alleviate the problem of small object information being submerged due to semantic gaps between different layers.At the same time,SODEM is introduced to strengthen the local features of the object,suppress background noise,enhance the information details of the small object,and fuse the enhanced features to other feature layers to ensure that each layer is rich in small object information,to improve small object detection accuracy.Our extensive experiments on challenging datasets such as Microsoft Common Objects inContext(MSCOCO)and Pattern Analysis Statistical Modelling and Computational Learning,Visual Object Classes(PASCAL VOC)demonstrate the validity of the proposedmethod.Experimental results show a significant improvement in small object detection accuracy compared to state-of-theart detectors.

ATwo-LayerOptimal Scheduling Strategy forRural Microgrids Accounting for Flexible Loads

In the context of China’s“double carbon”goals and rural revitalization strategy,the energy transition promotes the large-scale integration of distributed renewable energy into rural power grids.Considering the operational characteristics of rural microgrids and their impact on users,this paper establishes a two-layer scheduling model incorporating flexible loads.The upper-layer aims to minimize the comprehensive operating cost of the rural microgrid,while the lower-layer aims to minimize the total electricity cost for rural users.An Improved Adaptive Genetic Algorithm(IAGA)is proposed to solve the model.Results show that the two-layer scheduling model with flexible loads can effectively smooth load fluctuations,enhance microgrid stability,increase clean energy consumption,and balance microgrid operating costs with user benefits.

Short-term Wind Speed Prediction with a Two-layer Attention-based LSTM

Wind speed prediction is of great importance because it affects the efficiency and stability of power systems with a high proportion of wind power.Temporal-spatial wind speed features contain rich information;however,their use to predict wind speed remains one of the most challenging and less studied areas.This paper investigates the problem of predicting wind speeds for multiple sites using temporal and spatial features and proposes a novel two-layer attentionbased long short-term memory(LSTM),termed 2Attn-LSTM,a unified framework of encoder and decoder mechanisms to handle temporal-spatial wind speed data.To eliminate the unevenness of the original wind speed,we initially decompose the preprocessing data into IMF components by variational mode decomposition(VMD).Then,it encodes the spatial features of IMF components at the bottom of the model and decodes the temporal features to obtain each component's predicted value on the second layer.Finally,we obtain the ultimate prediction value after denormalization and superposition.We have performed extensive experiments for short-term predictions on real-world data,demonstrating that 2Attn-LSTM outperforms the four baseline methods.It is worth pointing out that the presented 2Atts-LSTM is a general model suitable for other spatial-temporal features.

Joint inversion of gravity and magnetic data for a two-layer model

Based on the synchronous joint gravity and magnetic inversion of single interface by Pilkington and the need of revealing Cenozoic and crystalline basement thickness in the new round of oil-gas exploration, we propose a joint gravity and magnetic inversion methodfor two-layer models by concentrating on the relationship between the change of thicknessI and position of the middle layer and anomaly and discuss the effects of the key parameters. Model tests and application to field data show the validity of this method.

基于LSTM网络的单台仪器地震烈度预测模型

烈度是地震预警系统的关键产出.如何实现快速预测目标场址的地震烈度是地震预警方法技术研究中的核心问题.本文提出了一种基于长短时记忆神经网络(Long Short-Term Memory,LSTM)的单台仪器地震烈度的预测模型(LSTM-Ⅰ).该模型以一个台站观测到地震动参数的时间序列特征为输入,实现动态预测该台站可能遭受的最大烈度.选取了日本K-NET台网记录的102次地震的5103条强震加速度记录训练了神经网络,利用89次地震的3781条数据检验了模型的泛化能力.利用准确率、漏报率以及误报率三个评价指标评价了LSTM-Ⅰ模型的性能.结果表明,当采用P波触发后3 s的序列进行预测时,模型出现漏报的概率为46.78%,出现误报的概率为1.25%;当采用P波触发后10 s的序列进行预测时,模型出现漏报的概率大幅降低到17.6%,出现误报的概率降低到1.14%.结果表明LSTM-Ⅰ模型很好把握住了时间序列中蕴含的特征.进一步基于LSTM-Ⅰ模型评估了Ⅵ度下台站所能提供的预警时间.本文模型能够提供的预警时间与P-S波到时差接近,说明LSTM-Ⅰ模型具有较高的时效性.

基于LSTM算法的玉米籽粒储藏温度预测

为减少储粮损失和虫霉等的发生,该文利用自制试验仓及检测系统检测储藏玉米籽粒不同位置的粮温,分析粮堆温度变化及整个粮堆热量的传递过程。试验结果表明,仓外环境温度对粮温影响较大,仓内粮食温度变化与仓外环境温度变化相比较为滞后,粮堆第一层2号位置温度在检测周期中一直处于较低状态,温度最高位置出现在第四层12号位置。基于粮堆温度变化分析,该文开展了基于长短时记忆网络(LSTM)算法的玉米籽粒储藏粮温预测研究。结果表明:(1)对比预测值与试验值可知,粮堆第一、二、三、四层测试集的粮温准确率分别为0.62、0.89、0.83、0.79;(2)位于粮堆第二层和第三层的预测结果精度较高,试验仓粮堆底层和顶层温度易受环境温度影响,粮堆热量交换速度快,温度变化迅速,导致第一层和第四层预测结果精度偏低。该研究可为粮食储藏温度预测研究提供新思路。

基于Bayes-LSTM的公路隧道围岩变形预测方法研究

在公路隧道施工过程中,围岩的稳定性对隧道施工的影响较大。因此公路隧道围岩变形的监控量测与准确预测是保障隧道施工安全的关键。针对当前隧道围岩变形的预测精度较低以及泛化能力较差等问题,该文提出一种基于贝叶斯(Bayes)优化长短期记忆网络(LSTM)的方法,该方法首先对拱顶沉降和周边收敛的原始监测数据进行预处理,而后构建公路隧道拱顶沉降与周边收敛的初始LSTM模型,并利用Bayes优化模型中的超参数,最终得出预测结果。利用该模型对某公路隧道拱顶沉降和周边收敛进行预测,将预测结果以均方根误差为评价指标与神经网络(CNN)和支持向量回归(SVR)进行对比。预测拱顶沉降时,Bayes-LSTM模型的平均预测精度相较于CNN与SVR模型分别提高了1.0与1.26;预测周边收敛时,Bayes-LSTM模型平均精度相较于CNN与SVR分别提高了0.3与0.32。表明Bayes-LSTM模型的预测精度较高,同时其能在训练模型过程中对历史信息进行判断和取舍,极大地提高了时序数据处理的效率,为公路隧道围岩变形预测提供了新的思路和探索。

基于LSTM的高压输电电缆接头异常预警系统设计

高压输电电缆接头异常预警容易受到谐波电流影响,导致预警效果不佳,为此设计了基于LSTM的高压输电电缆接头异常预警系统。检测高压输电电缆接头异常电流谐波,获取电缆接头介质状态。设计CC2430微处理器预警模块,负责处理预警区域内电流数据。构建接地线电流检测模块,负责评估电缆接头异常。构建LSTM神经网络预警模型,对比分析实时状态与模型中检测点状态,获取当前电缆接头状态的预测值。采用移动平均技术分析电流数据的整体变化走向,判断电流预测数据与阈值关系,根据系统指示灯颜色区分异常等级。由实验结果可知,该系统电流数值最大值为125 A,变化曲线与实际曲线一致,说明该系统的预警效果较好。

基于CNN-LSTM的大坝变形组合预测模型研究

为了提高大坝变形预测模型精度和泛化能力,建立了一种基于卷积神经网络(Convolutional neural networks,CNN)与深度学习长短期记忆(Long short-term memory,LSTM)神经网络的组合预测模型CNN-LSTM。该模型先利用CNN提取大坝变形监测时间序列的特征,再利用LSTM生成特征描述,该模型精度高、泛化能力强。以柏叶口水库混凝土面板堆石坝为例,经过CNN-LSTM模型计算,将模型变形预测值与原型监测资料进行对比,再与LSTM模型及CNN模型的预测结果进行对比。结果表明,CNN-LSTM模型预测值最接近监测资料实测结果。

基于融合注意力机制LSTM网络的地下水位自适应鲁棒预测

地下水水位是旱天污水管网地下水入渗量的重要影响因素,快速精准地预测地下水水位能有效提升旱天污水管网地下水入渗量估算准确度,辅助优化管网病害治理与维护策略。针对目前城市复杂水文预测存在的准确度低、灵敏度低、泛化能力弱等问题,本文提出了一种新的鲁棒自适应水位预测算法。首先,对水文数据进行预处理,解决了数据时间跨度大、噪声多、缺失及异常、非平稳等问题。其次,针对不同输入特征对预测指标的影响,在模型训练阶段提出一种新的空间变量注意机制,可快速识别与水位关联的关键变量,并对输入特征赋予不同的影响权重。然后,针对不同序列长度对预测效果的影响,还设计了自适应时间注意力机制,帮助网络自适应地找出与不同时间序列长度预测指标相关的编码器隐藏状态,以更好地捕捉时间上的依赖关系。在此基础上,以上下文向量作为输入,提出一种融合注意力机制的长短时记忆网络水文预测算法。最后,通过意大利Petrignano水文数据验证了所提算法的有效性,并与GRU、Elman、LSTM、VA–LSTM和S–LSTM等方法进行预测性能比较。结果表明,基于融合注意力机制的LSTM网络在面临大规模、噪点多的复杂数据时有优于其它几种算法的预测效果,表明该算法具有强自适应性和鲁棒性。本文研究结果可以为市政排水策略合理调整、及时控制提供参考。

基于FFT-LSTM的抽水蓄能发电机定子匝间短路故障诊断方法

定子短路故障是抽水蓄能发电机常见的故障之一,其会对发电机的性能和安全性产生严重影响,为确保抽水蓄能发电机安全稳定运行,提出基于FFT-LSTM的抽水蓄能发电机定子短路故障诊断方法。建立抽水蓄能发电机定子绕组匝间短路故障模型,分析定子绕组匝间短路故障时,发电机定子电、磁相关状态。以抽水蓄能发电机定子绕组匝间短路故障时的三相电流信号为依据,基于磁势相等原理将三相电流变换成两相电流后,利用FFT转换定子两相电流的时域信号为频域信号,获取故障电流频谱图输入LSTM网络中进行处理,输出抽水蓄能发电机定子绕组匝间短路故障诊断结果。实验结果表明,该方法可以更好地区分抽水蓄能发电机正常与故障状态,实现抽水蓄能发电机定子绕组匝间短路故障诊断,且故障诊断的交叉熵损失低。

基于AM-LSTM的飞行区航空器滑行轨迹预测与冲突识别

为解决航空器点源定位难以有效预测而引发冲突风险愈来愈多的问题,构建基于注意力机制(AM)和长短期记忆网络(LSTM)的时间序列轨迹预测模型AM-LSTM,预测未来短时间内飞行区航空器的瞬时点源位置;在此基础上,根据航空器型号和滑行航向对其进行轮廓扩展,以航空器速度作为安全距离权重,通过射线法实现轮廓冲突的判定;并以乌鲁木齐地窝堡机场为例进行验证,利用训练完成的轨迹预测模型预测飞行区航空器滑行轨迹,以识别航空器轮廓间的滑行冲突。结果表明:AM-LSTM预测模型能够准确预测飞行区航空器运动轨迹。未来3 s内轨迹位置预测的平均位移误差为1.05 m,轨迹点位置预测精准性可达94.37%,故能在轨迹预测的基础上精确识别滑行冲突风险,有利于保障飞行区的安全运行。

基于注意力机制的CNN-BiLSTM的IGBT剩余使用寿命预测

针对绝缘栅双极型晶体管(IGBT)可靠性问题,提出了一种融合卷积神经网络(CNN)、双向长短期记忆(BiLSTM)网络和注意力机制的剩余使用寿命(RUL)预测模型,可用于IGBT的寿命预测。模型中使用CNN提取特征参数,BiLSTM提取时序信息,注意力机制加权处理特征参数。使用IGBT加速老化数据集对提出的模型进行验证。结果表明,对比自回归差分移动平均(ARIMA)、长短期记忆(LSTM)、多层LSTM(Multi-LSTM)、 BiLSTM预测模型,在均方根误差和决定系数等评价指标方面该模型的性能最优。验证了提出的寿命预测模型对IGBT失效预测是有效的。

基于GPR-LSTM的地震热红外背景场的构建方法

地震监测是一项非常重要且具有挑战性的任务,遥感技术的不断发展加强了在宏观尺度上对地球表面的监测能力。研究表明,地震前通常都会出现地表温度异常升高的现象,因此各种异常提取算法被应用于地震热异常研究中。其中,基于背景场的提取方法由于具有较强的机理解释性而受到广泛应用。然而,以往基于背景场的异常提取方法更多将背景场限定于某一固定阈值,忽略了受外界因素(非震)影响导致的地表温度的小范围正常波动。据此,文中提出了一种基于GPR-LSTM的地震热红外背景场的构建方法。该方法包括两大部分:震期年变基准场的建立、实际LST的残差波动范围计算及背景场的构建。基于MODIS地表温度产品,以2008年四川汶川和新疆于田地震为研究对象,使用所述方法对地震前兆热异常信息进行提取与分析,经过实验得出以下结论:1)地震热异常通常沿青藏高原的断层分布,这不仅证明了文中方法能够减弱地表温度数据中噪声的干扰,同时也证明该方法在热异常信息提取方面的有效性;2)地震年份的构造活动比非地震年份更加活跃,导致地表温度的异常增温更加明显;3)不同地震案例震前的热异常时空特征各不相同。

Effective Moduli of Two-Layered Spherical Inclusions Reinforced Composites

A model is proposed to evaluate the,effective modufi of a composite reinforced by two-layered spherical inclusions.This model is based on the localisation problem of a two- layered spherical inclusion embedded in an infinite matrix.The interations of the reinforced phases are taken into account by using the average matrix stress concept.When the external layer vanishes,the proposed model reduces to the classical Mori-Tanaka's model for spherical inclusions.Theoretical results for the composite of polyester matrix filled by hollow glass spheres and voids show excellent agreement with experimental results.

Prophet-LSTM组合模型在运输航空征候预测中的应用

为准确预测中国运输航空征候万时率,提出了一种将时间序列模型和神经网络模型组合的预测方法。首先,利用2008年1月—2020年12月的运输航空征候万时率数据建立Prophet模型,使用RStudio软件进行模型拟合,获取运输航空征候万时率的线性部分;其次,利用长短期记忆网络(Long Short-Term Memory,LSTM)建模,获取运输航空征候万时率的非线性部分;最后,利用方差倒数法建立Prophet-LSTM组合模型,使用建立的组合模型对2021年1—12月运输航空征候万时率进行预测,将预测结果与实际值进行对比验证。结果表明,Prophet-LSTM组合模型的EMA、EMAP、ERMS分别为0.0973、16.1285%、0.1287。相较于已有的自回归移动平均(Auto Regression Integrated Moving Average,ARIMA)+反向传播神经网络(Back Propagation Neural Network,BPNN)组合模型和GM(1,1)+ARIMA+LSTM组合模型,Prophet-LSTM组合模型的EMA、EMAP、ERMS分别减小了0.0259、10.4874百分点、0.0143和0.0128、2.0599百分点、0.0086,验证了Prophet-LSTM组合模型的预测精度更高,性能更优良。

基于SSA−LSTM的风速异常波动检测方法

针对传统统计方法对风门开闭导致传感器监测数据异常波动的漏报率和误报率高的问题,通过挖掘风速传感器中时间序列数据中的数据特征,提出了一种基于奇异谱分析法(SSA)与长短期记忆神经网络(LSTM)组合的SSA−LSTM风速异常波动检测方法。首先利用SSA对风速传感器监测数据进行预处理,将风速数据分解为趋势分量、周期分量和噪声分量,通过重组趋势分量和噪声分量去除因湍流脉动产生的数据噪声;然后对LSTM进行参数优化,利用优化后的LSTM模型对预处理数据进行预测并得到重构风速;最后以对数概率密度函数计算监测风速与重构风速的异常分数,通过计算训练集数据样本的异常分数设定阈值对监测风速进行异常检测。试验结果表明:SSA去除因湍流脉动产生的数据噪声效果较好,在不影响数据波动情况下去除噪声分量,有助于提高风速重构效果和异常检测准确率;LSTM在无异常波动时能正确重构因湍流脉动导致的小幅波且与实际数据拟合效果较好,在有异常波动时根据历史波动趋势对异常波动段进行重构,可有效提高异常检测的准确率。通过对比分析,所提方法比ARIMA、BP、CNN模型的重构效果更好,异常检测准确率为99.2%,F1-Score为0.97,验证了所提方法的可靠性。表明本文所提方法在检测因风门开闭导致的风速异常波动上具有一定的应用价值。

改进MFO-LSTM网络的风电机组齿轮箱故障预警研究

风电机组齿轮箱在数据采集与监控系统(SCADA)的帮助下,通过监控齿轮箱油温是否超过阈值实现故障报警,其判断精度不高且问题发现不及时,因此使用长短期记忆网络模型(LSTM)融合SCADA数据实现对齿轮箱油温状态的预测。用齿轮箱正常运行状态下的数据训练LSTM模型,计算油温预测值与真实值之间的残差,根据正态分布的原则设置残差的上下预警阈值,用来对齿轮箱故障进行预警。为简化训练模型的复杂度,在SCADA数据中选用与齿轮箱油温相关性较为密切的参数作为LSTM模型的输入项。为降低因LSTM模型超参数设置不当造成的预测准确度表现不佳,提出改进飞蛾火焰算法(MFO)与LSTM的组合模型,在保留MFO算法强大的全局搜索能力的同时,使其避免陷入局部搜索的陷阱,通过改进MFO对LSTM模型参数进行迭代优化,最终构建合适的模型。最后通过某风电机组SCADA数据验证该方法能够有效预警齿轮箱的故障,并且与其他方法相比准确度更高,预警更及时,迭代效果更好。