基于DAM和CNN-LSTM的辅助动力装置性能参数预测模型

针对辅助动力装置(auxiliary power unit, APU)性能参数难以准确预测的问题，提出一种基于特征与时序的双阶段注意力机制(dual-stage attention mechanism,DAM)和卷积神经网络(CNN)-长短期记忆网络(LSTM)的混合模型。所提的方法在特征提取阶段加入了通道注意力机制(Channel attention mechanism ,CAM)；输出阶段加入了时序注意力机制(Temporal attention mechanism ,TAM)，加强了CNN对重要特征的提取能力和历史关键信息对预测输出的影响，并利用改进的粒子群算法对模型关键参数寻优，提高预测精度。实验结果表明，所提出的新方法在多变量输入和多步长的APU排气温度(exhaust gas temperature，EGT)预测中均取得了很好的效果，预测精度大幅提高。     

An optimized long short-term memory network based fault diagnosis model for chemical processes

With the development of the chemical industry, fault diagnosis of chemical processes has become a challenging problem because of the high-dimensional data and complex time correlation caused by the more complex chemical processes and increasing number of equipment. However, the ordinary feedforward neural network cannot solve these problems very well. Therefore, this paper proposes a fault diagnosis model based on the optimized long short-term memory (LSTM) network. Since the number of hidden layer nodes in the LSTM network has a great influence on the diagnosis result, the link of determining the optimal number of hidden layer nodes by the iterative method based on the LSTM network is added. Then the LSTM is optimized to get higher chemical process fault diagnosis accuracy. Finally, through the simulation experiment of the Tennessee Eastman (TE) chemical process, the results verify that the optimized LSTM network has better performance in chemical process fault diagnosis than the BP neural network, the multi-layer perceptron method and the original LSTM network.     

一种TCN的改进模型及其在短期光伏功率区间预测的应用

为了提高光伏功率预测的精度,提出了一种基于时序卷积网络(temporal convolutional network, TCN)的新型短期光伏功率区间预测模型。首先,采用深度残差收缩网络(deep residual shrinkage network, DRSN)的软阈值和注意力机制来改进TCN的残差模块以增强其对有用特征提取能力,并削弱冗余特征的不利影响;然后,利用樽海鞘群算法(slap swarm algorithm, SSA)对TCN的卷积层的卷积核大小和TCN层数等超参数进行自动寻优,以克服原TCN感受野不足的问题;接着,采用核密度估计(kernel density estimation, KDE)方法对所建改进TCN短期光伏功率预测模型的点预测结果进行误差分析,获得模型预测输出的区间。最后,通过对比仿真实验得到的结果表明,提出的SSA-DRSN-TCN模型的RMSE平均值为0.27,优于LSTM、GRU、CNN-LSTM和TCN等模型;而且,KDE方法能够在80%、90%和95%的置信度下准确描述光伏功率波动区间,验证了所提模型在提高光伏功率预测性能上的有效性。     

A novel hybrid model for short-term wind power forecasting

Wind energy prediction has a significant effect on the planning, economic operation and security maintenance of the wind power system. However, due to the high volatility and intermittency, it is difficult to model and predict wind power series through traditional forecasting approaches. To enhance prediction accuracy, this study developed a hybrid model that incorporates the following stages. First, an improved complete ensemble empirical mode decomposition with adaptive noise technology was applied to decompose the wind energy series for eliminating noise and extracting the main features of original data. Next, to achieve high accurate and stable forecasts, an improved wavelet neural network optimized by optimization methods was built and used to implement wind energy prediction. Finally, hypothesis testing, stability test and four case studies including eighteen comparison models were utilized to test the abilities of prediction models. The experimental results show that the average values of the mean absolute percent errors of the proposed hybrid model are 5.0116% (one-step ahead), 7.7877% (two-step ahead) and 10.6968% (three-step ahead), which are much lower than comparison models.     

基于CNN-LSTM的短期风电功率预测

短期风电功率预测对电力系统的安全稳定运行和能源的优化配置具有重要意义。鉴于卷积神经网络(CNN)高效的数据特征提取能力,以及长短期记忆网络(LSTM)描述时间序列长期依赖关系的能力。为了提高短期风电功率预测的精度,设计了一种基于CNN和LSTM的风电功率预测模型。该模型利用卷积神经网络对风电功率、风速、风向数据进行多层卷积和池化堆叠计算,提取风电功率相关数据的特征图谱。为了描述风电功率序列的时序依从关系,将图谱特征信息作为长短期记忆网络的输入信息,计算得到风电功率的预测结果。采用西班牙某风电场的实测数据进行模型预测精度验证。结果表明,该模型较LSTM、Elman模型具有更好的预测性能。     

结合注意力机制与LSTM的短期风电功率预测模型

风力发电预测在电力系统的运行中发挥着重要作用。现有风电功率的短期预测模型因风速的复杂性和随机性,难以确定风速与风电功率的非线性映射关系,导致预测精度降低。提出一种结合变分模态分解、双阶段注意力机制、误差修正模块与深度学习算法的短期风电功率预测模型。通过对原始数据进行互信息特征选择,获得与风电功率相关性较强的特征,并对其进行信号预处理,利用变分模态分解对多维特征序列进行分解,得到具有一定中心频率的模态分量,以降低各个特征序列的复杂性和非平稳性。采用基于双阶段注意力机制与编解码架构的长短时记忆（LSTM）神经网络对模态分量进行训练与预测,得到初始预测误差。在此基础上,利用误差修正模块对初始预测误差进行变分模态分解和修正,从而提高模型的预测精度。实验结果表明,与自回归移动平均模型、标准编解码结构的LSTM模型相比,该预测模型的平均绝对误差最高可降低约87%,具有较优的预测性能。     

LSTM模型在电力系统负荷预测中的应用

为解决数据随时间变化的电力系统短期负荷预测问题,本文阐述、采用和总结线性回归,普通BP神经网络,GRU神经网络,LSTM神经网络在电力系统短期负荷预测中应用的理论基础和方法,分别针对神经网络的多种预测方法进行归纳总结.LSTM即长短期记忆网络,拥有门机制,可以选择性的遗忘和记忆过去的信息,特别能记忆一个序列时间段的信息...     

基于PSO和BP网络的LF炉钢水温度智能预测

研究将粒子群优化算法与BP神经网络相结合,建立新的钢水终点温度预报模型.确定加热功率、初始温度、精炼时间等8个影响钢水终点温度的主要因素作为神经网络的输入量.用粒子群优化算法优化神经网络参数,改善神经网络温度预测模型的收敛性能.实验结果表明,该算法可以提高预测速度和精度,预测结果误差不大于±5℃的炉次大于90%.     

Fuzzy-Pareto-dominance driven possibilistic model based planning of electrical distribution systems using multi-objective particle swarm optimization

This paper presents a fuzzy-Pareto dominance driven possibilistic model based planning of electrical distribution systems using multi-objective particle swarm optimization (MOPSO). This multi-objective planning model captures the possibilistic variations of the system loads using a fuzzy triangular number. The MOPSO based on the Pareto-optimality principle is used to obtain a set of non-dominated solutions representing different network structures under uncertainties in load demands and these non-dominated solutions are stored in an elite archive of limited size. Normally, choosing the candidate non-dominated solutions to be retained in the elite archive while maintaining the quality of the Pareto-approximation front as well as maintaining the diversity of solutions on this front is very much computationally demanding. In this paper, the principles of fuzzy Pareto-dominance are used to find out and rank the non-dominated solutions on the Pareto-approximation front. This ranking in turn is used to maintain the elite archive of limited size by discarding the lower ranked solutions. The two planning objectives are: (i) minimization of total installation and operational cost and (ii) minimization of risk factor. The risk factor is defined as a function of an index called contingency-load-loss index (CLLI), which captures the effect of load loss under contingencies, and the degree of network constraint violations. The minimization of the CLLI improves network reliability. The network variables that are optimized are: (i) number of feeders and their routes, and (ii) number and locations of sectionalizing switches. An MOPSO (developed by the authors), based on a novel technique for the selection and assignment of leaders/guides for efficient search of non-dominated solutions, is used as the optimization tool. The proposed planning approach is validated on a typical 100-node distribution system. Performance comparisons between the planning approaches with the possibilistic and deterministic load models are provided highlighting the relative merits and demerits. It is also verified that the proposed solution ranking scheme based on the fuzzy-Pareto dominance is very much better from both quality and computational burden point of view in comparison with the other well-known archive truncation techniques based on clustering and solution density measurement etc.     

基于自适应模糊神经网络的功放预失真新方法

针对无线通信系统中记忆非线性功率放大器预失真结构不足和精度不高等问题,提出了一种基于模糊神经网络模型识别的双环学习结构自适应预失真方法。该方法以实数延时模糊神经网络模型为基础,采用改进的简化粒子群优化（Simplified Particle Swarm Optimization,SPSO）算法进行间接学习结构离线训练模糊神经网络来确定模型参数,作为预失真器的初值,再利用最小均方（Least Mean Square,LMS）算法进行直接学习结构在线微调整预失真器参数,拟合功放的非线性和记忆效应。该方法结构简单,收敛速度快且精度高,避免了局部最优。实验结果表明,该方案邻信道功率比经典的双环结构预失真方法约改善7 dB,功放的线性化性能明显提高,由此验证了其可行性。     

消纳大规模风电的电–气互联系统鲁棒区间调度模型与方法

现有的电–气互联系统联合调度模型在利用燃气机组的快速调整能力应对风电波动时,难以考虑到对气网运行约束的影响,同时往往忽略了电转气(P2G)装置进一步消纳风电的能力.为解决上述问题,本文计及燃气机组和P2G装置追踪风电波动提出了一种鲁棒区间调度模型,所提模型在保证风电消纳区间最大化的同时,将优化得到的燃气机组和P2G装置允许出力区间转换为最大、最小进气量场景以进一步校验气网的可行性.为进一步扩大风电消纳区间,本文将AGC机组和P2G装置的风电承担系数作为待优化变量,并对因此引入的非线性项进行松弛处理.最后,通过一种计及再调整的日后校正模型对各个算例场景进行蒙特卡洛模拟,验证了本文模型和方法的有效性.     

PSO优化BP神经网络的混沌时间序列预测

针对于BP神经网络预测模型,收敛速度慢,精度较低,容易陷入局部极小值等缺点,提出了一种改进粒子群优化BP神经网络预测模型的算法。在该算法中,粒子群采用改进自适应惯性权重和改进自适应加速因子优化BP神经网络预测模型的初始权值和阈值,然后训练BP神经网络预测模型并预测。将该算法应用到几个典型的混沌时间序列预测。实验结果表明,该算法明显提高BP神经网络预测模型的收敛速度和预测模型的精度,减少陷入局部极小的可能。     

基于分组混沌PSO算法的模糊神经网络建模研究

为改善记忆功放建模的精度,且针对粒子群算法早期收敛速度较快,但在后期易陷入早熟收敛,局部最优等特点,提出了一种分组并行混沌粒子群优化算法（Grouping Parallel-Chaotic Particle Swarm Optimization,GP-CPSO）,将分组粒子群优化算法与混沌思想相结合,并用该算法优化动态模糊神经网络（Dynamic Fuzzy Neural Network,DFNN）参数,建立DFNN功放模型。引入分组的CPSO群算法,将种群划分为若干个组,每组单独计算,大大提高了收敛速度,同时将混沌思想运用到每个粒子当中去,避免早熟和局部最优,缩短了迭代时间。通过仿真结果可以看到,GP-CPSO优化后的动态模糊神经网络建模的训练误差减小到0.1以内,收敛速度提高32.5%,从而验证了这种建模方法有效且可靠。     

Prediction of silicon content in hot metal using support vector regression based on chaos particle swarm optimization

The prediction of silicon content in hot metal has been a major study subject as one of the most important means for the monitoring state in ferrous metallurgy industry. A prediction model of silicon content is established based on the support vector regression (SVR) whose optimal parameters are selected by chaos particle swarm optimization. The data of the model are collected from No. 3 BF in Panzhihua Iron and Steel Group Co. of China. The results show that the proposed prediction model has better prediction results than neural network trained by chaos particle swarm optimization and least squares support vector regression, the percentage of samples whose absolute prediction errors are less than 0.03 when predicting silicon content by the proposed model is higher than 90%, it indicates that the prediction precision can meet the requirement of practical production.     

基于改进的蝗虫优化算法的红细胞供应预测方法

当前国内"血荒"问题比较严峻,血站与用血单位之间存在着血液供不应求的现象。针对这个问题,提出了一种基于改进的蝗虫优化算法的LSTM预测方法,用于对未来的红细胞供应情况进行预测,为血站工作人员在制定采血计划以及制备计划时提供有效的指导。该预测模型通过使用长短期记忆网络(Long-Short Term Memory Network,LSTM)来捕捉历史红细胞库存数据之间的潜在规律,以达到对未来的供应情况进行预测的效果。首先,针对蝗虫优化算法容易陷入局部最优、收敛速度较慢的问题,通过加入基于折射原理的反向学习机制与混沌映射,加快蝗虫优化算法的收敛速度,使其具备更强的搜索能力。其次,为提高LSTM的预测性能,将改进的蝗虫优化算法与LSTM相结合,并使用某地区的红细胞库存真实数据作为实验数据,用于验证改进的LSTM预测模型的性能。与标准LSTM相比,所提方法的MAE,MAPE,RMSE分别降低了39.827 8,1.10%,55.819 1。实验结果证明,提出的方法具有较高的可靠性。     

基于PSO—BP神经网络的混凝土抗压强度预测

为了有效提高混凝土抗压强度的预测精准度,利用粒子群算法优化BP神经网络初始权值和阈值,建立了混凝土抗压强多因子PSO-BP预测模型。模型以每立方混凝土中水泥、高炉矿渣粉、粉煤灰、水、减水剂、粗集料和细集料的含量以及置放天数为输入参数,混凝土抗压强度值作为输出参数,不仅可以克服BP算法收敛速度慢和易陷入局部极值的缺陷,而且模型的学习能力、泛化能力和预测精度都有了很大的提高。以UCI数据库中的Concrete CompressiveStrength数据集为例进行仿真测试,结果表明：PSO-BP模型预测精度较BP、GA—BP模型分别提高了8．26％和2．05％,验证了PSO—BP模型在混凝土抗压强度预测中的有效性。     

基于双向LSTM卷积网络与注意力机制的自动睡眠分期模型

针对现阶段深度睡眠分期模型存在的梯度消失、对时序信息学习能力较弱等问题,提出一种基于双向长短时记忆卷积网络与注意力机制的自动睡眠分期模型。将少样本类别的睡眠脑电数据通过过采样方式进行数据增强后,利用带残差块的卷积神经网络学习数据特征表示,再通过带注意力层的双向长短时记忆网络挖掘深层时序信息,使用Softmax层实现睡眠分期的自动判别。实验使用Sleep-EDF数据集中19晚单通道脑电信号对模型进行交叉验证,取得了较高的分类准确率和宏平均F₁值,优于对比方法。该方法能够有效缓解睡眠分期判别中少数类分类性能较低的问题,并提高了深度睡眠分期模型的整体分类性能。     

改进粒子群算法在无功优化中的应用

针对电力系统无功优化的特点,本文提出以有功网损最小为目标函数,以负荷节点电压质量和PV发电机节点无功出力为罚函数．以有功功率和无功功率为约束条件的数学模型,并应用改进的粒子群算法对无功优化问题进行求斛。该算法在权重系数和不活动粒子两方面进行改进,有效地解决了进化过程中陷入局部最优和搜索精度差的缺点。最后,将改进后的粒子群算法应用于IEEE14节电力系统进行无功优化算例分析,仿真结果验证了该算法解决电力系统无功优化问题的有效性和可行性。     

Parameter optimization of sub-35 nm contact-hole fabrication using particle swarm optimization approach

The major research focus on integrated circuits (ICs) mainly deals with increasing circuit performance and functional complexity of circuit. The lithography process is the most critical step in the fabrication of nanostructure for integrated circuit manufacturing. The most important variable in the lithography process is the line-width or critical dimensions (CDs), which perhaps is one of the most direct impact variables on the device performance and speed. This study presents a hybrid approach combining Taguchi’s robust design, back-propagation neural network modeling technique and particle swarm optimization (PSO) for sub-35 nm contact-hole fabrication in the lithography process. The BP neural network is employed to model the functional relationship between the input parameters and target responses. Particle swarm optimization is adopted to optimize the parameter settings through the well-trained BP model, where each particle is assessed using fitness function. The proposed PSO algorithm applies the velocity updating and position updating formulas to the population composed of many particles such that better particles are generated. Compared with realistic fabricated and measured data, this approach can achieve the optimal parameter settings for minimized CDs or target CDs. Meanwhile, it reduces the CD variation through the design of experiment. The experimental results show that the proposed approach dealing with the process modeling and parameter optimization demonstrates its feasibility and effectiveness for sub-35 nm contact-hole fabrication.     

基于神经网络的粒子群算法优化SVM参数问题

对支持向量机的核参数选取到目前仍没有形成一套成熟的理论,严重影响了其广泛的应用。对核参数的选取做了一定的探讨。将神经网络与粒子群优化算法相结合并用于支持向量机核函数的参数优化。该方法能够同时具有神经网络较强的非线性拟合能力和粒子群优化算法的寻优能力。数值实验结果表明该算法对支持向量机核参数的优化是可行的、有效的,并且具有较高的分类准确率和较好的推广性能。