NEURAL NETWORK INTELLIGENT SYSTEM FOR THE ON-LINE OPTIMIZATION IN CHEMICAL PLANTS

A strategy of developing on-line optimization intelligent systems based on combiningflowsheeting simulation and optimization package with artificial neural networks(ANN)is presented inthis paper.A number of optimization cases for a certain chemical plant are obtained off-line byusing PROCESS-Ⅱ or other flowsheeting programming with optimization.Then,taking these cases astraining examples,we establish a neural network systems which can be used on-line as an optimizer toobtain setpoints from input data sampled from distributed control system through gross error detectionand data reconciliation procedures.Such an on-line optimizer possesses two advantages over nonlinearprogramming package:first of all,there is no convergence problem for the trained ANN to be usedonline;secondly,the frequency for setpoints updating is not limited because only algebraic calculationrather than optimization is required to be carried out on-line.Here two key problems ofimplementing ANN approaches to the on-line optimization

Transient Fault Locating Method Based on Line Voltage and Zero-mode Current in Non-solidly Earthed Network

为解决配电网小电流接地故障时故障点定位困难的问题，提出一种基于线电压和零模电流暂态分量的故障定位新方法。该方法借助馈线自动化（feeder automation，FA）系统实现，不需要额外增加设备。馈线终端检测线电压和零模电流信号，利用暂态线电压的希尔伯特变换与暂态零模电流的乘积计算故障方向参数，FA主站根据故障点前后方向参数极性相反的特征确定故障点所在的线路区段。介绍基于线电压的小电流接地故障检测和接地故障相确定的算法。数字仿真与试验结果表明该方法是正确可行的。

Power Line Monitoring Data Transmission Using Wireless Sensor Network

The WSN used in power line monitoring is long chain structure, and the bottleneck near the Sink node is more obvious. In view of this, A Sink nodes’ cooperation mechanism is presented. The Sink nodes from different WSNs are adjacently deployed. Adopting multimode and spatial multiplexing network technology, the network is constructed into multi-mode-level to achieve different levels of data streaming. The network loads are shunted and the network resources are rationally utilized. Through the multi-sink nodes cooperation, the bottlenecks at the Sink node and its near several jump nodes are solved and process the competition of communication between nodes by channel adjustment. Finally, the paper analyzed the method and provided simulation experiment results. Simulation results show that the method can solve the funnel effect of the sink node, and get a good QoS.

ON-LINE STATE PREDICTION OF ENGINES BASED ON FLAT NEURAL NETWORK

A flat neural network is designed for the on line state prediction of engine. To reduce the computational cost of weight matrix, a fast recursive algorithm is derived according to the pseudoinverse formula of a partition matrix. Furthermore, the forgetting factor approach is introduced to improve predictive accuracy and robustness of the model. The experiment results indicate that the improved neural network is of good accuracy and strong robustness in prediction, and can apply for the on line prediction of nonlinear multi input multi output systems like vehicle engines.

ANALYSIS OF DYNAMIC CHARACTERISTICS FOR RCG LINES USING THEORY OF LADDER NETWORKS

The theory of RC uniform ladder networks based upon the recurrence of voltage and cur-rent functions is extended as a vehicle to analyse the dynamic characteristics of reg lines. Meth-ods for computing the time constants and simplifying the transfer functions for reg lines are alsopresented.

Hybrid Neural Network Architecture for On-Line Learning

Approaches to machine intelligence based on brain models use neural networks for generalization but they do so as signal processing black boxes. In reality, the brain consists of many modules that operate in parallel at different levels. In this paper we propose a more realistic biologically inspired hybrid neural network architecture that uses two kinds of neural networks simultaneously to consider short-term and long-term characteristics of the signal. The first of these networks quickly adapts to new modes of operation whereas the second one provides more accurate learning within a specific mode. We call these networks the surfacing and deep learning agents and show that this hybrid architecture performs complementary functions that improve the overall learning. The performance of the hybrid architecture has been compared with that of back-propagation perceptrons and the CC and FC networks for chaotic time-series prediction, the CATS benchmark test, and smooth function approximation. It is shown that the proposed architecture provides a superior performance based on the RMS error criterion.

Fault Line Selection Method Considering Grounding Fault Angle for Distribution Network

In the distribution network system with its neutral point grounding via arc suppression coil, when single-phase grounding fault occurred near zero-crossing point of the phase voltage, the inaccuracy of the line selection always existed in existing methods. According to the characteristics that transient current was different between the fault feeder and other faultless feeders, wavelet transformation was performed on data of the transient current within a power frequency cycle after the fault occurred. Based on different fault angles, wavelet energy in corresponding frequency band was chosen to compare. The result was that wavelet energy in fault feeder was the largest of all, and it was larger than sum of those in other faultless feeders, when the bus broke down, the disparity between each wavelet energy was not significant. Fault line could be selected out by the criterion above. The results of MATLAB/simulink simulation experiment indicated that this method had anti-interference capacity and was feasible.

On-line Condition Monitoring and Diagnostic Network System for Rotating Machine Set

Several industrial computers and a server are combined to set up the on-line monitoring and diagnostic system of turbo-generator sets. The main function of the system is to monitor machine sets' running condition. Through analyzing running data, technicians can detect whether there exist faults and where they occur. To share and transmit the dynamic information of the turbo-generator sets, a distributed network system is introduced. NetWare network operating system is used in the LAN (Local Area Network) system. The LAN is extended to realize the sharing of data and remote transmission of information. Furthermore, functions of monitoring and diagnostic clients are listed.

基于信号图像化和CNN-ResNet的配电网单相接地故障选线方法

配电网发生单相接地故障时,零序电流呈现较强的非线性与非平稳性,故障选线较为困难,针对此问题,提出一种基于信号图像化和卷积神经网络-残差网络的配电网单相接地故障选线方法。首先,利用排列熵优化变分模态分解算法的参数,将零序电流信号分解成一系列固有模态函数;其次,引入新的数据预处理方式,将固有模态函数转成二维图像,获得零序电流信号的时频特征图;最后,利用一维卷积神经网络提取零序电流信号的相关性和特征,利用残差网络提取时频特征图的特征,将两个网络融合,构建混合卷积神经网络结构,实现故障选线。仿真与实验结果表明,该方法能够在高阻接地、采样时间不同步、强噪声等情况下准确地选择出故障线路,可满足配电网对故障选线准确性和可靠性的需求。

考虑柔性负荷整形的含源配电线路升级与开关配置综合优化

针对高比例分布式光伏接入造成配电线路重过载的问题,提出一种基于时序互补线路互联和柔性负荷整形的配电线路升级与开关配置综合优化方法。考虑柔性负荷功率调节意愿、线路负载率与功率反送约束,提出柔性负荷整形潜力和线路分段整形需求计算方法。以此为基础,建立了考虑柔性负荷整形的含源配电线路升级与开关配置综合优化双层模型。其中,上层模型以配电线路年规划成本最低、线路段净负荷功率和柔性资源均衡为目标进行线路选型和开关配置,下层模型以线路年运行成本最低为目标进行网络重构和负荷曲线整形,通过上、下层模型交互迭代获得互联线路改造优化方案。上、下层模型分别采用非支配排序遗传算法、二阶锥与大M法进行求解。算例仿真结果表明,所提方法可以提高互联线路资产利用率、规划方案的经济性和可靠性,对互联线路进行柔性负荷整形有利于突破线路负载率50%的限制。

10kV配电线路单相树线故障选相研究

针对配电网三相电压不平衡发生树-线故障难以准确判别故障相而引发山火的问题,提出了一种基于电压信号突变值和变化趋势相结合的选相方法。在传统三相电压平衡配电网单相接地故障选相理论基础上,搭建了10 kV单相树-线故障试验平台。首先,在中性点不接地、消弧线圈接地欠补偿和过补偿3种情况下进行了多种树木、不同残余电流单相树-线故障试验研究。然后,根据树-线故障时零序电压和各相电压幅值变化特征,定义了故障时刻及故障相识别码,建立了某一相电压有效值的幅值增加最大则其滞后相为故障相的判据。通过故障时刻前后0.1 s内的各相电压平均有效值改变大小对比及相应判据,将生成对应的识别码进行匹配,实现精确快速判别故障相。最后,通过试验验证该方法能够有效解决单相树-线故障相识别问题且不受电压不平衡影响,具有较好的工程应用价值。

Power Line Communications Networking Method Based on Hybrid Ant Colony and Genetic Algorithm

When solving the routing problem with traditional ant colony algorithm, there is scarce in initialize pheromone and a slow convergence and stagnation for the complex network topology and the time-varying characteristics of channel in power line carrier communication of low voltage distribution grid. The algorithm is easy to fall into premature and local optimization. Proposed an automatic network algorithm based on improved transmission delay and the load factor as the evaluation factors. With the requirements of QoS, a logical topology of power line communication network is established. By the experiment of MATLAB simulation, verify that the improved Dynamic hybrid ant colony genetic algorithm (DH_ACGA) algorithm has improved the communication performance, which solved the QoS routing problems of power communication to some extent.

基于电压补偿的双端直流配电网电压就地协调控制

双端直流配电网是一种双端电源供电的网络结构,可为直流负荷提供更加稳定和可靠的电源电压。然而,负荷的功率波动和不平衡使线路电压跌落增大,可能导致直流负荷的电压不平衡度和电压偏差指标越限,影响直流负荷的正常运行。文中基于电压源型换流器(VSC)外环电压控制,考虑中线电压补偿,提出基于电压补偿等效模型的直流配电网电压偏差及不平衡度联合抑制策略,实现直流配电网电压就地协调控制。首先,建立双端直流配电网潮流模型,将VSC电压下垂控制引入潮流模型,分析不同控制策略下的电压偏差和不平衡度的特性。其次,在此基础上,以电压最低点为分点获得双端电源供电回路压降的简化等效模型,并利用最小二乘法实现等效阻抗参数辨识,构建双端直流配电网的电压补偿等效模型。以参数辨识结果作为电压外环控制输入,提出考虑中线电压补偿的双端直流配电网电压就地协调控制策略。最后,在MATLAB/Simulink中搭建仿真模型,验证了双端直流配电网潮流模型的正确性和控制策略的有效性。

配电网理论线损概率分布函数的计算与分析

线损是衡量配电系统节能减排和运行效率的重要指标。随着“双碳”目标的提出,大量随机性和间歇性的可再生能源接入配电网,给配电网的线损计算带来新的挑战。该文从概率角度出发,在基于高斯混合模型的节点注入电流基础上,结合广义非中心卡方分布推导了任意电流分布下配电网理论线损概率分布的解析表达式,并首次给出配电网理论线损的均值、方差和矩生成函数的准确表达式。考虑到理论线损概率分布难以直接计算,提出一种基于Raphaeli级数的线损概率密度函数和累积概率分布函数的数值计算方法。基于IEEE 33节点系统的算例分析表明,提出的线损计算公式能够准确、高效地计算线损均值和概率分布函数。

注意力感知的边−节点交换图神经网络模型

提出了一种注意力感知的边-节点交换图神经网络(attention aware edge-node exchange graph neural network,AENN)模型,在图结构化数据表示框架下,使用边-节点切换卷积的图神经网络算法进行图编码,用于半监督分类和回归分析。AENN是一种通用的图编码框架,用于将图节点和边嵌入一个统一的潜在特征空间。具体地,基于原始无向图,不断切换边与节点的卷积,并在卷积过程中通过注意力机制分配不同邻居的权重,从而实现特征传播。在3个数据集上的实验研究表明,所提方法较已有方法在半监督分类和回归分析中具有明显的性能提升。

线卷积网络在二维线检测和三维线框重建中的应用

本文提出了一个线卷积网络,可以从图像中检测线段.通过预测每条线段的中心位置、长度和角度,该网络能够以端到端的方式检测线段.此外,根据线中心检测的特点,本文设计了一维旋转卷积模块,并通过分析,证明其满足等变性质.在实现上,本文采用改进的数据增强和非极大值抑制方法,有效的提高了线检测的性能.通过大量消融实验以及对比实验,本文提出的线卷积网络相比于当前最优算法HAWP,在精度上高出3个百分点,达到了当前最优.最后,将该方法应用于单张图三维线框重建任务,从可视化结果上看,明显优于之前的方法.

典型神经网络串联型电弧故障检测及选线方法研究

为解决传统特征分析加机器学习的电弧故障检测方法的准确率和实时性受到特征参数选取的主观性及特征分析过程的影响问题,搭建了三相多负载并联的串联型电弧故障实验系统,对不同支路、不同相发生电弧故障时的干路电流信号时间序列进行分析。将电流信号进行分类、分段、标准化处理并作为检测模型样本;对深度卷积神经网络模型、长短期记忆网络模型、普通神经网络模型进行架构及训练;通过差分处理对网络模型在线分类结果进行优化分析;以准确度和损失函数值、在线测试速度、优化后多分类识别准确率为评价指标,对比分析了3种模型故障检测及选线效果。研究结果表明:基于深度卷积神经网络的串联型电弧故障检测及选线模型对电机类负载故障检测及选线准确率可达96.77%,对变频器类负载故障检测及选线准确率可到98%,准确率高于近几年其他三相回路电弧故障检测模型。

融合稳态和暂态特征量的接地故障选线方法研究

单相接地故障是配电网运行时发生概率最高的故障,但接地时存在电气故障特征弱、外界干扰大的情况,使得配电网存在接地选线困难的问题。对此,提出基于稳态和暂态故障特征量相融合的配电网接地选线方法,分析单相接地时接地故障线路与非故障线路对地电容电流突变量五次谐波分量在幅值和相位上的差异性,利用经验小波变换提取故障零序电流的低频分量综合相关系数和高频分量相对权重系数,并利用模糊神经网络实现融合特征量与故障的非线性映射诊断。建立配电网接地故障仿真模型,通过多重干扰下的选线对比分析,验证了本文方法的有效性和优越性。

考虑超载和纵坡效应的隧道开挖面破坏分析

针对地表超载和纵向坡度影响下的隧道开挖面稳定性问题,提出源于构造破坏机制的初始三角形网格更新策略,应用刚体平动运动单元上限分析(UB-RTME)方法开展非线性规划运算,获得地表临界超载系数σs/c图表与滑移线网破坏模式。结果表明:①以网格更新得到的UB-RTME滑移线网与OptumG2自适应加密网格,均可反映隧道开挖面破坏影响范围及承载关键区域,且UB-RTME以有效间断线及单元速度确定破坏范围,可揭示相对破坏范围S/D2定量数据及滑移线网破坏形态与影响因素的关联规律。②临界超载系数σs/c随纵向坡度角θ增大而显著降低,随埋深比C/D增大而增大,随重度系数γD/c增大而不断减小,随内摩擦角ϕ增大而非线性增大。③对于C/D和ϕ数值较大的情况,网格密度将显著影响UB-RTME运算得到的σs/c数值解精度;网格密度增加后,σs/c计算结果精度显著提高,且体现破坏特征的网状滑移线变得更加光滑。

基于SA-Apriori的道路事故风险评估研究

道路事故风险评估是提高道路安全水平的有效方式。为实现在道路网络特征视角下的道路事故风险评估,融合线段分析与频繁集算法,提出SA-Apriori的特征提取算法,进而构建SA-Apriori道路事故风险评估模型,以重庆市大渡口区作为分析对象,建立以道路网络特征变量为指标的道路事故风险等级划分体系,评估其道路事故风险水平,结合商业点位分布情况对该评估方法的有效性进行验证。并针对大渡口区道路事故风险情况,提出道路事故风险的管理对策与建议。结果表明:该道路事故风险评估方法具有较高的可靠性,可为基于路网特征的道路事故风险评估提供参考。