神经网络

The thesis contains five chapters. The first chapter summarizes the development and research of switched reluctance drive, discusses the main research direction, and explains the main work in this paper. In chapter 2, hardware system which includes the design of power converter and various control circuit is discussed. TMS320LF2407 DSP is used to design the hardware circuits of SRM control system, and design details including the current detection, position sensing, fault protection, speed detecting, keyboard and display etc. are provided. Because of the full use of the abundant peripheral resources of DSP, it comes to the aim simplifying the circuit structure and heightening the reliability. In chapter 3, the application of neural network on SRM control system is introduced, a new the application of neural network and the method of sampling is proposed and a simulation system based on Matlab/Simulink is established. Chapter 4 discusses the routine designing issue. Because the modularized programming method is adopted, and multi-interrupt processing technique is used, operation efficiency of the control software is highly raised. At last, the foregoing SRM control system is tested. Speed adjustment is realized, and other targets on the research and design of SRM control system are reached, which establishes a good foundation for further research.

本文共分五章，第一章概述了开关磁阻电机调速系统(Switched Reluctance Drive，简称SRD)及其发展和研究现状，论述了其主要研究方向，并说明了本文的主要工作；接下来的一章主要讨论了SRM的硬件系统，主要包括功率变换器的设计和各种控制电路的设计，本文以TMS320LF2407为核心设计了开关磁阻电机控制系统硬件电路，给出了包括电流检测、位置检测、故障保护、测速电路及键盘和显示电路等部分电路的详细设计，充分利用了DSP的丰富外设资源，达到了简化电路结构、提高运行可靠性的目的；第三章介绍了人工神经网络在开关磁阻电机控制系统中的应用，提出了一种新的应用方式(利用GPFN神经网络调节PI参数)以及如何采取样本，搭建了基于Matlab/Simulink的仿真系统；在第四章中，讨论了开关磁阻电机控制软件的设计，采用模块化编程方法，采用基于多中断的控制程序，提高了控制软件的效率；最后，对所设计的开关磁阻电机控制系统进行了实验，实现了电机调速，达到了开关磁阻电机控制系统研究和设计的预期目标，为更近一步研究打下了基础。

It is necessary to train the already known annual reduction quantity of arable land resources in Shaanxi Province through BP nerve network ＆ to establish a forecasting model for this reduction according to time order.

采用MATLAB软件建立用于预测的BP神经网络，将已知陕西省的耕地资源年减少量通过BP神经网络进行训练，建立陕西年耕地面积减少的时间序列的预测模型。

It has been proved that if example set is of lattice point distribution the example set can be memorized by a fuzzy neural network, otherwise the fuzzy neural network will lost its memory.

通过定理，证明了在这种情况下，样本组可被模糊神经网络所记忆；证明了若样本组前n-1个样本和整个样本组的模糊矩阵的秩相等，则权向量不能调整，否则会使模糊神经网络"丧失记忆"。

And point out that laxation type and densification type of wavelet-neural network will play an important role in non-destructive testing of wood material.

并提出小波、人工神经网络以及松散型和紧致型小波神经网络将在木质材料的无损检测中起重要的作用。

And the reasonable proposal for the construction the region electrical power system was put forwardThe power load forecast is the basis of the power system planning.With the actual planning of regional development,uses a scientific and effective load forecasting methods forecast plans the local load electric quantity,is an important work to the regional power system planning.This dissertation uses the BP algorithm, proposed a double hidden layer structure artificial neural network to the forecast ultra-short-term power system load.After the simulation with the actual data,the method which this dissertation proposed compared to the traditional linear extrapolation Prediction method average absolute percent error of 2.5 percentage points,ten minute load forecasts achieve the average absolute percentage error are smaller than 2.5%accuracy.

本文利用人工神经网络BP算法，提出了一种双隐层结构的电力系统超短期负荷预测方法，经过对实际数据的仿真实验，本文提出的神经网络超短期负荷预测方法比传统的线性外推法预测平均绝对百分误差降低2.5个百分点，十分钟负荷预测达到平均绝对百分误差小于2.5%的精度。

In this paper continuous differentiable conditions of output response functions of bidirectional associate memory neural networks are reduced to Lipschitz condition.

因此，本文将双向联想记忆神经网络的输出响应函数连续可微的假设削弱为满足Lipschitz条件，通过引入Lyapunov函数，利用不等式的方法，证明了双向联想记忆神经网络全局指数稳定性的一个定理。

W ithout assuming the boundedness, monotonicity, differentiability and Lipschitz continuity of the active functions, an algebraic criterion to ensure existence and globally exponential stability of periodic solutions was obtained.

在 放松该类神经网络激活函数的有界性、单调递增性、可微性及Lipschitz连续等条件下，得到了该类神经网络周期解的存在性与全局指数稳定的代数判据。

In order to enhance the training speed and the precision of neural networks, the daily load curve is divided and proper input and output variables are chose in each model.

为了提高神经网络的训练速度和精度，对每种不同的神经网络模型，将日负荷模型分段，选择出与预测日负荷相关的输入变量，比较得出最佳的输出变量。

Shape and color characteristics were used to form the separated pattern vector and RBF network was used as classifier for each pattern. Each classifier was linearly combined to give the final identification result. Shape pattern consists of rectangularity, elongation, roundness, sphericity, lobation and seven movement invariants. R, G, B, L histogram and R.G chromaticity histogram were treated as separated color characteristic vector.

昆虫图像的形态特征值由矩形度、延长度、球状型、叶状型、似圆度和 7个Hu不变矩等12个特征值组成，颜色特征值由红、绿、蓝、灰度真方图及基于红、绿的二维色度直方图特征值分别组成，然后建立径向基神经网络分类器，每一特征向量由独立的径向基神经网络做为分类器，最终识别由每个分类器识别结果的线性组合而成。

Transformer; magnetizing inrush current; fault current; BP neural network; RBF network

变压器；励磁涌流；故障电流； BP神经网络； RBF神经网络

We have no common name for a mime of Sophron or Xenarchus and a Socratic Conversation; and we should still be without one even if the imitation in the two instances were in trimeters or elegiacs or some other kind of verse--though it is the way with people to tack on 'poet' to the name of a metre, and talk of elegiac-poets and epic-poets, thinking that they call them poets not by reason of the imitative nature of their work, but indiscriminately by reason of the metre they write in.

索夫农 、森那库斯和苏格拉底式的对话采用的模仿没有一个公共的名称；三音步诗、挽歌体或其他类型的诗的模仿也没有——人们把&诗人&这一名词和格律名称结合到一起，称之为挽歌体诗人或者史诗诗人，他们被称为诗人，似乎只是因为遵守格律写作，而非他们作品的模仿本质。

The relationship between communicative competence and grammar teaching should be that of the ends and the means.

交际能力和语法的关系应该是目标与途径的关系。