基于细胞神经网络的逻辑函数设计

基于细胞神经网络构造动态逻辑门是近年来一个全新的研究方向。由于非线性系统状态演化具有很强的非线性特征和丰富的动态模式,细胞神经网络在构建灵活、可重构的逻辑门电路中具有独特的优势。本文提出基于细胞神经网络的逻辑函数设计,首先设计了两输入线性可分布尔函数“与”门和“或”门的标准非耦合细胞神经网络的模板参数的求解过程,然后给出了使用运放实现的细胞电路设计以及功能之间转换的时序仿真结果。同时以此方式设计了另外12种两输入线性可分布尔函数的模板参数,实现了在电路结构不变的情况下,改变参数即能动态调整布尔逻辑的功能。     

Investigation for Improved Artificial Intelligence Techniques for Thyristor-controlled Series-compensated Transmission Line Fault Classification with Discrete Wavelet Packet Entropy Measures

A new transmission line fault-classification algorithm based on half-cycle post-fault current data is presented for an advanced series-compensated transmission line equipped with a thyristor-controlled series compensator. The proposed scheme was developed with the signal feature enhancement tool of discrete wavelet packet entropy measures. The Chebyshev neural network is presented as network-growing technique for protective classification, the single-layer structure of which is a more powerful classifier that eliminates the need for complicated network design. A comparative implementation study of the multi-layer perceptron and Chebyshev neural network authenticates benefits gained by the Chebyshev neural network. To demonstrate the advantage gained by Chebyshev neural networks compared to support vector machines, a comparative study is presented with a support vector machine based classification technique. The fault datawere obtained by dynamic simulation of a sample system using the real-time power system simulator PSCAD (Manitoba HVDC Research Centre, Winnipeg, Manitoba, Canada). Extensive testing reveals the effectiveness of the Chebyshev neural network for fault classification; a comparative study brings out the superiority of the Chebyshev neural network for neural network design and implementation against the multi-layer perceptron. The Chebyshev neural network proved advantageous against support vector machines as being insensitive to the classification parameter.     

基于神经网络的指令驾驶系统设计

针对某型飞机的高度保持工作方式,依据BP神经网络良好的函数逼近功能,提出了一种基于神经网络的指令驾驶系统控制律设计方法。对控制律采用神经网络建模,并借助于自动驾驶仪系统在各种工作状态下的数据对神经网络进行训练,在不同飞行状态点建立了一个指令驾驶系统控制律的神经网络模型。人机闭环系统仿真结果表明,基于神经网络的指令驾驶系统控制律可以实现预定的高度保持飞行控制任务。     

基于MATLAB的BPNN在变压器故障诊断中的仿真参数研究

首先简要介绍了基于油中溶解气体DGA( Dissolved Gas - in - oil Analysis)的变压器故障诊断机理,然后介绍了反向传播神经网络BPNN( Back - propagation Neural Network)的网络结构、学习算法和训练流程,并结合变压器故障实际特点,分析了输入输出模式的确定、隐含层设计、传递函数和训练函数的选择对于整个网络设计的重要性,通过在MATLAB中神经网络工具箱平台上的仿真比较找出合理的参数,从而建立基于BPNN的变压器故障诊断模型,最后通过对验证样本的仿真诊断结果对比,说明了该模型在实际应用中的有效性.     

基于人工神经网络的电网频率测量方法

基于多层感知器可以任意精度逼近任何线性或非线性 基本原理,提出了一种采用多层前馈神经网络的频率测量方法,并给出相应反向传播学习算法（ＢＰ）神经网络的构造过程和训练方法。仿真结果表明,基于人工神经网络的频率测量方法具有实时性、高精度和鲁棒性,有实用价值。     

Power System Control With an Embedded Neural Network in Hybrid System Modeling

Output limits of the power system stabilizer (PSS) can improve the system damping performance immediately following a large disturbance. Due to nonsmooth nonlinearities arising from the saturation limits, these values cannot be determined by the conventional tuning methods based on linear analysis. Only ad hoc tuning procedures can been used. A feedforward neural network (with a structure of multilayer perceptron neural network) is applied to identify the dynamics of an objective function formed by the states and, thereafter, to compute the gradients required in the nonlinear parameter optimization. Moreover, its derivative information is used to replace that obtained from the trajectory sensitivities based on the hybrid system model with the differential-algebraic-impulsive-switched structure. The optimal output limits of the PSS tuned by the proposed method are evaluated by time-domain simulation in both a single-machine infinite bus system and a multimachine power system.      

面向地区电网的调度员培训仿真系统

传统的调度员培训仿真系统（DTS）面向发输电系统,主要在大区电网和省网中应用。文中设计和开发了面向地区电网的DTS,在功能、算法和应用方法上都具有地调特色。例如：对重要配网进行合环潮流模拟,解列与否采用不同的频率模型,逻辑保护中考虑辐射状小电流接地系统的特殊性,采用故障仿真来代替动态仿真,考虑消弧线圈防谐振操作,利用专家系统考虑操作规程等。该系统已投入实际运行。     

On-line voltage stability assessment using radial basis function network model with reduced input features

In recent years, voltage instability has become a major threat for the operation of many power systems. This paper presents an artificial neural network (ANN)-based approach for on-line voltage security assessment. The proposed approach uses radial basis function (RBF) networks to estimate the voltage stability level of the system under contingency state. Maximum L-index of the load buses in the system is taken as the indicator of voltage stability. Pre-contingency state power flows are taken as the input to the neural network. The key feature of the proposed method is the use of dimensionality reduction techniques to improve the performance of the developed network. Mutual information based technique for feature selection is proposed to enhance overall design of neural network. The effectiveness of the proposed approach is demonstrated through voltage security assessment in IEEE 30-bus system and Indian practical 76 bus system under various operating conditions considering single and double line contingencies and is found to predict voltage stability index more accurate than feedforward neural networks trained by back propagation algorithm and AC load flow. Experimental results show that the proposed method reduces the training time and improves the generalization capability of the network than the multilayer perceptron networks.     

基于混合神经网络的短期电力负荷预测方法

电力负荷预测对电网的经济运行至关重要,为提高短期负荷预测精度并降低混合神经网络模型的训练时间,提出了一种基于多层感知器（MLP）的基础子网、简单循环单元（SRU）与主成分分析（PCA）的短期电力负荷预测模型。首先,考虑影响电力负荷变化的各种因素,建立负荷预测输入特征集;其次,利用PCA对输入网络的部分特征进行变换并降维;最后,将经过PCA处理后得到的全新数据信息作为模型的输入,并结合Adam梯度下降算法进行训练,输出负荷预测的结果。通过仿真实验结果表明,包含SRU的混合模型在全部测试集样本上的MAPE为2.126%,远低于仅有子网的单一模型与包含DNN的混合模型,而与包含LSTM的混合模型相比,训练时间却降低了22.74%,同时PCA的应用也使得模型的收敛速度加快,极大地减小了训练轮数。     

基于人工神经网络的电力系统暂态安全域估计

本文提出了一种利用人工神经网络来描述和拟合电力系统暂态安全性能的方法方法，介绍了一种将集总学习规则和最小二乘法结合起来的快速学习算法。利用某一简化的电力系统。对本方法进行了验证，并将所提出的快速学习方法和传统的B-P学习法进行了比较。     

基于多输出衰减径向基函数神经网络的电网故障诊断

为提高电网故障诊断神经网络模型的构建速度,提出了一种基于多输出衰减径向基函数（Multi-output Decay Radial Basis Function, MDRBF）神经网络的故障诊断方法。DRBF神经网络不需训练即能以任意精度一致逼近任意连续多变量函数。介绍了单输出DRBF（Single-output DRBF, SDRBF）神经网络,分析了其存在的不足,即只能处理单输出变量问题,不能直接应用于电网故障诊断。在此基础上,根据电网元件的故障特点,提出了将SDRBF神经网络演变为多输出DRBF（Mu     

Single‐layer perceptron and dynamic neuron implementing linearly non‐separable Boolean functions

This paper presents a single‐layer perceptron (SLP) scheme with an impulse activation function (IAF) and a dynamic neuron (DN) with a trapezoidal activation function (TAF). Combining with some interesting properties of the offset levels, it is shown that many linearly non‐separable Boolean functions can be realized by using only one SLPwIAF or one DNwTAF. In the present work, a few appropriate IAF and TAF are adopted, and the inverse offset level method is used for the design of the SLPwIAF synaptic weights and the DNwTAF templates. The XOR and NXOR Boolean operations with two inputs and all 152 non‐separable Boolean functions with three inputs can be easily implemented by one SLPwIAF or one DNwTAF. Finally, the entire set of 152 DNwTAF templates associated with 152 non‐separable Boolean functions of three inputs is completely listed. Copyright © 2008 John Wiley & Sons, Ltd.     

Functional adaptive controller for multivariable stochastic systems with dynamic structure of neural network

The article deals with a challenging problem of adaptive control design for multivariable stochastic systems with a functional uncertainty. Model of the system is based on multi‐layered perceptron neural networks where both the unknown parameters and the structure are found in real time without a necessity of any off‐line training process. The unknown parameters are estimated by a global estimation method, the Gaussian sum filter, and the structure of the neural network model is optimized by a proposed pruning method. The control law is based on a bicriterial approach to the suboptimal dual control. Two individual criteria are designed and used to introduce conflicting efforts between the estimation and control; probing and caution. A comparison of the proposed dual control and its alternative with an implementation of the pruning algorithm is shown in a numerical example. Copyright © 2011 John Wiley & Sons, Ltd.     

A novel fault classification-based fault-tolerant control for two degree of freedom helicopter systems

Fault detection and diagnosis (FDD) plays an essential role in identifying and isolating various faults in a system. In general, fault detection is attained by monitoring the extent of matching between the actual operating condition and an analytical model prediction. This process aids in achieving enhanced performance, and for operating the system within the acceptable bounds. In this article, a neural network-based classification method and fuzzy-based control strategy are adapted to perform FDD on a two degree of freedom (2DoF) helicopter system. The operating voltage, pitch, and yaw outputs of the 2DoF helicopter system were considered for developing the algorithm. The signal processing properties of the discrete wavelet transform and pattern recognition properties of a multilayer perceptron neural network are adapted to design the classification algorithm. The developed algorithm improves training and testing efficiency. In order to reinstate the normal operation of the system, the classifier output is integrated with a hybrid fuzzy-proportional integral derivative controller. This control technique enhances the 2DoF helicopter response as the time taken by the pitch and yaw angle to settle trajectory is reduced. The results depicted validate the efficiency of the projected approach.     

BP人工神经网络负荷预测模型的L-M训练算法

根据电力系统短期负荷预测的需要,用C 开发了单隐含层BP人工神经网络程序。程序用Levenberg-Marquardt训练算法实现神经网络训练,大大提高了训练速度。采用24个单输出人工神经网络模型分别预测每天的整点负荷。该预测模型可动态生成,提高了预测模型的自适应性。实际算例结果表明,采用该算法及其程序进行短期负荷预测,可得到令人满意的训练速度及预测精度。     

小波神经网络在电磁场优化问题中的应用

本文利用基于小波变换的单隐层前馈型神经网络，模拟高度复杂的非线性映射，针对电磁机构的优化问题，小波神经网络使用来自有限元分析的训练信息，通过选取一簇适当的小波基函数合成具有一定拓朴结构的小波网络，且网络的训练过程是对一个凸函数的优化过程，从而能得到全局最优解，学习的收敛速度很快。我们将之应用于交流真空接触器直流激磁系统的优化设计中，取得了较好的效果。     

基于深度学习的特高压直流闭锁故障智能调度决策

针对特高压直流闭锁故障的处置策略问题,提出一种基于深度学习的故障特征建模方法及故障后电网调度策略生成方法,所提智能调控决策依据电网直流故障特征和运行环境信息,通过大数据驱动模型训练得到故障后的调度策略。首先根据故障环境信息,利用故障影响相关性提取有效故障信息,构建故障特征模型。然后介绍深度学习类神经网络原理和多层感知器模型,提出利用深度网络提取训练故障前后运行特征,自动生成调控策略的思路。之后利用反向传播算法构建深度学习框架,通过不断计算损失函数和准确率修正训练模型,自动生成有效故障处置策略。最后利用锦苏直流特高压线路相关的电力系统验证了所提方法的有效性。     

Neural Network-Based Faulty Line Identification in Power Distribution Systems

Artificial neural networks (ANNs) have large input-error tolerance ranges and can be used as classifiers. Utilizing this property, a neural network-based detector, which identifies the faulty line directly by taking current and voltage patterns as feature vectors, has been designed. The quality of classification is not dependent on the transmission model, but rather on the net topology, training set, and the choice of learning law. A feed-forward multilayer perceptron, using the Back-Propagation Learning Algorithm, has been used to realize an optimal classifier. The classification quality, by simulating certain faults on the lines, has demonstrated the capability of the proposed approach for distribution power system protection.     

Self tuning control of wind turbine using neural network identifier

The nonlinear characteristics of the wind turbines and electric generators necessitate that grid connected wind energy conversion systems (WECS) use nonlinear controls. The present paper proposes an adaptive self tuning control strategy with neural network Morlet wavelet for WECS control. The proposed strategy is based on single layer feedforward neural networks with hidden nodes of adaptive Morlet wavelet functions controller and an infinite impulse response recurrent structure. The neuro controller is based on a certain model structure to approximately identify the system dynamics of WECS, and control its response. The proposed controller is studied in three situations: without noise, with measurement input noise and with disturbance output noise. Finally, the results of the performance of the new controller were compared with a multilayer perceptron network proving a more precise modeling and control of WECS.     

Power system transient stability margin estimation using neural networks

This paper proposes a methodology for estimating a normalized power system transient stability margin (ΔV_n) using multi-layered perceptron (MLP) neural network with a fast training approach. The nonlinear mapping relation between the ΔV_n and operating conditions of the power system is established using the MLP neural network. The potential energy boundary surface (PEBS) method along with a time-domain simulation technique is used to obtain the training set of the neural network. Results on the New England 10-machine 39-bus system demonstrate that the proposed method provides a fast and accurate tool to evaluate online power system transient stability with acceptable accuracy. In addition, based on the examination of generators rotor angles after faults, a method is presented to select the power system operating conditions that most effect the ΔV_n for each fault.