Radial basis probabilistic neural network for differential protection of power transformer

Protection of medium- and large-power transformers has always remained an area of interest of relaying engineers. Conventionally, the protection is done making use of magnitude of various frequency components in differential current. A novel technique to distinguish between magnetising inrush and internal fault condition of a power transformer based on the difference in the current wave shape is developed. The proposed differential algorithm makes use of radial basis probabilistic neural network (RBPNN) instead of the conventional harmonic restraint- based differential relaying technique. A comparison of performance between RBPNN and heteroscedastic-type probabilistic neural network (PNN) is made. The optimal smoothing factor of heteroscedastic-type PNN is obtained by particle swarm optimisation technique. The results demonstrate the capability of RBPNN in terms of accuracy with respect to classification of differential current of the power transformer. For the verification of the developed algorithm, relaying signals for various operating conditions of the transformer, including internal faults and external faults, were obtained through PSCAD/EMTDC. The proposed algorithm has been implemented in MATLAB.     

Application of probabilistic neural network for differential relaying of power transformer

Investigations towards the applicability of probabilistic neural networks (PNNs) as core classifiers to discriminate between magnetising inrush and internal fault of power transformer are made. An algorithm has been developed around the theme of conventional differential protection of transformer. It makes use of the ratio of the voltage-to-frequency and the amplitude of differential current for the detection of the operating condition of the transformer. The PNN has a significant advantage in terms of a much faster learning capability because it is constructed with a single pass of exemplar pattern set and without any iteration for weight adaptation. For the evaluation of the developed algorithm, transformer modelling and simulation of fault are carried out in power system computer-aided designing PSCAD/EMTDC. The operating condition detection algorithm is implemented in MATLAB     

滚动轴承的MSE和PNN故障诊断方法

针对滚动轴承不同运行状态振动信号具有不同复杂性的特点,提出一种新的基于多尺度熵(multiscale entropy, MSE)和概率神经网络(probabilistic neural networks, PNN)的滚动轴承故障诊断方法。该方法首先利用MSE方法对滚动轴承振动信号进行特征提取,并将其作为PNN神经网络的输入,再利用PNN自动识别轴承故障类型及故障程度。实验数据包括不同故障类型和不同故障程度样本,结果表明,相比于小波包分解和PNN结合的诊断方法,提出的方法具有更高的诊断精度,能有效实现滚动轴承故障类型及程度的诊断。     

High impedance fault detection in power distribution networks using time?frequency transform and probabilistic neural network

An intelligent approach for high impedance fault (HIF) detection in power distribution feeders using advanced signal-processing techniques such as time-time and time-frequency transforms combined with neural network is presented. As the detection of HIFs is generally difficult by the conventional over-current relays, both time and frequency information are required to be extracted to detect and classify HIF from no fault (NF). In the proposed approach, S- and TT-transforms are used to extract time-frequency and time-time distributions of the HIF and NF signals, respectively. The features extracted using S- and TT-transforms are used to train and test the probabilistic neural network (PNN) for an accurate classification of HIF from NF. A qualitative comparison is made between the HIF classification results obtained from feed forward neural network and PNN with same features as inputs. As the combined signal-processing techniques and PNN take one cycle for HIF identification from the fault inception, the proposed approach was found to be the most suitable for HIF classification in power distribution networks with wide variations in operating conditions.     

基于多源数据融合的掘进机截割岩壁硬度识别方法

针对煤矿井下掘进机截割岩壁硬度识别难度大的问题,利用其悬臂振动信号、升降油缸和回转油缸压力信号、截割电机电流信号,提出了一种基于多源数据融合的截割岩壁硬度识别方法。该方法首先对各类信号进行小波包分解,单支重构各频带信号并组建时频矩阵,通过奇异值分解得到包含时频信息的若干特征奇异值,以构造特征向量;再利用LDA算法实现数据特征级融合,得到类可分性更好的低维特征。为解决概率神经网络(PNN)平滑参数无法确定和网络结构复杂的问题,提出了基于差分进化算法(DE)和QR分解的PNN优化方法,并通过优化PNN对低维特征进行硬度识别。实验结果表明:所提出的特征量提取和模式识别方法是有效的,与目前常用的其它模式识别算法相比,优化PNN在掘进机三种工况下均有更高的硬度识别准确率。     

Busbar differential protection in conjunction with a current transformer compensating algorithm

The design, evaluation and implementation of a busbar differential protection relay that operates in conjunction with a current transformer (CT) compensating algorithm are described. Prior to saturation, the secondary current of a CT is not compensated. The compensating algorithm detects the start of first saturation on the basis of the third-difference function of the current and estimates the core flux at the first saturation start by inserting the negative value of the third- difference function of the current into the magnetisation curve of a CT. Thereafter, it calculates the core flux and then the corresponding magnetising current in conjunction with the magnetisation curve. The calculated magnetising current is added to the measured secondary current to obtain the correct secondary current. The algorithm can estimate the correct current irrespective of the level of the remanent flux. In the proposed busbar protection scheme, a current differential relay with the single-slope operating characteristic is used on the basis of the compensated current of the saturated CT. Test results indicate that the relay shows satisfactory performance for the various external and internal faults with CT saturation, particularly in the case of a progressive fault from a feeder fault to a busbar fault. The algorithm is implemented in a prototype relay based on a digital signal processor. The relay achieves greater stability on external faults, enhanced sensitivity on internal faults and fast operation on internal faults with CT saturation.     

Algorithm to identify the excitation inductance of power transformer with wye-delta connection

An algorithm to identify the excitation inductances of three-phase power transformer with wye-delta connection is proposed. Existing methods of determining the excitation inductances of three-phase transformers require that all winding currents to be known, making them impractical on some wye-delta transformers where the delta winding currents are not measured. Based on the transformer equivalent circuit, the proposed algorithm eliminates the influence of the delta circulating current, allowing the excitation inductances to be calculated using the line currents of the delta connection side directly. The algorithm?s accuracy has been verified by electromagnetic transients program including direct current (EMTDC) simulations, which show that the proposed algorithm is able to differentiate accurately and sensitively between transformer inrush and fault conditions. This lends itself to applications in the area of transformer protection.     

旋转机械故障诊断的神经网络方法研究

BP神经网络具有较好的非线性映射能力,可以描述频率特征和故障之间的关系,而概率神经网络学习规则简单、训练速度快、避免局部极小和反复训练的问题。根据两种神经网络的原理选择合适的参数建立两个旋转机械故障诊断模型,并利用模型对某旋转机械的故障数据进行处理,结果显示两种网络在故障诊断方面的实用价值。通过对故障数据的结果对比可以看到PNN网络比BP网络具有更好的容错能力。     

基于微分进化神经网络的模拟电路故障诊断

BP神经网络已在模拟电路故障诊断领域得到广泛应用,但BP神经网络存在训练速度慢且容易陷入局部最优的问题.由此,本文提出了一种基于混合变异策略的微分进化改进算法,描述了利用微分进化改进算法进行神经网络权值训练的过程和方法,并将微分进化神经网络用于模拟电路故障诊断,文中还对微分进化神经网络与BP神经网络进行了比较.实验结果表明,微分进化神经网络的训练时间和训练精度均优于BP神经网络,其在模拟电路故障诊断中的准确度比BP神经网络提高了7%.     

A neuro-evolutive technique applied for predicting the liquid crystalline property of some organic compounds

A simple self-adaptive version of the differential evolution algorithm was applied for simultaneous architectural and parametric optimization of feed-forward neural networks, used to classify the crystalline liquid property of a series of organic compounds. The developed optimization methodology was called self-adaptive differential evolution neural network (SADE-NN) and has the following characteristics: the base vector used is chosen as the best individual in the current population, two differential terms participate in the mutation process, the crossover type is binomial, a simple self-adaptive mechanism is employed to determine the near-optimal control parameters of the algorithm, and the integration of the neural network into the differential evolution algorithm is performed using a direct encoding scheme. It was found that a network with one hidden layer is able to make accurate predictions, indicating that the proposed methodology is efficient and, owing to its flexibility, it can be applied to a large range of problems.     

Power transformer protection using S-transform with complex window and pattern recognition approach

A new approach for power transformer protection using S-transform with complex window to distinguish between inrush current and internal fault is presented. The S-transform with complex window is used to extract patterns of transient current samples during inrush and faults. S-transform is a very powerful tool for non-stationary signal analysis giving the information of transient currents both in time and in frequency domains. The spectral energy is calculated for inrush and internal faults and an energy index is found out to distinguish between inrush magnetising current and internal faults. The simulation results and the results obtained using real-time data from a transformer in the laboratory environment indicate the robustness of the proposed technique     

Improved probabilistic neural network algorithm for chemical sensor array pattern recognition

An improved probabilistic neural network (IPNN) algorithm for use in chemical sensor array pattern recognition applications is described. The IPNN is based on a modified probabilistic neural network (PNN) with three innovations designed to reduce the computational and memory requirements, to speed training, and to decrease the false alarm rate. The utility of this new approach is illustrated with the use of four data sets extracted from simulated and laboratory-collected surface acoustic wave sensor array data. A competitive learning strategy, based on a learning vector quantization neural network, is shown to reduce the storage and computation requirements. The IPNN hidden layer requires only a fraction of the storage space of a conventional PNN. A simple distance-based calculation is reported to approximate the optimal kernel width of a PNN. This calculation is found to decrease the training time and requires no user input. A general procedure for selecting the optimal rejection threshold for a PNN-based algorithm using Monte Carlo simulations is also demonstrated. This outlier rejection strategy is implemented for an IPNN classifier and found to reject ambiguous patterns, thereby decreasing the potential for false alarms.     

应用概率神经网络诊断自行火炮发动机的故障

目的　研究概率神经网络模型 ,并应用于故障诊断 .方法　对基于概率统计思想和 Bayes分类规则的概率神经网络模型、网络结构、算法及其特点进行分析 ,利用其进行故障诊断 ,并提出一种优化估计平滑因子的方法 .结果　概率神经网络可很好地诊断自行火炮发动机进行中油路和气路的故障 .结论　概率神经网络在模式识别和故障诊断领域中可取得良好地应用效果     

Active control strategy of structures based on lattice type probabilistic neural network

A new neuro-control scheme for active control of structures having a basic structure similar to the Probabilistic Neural Network (PNN) is proposed. It utilizes the lattice pattern of state vector as the training data of PNN, and thus it is called the Lattice Probabilistic Neural Network (LPNN). Comparing the two schemes, PNN takes much time to obtain a control force in the application because it uses all the training patterns. This may delay the control action inevitably. However, in LPNN, the control force is calculated by using only the adjacent information of LPNN input, making the response of LPNN greatly faster than that of PNN. To investigate the general control capability of the proposed algorithm, one-story and three-story buildings under California, El Centro, and Northridge earthquakes are used as test models. Control results of the LPNN are compared with those of the conventional PNN, and these show that the structural responses have been suppressed effectively by the proposed algorithm.     

ANN–GA approach for predictive modeling and optimization of NOx emission in a tangentially fired boiler

An artificial neural network (ANN) and genetic algorithm (GA) approach to predict NOx emission of a 210 MW capacity pulverized coal-fired boiler and combustion parameter optimization to reduce NOx emission in flue gas, is proposed. The effects of oxygen concentration in flue gas, coal properties, coal flow, boiler load, air distribution scheme, flue gas outlet temperature, and nozzle tilt were studied. The data collected from parametric field experiments was used to build a feed-forward back-propagation neural net. The coal combustion parameters were used as inputs and NOx emission as outputs of the model. The ANN model was developed for full load conditions and its predicted values were verified with the actual values. The algebraic equation containing weights and biases of the trained net was used as fitness function in GA. The genetic search was used to find the optimum level of input operating conditions corresponding to low NOx emission. The results proved that the proposed approach could be used for generating feasible operating conditions.     

Automatic Fault Isolation by Cultural Algorithms With Differential Influence

An evolutionary algorithm with a cultural mechanism of evolution influence for effectiveness and efficiency higher than classical genetic algorithms is proposed for industrial fault isolation. Moreover, the evolution influence is based on a differential concept in order to move toward better zones of the solution space by sensing the fitness gradient. The proposed cultural algorithm is designed in order to be portable and easily configurable in different diagnostic applications. On-field results of an industrial application to motor-vehicle fleet remote monitoring and automatic fault isolation of vehicle wear, operating danger, and fraud in a company that transports dangerous goods are shown.     

基于特征参数迁移的滚动轴承故障诊断

针对变工况下的滚动轴承无法获得大量带标签样本数据以及传统深度学习诊断方法识别率低的问题,提出一种基于迁移学习的卷积神经网络模型滚动轴承故障诊断方法.首先,采用短时傅里叶变换处理滚动轴承振动信号获得源域、目标域样本集;其次,利用源域样本预训练卷积神经网络模型;最后,通过目标域样本微调卷积神经网络模型实现滚动轴承故障诊断....     

皮带秤动态称重用神经网络算法设计

针对皮带秤在使用中难以保持标称计量精度的缺点,提出将过程神经网络引入皮带秤动态称重误差的补偿中。将动态称量过程中皮带秤单位长度上的重量、皮带速度、皮带垂度变化作为模型输入,设计了应用于皮带秤动态称重误差研究的单隐层过程神经网络误差反传播学习算法,利用Matlab软件对算法模型进行训练和测试,模型经过149次学习优化达到网络精度要求,测试组误差为1%,较使用网络前的原误差明显降低,验证了算法的可行性和有效性。     

基于遗传神经网络的异步电动机故障诊断研究

提出一种基于遗传神经网络进行异步电机故障检测的新方法,仅利用一个振动传感器来获取异步电机的特征信息,建立电机动态非线性神经网络检测诊断模型,并利用该模型进行电机的故障检测,为减少网络权值学习搜索空间,解决神经网络权值学习中易于陷入局部最小点的问题,本文采用遗传算法实现模型权值的修正,实际使用证明利用该方法可以方便的实现在线故障诊断,且方法简单,易于实现。