Ganyun Lv

Application of Adaptive Least Square Support Vector Machines in Nonlinear System Identification

Training problem of least squares support vector machine (LS-SVM) is solved by finding a solution to a set of linear equations. This makes online adaptive implementation of the algorithm feasible. In this paper, an adaptive algorithm for the purpose of nonlinear system identification is proposed. Using this training algorithm, a variant of support vector machine has been developed called adaptive LS-SVM. The adaptive LS-SVM is especially useful on online system identification. Several pertinent numerical simulations have shown the validity of the proposed method

A Reducing Multi-Noise Contrast Enhancement Algorithm for Infrared Image

A kind of infrared image contrast enhancement algorithm based on discrete stationary wavelet transform (DSWT) and nonlinear gain operator is proposed. Having implemented DSWT to an infrared image, de-noising is done by the method proposed in the high frequency sub-bands which are in the better resolution levels and enhancement is implemented by combining de-noising method with incomplete Beta transform (IBT) in the high frequency sub-bands which are the worse resolution levels. According to experimental results, the new algorithm can reduce effectively the correlative noise (1/f noise), additive Gauss white noise (AGWN) and multiplied noise (MN) in the infrared image while it also enhances the contrast of infrared image well. In visual quality, the algorithm is better than the traditional unshaped mask method (USM), histogram equalization method (HIS)

Voltage Sags Detection and Identification Based on Phase-Shift and RBF Neural Network

Voltage sags are probably one of the most important power quality problems because of its impact on malfunctioning electrical equipment in industrial and commercial installations and high frequency. This fact highlights the need for an effective technique of detection, evaluation and classification of the sags problems. This paper proposed a voltage sags detection and identification method based phase-shift and RBF neural network. The voltage sag magnitude, duration and shape were extract out with the proposed phase-shift method, according to instantaneous virtual peak value. The proposed technique has good performance of real-time. Through a data dealing process of detecting outputs by the phase-shift method, a set of features is extracted for identification of voltage sags. Finally, a RBF network was developed for voltage sags classification according to the Cause. The proposed method is simple and reach 92% identification correct ratio even under noise. The results are useful for the diagnosis of the sags cause.

Particle Swarm Optimization for Calibrating Stream Water Quality Model

Parameter estimation of water quality model is generally obtained through the comparison of measured histories and model predictions of water bodies. In this paper, we present a new calibration tool of solving the parameter estimation problem by using particle swarm optimization (PSO) technique, which is a swarm intelligence method developed to perform heuristic direct search in a continuous parameter space without requiring any derivative estimation. To demonstrate the scheme, experiments with synthetic data were conducted on one-dimensional steady state stream water quality model. The experimental results show that the PSO approach still performs very well even if the exact solution is corrupted by noise.

Solving Parameter Identification Problem of Nonlinear Systems Using Differential Evolution Algorithm

A new technique, based on differential evolution algorithm, is proposed for solving the parameter identification problem of nonlinear systems. The technique improves the accuracy of parameter identification. Two kinds of process systems have been used as examples for demonstration. The effectiveness of differential evolution algorithm is compared with that of genetic algorithms in terms of obtained parameter accuracy and objective function value. The simulation results show that the accurate estimation of unknown system parameters and small values of objective function can be achieved by the proposed technique.

Constructive Robust Control Lyapunov Functions for Nonlinear Systems with Disturbance

This paper develops a method by which robust control Lyapunov functions of a class of nonlinear systems with disturbance can be constructed systematically. By using the robust control Lyapunov function a feedback control is established to stabilize the nonlinear system. A simulation shows the effectiveness of the method developed in this paper

Asymptotic controllability of discrete-time systems with disturbances

This paper demonstrates that the existence of a smooth control Lyapunov function for discrete-time systems with disturbances is equivalent to the system is uniformly globally asymptotic controllability with respect to a closed set. It is obtained that this control Lyapunov function may be used to construct a feedback law to stabilize the closed-loop system. In addition, it is shown that for periodic discrete systems, the resulted control Lyapunov functions are also time periodic.

Fault Diagnosis of Power Equipment Based On Dissolved Gas Analysis And LS Fusion Combining Neural Network

In this paper, a new method for power equipment fault diagnosis is presented based on a least square (LS) fusion combining neural network and dissolved gas analysis (DGA). Contents of five characteristic gases obtained by DGA are preprocessed through a special dada dealing process, and 6 features for fault diagnosis are extracted. Then five child back- propagation (BP) artificial neural networks (ANNs) with different structure are applied to diagnosis the fault respectively. The diagnosing results of the child ANNs are fused by the LS weighted fusion algorithm. The fault is identified based on the fused results at last. Compared with single neural network, the LS fusion combining network can identify fault type safely when the fault is deceptive, however, a single neural network may fail in this case. Furthermore, the combining neural network is more reliable than single neural network. The test results of power transformer fault diagnosis proved the conclusions.

Power quality identification based on s-transform and RBF neural network

This paper presents a new power quality (PQ) disturbances identification method based on S-transform time-frequency analysis and RBF network. The proposed technique consists of time-frequency analysis, feature extraction, and pattern classification. Though there are several time-frequency analysis methods existing in the literature, this paper uses S-transform to obtain the time-frequency characteristics of PQ events because of its superior performance under noise. Using the time-frequency characteristics, a set of features is extracted for identification of power quality disturbances. Finally, a RBF network is developed for classification of the power quality disturbances. The proposed method is simple and reached 97.5% identification correct ratio under high signal to noise ratio for those most important disturbances in power system.

S-transform based LS-SVM recognition method for identification of PQ disturbances

This paper presents a new method based on S-transform time-frequency analysis and multi-lay SVMs classifier for identification of power quality (PQ) disturbances. The proposed technique consists of time-frequency analysis, feature extraction and pattern classification. Though there are several time-frequency analysis methods existing in the literature, this paper uses S-transform to obtain the time-frequency characteristics of PQ disturbances because of its superior performance under noise. With the time-frequency characteristics of ST result, a set of features is extracted for identification of PQ disturbances. Then PQ disturbances training samples were contrsucted with the features, and a multi-lay LS-SVMs classifier was trained by the training sample. Finally, the trained multi-lay LS-SVMs classifier was developed for classification of the PQ disturbances. The proposed method has an excellent performance on training speed and correct ratio. The correct ratio of identification could reach 98.3% and the training time of the N-1 classifier was only about 0.2s.