Based on the powerful nonlinear mapping ability of kernel learning, and in combination with the partial least square (PLS) algorithm for linear regression, a wavelet kernel partial least square (WKPLS) regression method is proposed. By the method, the input-output data are firstly mapped to a nonlinear higher dimensional feature space, a linear PLS regression model is then constructed by the classic kernel transformation trick used in support vector machines. The PLS approach utilizes the covariance between input and output variables to extract latent features, and the wavelet kernel which is an admissible support vector kernel function is characterized by its local analysis and approximate orthogonality. Hence, the proposed WKPLS method combining PLS approach with wavelet kernel function shows excellent learning performance for modeling nonlinear dynamic systems. The WKPLS is then applied to modelling of several chaotic dynamical systems and compared with the kernel partial least squares(KPLS) method using Gaussian kernel function. Simulation results confirm that the WKPLS identifier is fast and can accurately approximate unknown chaotic dynamical system, and its approximation accuracy is higher than the KPLS under the same conditions.

Abstruct: In this paper, we propose a new method for diirect identifying the stator flux linkage and electromagnetic torque of induction motors, which uses artificial neural networks. Because the multilayer feedforward network has the good capability of approximating any linear or nonlinear function, which is trained using the back propagation (BP) algorithm, it can observe the stator flux linkage and electro-magnetic torque of induction motors accurately. Based on this principle, we built a new direct torque control (DTC) system. With simulation experiment, the results show that this method can observe accurately the stator flux linkage and electro-magnetic torque of induction' motors, and the system has good dynamic and static performance.Keywords-About four, direct torque control, artificial neural network, multilayer feedforward network, error back propagation algorithm

Aimed at the problem of DC micro grid rectifier control delay and the DC bus voltage stability, the method of two-step model predictive direct power control (TMPDPC) combined with neutral point potential control is proposed. Model predictive direct power control (MPDPC) is designed for rectifier. A cost function is then used to evaluate the active and reactive power ripples, from which the vector that generates the lowest power ripple will be applied during the next sampling interval. Two-step predictive control is designed to compensate for the delay of one-step predictive control. On the basis of this, the two-capacitor voltage unbalance problem in DC side is considered, and the neutral point potential control is added. In the Matlab/Simulink simulation, compared with control effect of direct power control (DPC) and one-step model predictive direct power control (OMPDPC), the control strategy of neutral point potential added to TMPDPC can make the system stability and control accuracy better. The validity of this scheme was validated by physical simulation at last.

The model and approach of hierarchical fault diagnosis for substation based on rough set are presented, and the information of circuit breakers and relays are used in this method to mine fault diagnosis rules. On the one hand it can set up rational and succinct diagnosis model for large complicated power system, on the other hand it can dispose uncertainty information of substation with rough set, and get correct diagnosis result under ii:complete information. At last an example is given, the result indicates that hierarchical fault diagnosis based on rough set for substation is an effective method. Index Terms -hierarchical fault diagnosis, rough set, substation.