Shihe Chen

Fault detection and diagnosis for steam turbine based on kernel GDA

A novel fault detection and diagnosis method based on kernel generalized discriminant analysis (kernel GDA, KGDA) is proposed in order to solve the problem of turbine fault detection and diagnosis. Through kernel GDA, the data is mapped from original space to the high-dimensional feature space. Then the statistic distance between normal data and test data is constructed to detect whether a fault is occurring. If a fault has occurred, similar analysis is used to identify type of the faults. The proposed method is scalable to different steam turbine and rotating machineries. Its effectiveness is evaluated by simulation results of vibration signal fault dataset.

A novel soft sensor modelling method based on kernel PLS

A novel soft sensor modeling method based on kernel partial least squares (kernel PLS, KPLS) was proposed. Kernel PLS is a promising regression method for tackling nonlinear problems because it can efficiently compute regression coefficients in high-dimensional feature space by means of nonlinear kernel function. Application results to the real data in a fluid catalytic cracking unit (FCCU) process show that the proposed method can effectively capture nonlinear relationship among variables and have better estimation performance than PLS and other linear approaches.

Application of Multivariable Model Prediction Control to Ultra-supercritical Unit

In this paper, development of ultra supercritical unit control is summarized. Based on analyzing the control difficulties and the input-output relationship of Ultra-supercritical Units, a model predictive control scheme for Ultra-supercritical Unit is proposed. The input variables are fuel flow, turbine valve opening and water flow and output variables are load, steam temperatureteam pressure. The algorithm and implementation method are also given in details.

Research and Application of Multiple Model Predictive Control in Ultra-supercritical Boiler-Turbine System

Based on analyzing the characteristics of Ultra-supercritical unit, this paper introduced a multiple model MCPC (Multivariable Constrained Predictive Control) structure with three inputs and three outputs for coordination control of Ultra-supercritical unit. In the structure, double-layer structure of optimization was used to obtain good steady and dynamic performance, and piecewise linear models at the different operating points of Ultra-supercritical unit were used to deal with nonlinearity. In the real-time simulation, nonlinear model of 1000MW Ultra-supercritical unit in [1] was considered. Finally, the result of real-time simulation was given in the paper.

A Novel Wind Power Prediction Technique Based on Radial Basis Function Neural Network

To ensure the stability of power system and wind farm operation, it is important for power system dispatch in to forecast wind power outputs exactly. The historical data are acquired from an operating wind farm. According to a well-developed Radial Basis Function (RBF) neural network, a wind power predictive model is established, using the historical data such as wind speed, environmental temperature, wind power and so on. Comparing with the actual power output of the wind, the forecasting results show that the proposed method can predict a comparatively accurate and lead to stable results. The proposed power prediction method can be used to make more reasonable dispatching plans.

A Stimulated Simulation System Based on Ovation Virtual DCS

In full accord function with reality power unit, Operator station and engineer station provided by Ovation virtual DCS is accepted by simulation users. The development of interface between virtual DCS and simulation model is key technology for Stimulated Simulation System. The application in Guangdong Gao Lan Gang power plant of a stimulated simulation System Based on Ovation Virtual DCS is introduced here. Key technology of Communication Interface is discussed in this paper.

Development of performance assessment and fault identification strategy based on kernel GDA

Statistical performance monitoring aims at improving process operation by distinguishing abnormal process conditions from common cause variations. However, for some complicated cases in industrial process, because of the nonlinear correlations between process variables, conventional linear statistical methods often have poor ability for monitoring these processes. In this paper, a novel nonlinear on-line performance monitoring and fault identification method based on kernel discriminant analysis (kernel GDA) was proposed. The basic idea of this method is to first map data in the original space into high-dimensional feature space via nonlinear kernel function and then extracts the optimal Fisher feature vector and discriminant vector to perform process monitoring. If faults occurred, it uses the similar degree between the present discriminant vector and the optimal vector of fault in historical dataset to diagnosis. The proposed method can effectively capture nonlinear relationship in process variables. It was evaluated by the application to the CSTR process and its effectiveness was demonstrated.

Study of Condensate Energy Saving Control Technology Based on Pressure Self-Adaptive

A novel energy saving control technology of condensate based on pressure-adaptive is proposed. Through the improved condensate pump and deaerator level control strategy, the fluctuation of condensate pump outlet pressure and deaerator level is small. Through decreasing the outlet pressure set-point to constraints vicinity of operators, current of condensate pump is also decreased. Application results of some 1000MW unit show that the proposed strategy can achieve good energy-saving results. It is worth to popularizing and application.

Nonlinear parameter prediction of fossil power plant based on OSC-KPLS

In order to solve problems of the failure of measured parameters and realize online optimal running in fossil power plant, a novel parameter prediction and estimation method based on orthogonal signal correction (OSC) and kernel partial least squares (KPLS) is proposed. OSC is a data preprocessing method that remove from X information not correlated to Y. Kernel partial least square is a promising regression method for tackling nonlinear problems because it can efficiently compute regression coefficients in high-dimensional feature space by means of nonlinear kernel function. In this paper, the prediction performance of the proposed approach (OSC-KPLS) is compared to those of PLS, OSC-PLS and KPLS using industrial example. OSC-KPLS effectively simplifies both the structure and interpretation of the resulting regression model and shows superior prediction performa- nce compared to PLS, OSC-PLS and KPLS.