Font Size: a A A

The Research On Nonlinear Predictive Control Based On Least Squares Support Vector Machines

Posted on:2007-07-28Degree:MasterType:Thesis
Country:ChinaCandidate:Z Q PanFull Text:PDF
GTID:2178360215970021Subject:Control Science and Engineering
Abstract/Summary:PDF Full Text Request
Missile control system can control the flight posture commendably and assure the precision of hitting by controlling some important parameters efficiently and making sure that they can fluctuate in the permitted range. It is difficult to construct the predictive control model because of the high nonlinearity of the control system and the complex relation between input & output; although some modeling methods can relieve this difficulty to some extent, however, there exists some problems in the process of modeling such as over-study, local optimization and high dimensions.Based on the background and on the support of spaceflight project- "XX missile launch decision support system research", this paper makes use of the ground test data during the launching and flight of the missile to identify and build predictive control model so that we can predict and control the system parameters efficiently. The primary research contents are as follows:Having discussed two key problems of the least squares support vector machines (LS-SVM)—tuning of kernel parameter and pruning of support vectors. This thesis introduces new algorithm of tuning of kernel parameter and pruning of support vectors and utilizes the strong nonlinearity feature and studying ability of LS-SVM to identify nonlinear system.The thesis proposes the dynamic matrix predictive control based on LS-SVM. It takes LS-SVM as predictive model and combines with dynamic matrix control algorithm to build parameter's predictive control model. The chosen methods of model's parameters and the implementing of algorithm have been studied in-depth. Lastly, we show the effectiveness of the model by simulation.After identifying system and building its predictive model by LS-SVM, we make use of predictive error control algorithm to implement the model's predictive control and propose error predictive control based on LS-SVM. In the end, we compare this model with that of the error predictive control based on RBF neural network and exhibit that our model has better performance of predictive control and anti-jamming.According to the ideas mentioned above, we apply the two methods to research the posture control of "XX" missile.
Keywords/Search Tags:least squares support vector machines, predictive control, dynamic matrix control, predictive error, RBF neural network
PDF Full Text Request
Related items