Data reconciliation of nonlinear dynamic process based on LSSVM

A new data reconciliation algorithm based on least squares support vector machines (LSSVM) for nonlinear dynamic process is proposed in this work. Firstly, response of processes and training data is obtained by computation tools or simulation software. Secondly, the local models of processes are identified by LSSVM. Finally, process data reconciliation is transformed to nonlinear program problem with constraint equation. Compare to the prevailing data reconciliation method for nonlinear dynamic processes- NDDR. The proposed algorithm is simple, easy to realize. In addition, the mechanism model does not be involved, which is substituted by the LSSVM model identified by sampling data online. Simulation results demonstrate the superiority of the proposed algorithm.