An Application Based on Levenberg-Marquardt Bayesian Regulation Algorithm in Power Plate

In order to overcome the insufficiencies of weak extensive ability and the problem of easily trapping into the local extreme value BP Neural Networks (BPNN), a new hybrid algorithm-LMBR, which uses the Levenberg-Marquardt(LM) Bayesian Regularization Algorithm to optimize BPNN, is proposed. The network model, which met the requirements after being examined by the test samples, was applied to predict first-grade overheating desuperheater A side desuperheating water flow. By Simulation Experiments analysis, the LMBR network model was trained until getting the converged network only by 13 steps, while the BP network model didn't achieve the same desired network by 1000 steps, which obviously shows that the LMBR algorithm has faster convergence rate. The average relative forecasting error (ARFE) of the predictive results obtained by the LMBR algorithm is 1.77%, while the ARFE of the predictive results obtained by LM Algorithm and BP is respectively 2.47% and 67.81%. From above comparison, this clearly indicates that the forecast precision of the LMBR algorithm is higher.