Soft sensor modeling method of extreme learning machine ridge regression
Chengli Su · Journal of Hefei University of Technology · 2011
As a new type of feedforward neural network,the extreme learning machine(ELM) is applied to dealing with function regression.A soft sensor modeling method of ELM Ridge Regression(ELMRR) is proposed to solve the multicollinearity problem existing in the output layer nodes of ELM learning algorithm.The new algorithm makes use of the ridge regression instead of the previous linear regression,and uses the particle swarm optimization algorithm to optimize ridge parameters according to root mean square error(RMSE) so as to overcome the main flaw in traditional ridge regression that is difficult to obtain the global optimal ridge parameter.The effectiveness and feasibility of the algorithm are shown by an example.The proposed ELMRR soft sensor modeling method has been applied to predicting the gasoline end point in delayed coking and satisfying results are obtained.The experimental results show that the ELMRR modeling method has better prediction precision and good prospect of application compared to ELM method.