A Novel Multi-Step Prediction for Dam Deformation Prediction Based on EEMD
1 College of Geomatics and Geoinformation, Guilin University of Technology, 12 Jiangan Road, Guilin 541004,China
2 Guangxi Key Laboratory of Spatial Information and Geomatics, 12 Jiangan Road, Guilin 541004,China
Abstract Multi-step prediction, a new algorithm based on ensemble empirical mode decomposition (EEMD) for dam deformation prediction, is presented. Firstly, starting from the time-frequency analysis, with the use of a collection of empirical mode decomposition, deformation time series are broken down into characteristic components of different frequencies. Secondly, run-determination act is used to reconstruct volatility components similar to high, medium and low ones. This is effective in centralizing model predictive feature information and reducing the difficulty. Finally, multi-step prediction model of the three is established separately and the predictive values are overlayed as the final prediction result. The calculation result is analyzed and compared with the AR model, BP neural network and SVM. At the same time, different prediction verification instructions are established to prove our algorithm. The results show that the prediction accuracy of EEMD is higher, it can guarantee better prediction of dam deformation in volatile periods| it is feasible to apply to dam deformation prediction.