Two-phase extreme learning machines integrated with the complete ensemble empirical mode decomposition with adaptive noise algorithm for multi-scale runoff prediction problems

Two-phase extreme learning machines integrated with the complete ensemble empirical mode decomposition with adaptive noise algorithm for multi-scale runoff prediction problems

Expert systems in multi-scale runoff prediction are useful decision-making tools but the stochastic nature of hydrologic variables can pose challenges in attaining a reliable predictive model. This paper advocates a data-driven approach to design two-phase hybrid model (CVEE-ELM). The model utilizes complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) coupled with variational mode decomposition (VMD) for frequency resolution of the input data and extreme learning machines (ELM) as the objective model. In the first stage of the presented model design, frequencies in predictor-target series are uncovered, utilizing CEEMDAN where inputs are decomposed into Intrinsic Mode Functions (IMFs) and Residual (Res) series. The second stage entails a VMD approach, to decompose the yet-unresolved high frequencies (IMF1) into variational modes, discerning and establishing data attributes to be incorporated in ELM to simulate IMF, Res and VM series, aggregated as an integrative for runoff prediction. In evaluative phase, hybrid CVEE-ELM is cross-validated with a single-phase hybrid ELM and an autoregressive integrated moving average (ARIMA) model to benchmark its accuracy for predicting 1-, 3- and 6-month ahead runoff in Yingluoxia watershed, Northwestern China. Two-phase hybrid model exhibits superior accuracy at all lead times to accord with high correlations between observed and forecasted runoff, a relatively large Nash-Sutcliffe and Legate-McCabe Index. Taylor diagram depict the two-phase hybrid CVEE-ELM forecasts located close to a reference (perfect) model, with lower root-mean square centered difference, and a correspondingly large correlation for all forecast horizons, although the accuracy for shorter lead times (1-month) are better than the 3- and 6-month horizon. The study shows that the two-phase hybrid model is a preferred data-driven tool for decision-systems, particularly for hydrologic problems with stochastic data features, and thus, require reliable forecasts at multi-step horizons.

Files associated with this item cannot be displayed due to copyright restrictions.