Short Term Load Forecasting Based on SBiGRU and CEEMDAN-SBiGRU Combined Model

With the continuous development of global science and technology industry, demand for power is increasing, so short-term load forecasting particularly important. At present, a large number models have been applied to forecasting, but most them ignore error accumulation in iterative training process. To solve this problem, article proposes combined measurement model which combines stacked bidirectional gated recurrent unit (SBiGRU), complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) correction. In first stage, SBiGRU established study time series characteristics under influence temperature holiday types. The generated prediction process reflects series; second sequence decomposed into several intrinsic functions (IMF) components trend by CEEMDAN algorithm. again each component learn predict, predicted values all are reconstructed get results; Finally, sum two-stage results correct error. accuracy SBiGRU-CEEMDAN-SBiGRU combination evaluated two public data. experimental show that has better stability than traditional model.