Short-Term Electrical Load Forecasting Based on VMD and GRU-TCN Hybrid Network

With the continuous increase in user-side flexible controllable resources connected into a distribution system, components of electrical load become too diverse and difficult to be accuracy forecasted. A short-term forecast method that integrates variational modal decomposition (VMD), gated recurrent unit (GRU) time convolutional network (TCN) hybrid is proposed this paper. Firstly, original sequence data with noise are decomposed intrinsic IMF different frequencies amplitudes based on VMD method. Secondly, combined forecasting GRU TCN for high low-frequency subsequent signals, respectively. Finally, signals results rebuilt final forecasting. The experiment actual operation (data set 1) simulation 2), which show can reduce error by 36.20% 10.8%, respectively, comparison VMD-GRU. reliability verified through other methods such as LSTM, Prophet XG Boost.