Broad learning system based on Savitzky—Golay filter and variational mode decomposition for short-term load forecasting

Global demand for electricity is increasing dramatically, because of population and electrical commodities growth. Therefore, accurate forecasting of electricity consumption is of great significance for formulating energy plans and ensuring the safe operation of power systems. However, due to the non-stationarity and non-linearity of electricity consumption time series, traditional forecasting methods can not capture the dynamic changes of load curves effectively. To solve this problem, we propose a novel Broad Learning System (BLS) based on Savitzky-Golay (SG) and Variational Mode Decomposition (VMD) for short-term load forecasting. First, we apply SG filter to eliminate the non-stationarity of the data. Then, VMD is used to decompose time series according to time frequency characteristics and extract the non-linear characteristics in the series. Finally, since BLS has a fast training process due to its single-layer network structure, we combine the developed filtering and decomposition algorithm with BLS for electricity forecasting. The study establishes empirical experiments with hourly electricity consumption data from the Los Angeles area. Experimental results show our framework achieves promising results and outperforms the state-of-the-art approaches on extensive public datasets.