A multi-stage ensemble model for power load forecasting based on decomposition, error factors, and multi-objective optimization algorithm

Abstract

Short-term power load forecasting plays an important role in ensuring the stable operation of power systems and improving economic benefits. However, most of the previous studies ignored the limitations of a single prediction model and the useful information in the error factors, resulting in low prediction accuracy. Therefore, this paper proposes a multi-stage integrated model based on decomposition, error factors, and a multi-objective evolutionary algorithm based on decomposition (MOEA/D). The proposed model consists of three stages: in the first stage, the gated recurrent unit (GRU) is used to predict the components of complete ensemble empirical modal decomposition with adaptive noise, and new data sets are obtained by combining them with the original data sets to fully mine the data characteristics. In the second stage, the MOEA/D based on angle and distance selection strategy and adaptive population generation strategy is used to optimize GRU network parameters with accuracy and diversity as the objective functions, obtaining several load forecasting models and error forecasting models that consider accuracy and diversity. In the third stage, a new nonlinear integration method based on GRU optimized by MOEA/D is used to integrate load forecasting values and error forecasting values, considering error factors to further improve forecasting accuracy. Experimental results on the Australian wholesale electricity market and energy market datasets show that the proposed model outperforms the comparative model in terms of accuracy and generalization and can be widely applied in load forecasting.