Main content area

A hybrid forecasting system based on a dual decomposition strategy and multi-objective optimization for electricity price forecasting

Yang, Wendong, Wang, Jianzhou, Niu, Tong, Du, Pei
Applied energy 2019 v.235 pp. 1205-1225
algorithms, case studies, data collection, electricity costs, electricity generation, models, politics, New South Wales
Electricity price forecasting plays a crucial role in balancing electricity generation and consumption, which is of great political and economic significance for all of society but is still a challenging task. However, in previous studies, most researchers have focused on improving either forecasting accuracy or stability while ignoring the significance of performing these tasks simultaneously. More importantly, few researchers have deeply studied the data preprocessing strategy, only focusing on the application of individual decomposition approaches. Therefore, a novel hybrid forecasting system based on a dual decomposition strategy and multi-objective optimization is developed for electricity price forecasting that includes four modules: a data preprocessing module, optimization module, forecasting module and evaluation module. In this system, an effective multi-objective optimization algorithm is employed to guarantee simultaneous improvements in accuracy and stability. In addition, an improved data preprocessing approach named the dual decomposition strategy is developed, which successfully overcomes the potential drawback of the individual decomposition approach and further improves the effectiveness of the developed forecasting system. Moreover, the evaluation module is incorporated to verify the superiority of the developed forecasting system. Case studies utilizing half-hourly electricity price data collected from New South Wales, Australia are employed as examples. The results prove the superiority of the multi-objective optimization algorithm and the developed dual decomposition strategy and reveal that the developed forecasting system outperforms all of the considered comparison models, which shows its better ability to forecast future electricity prices with better accuracy and stability.