Combined Wind and Photovoltaic Power Forecasting Based on Attention-BiLSTM Multitask Learning for Renewable Energy System

• Main Authors: He, Yingjing, Wang, Cenfeng, Zhu, Keping, Chen, Yuejiang
• Format: Conference Proceeding
• Language: English
• Subjects: Adaptation models; attention mechanism; combined power forecasting; Correlation; Couplings; Hybrid power systems; multitask learning; Photovoltaic systems; Predictive models; Renewable energy sources; spatio and temporal correlation
• ISSN: 2768-0525
• DOI: 10.1109/ICPRE59655.2023.10353765

In the renewable energy system with renewable energy as the main body, the random fluctuation of a high proportion of renewable energy intensifies the instability of the power system after grid connection. To solve this problem, this paper proposes a combined forecasting model of wind and photovoltaic power based on Attention-BiLSTM hybrid model and Progressive Layered Extraction (PLE) multitask learning method. Meteorological and historical power data are used as input, and measured power data are used as the forecasting target. The model can reflect the spatio-temporal correlation between wind and photovoltaic power, and extract the nonlinear coupling information between them to improve the forecasting accuracy. The proposed method is validated using real power generation data. The analysis results show that the proposed model can not only improve the power forecasting accuracy, but also obtain the forecasting results of wind and photovoltaic power at the same time, which reduces the workload and has certain engineering application value.