AUTHOR=Jin Shi , Liu Qian , Zhang Wenlu , He Zhihong , He Yuxiong , Zhang Lihong , Liu Yuan , Xu Peidong , Zhang Xiao , He Yuhong 

TITLE=Data-driven methods for situation awareness and operational adjustment of sustainable energy integration into power systems

JOURNAL=Frontiers in Energy Research

VOLUME=Volume 11 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/energy-research/articles/10.3389/fenrg.2023.1253206

DOI=10.3389/fenrg.2023.1253206

ISSN=2296-598X

ABSTRACT=In the context of increasing complexity in power system operations due to the integration of renewable energy sources, accurate short-term wind power forecasting plays a crucial role in power system scheduling. This study proposes a concise short-term wind power generation prediction model that combines a feature selection-based convolutional neural network-bidirectional long shortterm memory network (CNN-BiLSTM) model. By effectively screening multidimensional feature datasets, the model optimizes the selection of highly correlated feature parameters. Additionally, it assigns weights to input data based on feature correlation, resulting in multidimensional feature datasets. The CNN-BiLSTM combination model is then employed to establish a single predictive model for power generation based on multiple features. Training a linear model further refines the wind power generation prediction, enabling accurate awareness of new energy integration into power systems. Moreover, this study introduces an automatic adjustment model for power flow convergence using reinforcement learning, addressing the challenge of power imbalance leading to flow nonconvergence. The utilization of the D3QN (Double Dueling Q Network) reinforcement learning algorithm enables effective control of power flow convergence and adaptive adjustment of operating modes. To validate the proposed method, experiments are conducted using the KDD Cup 2022 wind power prediction dataset. The results demonstrate that the CNN-BiLSTM model effectively utilizes time-series data, surpassing other neural networks in prediction accuracy. Furthermore, simulation results based on the PYPOWER case39 standard case reveal that the model's reward value increases with training rounds and stabilizes at 40. Remarkably, more than 72% of abnormal flow samples achieve rapid convergence within 10 steps, affirming the proposed method's efficacy and computational efficiency.