基于LSTM-CNN-CGAN的风储微电网日前风险优化调度方法OA

To accurately characterize the power imbalance risk of a wind-storage microgrid and reduce its dependence on the upstream grid,a day-ahead risk optimal scheduling method for the wind-storage microgrid is proposed,which is based on long short-term memory-convolutional neural network-conditional generative adversarial network(LSTM-CNN-CGAN).First,the short-term forecasted wind power is taken as input,the LSTM neural network is used to enhance the CGAN model's capability to capture temporal features,and CNN is applied to improve the CGAN model's accuracy in identifying local features.This generates a day-ahead wind power uncertainty sample set that satisfies both the condition-al correlation and the temporal autocorrelation.Second,chance constraints are employed to quantify the microgrid's im-balance risk.With the consideration of the self-balancing rate constraint,the electricity and reserve interaction strate-gies between the microgrid and the upstream grid are optimized.By fully leveraging the regulation potential of energy storage in both the electricity regulation and the reserve response,the system's reliance on the upstream grid is further reduced.Finally,a practical microgrid in north China is used as a case study for simulation validation.Results demon-strate that the proposed method can accurately capture the probability distribution characteristics of short-term wind power forecast errors,effectively enhance the self-balancing capability of the wind-storage microgrid and maintain the power imbalance risk within a preset confidence level.