风电、光伏具有较强的随机性与波动性，提高其预测精度对构建新型电力系统具有重要意义。位于同一地区的风光出力具有明显的时空相关性规律，鉴于这些相关特征提出了一种基于相似日聚类与基于鲸鱼优化算法的双向长短时记忆神经网络Copula（whale optimization algorithm-bidirectional long short-term memory neural network,WOA-BiLSTM-Copula）算法的短期风光功率相关性概率区间预测模型。首先，采用K-means聚类算法划分数值天气预报（numerical weather prediction，NWP）数据集，依据Kendall、Spearman相关性系数提取具有相关性的风光联合出力典型相似日场景。其次，针对相关性相似日场景，采用非参数核密度估计法进行Copula建模，确定风光出力最优Copula函数类型。之后，训练鲸鱼算法优化的双向长短期记忆神经网络，并对风电、光伏功率进行点预测。最后，使用蒙特卡洛法对最优Copula函数采样，基于风光点预测值生成相关性概率预测区间。仿真结果表明所提模型可以有效提取风光出力相关性特征，与现有模型相比精度更高，验证了模型的有效性。

Wind and photovoltaic power generation exhibit strong randomness and volatility. Improving the prediction accuracy is of great significance for constructing new power systems. Considering the certain correlation between wind and photovoltaic output in the same region, a short-term wind and photovoltaic power correlation probability interval prediction model based on similar day clustering and whale optimization algorithm-bidirectional long short-term memory neural network-Copula（WOA-BiLSTM-Copula) algorithm is proposed. Firstly, the K-means clustering algorithm is adopted to divide the numerical weather prediction （NWP） dataset, and wind and photovoltaic joint output typical similar day scenarios with correlation are extracted based on Kendall and Spearman correlation coefficients. Secondly, the non-parametric kernel density estimation method is used to establish Copula model for similar day scenes, and the optimal Copula function types of wind and photovoltaic output are determined. Then, the WOA is trained to optimize the BiLSTM and perform point predictions on wind power and photovoltaic power. Finally, the Monte Carlo method is used to sample the optimal Copula function, and the correlation probability prediction interval is generated based on wind and photovoltaic point prediction values. The simulation results show that the proposed model can effectively extract the correlation characteristics of wind and photovoltaic output, and the accuracy is higher than the existing models, which verifies the effectiveness of the model.