针对光伏发电在实际电力系统中的波动性和不确定性问题，建立了基于CEEMD-WOA-LSTM的光伏发电功率预测模型。首先，采用皮尔逊相关系数法确定辐照度、湿度、温度和风速为光伏功率的关键影响因素，进而基于高斯混合模型聚类将数据集分为晴天、多云、雨天三种天气类型，以降低训练集与测试集之间的差异并提高预测模型的泛化能力，从而完成数据预处理。其次，采用互补集合经验模态分解对预处理后的数据进行分解并重构，降低其强随机性和复杂性，然后通过利用长短期记忆神经网络对分解所得的各本征模态函数分量进行功率预测，并利用鲸鱼优化算法优化网络参数以提升预测精度，从而叠加各分量的预测结果以确定最终预测值。最后通过实验验证所提方法的有效性。特别地，与现有方法相比，CEEMD-WOA-LSTM在不同天气条件下的预测精度均有所提高，且在复杂天气条件时，展现出更好的稳定性和鲁棒性。

To tackle the fluctuation and uncertainty issues of photovoltaic (PV) power generation in practical power systems, a PV power prediction model based on CEEMD-WOA-LSTM was established. Firstly, the Pearson correlation coefficient method was employed to determine irradiance, humidity, temperature, and wind speed as the key influencing factors of PV power. Subsequently, the dataset was clustered into three weather types: sunny, cloudy, and rainy, using the Gaussian Mixture Model to reduce differences between the training and testing sets and enhance the generalization ability of the prediction model, completing data preprocessing. Secondly, complementary ensemble empirical mode decomposition was applied to decompose and reconstruct the preprocessed data, reducing its strong randomness and complexity. Then, the Long Short-Term Memory neural network was used to predict each intrinsic mode function component obtained from the decomposition. The whale optimization algorithm was employed to optimize network parameters to improve prediction accuracy. Finally, the predicted results of each component were superimposed to determine the final prediction value. The effectiveness of the proposed method was validated through experiments. Particularly, compared to existing methods, CEEMD-WOA-LSTM showed improved prediction accuracy under different weather conditions and demonstrated better stability and robustness in complex weather conditions.