基于特征工程的集成学习短期光伏功率预测

短期光伏功率预测对于电网稳定运行具有重要意义。为了解决单一模型预测精度不佳的情况,提出了一种在Stacking集成学习框架下融合Bagging和Boosting算法的短期光伏功率预测模型。首先,引入Copula函数的相关性分析和轻量级梯度提升机的特征贡献度计算来进行特征筛选;然后,选取泛化性能较优的模型作为基学习器,并采用贝叶斯优化算法来对基学习器模型参数进行优化,最后,定义一个超级学习器,采用5折交叉验证,将基学习器与元学习器封装到超级学习器中训练。算例结果表明,在不同季节和不同天气条件下,Stacking模型相较于单一模型有着更高的预测精度。

Short-term PV Power Prediction by Ensemble Learning Based on Feature Engineering

Short-term PV power prediction is of great importance for stable grid operation. In order to solve the poor prediction accuracy of a single model, a short-term PV power prediction model with Bagging and Boosting algorithm under the framework of Stacking ensemble learning was proposed. Firstly, the correlation analysis of Copula function and the calculation of LightGBM feature contribution were introduced to carry out feature screening, and then, the models with better generalization performance were selected as the base learner, and the Bayesian optimization algorithm was used to optimize the model parameters of the base learner. Finally, a super learner was defined, and the base learner and the meta learner were encapsulated into the super learner for training by 5 fold cross validation. The results show that the Stacking model has higher prediction accuracy than the single model under different seasons and different weather conditions.