基于极端梯度提升-长短期记忆神经网络组合模型的短期电力负荷预测

Short-term Power Load Forecasting Method Based on eXtreme Gradient Boosting-Long Short Term Memory Neural Network

为了提高短期电力负荷预测的精度,提出一种基于极端梯度提升和长短期记忆网络的组合预测方法。首先采用Spearman相关系数法对负荷与气象因素进行相关性分析,提取模型输入特征。然后分别建立XGBoost、LSTM预测网络,并采用遗传算法优化网络的参数。最后利用模拟退火算法对各网络的预测结果分配最优权重系数,通过加权组合得到最终的集成预测结果。实验结果表明,XGBoost和LSTM组合模型对短期电力负荷预测的平均绝对百分比误差为0.88%,与XGBoost模型、LSTM模型相比,误差分别降低了2.17%、1.99%,在负荷预测领域更具有优势。

In order to improve the accuracy of short-term power load forecasting,a combined forecasting method based on eXtreme Gradient Boosting(XGBoost)and Long Short Term Memory(LSTM)Neural Network is proposed.Firstly,the Spearman correlation coefficient method is used to analyze the correlation between load and meteorological factors,and the input characteristics of the model are extracted.Then the XGBoost and LSTM prediction networks are established respectively,and the genetic algorithm is used to optimize the parameters of the network.Finally,the simulated annealing algorithm is used to assign the optimal weight coefficients to the prediction results of each network,and the final integrated prediction results are obtained through weighted combination.Experimental results show that the combined model of XGBoost and LSTM used in this paper has a MAPE of 0.88%for short-term power load forecasting.Compared with the XGBoost model and the LSTM model,the error is reduced by 2.17%and 1.99%,respectively,and it has more advantages in the field of load forecasting.