基于EMD-LSTM模型的台区负荷短期预测方法

Short-term Load Forecasting Method Based on EMD-LSTM Model

深度学习算法是目前台区负荷预测的主要方法。为解决深度学习方法在逻辑拟合、特征冗余方面的问题,提出一种基于经验模态分解和长短记忆神经网络算法的短期台区负荷预测模型。利用经验模态分解将台区负荷分解为多个本征模函数,使用相关性分析法从特征集合中选择各本征模函数的特征子集,用长短记忆神经网络对这些本征模函数分别进行预测。最后采用某地市台区历史数据对提出的预测方法进行了验证,结果表明,提出的方法较目前主流的深度学习算法具有更高的预测精确度和较低的训练时间。

Deep learning algorithm is currently the main method for load forecasting.To tackle the problems of deep learning methods in logic fitting and feature redundancy,this paper proposes a short-term load forecasting model based on empirical mode decomposition(EMD)and long-short memory algorithm(LSTM).First,the EMD is used to decompose the load data into multiple intrinsic mode functions(IMF).Then the correlation analysis is used to select the feature subset of each IMF.After that,LSTM model is used to forecasting these IMFs separately.Finally,the proposed model is verified by real-world distribution transformer data.The results show that the proposed method has higher prediction accuracy and lower training time than other deep learning algorithms.