基于VMD和LSTM-CNN的短期负荷预测模型研究

Research on Short-term Load Forecasting Model Based on VMD and LSTM-CNN

针对传统长短期记忆神经网络结构上缺乏纠错设计的问题,提出了一种基于变分模态分解(VMD)和长短期记忆卷积神经网络(LSTM-CNN)的短期负荷预测模型。该模型通过在长短期记忆神经网络中引入一维卷积模块,使模型具备纠正隐藏状态向量错误的能力。同时利用VMD对温度、湿度等气象数据进行分解,提取与建筑负荷相关性最高的模态分量作为输入特征向量,增强输入特征向量与负荷数据的关联度。实验结果表明,该模型预测结果的均方根误差、平均绝对百分误差相比于传统LSTM模型分别降低了102.16和0.93%,验证了该模型在短期负荷预测中的有效性。

Aiming at the problems of lacking error correction modules in the traditional long and short-term memory networks,a short-term load forecasting model based on VMD and LSTM-CNN is proposed.The model introduces the one-dimensional convolution module into the LSTM to correct errors appearing in hidden state vectors.Meanwhile,the model takes advantage of the VMD to decompose the temperature,humidity and other data,and the modal components with the highest correlation to the building load as input feature vectors to enhance the correlation between the input feature vectors and the load data.The experimental results show that the root mean square error and the mean absolute percentage error of the model's prediction results are reduced by 102.16%and 0.93%respectively compared to the traditional LSTM model,which verifies the effectiveness of the model in the short-term load prediction.