基于VMD-AHA-LSTM的短期电力负荷预测

Short-term Power Load Forecasting Based on VMD-AHA-LSTM

为了更好地提取负荷序列特征,降低噪声干扰,提高短期电力负荷预测精度,提出了一种由变分模态分解(VMD)、人工蜂鸟算法(AHA)与长短期记忆网络(LSTM)相结合的预测方法。首先通过VMD将负荷数据分解为多个平稳的子序列,然后采用AHA对LSTM超参数进行寻优,将各个序列分别送入到优化后的模型中进行预测,最后对各序列进行重组,得到预测结果。通过我国南方某巿真实数据仿真可知:相较于BP等传统模型,VMD-AHA-LSTM模型的预测效果更好。

In order to extract load sequence features better,reduce noise interference and improve the shortterm power load forecasting accuracy,a forecast method is proposed that combines Variational Mode Decomposition(VMD),Artificial Hummingbird Algorithm(AHA)and Long Short Term Memory(LSTM)network.The prediction results are obtained by first decomposing the load data into several smooth sub-series by VMD,then using AHA to find the optimal LSTM hyperparameters,feeding each series into the optimised model for prediction,and finally reorganising each series.The VMD-AHA-LSTM prediction results are better than the traditional models such as BP,as shown by the simulation of real data from a city in southern China.