基于VMD-ISSA-LSTM-EC的抽水蓄能负荷预测研究

Research on load forecasting of pumped storage unit based on VMD-ISSA-LSTM-EC

针对短期抽水蓄能负荷预测方法误差大的问题,为提高抽水蓄能负荷预测精度,本文提出了一种基于VMD-ISSA-LSTM-EC的短期抽水蓄能负荷预测模型。首先,根据负荷序列强非线性、非平稳性的时序性特点,采用变分模态分解(VMD)对负荷数据进行分解,得到模态分量;其次,针对长短期记忆网络(LSTM)需手动选择参数的不准确性,采用改进的麻雀搜索算法(ISSA)对LSTM超参数进行优化,结合Tent映射策略以及黄金正弦策略对麻雀搜索算法(SSA)进行改进,提升其算法的寻优能力,再将VMD分解得到的各分量输入ISSA-LSTM网络进行预测,所得各预测值叠加得到初始预测功率;最后,采用ISSA-LSTM网络对误差序列进行误差修正功率预测,将初始预测功率和误差修正功率求和得到最终结果。实验结果表明,本文提出的模型预测精度最高,模型的RMSE和MAPE分别为60.53MW和0.96%,验证了模型在抽水蓄能负荷预测上的有效性。

Aiming at the problem of large errors in short-term pumped storage load forecasting methods,in order to improve the accuracy of pumped storage load forecasting,this paper proposes a short-term pumped storage load forecasting model based on VMD-ISSA-LSTM-EC.Firstly,for the strong nonlinearity and nonsmoothness of the load sequence,the modal components are obtained by decomposing the load data using the variational modal decomposition(VMD).Secondly,to avoid the inaccuracy of the long-and short-term memory network(LSTM),which requires to select the parameters manually,the improved sparrow search algorithm(ISSA)is adopted to optimize the hyper-parameters of the LSTM,which is combined with the Tent mapping strategy and the golden-sine strategy to improve the sparrow search algorithm(SSA),to improve the optimization ability of the SSA algorithm.The components obtained from the VMD decomposition are input into the ISSA-LSTM network for prediction,and the predicted values are superimposed to obtain the initial prediction power.Finally,the ISSA-LSTM network is used to predict the error correction power for the error sequence,and the final result is obtained by summing up the initial prediction power and the error correction power.The experimental results show that the model proposed in this paper has the highest prediction accuracy,and the RMSE and MAPE of this paper’s model are 60.53 MW and 0.96%,respectively,which verifies the validity of this paper’s model in pumped storage load prediction.