针对光伏发电功率存在随机波动性的问题,提出基于变分模态分解(variational mode decomposition,VMD)和改进麻雀搜索算法(improved sparrow search algorithm,ISSA)优化长短期记忆(long short term memory,LSTM)神经网络的短期光伏发电功率预测方法.首先,通过VMD算法将多维光伏特征数据分解为若干不同频率的本征模态和残差分量,以降低原始序列的非平稳性;然后,采用ISSA对LSTM神经网络超参数进行全局寻优,建立了不同模态序列分量下的ISSA-LSTM组合模型;最后,使用训练好的组合模型对各分解的子序列模态特征分量进行多维预测,并将各层模态预测序列叠加组合成最终的输出结果.仿真结果表明,构建的VMD-ISSA-LSTM组合模型相较于常规的短期光伏发电功率预测模型,具有更强的鲁棒性和高精度性.
Short Term Photovoltaic Power Generation Prediction Based on VMD-ISSA-LSTM
Aiming at the problem of random fluctuations of the photovoltaic(PV)power generation,this paper proposes a short-term PV power prediction method based on variational mode decomposition(VMD)and improved sparrow search algorithm(ISSA)to optimize the long short term memory neural network(LSTM).Firstly,it decomposes the multidimensional PV feature data into several intrinsic modes and residual components of different frequencies through VMD to reduce the non-stationary nature of the original feature sequence.Then,it uses the ISSA algorithm to optimize the hyperparameters of the LSTM network,and establishes a VMD-ISSA-LSTM combination prediction model for different subsequence components.Finally,the paper uses the trained optimized prediction model to predict the characteristic components of each decomposed subsequence,and combines the prediction set to form the final prediction result.The simulation results show that this high-dimensional characteristic combination prediction model has stronger robustness and high accuracy compared to conventional combination prediction models.
photovoltaic power generationvariational mode decomposition(VMD)improved sparrow search algorithm(ISSA)long short term memory neural network(LSTM)
万方数据
