基于残差-密集连接-双向长短期记忆融合网络的光伏出力短期预测

Short-Term Forecast of PV Output Based on Residual-Dense Connection-BiLSTM Fusion Network

光伏出力受随机气象因素的影响而具有极大的不确定性,预测不准将影响配电网系统运行的安全性和稳定性。提出基于残差-密集连接-双向长短期记忆融合网络的光伏出力短期预测模型,按天气类型分类进行模型训练。算例结果表明,所提模型在时间序列预测能力及峰值准确性方面都具有良好的性能。

Photovoltaic output is affected by random meteorological factors and has great uncertainty,whose inaccurate prediction will affect the safety and stability of system operation.This paper proposes a short-term prediction model of PV output based on residual-dense connection-bidirectional long and short-term memory(BiLSTM)fusion network.The model was trained according to weather types.The calculation results show that the proposed model has good performance in time series prediction ability and peak accuracy.