基于Attention-GRU的短期光伏发电功率预测

针对传统长短时记忆神经网络（LSTM）参数量较多以及在处理长时间序列时容易忽略重要时序信息的不足,提出一种结合注意力机制（attention）与门控循环单元（GRU）的Attention-GRU短期光伏发电功率预测模型。首先,基于改进相似日理论建立新的数据集;然后,利用门控循环单元提取光伏发电功率的时序特征,引入注意力机制加强对时序输入中重要信息的关注;最终构建针对不同天气类型的预测模型。仿真结果表明,提出的模型与对比模型相比,预测精度更高。

SHORT-TERM PHOTOVOLTAIC POWER FORECASTING BASED ON ATTENTION-GRU MODEL

In view of the large number of parameters of traditional long-short term memory neural network （LSTM） and lack of important timing information when processing long-term sequences, an Attention-GRU short-term photovoltaic power forecasting model combining attention mechanism and gated recurrent unit （GRU） is proposed. Firstly, a forecasting model for different weather types is established. Then, GRU is used to extract the time series characteristics of photovoltaic power generation and the attention mechanism is introduced to strengthen the attention to important information in the time series input. Finally, a forecasting model for different weather types is established. Simulation results show that the proposed Attention-GRU model has higher forecasting accuracy than the comparison models.