MEDIUM AND LONG TERM PHOTOVOLTAIC POWER PREDICTION BASED ON PERIODIC ATTENTION MECHANISM
In order to solve the problem that the self-attention mechanism is difficult to capture the potential periodic information in the photovoltaic power sequence,a photovoltaic power prediction model Periodformer based on periodic attention mechanism is proposed.Firstly,the potential period of the original sequence is obtained by frequency domain transformation of the period slice and embedding module,and a series of 3D sequence blocks are obtained by slicing and stacking according to the potential period.Secondly,in the oncoding stage,a period strength block is proposed to capture the features during and within the sub-cycle of the sequence.Finally,in the decoding stage,a period cross-attention module is proposed to predict each 3D sequence block respectively,and the prediction results are convoluted and fused to get the prediction results.Experimental analysis shows that the proposed model can well capture the potential characteristics during and within cycles,and has higher prediction performance,and can still maintain a better effect compared with the benchmark model in the medium-and long-term photovoltaic power prediction task.
万方数据
维普
