利用相空间重构短时间间隔的吸收太阳能预测方法

Harvested Solar Energy Short Term Prediction Method Utilizing Phase Space Reconstruction

考虑太阳能吸收能量的混沌特性和递归小波网络学习能力强、动态适应性强的优点,提出采用相空间重构与递归小波网络结合的预测模型。分别用该模型与未进行相空间重构、普通递归网络等模型分析实测数据比较,发现采用相空间重构后数据进行模型训练与预测对各种天气、环境的适应性强;而且预测间隔越短,计算量降低的效果越明显;当预测间隔为1.25 min时,计算量降低25%以上。此外,该模型具有相对简单、需要历史数据少、存储空间少的优点,而且预测精度高。

In this paper,considering the chaotic characteristics of solar energy harvesting,and the advantages of strong learning ability and strong dynamic adaptability,a prediction model combining phase space reconstruction and recurrent wavelet network is proposed.The measured data are analyzed by the proposed method and the models of the non-phase space reconstruction and the ordinary recursive network.It is found that because the proposed model is trained and predicted by the phase space reconstruction data,the adaptability to various weather and environment is strong.Moreover,the shorter is the prediction interval,the more obvious is the effect of reducing calculation cost.When the prediction interval is 1.25 min,the computation amount is reduced by more than 25%.In addition,the model is relatively simple with less historical data and less storage space requirement.However,the prediction accuracy is high.