基于数据优化的改进深度学习方法光伏出力超短期预测模型

Improved deep learning method based on data optimization for ultra-short-term prediction model of photovoltaic output

为解决光伏出力超短期预测模型精度不足和运算速度慢的问题,本文提出了一种基于数据优化的改进深度学习方法光伏出力超短期预测模型。首先,为提升模型的计算效率,通过数据预处理和动态指数平滑法对样本数据进行优化;随后,应用卷积神经网络算法(CNN)构建的多阶卷积通道合并运算挖掘不同光伏电场间的时空耦合关系,得到反映多光伏电场光伏出力的融合特征值,将得到的融合特征值作为输入,利用改进深度学习算法进行分析,输出不同天气情况下的光伏超短期预测结果,以提高模型的预测精度;最后,基于实测光伏出力数据进行超短期预测,验证所提模型的有效性和准确性。算例分析表明,所提预测模型相比传统的超短期模型具有计算速度快和预测准确度高的优点。

In order to improve the ultra-short-term prediction accuracy and training speed of photovoltaic output,an improved deep learning method based on data optimization is proposed for ultra short-term forecasting of photovoltaic output.Firstly,in order to improve the computational efficiency of the model,the sample data are optimized by data preprocessing and dynamic exponential smoothing method.Then,the multi-order convolution channel combination operation constructed by convolution neural network algorithm(CNN)is applied to mine the spatio-temporal coupling relationship between different photovoltaic electric fields,and the fusion eigenvalues reflecting the photovoltaic output of multiple photovoltaic electric fields are obtained.The obtained fusion eigenvalues are used as input,analyzed by improved depth learning algorithm,and the photovoltaic ultra short-term prediction results under different weather conditions are output,to improve the prediction accuracy of the model.Finally,the ultra short-term prediction is carried out based on the measured photovoltaic output data to verify the effectiveness and accuracy of the proposed model.An example shows that the proposed prediction model has the advantages of fast calculation speed and high prediction accuracy compared with the traditional ultra-short-term model.