Short-term Wind Power Forecasting Based on Convolutional Neural Networks

2nd International Symposium on Resource Exploration and Environmental Science

生态环境科学

Dou, Jinli^1 ; Liu, Chun^1 ; Wang, Bo^1

State Key Laboratory of Operation and Control of Renewable Energy and Storage Systems, Beijing

100192, China^1

关键词: Convolution neural network;    Convolutional neural network;    Convolutional Neural Networks (CNN);    Multilayer feedforward neural networks;    Numerical weather prediction;    Short-term wind power forecasting;    Short-Term wind power predictions;    Spatial-temporal data;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/170/4/042023/pdf
DOI  :  10.1088/1755-1315/170/4/042023

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

Traditional methods of short-term wind power prediction are mostly based on NWP (Numerical Weather Prediction) data on a single station in single-time cross-section and lack in spatiotemporal correlation mining of data. Therefore, a CNN-based prediction method is proposed. Firstly, based on the theoretical analysis of convolution neural network, the input was modelled considering the time correlation, and a variety of convolution neural network structures were designed. Then, a variety of error evaluation criteria were used to evaluate the correlation between single-layer and multilayer feedforward neural networks as well as the convolutional neural networks prediction method. The error and the actual prediction results were analyzed. Prediction error analysis shows that the convolution neural network model can effectively mine the time correlation between data and improve the accuracy of short-term wind power predictions.