基于CNN-LSTM-Attention的短期风向预测研究

Short-Term Wind Direction Prediction Based on CNN-LSTM-Attention

关键词:  偏航系统  风向预测  注意力机制  CNN  LSTM网络

Key words:  yaw system  wind direction prediction  attention mechanism  CNN  LSTM network

摘要:

风向预测对提高风能转化率、保障风力发电机偏航系统安全运行及增加风力发电效益具有重要意义。为准确预测风向，提出一种基于CNN-LSTM-Attention的短期风向预测模型。首先，利用卷积神经网络(Convolutional Neural Network, CNN)提取风向数据动态变化特征，然后将所提取的特征向量构成时间序列作为长短期记忆(Long Short-Term Memory，LSTM)网络的输入，最后使用注意力机制(Attention mechanism)分配LSTM隐含层不同权重，增强重要特征的作用，完成风向预测。采用北部湾海域历史风向数据，通过实验与其他神经网络预测模型进行对比，结果显示，CNN-LSTM-Attention模型的相对平均误差(MAPE)值为3.2119%，R²为0.982 6, 优于其他对比模型。所得结果为广西北部湾海域海上风电探索发展提供参考。

Abstract:

Wind direction prediction is of great importance to improve the conversion rate of wind energy, ensure the safe operation of wind turbine yaw system and increase the benefits of wind power generation. In order to accurately predict wind direction, a short-term wind direction prediction model based on CNN-LSTM-Attention is proposed. Firstly, the Convolutional Neural Network (CNN) is used to extract the dynamic change features of wind direction data. Then, the extracted feature vectors are used to form a time series as the input of the Long Short-Term Memory (LSTM) network. Finally, attention mechanism is used to allocate different weights of the LSTM hidden layer to enhance the role of important features and complete the wind direction prediction. The historical wind direction data of Beibu Gulf waters are used to compare with other neural network prediction models through experiments. The results show that the MAPE value of CNN-LSTM-Attention model is 3.2119%, and R² is 0. 982 6, which is better than other comparison models.The results provide a reference for the exploration and development of offshore wind power in Beibu Gulf of Guangxi.