基于GRU网络的风功率短期预测模型

随着大容量风机并网需求的急速增长,现有电网的安全可靠运行面临着巨大挑战。为解决这个难题,提出一种基于深度学习的风功率预测模型。该模型以GRU网络为核心,将风电场的历史功率数据及功率相关的天气数值数据输入到模型中进行预测。鉴于风功率预测模型输入数据维度高、高波动性等特点,为了让GRU网络模型得到更精准的预测结果,在数据处理阶段引入CNN网络,降低输入数据维度。为了克服训练后的预测模型过拟合,引入了dropout技术。最后,通过实验验证该预测模型在预测速度和精度方面均有良好的表现。

Short-term Forecast Model of Wind Power Based on GRU Network

With the rapid growth of the demand for large-capacity wind turbines connected to the grid, the safe and reliable operation of the existing power grid is facing great challenges. In order to solve this problem, this paper proposes a wind power prediction model based on depth learning. The GRU network is the core of the model. The historical power data and power-related weather data of the wind farm are input into the model line for prediction. In view of the characteristics of high dimension and high volatility of input data from the wind power prediction model, the CNN network is introduced in the data processing stage to reduce the dimension of input data in order to make GRU network model get more accurate prediction results. In order to overcome the over-fitting of the trained prediction model, the dropout technology is introduced. Finally, experiments are conducted to verify the prediction speed and accuracy of the prediction model.