为了解决高比例不确定性风电接入电力系统带来强烈调频需求的问题，提出了基于混合深度学习模型的风电功率预测及其一次调频应用方法。首先，采用孤立森林（Isolated Forest，IF）对历史数据进行异常值处理，提高数据质量，其次，构建卷积神经网络（Convolutional Neural Network，CNN）、双向长短期记忆（Bidirectional Long Short Term Memory，BiLSTM）和注意力机制（Attention Mechanism，AM）的混合深度学习模型对风电功率进行预测。最后，依据功率预测精度配置超级电容器储能，设计储能调频控制原则，弥补风电机组自身预测误差，并协同风电机组参与电力系统一次调频。基于预测结果为4台风电发电机组2个负荷区域仿真系统配置超级电容器储能系统，利用digsilent平台进行了风预测误差和负荷波动下的一次调频仿真。结果表明：所提IFCNNBiLSTMAM模型比BP和LSTM基准模型预测误差(MSE)降低了81.53%和51.44%，具有最优的预测性能；设计的风储一次调频模型与原则可有效应对风电预测误差和负荷波动带来的一次调频问题。

To solve the strong frequency regulation demand caused by the high proportion of uncertain wind power integrated into the power system, a hybrid deep learning modelbased approach for wind power prediction and its primary frequency regulation application was proposed. Firstly,the isolated forest (IF) was applied to process the outliers of historical data to improve the data quality. Secondly, a hybrid deep learning model of convolutional neural network (CNN), bidirectional long short term memory (BiLSTM) and attention mechanism (AM) was constructed to predict wind power. Finally, the supercapacitor energy storage was configured according to the power prediction accuracy, and the energy storage frequency modulation control principle was designed to make up for the prediction error of the wind turbine itself, and cooperate with the wind turbine to participate in the primary frequency modulation of the power system. Based on the prediction results, the super capacitor energy storage system was configured for the 4generator and 2area simulation system, and the primary frequency modulation simulation under wind prediction error and load fluctuation was carried out by using the digsilent platform. The results show that the proposed IFCNNBiLSTMAM model reduces the prediction error (MSE) by 81.53% and 51.44% compared with the benchmark models of BP and LSTM, and has the best prediction performance; the primary frequency regulation model and principle designed in this paper can effectively deal with the primary frequency problem caused by wind power prediction error and load fluctuation.

TK221

国网河北能源技术服务有限公司科技项目（TSS2020-12）