运用广义回归神经网络预测风电场功率

运用广义回归神经网络对风电场出力提前了24h预测。对引入数值气象预报信息与不引人数值气象预报信息两种情况的预测结果进行了比较分析。首先，对前15d的风功率数据进行训练，通过交叉验证，建立模型，预测了未来一天的风电场出力。然后加入历史风速数据，对历史风速和风功率进行训练，利用数值气象预报信息，预测未来1d的风功率。通过算例表明，使用广义回归神经网络模型预测未来1d的风电场出力，预测结果能够跟踪实际风功率，同时加入数值气象预报信息的预测结果较不加入数值气象预报信息的神经网络预测，精度有所提高。

Wind Power Forecasting Using Generalized Regression Neural Network

The generalized regression neural network is used to predict the wind farm output 24 hours in advance. In this paper, comparisons are made between the two prediction cases （with and without the numerical weather prediction added）. Firstly, the wind power data of the previous 15 days are trained and the model is established though cross-validation and the wind power of the next day is predicted. Secondly, the history wind speed data is added and both the history wind speed and wind power are trained and the wind power of the next day is predicted using the numerical weather prediction （NWP） information. The example shows that prediction of the wind power output of the next day is effective, and the prediction result can track the actual wind power. Besides, with addition of the numerical weather prediction information to the neural network prediction, the forecast accuracy is improved.