基于人工神经网络模型的风速预测

WIND SPEED FORECASTING WITH ARTIFICIAL NEURAL NETWORKS MODEL

利用在达坂城风电场30m轮毂高处的1min实测风速数据,采用人工神经网络模型ANN对未来短时间风速进行预报。通过对风速反复训练与检测来选择一组合适的模型参数,并对模型进行了误差分析。研究结果表明,使用BP神经网络对未来风速进行短时间预测能够达到较好的效果,误差较ARMA模型更精确,但是对于突变信息的处理能力仍然有限。

In order to evaluate the total short-term output power from wind farm in the future, the artificial neural network model was used to forecast a short-term wind speed with actual one-minute interval wind speed data at 30m hub height in Dabancheng wind farm, Xingjiang Region. Several suitable model parameters were selected by recycle training and validation. The errors between the measurement data and forecasting data were analyzed. The results show that by ANN model can satisfy the requirements of short-term wind speed forecasting. The forecasting errors with ANN model are relatively smaller comparing with ARMA model. However, the ANN model also had little ability to deal with the mutation information.