考虑气象相似性与数值天气预报修正的海上风功率预测

风功率的准确预测对于海上风电的并网运行意义重大。相比于陆上风电,海上数值天气预报信息(numerical weather prediction,NWP)不准确、爬坡事件频发等因素,使得其预测精度难以满足实际需求。在对海上功率预测的技术难点进行总结的基础上,提出了基于气象相似性和NWP修正的海上短期模型,提前24～72 h进行风功率预测。首先,模型通过对数据样本进行初步气象相似性分类以提高预测精度;然后对未来时刻的风功率波动范围进行预测归类,以降低爬坡事件对预测造成的影响;接着通过对NWP的精确度与模型误差之间的内在联系进行挖掘,实现预测结果的NWP修正。算例部分对国内某海上风电场进行仿真预测,结果表明所提模型对于功率爬坡、NWP信息异常等情况都有较好的预测精度。

Offshore Wind Power Forecasting Considering Meteorological Similarity and NWP Correction

Accurate power prediction is of great significance for integration of offshore wind power. Compared with onshore wind power, offshore wind power has larger fluctuation and more inaccurate NWP information, making offshore wind power difficult to forecast. Firstly, in order to improve prediction accuracy, this paper preliminarily classifies the samples. Then, on basis of classifying the range of historical wind power fluctuation, extreme learning machine(ELM) is used to predict categories of power fluctuation in future time. Finally, this paper uses multi-layer perceptron(MLP) to mine intrinsic relationship between accuracy of NWP information and model error. An NWP correction model is established to predict and correct the errors caused by inaccuracy of NWP information. In example part, the proposed model is used to simulate and predict an offshore wind farm, verifying its effectiveness and practicality.