基于OS-ELM的风速修正及短期风电功率预测

Wind speed correction and short-term wind power prediction based on OS-ELM algorithm

随着时间的推移,风电场风电功率预测模型的适用性逐渐降低,导致预测精度下降。为了解决该问题,基于在线序列-极限学习机(OS-ELM)算法提出了风电场短期风电功率预测模型的在线更新策略,建立的OS-ELM模型将风电场的历史数据固化到隐含层输出矩阵中,模型更新时,只需将新产生的数据对当前网络进行更新,大大降低了计算所需的资源。采用极限学习机(ELM)算法对数值天气预报(NWP)的预测风速进行修正,并根据风电功率的置信区间对预测功率进行二次修正。实验结果表明,采用OS-ELM算法更新后的模型适用性增强,预测精度提高;采用基于风电功率置信区间的功率修正模型后,风电功率的预测精度明显提高。

As time goes on, the applicability of the wind power prediction model is gradually reduced, which causes decline of prediction accuracy. To solve this problem, online update strategy of short- term wind power prediction model is proposed in this paper based on online sequential- Extreme Learning Machine（ OS- ELM） algorithm, OS- ELM model established solidify the historical data of wind farm to the implied layer output matrix, and when updating model, simply use new produced data to update current network, which greatly reduces the resources required for the calculation. Extreme Learning Machine（ ELM） is used to correct predicted wind speed of numerical weather prediction（ NWP） and make secondary correction for the predicted power based on wind power and confidence intervals. Experimental results show that the applicability of updated model by OS- ELM is enhanced, and the prediction accuracy is improved. With wind power confidence intervals of the power correction model, the prediction accuracy of wind power has improved significantly.