基于LightGBM-Seq2Seq的异常天气下的风电功率预测

Wind Power Forecasting Based on LightGBM-Seq2Seq Model Under Abnormal Weather

异常天气下新能源出力剧烈变化会严重威胁电网的安全运行,针对气象因素的异常变化导致的风电功率预测准确率低的问题,文章提出了一种基于LightGBM-Seq2Seq的异常天气下的风电功率预测方法。首先,由于目前新能源发电中缺乏有关异常天气的定量判据,文中设计了异常天气判别标准,并采用多尺度滑动窗口进行异常样本提取。其次,针对异常天气下气象波动和功率波动的匹配性差、风电出力情况难以估测的问题,提出基于LightGBM的功率基准值预测模型计算异常天气下的基准功率,同时针对异常气象波动引起的实际功率与基准功率的偏差,提出基于Seq2Seq的功率增量预测模型,通过功率增量对功率基准值进行修正,以实现异常时段的风电功率预测。最后通过实际算例验证了所提方法能够有效提高异常天气下的风电功率预测精度。

The sharp fluctuations of new energy output under abnormal weather will seriously threaten the safe operation of the power grid.Aiming at low accuracy of wind power prediction caused by abnormal changes of meteorological factors,this paper proposes a wind power forecasting method for abnormal weather period based on LightGBM-Seq2Seq.Firstly,due to the lack of quantitative criteria about abnormal weather in the field of renewable generation,this paper designs a judging criteria for abnormal weather and uses multi-scale sliding windows to extract abnormal sample.Secondly,poor matching between meteorological fluctuations and power fluctuations under abnormal weather will make the wind power output forecasting more difficult.Aiming at this problem,this paper proposes a benchmark power prediction model based on LightGBM to calculate the benchmark power under abnormal weather.Meanwhile,this paper takes the deviation of actual power from the benchmark power caused by abnormal meteorological fluctuations into consideration and proposes an incremental power prediction model based on Seq2Seq,in order to correct the benchmark power value and achieve wind power prediction under abnormal weather.Finally,the actual example verified that the proposed method can effectively improve the accuracy of wind power prediction under abnormal weather.