摘要
为丰富短期电力负荷预测方法并提高预测准确度,提出一种基于变分模态分解(variational mode decomposition,VMD)和蝙蝠算法(bat algorithm,BA)优化随机森林(random forest,RF)的短期电力负荷预测方法。该方法首先利用VMD对实际负荷数据进行分解处理,得到多组具有不同特征的模态函数分量;然后基于BA算法优化随机森林回归模型中的决策树和分裂特征数,进而分别对每组模态函数分量进行预测;最后对所预测的模态分量函数进行重构获取最终预测结果。与此同时,将VMD-BA-RF与Spark平台的优秀计算能力相结合,节省负荷预测过程所需要的时间。算例结果表明:VMD-BA-RF方法可有效进行短期电力负荷的预测,其预测平均绝对误差为31.99 MW,均方根误差为53.28 MW,平均绝对百分比误差为0.60%,而传统RF方法、VMD-RF方法和BA-RF方法的3种指标均大于VMD-BA-RF方法,因此所提方法具有较高预测准确度。
To enrich short-term load forecasting methods and improve forecasting accuracy,a short-term load forecasting method based on variational mode decomposition(VMD)and bat algorithm(BA)to optimize random forest(RF)is proposed.Firstly,VMD is used to decompose the actual load data to obtain multiple groups of modal function components with different characteristics;secondly,BA is used to optimize the decision trees and split features in the RF regression model,and then each group of modal function components is predicted respectively;finally,the predicted modal component functions are reconstructed to obtain the final prediction results.At the same time,VMD-BA-RF is combined with the excellent computing performance of the spark platform to reduce the operation time of the load forecasting process.The results show that the VMDBA-RF method can effectively predict the short-term power load,the average absolute error is 31.99 MW,the root mean square error is 53.28 MW,the average absolute percentage error is 0.60%,while the three indexes of the traditional RF method,VMD-RF method,and BA-RF method are larger than VMD-BA-RF method,so the proposed method has higher prediction accuracy.
作者
刘成龙
高旭
曹明
LIU Chenglong;GAO Xu;CAO Ming(State Grid Hebei Electric Power Co.,Ltd.,Shijiazhuang 050000,China;Information and Communication Branch of State Grid Hebei Electric Power Co.,Ltd.,Shijiazhuang 050000,China)
出处
《中国测试》
CAS
北大核心
2022年第4期159-165,共7页
China Measurement & Test
基金
国网河北省电力有限公司科技项目(SGHEXT00DDJS2000080)。
关键词
变分模态分解
蝙蝠算法
随机森林
短期负荷预测
SPARK
variational mode decomposition
bat algorithm
random forest
short term load forecasting
Spark
