基于大数据技术的广州市台风负荷影响分析和预测
|
摘要
论文通过研究台风期间天气气象指标的变化规律,分析各气象指标变化对广州市日最大负荷的影响,进而建立台风负荷预测模型,研究台风期间气象变化对广州市小时最大负荷的影响。论文分别从台风气象、广州气象及广州电力负荷三个维度构建了台风预测模型,通过相关分析技术筛选台风期间对广州电力负荷有显著影响的指标,最后采用多种机器学习方法进行预测。实验结果表明XGBOOST方法优于其他机器学习方法且模型也通过假设检验,模型对数据的拟合程度达到68.1%,台风期间负荷的外推预测平均准确率能达到85.22%。
Through studying the variation rule of weather and meteorological indexes during typhoon, the paper analyzes the influence of the changes of meteorological indexes on the maximum dailyload in Guangzhou, and then establishes a typhoon load forecasting model, so as to study the influence of meteorological changing on the maximum hourly load in Guangzhou during typhoon. The typhoon forecasting model is established from three dimensions: typhoon meteorology, Guangzhou meteorology and Guangzhou power load. Through correlation analysis technology, the index has significant influence on Guangzhou electric power load is filtered. Finally, a variety of machine learning methods are used to predict the power load. The experimental results show that XGBOOST method is superior to other machine learning methods and the model also passes hypothesis test. The fitting degree of the model to data reaches 68.1%, and the average accuracy of load extrapolation during typhoon can reach 85.22%.
Through studying the variation rule of weather and meteorological indexes during typhoon, the paper analyzes the influence of the changes of meteorological indexes on the maximum dailyload in Guangzhou, and then establishes a typhoon load forecasting model, so as to study the influence of meteorological changing on the maximum hourly load in Guangzhou during typhoon. The typhoon forecasting model is established from three dimensions: typhoon meteorology, Guangzhou meteorology and Guangzhou power load. Through correlation analysis technology, the index has significant influence on Guangzhou electric power load is filtered. Finally, a variety of machine learning methods are used to predict the power load. The experimental results show that XGBOOST method is superior to other machine learning methods and the model also passes hypothesis test. The fitting degree of the model to data reaches 68.1%, and the average accuracy of load extrapolation during typhoon can reach 85.22%.
引文
[1]刘思捷,刘昌,李小燕,等.台风对广东电力负荷特性的影响[J].云南电力技术,2012,40(1):91-93.
[2]莫若慧,符子星,曾炎,等.台风天气下电网负荷对气象因素的敏感度分析[J].南方电网技术,2015,9(2):79-83.
[3]李小燕,文福拴,卢恩,等.基于相似日负荷修正的台风期间短期负荷预测[J].电力系统及其自动化学报,2013,25(3):82-89.
[4]刘晶.基于相似日和支持向量机的短期负荷预测研究[D].广州:华南理工大学,2010.
[5]林辉,刘晶,郝志峰,等.基于相似日负荷修正的节假日短期负荷预测[J].电力系统保护与控制,2010,38(7):47-51.
[6]于龙,郑益慧,王昕,等.基于SVM与相似日的短期电力负荷预测[J].电工技术学报,2013,28(1增):217-223.
[7]Taylor J W,Buizza R.Neural Network Load Forecasting with Weather Ensemble Predictions[J].IEEE Power Engineering Review,2007,22(7):59-59.
[2]莫若慧,符子星,曾炎,等.台风天气下电网负荷对气象因素的敏感度分析[J].南方电网技术,2015,9(2):79-83.
[3]李小燕,文福拴,卢恩,等.基于相似日负荷修正的台风期间短期负荷预测[J].电力系统及其自动化学报,2013,25(3):82-89.
[4]刘晶.基于相似日和支持向量机的短期负荷预测研究[D].广州:华南理工大学,2010.
[5]林辉,刘晶,郝志峰,等.基于相似日负荷修正的节假日短期负荷预测[J].电力系统保护与控制,2010,38(7):47-51.
[6]于龙,郑益慧,王昕,等.基于SVM与相似日的短期电力负荷预测[J].电工技术学报,2013,28(1增):217-223.
[7]Taylor J W,Buizza R.Neural Network Load Forecasting with Weather Ensemble Predictions[J].IEEE Power Engineering Review,2007,22(7):59-59.