新冠肺炎疫情影响下基于多源数据驱动的电力系统负荷预测方法

Multi-source Data-driven Load Forecasting Method for Power System Under the Influence of COVID-19 Epidemic

新冠肺炎疫情的爆发及快速传播给中国乃至全球造成了深远的影响,对社会带来全方面影响的同时也改变了大众的能源消费习惯,催生了新的电力需求侧的变革与契机。疫情对社会各方面的影响可以通过病例数、在线办公人数、电力市场数据、外卖快递数据、社交距离数据、城市亮度数据与二氧化氮等数据得到更为直接的展现,利用大数据相关系数算法,结合历史数据与天气数据,表征疫情下各个社会属性量与电力系统负荷之间的强弱联系。提出了疫情下基于极端梯度增强算法(extreme gradient boosting,XGBoost)分类器的电力系统负荷预测模型,利用新冠肺炎病例数据将地区划分为重度疫情期、轻度疫情期与疫情恢复期,探讨不同时间段新冠肺炎疫情影响下,各个社会属性量数据与电力系统负荷之间的关系,提出了多源数据驱动下的电力系统负荷预测模型,相关试验结果验证了提出方法的有效性与科学性。

COVID-19 epidemic outbreak and rapid spread have a profound impact on our country and even the whole world.It has brought about changes in the whole society and changed the public’s energy consumption habit,which has brought about new changes and opportunities for the demand side.The outbreak of COVID-19 epidemic is difficult to end in the short term.During the outbreak and recovery period,due to the impact of administrative policies and social psychology,the load change of power system not only has a short-term rapid decline,but also has a long-term change caused by the change of people’s energy consumption habits.The impact of the epidemic on all aspects of society can be more directly shown by the number of cases,number of online office workers,power market data,take out and express delivery data,social distance data,city brightness data and nitrogen dioxide data.By using the big data correlation coefficient algorithm,combined with historical data and weather data,the relationship between various social attributes and power system load under the epidemic situation is characterized strong and weak ties.This paper proposes the COVID-19 epidemic load forecasting model based on XGBoost classifier.According to the COVID-19 data,the area was divided into severe epidemic period,mild epidemic period and epidemic recovery period.The relationship between various social attribute quantity data and power system load under the influence of COVID-19 epidemic in different time periods is discussed,and a power system load forecasting model driven by multi-source data is proposed.The validity and scientificity of the proposed method are verified by the related test results.