摘要
为向大工业提供更为精确的月度负荷预测,提出将XGBoost(极端梯度上升)算法引入电网负荷预测,对负荷及相关影响因素指标进行异常识别填补、指标转换、独热编码等数据预处理工作;结合关联分析,对相关影响因素指标进行筛选;最后采用XGBoost算法进行月度负荷建模预测。结合实例,对比XGBoost与支持向量机、神经网络模型算法在大工业用户近几年历史月度负荷数据建模预测中的应用,发现XGBoost模型应用效果较佳,具有一定的实用性。
In order to provide more accurate monthly load forecasting for large industries, XGBoost (extreme gradient rise) algorithm is introduced into load forecasting of power grid, and the data preprocessing work such as anomaly identification, index conversion and unique heat coding is carried out;the related influencing factor indexes are screened by combining the correlation analysis;Finally, XGBoost algorithm is used for modelling and forecasting of monthly load. In an example, by comparing the application of XGBoost with that of support vector machine and neural network model algorithm in the past monthly load forecasting in the last years, it is found that XGBoost model has better application effect and practicability.
作者
钱仲文
陈浩
纪德良
QIAN Zhongwen;CHEN Hao;JI Deliang(State Grid Zhejiang Electric Power Co., Ltd., Hangzhou 310007, China;Zhejiang Huayun Information Technology Co., Ltd., Hangzhou 310008, China)
出处
《浙江电力》
2019年第5期77-82,共6页
Zhejiang Electric Power
基金
国网浙江省电力有限公司科技项目(5211HZ17000C)
