基于智能相似日识别及偏差校正的短期负荷预测方法

在传统负荷预测理论的基础上,提出了基于智能相似日识别及偏差校正的新型短期负荷预测方法。首先构建地市—相关因素特征矩阵,通过判断矩阵相关性智能选取负荷相似日,从而实现负荷曲线的一次预测。在此基础上,建立了实时气象偏差校正策略,采用XGBoost算法进行负荷曲线的二次偏差校正,达到短期负荷预测的目标。算例研究表明,该策略能够有效提升短期负荷预测精度,而且具有较好的自适应特性,可以应用于电力系统短期负荷预测实践。

A short-term load forecasting method based on intelligent similar day recognition and deviation correction

Based on the traditional load forecasting theory, this paper proposes a new short-term load forecasting method based on intelligent similar day recognition and deviation correction. Firstly, the characteristic matrix of prefecture-city and correlation factors is constructed to select the most similar day of load curve through calculating matrix correlation coefficient. On this basis, the real-time meteorological deviation correction strategy which adopts the XGBoost algorithm is established to carry out the secondary deviation correction of the load curve, so as to achieve the goal of short-term load prediction. An example study shows that this strategy can effectively improve accuracy of short-term load forecasting, and also has good adaptive characteristics. Therefore, this method can be applied to the short-term power load forecasting practice. This work is supported by Science and Technology Project of State Grid Corporation of China (No. 52150016006B).