基于Box-Cox变换分位数回归与负荷关联因素辨识的中长期概率密度预测

中长期电力负荷预测是电力部门制定电力系统发展规划和稳定运行的重要前提.针对影响中长期电力负荷预测精度的多个因素,本文利用逐步回归方法,从众多影响负荷预测精度的关联因子中,对关键的影响因子进行辨识,并提出基于Box-Cox变换分位数回归和核密度估计相结合的概率密度预测方法,得出不同分位点下未来连续几年的概率密度预测结果,实现了对未来年用电量准确波动区间的预测.以安徽省的历史用电量和社会经济数据为例,进行仿真实验.结果表明:该方法不仅实现了中长期电力负荷概率密度预测,而且利用强关联因素提高了中长期电力负荷概率密度预测的精度,有效解决了考虑多因子的中长期电力负荷概率密度预测问题.

Medium and long term probability density forecasting based on Box-Cox transformation quantile regression and load relation factor identification

Medium and long term load forecasting is an important prerequisite for the power sector's development planning and stable operation. According to the multiple factors of influencing the medium and long term power load forecasting accuracy, this paper uses the stepwise regression method to identify the key influencing factors from a number of factors associated load forecasting, and proposes a probability density forecasting method based on the Box-Cox transformation quantile regression combined with kernel density estimation. The probability density forecasting results of load under the different quantiles at any year in the next few years are evaluated. The proposed method is likely to realize the accurate range prediction of future annual electricity consumption. The historical load and socio-economic data of Anhui province are adopted as simulation experiment. The results show that the proposed method not only realizes the medium and long term load forecasting, but also well improves the precision of medium and long-term power load probability density forecasting by means of introducing strong relation factors, and effectively solves medium and long term power load probability density forecasting problem considering multiple factors.