摘要
针对温度因素对中期电力负荷的影响,在现有的神经网络预测、区间预测和概率密度预测方法的基础上,研究在不同分位点上温度和历史负荷对电力系统中期负荷分布规律的影响,提出基于神经网络分位数回归的中期电力负荷概率密度预测方法。根据连续的条件分位数函数预测中期负荷在某天的概率密度,获得更多关于中期负荷预测信息。同时,通过比较在考虑温度因素下和不考虑温度因素下的条件概率密度预测曲线以及峰值对应的点预测值,可以得出,预测当天温度对中期负荷预测有较重要的影响,这为降低中期电力负荷预测的不确定因素提供了更多的决策信息和预测结果。
In allusion to the influence of temperature factor on medium-term power load, based on exiting neural network prediction, interval prediction and probability density prediction the influences of temperatures and historical loads at different quantiles on the distribution rule of power system medium-term load is researched and a method to predict neural network quantile regression based medium-term probability density of power load is proposed. According to continuous conditional quantile functions the probability density of medium-term power load on a certain day is predicted to obtain more information related to medium-term power load. Meanwhile, the comparison results of conditional probability density prediction curves, in which the temperature factors are considered and not considered respectively, and the prediction values corresponding to peak load points show that the temperature at the very predicted day evidently influences the predictive result of medium-term power load, so it offers more decision information and prediction results, in which the uncertain factors in medium-term power load prediction are decreased.
出处
《电网技术》
EI
CSCD
北大核心
2015年第1期176-181,共6页
Power System Technology
基金
国家自然科学基金项目(71401049)
高等学校博士学科点专项科研基金资助项目(20130111120015)
安徽省自然科学基金项目(1408085QG137)
全国统计科研计划重点项目(2012LZ041)~~
关键词
温度
概率密度预测
神经网络分位数回归
中期负荷
temperature probability density prediction neural network quantile regression medium-term load
