中国电机工程学报2016，Vol.36Issue(23)：6409-6417,9.DOI:10.13334/j.0258-8013.pcsee.160047
考虑负荷自适应检测和修复的鲁棒极限学习机短期负荷预测方法
Short-term Load Forecasting Method Based on Outlier Robust Extreme Learning Machine Considering Adaptive Load Detection and Repair
彭显刚 1郑伟钦 1林利祥 1刘艺1
作者信息
- 1. 广东工业大学自动化学院，广东省 广州市 510006
- 折叠
摘要
Abstract
There is low forecasting accuracy when predicting the power load influenced by the accumulation effect of temperature and humidity. For this reason, a short-term load forecast method based on outlier robust extreme learning machine(ORELM) algorithm was proposed. Load detection was implemented by the time vary Cook distance (TVCD) statistic and abnormal load data was repaired self-adaptively based on the nonparametric probability density function (NPDF) estimation. First, the history load and weather data were grouped seasonally. Then the TVCD statistic based on the adaptive forgetting factor (FF), which obtained by using Recursive Least Squared (RLS) algorithm, was constructed for anomaly (or influential value) detection depending on the seasonal real time load and the meteorological factors. Load repair was further analyzed by using the NPDF estimation which depends on the stochastic response of load to the weather. Considering the impact of the abnormal power load data on the forecasting accuracy, a regressive analysis was made by adopting the ORELM algorithm. The genetic algorithm (GA) was adopted to get the optimal parameters of the modification model to improve the forecasting accuracy. Case study shows that the novel load forecast method can improve the accuracy of prediction effectively.关键词
短期负荷预测/时变Cook距离/非参数概率密度函数估计/累积效应/鲁棒极限学习机算法Key words
short-term load forecast/time vary Cook distance/non-parameter probability density function estimation/accumulation effect/outlier robust extreme learning machine (ELM) algorithm分类
信息技术与安全科学引用本文复制引用
彭显刚,郑伟钦,林利祥,刘艺..考虑负荷自适应检测和修复的鲁棒极限学习机短期负荷预测方法[J].中国电机工程学报,2016,36(23):6409-6417,9.
