摘要
在基于支持向量机(SVM)的电力系统中期负荷预测的基础上,针对SVM参数难以确定的问题,在引进微分进化(DE)算法优化SVM参数的基础上,为了减少DE的寻优时间,提高全局搜索能力,用基于学习样本集噪声估计的方法确定SVM参数的范围作为DE的寻优范围,以指导DE寻优。同时,引进自适应算子,采用参数自适应DE(ADE)算法选择SVM参数。由于影响负荷的气温因素是模糊的,利用隶属度函数对气温因素进行模糊化处理,进一步提高了预测精度。将上述方法用于欧洲智能技术网络(EUNITE)竞赛数据的中期电力负荷预测,结果表明,该方法能够准确预测负荷变化,且比其他算法具有更高的预测精度,为电力系统负荷预测提供了重要手段。
Based on mid-term load forecasting with support vector machine (SVM) for power system, since SVM parameters are hard to be determined, the differential evolution (DE) algorithm is introduced. And to reduce the optimal time of DE and improve the global search ability, the method based on noise estimation of training sample is applied to determine the scope of SVM parameters as the scope of DE optimal, to guide optimal of DE. Meanwhile, the adaptive operators are introduced, and self-adaptive DE (ADE) algorithm is applied to optimize SVM parameters. Due to the influence of the temperature to load is vague, the membership function is used to fuzz the temperature to further improve the prediction accuracy. This method is used to mid-term load forecasting of European intelligence technology network (EUNITE) data, the results show that the method can accurately predict load changes, having higher prediction precision than other algorithms, which provides an important method for the power system load forecasting.
出处
《电力系统保护与控制》
EI
CSCD
北大核心
2012年第8期110-115,120,共7页
Power System Protection and Control
基金
华北电力大学校内基金(200814003)
关键词
中期负荷预测
支持向量机
微分进化算法
自适应
模糊理论
mid-term load forecasting
support vector machine(SVM)
differential evolution (DE)
self-adaptive
fuzzy theory