摘要
概率神经网络主要应用于模式识别、分类等问题,有很多优良的性质,但结构上固有的2个缺点限制了其应用。文中通过把测试样本由线性不可分转化为线性可分,在不改变概率神经网络结构的条件下,对其2个缺点进行了改进,扩大了方法的应用范围。经电压暂态扰动分类的仿真验证了改进后的方法可取得更高的正确率。
Probabilistic neural networks are commonly used to pattern recognition and classification. It has many well properties, but two structural shortcomings limit its applications. Keeping the optimal properties, an improved method turns the linear indivisible samples to linear divisible ones. Not changing the structure of probabilistic neural networks, this method improves its two shortcomings to extend its applications. The simulation results for classification of the transient voltage disturbances verify that the method gives a very good accuracy of detection.
出处
《电力系统自动化》
EI
CSCD
北大核心
2007年第18期34-38,共5页
Automation of Electric Power Systems
关键词
电能质量
概率神经网络
电压暂态扰动分类
power quality
probabilistic neural networks
disturbance classification
