摘要
结合自适应遗传算法(AGA)和BP算法各自的优点,本文构造了AGA BP混合算法作为神经网络的学习算法。分别采用BP、AGA和AGA BP神经网络对实验室中变压器超高频局部放电自动识别系统检测到的五种放电类型进行了模式识别。实验结果表明,AGA BP神经网络既解决了BP神经网络对初始权值敏感和容易局部收敛的问题,又提高了AGA神经网络的收敛速度、稳定性和求解质量,具有较高的识别率和较强的推广能力。
A pattern recognition system of ultrahighfrequency (UHF) PD designed by authors has been put forward newly to study the discharge properties in transformers. By combining adaptive genetic algorithm (AGA) with backpropagation (BP) algorithm, this paper presents AGABP hybrid algorithm to train neural network (NN). Using BPNN, AGANN and AGABPNN, we distinguish between basic types of defects appearing in transformers, such as corona, void, bubble, creeping discharge and floating discharge. The classification of defects is based on the calculated statistical operators extracted from discharge patterns. Tests in laboratory give satisfactory results of the classification process. Compared with BPNN and AGANN, AGABPNN can overcome the entrapment in local optical optimum of BPNN and the premature of AGANN. Thus, the convergence, discrimination and generalization ability of AGABPNN is improved remarkably.
出处
《电工电能新技术》
CSCD
2003年第2期6-9,55,共5页
Advanced Technology of Electrical Engineering and Energy