摘要
在训练样本不够丰富的情况下,采用交叉验证法可以有效地减少BP神经网络因训练样本不同的输入顺序和数量不足造成的误差。针对汽轮机故障诊断的试验数据证明了所提出方法的可靠性和实用性。
It introduced BP algorithm and the method of establishing BP neural network in Matlab. Under the condition of the sample not enough, using the cross-validation can effectively reduce the error caused by input sequence and amount of train sample. Test data proved the reliability and practicability of BP neural network.
出处
《电力科学与工程》
2008年第3期31-34,共4页
Electric Power Science and Engineering
