自适应学习速率法在变压器故障诊断中的应用

为了提高电力变压器故障诊断的准确率,针对油中溶解气体分析,提出了一种基于误差自动调节修正因子的自适应学习速率法,使神经网络通过自身的误差变化过程自动调整学习速率修正因子,保证网络总是以最大的可接受学习速率进行训练,从而提高网络收敛速度。针对电力变压器故障气体及故障类型的特点,建立了电力变压器故障诊断BP(Back-Propagation)网络模型,应用该算法和原算法对该故障诊断网络模型进行训练。仿真结果表明,该算法的训练次数减少了35.4%,收敛速度提高了44.9%,有效地改善了网络模型的性能。将该算法应用于电力变压器故障诊断,能较为精确地判断出电力变压器的故障类型,故障诊断准确率达90.8%。

Application of Adaptive Learning Rate Method in Power Transformer Fault Diagnosis

To improve the accuracy of power transformer fault diagnosis,aiming at dissolved gases analysis,an algorithm of adaptive learning rate method with error-based self-regulation amendment factors is given.During the network training process, the neural network automatically adjusts the amendment factors of learning rate by the changing process of network error.It can ensure that the network is trained at the largest accepted learning rate.So the convergence speed is improved.According to the fault gases and fault types of power transformer,BP(Back-Propagation) neural network model for power transformer fault diagnosis is established and trained by the new algorithm and the original algorithm.Comparing with the original algorithm,the training results showed that the training times are decreased by 35.4% and the convergence speed is increased by 44.9%.The new algorithm effectively improved the performance of network model,decreased the training times and improved the network convergence speed.Applying the algorithm to power transformer fault diagnosis can detect the fault types accurately.The accuracy of fault diagnosis is 90.8%.