一种融合分形与BP网络PCA的海流机故障检测方法

Fault detection method for marine current turbine using fractal and BP network principal component analysis

针对直驱式永磁同步海流发电机在非稳定工况下故障检测问题,提出了一种融合分形与BP网络PCA的故障检测方法。该方法通过计算历史监测数据的分形维数,实现历史数据与待测数据的轨迹同步和工作模式统一。通过BP网络PCA建立预测模型,去除流速变化对监测数据的影响,判断数据变化趋势。利用统计方法获取残差函数,进行故障检测。为了验证所提方法的有效性和可靠性,搭建了海流样机实验平台,用于模拟多工况运行环境,并测试多种传统方法。实验结果表明:在海流机变转速同时变载荷工况下,该方法对比其它方法能够快速准确的检测到故障,适用于解决非稳定工况的故障检测问题。

To solve the fault detection problem of direct drive permanent magnet synchronous generator marine current turbine under unstable conditions, a fault detection method based on fractal and back propagation（BP） network principal component analysis（PCA） is proposed.The method was used to realize the synchronization of the test data with historical data and unify working mode by calculation of fractal dimension. Prediction model, established by PCA and neural network, was to remove the influence of water velocity changes and to evaluate the trend of data. The fault detection can be carried out by utilizing the residual characteristics of statistics.In order to verify the validity and reliability of the proposed method, the experimental platform of marine current turbine was built to simulate the operation environment and to test detection methods. Experimental results show that the fault under variable speed and variable load conditions can be detected rapidly with high accuracy. Compared with other methods,the proposed method is suitable for solving the detection problem under unstable work conditions.