基于EEMD和LSTM的水电机组劣化度预测方法研究

针对水电机组作为低速旋转设备具有复杂的运行机理,在目前缺乏先验知识、故障样本较少的情况下,运用传统故障诊断方法很难对水电机组的运行状况做出正确判断的问题,提出一种基于集合经验模态分解法和长短期记忆神经网络相结合的水电机组劣化度预测方法.利用水电机组非故障运行期间的数据计算不同工况下的特征参数健康值标准,使用劣化度描述机...

RESEARCH ON FORECASTING METHOD OF HYDROPOWER UNIT DETERIORATION BASED ON EEMD AND LSTM

As a low-speed rotating equipment, the hydroelectric generator has a complicated operating mechanism. In the absence of prior knowledge and few fault samples, it is difficult to make a correct judgment on the operating status of a hydroelectric generator using traditional fault diagnosis methods. In view of the above problems, a method for predicting the degradation degree of hydroelectric generators based on the combination of ensemble empirical mode decomposition （EEMD） and long short-term memory （LSTM） is proposed. Using the data of the hydroelectric generator during non-failure operation to calculate the standard of the health value of the characteristic parameters under different working conditions, using the degree of degradation to describe the degree to which the characteristic value deviates from the health value during the operation of the generator. Furthermore, the EEMD method is used to decompose the original non-stationary degradation time series into several stationary component sequences. Finally, the LSTM prediction algorithm is used to predict the deterioration degree of the generator. The prediction results show that the method has good prediction accuracy and can accurately predict the deterioration trend of hydroelectric generators.