基于MDN和VMD-TPA-LSTM的泵站主机组劣化趋势预测

Deterioration trend prediction of pump station host group based on MDN and VMD-TPA-LSTM

为提高泵站主机组的安全稳定运行能力,解析其运行状态,获取机组设备的健康状况,准确预测其未来发展趋势,提出一种基于混合密度网络(Mixture Density Networks,MDN)和融合变分模态分解(Variational Mode Decomposition,VMD)与基于时序模式和注意力机制的长短时记忆网络(Temporal Pattern Attention-Long Short-Term Memory Network,TPA-LSTM)的泵站主机组劣化趋势预测模型。模拟结果表明,此法能够准确地表达机组的劣化趋势并可有效提高其预测精度。

In order to improve the safe and stable operation capability of the pump station host group,analyze its operating status,obtain the health status of the unit equipment,and accurately predict its future development trend,a pump station host group degradation trend prediction model based on Mixed Density Networks(MDN),Variational Mode Decomposition(VMD),and Temporal Pattern Attention Long Short Term Memory Network(TPA-LSTM)based on temporal pattern and attention mechanism is proposed.The simulation results show that this method can accurately express the deterioration trend of the unit and effectively improve its prediction accuracy.