基于大数据分析技术的风电机组齿轮箱故障预警

Wind Turbine Gearbox Fault Warning based on Big Data Analysis Technology

齿轮箱的损伤和失效会直接或间接导致风电机组的故障或停机,造成重大的经济损失。为了能对齿轮箱的故障状态进行预警,本文提出了一个基于大数据分析技术的风电机组齿轮箱故障预警模型。首先利用最大互信息系数分析各工况参数与齿轮箱输出轴温度的相关性,提取主要影响因子作为具有记忆时序信息的循环神经网络的输入特征,构建正常状态下齿轮箱温度的残差分布,设置故障预警阈值,从而实现对风电机组齿轮箱故障状态预警。

Damage and failure of gearboxes will directly or indirectly lead to the shutdown of wind turbines, resulting in significant economic loss. This paper proposes an early warning model of the wind turbine gearbox fault based on big data analysis technology to provide early warning for gearbox faults. The maximal information coefficient is first used to analyze the correlation between operating parameters and the output shaft temperature of the gearbox. The main influencing factors are then extracted as the input features of the recurrent neural network with memory time sequence information to construct the residual distribution of the gearbox temperature under normal conditions. Finally, the warning threshold is set to realize the fault warning of the wind turbine gearbox.