Pre-Calculation of Wind Turbine Blade Fatigue Damage by Machine Learning

Abstract

Blade fatigue failure poses a great threat to the safe and stable operation of wind turbines. To solve the problem that the existing fatigue damage calculation methods are difficult to balance the calculation accuracy and calculation speed, a LightGBM machine learning fatigue damage calculation method is proposed. The method mainly analyses the blade 3D model by finite element software to determine the blade fatigue dangerous nodes and the stress magnitude corresponding to the unit load at the nodes, then uses Von Mises yield criterion to convert the load time series data into equivalent force time series data, and applies the rain-flow counting method and linear fatigue cumulative damage theory to calculate the blade fatigue damage amount. Finally, on the basis of the obtained minute-level blade fatigue damage amount, LightGBM machine learning algorithm is applied to train the fatigue damage calculation model, inputting wind speed, air density, turbulence intensity, wind shear and other wind parameters as well as the inflow angle and other wind turbine operation conditions, which can output the blade fatigue damage amount correspondingly. The proposed fatigue damage calculation method is evaluated to be highly accurate and feasible.