Fault Diagnosis Of Wind Turbine Driven By Digital Twin

As wind turbine design gradually moves towards higher single-unit capacity and larger rotor diameters,issues such as complex structure,high maintenance difficulty and relatively high failure rate become increasingly prominent.The maintenance cost of onshore wind turbines accounts for a significant proportion of total investment in wind farms while that of offshore wind turbines is several times that of onshore ones.Therefore it is particularly important to explore more universal,accurate and flexible fault monitoring,diagnosis and early warning models.Wind turbines are typical complex electromechanical-hydraulic-magnetic coupling devices whose internal physical quantities often interact in ways that are not always clear.This has led to data-driven fault warning models often being unexplainable or having poor universality or requiring a lot of manual adjustment before they can be adapted to a specific model type.This is one of the biggest challenges in fault diagnosis for wind turbines:low utilization rates for fault warning models and repetitive work required.Digital twins serve as a link between physical and digital worlds with modeling methods that combine mechanisms with data having potential beyond previous models in terms of universality accuracy and timeliness.Therefore introducing digital twin theory into fault diagnosis for wind turbines may help solve this problem.This paper first establishes an aerodynamic model for a wind turbine rotor based on blade element momentum theory to simulate how it captures wind energy from its surroundings where blades are simplified into a series of actuator rings separated by ten cross-sections along chord lines defined entirely by four sets of airfoil data sets.A method for establishing transmission chain and structural models for a wind turbine based on multibody dynamics theory is also proposed;secondly using design information about units and prior knowledge about mechanisms historical data to deduce control logic for the torque and pitch control model;thirdly combining established aerodynamic model,structural multibody dynamic model,torque and pitch control model into complete digital twin model,Then use the above model for simulation testing,comparing simulation results with historical monitoring data iterating correcting digital twin model ultimately verifying model accuracy.Further targeting two cases involving faults in pitch and generator systems at two different wind farms residual error constructed from differences between simulation results actual monitoring data using established digital twin model used as fault characteristic variables with complete interpretability physical meaning using residual error as fault characteristic variables diagnosing analyzing pitch and generator system faults at two aforementioned cases verifying universality effectiveness feasibility.