Symmetry-Based Fault Detection and Diagnosis for Hydraulic Wind Turbine Pitch Systems

Abstract

Blandt de mest betydningsfulde og hyppigt forekommende fejl i hydrauliske pitch-systemer til vindmøller er intern lækage. Denne afhandling præsenterer en ny tilgang, der udnytter vindmøllens symmetri ved at sammenligne adfærden af signaler fra de tre vinger, hvilket muliggør fejldetektion og -diagnose gennem relative afvigelser. For at muliggøre simulering og analyse af systemets adfærd under forskellige driftsforhold er der udviklet en dynamisk model af et pitch-system, som er blevet valideret gennem tests udført på en fysisk testopstilling. Modellen er efterfølgende udvidet til at simulere alle tre vinger, hvor OpenFAST anvendes til at gengive realistiske turbinedynamikker. For at muliggøre simulering og test af systemet i drift er der designet og implementeret en kontrolstrategi til styring af pitch-vinklen. Ved at transformere signaler til vingevinkel-domænet er karakteristiske mønstre forbundet med intern lækage, friktion og ekstern lækage blevet identificeret, hvilket muliggør fejlisolering baseret på signalernes tendenser. Det er særligt bemærkelsesværdigt, at intern lækage medfører et fald i begge kammertryk i cylinderen udsat for fejl sammenlignet med de to raske cylindre. Validering på den fysiske testopstilling viser lignende mønstre i vinkel-domænet, når intern lækage introduceres. Baseret på disse observerede afvigelser foreslås en residualbaseret algoritme til fejldetektion og -diagnose. Algoritmen kombinerer statistiske og logiske metoder, hvor en residualtærskel anvendes til at detektere fejl, og logik benyttes til at diagnosticere typen af fejl baseret på det observerede mønster. Ved hjælp af denne algoritme blev lækagefejl helt ned til 0,06 L/min under møllens nominelle hastighed og 0,3 L/min over nominel hastighed med succes diagnosticeret i simulering. Lækageflowet over nominel hastighed blev estimeret med en maksimal fejl på 0,3 L/min. Eksperimentel validering bekræftede algoritmens evne til at detektere intern lækage i et fysisk system. Metodens enkelhed og effektivitet peger på et potentiale for implementering i idriftsatte vindmøller, forudsat yderligere forskning.

Among the most significant and frequently occurring faults in hydraulic pitch systems of wind turbines is internal leakage. This thesis introduces a new approach that utilises the symmetry of the wind turbine by comparing the behaviour of signals associated with each of the three blades, enabling fault detection and diagnosis through relative deviations. To allow for simulation and analysis of the system behaviour under various conditions, a dynamic model of a pitch system has been developed and validated through tests conducted on a physical setup. The model has been extended to simulate all three blades, utilising OpenFAST to replicate realistic turbine dynamics. To facilitate simulation and testing of the system in operation, a control strategy for the pitch angle has been designed and implemented. By transforming signals into the blade-angle domain, distinct patterns associated with internal leakage, friction and external leakage have been identified, enabling fault isolation based on signal tendencies. Notably, internal leakage is found to cause decrease in both chamber pressures of the cylinder subjected to the fault, compared to the two healthy cylinders. Validation on the physical test setup shows similar tendencies in the angle domain when internal leakage is introduced. Based on these observed deviations, a residual-based detection and diagnosis algorithm is proposed. The algorithm combines statistical and logic-based methods, using a residual threshold to detect faults and logic to diagnose the fault based on the observed pattern. Using this algorithm, leakage faults as low as 0.06 L/min under rated wind speed and 0.3 L/min above rated speed were successfully diagnosed in simulation. The leakage flows above rated wind speed were successfully estimated with a maximum estimation error of 0.3 L/min. Experimental validation demonstrated the algorithm's capability to detect internal leakage faults in a physical system. The simplicity and effectiveness of this method suggest potential for implementation in commissioned wind turbines, pending further research.

A master programme thesis from Aalborg University

Symmetry-Based Fault Detection and Diagnosis for Hydraulic Wind Turbine Pitch Systems