Optimizing the operation and maintenance of wind turbines is crucial as the wind energy sector continues to expand. Predicting the mass imbalance of wind turbines, which can seriously damage the rotor blades, gearbox, and other components, is one of the key issues in this field. In this work, we propose a machine learning-based method for predicting the mass imbalance of wind turbines utilizing information from multiple sensors and monitoring systems. We collected data and trained the model from Adwen AD8 wind turbine model and evaluated on the real wind turbine SCADA data which is located at Fraunhofer IWES, Bremerhaven. The data included various parameters such as wind speed, blade root bending moments and rotor speed. We used this data to train and test machine learning classification models based on different algorithms, including extra-tree classifiers, support vector machines, and random forest. Our results showed that the machine learning models were able to predict the mass imbalance percentage of wind turbines with high accuracy. Particularly, the extra tree classifiers with blade root bending moments outperformed other research for multiclassification problem with an F1 score of 0.91 and an accuracy of 90%. Additionally, we examined the significance of various features in predicting the mass imbalance and observed that the rotor speed and blade root bending moments were the most crucial variables. Our research has significant effects for the wind energy sector since it offers a reliable and efficient way for predicting wind turbine mass imbalance. Wind farm operators can save maintenance costs, minimize downtime of wind turbines, and increase the lifespan of turbine components by identifying and eliminating mass imbalances. Also, further investigation will allow us to apply our method to different kinds of wind turbines, and it is simple to incorporate into current monitoring systems as it supports prediction without installing additional sensors. In conclusion, our study demonstrates the potential of machine learning for predicting the percentage of mass imbalance of wind turbines. We believe that our approach can significantly benefit the wind energy industry and contribute to the development of sustainable energy sources.

För att en kvalitativ och driftsäker kraftomvandling från vind till elkraft skall erhållas ställs krav på en mängd olika faktorer. Klimatologiska och tekniska faktorer kräver korrekt dimensionering och anpassning av omvandlingstekniken för att största möjliga energimängd skall kunna omvandlas på ett driftsäkert och energieffektivt vis som vindkraftsägare, nätägare kräver. Vinden är som bekant en oberäknelig kraftkälla. Variationerna i styrka och tid kan innebära en hel del driftoptimeringsproblematik med efterverkningar för vindkraftsverket, nätet och belastningen. Konsekvenserna kan bero på vilken typ av teknik som är installerad i de olika delarna av energisystemet och i områdets elnät. Varierande effektbehov och effektfaktor i tiden har också en signifikant betydelse för elnätets stabilitet.  Anpassning av drift i elnätet mot den turbulenta vinden ger ofta avvikelser på spänning och effektflöden, speciellt i extrema situationer där svaga elnät existerar. God samverkan med vindkraftverket och nätets varierande aktiva och reaktiva effektbehov till energianvändaren kan ge förutsättningar för god elkvalitet och därmed optimerad och säker drift med få avbrott över tiden. Ofta kan och är det mekaniska, elektrotekniska val i vindkraftverket och tillhörande elsystem som spelar en avgörande roll för hur lönsam investeringen blir under verkets tekniska livslängd. Nätägaren eftersträvar en god interagering mellan elnät och elproduktion som ger upphov till få medeltalsfel mellan avbrott och elfel, detta kallas i branschen för ”Mean time between failures” MTBF.  Enligt Svensk energi skall långsamma och snabba spänningsvariationer samt övertoner och erforderlig kortslutningseffekt utredas och jämföras mot de krav och villkor som råder vid elektrisk integrering av elproduktion i elnätet. Förstudien har kommit fram till två lämpliga anslutningsförslag med vindkraftsprojektering. Det beskrivs senare i rapporten förslag på två tillhörande Smart Grid varianter med energilagring för anslutningsförslagen i det befintligt svaga lokala elnätet. / To achieve a reliable and qualitative power conversion from the wind into electric power, a variety of factors and demands need to be obtained. Climatological and technological factors requires proper dimensioning and adjustment of the conversion technology, to harvest the greatest possible amount of energy and to be converted in a reliable and energy efficient way, that windmill owners, power grid owners require. The wind is as familiar an unpredictable power supply. The variations in intensity over time could mean a number of drive optimization problems with after-effects of the wind turbine, power grid and load. The consequences may depend on which type of technology that is installed in the different parts of the energy system. The area's power grid and varying power needs with characteristics over time, also has a significant importance. The turbulent wind gives deviations of voltage and power flow, especially in various extreme situations in weak power grids. Good interaction between the wind turbine and power grid with varying active and reactive power demand for the energy users, provides conditions for a good power quality and thus, an optimal and safe operation with few interruptions over time. It can be, and often is the mechanics, electro-technical choices in the wind turbine and associated electrical systems that play a critical role in how profitable installation is during the wind turbines technological life. The power grid owner strives for a good interaction between the power grid and electrical generation which rise for few faults between interruptions and errors. In the branch this is known as "Mean time between failures" MTBF.  According to the Swedenergy, harmonics, slow and fast voltage variations including required short-circuit power should be investigated and compared with those requirements and terms that prevails with electrical integration of power into the grid. The feasibility study has concluded two suitable power connection proposals including wind mapping research, later in this report it is described and suggested two related Smart Grid variants with energy storage for the two power connection proposals in the existing weak local grid. / Judith Saari var betygsättare på muntlig presentation.

Oberschwingungen als ein Merkmal der Elektroenergiequalität gewinnen durch die starke Netzintegration leistungselektronisch geregelter Anlagen wie Windenergieanlagen und nichtlineare Lasten zunehmend an Bedeutung. Bestehende Normen entsprechen nicht den zukünftigen Erfordernissen des Elektroenergiesystems und bedürfen diesbezüglich einer Überarbeitung. In der Arbeit werden wesentliche Einflussfaktoren auf das Oberschwingungsverhalten von Windparks identifiziert, beschrieben und modelliert. Dabei wird der stochastische Charakter der Oberschwingungen mithilfe von probabilistischen Ansätzen erfasst. Des Weiteren wird ein neuer Ansatz zur Untersuchung der Wechselwirkung zwischen dem Windpark und dem vorgelagerten Netz entwickelt. Der Ansatz ermöglicht, die vom Windpark verursachte Änderung der Oberschwingungsspannung am Netzanschlusspunkt zu ermitteln. Diese Arbeit liefert einen Beitrag zur Verbesserung bestehender Normen für die Anbindung von Windparks. / Harmonics, as one of the power quality criteria, are increasingly gaining attention due to the progressive contribution of renewable energy resources and the application of the nonlinear load in the power system. Current standards do not conform to the future requirements of the power system, thus requiring a revision. In this work, main influence factors on the harmonic behavior of wind farms are identified, explained, and modelled. Thereby, the stochastic nature of harmonics is taken into account using probabilistic approaches. Moreover, a novel approach is developed to investigate the interaction between the wind farm and the upstream grid. With the aid of this approach, it is possible to determine the voltage change caused by the wind farm at the point of connection. This work contributes to improve the existing standards for the connection of wind farms.