Convolutional Neural Network for Wind Turbine Failure Classification Based on SCADA Data

Abstract

Listen

Translate article

As a renewable energy source and an alternative tofossil fuels, the wind power industry is growing rapidly. However,due to harsh weather conditions, wind turbines (WT) still face manyfailures that raise the price of energy produced and reduce thereliability of wind energy. Hence, the use of reliable monitoringand diagnostic systems of WTs is of great importance. Operationand maintenance expenses represent 30% of the total cost of largewind farms. The installation of offshore and remote wind farmshas increased the need for efficient fault detection and conditionmonitoring systems. In this work, without using specific customdevices for monitoring conditions, but only increasing the samplingfrequency in the sensors already available (in all commercial WT) ofthe supervisory control and data acquisition system (SCADA), data-driven multiple fault detection is performed, and a classificationstrategy is developed. The data is processed, and subsequently,using a convolutional neural network (CNN), six faults are classifiedand evaluated with different metrics. Finally, it should be noted thatthe classification speed allows the implementation of this strategyto monitor conditions online in real under-production WTs.