Exploring Tensor Completion for Missing Data Estimation in Wind Farms

Abstract

The large number of greenhouse gas emissions caused by human activities, and their harmful effect on the earth’s climate, have reached a point where actions are needed. Wind energy is one of the available green energies that can be used to mitigate this problem. Predictive maintenance is of vital importance to ensure continuous wind power generation and is typically based on the use of sensor data from all wind turbine systems. But in some cases, data contain outliers or are not available at all due to sensor or system failures. In this letter, we explore the use of tensor completion methods to estimate missing data in this field. Experimental results demonstrate the usefulness of the proposed tensor completion algorithms, especially the high-accuracy low-rank tensor completion (HaLRTC) method, which outperforms the interpolation method used as a reference.