Oehmcke S. Deep learning of virtual marine sensors

University of Oldenburg, 2018. — 169 p.This thesis proposes an approach to build virtual sensors based on machine learning to replace broken physical sensors. These virtual sensors are trained with sensor data from the Time Series Station Spiekeroog (TSS) and the Biodiversity-Ecosystem Functioning across marine and terrestrial ecosystems (BEFmate) project in the Wadden Sea. In the first part of my work, I begin by explaining the data and its preprocessing. Next, an unsupervised extreme event detection task on the TSS data with a subsequent expert evaluation is presented. I propose an imputation method for longer consecutively missing values, which penalizes previously interpolated values based on the length of the gaps with a k-nearest neighbors approach. In the second part, I design a neural network architecture to model broken sensors. The foundation is a bidirectional recurrent neural network with long short-term memory (bLSTM) that utilizes my time dimensionality reduction method exPAA. Then, I introduce convolutional layers, uncertainty predictions, and my input quality based dropout layer to the architecture, which proves to outperform the architecture with only bLSTM layers.