Homological Time Series Analysis of Sensor Signals from Power Plants

TL;DR

This paper introduces a novel topological data analysis approach combined with deep learning architectures to classify power plant sensor signals, leveraging persistence diagrams and residual networks for improved data visualization and processing.

Contribution

It presents a new method integrating topological data analysis with deep neural networks for power plant sensor signal classification, including hyper-parameter derivation and multi-input architecture.

Findings

Effective classification of power plant sensor signals achieved.

Persistence diagram representations enhance data preprocessing and visualization.

Deep residual networks improve learning from topological features.

Abstract

In this paper, we use topological data analysis techniques to construct a suitable neural network classifier for the task of learning sensor signals of entire power plants according to their reference designation system. We use representations of persistence diagrams to derive necessary preprocessing steps and visualize the large amounts of data. We derive deep architectures with one-dimensional convolutional layers combined with stacked long short-term memories as residual networks suitable for processing the persistence features. We combine three separate sub-networks, obtaining as input the time series itself and a representation of the persistent homology for the zeroth and first dimension. We give a mathematical derivation for most of the used hyper-parameters. For validation, numerical experiments were performed with sensor data from four power plants of the same construction type.