Deep Unsupervised Learning for Condition Monitoring and Prediction of High Dimensional Data with Application on Windfarm SCADA Data

Abstract

In this work we are addressing the problem of statistical modeling of the joint distribution of data collected from wind turbines interacting due to collective effect of their placement in a wind-farm, the wind characteristics (speed/orientation) and the turbine control. Operating wind turbines extract energy from the wind and at the same time produce wakes on the down-wind turbines in a park, causing reduced power production and increased vibrations, potentially contributing in a detrimental manner to fatigue life. This work presents a Variational Auto-Encoder (VAE) Neural Network architecture capable of mapping the high dimensional correlated stochastic variables over the wind-farm, such as power production and wind speed, to a parametric probability distribution of much lower dimensionality. We demonstrate how a trained VAE can be used in order to quantify levels of statistical deviation on condition monitoring data. Moreover, we demonstrate how the VAE can be used for pretraining an inference model, capable of predicting the power production of the farm together with bounds on the uncertainty of the predictions. Examples employing simulated wind-farm Supervisory Control And Data Acquisition (SCADA) data are presented. The simulated farm data are acquired from a Dynamic Wake Meandering (DWM) simulation of a small wind farm comprised of nine 5MW turbines in close spacing using OpenFAST. The contribution of this work lies in the introduction of state-of-the-art machine learning techniques in the general context of condition monitoring and uncertainty quantification. We show how the high dimensional joint probability distribution of condition monitoring parameters can be analyzed by exploiting the underlying lower dimensional structure of the data imposed by the physics of the problem. The process of making use of the trained joint distribution for the purposes of inference under uncertainty and condition monitoring is clearly exposed.

Conditional variational autoencoders for probabilistic wind turbine blade fatigue estimation using Supervisory, Control, and Data Acquisition data

“…In the offered case study, we demonstrated this idea on short‐term fatigue damage equivalent loads for wind turbines. Other wind‐energy related problems where CVAEs may find application are, for instance, wind farm level SCADA data modeling 48 or turbulence modeling 84,85 where finer turbulence scales are resolved in a data‐driven manner in place of the mostly empirical and mathematically simplified standard classical approaches.…”

“…The VAE is used as a means of learning features from the monitoring data. In Mylonas et al 48 we have applied conditional variational autoencoders n the problem of modeling the SCADA summary statistics data, on a simulated wind farm, with the purpose of establishing VAEs as a tool for condition monitoring.…”

On generative models as the basis for digital twins

“…A very recent methodology StatFEM (Duffin et al, 2021;Girolami et al, 2021), provides an elegant Bayesian framework for both building and updating generative FE models. In terms of generative black-box models, GANs are by no means the only option; in fact, one alternative-the variational auto-encoder (Kingma and Welling, 2014)-has already proved to be generally useful in engineering problems; particularly in condition monitoring problems (e.g., Mylonas et al, 2020). Another versatile generative framework is provided by Gaussian processes (GPs) (Rasmussen and Williams, 2005).…”

Low-Cost Design and Implementation of Cloud SCADA System

“…At a supervisory control system, the major task of SCADA systems is to monitor a system's processes and apply the appropriate controls accordingly. SCADA systems are basically CPSs used in industries which included in a wide number of application areas [1][2][3][4].This solution present in the industry is Web SCADA, which provides multiple benefits that include anywhere/anytime accessibility to the system through a secure web browser connection. When the future Internet is considered, new technologies replace old technologies, hence, we can say that the integration of industrial business systems and the cloud concept has made the integrated SCADA systems more vulnerable compared with classical SCADA systems.…”