Session: 5.1 - Advanced Tools for Cyber-Physical Systems and Digital Twins

Paper Number: 118908

118908 - Hybrid Digital Twins Technology for Energy Industry 

Digital Twins have been offered as a solution for connected power systems to solve issues related to model-predictive operations, asset management, field management support, post-mortem analyses, and collaborative decision-making. However, most implementations rely heavily on ingesting large amounts of data to build black box machine learning (ML) models and training artificial intelligence (AI) algorithms to make decisions based on such black-box models which lack explainability and limited predictive capabilities in untested domains. Moreover, these implementations do not provide a consistent co-simulation framework to integrate multi-domain, multi-fidelity and multi-scale simulations. The use of multiple simulation tools along with limited capability to validate and enhance simulation models with measurement data coupled with the difficulty to incorporate such models in Industrial Internet of Things (IIoT) platforms poses a major barrier to the deployment of Digital Twins and reaping their benefits.

Ansys has integrated a complementary suite of technologies to enable Hybrid Digital Twins to resolve these issues by combining the inferential and deterministic capabilities of physics-based simulations with the predictive capabilities of ML models to deliver unique grey- and white-box models to inform real world operations of power systems. This involves an advanced reduced-order-modeling technology that leverages statistical and physics-informed machine learning techniques to build fast executing models which achieve superior accuracy using small datasets even outside the original training range. We also have a novel Fusion Modeling technique which enables engineers to improve simulations through comparison with measurement data by compensating for missing physics in simulations. This combination approach is further bolstered by uncertainty quantification techniques like Bayesian Inference for parameter estimation and scatter identification to model performance drifts, equipment aging, and anomalies for fault diagnostics and predictive maintenance. Ansys provides a comprehensive co-simulation platform to integrate multi-fidelity simulations from different domains and vendors in a single environment to facilitate dynamic digital twins and also, ensures data security within the Hybrid Digital Twin through the use of encryption to protect model intellectual property (IP) and critical infrastructure data to expose just the requested or relevant variables. Ansys partners with IIoT platform vendors like PTC, Amazon, and Microsoft to incorporate Ansys Digital Twins within existing data analytics infrastructure. This presentation will provide an overview of Ansys Digital Twin technology along with customer examples of Ansys Digital Twin technology from a variety of industries including power generation in use cases such as asset monitoring, fault diagnostics, virtual sensors, equipment performance drift and aging, operations dashboarding, predictive maintenance, uncertainty quantification and propagation and improve performance and operational maps using Fusion Models.

Presenting Author: KALYAN CHAKRAVARTHY Sharna Ansys

Presenting Author Biography: Kalyan is a systems engineer working at the intersection of systems simulations, predictive analytics, and process automation to enable virtual product development and deploy simulation-based sustainment practices. At Ansys, he works with customers across industries to develop simulation processes to enable digital twin technologies. Before joining Ansys, he worked at Ford Motor Company where his simulation based virtual validation efforts led to multi-million-dollar savings annually across multiple product lines. His interest in this area started while developing and deploying a simulation-based virtual prototype platform for the world's largest offshore wind turbine test rig at Department of Energy and Clemson University.