Emerging Technologies for Fuel Management Application: Optimization of Core Designs using Artificial Intelligence

Price: $49

Organizers: Westinghouse/INL

Time: 4 hrs total, 1 hr ACE and 2.5 hr RAVEN and 0.5 hr both presenters discuss comparison of surrogate modeling techniques and GA for this application.

This will be a highly interactive demonstration, but the participants will not be able to run the codes due to license complications. So, they are welcome to bring laptops if they want to download the open source Raven code as example, but they won’t have access to the nodal simulation codes to be able to actually generate any LPs.

Introduction to RAVEN: RAVEN is a multi-purpose stochastic platform that integrates uncertainty propagation, machine learning, optimization, and data analysis methods, and it provides a unique language to apply these methods to user-provided simulation models. With RAVEN, users can create customizable statistical analysis/optimization workflows where the response of simulation models is explored (e.g., for uncertainty propagation, model optimization, model calibration and model validation) for a variety of initial and operating conditions and the resulting data can be analyzed using machine learning, data mining and artificial intelligence algorithms. RAVEN orchestrates these machine learning/digital twinning pipelines on multiple operating systems and hardware configurations, ranging from laptops to high performance computing (HPC) environments. RAVEN also provides a plug-in interface that has already been leveraged by many system analysis and design tools, which enable simple multi-code integration across simulation tools.

An overview of the software is available at https://github.com/idaholab/raven/wiki

The software is open source and can be downloaded at: https://github.com/idaholab/raven

Training Objectives: The first objective is to provide a general understanding of the RAVEN package and its main capabilities. Second, a series of practical examples will be provided in ascending level of complexity, starting from the simplest statistical analysis to the generation of the complex machine learning models and their utilization in system analysis and uncertainty quantification. Third, the system optimization, especially, plant fuel reload optimization with genetic algorithms will be covered. This training section will include a theoretical/code usage overview of the subject capability and demonstrations. If the attendees would like to try some demonstrations, we recommend the attendees have their own laptop ready and follow the installation procedures provided in https://pypi.org/project/raven-framework/ before the workshop.

Detailed agenda will be provided as we come closer to the workshop date.