To address the problems at hand, we are going to start with Theory-Trained Neural networks (TTNs). Shallow theory-trained neural networks have so far been successfully used for learning the solution of highly coupled differential equations in small systems. This project is to scale up both TTNs and Physics-Informed Machine Learning methods for simulating wind dynamics and to build a real-time forecasting platform to optimize wind power generation. Furthermore, we develop even more efficient algorithms by employing theory-trained quantum neural networks, instead of or in addition to the classical neural network. The algorithms we will be employing includes convolutional neural network, Long-Short-Term-Memory network, Variational Quantum Eigensolver, Quantum Monte Carlo, Quantum Approximate Optimization Algorithm. Applying such techniques to large data sets, training deep neural networks, and computational fluid dynamics simulations are computationally heavy and can only be run on HPC in a timely manner.

Industry Partner(s):ForeQast Technologies Limited

Academic Institution:University of Waterloo

Academic Researcher: Achim Kempf

Focus Areas: Advanced Manufacturing, Clean Tech, Quantum

Platforms: Cloud, GPU, Parallel CPU