With support from the Quantum Computer User Program (QCUP) at the Oak Ridge Leadership Computing Facility (OLCF), researchers from Cornell University have developed a new quantum computer–based artificial intelligence (AI) system for identifying and diagnosing faults in electrical power grids. The framework promises much faster response times and smarter solutions than current state-of-the-art systems and hints at the impact that quantum computers may have on daily life once the technology matures.
Fengqi You, the Roxanne E. and Michael J. Zak Professor in Energy Systems Engineering in Cornell’s College of Engineering, along with doctoral student Akshay Ajagekar, co-authored “Quantum Computing-based Hybrid Deep Learning for Fault Diagnosis in Electrical Power Systems,” published in Applied Energy on December 1. Their paper lays out a proposed AI system that combines the feature extraction capabilities of a deep learning method, called the conditional restricted Boltzmann machine, with a classifying artificial neural network.
Deep learning is a subfield of machine learning that enables computers to learn patterns from huge amounts of data and make decisions based on those patterns. Deep learning is enabled by artificial neural networks—algorithms that imitate human neurons but can operate tirelessly on vast amounts of data. However, designing and training effective neural networks can take even the most talented coders up to a year or more. Setting the Cornell team’s proposed system apart from current fault-detection systems that use neural networks is an accelerated training process that leverages both quantum computers and classical computers to achieve faster, smarter results.
“Quantum computers can help improve certain aspects of the training methodology by providing better gradient estimates compared to their classical counterparts, thus improving convergence and resulting in better generalization,” You said. “Apart from the improvement in training techniques, the proposed fault diagnosis framework also demonstrates faster response times than popular fault detection methods.”
You and Ajagekar tested their proposed system on a quantum computer from D-Wave Systems. Using quantum bits (qubits) to encode data as 1s, 0s, or both, quantum computers have the potential to solve certain problems much faster than classical computers—the bits of which only use 1s and 0s. The team’s largest case study on the D-Wave computer consisted of 30 buses of electrical power systems. This proof-of-concept experiment demonstrated the efficacy of the system’s practical application, thereby pointing the way for further study as quantum computers continue to be developed.
“The proposed fault diagnosis models and methods are scalable in theory. However, a larger number of qubits than what we used in the D-Wave quantum device is required to exhibit computational improvements associated with quantum computer–based techniques applied to real-world electrical power systems,” You said.
Access to D-Wave’s quantum computer was provided through QCUP, which is administered by the OLCF, a US Department of Energy Office of Science user facility at Oak Ridge National Laboratory (ORNL). QCUP awards time on multiple quantum computers owned by companies such as IBM, Rigetti Computing, and Honeywell to scientists with quantum-specific projects. The merit-based allocations are made in 6-month increments, with extensions predicated on the progress or results of each project. The 4-year-old program has supported over 100 users running more than 70 projects by providing compute time and technical and scientific support.
The science being conducted through the QCUP projects is providing valuable knowledge about the operation of quantum computers for open science, said Travis Humble, director of the Quantum Computing Institute at ORNL.
“As a facility, we’re trying to understand when quantum computing as a technology will be ready for primetime. At the moment, it’s very experimental with a focus on exploration and discovery. We established QCUP to monitor the technology and track the progress that the users are making,” said Humble, who also manages QCUP.
Another factor that makes QCUP unique is that it’s not tied to one particular system vendor—the program connects scientists with a variety of companies that are producing quantum computers.
“One of the things that differentiates QCUP is that we’re not advocating for a particular quantum computing system. We’re very agnostic to the individual devices,” Humble said. “We really just want people to do the best science.”
UT-Battelle LLC manages Oak Ridge National Laboratory for DOE’s Office of Science, the single largest supporter of basic research in the physical sciences in the United States. DOE’s Office of Science is working to address some of the most pressing challenges of our time. For more information, visit https://energy.gov/science.
