The U.S. Energy Department’s research arm is leveraging machine learning technologies to simplify the design process for energy systems ranging from photovoltaics and wind turbines to aircraft engine compressors.

The Advanced Research Projects Agency-Energy, or ARPA-E, last month announced 23 research contracts totaling $15 million to incorporate machine learning into energy product designs. The first-phase contracts are part of an ARPA-E initiative dubbed DIFFERENTIATE, standing for—take a breath—Design Intelligence Fostering Formidable Energy Reduction (and) Enabling Novel Totally Impactful Advanced Technology Enhancements.

David Tew, an ARPA-E program director, said the two-year machine learning effort is focused on the engineering design process with the goal of optimizing power generation systems. Along with wind turbines and photovoltaics, DIFFERENTIATE also will focus on power conversion and heat transfer systems, aerodynamics, photonics and range of foundational energy technologies.

Detailed project descriptions are here.

At a fundamental level, the energy ML initiative targets math optimization problems common to energy and other system design processes. According to the program summary, energy researchers have “conceptualized machine learning and artificial intelligence-based solutions to help engineers execute and solve these problems in a manner that dramatically accelerates the pace of energy innovation.”

Besides applying machine learning techniques ranging from deep neural networks and “context-aware” learning to energy system architectures, the effort will utilize high-fidelity simulations to generate design details that can be used to improve components. Those computational fluid dynamic tools can be used both to predict system performance while generating costly training data needed to improve machine learning models.

“Our focus is really on algorithm development,” Tew said in an interview. Among the outcomes will be software packages, some open source, others proprietary. “We want the software to have a commercial impact,” Tew stressed.

Another component of the energy effort centers on inverse design, that is, using machine learning to arrive at an explicit definition of a desired function. That approach involves declaring upfront a specific functionality. Then, for example, machine learning techniques are used to identify the appropriate material for, say, a photovoltaic design.

The approach could then be used to come up with a complete design specification for a given component or system architecture. Ultimately, the application of machine learning to the inverse design approach could eliminate the cumbersome design iteration process, the energy agency noted.

Tew said the initial focus will be demonstrating machine learning capabilities in selected energy systems over the next nine months. At that point, ARPA-E expects to down-select the most promising machine learning projects with the goal of generating commercial software packages.

Other components include enhancing the conceptual design process and accelerating the application of high-fidelity analysis and optimization in the design of energy systems.

While DIFFERENTIATE spells out ambitious goals, Tew noted the energy agency has adopted a measured approach to applying machine learning to energy innovation. “We’re taking baby steps before we take big steps,” he said.