Unlocking Innate Computing Abilities in Electric Grids

TL;DR

This paper proposes a novel approach to transform electric power grids into neural network-like computational systems by programming power electronic converters, enabling data processing without disrupting energy delivery.

Contribution

It introduces a method to harness innate computational abilities of electric grids by programming PECs to mimic neurons, merging energy and computing optimization.

Findings

Successfully encoded data into PEC control for computation

Electric grid operation remains unaffected during computation

Demonstrated affine transformation in a microgrid

Abstract

High energy consumption of artificial intelligence has gained momentum worldwide, which necessitates major investments on expanding efficient and carbon-neutral generation and data center infrastructure in electric power grids. Going beyond the conventional ideation, this article unleashes innate computational abilities in the power grid network circuits itself. By programming power electronic converters (PECs) to mimic biological neurons, we sustainably transform power grids into a neural network and enable it to optimize, compute and make data-driven decisions using distributed PECs. Instead of seen merely as an energy delivery platform, this article conceptualizes a novel application for electric grid to be used as a computing asset without affecting its operation. To illustrate its computational abilities, we solve a affine transformation task in a microgrid with five PECs. By encoding the digital data into the control of PECs, our preliminary results conclude that computing using electric grids does not disturb its operation. From a scientific perspective, this work fundamentally merges energy and computing optimization theories by harnessing inherent high-dimensional computational relationships in electric grids.