Skip to content
Log in
#graph neural networks Open access

An edge AI‐powered power converter hardware platform for practical end‐to‐end AI‐to‐converter closed‐loop deployment

Pingyang Sun Fan Zhang Zihang Qiu Junjie Chang Hanwen Zhang Gen Li
Aug 2026 · Energy Conversion and Economics · 21 references

Abstract

Abstract Modern power systems require increasingly flexible and responsive power conversion, driving the need for power converters with enhanced controllability and adaptability, where artificial intelligence (AI) plays a key role in enabling intelligent decision making, adaptive control and system level optimization. However, current AI applications in power systems remain largely restricted to cloud‐based post processing or device‐level demonstrations, lacking end‐to‐end hardware platforms that support the full pipeline from programming to deployment within a closed loop formed by physical converter hardware and AI modules. To address this gap, this paper proposes an edge AI‐powered power converter (EAI‐PC) hardware platform that utilizes locally deployed edge intelligence to establish bidirectional data exchange between converter low‐level control and EAI upper‐layer command dispatch. The platform features an end‐to‐end AI‐to‐converter closed‐loop deployment framework, enabling AI‐generated commands to be directly executed on physical converter hardware. By interconnecting converters (AC/DC and DC/DC) with an EAI module (using NVIDIA Jetson Orin Nano) via an industrial CAN protocol, this platform supports hardware‐level algorithm verification in the power system domain. The proposed EAI‐PC platform is experimentally evaluated across system configurations from single to multiple interconnected converters, assessing converter performance, command delivery and three AI deployment cases (two offline and one online): a multilayer perceptron (MLP) for power‐sharing prediction, a spatio‐temporal graph neural network (STGNN) for photovoltaic (PV) and load forecasting, and a bidirectional long short‐term memory (Bi‐LSTM) for online battery dispatch in load management and peak shaving.

DOI
View source

Similar papers

#graph neural networks Oct 2026

Locality-Preserving Graph Laplacian Manifold Learning Based Model Predictive Control for Three-Phase Inverters

This article presents a model predictive control (MPC) strategy for three-phase inverters based on locality preserving projections (LPPs). Unlike conventional machine learning–based MPC approaches that rely on predefined or high-dimensional input features, the proposed LPP-MPC automatically extracts compact, informative representations by preserving the data’s intrinsic geometric structure. This dimensionality reduction enables fast linear control-law evaluation with computational complexity O(1), making the controller well-suited for real-time implementation. Experimental results demonstrate that the LPP-MPC achieves lower total harmonic distortion (THD) and reduced tracking error compared to quadratic-programming MPC under both linear and nonlinear load conditions, and the proposed controller maintains consistently lower THD throughout load transients than other methods such as two-degree-of-freedom MPC. Compared to existing model-free MPC and deep learning neural network, the LPP-MPC has the lowest THD and root mean square error with the least computational time owing to its efficient linear structure and strong generalization capability.

Jianwu Zeng, Lizheng Cheng, V. Winstead et al. · 1 citation

Related blog posts