Hardware-in-the-loop validation of a real-time receding-horizon enhanced Dung beetle optimization algorithm for solar-connected smart-grid energy management

Abstract

The Enhanced Dung Beetle Optimization (EDBO) algorithm, which augments the Dung Beetle Optimizer with the Harris Hawks Optimizer and the Nelder–Mead simplex, has demonstrated strong performance for offline energy-management optimization in solar-connected residential micro-grids. However, its practical deployment ability under real-time, embedded and communication-constrained conditions has not been established, because the offline formulation solves a single 168-variable weekly problem with full future knowledge and a very large computational budget. This paper reformulates the EDBO-based energy management system (EMS) as a real-time receding-horizon controller and validates it on a Hardware-in-the-Loop (HIL) platform in which the controller and a real-time smart-grid plant model exchange data over a real Modbus TCP link. The controller re-optimizes a short look-ahead window each hour from the live battery state of charge, reducing the per-cycle dimensionality from 168 to 48 variables. Across a six-scenario matrix, the controller met the real-time deadline on all 1,008 control cycles with sub-millisecond communication latency, reproduced the published no-storage baseline exactly (40.012 USD/week), achieved a 6.36% weekly cost reduction on the nominal scenario — within 1.3% of the previously published offline optimum (36.981 USD/week) — and remained stable and robust under price spikes, load steps, 20% forecast error and 15% communication packet loss. The results indicate that the receding-horizon reformulation recovers essentially all of the offline economic benefit while enabling real-time, embedded, communication-constrained deployment.