Digital-twin-enabled design and machine-learning-assisted energy management of a solar refrigerated van with phase-change thermal storage.

Rising ambient temperatures and tightening decarbonisation targets are compelling the cold-chain sector to adopt self-sustained cooling units. This study develops a high-fidelity digital twin of an electric refrigerated van that integrates roof-mounted photovoltaic (PV) modules, a 10 kWh lithium-ion battery, and a -21 °C organic phase-change material (PCM) thermal buffer. The system is modelled in Modelica and dynamically driven by typical meteorological year data for Birmingham (UK). Simulation outputs are used to train Random-Forest regressors that predict battery state-of-charge (SoC) and compressor speed in real time. Results for a 12-h delivery mission show that adding 4 kg of PCM halves compressor cycling frequency and improves final SoC by 11 %, while solar charging recovers up to 18 % of daily refrigeration energy. The Random-Forest model attains an R² of 0.9990 for SoC and 0.9683 for rotor-speed prediction, enabling proactive energy management. The proposed architecture therefore offers a practical pathway to extend BEV range and resilience in food logistics.