Answers are grounded in this paper and cite their sources.
DC-link capacitors in industrial drive systems are subjected to significant harmonic currents, which increase thermal stress, accelerate ageing and reduce capacitor lifetime. In multi-drive industrial systems, the interaction of multiple converters connected to a common supply further complicates the harmonic behaviour, making accurate estimation of the DC-link capacitor current harmonics particularly challenging, while direct measurement requires a dedicated sensor inside the converter. This paper proposes a data-driven framework for estimating the dominant DC-link capacitor current harmonics in multi-drive industrial systems using a Mixture-of-Experts architecture, in which multilayer perceptron experts are combined by a Top-2 gating network with load-balancing regularisation. Voltage harmonics at the point of common coupling, together with grid-related and operating parameters, are used as input features to predict the amplitudes of the 6 th , 12 th and 18 th capacitor current harmonics. The model is trained and evaluated on a simulation dataset covering a wide range of background harmonic, grid impedance and loading conditions, and the simulation model is verified through laboratory measurements of the DC-link capacitor current and DC-link voltage. Over ten training runs, the proposed model achieves the highest mean prediction accuracy for all three harmonics compared with the benchmark models, including the TabTransformer. The framework enables the harmonic stress of DC-link capacitors to be estimated without a dedicated current sensor, supporting capacitor stress assessment, lifetime estimation and predictive maintenance in multi-drive industrial systems.
No takes yet. Share an insight, caveat, or question.
Jayathissa et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: