Work

Model Predictive Ozone Dose Regulation in Surface Dielectric Barrier Discharge for Plasma Medicine Under Environmental Disturbances

Year: 2026
Type: article
Abstract: Non-thermal atmospheric plasma has attracted increasing interest in plasma medicine, where precise and reliable dose delivery remains a key challenge under variable operating environments. In surface dielectric barrier discharge (SDBD) systems, ozone is a practically measurable dose-related output that is strongly affected by excitation frequency and environmental temperature and humidity. This work proposes an intelligent control framework that separates offline modeling from embedded real-time control. A long short-term memory (LSTM) network was trained offline to capture ozone-generation dynamics under coupled temperature-humidity variations, and Shapley additive explanations (SHAP) analysis was used to examine consistency with expected physical trends, yielding an interpretable surrogate model for simulation-based benchmarking. For embedded implementation, a lightweight model predictive controller using a nonlinear autoregressive network with exogenous inputs (NARX-MPC) was developed to balance accuracy and computational cost. With the LSTM surrogate model as the plant model, NARX-MPC was benchmarked against a baseline proportional-integral-derivative (PID) controller over a broad range of coupled temperature-humidity disturbance scenarios. Experiments under representative operating conditions showed that NARX-MPC markedly improved ozone regulation versus PID, reducing overshoot by 82%, settling time by 68%, and mean absolute error by 25%. These results support the feasibility of ozone dose regulation in plasma medicine and provide a basis for precise and reliable dose delivery under therapeutically relevant environmental disturbances.
Language: English
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7204476258
FWCI:
Cites:
Cited by:
Related to: 10