Answers are grounded in this paper and cite their sources.
This paper presents an experimental case study of thermoelectric (Peltier) cooling in which the control reference is the dew point, computed in real time from ambient temperature and relative humidity, and the controller is a finite-memory discrete fractional-order PID (FOPID) tuned by multi-objective optimization. The FOPID parameters were selected from an NSGA-II Pareto front that trades off tracking error, electrical energy and control-signal variation. The resulting controller (OPT-FOPID) was compared on the same embedded platform with a PID and with a FOPID tuned by particle swarm optimization (PSO-FOPID), in five 300 s repetitions per controller, run in randomized order with a single firmware and calibration. The three controllers reached the dew point in 6–7 s and held it with an error dispersion of 0.03–0.04 ^∘ C; no statistically significant difference was detected among them in tracking error, settling time or energy per degree of cooling demand. The OPT-FOPID reduced the total variation of the PWM signal by 35–38% with respect to both baselines ( $$p<0.001$$ ), while the differences in peak electrical power were small (5–12%). In simulation, a multi-objective PID tuned with the same procedure reached a comparable trade-off, so the smoother actuation is attributed to the multi-objective tuning rather than to the fractional orders. A two-node physical model of the module, identified from a characterization test ( R²=0.992 ) and validated on independent trials, gives a model-based coefficient of performance of about 0.1. On this platform, the multi-objective design smoothed the actuation without a detectable loss of tracking accuracy, but did not reduce the energy consumption.
No takes yet. Share an insight, caveat, or question.
Pareja et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: