Demand response in DC microgrids using particle swarm optimization-optimized model predictive control

Efficient energy management in DC microgrids requires advanced control strategies capable of handling renewable variability, storage limitations, and dynamic load demands. The proposed strategy employs a Particle Swarm Optimization (PSO)-tuned custom cost function within Model Predictive Control (MPC) to implement demand response. The controller tracks the uncurtailed demand reference outside scheduled curtailment windows and permits targeted demand reduction within them. The baseline demand profile is treated as the uncurtailed service reference, and demand response is implemented as a scheduled and controlled deviation from that reference rather than as continuous load minimization. The study also provides a structural comparison between the default fixed-weight MPC cost function and the PSO-tuned, mode-dependent custom cost function, interpreting their weights through the control priorities they encode among load tracking, grid use, and hybrid-storage operation. PSO is applied offline to determine two mode-dependent weight sets, while the online MPC uses these fixed weights to respond to changing load and renewable-generation conditions. Simulations under sunny and cloudy conditions show that the custom formulation preserves baseline demand tracking outside the demand-response windows and enables targeted curtailment within them. Quantitatively, the scheduled demand is reduced by 56.4– \(61.0\%\) during the morning curtailment periods and by \(58.4\%\) during the afternoon periods, while the tracking error remains below \(1.6\%\) outside the demand-response windows. Net grid-energy exchange decreases by \(64.3\%\) under sunny conditions and \(24.1\%\) under cloudy conditions, and net daily electricity cost decreases by \(49.1\%\) and \(29.5\%\) , respectively, while both storage states remain within their prescribed limits.