Prediction errors of different models.

Microgrid systems incorporating carbon trading mechanisms and demand response (DR) demonstrate significant potential for facilitating low-carbon societies and advancing sustainable energy development. The optimal operation of microgrid systems faces challenges due to: (1) response rate disparities among cooling, heating, and power equipment, (2) load prediction inaccuracies, and (3) complex interdependencies in multi-energy device coupling. To address these challenges, we propose a two-layer rolling optimization framework with multi-time scale scheduling for CCHP microgrid systems. First, wind and photovoltaic power generation are predicted using a CNN-ATT-BiLSTM model, with comparative analysis against standalone CNN, BiLSTM and CNN-LSTM models. Second, we establish a multi-time scale optimization model for CCHP-MG systems, with operating cost minimization as the objective function. Finally, we evaluate four operational scenarios incorporating DR and carbon trading mechanisms, with comparative cost analysis. Case study results demonstrate that the proposed model simultaneously satisfies cooling/heating/power demand while mitigating stochastic supply-demand fluctuations through multi-temporal resolution coordination.