A forecasting-informed rolling demand response strategy for industrial microgrids

With the increasing penetration of renewable energy in industrial microgrids, enhancing the operational flexibility of industrial loads has become essential for improving renewable energy utilization and reducing electricity costs. This paper proposes a forecasting-informed rolling-horizon demand response framework for industrial integrated energy systems with photovoltaic (PV) generation. A bidirectional long short-term memory (BiLSTM) based PV forecasting model is first developed to predict short-term renewable generation. Subsequently, a mixed-integer optimization model incorporating industrial machine scheduling, material-flow coupling, and intermediate buffer dynamics is established to coordinate industrial production and renewable energy utilization under time-varying electricity prices. To improve adaptability to forecasting uncertainty, a rolling-horizon scheduling strategy is introduced to update operational decisions based on real-time PV forecasts continuously. Case studies demonstrate that the proposed BiLSTM model achieves high forecasting accuracy with an NMAE of 0.0387 and an NRMSE of 0.0891. Compared with conventional operation strategies, the proposed framework reduces electricity costs by approximately 17.8% while maintaining production feasibility. Under the rolling-horizon strategy, the total operating cost is reduced to 2190.7 Yuan, corresponding to a 16.0% cost reduction compared with the no-demand-response case. Sensitivity analysis further shows that even under 30% PV forecasting error, the operating cost has slight fluctuation around baseline, showing that the proposed method maintains stable economic performance. RT-LAB experiments verify the real-time feasibility and robustness of the proposed framework under renewable uncertainty.