Research on microgrid optimal scheduling method based on adaptive stochastic model predictive control
Fang, Wensong (2026)
Kandidaatintyö
2026
School of Energy Systems, Sähkötekniikka
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe2026052755581
https://urn.fi/URN:NBN:fi-fe2026052755581
Tiivistelmä
As the penetration of renewable energy is increasing, microgrids have been an effective solution for accommodating and utilizing distributed energy resources. However, renewable power generation and load demand sometimes can be vulnerable, since it can be influenced highly by weather conditions, environmental factors, and human activities, exhibiting strong uncertainty, volatility, and temporal correlation. Traditional fixed-period or deterministic scheduling methods are often incapable of accurately capturing short-term time-series variations of sources and loads, which may lead to power imbalance and reduced operational economy of microgrid systems.
To mitigate the existing issues, this thesis proposes an optimal microgrid scheduling method based on adaptive stochastic model predictive control (ASMPC). By constructing probabilistic models to describe the uncertainties of renewable generation and load demand, stochastic optimization is integrated into the model predictive control framework. An adaptive mechanism is further introduced to dynamically update scenario-based uncertainty representations and perform adaptive segmentation within a fixed prediction horizon according to real-time system operating conditions. The introduced method facilitates rolling optimization and feedback correction, thereby mitigating the impact of uncertainty disturbances effectively.
Simulation results from 5 repeated experiments show that all three optimized strategies maintain a 100% optimization success rate and near-unity renewable energy utilization. Compared with the non-optimized grid-only baseline, the optimized strategies reduce the mean operating cost by about 5.60% to 6.23%. Among the optimized methods, Adaptive SMPC does not achieve the minimum total cost, but it preserves economic competitiveness under uncertainty while reducing the upper-level decision dimensionality by 24.86% relative to Fixed-Horizon SMPC. Fixed-Horizon SMPC refers to the SMPC method using a fixed prediction horizon and uniform time intervals. The study therefore supports ASMPC as a feasible and computationally efficient scheduling framework for microgrids with high renewable energy penetration.
To mitigate the existing issues, this thesis proposes an optimal microgrid scheduling method based on adaptive stochastic model predictive control (ASMPC). By constructing probabilistic models to describe the uncertainties of renewable generation and load demand, stochastic optimization is integrated into the model predictive control framework. An adaptive mechanism is further introduced to dynamically update scenario-based uncertainty representations and perform adaptive segmentation within a fixed prediction horizon according to real-time system operating conditions. The introduced method facilitates rolling optimization and feedback correction, thereby mitigating the impact of uncertainty disturbances effectively.
Simulation results from 5 repeated experiments show that all three optimized strategies maintain a 100% optimization success rate and near-unity renewable energy utilization. Compared with the non-optimized grid-only baseline, the optimized strategies reduce the mean operating cost by about 5.60% to 6.23%. Among the optimized methods, Adaptive SMPC does not achieve the minimum total cost, but it preserves economic competitiveness under uncertainty while reducing the upper-level decision dimensionality by 24.86% relative to Fixed-Horizon SMPC. Fixed-Horizon SMPC refers to the SMPC method using a fixed prediction horizon and uniform time intervals. The study therefore supports ASMPC as a feasible and computationally efficient scheduling framework for microgrids with high renewable energy penetration.