.png){kind=icon}
.png){kind=icon}
Return
Integrating learning and explicit model predictive control for unit commitment in microgrids
.png){kind=icon}
PRE
AI
DOI：10.1016/j.apenergy.2021.118026.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In this paper, we apply flexibility-based operational planning method to microgrid (MG) unit commitment (UC). The problem is formulated based on model predictive control (MPC) paradigm. Conventionally, such paradigm requires the online solution of a mixed-integer optimization problem, which faces difficulties in practical field implementations in a low-power computing microgrid controller. The explicit model predictive control (EMPC) can address this problem of MPC by computing the control laws of MPC in an explicit form offline to enable fast online computation. However, the complexity of the explicit control laws usually grows exponentially with the dimension of the problem, which hinders the application of EMPC in larger systems. In this paper, we integrate learning and EMPC to develop a computationally efficient and rigorous approach for implementing flexibility-based unit commitment paradigms in a microgrid controller with limited computational power. The computational complexity of the proposed approach can be adjusted to meet the hardware limitation of any given microgrid controller, while preserving as much as possible the optimality of the full-fidelity EMPC. This overcomes the drawbacks of traditional EMPC. Moreover, compared to the existing learning-based methods for accelerating optimization algorithms, the proposed approach is able to handle the variables and constraints of the original unit commitment problem systematically, which guarantees the feasibility of its output unit commitment schedules. We conduct case studies to demonstrate the effectiveness of the proposed approach.
Keywords:
Explicit model predictive control
Flexibility
Machine learning
Microgrid
Microgrid controller
Unit commitment