Closed-Loop Predict-and-Optimize Framework

Update (December 2024)

We have published a more comprehensive work on this topic, including a bi-level programming-based approach and a literature review. If you have any questions, feel free to reach out!

• X. Chen, Y. Liu, and L. Wu, "Towards Improving Unit Commitment Economics: An Add-On Tailor for Renewable Energy and Reserve Predictions," IEEE Transactions on Sustainable Energy, vol. 15, no. 4, pp. 2547-2566, Oct. 2024.

Update (September 2024)

Some colleagues have pointed out that the code might produce NaN results. This occurs due to limitations in YALMIP's support for solver parameter tuning.

To fix this, please comment out the following lines in Step 01 and Step 02:

ops.gurobi.MIPGap    = Solver_gap/100;
ops.gurobi.TimeLimit = Solver_time*60;
ops.cplex.mip.tolerances.mipgap = Solver_gap/100;
ops.cplex.timelimit = Solver_time*60;
ops.mosek.MSK_DPAR_MIO_TOL_REL_GAP = Solver_gap/100;
ops.mosek.MSK_DPAR_MIO_MAX_TIME = Solver_time*60;

We have updated the code accordingly. Thank you for your feedback!

---

Dataset: Load, Renewables, and Feature System

Below are the datasets and codes accompanying the paper:

• X. Chen, Y. Yang, Y. Liu, and L. Wu, "Feature-Driven Economic Improvement for Network-Constrained Unit Commitment: A Closed-Loop Predict-and-Optimize Framework," IEEE Transactions on Power Systems, vol. 37, no. 4, pp. 3104–3118, July 2022.

If you find these resources helpful in your research, please cite our paper.

Contents

1. Load.xlsx
   Day-ahead forecasts and actual realizations of Belgian load from 2018/01/01 to 2020/12/31.

2. Solar_power_farm.xlsx
   Day-ahead forecasts and actual realizations of 13 solar farms in Belgium (2018/01/01 to 2020/12/31).

3. Wind_power_farm.xlsx
   Day-ahead forecasts and actual realizations of 7 wind farms in Belgium (2018/01/01 to 2020/12/31).

4. System_IEEE_24_bus.xlsx
   Configurations of the modified IEEE RTS 24-bus system.

5. System_ISO_5655_bus.xlsx
   Configurations of the 5655-bus system.

6. Feature.xlsx
   Well-collected feature vectors.

Note: The load and renewable energy data are collected at 15-minute intervals, hence the sub-hour labels.

---

Additional Files

• Map.pdf
  Illustrates modifications made to the IEEE RTS 24-bus system.

• OR_Presentation.pdf
  Our recent presentation on this work.

---

Code and Requirements

We have made our code open-access to foster further research. The provided MATLAB scripts solve the 24-bus case study and require:

• MATLAB
• GUROBI (as the solver)
• YALMIP (for calling GUROBI from MATLAB)

If you encounter any issues or would like the datasets and codes for larger systems, please feel free to contact me at [email protected]. It’s my pleasure to share and discuss these resources!

---

Thank you for your interest in our work!
If you use these datasets or the code in your research, kindly cite the relevant publications.