Classical-Equivalent Bayesian Portfolio Optimization for Electricity Generation Planning.

Area of Science:

• Energy economics
• Optimization under uncertainty
• Computational finance

Background:

• Electricity generation planning relies on cost estimations, which are subject to uncertainty.
• Traditional mean-variance optimization assumes known cost parameters, leading to non-robust resource allocation.
• Expert prior information is valuable but often not integrated into planning models.

Purpose of the Study:

• To develop a robust method for electricity generation planning that accounts for parameter uncertainty.
• To integrate expert prior knowledge into resource allocation decisions.
• To compare a novel Bayesian approach with classical methods.

Main Methods:

• Introduced classical-equivalent Bayesian mean-variance optimization.
• Utilized both improper and proper prior distributions for uncertain cost parameters.
• Applied the method to electricity generation planning.

Main Results:

• The Bayesian approach provides more robust resource allocation compared to naive mean-variance optimization.
• Demonstrated the effectiveness of incorporating prior information and parameter uncertainty.
• Showcased the application through a comparative analysis of optimal portfolios.

Conclusions:

• Bayesian mean-variance optimization offers a superior framework for electricity generation planning.
• Accounting for parameter uncertainty and expert knowledge enhances the reliability of energy strategies.
• The proposed method provides a more realistic and robust solution for energy resource allocation.