Comparison of Priority Service With Multilevel Demand Subscription

Gérard, Céline; Ávila, Daniel; Mou, Yuting; Papavasiliou, Anthony; Chevalier, P

doi:10.1109/tsg.2022.3153186

Cited by 2 publications

(1 citation statement)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…S TOCHASTIC dual dynamic programming (SDDP) is a popular technique for optimizing power system operations. Industrial applications include: long-term hydro-thermal planning [1]- [5], it application to power systems of various countries, e.g., Brazil [6] and Norway [7]; risk management in hydro-thermal scheduling [8]; emission constraints in hydrothermal planning [9]; short-term dispatch [10]; demand response [11]; optimal power flow [12]; gas storage valuation [13]; hydropower expansion planning [14]; and dairy farm operations [15]. While SDDP has exhibited superior performance in solving large-scale instances of difficult optimization problems, the algorithm often also fails to converge due to computational limitations, e.g., [16], despite being able to converge theoretically [17].…”

Section: Introductionmentioning

confidence: 99%

Batch Learning SDDP for Long-Term Hydrothermal Planning

Ávila

Papavasiliou

Löhndorf

2024

IEEE Trans. Power Syst.

View full text Add to dashboard Cite

We consider the stochastic dual dynamic programming (SDDP) algorithm -a widely employed algorithm applied to multistage stochastic programming -and propose a variant using experience replay -a batch learning technique from reinforcement learning. To connect SDDP with reinforcement learning, we cast SDDP as a Q-learning algorithm and describe its application in both risk-neutral and risk-averse settings. We demonstrate the superiority of the algorithm over conventional SDDP by benchmarking it against PSR's SDDP software using a large-scale instance of the long-term planning problem of interconnected hydropower plants in Colombia. We find that SDDP with batch learning is able to produce tighter optimality gaps in a shorter amount of time than conventional SDDP. We also find that batch learning improves the parallel efficiency of SDDP backward passes.

show abstract