09h00 - 10h00
Multiscale stochastic co-optimization of reserves, storage, renewables, and flexibility resources in large scale systems
The very fast insertion of variable energy resources (VER) such as wind and solar in several countries has been essential for decarbonization. However, it has posed significant modeling and computational challenges for power system planning: (i) it becomes necessary to represent both the stochasticity and the hourly resolution of VER production; (ii) management of VER production variability requires new probabilistic methodologies for generation reserve; (iii) the short VER construction time and widespread geographical siting impacts transmission networks, which are better managed by flexible resources such as “smart wires” and storage instead of new transmission reinforcements; (iv) utility-scale storage solutions require the joint optimization of traditional hydro reservoirs, pumped storage and batteries.
This talk describes how to combine large-scale optimization techniques (multistage stochastic and robust optimization); data driven modeling (dynamic probabilistic reserve, Bayesian networks for producing VER and inflow scenarios); and massive cloud-based computing to produce expansion plans for large scale system such as Brazil (construction of 60 GW of wind, 30 GW of utility scale solar and 25 GW of distributed generation in the next ten years) and the US Pacific Northwest (solving 230 million large MIPs with 30 thousand CPUs).