PSR
SDDP Platform
SDDP Platform · Expansion planning

OptGen

Generation and transmission expansion planning under uncertainty

An expansion plan has to work across many scenarios of hydrology, renewables and demand. Many planning tools reduce the number of scenarios or the time resolution to keep the problem solvable. OptGen decides which projects to build, where and when, at minimum total cost, integrated with SDDP. Its decomposition algorithms scale with high-performance computing, so studies can keep many scenarios and a detailed time resolution.

01Methodology

Two strategies on one platform

OptGen decides which technologies to build, where and when, at minimum total cost. It offers two strategies, both running on the SDDP database.

  • OptGen 1, Benders decomposition. An investment module decides the plan, SDDP evaluates its stochastic operation and returns Benders cuts, and the two iterate until the lower and upper bounds converge. Suited to long horizons with many scenarios and stages.
  • OptGen 2, co-optimization. Investment and operation are solved together in a single problem, with hourly resolution on typical days and a rolling horizon, year by year. Suited to hourly variability, intra-day storage and integrated generation and transmission studies.
OptGen 1 Benders decomposition OptGen 2 Co-optimization Investment module Which, where and when Operational module SDDP stochastic dispatch Plan Cuts Iterates until the bounds converge One problem, solved jointly Investment Which, where and when Operation Hourly, on typical days Rolling horizon, year by year
The two OptGen strategies.
02New in SDDP 19

Data centers and large loads

Large loads such as data centers are now decision variables. OptGen decides whether, where and when they are built, and treats them in the same way as renewable candidates.

Investment cost

The construction cost in million USD, as for any expansion candidate.

Demand profile

The hourly profile of the load and its flexibility window.

Willingness to pay

A price in USD/MWh that represents the value of compute for each MWh consumed.

03New in SDDP 19

Battery and solar degradation

SDDP usually handles this kind of data per calendar year, but degradation depends on the entry date of each project, and in OptGen that date is a decision. A dedicated screen now relates degradation to the entry date OptGen chooses, so it does not bias the investment decisions.

End of warranty 50% 70% 90% 100% 2026 2030 2034 2038 2042 2046 Battery capacity Built in 2028 Built in 2035
Illustration. Battery capacity fades from the year each project enters, and OptGen chooses that year.
100% 95% 90% 85% 0 5 10 15 20 25 Years of operation Solar output Initial drop in the first year Gradual decline
Illustration. Solar output drops in the first year and then declines gradually.
04New in SDDP 19

Automatic firm capacity of renewables

An expansion plan has to meet a firm capacity requirement: enough capacity that the system can count on in the hours of highest need. For wind and solar, that amount depends on the mix being built.

The critical hours move with the mix

Thermal and hydro plants count for most of their installed capacity. Wind and solar count only for what they produce in the hours of highest net demand, that is, demand minus renewable generation.

Those hours change as the mix changes. With little solar, net demand peaks in the early evening, while solar plants still produce. As more solar is built, net demand falls at midday and its peak moves past sunset, and the firm capacity of solar falls.

0 50 100 0 h 6 12 18 24 h Net demand 19 h 20 h 21 h Critical hour moves later 2026, little solar 2032 2040, much solar
Illustration. As solar grows, the hour of highest net demand moves later in the day.

Calculated by OptGen, year by year

A fixed value, entered once, cannot follow this change. Kept high, it lets the plan count on capacity that will not be there at the peak. Updated by hand, it has to be recalculated for every year and every mix, while the mix is what OptGen is deciding.

OptGen now calculates it year by year: it finds the hours of highest net demand of the previous year and averages the available generation of each renewable candidate over those hours. The firm capacity follows the system and the renewable mix being built.

The calculation is already used in planning studies and becomes a standard OptGen feature in SDDP 19.

0% 20% 40% Firm capacity of solar (share of nameplate) 2026 2028 2030 2032 2034 2036 2038 Solar output matches the critical hour Critical hour after sunset
Illustration. The firm capacity of solar falls once the critical hour moves past sunset.
05New in SDDP 19

Resilience days

OptGen 2 represents the hourly operation with typical days. A typical day is an average of many days and scenarios: it keeps the usual shape of the day, with the valley at midday when the sun is high.

The average also hides the unusual days. On a gloomy day, with little sun, net demand stays high from morning to night, and a system sized for the typical day may not cope with it.

Resilience days add real days to the OptGen 2 expansion problem. Each one is identified by its date and scenario, keeps all its hours, and its operation must be feasible with the planned system.

0 h 6 12 18 24 h Net demand over one day Gloomy day, little sun: the average leaves it out Real days Typical day (average) Gloomy day
Illustration. The typical day is the average of many days, and it hides a gloomy day with little sun.
06Performance

Faster convergence

Integrated generation and transmission studies with a detailed network are the hardest cases for the decomposition. SDDP 19 helps them converge.

Regularization and relaxed Benders cuts

A regularization scheme stabilizes the investment problem, and relaxed cuts, computed without the second Kirchhoff law, give it early guidance. Exact cuts then sharpen the bound. On hard cases with a detailed network, this can be the difference between converging and not.

Automatic precedence constraints for transmission projects also help these studies converge.

Investment layer Operational layer Regularized investment problem Investment plan and cost Original investment problem Lower bound Relaxed operational problem No second Kirchhoff law Exact operational problem Exact cuts, upper bound Relaxed cuts Plan and cost Exact cuts Lower bound Upper bound
The regularized and relaxed Benders scheme.
07New in SDDP 19

Transmission expansion with OptNet

OptGen already plans transmission together with generation. OptNet, its new network module, refines the transmission plan against every dispatch scenario computed by SDDP, so the plan does not rely on a few snapshots of the system.

Explore OptNet
08Outputs

Results

A new dashboard

A more flexible dashboard, integrated with the Results tab, shows the operational dispatch behind each expansion decision.

OptGen 2 results in the Results tab

A new Results module loads, plots and navigates every opt2_*.csv result of the OptGen 2 operational module, which used to need custom scripts. The SDDP results workflow now covers OptGen 2.

SDDP 19 · OptGen

OptGen is part of SDDP 19

Install SDDP 19 to run OptGen, or talk to our team about your study.