Temporal dimension - WITNESS: Difference between revisions
mNo edit summary |
No edit summary |
||
| (2 intermediate revisions by the same user not shown) | |||
| Line 5: | Line 5: | ||
= General principles = | = General principles = | ||
WITNESS temporal granularity is currently and typically 1 year steps. | WITNESS framework default temporal granularity is currently and typically 1 year steps. | ||
The underlying simulation mechanism could accommodate with smaller or bigger time steps (or adaptative with some developments to better capture shocks), but for models calibration purposes it was convenient to stick to 1 year steps as many data are available at a 1 year interval. | The underlying simulation mechanism could accommodate with smaller or bigger or variable time steps (or dynamically adaptative with some developments to better capture shocks by automatically switching to smaller granularity when high evolution gradients are detected), but for models calibration purposes it was convenient to stick to 1 year steps as many data are available at a 1 year interval. | ||
Default framework typically run from current year (default 2023 today) to 2100 (start and end date can also be changed). Stability of evolutions can easily be evaluated as access to gradients of all values are available, allowing to estimate if current timestep situation is stable of highly unstable / evolutive. | |||
All systems are at equilibrium at each steps, using an [https://gemseo.readthedocs.io/en/stable/mdo/mdo_formulations.html Multi-Disciplinary Feasible (MDF) formulation] (in short forcing the convergence of a fixed-point algorithm using optimization). | |||
= Optimization = | = Optimization = | ||
Optimization mechanisms typically will consider design variables as a vector of per | Optimization mechanisms typically will consider design variables as a vector of per-timestep values (not optimizing for a timestep and pushing to next, but globally optimising all timesteps simultaneously to avoid staying blocked in deadends), potentially driven by splines to reduce the actual # of control parameters and ensuring some continuity where needed. | ||
For example in the basic framework settings, which optimize investment available from economy in various energy generation and CCUS technologies on a yearly timestep from now to 2100, you have a matrix of investment per | For example in the basic framework settings, which optimize investment available from economy in various energy generation and CCUS technologies on a yearly timestep from now to 2100, you have a matrix of investment per year and per technology (80+) as design variables. | ||
Latest revision as of 16:02, 19 September 2024
Note: The documentation of WITNESS is 'in preparation' and is not yet 'published'!| Corresponding documentation | |
|---|---|
| Previous versions | |
| No previous version available | |
| Model information | |
| Model link | |
| Institution | Open-Source for Climate (OS-C), N/A, https://os-climate.org/transition-analysis/., Linux Foundation (LF), N/A, https://www.linuxfoundation.org/. |
| Solution concept | Systems dynamics based approach |
| Solution method | OptimizationSimulation-based optimization |
| Anticipation | |
General principles
WITNESS framework default temporal granularity is currently and typically 1 year steps.
The underlying simulation mechanism could accommodate with smaller or bigger or variable time steps (or dynamically adaptative with some developments to better capture shocks by automatically switching to smaller granularity when high evolution gradients are detected), but for models calibration purposes it was convenient to stick to 1 year steps as many data are available at a 1 year interval.
Default framework typically run from current year (default 2023 today) to 2100 (start and end date can also be changed). Stability of evolutions can easily be evaluated as access to gradients of all values are available, allowing to estimate if current timestep situation is stable of highly unstable / evolutive.
All systems are at equilibrium at each steps, using an Multi-Disciplinary Feasible (MDF) formulation (in short forcing the convergence of a fixed-point algorithm using optimization).
Optimization
Optimization mechanisms typically will consider design variables as a vector of per-timestep values (not optimizing for a timestep and pushing to next, but globally optimising all timesteps simultaneously to avoid staying blocked in deadends), potentially driven by splines to reduce the actual # of control parameters and ensuring some continuity where needed.
For example in the basic framework settings, which optimize investment available from economy in various energy generation and CCUS technologies on a yearly timestep from now to 2100, you have a matrix of investment per year and per technology (80+) as design variables.