Perfect Foresight Optimization

From EPRI Storage Wiki
< DER VET User Guide‎ | Model Details
Revision as of 16:41, 31 March 2021 by Mevans (talk | contribs) (1 revision imported)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

DER-VET's core is an optimization problem that simultaneously solves for all optimization variables present in the problem. In the case where any technology is being optimally sized by DER-VET, the DERs' size parameters are treated as optimization variables in the optimization problem that are solved concurrently with the dispatch of all DERs. If no technologies are being optimally sized, the only optimization variables are the dispatch (service participation) of each DER within the optimization window.

Because the optimization problem solves for the optimal dispatch of all DERs within an optimization window concurrently, it is considered a "perfect foresight" optimization. The optimization solver has access to errorless information about the whole optimization window when it decides the optimal dispatch for the whole optimization window. This is a useful approach in many cases because the results are clearly known as the upper bound on value possible given the circumstances. They do not incorporate sources of error that might be different between scenarios, such as controller performance, enabling apples-to-apples comparisons between scenarios.

Energy storage systems and similar technologies have a state of energy [kWh], also expressed as state of charge [%] that couples every time step in an optimization window. Operating a storage system in time step t impacts the ability to operate the storage system at time t+1 and vice versa. When a storage system participates in a service, it could be precluding itself from participating in the future based on its limited energy capacity. As such, knowledge of future conditions (market prices, customer load, solar generation, etc.) is very relevant for realizing the full value of a storage system. In the real world, different controllers will perform differently under different conditions. It is often useful to ignore this inefficiency in the planning phase of a project in favor of maximizing the potential of the project.

This perfect foresight approach means that the default result is an upper limit on the value a DER mix could achieve under those conditions. There are built-in methods for users to address this assumption and more closely bound the value of DERs in their applications (see below).

Optimization vs Evaluation Inputs

DER-VET has the ability to evaluate the financial performance of DERs using different data than was used in the optimization problem. Using this feature breaks the perfect foresight assumption. By solving the optimization using imperfect inputs and then calculating the financial value of the DERs with the reference ("errorless") inputs, the DERs will be operated sub-optimally.

Not all inputs are compatible with separate optimization and evaluation values. Any input that is central to the optimization (e.g. time step) cannot be changed, but prices, financial inputs, and anything that is used in post-optimization calculations can be changed.

Persistence Method

A common method for setting a lower bound on the value of a resource providing wholesale energy is called the persistence method. In this method, historical data is used as an imperfect representation of the present. This could be as simple as using yesterday's energy prices to self-schedule charge/discharge/generation for today. More complexity can be layered in by using data from the last weekday if it's currently a weekday or weekend if it's currently a weekend, considering holidays, considering only the same day of the week, etc.

In DER-VET, this can be implemented by entering the "real" energy prices as the "evaluation" data and entering a shifted timeseries data file as teh "optimization" data