Author: @Tom Thunell | [email protected]
Date Published: October 6, 2022
Modeling energy production is a core task for those developing, financing, and operating renewable energy projects. However, with energy storage, since no energy is being produced, the main consideration to model is when should the battery charge and discharge? To answer this we need to define what we mean by “should” and come up with a way to represent the physical attributes of energy storage and the commercial/market rules for how they are allowed to operate. This type of problem lends itself to a method called optimization. We define an objective (the “should”), apply constraints (which captures the physical and commercial/market attributes), and determine the decisions that lead to the best outcome. This blog post describes the intuition behind optimization and how applying different constraints and optimization approaches impacts results.
Summary Contents
Optimization is the process of maximizing or minimizing an objective function by finding the best available decisions across a set of inputs. When applied to energy storage modeling, the objective function is to maximize revenue* given a set of constraints.
Those constraints are driven by both physical battery and electrical properties (ex: limited amount of charge, or energy capacity) and market-based (ex: ISO requirements for certain reserved charge to serve a previous commitment). To understand how to maximize revenue, we also need to have some price expectation. For the rest of this blog, we consider price to be a fixed input (see this blog on price uncertainty) and focus on the decision of when to discharge and charge to make the most revenue, subject to constraints. The charge/discharge is therefore the quantity of the Price x Quantity that makes up our revenue calculation that we are maximizing. Finally, because charging requires buying energy from the grid (or charging off of a co-located resource) and later supplying energy, the optimization captures a simple economic approach: buying energy at the lowest cost and selling it at the highest possible value.
All that said, optimization can be complicated and computationally intensive. Therefore, it is useful to have a rough approximation of how an optimization process works. For energy storage, a common approximation is the TxBx method. With this, modelers take the sum of the top X priced hours in a day and subtract the sum of the bottom X priced hours in a day to estimate the revenue you could generate from a storage project with X hours of duration. The benefit of this method is that it captures the simple economic approach of a battery: buying low and selling high. It also replicates some of the core constraints, namely the limited amount of charge capacity. Finally, it is simple to replicate in Excel, and can serve as an effective indicator of energy price volatility and a high potential energy storage opportunity.
Of course, in the field, the physical and market constraints are not optional, and need to be considered when modeling how a battery will operate. A true optimization, such as Tyba’s approach, ends up being required to align modeled battery behavior with realistic battery operations. Nevertheless, the TxBx method has useful explanatory power. Thus, we wanted to outline the broader set of relevant constraints for batteries and to what extent the TxBx method captures them — and the table below includes a (non-exhaustive) set of common physical and commercial constraints you may want to consider.
| Constraint | Category | TxBx Approach | Tyba Approach |
|---|---|---|---|
| Size | Physical | Yes | Yes |
| Duration | Physical | Yes | Yes |
| Cycling Limits | Physical | Limited | Yes |
| Degradation | Physical | Limited | Yes |
| State of Charge (SOC) Compliance | Physical | No | Yes |
| Single Energy Market | Commercial | Yes | Yes |
| Multiple Energy Markets | Commercial | Limited | Yes |
| Energy & Ancillaries | Commercial | Limited | Yes |
| Hurdle Rates | Commercial | Limited | Yes |
| Day-Ahead Commitment Constraints | Commercial | No | Yes |
| ISO Rules | Commercial | No | Yes |
| Price Uncertainty | Commercial | No | Yes |
| Ancillary Utilization Uncertainty | Commercial | No | Yes |
| Bid/Clear (SCED) Logic | Commercial | No | Yes |
| Virtual Trading | Commercial | No | Yes |
We want to understand how constraints impact optimization results (planned operations and revenue) and to ensure that model assumptions accurately reflect our distinct point of view on the market. To help illustrate this, we modeled a standalone storage project ERCOT over the course of a reference year under both the TxBx** and Tyba optimization approaches and varied the constraint assumptions around the:
Hurdle rate, which is the delta between buying and selling energy required to be active with storage resource
Market participation strategy
Utilization uncertainty associated with providing RRS
What is utilization uncertainty?
For ancillary service products like RRS, the amount the ISO will call on a resource to provide the service during an interval is typically not known. This leads to uncertainty in the exact amount of energy that is needed to deliver the service. To address this, Tyba allows users to set utilization bands that ensure you have a minimum of X MWh and maximum of Y MWh of ability to discharge / stop charge (for an up service) or charge / stop discharge (for a down service).
Summary Results