Description
Project: Climate ToolSuite
Technical Leads: Osten Anderson, Kostas Oikonomou
Objective/Summary
This tool provides functionality for sizing energy storage based on the principle of increasing reliability through the reduction of unserved energy. The tool optimizes the technology, capacity, and duration of energy storage investment alongside dispatch. It relies on operational characteristics of a specified generation portfolio, obtainable as outputs from a production cost model. Energy storage dispatch is modeled against time series of unserved energy, curtailment, and energy prices. By isolating storage sizing from generation expansion, the tool is more computationally efficient than traditional methods, and suitable for considering a wide candidate-space of storage technologies at discrete durations.
Methods
Overview
The primary objective of this tool is to determine the optimal capacity, duration, and technology mix of energy storage needed to ensure reliable grid operation. The proposed method formulates this problem as an LP optimization problem that simulates energy storage dispatch and investment, with the objective of reducing unserved energy. Example operational inputs to the model and energy storage outputs of the model are provided. The tool relies on cost and performance characteristics of various energy storage technologies from the Pacific Northwest National Laboratory’s Energy Storage Grand Challenge database. The model is developed as a linear problem in the Pyomo package, solvable using commercial or open-source optimization solvers.
Data and Study Design
The proposed method is intended to determine the optimal sizing of storage independently, considering the generation fleet as an exogenous feature. Accordingly, the need for energy storage is evaluated relative to the performance of that generation fleet. Energy storage improves the overall performance of the power system by shifting energy between periods of abundance to periods of shortage. Operational performance characteristics are quantified by the amount of unserved energy and the amount of renewable curtailment, when evaluated without storage. Such operational metrics can be obtained from the outputs of production cost models (PCM), although the approach developed here is generic to the source the inputs.
The operational features of interest include hourly time series of renewable curtailment, unserved energy, and locational marginal prices (LMPs). Curtailment and unserved energy respectively represent abundance and shortage, which manifest as energy imbalances that energy storage can mitigate. LMPs represent the marginal cost of energy in a given area, accounting for generation costs, transmission congestion, and losses. Low LMPs typically indicate energy abundance or limited local demand, while high LMPs signal scarcity or network constraints. This allows each area to model charging behavior using a surrogate for imports, without the need for computationally expensive modeling of the transmission system.
The proposed storage sizing model is developed over a time horizon defined by a set of time points t∈T, due to the strong temporal dependence of storage. Each time point is associated with a duration d (in hours). Although the formulation is general, it is recommended that T span a full year at hourly resolution (i.e., d=1h). This resolution provides a multi-timescale perspective, with granular hourly resolution to capture short-duration storage behavior and a full year to capture long-duration storage behavior. These temporal characteristics are defined by the horizon and resolution in the input operational features. Storage characteristics are captured by technical parameters, including rated duration, charge and discharge efficiency, and associated costs.
Documentation and demonstration: manuscript submitted
Software: https://github.com/pnnl-int/StorageSizing