State of Charge Evolution Equations for Flywheels

Dariush Fooladivanda, Guillaume Mancini, Siddharth Garg, Catherine RosenbergarXiv 2014 · arXiv

Introduction

Flywheel energy storage systems (FESS) are increasingly vital for modern power grids, particularly for applications requiring rapid power injection and absorption, such as frequency regulation. Despite their operational advantages, accurately tracking the State of Charge (SoC) in these systems remains a challenge. Researchers at the University of Waterloo have developed a rigorous analytical framework to address this, moving beyond simplistic "bucket" models to capture the true physical dynamics of flywheels.

The Problem: Accuracy vs. Complexity

Existing FESS models typically fall into two extremes. Simple models often assume that energy change is merely the product of power and efficiency, neglecting "standing losses" like air friction (windage) and magnetic drag in bearings. Conversely, high-fidelity models rely on complex circuit simulations that are too computationally expensive for large-scale grid optimization.

The Waterloo team's research fills this gap by providing a middle-ground analytical model. It accounts for mechanical self-discharge and the inherent physical lag of electrical power controllers without the overhead of full-system simulation.

Analytical Framework: System Decomposition

The core of the proposed model lies in decomposing the FESS into two distinct but interacting subsystems.

Schematic of a Flywheel Energy Storage system showing the interaction between the flywheel (FW), motor/generator (M/G), and active power controller.
  1. The Mechanical Subsystem: This includes the flywheel and the induction machine rotor. The researchers observed that windage and eddy current losses are proportional to the square of the rotational speed ($\omega^2$). Since stored energy ($E$) is also proportional to $\omega^2$, these losses can be simplified into a first-order differential equation: $dE/dt = -E/T_{loss}$, where $T_{loss}$ is the energy decay constant.
  2. The Electrical Subsystem: This encompasses the active power controller and the induction machine. This subsystem is modeled as a first-order system with a time constant ($T_{cont}$), typically between 0.05 and 0.4 seconds, representing the response lag of the controller.

Key Insights: The Role of History and Lag

A significant finding of this work is that the SoC at any given moment is not solely a function of current power input; it is path-dependent. Because the electrical controller takes time to transition between power levels, the current energy state depends on the previous power state.

By applying Laplace transforms and convolution to these subsystems, the authors derived recursive formulas to calculate energy at discrete time steps. They also introduced an "Approximate Method" for scenarios where the loss time constant is significantly larger than the control time step—a common occurrence in grid applications. This approximation is backed by a formal proof of error bounding, ensuring reliability in practical use.

Conclusion and Future Directions

The derived evolution equations provide a computationally efficient yet physically grounded method for modeling flywheels in grid-level applications. By accounting for $T_{cont}$, the model avoids significant errors in high-frequency scenarios where power commands change rapidly.

While the current model assumes constant charging and discharging efficiencies, it sets the stage for future refinements. Subsequent work may involve validating these equations against experimental data from commercial deployments and exploring variable efficiency models based on rotational speed or thermal conditions.

Summary updated

BibTeX
@misc{fooladivanda2014state,
  author = {Dariush Fooladivanda and Guillaume Mancini and Siddharth Garg and Catherine Rosenberg},
  title = {{State of Charge Evolution Equations for Flywheels}},
  year = {2014},
  eprint = {1411.1680},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1411.1680},
}
Figure