A multi-objective spatio-temporal pricing method for fast-charging stations oriented to transformer load balancing

LiangSong Guo, Ming Jin, Bin Jing, Leilei Lv, Min Guo, Hao Ding, JiaBin Zhang, Jianping WangFrontiers in Energy Research 2024 · DOI

TL;DR

This paper proposes a dynamic, "spatio-temporal" pricing strategy for electric vehicle (EV) fast-charging stations to prevent power grid imbalances. By adjusting prices based on both time and location, the system nudges EV drivers toward underutilized charging stations, successfully balancing transformer loads and stabilizing grid voltage.

What Problem Does This Paper Address?

As electric vehicles become mainstream, their demand for "fast-charging" creates a massive headache for grid operators. Unlike slow home charging, fast chargers pull enormous amounts of power in short bursts. If too many EVs congregate at a single charging hub—perhaps because it’s near a popular highway or commercial center—the local distribution feeder can become overloaded, leading to transformer strain and dangerous voltage drops.

Current solutions often rely on Time-of-Use (TOU) pricing, which tells drivers when to charge but not where. This paper addresses the "spatial" gap: how to use price signals to distribute EV charging loads across a geographic area to ensure the grid remains stable and efficient without requiring expensive physical infrastructure upgrades.

Methods

The authors developed a two-layer optimization framework that couples a traffic system model with a power system model:

  1. The EV Behavioral Model (Traffic Layer): To predict where EVs will go, the authors used Monte Carlo simulations and Dijkstra’s algorithm to model travel paths. Crucially, they included a "preference model" that weighs two factors: travel time and charging cost. This simulates how a driver might choose a slightly further station if the price is low enough.
  2. The Multi-Objective Pricing Model (Grid Layer): The grid side uses the DistFlow model (a standard for radial distribution networks) to calculate power flow. The objective is two-fold: minimize the spatio-temporal imbalance of voltage and minimize the imbalance of the load.
  3. The Solver: Because human behavior and grid physics create a non-linear, complex problem, the authors employed an Improved Strength Pareto Evolutionary Algorithm (SPEA2). This heuristic approach finds a "Pareto front"—a set of optimal solutions where you can't improve voltage stability without slightly compromising load balance, and vice versa—allowing operators to choose the best compromise.

Datasets

The methodology was validated using two distinct simulation environments:

  • Case 1 (Small Scale): A self-built 29-node road network paired with a 9-node distribution power network.
  • Case 2 (Large Scale): A 66-node road network and a 33-node distribution network based on real-world topology from the Xinjiang region in China.
  • Traffic Data: The models incorporated statistical data from the 2011 Beijing Transportation Development Annual Report to ground the EV travel patterns in reality.

Key Insights and Ideas

  • Price as a "Nudge": The paper demonstrates that EV drivers are "elastic" responders. By creating a price differential between stations, the grid can effectively "steer" traffic.
  • Spatial vs. Temporal Flexibility: Most research focuses on shifting load to nighttime (temporal). This paper highlights that moving load to a different neighborhood (spatial) is equally vital for preventing local transformer failure.
  • Superiority over DLMP: The proposed method outperformed standard Distribution Locational Marginal Pricing (DLMP). While DLMP is good at reflecting marginal costs, it often fails to promote regional load balance as effectively as a multi-objective spatio-temporal approach.
  • The "Aha" Moment: The results showed that the benefits of this pricing strategy actually increase as the network scale grows. In the larger Xinjiang case, the system was significantly more effective at smoothing out "load spikes" than in the smaller test case.

Key Results

  • Load Balancing: In Case 2, the proposed method reduced the power variance (a measure of load imbalance) from approximately 69.8 million kW² (under standard DLMP) to 62.5 million kW².
  • Voltage Stability: The voltage variance was reduced from 0.059 kV² to 0.046 kV², indicating a much steadier and safer operating environment for the grid.
  • Peak Shaving: The heat maps generated by the study (Figures 8 & 9) visually confirm that the spatio-temporal method successfully shifted charging peaks away from congested nodes and high-demand time slots compared to uncontrolled charging.

Limitations and Future Work

While the results are promising, the authors note a few areas for improvement:

  • Computational Intensity: As the system scale increases, the simulation time for the spatio-temporal distribution model grows. Future work is needed to simplify these models for real-time application.
  • Human Stochasticity: While Monte Carlo simulations are robust, real-world human behavior is even more unpredictable. The authors suggest a "data-driven" approach (likely involving machine learning) could better capture the nuances of driver choices.
  • Integration of Renewables: The current model focuses on the grid and EVs. Future iterations could incorporate distributed energy resources (like solar panels and batteries located at the charging stations) to further buffer the grid.

Summary updated

Abstract

To address the challenges posed by the fast-charging demand of electric vehicles, causing feeder load and voltage imbalances during operation, this paper introduces a spatio-temporal pricing strategy tailored to enhance feeder operation equilibrium. This approach facilitates the spatio-temporal guidance of fast-charging loads for electric vehicles in operation. This paper begins by formulating a spatio-temporal distribution model for electric vehicle fast-charging loads, considering owners’ preferences. It further develops a behavioral model for the travel choices of electric vehicles, illustrating the impact of spatio-temporal electricity pricing at fast-charging stations on load distribution. Next, it proposes a multi-objective spatio-temporal pricing model and its solution method specifically designed for feeder-balance-oriented fast-charging stations. This model targets the minimization of the spatio-temporal imbalance in feeder voltage and load. It takes a comprehensive approach, considering the constraints of the spatio-temporal load distribution model and optimal power flow model. The resulting spatio-temporal pricing model for fast-charging stations is effectively solved using the extended Pareto evolutionary algorithm. To validate the effectiveness of the proposed method in achieving feeder balancing, this paper analyzes two examples: a self-built 29-node road network and a 9-node distribution network, as well as a 66-node road network and a 33-node distribution network in the Xinjiang region. The results show that the proposed method can effectively guide the charging of electric vehicles and make the load distribution more balanced.
BibTeX
@inproceedings{guo2024multi,
  author = {LiangSong Guo and Ming Jin and Bin Jing and Leilei Lv and Min Guo and Hao Ding and JiaBin Zhang and Jianping Wang},
  title = {{A multi-objective spatio-temporal pricing method for fast-charging stations oriented to transformer load balancing}},
  booktitle = {Frontiers in Energy Research},
  year = {2024},
  doi = {10.3389/fenrg.2024.1362343},
}
Figure