Optimal Day-Ahead Scheduling and Real-Time Control of Electric Vehicle Charging Stations for Dispatching Active Distribution Networks

Core Concepts

The proposed framework determines an optimal day-ahead power schedule (dispatch plan) at the grid connection point by leveraging the flexibility offered by electric vehicle charging stations and battery energy storage systems, while accounting for the uncertainties in load, photovoltaic generation, and electric vehicle charging demand.

Abstract

The paper presents a two-stage framework for dispatching active distribution networks (ADNs) using the flexibility from electric vehicle charging stations (EVCSs) and battery energy storage systems (BESSs).
Day-Ahead Stage:

A day-ahead dispatch plan is computed by solving a stochastic optimization problem that accounts for the uncertainties in load, photovoltaic (PV) generation, and EVCS demand.
The EVCS demand is forecasted using a Gaussian Mixture Model (GMM) that captures the multivariate stochastic behavior of EV arrival/departure times, battery capacities, and state-of-charge targets.
The grid constraints are modeled using a linearized optimal power flow (LOPF) approach, which ensures the feasibility of the dispatch plan with respect to voltage limits and branch power flow limits.
The objective is to minimize the difference between the dispatch plan and the actual power at the grid connection point, while also minimizing the reactive power flow.
Real-Time Stage (described in Part II):

A real-time model predictive control (RT-MPC) is used to track the day-ahead dispatch plan by controlling the EVCS and BESS flexibilities.
The RT-MPC accounts for the updated grid state and short-term forecasts of the uncontrollable injections to compensate for any deviations from the day-ahead plan.
The proposed framework is numerically validated on a real-life ADN at the EPFL campus, demonstrating the effectiveness of using EVCS flexibility in reducing the dispatch tracking error and the required BESS capacity.

Stats

The maximum absolute error (MAE) between the dispatch plan and the power at the grid connection point is reduced by 96% when using both BESS and EVCS control, compared to the case without any control.
The maximum power at the grid connection point (MPP) is reduced by 77% when using both BESS and EVCS control, compared to the case without any control.

Quotes

"The dispatch plan accounts for the uncertainties of vehicles connected to the EVCS along with other uncontrollable power injections, by day-ahead predicted scenarios."
"The framework ensures that the grid is operated within its voltage and branches power-flow operational bounds, modeled by a linearized optimal power-flow model, maintaining the tractability of the problem formulation."

Key Insights Distilled From

Grid-aware Scheduling and Control of Electric Vehicle Charging Stations for Dispatching Active Distribution Networks. Part-I: Day-ahead and Numerical Validation

by Rahul K. Gup... at arxiv.org 04-22-2024

https://arxiv.org/pdf/2404.12863.pdf

Deeper Inquiries

How can the proposed framework be extended to consider the impact of vehicle-to-grid (V2G) capabilities on the dispatch plan and real-time control

To incorporate vehicle-to-grid (V2G) capabilities into the proposed framework, the control strategy would need to be adapted to allow for bidirectional power flow from the electric vehicles back to the grid. This would involve modifying the optimization algorithms to consider the potential for EVs to discharge power during peak demand periods or provide ancillary services to the grid.
In the day-ahead scheduling, the forecasting models would need to account for V2G capabilities, predicting not only the charging behavior of EVs but also their potential to discharge power. This would require additional data on the availability of EVs for V2G services, their state of charge, and the grid conditions that would trigger V2G operations.
In the real-time control stage, the model predictive control algorithm would need to be updated to include V2G as a controllable resource. The optimization problem would need to balance the benefits of using EVs for V2G against the impact on the battery life and the driving needs of the EV owners. The control strategy would need to dynamically adjust the power flow from EVs based on real-time grid conditions and V2G availability.
Overall, integrating V2G capabilities into the framework would enhance the flexibility of the system, allowing EVs to not only consume energy but also contribute to grid stability and reliability.

What are the potential challenges and limitations in implementing the proposed framework in a real-world distribution network with a large number of EVCSs and other distributed energy resources

Implementing the proposed framework in a real-world distribution network with a large number of EVCSs and other distributed energy resources may face several challenges and limitations:

Scalability: Managing a large number of EVCSs and other resources can increase the complexity of the optimization problem, leading to longer computation times and potentially suboptimal solutions. Scaling the framework to handle a large number of devices while maintaining efficiency and accuracy is a significant challenge.

Data Quality and Availability: The accuracy of the day-ahead forecasting models relies heavily on the quality and availability of historical data. In a real-world scenario, obtaining reliable data on EVCS demand patterns, PV generation, and load profiles for a large number of devices can be challenging.

Communication and Coordination: Coordinating the operation of multiple EVCSs and other resources requires robust communication infrastructure and coordination mechanisms. Ensuring seamless communication between the control system and the distributed resources can be a logistical challenge.

Regulatory and Market Barriers: Regulatory constraints and market structures may pose limitations on the implementation of the framework. Adhering to grid codes, market rules, and regulatory requirements while optimizing the dispatch of resources can be a complex task.

Cybersecurity and Privacy Concerns: Managing a large number of connected devices introduces cybersecurity risks and privacy concerns. Ensuring the security of data transmission, protecting against cyber threats, and addressing privacy issues related to customer data are critical considerations.

Addressing these challenges will require a comprehensive approach that considers technical, operational, regulatory, and cybersecurity aspects of implementing the framework in a real-world distribution network.

How can the day-ahead forecasting models be further improved to better capture the complex dependencies and correlations between the various stochastic inputs (load, PV, EVCS demand)

Improving the day-ahead forecasting models to better capture the complex dependencies and correlations between stochastic inputs can enhance the accuracy and reliability of the dispatch plan. Here are some ways to enhance the forecasting models:

Advanced Machine Learning Techniques: Utilize advanced machine learning algorithms such as deep learning, recurrent neural networks, or ensemble methods to capture nonlinear relationships and temporal dependencies in the data. These techniques can improve the forecasting accuracy by learning from historical patterns and trends.

Feature Engineering: Enhance the feature set used in the forecasting models by incorporating additional relevant variables such as weather data, traffic patterns, and grid conditions. Feature engineering can help capture the complex interactions between different factors influencing the stochastic inputs.

Ensemble Forecasting: Combine multiple forecasting models, each capturing different aspects of the data, to create an ensemble forecast. By aggregating the predictions from diverse models, the ensemble approach can provide more robust and accurate forecasts.

Dynamic Updating: Implement a mechanism to dynamically update the forecasting models with real-time data as it becomes available. This adaptive approach can improve the accuracy of the forecasts by incorporating the latest information into the models.

Probabilistic Forecasting: Move towards probabilistic forecasting techniques that provide not only point estimates but also uncertainty intervals. Probabilistic forecasts can offer valuable insights into the range of possible outcomes, helping decision-makers assess risks and make more informed decisions.

By incorporating these enhancements, the day-ahead forecasting models can better capture the intricate relationships and dependencies between load, PV generation, and EVCS demand, leading to more reliable and accurate predictions for optimizing the dispatch of resources in the distribution network.