Bidirectional EV Charging Control for Day-Ahead and Intra-Day Arbitrage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for managing bidirectional charging at electric vehicle (EV) charging stations face challenges due to limited charging and trading flexibility, high computation complexity, and electricity price volatility, particularly in day-ahead and intra-day markets, which hinders optimal energy arbitrage and fleet operation.
Innovation Solution
A method and system utilizing a day-ahead planning module and an intra-day planning module, with a learning agent and Graph Neural Network (GNN) to optimize vehicle-charger allocation and prioritize energy trading, leveraging price fluctuations across markets to minimize charging costs and maximize discharging profits.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If a day-ahead planning module is used to prepare charging schedules, then charging cost optimization is improved, but computation complexity increases due to fleet-level planning with heterogeneous vehicles and chargers
Solution Approach 1:
The fleet management problem is segmented into individual vehicle-level planning problems. Each vehicle is assigned to chargers independently using a learning agent that makes local decisions based on vehicle-specific parameters (battery capacity, charging rate requirements) and charger availability, rather than solving a monolithic fleet-level optimization problem.
Solution Approach 2:
Each vehicle effectively serves its own charging planning needs through the learning agent that autonomously determines optimal charging schedules based on trip plans, battery state, and predicted electricity prices, without requiring centralized fleet-wide coordination and complex global optimization.
2Ease of manufacture
If limited charging points are provided to reduce capital expenditure, then asset utilization is improved, but charging flexibility decreases leading to planning difficulties
Solution Approach 1:
The system dynamically assigns vehicles to available charging points based on real-time charger availability, vehicle battery state, and predicted electricity prices. The learning agent adapts charging schedules flexibly to match limited charger availability with vehicle charging needs, optimizing the use of fewer charging points without requiring rigid infrastructure expansion.
3Productivity
If intra-day market trading is prioritized for arbitrage opportunities, then profit potential is improved, but day-ahead trip fulfillment reliability may worsen due to limited charging flexibility
Solution Approach 1:
The system changes the pricing parameter dynamically by switching between day-ahead and intra-day electricity markets. The learning agent monitors electricity price predictions and automatically adjusts the charging/selling strategy - charging in the day-ahead market when prices are favorable and ensuring trip fulfillment, then utilizing intra-day market opportunities for arbitrage when price differentials exist, without compromising vehicle availability.
4Productivity
If multiple charging/discharging operations are scheduled per vehicle per day, then energy arbitrage opportunities are improved, but planning error propagation increases
Solution Approach 1:
The system implements feedback mechanisms where the learning agent continuously monitors actual charging/discharging outcomes, electricity price realizations, and vehicle trip completion status. This feedback allows the system to learn from past decisions and adjust future scheduling to minimize error propagation, ensuring that multiple charging/discharging operations per day are coordinated to maintain trip fulfillment while capturing arbitrage opportunities.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This disclosure relates generally to a bidirectional charging at an electric vehicle (EV) charging station by an energy model that uses electricity bought from the day-ahead market for charging the fleet of electric vehicles (EVs) and uses the intra-day market for arbitrage. The competitive pricing of wholesale electricity markets and distributed energy resource capability of EV fleets (in addition) provide a revenue channel through energy arbitrage. To effectively handle electricity price variations and the energy demand of the EV fleet, the present disclosure utilizes a graph representation-based learning agent (LA3_D) with two-stage encoding for day-ahead charge planning; and a priority order based greedy heuristic (GH_I) for intra-day arbitrage planning. Because the agent learns the planning policy of mapping EVs to charging operations over several problem instances, it is able to solve a given instance with limited sub-optimality when put to test at different levels of scale.