EV Charge Scheduling Using Category-Based Demand Flexibility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current EV charging scheduling algorithms are sub-optimal due to high computational complexity and failure to incorporate past data for large-scale EV charging, leading to inefficiencies in managing peak demand and grid strain, especially with the rapid growth of EV adoption.
Innovation Solution
Implementing a stochastic dynamic program with approximate dynamic programming (ADP) for preemptive scheduling of EV charging, categorizing EVs based on arrival/departure times and SoC, allowing flexible charging demands, and optimizing profit through multi-stage algorithms to manage peak hours effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If deterministic optimization algorithms are used for EV charging scheduling, then charging profits can be optimized, but computational complexity increases exponentially with the number of EVs
Solution Approach 1:
The patent segments the EV charging scheduling problem by categorizing EVs into different types (e.g., flexible demand, rigid demand, different departure times, different SoC levels). This segmentation transforms the exponentially complex problem of scheduling individual EVs into a polynomial complexity problem of scheduling category groups, resolving the contradiction between optimization accuracy and computational complexity.
2Speed
If online algorithms like EDF or LLF are used for EV charging scheduling, then real-time scheduling is achieved, but future demand information from past data is not incorporated leading to sub-optimal scheduling
Solution Approach 1:
The patent applies preliminary action by pre-processing historical data to extract statistical patterns and characteristics of EV demand before the actual scheduling event. This allows the system to incorporate learned patterns from past data into real-time scheduling decisions, achieving both real-time responsiveness and data-driven optimality simultaneously.
3Reliability
If Model Predictive Control or scenario-based algorithms are used to optimize charging schedules, then charging profits are maximized, but the algorithms become computationally intractable for large-scale EV markets
Solution Approach 1:
The patent changes the parameters of the optimization problem by transitioning from continuous individual EV scheduling variables to discrete category-based aggregated variables. This parameter transformation reduces the problem dimensionality from exponential to polynomial scale, enabling scalable implementation in large EV markets while maintaining profit optimization through category-level decision variables.
4Reliability
If queue network frameworks are used to allocate chargers to different EV types, then service quality can be optimized, but the system cannot effectively manage aggregate demand and peak hours
Solution Approach 1:
The patent adds another dimension to the scheduling problem by incorporating time-based categorization (different departure times, different charging windows) and SoC-based categorization in addition to flexible/rigid demand classification. This multi-dimensional categorization enables the system to manage aggregate demand across multiple dimensions simultaneously, achieving both service quality and aggregate demand management.
Data Source
AI summary
A system and method for controlling charging of multiple electric vehicles (EVs) arriving at, and departing from, different charging stations at different times, includes scheduling charging of each EV of the multiple EVs responsive to which one of a plurality of categories each EV is assigned, each EV being assigned to one of the categories according to an arrival time at an associated one of the different charging stations, a departure time from the associated one of the different charging stations, an initial state of charge (SoC) of the EV, and a target SoC of the EV, and controlling charging of each EV responsive to the scheduling of charging for the assigned category of each EV. Charging demand may be selected by each EV as being reliable or flexible with flexible charging demand having a minimum target SoC and maximum target SoC.