The central objective for this project is to determine the feasibility of Vehicle Grid Integration (VGI) by quantifying the potential value, cost, complexity, and risks in different implementations of VGI. Allocating available value among stakeholders and determining pathways for electrification of transportation to enable beneficial grid services such as mitigating renewables intermittency.
Aggregate ex ante modeling of collections of PEVs
Consider a collection of PEVs, for example, arriving at and departing from a garage. We have access to their arrival and departure statistics from historical data archives, as well as an empirical distribution of their net charging demands. We can define a default or nominal charging profile for each PEV. Each of these PEVs is flexible in that they can accept perturbations around this nominal charging profile. They could be charged rapidly at high currents, or could defer charging or discharge for some time. These PEVs are indifferent to charging profiles as long as their net demand is largely met before departure.
A key challenge is to model the aggregate flexibility of the collection of PEVs. Our central thesis is that the aggregate flexibility of the collection can be approximately modeled as a virtual battery. While individual PEVs have physical batteries, our thesis is nuanced and not obvious. The virtual battery has a compact parametric representation. It is characterized by just three parameters – its capacity C, the dissipation α, and its maximum charge/discharge rate m. These parameters are random. They depend on exogenous stochastic processes such as arrival/departure rates and charging demands. If we forecast the virtual battery parameters ex ante, the collection of PEVs can offer local services and grid services. For example, the collection could offer frequency regulation capacity in conventional day-ahead ancillary service markets.
Two fundamental advantages of this virtual battery model representation are its simplicity and its portability. The system operator (SO) can readily digest battery models and is immunized from the details of how the virtual battery is physically delivered. Indeed, it is well understood that other flexible demand resources such as smart buildings or pool pumps can also be modeled as virtual batteries. These diverse virtual batteries are distinguished by their reliability, availability, capability, and cost. The SO can balance these aspects and determine optimal collections of resources that provide services at minimum cost with assured reliability.
Run-time control algorithms for collections of PEVs
The provision of local and grid services have to be decided on in advance, with a certain lead-time. Lead-times are necessary to organize available resources (e.g. PEVs) or to respect current market structures (e.g. ancillary services co-optimized with day ahead wholesale markets). Once a service is contractually agreed upon, the total power consumption of the collection of vehicles must follow a prescribed reference signal that depends on the nature of the commitment made. For example in frequency regulation, the total power consumption of the PEV collection must closely follow the automated generator control (AGC) command issued by the SO (as measured from a pre-defined nominal power trajectory). This reference signal is revealed in run-time to the aggregator.
At every sample, the aggregator must parse the reference signal, i.e. it must decide how much power to allocate to each vehicle. This allocation must respect diverse constraints (e.g. providing 95% of total charge, precluding deep discharge of any vehicle, fairness considerations, etc.). This real-time control or scheduling problem is a central algorithmic challenge faced by the aggregator. It is very different from control problems in industrial processes, which use well understood PID control strategies. The control problem faced by aggregators representing collections of PEVs more closely resembles processor time allocation problems that arise in the context of assigning computational tasks to CPUs. There are various creative control strategies that can be explored including Earliest Deadline First (EDF) or Least Laxity First (LLF). EDF allocates power to PEVs prioritized by their anticipated departure time, while LLF allocates power prioritized by the charging demands. The control algorithm for allocating and dispatching vehicles for grid services fundamentally affects the feasibility and value proposition of VGI.
The two key outcomes are:
- Developing and releasing the VGISoft co-simulation framework for examining vehicle-grid interactions in any implementation of VGI. VGISoft will enhance Grid Modernization Lab Call (GMLC) activities by coupling with GMLC foundational models in valuation, design and planning tools, and SETO models to address solar variability.
The purpose and applications of each toolkit is outlined briefly in Figure 2.
- Application of VGISoft to address critical knowledge gaps for VGI through targeted case studies that quantify the feasibility of VGI, quantify potential for VGI to provide grid services such as supporting renewables integration, and determine the optimal implementation approaches for VGI.
The major outcomes of this project are summarized in the Figure on the right.
Described in further detail:
Outcome 1: Development of open-source toolsets for VGI planning, analysis, and operations
Addressing challenges requires toolsets that capture the complex interactions between drivers, vehicles, aggregators, utilities, SOs, and other VGI stakeholders. The first outcome of this project is the creation of the VGISoft co-simulation framework, comprised of several distinct toolsets.
While the complete collection of toolkits in VGISoft provide a valuable framework for R&D, the CAT and RAT (i.e. without PET) toolkits enable real-world implementation of VGI. Together, these two toolkits allow collections of vehicles to be aggregated and enable real-time operation of vehicles to deliver grid services while satisfying the mobility needs of each driver.
Outcome 2: Application of toolsets to address critical knowledge gaps and barriers to VGI
The two toolsets will be applied in two broad sets of case studies that address the knowledge gaps and critical needs for stakeholders in deploying VGI. These case studies serve to illuminate how VGISoft can be used in a variety of contexts.
The project will be conducted in close coordination with the multi-lab team within Vehicle Technologies Office (VTO’s) EV Smart Grid Working Group, including LBNL (prime), ANL, INL, NREL, ORNL, and PNNL. The methodologies, case studies, and timeline for this project will be closely coordinated with four other proposals submitted by the EV Smart Grid Working Group (GM0085, GM0150, GM0062, and GM0163).