Adaptive EV load management for the smart building
A new paradigm for the efficient allocation of electricity to charge electric vehicles within the smart grid - executed in hardware and supported with software

Variablegrid - a subset of the electrical grid defined by automated demand control

Abstract: The portion of the electrical grid dealing with EV charging can be virtualized and separated from the remainder, because - unlike the loads currently on the grid - its loads can be automatically controlled. This control entails three co-operating processes: J1772 charger modulation, current load sensing and a feedback consumption indicator signal. The interface to the existing grid is defined by a final, feedforward command signal within the same hierarchical network. It can be issued from a transformer substation point, but location is discretionary.


Variablegrid is a tool to enable users of standardized electric vehicle (EV) chargers to maximize EV charge rates and reduce charging times. It is also a tool to allow providers of multi-point charging facilities to reduce equipment costs. And finally, it is a tool that gracefully integrates these new large-scale demands for electricity into the existing electrical grid.

Until now, it would have been reckless for an EV user with a 100A electrical service to install an 80A charging circuit. This is no longer true. Of even greater importance, it is now possible to connect multiple charging circuits to a central distributor when the aggregate rating of these circuits exceeds the rating of the central distributor, on a deterministic rather than probabilistic basis.

This second point is particularly significant to condominium developers offering EV charging facilities, to centralized commercial EV charging facilities and to electric utilities.


There is a reason why we have to consider these features now - any growing technology-based market is sooner or later subject to demands for increased capabilities. It is very likely that some catalytic event will precipitate increasing demand for EV's. History has shown this to be true of the microcomputer in the late 70's. To somebody who has experienced that phenomenon firsthand, the current EV market gives the impression that it is a market waiting for something to happen. It is also likely that with increasing demand will come technological gains that lead to more powerful batteries. EV charging loads can be expected to be both continuous and concurrent - unlike most loads on the electrical grid - and therefore pose new challenges. Variablegrid is designed to meet these challenges, primarily by taking measures that allow existing infrastructure to change in an evolutionary rather than revolutionary way.


The basic concept behind variablegrid is very simple: it carries out current sensing at the customer load point. It then computes the difference between the instantaneous measured current flow and the load point's current rating and assigns that value to the EV charger. The amount of current allocated to an active charger never exceeds its demand, because variablegrid measures its consumption. However simple this idea is, it has significant consequences.

By using a small amount of electronics and some software to realize the above technique, variablegrid can provide pinpoint variable (modulated) control over each charger - this is more desirable and in sharp contrast to on-off control over a whole region. This feature makes use of the J1772 standard and basically prioritizes EV chargers relative to other loads in a graceful way.

In defining control at the customer load point, variablegrid is able to provide very precise EV load control integration with existing equipment and without the need to resort to identification of, and feedback from, individual EV chargers to a remote central control.

As a result of its simple algorithm, variablegrid provides an adaptive load control mechanism: most of the current will be allocated to the EV charger during off-peak consumption times. This provides a fairly elegant and unique load-leveling mechanism over large service areas.

Since EV charging rates are inversely dependent on any concurrent loads on the measured distributor, the variablegrid adaptive control strategy directly (sooner or later) will influence customer consumption patterns without any need for specific customer education. The customer will learn automatically that turning on his hot tub will lengthen the charging time for his electric car, or that leaving on his plasma TV with the sound turned off will do the same.

Because variablegrid consists of standard, low cost equipment, a simple strategy and a deterministic control algorithm, it can be brought online quickly and without the need for regional load simulation studies.

However, the implementation of variablegrid does carry a serious responsibility. By potentially driving current distributors to their limits it can overload equipment feeding these distributors. This equipment typically consists of a distribution transformer node. Because the current rating for these nodes is based on historical consumption patterns - which assume that not all customers drive all their loads at the same time - their rating is lower than the aggregate ratings of the load points they support. As a result, they could become overloaded. However, distribution transformers are merely (larger) current distributors and can be handled as just another load point by variablegrid.

Up to this point everything that was said about variablegrid seems to make it simple, adaptive, automatic and self-contained. Nevertheless, variablegrid must be integrated with one final current distributor - the distribution substation. Here, although variablegrid cannot play any direct role, it does serve two functions: reporting and throttling EV consumption.

Please recall that above we said there are 2 parameters associated with a distributor: its current flow and its capacity (or rating.) Reporting EV consumption is a relatively simple task, but throttling it depends on a feature of variablegrid called virtual (or allocated) capacity. In essence, variablegrid does not use the load point's actual rating in its calculation, but rather its allocated rating. This is a software value that can vary from 0 to 100% of actual capacity. By sending a (very short) signal to any of the distributor nodes sensed by variablegrid, a central administration can reduce the EV consumption for that node, or bring it back up from some reduced level; the signal effectively changes the value of allocated capacity at the node. This method adds supply-side control to the already shown demand-side control.

This mechanism does not affect any other equipment deriving power from that distributor.

One final feature of variablegrid is worth mentioning. EV loads will represent a substantial demand on the grid. After a regional power failure there are tense moments during the so-called blackstart recovery phase, when large reactive currents flowing during low-voltage startup conditions can cause equipment to trip or fail. Variablegrid measures reactive currents in exactly the same way as it does resistive currents. Excessive current flows at a measured distributor will cause variablegrid to delay EV charging on that distributor until its current flows are once again nominal. This feature will apply not just to a utility, but also to residential or industrial sites that combine the use of motorized heavy equipment (such as air conditioners) with EV charging.


In order to implement all of the above, variablegrid relies on a single printed circuit board (PCB) with removable system-on-chip (SOC) electronic components and function-dependent software. This approach was taken to reduce production costs. Each board has the same number of discrete components, but the specific complement of SOC electronics on the board depends on the board's function as a node. The software depends on both the functions of the board and the network architecture it supports. A single node can be a sensor, a J1772 signal generator (control node) or a communications supervisor. A sensor node can support AC, DC and current or voltage output sensors. The J1772 control node also implements the J1772 safety protocols to prevent current flow to EV's that are either not plugged in, not demanding charge or that have non-standard, damaged or altered interfaces. The communications supervisor can be used as a single-node monitor or as a regional network control node. It uses an RJ45 plug into an SOC module that can connect over TCP/IP to a local variablegrid program, or to the same program running remotely over the Internet. Each board is optionally fitted with an SOC wireless communications unit. A sensor or a control node can be addressed locally with a USB connection to set its actual rating (actual ratings cannot be set remotely) or to monitor its operation.


Before discussing network architecture it should be clear that variablegrid can discharge its current sensing and EV charger control functions locally, without the use of a network. This form of operation disregards any existing infrastructure considerations and - within limits - only serves individual EV users. The phrase within limits here means that any degree of clustering by such independent users around the same load point may eventually cause it to trip or fail.

The simplest node possible combines both sensing and control in the same board. This is applicable to a single EV garage that also houses the main circuit breaker. A slightly more complex arrangement, where the EV charge station is separated from the main breaker location by obstructions that do not permit signal wiring, requires a wireless connection between two nodes. A user that wishes to monitor operation of the nodes remotely may add a TCP/IP communications node to either of the above. This covers all possible arrangements at a single user location. In an industrial or multi-unit residential setting, each EV service equipment (EVSE) or EV charger location requires one control node with wireless capabilities; each load point feeding multiple chargers or in turn feeding multiple sensors, requires a sensor node with a wireless feature. If the separation between the EVSE's and the sensor is significant or spatially complex (concrete or cement/steel obstructions) or if the number of EVSE's per sensor is large, there will have to be one or more intermediate communications nodes. Finally there is one communications node attached to a central processor that can execute a Java program - a 1 GHz net book machine executing Windows, Unix or OSx would be typical. It is this program that allows control over the allocated ratings described above. In a typical regional configuration each distribution transformer is assigned a sensor node, as is each residential customer load point; each customer also has an EVSE control node. The regional distribution substation employs one communication node, separated from the distribution transformer sensors by intermediate communications nodes.

A final note about communications in regional networks. Although variablegrid is being tested with only wireless communications, there may be challenges implementing this when distribution transformers are either underground or radio frequency (RF) shielded. A better form of communications would be power line carrier (PLC) based. However, this technology has in the past encountered difficulties in bridging transformer equipment. A promising new form of PLC called orthogonal frequency division multiplexing (OFDM) may be a solution, but low-cost chips and ancillary support seem to be still nebulous at present. Utilities have a great degree of experience in this area. Utilities are also planning a smart grid - if this plan includes Internet connections, then regional variablegrid nodes could be accessed as part of smart grid services.


We have presented a new way of matching nascent electrical grid loads to existing plant and equipment using simple but elegant techniques with a relatively flat learning curve and a short planning horizon. We believe that the EV's representing these loads will soon become a large market. History has shown that such a technology-driven market could be subject to sudden unexpected shifts to the upside, given just one major catalytic event. We believe that the shorter the planning horizon applied to it, the lower are the chances that such an event will cause market dislocations in regard to existing infrastructure.

Copyright 2010 IBX datasystems ltd Vancouver Canada