Origin of EvAuto EV PLM (Electric Vehicle Peak Load Management)
I started working on electric vehicle charging in 2009 when I was at Servidyne. At that time, we were approached by a client that had purchased a small fleet of electric delivery trucks using funds from a government grant. This was very early in the evolution of electric vehicles, and much of the systems were experimental. As part of the installation of electric vehicle supply equipment (EVSE), we were asked to install metering for the distribution panel feeding the EV fleet and the individual EVs themselves. As we were designing the metering system, we found that the charging equipment included connections that would allow us to control charging. Servidyne’s practice focused on energy efficiency and cost control. With that in mind, I added logic to control the charge cycle and add demand response functionality.
At first the solution to managing charging to avoid usage peaks and minimize costs seemed easy. All we would need to do was to schedule vehicles to charge at different times, and focus those times on after-hours (non-peak) periods. Implementation of the logic was easy, and we expected good results, but it quickly became apparent that a simple solution wouldn’t work for several reasons. While a few of the reasons related to technical challenges, most of the challenges result from how the operator’s want and, more importantly, need the vehicles to work.
It quickly became apparent that in order to manage the charge cycle, we would need to collect data on how vehicles charge. Our first data stream was to collect the power and energy usage. We started to track kW and kWh, and by observing those trends, we could see how individual vehicles and fleets consumed electricity. The data was immediately useful as we could see when a vehicle charged, how much energy (kWh) each cycle charge consumed, the power (kW) requirements for each vehicle. We could also see if no vehicle charged at an individual EVSE at all on a particular day. I started to get a sense of the charge and trickle behavior of these first generation vehicles.
At first I believed that we could shape our load simply on that data. We were told that the fleet operated the same hours each week day, and that they would follow a similar route. We tried to create schedules that would match usage patterns such that vehicles would neither charge during peak hours nor simultaneously. Despite our best efforts we failed. Our customer frequently called to ask to override the controls as they suspected (and experienced) failures with the vehicle’s internal charging system that they wanted to repair during business (peak) hours. Even when the vehicles operated correctly, using our simple schedule we found that we were still setting peaks. After much effort, we realized that there were other variations that our simple model didn’t accommodate, and that we’d need more data about how the vehicles were being used.
This was complicated by the fact that I’m based in Atlanta, and the closest vehicles to me were over 500 miles away. In order to determine what the vehicles were actually doing my only input was the power meter installed at each vehicle. While this provided significant insights into the charge cycle, I had no way to determine the operational flow of the vehicles. At this point, our customer lost interest in peak load management as their focus turned to operational issues with keeping the vehicles on the road.
However, I never lost interest in solving this interesting control issue. From time to time I would travel to sites where I could speak with the operators. I learned when they needed the vehicles to be ready, what would happen to the vehicles when they returned to the yard and when the vehicles would be connected for charging. I combined the meter data with this information on daily operations to create algorithms to reliably record when the vehicles leave, when they return, when they charge, when they trickle, and even measure how long it takes for the vehicles to be loaded with product. However, this data alone didn’t yield a control solution
EvAuto evolved from an unexpected quarter. After the initial pilot installations, my participation in electric vehicle charging slowed and I returned to programming commercial and industrial control systems. I was working on a program for an older gas-fired boiler system when it occurred to me that there was a parallel to charging electric vehicles. I filed that idea away in my mind and only returned to it when I got another EVSE opportunity.
The final piece for making EV charge control possible was the development and availability of low cost industrial-grade controllers. These controllers are much smaller than those available when we first launched our EVSE program, and they operate a much wider temperature and humidity range, and offer significant processing power and memory. By coupling these control devices with low-cost cellular modems we have created and EV charge controller at a cost that can yield an immediate financial payback to EV fleet operators.
I’m excited by the possibilities for savings and load control for EV fleet managers. Time will tell when the fleets arrive, but when they do, I believe that EvAuto’s low cost approach can yield significant savings.