Smart charging of electric vehicles

Smart charging of electric vehicles

By 2020, electric cars will be commonplace. Software agents play an important role in reaping the financial benefits from their battery capacity and in charging huge numbers of new electric cars without causing power failures. Konstantina Valogianni anticipates that charging electric cars without a co-ordinating mechanism will put a huge strain on the electric grid, creating a very real danger of causing power outages. Valogianni: “Customers want their cars to charge fast. They charge them at a maximum of 25 kW within an hour, while an average household usually consumes 3 kW up to 10 or 11 kW. At some point, there is no supply to cover this demand. That’s the main problem. Prices will skyrocket and blackouts could occur.”

So far, academic literature has only offered solutions that involve a central party co-ordinating the charging of electric cars. But Valogianni and her team take a decentralised approach. They propose an algorithm – Adaptive Management of EV Storage (AMEVS) – by which a software agent makes the decision whether to charge a car or to sell energy back to the retail market. The agent learns and tailors itself to individual car owners. Valogianni: “It focuses on the heterogeneity of the customers, who have different preferences and mobility habits. The algorithm learns our individual preferences and uses them to optimise charging times. It will allocate charging times differently within the time horizon of the day. That’s what makes the algorithm reduce peaks, and ultimately the demand on the grid becomes less volatile.”

The algorithm is run inside the car. “Ideally, you can schedule your driving routes and input your preferences into a chip in the car, for example, through a screen interface,” Valogianni says. This scenario is close to becoming a reality – Tesla Motors has already implemented a very simple version of personalised charging, where you can tell your car to charge during the night to benefit from lower tariffs. It paves the way for electric car management by algorithms such as AMEVS. “But of course, we need to test the algorithm in test beds like Power TAC before we can release it to the market.”