Advantages of artificial intelligence for charging electric vehicles

Table of contents

We firmly believe that AI can play an important role in supporting people. Charging solutions that use artificial intelligence could also win over most EV drivers in the future. With this Insight, we want to dive deeper into the potential of AI by looking at how it could improve the end-user experience of public charging.

Anna’s problem with public charging

To understand how EV drivers make decisions, let’s look at a real-life scenario of a typical EV driver. Imagine Anna, 32 years old, lives in a medium-sized German city and works as a software developer at a regional bank. She recently decided to go electric (with 200 km range) and definitely loves it.

On a sunny Thursday afternoon, she receives a call from a friend who lives in a town about 120 km away. Her friend was released without warning and would therefore enjoy some comfort from Anna. They both decide that they will have dinner at their friend’s house and Anna will stay until tomorrow. Before Anna starts her journey, she immediately thinks about loading, as she knows that some planning is required. After checking her favorite charging app, she finds a public 22 kW charger near her friend’s house – currently occupied – and an HPC charger right outside the highway exit. She decides to charge at the HPC charger because she doesn’t want to risk not being able to charge her car in the evening. She won’t have time before work the next morning.

No sooner said than done – she gets into her car, drives along the highway, charges at the HPC charger and arrives at her friend’s house. When she arrives, she sees that the AC charger is now available and that she could have charged there. She is frustrated because she has already paid a lot of money for charging at the HPC charger.

Why did she have this frustrating charging experience? Their decision, which turned out not to be optimal for them, was mainly based on three heuristics that people regularly rely on when making decisions: Availability, Representativeness and Adaptation. All these heuristics also influenced Anna’s decision-making:

  • Availability heuristic: She recently tried to charge at an AC charger but found it blocked. This is the most present memory of AC-Laden in her mind. She doesn’t want to experience this again.
  • Representativeness heuristic: The AC charger in her neighborhood is used a lot and her neighborhood is quite similar to that of her friend. At least that’s what she thinks. She therefore projects the characteristics of the AC charger near her onto the AC charger near her friend, without knowing whether this is actually true.
  • Adaptation heuristic: In their minds, AC chargers are typically blocked. Even if she invested more time to better understand the local situation with her friend, she probably wouldn’t adjust her estimate of availability enough to conclude that he would be available when she arrived.

The general lack of information and reliance on heuristics, combined with people’s typical high risk aversion, leads Anna to choose the HPC charger – even if it means she has to pay more for charging. But can an intelligent charging solution do something about it and help her?

Why displaying statistical information is not enough

The use of statistical data to help Anna is often discussed. A charging application could display a histogram of usage data based on the time of day and day of the week. In this way, Anna could get an estimate of how likely it is that the station will be available when she arrives. We all know this type of diagram from Google Maps.

Calculating the average occupancy on Thursday at 19:00 on historical data would give the a priori probability


of the availability of the charging station:

𝑃 (available)=number of available charging processes

Total number of charging processes

Anna’s charging app also knows the current status of the charger – it is occupied. The combination of this information could be expressed as an a posteriori probability that the AC charger will be available in one hour if it is occupied now:

𝑃(available in one hour∣occupied now)

The duration of the charging session could also be taken into account:

𝑃(available in one hour) = 𝑃(available) × 𝑃(duration of charging session)

But there is a lot more information available and some has already proven to be very important in changing the estimate of availability, such as weather, events in the area and traffic. Do you understand where this leads? Nobody wants Anna or her charging app to collect all this information and perform the calculations. The statistical approach would be laborious and may not end up being as accurate as it needs to be. We believe that artificial intelligence can take over this task by integrating information and making a valuable recommendation on future availability.

How AI can make the difference

Most people know from biology lessons that the human brain consists of neurons and axons that send action potentials. Although it is much more complex and scientists are still puzzling over it, this basic knowledge is enough to inspire computer science: In neural networks (just one part of AI), the basic structure of our brain is replicated. This is done in order to artificially create one of the most important functions for decision-making: learning from experience.

People can generalize or imitate from only a few examples. Depending on the task, artificial neural networks need thousands to millions of examples to learn from them. By focusing on a single task, they can discover correlations between different parts of the data that a human would overlook. The “input” is different here: People have their personal experiences, their knowledge and their situational perception, while neural networks can only read numbers. Both filter useful information to solve the given task. While humans tend not to change their filters for a given task (once decided), neural networks remain completely rational and consider all given information equally. Depending on the task, this can be a great advantage.

If you look at Anna, she has already learned from experience that the loader near her is often busy in the evening. She thought that the loader near her friend’s house was being used in a similar way. Maybe both are placed in the city center, but some surrounding features are different, such as the nearest charging station, restaurants, etc. Unlike the human decision-making process, an AI could integrate different types of data that influence charging availability but are impractical for an EV driver to consider manually. The historical use of charging stations is just one part of the unlimited possibilities of data from the surrounding area and similar locations. Such a system could be able to predict the availability of charging stations in the near future with high accuracy and give Anna a rational risk assessment for her charging decision. If her favorite charging app used such a system, she could avoid the excessive costs at the HPC charger.

We need to make AI accessible for all charging applications

We believe there are many great charging products already out there, but at the time of writing they are simply not good enough to help EV drivers with their charging decisions. Especially in times and areas where the number of public charging points is growing slower than the number of EVs on the road, we need to get better at helping drivers find free charging points.

The use case described in this blog post is just one of many where integrating advanced AI technology into a charging application could make the difference between a “frustrating” and a “best-in-class” charging experience. We know that AI can be challenging and requires significant upfront investment. However, Service4Charger has set itself the goal of continuously looking for innovative solutions to improve the charging experience for all users.