Testing thermal management in e-cars

If the battery of an electric car overheats or the interior does not get warm at minus 20 degrees, the problem often lies in the thermal management - a highly complex system with 200 possible circuits. Software testing for such systems means: test runs over several days, simulating sluggish temperature curves while remaining energy-efficient so that the range does not suffer. The path from expensive vehicle testing in extreme regions to virtual control units shows how Shift Left works in practice - and where the limits of simulation lie.

Podcast Episode: Testing thermal management in e-cars

In this episode, I talk to Patrick Meuth about thermal management software in electric cars. Many tests used to take place in the car, today they are shifting left: first hardware in the loop, then a virtual control unit that only executes the thermal functions. This saves journeys and detects faults earlier. The scenarios are demanding: protecting the battery, cooling the interior, prioritizing performance, keeping an eye on range.

“Then there are a total of 200 different interconnections as to how the entire system can ultimately be operated.” - Patrick Meuth

Patrick Meuth began his professional career at TKI Automotive in 2016 with his bachelor’s thesis in the field of model-based testing. He was then taken on as a test engineer for system validation on HIL test benches in 2017. Through his experience in various projects, he continuously developed his skills and was appointed Senior Test Engineer and then Test Manager in 2020. With the transition to CARIAD in 2021, he continued his work as a test manager. Since then, he has led the system validation team in an agile structure and assumed technical responsibility as Product Owner.

Highlights der Episode

• Thermal management testing moved from Africa testing to hardware in the loop to virtual ECU - earlier, cheaper, more targeted.
• Virtual ECU only tests thermal management function in isolation - no cross-influences from basic software or other ECUs.
• The same automated test cases run on the Hill and virtual ECU - only adapt mapping, not logic.
• Temperature control takes time: tests take two to three days because filters and timers are simulated realistically.
• AI-generated test specifications only work if requirements are structured and standardized beforehand.

Software tests for thermal management - How cars heat and cool properly

Modern electric cars not only have to drive, they also need to be kept at the right temperature at all times. The battery must not get too hot. The interior should be comfortable in all weather conditions. But how do you ensure that the software responsible for thermal management works safely, efficiently and reliably? That’s exactly what Richie and Patrick Meuth talk about in a new episode of the Software Testing podcast.

What is thermal management anyway?

Thermal management is the process of regulating how the temperature of different areas in the electric car is controlled. Patrick Meuth works at the Volkswagen Group, mainly for Audi and Porsche, and checks whether the software in the vehicle is doing everything right: cooling or heating the battery, keeping the interior at a pleasant temperature, but always remaining economical. After all, any energy that goes into heating or cooling is lost to the range - and nobody wants that.

The control unit, a small box in the car, controls everything. It receives information from various sensors, such as how warm the battery is, and then decides what to do. These are not just simple switch actions - up to 200 different combinations of states, outside temperatures and requirements can occur.

Testing past and present: Why Shift Left is so important

In the past, the development team tested the software almost exclusively directly on the vehicle. By car to Scandinavia for cold tests, to Africa for heat tests. That cost time and money. What’s more, you can’t just generate minus 20 degrees at the touch of a button when it’s 20 degrees outside.

The new approach is called Shift Left: testing as early as possible, discovering as many errors as possible before going into the real car and the real world. The first tool was a so-called HiL system (hardware in the loop). This is a huge cabinet in the laboratory in which the original control unit is installed. Sensors and actuators are simulated. For example, a drive up a mountain at an outside temperature of 40 degrees can be tested on the computer.

But even that is time-consuming. The aim is to replace the large cabinet with virtual control units. The thermal management software then simply runs on the computer, without any real hardware. This allows the team to check whether the control unit is doing its job even more quickly and cost-effectively.

Automation: less manual work, more control

Most tests today are fully automated. The same automation tool checks the software on both the HiL system and the virtual ECU variant. This means that one and the same test case can be executed at different points in the development process. Minor adjustments are made, but the basic principle remains the same.

Manual work is still sometimes required. For example, when very special tests are required that cannot be easily simulated or that require a real feel for the effect of the air conditioning system, for example.

New software all the time - how do you keep testing clear?

Electric car software is updated frequently these days. This means a lot of changes, often at short notice. Nevertheless, testing remains stable. The team knows exactly which requirements change and which test cases need to be appraised. If no impact on the function is expected, the tests are simply run again.

Complex scenarios in particular, where several requirements have to be met at the same time - such as cooling the battery AND the interior with limited energy - present challenges. But automatic and early testing can often identify problems before they occur in the real vehicle.

What’s left for the real car?

In the end, simulation is useless if you don’t test how the system feels in the real vehicle. Some things simply cannot be measured: How pleasant does the air conditioning really feel? Do the valves switch correctly? A dedicated team checks this in the car - but the focus is clearly on finding as many faults as possible in the laboratory beforehand.

A look into the future

Patrick Meuth plans to further expand the virtual test environment, introduce continuous integration (fully automated processes without manual work) and even specify tests with artificial intelligence. Development remains exciting - and the goal is clear: more efficient, safer and more pleasant electric cars for everyone.