New grant to accelerate testing of solar technology

A solar PV system must be able to operate under widely varying conditions. A dark and rainy day in January places one type of strain on the system, while a sunny day in July places another.

For companies developing inverters and other power electronics for solar PV systems, investigating how their equipment responds to several years of changing weather conditions can therefore be a considerable undertaking.

Researchers at the Centre for Industrial Electronics at the University of Southern Denmark now aim to make that process easier.

With a grant of DKK 336,000 from the EMD Foundation, Assistant Professor Mohammed Ali Khan will develop a method for selecting the most representative parts of large datasets. The aim is to convert several years of weather and operational data into a smaller, more representative dataset that can be fed into existing tests while still reflecting the stresses the equipment will experience throughout its lifetime.

– When we work with artificial intelligence and machine learning, data play a crucial role. Large volumes of data also require considerable computing power and can make testing a lengthy process. We want to identify the data points that provide the most valuable information, enabling companies to test their equipment more quickly and efficiently, says Mohammed Ali Khan.

A few days to represent an entire month

The project is called PRISM-Solar and will run for one year. Its total budget is DKK 374,000, including co-financing.
Among other things, the researchers will analyse data on solar irradiance, temperature, humidity and other weather conditions that affect energy production and the stresses experienced by a solar PV system.

Instead of processing every measurement from an entire month, the method will identify a small number of days that collectively represent the different conditions experienced during that month. These might include a sunny day, an overcast day and a day with highly variable weather.

The same principle will be applied to the remaining months of the year. A typical day in January looks very different from a typical day in July, and the model must therefore preserve the variations that influence the equipment’s performance and degradation.

– We need to find the right balance. The dataset should be as compact as possible while retaining the required level of accuracy. If we succeed, companies will be able to spend considerably less time investigating how their equipment will perform under different operating conditions, says Mohammed Ali Khan.

A faster route from development to market

Solar panels produce direct current, while the electricity grid uses alternating current. An inverter converts the electricity and is therefore a key component of a solar PV system.

When a new inverter or another power electronic component is developed, the manufacturer must investigate how the technology performs and degrades under changing loads. Even accelerated lifetime testing can take several months.

A smaller and representative input dataset could therefore allow existing tests to run faster and save energy and computing power. It could also give companies a better basis for assessing the equipment’s efficiency, lifetime and economic viability.

– The perspective is that companies will be able to test their solutions more quickly and thereby shorten the route from development to market. The method could also be used to compare different technologies and assess how they perform under local weather conditions, says Mohammed Ali Khan.

Data from Sønderborg to produce a local test profile

By the end of the project, the ambition is to have developed a flexible digital tool.

Users will be able to enter several years of data from a particular geographical area and generate a shorter set of representative operating profiles. Data from Sønderborg could, for example, produce one profile, while data from Copenhagen or a location in southern Europe could produce another.

This would allow companies to test their technology against the specific conditions in which it is intended to operate.
The method will initially be developed for solar energy, but the underlying principle could eventually also be applied within fields such as wind energy and electric motors.

CIE strengthens its investment in artificial intelligence

PRISM-Solar is also part of a broader development at the Centre for Industrial Electronics. Over many years, the centre has built up strong expertise in power electronics and energy systems and is now working strategically to combine these fields with artificial intelligence.

This work is taking place within the research area Intelligence for Dependable Power Engineering. CIE is also developing new teaching activities in artificial intelligence for power electronics and electrical power engineering.

– Congratulations to Mohammed Ali Khan on securing the grant. It is a great achievement for him and for CIE, says Professor Thomas Ebel, Head of the Centre for Industrial Electronics.