DNV GL on risks of renewables: “We need new tests and standards to prevent major failures”

Print Friendly
The largest man-made system on earth (photo DNV GL)

The largest man-made system on earth (photo DNV GL)

The transition to renewable energy is accompanied by the widespread use of power electronics, such as inverters, which require a whole new way of testing smart equipment, says Theo Bosma, Program Director Power Systems & Electrification at DNV GL, one of the largest technical consultancies in the world. According to Bosma, the new power electronics are not adequately tested at the moment. “New technologies such as solar, wind, batteries and smart grids are driven by software. But as a sector we are still testing individual components instead of the brains of these systems.” He warns that “if we don’t change our methods and standards, this will cause major problems in the future, including blackouts”. DNV GL calls on the electricity sector to jointly develop new industry standards based on “hardware-in-the-loop” techniques.

With the electric power system increasingly being taken over by software-driven “power electronics”, Theo Bosma, who leads a team of  20 researchers in labs in Singapore, Bristol, Copenhagen, Oslo and Arnhem, has a practical example showing how things can go seriously awry. In the business this is known as “the German 50.2 Hz problem”.

What happened was that in the southern part of Germany solar panels would all shut down when the frequency exceeded 50.2 Hz (this happens when the sun starts shining and all panels together produce too much electricity), and would all start producing again once the frequency went back to 50 Hz, after which they would all shut down again. This yo-yo effect was the result of many inverters reacting simultaneously to signals from the system. “Individually they all did what they had to do”, says Bosma, “but in combination the result was not what was expected or wanted.”

“You can’t tell by looking at the components what they are going to do. You need to test their brains”

Another example Bosma encountered in which “smart” components showed unexpected reactions was in the case of a medium-voltage transformer which contained power electronics keeping the voltage at a constant level. “This worked so well that a similar transformer a few kilometres away was put in place. But then the two started hunting each other. If one reduced the voltage, the other started increasing it, because there was no communication or coordination.”

What these examples show, says Bosma, is that the way electric power equipment is tested is not adequate anymore. “We only test to comply with the standard, not whether something is fit-for-purpose. We used to have only passive components in the system. You could simply test how they behaved and then install them. But new power electronic equipment is entering the system,  which is software-driven, such as inverters in wind turbines and solar panels. You can’t tell how they will react in the system if you test them in the old way, individually. You have to test how they behave within the system.”

Bosma says inspections used to be simply a matter of “opening the cabinet and looking at the components”. But nowadays “you might as well close the cabinet again, because you can’t tell by looking at the components what they are going to do. You need to test their brains. The software that controls them.”


Among researchers, including those at DNV GL, a company that has long been involved in developing quality control tests and standards, there has been a growing awareness that a fundamental change is needed in testing practices in the electricity sector. In fact, researchers have already developed new techniques enabling the testing of components at a systemic level. This is called hardware-in-the-loop testing.

What it amounts to, says Bosma, is that you simulate in the lab the real-life environment in which the components will be functioning. “In this way you are able to test the behaviour of the software. If you don’t do this, you will get what we call rogue software entering the system. This can have serious consequences down the line.”

White Paper: Power Cybernetics 

For more information on the need for hardware-in-the-loop testing, see this White Paper: Power Cybernetics – the future of Validation, published by DNV GL earlier this year. This discusses the implications of the introduction of power electronics for the testing and certification of electricity components. 

Cybernetics comes from a Greek word meaning “the art of steering” and can be described as the science of automatic control systems. It applies especially to systems that incorporate feedback loops, to trigger  system changes. The Norwegian company Marine Cybernetics, a subsidiary of DNV GL, is specialised in hardware-in-the-loop testing in the maritime sector.

Such hardware-in-the-loop testing makes use of supercomputers which operate at the tremendous speeds that are required to be able to calculate the behaviour of smart systems. DNV GL has one, as do several other labs in the world. But hardware-in-the-loop testing is far from standard industry practice at the moment, says Bosma. “It has been developed at the academic level. What we at DNV GL would like to see is that this becomes the new normal in testing. If the industry remains stuck in old-fashioned component testing, we see great risks.” What kind of risks? “Anything you can imagine that may go wrong.”

IEC community

The most important way of ensuring that future testing will be up to speed, says Bosma, is to change the international standards that components must comply with. In the electricity sector the most important norms are the IEC standards of the International Electrotechnical Commission. Contrary to what one might think, these standards are not imposed by governments or regulators, but developed by the industry itself through this international platform, in which different stakeholders from just about all countries in the world are represented.

So is the IEC community not sufficiently aware of the issue? For example, aren’t system operators, who are responsible for the smooth functioning of the electricity network, not paying enough attention to the risks caused by the spread of power electronics?

“If the industry remains stuck in old-fashioned component testing, we see great risks”

“People are becoming more aware of the risks”, says Bosma. “But as an industry this should be higher on the agenda . Some manufacturers are starting to use hardware-in-the-loop testing, for example those who are developing high-voltage direct current (HVDC) networks. They are so big you can’t afford to take chances there. But it’s not enough.”

Many system operators still believe that N-1 testing is sufficient, says Bosma. “As DNV GL we have always been a forward looking company. We see that this is not just about individual components anymore. We tell system operators that this is software. This is about the entire system, not about individual components.”

New procedures

The transition to a decarbonised electricity system increases the need for systemic testing, Bosma points out. “Solar and wind energy systems are both full of power electronics. So are battery systems and electric vehicles. In the future, both generation and use of electricity will be dominated by power electronics. If we want this transition to be successful, we need new standards and testing techniques.”

Bosma is convinced that it is possible to have an electricity system dominated by variable renewable sources, such as solar and wind power. “If you ask me, will we be able to integrate these sources, my answer is yes. Technically it can be done. You can reinforce and expand networks. Make demand more flexible. Implement storage. But the system will become much more complex. That’s why we are calling on the industry to jointly develop new testing procedures and standards based on hardware in the loop techniques.”


  1. Mike Parr says

    Some good points. The stability of networks relies on a transition from individual random behaviour both in terms of timing and size of load to an aggregated averaged load. Expressed another way, individual load points behave in a quantum fashion. Once aggregated – the quantum wave function collapses & one has “classical” load behavior. The introduction of embedded generation with all the same set points as the article rightly identifies – introduces collective behavior which was formerly absent. Same happens when, for example PV & storage becomes widespread – with a focus on charging the batt’ in the morning asap – & once that is done – large amounts of PV energy then hit the network. An alternative approach is to intro a randomizer function based on weather forecasts – the batt would be charged randomly though the day with the objective of max charge at the end of the day. This would lead to an average behavior from the PV+batt fleet. However, current trajectories in some EU projects focus on central control – which is a mistake. The 50hz – 50.2hz behavior could, likewise be partly addressed by taking a randomizer approach coupled to a bandwidth approach (49.9- 50.05 and 50.19 – 50.22) which would go some way to ameliorating the impact of PV drop-out/drop in. Of course none of this will happen because DNOs want to become DSOs (& enhancing their control of embedded generation) thereby enhancing their income streams.

  2. Frans Rusting says

    In discussions about future networks and especially about the advanced equipment that is needed a major subject is rarely mentioned.
    Most systems require some sort of remote control. Using the internet for this purpose seems attractive, but the importance of our energy provision for economy and well-being makes attacks (both by civilian and military hackers) highly probable.
    I believe that this important subject needs a lot of attention and discussion. Using the internet should in my thinking be avoided. For HV systems it may be possible to use the ‘natural protection’ provided by the high voltage cables, e.g., by using these also for remote control purposes.
    But for systems like Smart Grids such protection is not available. Using the internet may be too risky, and in this case there is no ‘natural protection’.

  3. Tilleul says

    Sounds like a solution trying to find a problem to solve.

    Inverters no more disconnect at 50.2 Hz in Germany so there are no 50.2 Hz problem ( see slide 9 for the droop of an inverter http://files.sma.de/dl/7418/Flyer_Niederspr-ADE123016w.pdf , they gradually reduce their power).

    Inverters don’t behave like the factory told them, they behave like their settings told them and the settings are decided by the grid operator… The 50.2 Hz problem is not a problem made by inverters, it’s a problem made by the settings of the inverters chosen by the grid operator.

    And grid operators being people who knows what they do they are perfectly well aware of when to change the settings according to the level of PV penetration as they did in Germany (http://www.pv-magazine.com/news/details/beitrag/germany-retrofits-200-000-pv-installations-to-meet-50-hz-requirement_100016243/#axzz4Rgakj67X ) or as they did in Hawai…

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>