- +61 7 3374 2877
- Email Us
Systems and sensors that use AI-supported software increase the service life and operational reliability of networked assets, but this is not possible without associated cybersecurity. Ideally, holistic systems are placed at three central levels. And this is already possible today.
Energy network operators and asset managers already have to compensate for the growing loss of personnel knowledge and the increasing maintenance requirements of aging operating equipment. At the same time, challenges and burdens are increasing: The urgent expansion of grids, the feed-in of renewable energies and the decentralization of energy supply in the face of ever increasing energy demands hardly seem manageable. A key role is therefore being played by new technologies that help monitor and diagnose assets. Such intelligent data-driven systems make it possible to use the existing infrastructure more efficiently and extend its service life. The perfect interaction of all automation levels is crucial for this: from the sensor to the data node on the transformer to the level of global asset management.
“This is no longer a dream of the future; it is already possible today,” explains Tobias Gruber, MR product manager who is involved in the development of algorithms and mathematical training methods. “Reinhausen already offers a large portfolio of self-learning sensors and systems. And their capabilities are constantly increasing through mutual networking and an ever-growing learning curve.” The digitalization of assets is a necessary step for the reasons mentioned. But this transformation process — the independent further development of the systems themselves – will bring about changes in all industries, and also bring benefits that are not yet imaginable today.
Since it is obvious that AI is here to stay, it makes sense to use it to support the energy industry at all levels. At the process level, various sensors record numerous signals — reliably and precisely — directly at the transformer and forward them to a central communication node at the field level for consolidation. This provides reliable information on the maintenance and health status of the respective transformers. These evaluations then come together at the control level where maintenance strategies for the entire fleet can be developed. However, this only works if the sensor technology is modularly expandable and manufacturer-independent so that it can react agilely to future needs in fleet maintenance and renewal. And with such a high level of data exchange between the levels, cybersecurity is of particular importance. There is more to the term than just protection against hacker attacks.
1. INTELLIGENT SENSOR TECHNOLOGY
Sensors collect information – that much is obvious. Smart sensors can also evaluate it. But what does that mean in concrete terms? Which data can be better evaluated using machine learning than with previous methods? An important standard measure, for example, is the analysis of the insulating oil in the gas phase (DGA – dissolved gas analysis). In all common analysis methods, the gas is extracted, which exposes it each time to fluctuating external factors such as ambient temperature, air pressure, humidity and much more. Ensuring that these values always remain constant for measurement would consume vast sums of money and significantly increase the complexity of DGA systems. This is where artificial intelligence comes into play. Using training data sets, the algorithm in the smart DGA system learns the relationships between the gas concentration in the insulating oil, the sensor signal and the possible disturbing influences. With the aid of mathematical-statistical methods from the machine learning toolbox, the DGA sensor is thus “trained” during its development phase. The new intelligent specialist is then installed on the transformer and immediately knows what to do. “With the laboratory analysis of the oil as a reference point, the DGA sensor then also learns about its transformer and continuously recalibrates itself,” says Gruber. “In this way, effects such as sensor drift, aging of the insulating oil and the like can be compensated for, and consistent measurement repeatability ensured.”
