Post by The Marconi Research and Innovations Lab | Feb. 13, 2023
Artificial Intelligence (AI) has been the ultimate buzz word over the past decade due to its disruptive effect in virtually all sectors including healthcare, marketing, banking, logistics, cybersecurity, and the list could go on and on.
However, there has been a slow adoption of AI in power systems. Could this be due to the high levels of skepticism exhibited by engineers!? Coming from an engineering background, I can attest to how little trust engineers have for unexplainable things. Yet their skepticism is justified because most AI solutions so far act as black boxes — we simply can’t tell exactly what is going on inside these systems.
What is artificial intelligence?
Depending on who you ask, you will get different answers to this question. So how about we go back to 1955 where it all started? John McCarthy coined the term artificial intelligence and defined it as follows;
Every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves.
In brief, you could say artificial intelligence is a field that aims to train machines to simulate human intelligence. Some of the fields suggested by McCarthy included automation, natural language processing, neural nets, self-improvement, randomness and creativity, abstraction and problem complexity calculation.
Yes… All this was discussed close to 70 years ago. We can agree that these guys could see the future. Because today, AI has disrupted almost every industry. It is being used in healthcare, marketing, e-commerce, banking, entertainment, robotics, and driving cars, to mention but a few.
AI technologies such as machine learning and deep learning have proved to achieve and in some cases surpass human intelligence on different tasks. For instance, AlphaGo, a computer program trained to play the board game Go, defeated a Go world champion in 2016. In 2022 ChatGPT, a groundbreaking conversational AI that can generate comprehensive answers to questions was released. It can answer questions about any topic in great detail with coherence. This proved the possibilities of AI that can actually generalize across multiple topics.
Well, how about power systems? you may say How is AI being used in power systems?
What is a power system?
A power system is a network of electrical components that supply, transfer and distribute electric power to consumers. This includes power plants (which generate electricity), transformers (which step up or down voltage), transmission lines (that carry the power to consumers), and many other systems in-between.
Since electricity should be consumed immediately after it is generated, power systems are designed and managed in such a way that supply is always equal to demand. Any imbalance can result in drastic damage to the power system. Yet the power system is expected to be safe, efficient, reliable and simple. Faults on the power system should be quickly detected and isolated.
The field of power systems has a rich domain of knowledge spanning over 120 years of research. To a large extent, power systems can be accurately modelled mathematically with great detail. However, some challenges remain very complex to model using simple mathematical equations. In addition, the rise of distributed renewable energy systems such as solar and wind systems and their interconnection to the already complex grid has made power production planning even more challenging. This is due to the intermittent nature of most renewable energy sources.
Recently, power systems have adopted the use of Internet of Things (IoT) technologies to interconnect assets and infrastructure in order to lower costs and power consumption. These include SCADA, smart metering, and smart grids among others. This had exponentially increased the amount of data available about power systems.
AI applications in power systems
AI can be adopted to harness the vast amount of information embedded within the data collected from such IoT technologies to gain further insights into the state of power systems. Below we look at some of the possible AI applications in power systems.
Expert systems
This was the first application of AI in power systems and started over 40 years ago. An expert system is a system that emulates the decision-making ability of a human expert by reasoning through a large database of knowledge in terms of if-then rules. Expert systems have been applied in the diagnosis, control, design planning and scheduling of power systems. They yield great results in situations where there is no established theory, or in a scarcity of human expertise.
Forecasting
Since power systems must ensure a balance between power supply and load demand, a great deal of forecasting is employed to estimate both the power supply and demand at future time periods. Given the complex and dynamic nature of consumer behaviours and power output from renewable energy sources, AI can be employed to find patterns in the already existing data to accurately predict both the power output and load demand at various time horizons, from intra-hour to day ahead horizons.
Maintenance
Power systems are designed to mechanically detect and isolate faults with great speed and accuracy. However, AI can help to detect possible faults even before they happen. For instance, various work has been done to monitor transformers and detect faults based on the sound from the transformer.
Drone technologies have also made it possible to monitor the health of transmission lines, especially in remote areas where manual inspection is impossible. Dangerous maintenance activities at high voltage can also be successfully completed with the use of robotics and computer vision with a high degree of accuracy and precision.
Challenges of AI in power systems
AI systems are data-hungry — they require a vast amount of data to learn. This is challenging given the limited data available on power systems. For instance, there is very little data on catastrophic faults as compared to that available in normal operating conditions. Further, the collection of this data can be expensive, for instance for every fault data sample collected, an actual machine must be damaged.
Most AI applications are still black-box in nature and non-deterministic. This is concerning especially since any error in a power system can cause catastrophic consequences.
In conclusion, AI has the potential to greatly improve power systems. And with the growing amount of data captured about power systems and the growing need for renewable and distributed energy sources, AI will be the enabler through automation, forecasting and expert systems.
I will leave you with a quote Bill Gates made in 2017,
“If I were starting out today and looking for the same kind of opportunity to make a big impact in the world, I would consider three fields. One is artificial intelligence… The second is energy… The third is the biosciences,…”
Article by Alvin Kimbowa
References
McCarthy, John, et al. “A proposal for the dartmouth summer research project on artificial intelligence, august 31, 1955.” AI magazine 27.4 (2006): 12–12.
Zhang, Z. Zhang, G. S. Hope, and O. P. Malik. “Expert systems in electric power systems-a bibliographical survey.” IEEE Transactions on Power Systems 4.4 (1989): 1355–1362. Fault diagnosis of transformer using artitificial intelligence: A review
Acoustic Emission-Based Fault Detection of Substation Power Transformer
Fault detection and diagnosis in power transformers — a comprehensive review and classification of publications and methods.
How AI will transform the energy sector
Artificial intelligence in the energy sector: hype, hallelujah or outdated?