How AI will transform the energy sector

Back in 2017, Bill Gates — co-founder of Microsoft and an avid philanthropist — penned a short blog post in the form of a commencement address for that year’s graduating cohorts. He outlined three fields that he thought graduates should apply themselves to if they truly wished to have an outsized impact on the future of civilization. The first two picks were artificial intelligence (AI) and energy (ahead of biosciences).

Interestingly, AI and energy are far more interrelated than people appreciate. AI is poised to transform the entire energy sector in the coming years, by helping overcome energy’s inherently variable and uncertain nature, and by accelerating the adoption of renewables.

One need only look at the power outages that crippled much of Texas in February of this year — attributed in part to freezing temperatures — to see that our electricity infrastructure is failing us, not to mention the uptick in extreme weather events is revealing the shortcomings of our current climate change strategy.

AI — broadly speaking the use of modern computing power to perform tasks that have traditionally required human intelligence —  is the true enabler of Industry 4.0. It will allow physical industrial assets to be interconnected and communicating with each other through the flow of vast amounts of data in real-time.

There are two main areas in which the implementation of AI methods can substantially improve the effectiveness of current solutions in the energy sector, and support faster integration of renewable energy sources:

Improving Prediction Accuracy

Machine learning (ML) algorithms can identify patterns and insights within large data sets, and predict outcomes given certain data inputs. This would allow energy companies to:

  • Anticipate energy demand. Increased accuracy in short-term demand forecasting can assist energy companies in improving production decision-making, enhancing dispatch efficiency, and reducing required operating reserves. For end-consumers, AI-based models may help reduce utility bills through the optimization of home solar and battery systems, while also shifting loads to periods when electric grids are less congested.
  • Mountain View-based startup Bidgely Inc., for example, uses AI to provide granular visibility and analytics into households’ energy generation and consumption. Through it, users can identify which electric appliances are being used and when, as well as how efficiently they are performing, enabling the offering of real-time energy saving plans, rate programs and rebates
  • Forecast weather conditions. Balancing different sources of energy in a renewable portfolio is no easy task, given their dependency on atmospheric conditions. Self-learning weather models, coupled with large-scale processing of satellite imagery and real-time measurements can significantly reduce renewable energy generation costs.
  • Jungle.ai, based in Lisbon, is one of the companies that have succeeded in producing highly accurate power forecasts based on deep learning models, asset telemetry and numerical weather predictions. In particular, solar and wind energy producers can leverage this offering to estimate generation capabilities and reduce the cost of variabilities in their power output.
  • Predict maintenance for infrastructure. Smart scheduling of check-ups for critical renewable infrastructure can improve their long-term performance and mitigate downtime revenue loss, estimated at around $50bn every year according to recent studies. Additionally, computer vision solutions that can virtually identify and predict defects have the potential to replace manual inspection procedures.
  • eSmart Systems, headquartered in Norway, is an up-and-coming player in this space, focused on the maintenance of electric utility infrastructure. It uses AI algorithms to analyze aerial and ground-based images of renewable energy units, enabling fast and reliable anomaly detection and repair.

 

Enabling Full Autonomy

The next step is capturing the output of all these predictions and acting accordingly, independent of human guidance.

The holy grail is achieving the full autonomy of energy systems.

Substantial advances in ML algorithms are opening possibilities beyond the mere automation of decisions based on the improved recommendations of AI-based models. The holy grail in the energy industry is achieving the full autonomy of energy systems – particularly of power grids, some of the most complex mechanised systems in the world.

They are becoming even more challenging to operate due to the advent of distributed energy assets (e.g., personal photovoltaic panels) and the rise of the prosumer, which are shifting supply and demand dynamics and turning the traditional energy value chain upside down.

AI-based deep learning models have the potential to automate the optimisation process of energy grids by analysing heaps of historic and real-time data, acting independently upon the output, and using feedback loops to self-learn and become even more accurate. This could be key to reducing grid congestion, integrating intermittent renewable energy sources, and enabling quick recovery in the wake of natural disasters.

During the Texas winter storms, AI-based models could have prepared for the subsequent outages.

 

For example, during the Texas winter storms, AI-based models could have predicted and prepared for the subsequent outages, autonomously triggering alternative energy source generators and swiftly dispatching the power to neighbourhoods that needed it the most.

GreenCom Networks — a leading provider of white-label solutions for distributed energy management — focuses on extending autonomous capabilities within energy systems. Its platform can independently optimise decentralised energy generation and consumption, and reduce overall grid congestion.

Expect resistance before widespread adoption

We know for a fact that AI can accelerate our shift towards renewable energies, but the road there is not clear of obstacles. First, these technologies are likely to face initial mistrust from sceptical consumers — both at the organization and individual level — due to their inherent “black-box” nature.

It is difficult to find entrepreneurs with the required expertise.

Second, there is currently a lack of in-depth knowledge of AI, given that it is still a relatively new technology. Today, it is strikingly difficult to find entrepreneurs with the required expertise to build holistic AI-powered software solutions that have real practical value to the energy industry.

Cyber-attack vulnerability and fear of critical infrastructure decentralisation will be challenges.

Third and most importantly, challenges will crop up on the regulatory side — cyber-attack vulnerability and fear of critical infrastructure decentralisation are set to be the main culprits. Just last February, the European Commission released a whitepaper calling for the regulation of AI in the energy sector – flagging its “high-risk” status, as well as inherent data security and governance issues.

Overall, the next few years are expected to witness an explosion in the number of new use-cases and areas for the application of AI models within the energy space. As costs plummet across all types of renewables, energy companies will hunt for technologies that can provide sustained competitive advantages over rivals. Interestingly, it is possible that the biggest opportunities will arise from developing countries where underlying infrastructure may not yet be built, and the lack of entrenched industry incumbents could help drive a relatively higher rate of adoption.

 

Original post: https://sifted.eu/articles/ai-energy-transform/

Leave a Reply

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