There is no day that goes by these days where the letters A and I, which are said consecutively, do not cross the human consciousness. Once a concept that was only known to tech-focused minds, it has now catapulted into the mainstream. Yes, AI has well and truly taken over, with its advantages being harnessed left, right, and center.
And it is no different when it comes to the energy sector.
Historically regarded as an industry that is slow to take on and make use of emerging technologies in spite of literally powering modern life, the energy vertical has come on leaps and bounds in recent times, not just in terms of the sources of energy used but also when it comes to harnessing the advantages that come along with digitization as well as modern technology.
This is where predictive AI comes in. Not to get confused with generative AI- Gen AI, which has its main focus on creating new as well as original content, predictive AI happens to be a form of intelligence that makes use of patterns in historical data so as to forecast or classify, thereby offering actionable insights and also support when it comes to decision-making along with strategy formulation.
AI’s role when it comes to changing the energy sector
AI, as per a 2023 GlobalData report, is all set to revolutionize the potential when it comes to predictive maintenance, thereby enhancing system efficiency and even international energy security, all thanks to its capacity to pinpoint as well as resolve challenges before they even have an option to disrupt functions.
Thematic Intelligence: It is well to be noted that Artificial Intelligence in the Energy report shines puts forth the fact on how AI is altering the energy sector, thereby suggesting that AI will growingly be used to evaluate real-time data. Major findings go on to show that all this can be leveraged so as to detect as well as repair faults by way of using monitoring, thermography, and analytical technology.
Predictive AI plays a key role when it comes to increasing energy’s reliability
Colin Gault, who happens to be the Head of Product at Scottish energy software firm POWWR, has had a front-row seat when it comes to witnessing the emergence of applications for predictive AI in terms of maintaining and optimizing energy assets. He says that the advances in technology have been quick, but the issue has been when it comes to supplying the actual data. This is now being overcome because of the wider digital transformation of the industry.
Gault says that predictive AI is enabling to increase the reliability of the overall energy system, as it not just informs when an asset happens to be at a higher risk of sustaining damage and also in requirement of preventative maintenance but blends this with other data such as weather and traffic so as to support dispatching engineers to the site in the most optimum way possible.
These thoughts are also echoed by the Chief Environmental Sustainability Officer and Head of Data and Intelligence Solutions at NTT DATA in the UK and Ireland, Bill Wilson. Bill, who has 25 years of experience when it comes to consulting and extensive experience throughout the energy landscape pertaining to working for energy giants such as BP, EDF, and Trafigura, has also gone on to observe the power of predictive maintenance across other industries like transport, defense, as well as telecoms.
He adds that he would like to witness similar optimizations as well as problem-solving methods that are applied in the energy sector, thereby acknowledging its power and usage in the industry thus far, while at the same time noting it has a fair way to go when it comes to being used to its entire potential.
When it comes to manufacturing, audio information and audio learning models can partly take the place of the trained engineer who knows when a machine is not sounding healthy, he further explains. The same predictive techniques can be made use to gather data from numerous sources so as to inform condition maintenance. This needs less predictive power as it is more concerned with precisely understanding the entire condition of an asset based on many inputs. Sensors can go out to give false readings, but techniques like cohort analysis enable these anomalies to be identified by way of comparing each sensor with others that go on to do a similar role.
Part of this, and the one that can be argued as a small teething problem, is how the training data gets sourced and subsequently executed. As Wilson goes on to explain, traditional predictive AI goes on to rely on having enough exact telemetry data, clean information when it comes to fault occurrence, and also relevant context. If the incidence of faults happens to be rare, this makes it even harder to train AI models, but techniques like random forest can be made use of to mitigate this issue.
Recently, as per Wilson, drone surveys when it comes to visible assets have begun to complement this approach, while generative AI can go on to produce additional training data that helps computer vision models to be trained to detect faults.
Predictive AI’s role when it comes to predicting energy forecast patterns
Gault happens to be a firm believer in the idea that the shift to net zero disrupts the supply side as well as the demand side of the energy system.
For instance, EVs, residential solar, and electric heating happen to be continuously changing demand shifts, he shares. At the same time, a surge when it comes to renewables on the grid is leading to fluctuations in supply capacity. Apart from this, more frequent extreme weather events and one can create even more challenging supply and demand patterns.
By being able to being better forecast when the energy system will experience imbalances in supply as well as demand means that the EV charging can be scheduled better so as to ensure the balancing of the grid, with the reward being cheaper electricity, and if the charging can be in sync with when there happens to be a renewable energy supply, then the output of CO2 can also be decreased.
Wilson says that the changing dynamics of battery as well as EV storage, and also the activities of prosumers happen to be making the electricity grid increasingly unpredictable. In a spectrum such as this where statistical models are not able to deal with the number of unknowns, one can use machine learning models and begin to explore more novel or even alternative sources of data.
When it comes to his role, Wilson oversees how data that is generated by other sectors can be shared with the energy supply sector. For instance, a telco client is experiencing unanticipated extremes when it comes to power demand.
AI and energy efficiency happen to be a topic all by themselves, and there are numerous mature solutions. They happen to be a partner with a firm that makes use of AI so as to predict when file servers will be in demand and turn down the clock speed at other times, hence saving 15-30% of energy requirements.
He took part in an energy industry round table in which strong consensus was that AI happens to be more than a measure to enhance efficiency. It was indeed clear that executives looked at AI-driven optimization as a mandatory stop-gap while physical infrastructure catches up, thereby taking them towards a more flexible, higher capacity as well as a lower-carbon future.
Can predictive AI go on to mitigate risks to the energy sector?
A major part of working toward a lower-carbon future can be attributed to its effectiveness in other areas like the supply chain. This, in addition to other unpredictable challenges such as extreme weather events along with cybersecurity threats, happens to have a circular impact when it comes to the slickness of the sector’s operations and also its green credentials as a result.
Wilson and Gault both firmly believe that technology, like predictive AI, goes on to enhance the sector’s total resilience and subsequent operations.
Gault says that a big risk to the energy sector happens to be an energy imbalance. The ability to go ahead and precisely forecast is important to being able to mitigate the supply-demand imbalance. Extreme weather not only impacts supply as well as demand profiles but can also go on to damage power lines and prevent power plants from functioning properly. There are some innovative projects, like one by Scottish Power, that look to better predict when extreme weather events shall lead to power outages and where such outages will go on to occur so as to provide them with enhanced intelligence.
Wilson concurs, signposting the sector to understand that and to be conscious of the application of predictive AI, there has to be a process of subdividing the categories of risk on top of the others that have been mentioned before, such as the physical risks of climate change, which go on to put infrastructure at risk.
He goes on to elaborate, combining this with geology information that shows some buried assets happen to be more at risk than others.
It is well to be noted that other AI systems will be responsible when it comes to predicting the greatest pinch points in the grid, so that fresh developments do not create local resilience challenges. The UK energy grid happens to be highly resilient today, but AI can go on to help maintain that resilience while at the same time dealing with rising renewable generation and its fluctuating output. One only requires fossil fuel capacity online when it is actually needed, but it takes time so as to bring this capacity online.
This is where a detailed AI forecast on supply and demand can go on to help optimize this process, Wilson says, adding that as this trend happens to be more widely adopted, the energy sector will be able to make alternative dataset usage from outside the sector so as to make it stronger and even more resilient than before.
Gault concludes that the usage of predictive AI happens to be already prevalent in numerous new projects, but it is still in the emerging technology bracket and requires overcoming the challenges in terms of scaling up. That is where the digitalization of the energy sector will focus in the coming year. The sector has started to envision a digital twin of the energy system in which predictive AI as well as open data go on to combine to better plan as well as function a much more distributed and flexible energy system.
