AI for the Renewable Energy Industry: The Present and Future

Intro

Artificial Intelligence is a new technological development with the potential to disrupt industries all over the world. Not since the arrival of the internet has there been a new technology with such broad applications and such disruptive potential. But just like the early days of the internet, it is difficult for us to see into the future how this new technology will transform our lives and our work.

As a renewable energy specialist, I am excited to understand how Artificial Intelligence can be used in our industry to make our lives easier and give us new tools to assist the renewable energy transition.

?

In this article I will first define and give context to the term AI so we can understand the ways it can be applied towards the renewable energy industry, next I will describe some of the ways AI is already transforming the renewable energy industry, and I will highlight some areas with high potential for AI to disrupt the renewable energy industry in the future.

?

This article can be looked at in two halves: the first half is defining AI and making the term clear, and the second half focuses on the renewable energy industry. If you are not interested in the definition of AI and some of the technical aspects of this new technology, feel free to jump straight to the section about AI for the Renewable Energy Industry on page 6.

2. What is AI?

AI is the acronym for Artificial Intelligence. AI is a transformative step in the evolution of technology that gives computers much greater ability to analyze data and respond in a way that was previously the sole domain of human intelligence. AI gives computers the ability to take large and diverse datasets and to, very quickly, make an “intelligent” analysis of the data and respond accordingly with “decisions” based on the analysis.

?

According to TechTarget online magazine,

“Artificial intelligence is the simulation of human intelligence processes by machines, especially computer systems.”

?

AI is not a single technology, but rather a set of mathematical models and algorithms that comprise certain techniques of extracting insights from large datasets, and to identify patterns within the datasets, and to use logic to predict the probabilities of potential outcomes of complex scenarios. Because of the massive data sets that AI can process in a very short time AI is effective at spotting patterns in large data sets and optimizing processes, and AI can also give insights into operations and industries that might not have been clear before.

?

AI is often confused with automation, but they are not the same. Automated systems perform repetitive tasks following a set of programmed rules and conditions, while AI can identify patterns and insights in massive data sets. AI also has the ability to “learn” through feedback over time, to identify patterns more accurately and effectively. This feedback mechanism is a subset of AI known as “Machine Learning” or “Deep Learning” and represents some of the latest developments in AI to this day.

?

Recent developments in AI, like machine learning, have unlocked a world of new potential for applications of AI, that have actually been built on decades of progress to reach this point. The 1990s saw an explosion of computational power available and the collection of data sets that are too large or complex to be dealt with by traditional data-processing application software, known as “big data”. This sparked a renewed push for the development of computer processes that could mimic human intelligence with the added memory and instant analysis benefits of a computer.

?

These developments led to such memorable milestones in AI as IBM’s Deep Blue defeating the Russian chess champion Garry Kasparov in 1997, becoming the first computer program to beat a world chess champion. Perhaps this provides a good illustration of AI because we can understand that the program did not have true intelligence, but rather it could analyze the many complex patterns and probabilities on the chess board and make probabilistic decisions according to pre defined goals and rules.

?

In the 2000s further developments in AI allowed the launch of Google’s search engine which is able to interpret, analyze and match user input to millions of sources. At this time AI developments were also used for Amazon’s product recommendations feature which could analyze and interpret massive data about a customer’s preferences and search history and other factors to suggest products that this customer might have a strong interest in. Text-to-speech features, and some of the first self driving cars were soon to follow. These examples illustrate the ability of AI to work very quickly with datasets that are just too large for humans or for traditional software to handle effectively.

?

Now in the 2020s we have reached the development of Generative, Pre-Trained Transformers (think Chat GPT). I will share a bit more about GPT tools and the modern AI applications in a bit.

?

Data Inputs

AI applications can be classified and further understood based on the data inputs they use. AI can use many forms of input data, including audio, speech, images, videos, data gained from sensors, data collected manually or robotically, etc.

?

?

The majority of AI applications fall in the following categories of data:

·???????? Market data / Equipment and sensor data: AI can use data collected from a variety of sources. This data could be from readily available data like the stock markets or weather data, and it could also come directly from smart sensors and other equipment that can provide real-time dynamic data. As already mentioned AI is able to recognize patterns and make probabilistic predictions based on patterns identified within the data. When used with sensors and smart equipment that have networking capabilities and digital connectivity, this gives AI the potential to coordinate physical assets over a wide geographic area almost instantly based on complex data collected in real time.

?

?

·???????? Images and videos: Also known as “Machine vision”, AI has the ability to use images and video data to recognize objects or patterns or conditions and make analysis based on this data. For example weather patterns and cloud conditions can be recognized and used to predict future weather conditions in a certain location. Some advantages of this AI over human vision are that it can happen 24 hours a day, 7 days a week with no breaks or dips in accuracy due to fatigue. Machine vision is also not bound by our biology, and it can be programmed to use data that would be invisible to the human eye, for example thermal imaging and infrared can be used to “see” through walls and other phenomena normally invisible to us.

?

·???????? Natural Language Processing. NLP allows the computer to interpret and analyze human language. This could be speech, or written text. A well known example of NLP is spam detection; it looks at the text in the subject line and email and decides if it is spam or not. NLP is a feature made possible by advances in machine learning. Now NLP tools include speech recognition, speech to text tools, translation tools, and the natural language prompts and responses that we see now with the advent of GPTs. (see below).

?

?

Data Outputs: Automation vs Aid

Another way of understanding the applications of AI is to be aware that AI can aid human decision making by providing information and recommendations for a human to act on, or that AI that can make decisions and take actions by itself without human involvement: automated decision making, or aided decision making.

?

Automated decision-making allows the AI algorithms to make decisions and take action based on the insights and patterns recognized within the data. This process eliminates the need for human intervention, which enables the AI to tackle complex tasks much faster than a human could.

?

However, automated decisions can be susceptible to various forms of bias, potentially leading to erroneous outcomes. Such inaccuracies can pose significant risks in critical systems, highlighting the limitations of automated decision-making and its unsuitability for certain applications.

Aided decision-making mitigates the drawbacks of automated decision-making by placing the responsibility of the final decision on human judgment. In this approach, AI serves as a valuable tool to enhance the decision-making process by providing insightful data and analysis. While the huge volume of data created on a daily basis would bury a human researcher, AI applications using machine learning can take that data and quickly turn it into actionable information

This collaborative approach leverages the strengths of both AI and human expertise, allowing for greater flexibility and informed decision-making.

?

Most implementations of AI will be not fully automated, and will simply improve products or services by providing insights and recommendations to the users. This has the benefit of applying the user’s human knowledge and experience on top of the patterns and probabilities that AI has recognized.

?

?

Modern AI

Two new technologies that come under the umbrella of AI are Machine Learning and Deep Learning. Machine learning enables software applications to become more accurate at predicting outcomes without being explicitly programmed to do so. Machine learning algorithms use historical data as input to predict new output values. This approach became vastly more effective with the rise of large data sets to train on.

?

Deep learning, a subset of machine learning, is based on our understanding of how the brain is structured. In very simple terms, can be thought of as the automation of predictive analytics. Deep Learning makes use of digital neural networks. A Nueral network is a way of processing data that is inspired by the human brain.

?

According to Amazon AWS:

[A Neural Network] “is a type of machine learning process, called deep learning, that uses interconnected nodes or neurons in a layered structure that resembles the human brain. It creates an adaptive system that computers use to learn from their mistakes and improve continuously. Thus, artificial neural networks attempt to solve complicated problems, like summarizing documents or recognizing faces, with greater accuracy.”

?

A neural network is much better suited to learn and model relationships between input and output that are complex and nonlinear. For example, a neural network is what could allow a computer to know that Arthur Street is a place, but Arthur Smith is a person’s name. A neural network is what could allow a computer to understand that two different sentences could have a very similar meaning, like:

“Can you tell me how to get to the hospital?”

“How do I get to the hospital?”

And a neural network can also make possible self driving cars, and natural language processing. In fact neural networks are the foundation of many of the technologies and features we have discussed, like machine vision, speech recognition, and recommendation engines.

This has led to AI that can do a wide variety of tasks, including

·???????? produce extremely detailed and realistic images and videos based on simple prompts, many of these are virtually indistinguishable from a real image or video,

·???????? camera technology that can recognize and interpret facial expressions,

·???????? real time audio translation, and

·???????? technology that could allow robots to move and interact with objects in a factory or warehouse setting, and much, much more.

A recent development in AI technology that has brought AI to the fore front of discussions was the introduction of the GPT, most notably Chat-GPT.

OpenAI?launched ChatGPT in November of 2022 and almost immediately became the fastest growing user base of any application in history, reaching over 100million users in less than two months. The introduction of Chat GPT also triggered a chain of new innovations in AI. Vendors such as OpenAI, Nvidia, Microsoft, Google, and others provide generative pre-trained transformers (GPTs), and countless other tools are appearing everyday that make use of these GPTs in some way or form

But what is a GPT, and why was this introduction so important?

The reason that the introduction of GPT was such a big event is that a GPT allows almost anyone to have access to AI. GPT stands for Generative Pre-trained Transformers. Previously, enterprises would have to build and train their own AI models from scratch. This is very difficult, and very expensive, and the expertise is still lacking so there is high risk with this challenge.

A GPT, however, has a built-in large language model behind it, and allows users to use natural language prompts to assign tasks to the AI tool. In other words, the tool is pre-trained on a large language model (hence the P in GPT). And the responses can be entirely new creations, not just copy and paste from an existing source, meaning the AI can generate all new answers unique to the prompt (hence the G in GPT).

So now, for the first time, all of us are able to use an existing AI tool that is pre trained on a large language model and can generate new responses based on how well we are able to fine tune our prompt.? The arrival of GPTs have dramatically reduced the cost, the difficulty, and the risk of developing proprietary AI tools, and everyday more industries, businesses and individuals are using AI.

Before we look at ways to apply AI to the renewable energy industry, lets look at how AI is being used in a few other industries.

AI in customer service.? Chatbots using AI are now quite common, and are available for businesses at a lower cost and lower implementation risk than ever before. AI chat bots are trained on large language models to be able to understand natural human language questions, and to use data and rules specific to the company that is using the tool, and therefore able to make probabilistic predictions about how to help the customer based on analysis of the order of their words. AI chatbots have scored higher customer satisfaction scores than humans and certainly shorten response times in many cases.

AI in healthcare. Hospitals and Doctors and private companies in the healthcare industry are using AI to make better and faster medical diagnoses than humans can. AI can analyze images including X-rays and MRI’s and compare vast amounts of data to provide an insight or make a recommendation that could have been slow and difficult and expensive for a human doctor to make. There are also obscure patterns and data points that the AI could possibly identify to help make a more accurate diagnosis.

AI in education. AI can assist with generating original questions and examples to illustrate a concept for students to learn. AI can automate or assist with grading, even essays and long form responses. AI can?provide a more comprehensive assessment of student’s proficiencies and weaknesses and help them work at their own pace in their areas of need.

AI in finance.? AI applications are able to collect personal data and provide financial advice. AI can provide insights into stock markets and currency markets.

AI in law.? AI could sift through lengthy legal documents and provide summaries and insights based on the text. AI could interpret requests for information, and other time consuming processes, potentially even assisting lawyers in making a legal argument based on the context of the case.

AI in software and coding AI is able to generate working code based on natural human language inputs. This is already disrupting the software coding space and will continue to get more and more complex.

AI in transportation. Beyond its pivotal role in powering autonomous vehicles, AI technologies are revolutionizing various aspects of transportation, including traffic management, flight delay prediction, and optimization of ocean shipping safety and efficiency. In supply chains, AI is supplanting conventional forecasting methods, enabling more accurate demand predictions and disruption predictions.

?

Now that we have established this understanding of what AI is and what are some of its parameters and use cases, we can finally address the question that inspired this article:

?

3. How can this new technology be leveraged by the solar and renewable energy industry?

The first important part of applying AI to the renewable energy (or any industry) is industry is DATA. AI can make its biggest contributions when it is fed enormous amounts of data. For the renewable energy industry this means using “smart” equipment that has networking capabilities, and the installation of more sensors throughout the entire renewable energy ecosystem—from component level sensors in residential energy consumers, through the distribution network, and the transmission network, and within generation assets. For AI, the more sensors the better, and the more accurate and fast the data can be collected the better the AI insights and AI value will be.

?

The second important part of applying AI to the renewable energy industry will be to use “smart” equipment that has two-way communication technologies and remote control systems. This will allow AI computer systems to automate some important functions much faster than a human would be able to.

?

Data, software, and automation already play a significant role in the energy sector; however, the AI revolution will allow the data to be processed and analyzed and acted upon at real-time speeds and with astonishing accuracy.

?

With these two principles in place, let us now discuss specifically how AI will transform the renewable energy industry.

?

According to the World Economic forum:

AI has tremendous potential to support and accelerate a reliable and lowest-cost energy transition, with potential applications ranging from optimizing and efficiently integrating variable renewable energy resources into the power grid, to supporting a proactive and autonomous electricity distribution system, to opening up new revenue streams for demand-side flexibility - WEF

?

You may be surprised to learn that AI is already at use in several aspects of the renewable energy industry, and there is ongoing research into new applications that have transformative potential. I have identified six areas where AI has high potential to reduce costs and provide value:

?

·???????? Renewable power generation and yield forecasting:

As renewable power generation grows, both in absolute terms and as a share of the power supply, it will become essential to have more accurate predictions of the solar and wind power generation, to improve capacity factors, improve return on investments, reduce risks by accurately forecasting power supply.

?

·???????? Reduction of costs in planning, developing, construction

The planning, developing and construction stages of renewable energy plants involve complex variables and timelines, and present a strong opportunity for cost reduction with the use of AI.

?

·???????? Reduction of costs of O&M / Safety

Ai can be used to do predictive maintenance, and condition based maintenance that can be more accurate, more efficient, and more cost-effective than calendar based maintenance. This also has implications for safety as disruptions and disasters can predicted, reduced, and avoided.

?

·???????? Grid operation and optimization

Current plans to reach net zero by mid-century imply a massive increase in renewable generation capacity and expansion of transmission infrastructure within a relatively short period of time. Using AI to optimize grid operation and enhancing the capacity of existing transmission and distribution lines, as well as extending the lifetime of existing equipment, will be key to supporting the energy transition. In addition, in an integrated and decentralized energy system, responsibility for system optimization happens at both the higher and lower voltage levels, distribution grids become more important, and maintaining grid stability and ensuring the security of supply become more complex. AI can be helpful in grid planning to optimize infrastructure by extending the lifetime of expensive grid equipment and keeping the whole grid system stable, even as more renewables are integrated.

?

·???????? Energy Demand Management

Demand management is an important part of the renewable energy transition as it is a part of the shortest and least cost roadmap to a decarbonized energy sector. AI can be used to analyze massive amounts of data in the generation, production, transmission networks, distribution distributions networks, and to compare against time of day, time of year, geographic locations, and real-time weather patterns to optimize demand side management in ways that are simply not possible for humans alone.

?

·???????? and Materials Discovery and Product Innovation

AI can be used to model complicated new products and new performance parameters and new products and product features, which can dramatically reduce the cost of developing these new products and can ultimately accelerate the renewable energy transition.

?

A.??? RENEWABLE POWER GENERATION AND YIELD FORECASTING

Forecasting and analyzing solar system performance

With renewable energy like wind and solar that is intermittent, generating an accurate yield projection is very difficult. Even more difficult is to forecast what the energy generation might be at any particular time in the future.

This has some significant financial implications, as the yield over time is often a big part of Power Purchase Agreements (PPA) and Independent Power Producer (IPP) contracts. The inability to make more accurate forecasts also causes inefficiencies in the energy mix of a utility. Being unable to accurately forecast renewable energy generation prevents renewable energy from taking a larger part of the overall energy mix because more spinning reserve is required to cater for sudden spikes in demand or sudden dips in renewable generation. AI now has the ability to consume an incredible amount of data, from thousands of renewable energy sites around the world, and massive amounts of weather data, and a wide range of equipment data such as maintenance records, and repair needs, or likely upgrades, and it can find patterns to become more and more accurate, and it can learn to predict more accurately the renewable energy generation into the future.

The AI tools could also monitor multiple power generation sites at the same time and make predictions about the demand to find the optimal balance of energy production across a region and across multiple sources of energy. This would provide financial returns in the form of more cheap electricity consumed and less cheap renewable electricity curtailed, and less expensive electricity consumed. There would also be financial returns to the industry in the form of more investor confidence in the yield projections of existing and proposed renewable power plants, which could help to mobilize investments around the world.

This effort to improve renewable energy forecasting is an effort that has been on going for some years now. In 2016, IBM acquired The Weather Company with the aim of providing the “most accurate weather forecasts globally with personalized and actionable results” by using advanced AI and big data.

The National Centre for Atmospheric Research (NCAR) is also a contributor to the effort to use AI to make advances in renewable energy forecasting, to produce outcomes that have an economic value.

A more specific example is regarding solar plants that use bifacial modules and a single axis trackers. According to several research papers, using bifacial modules and single axis trackers can produce the lowest LCOE of any other solar plant configuration. However, the bifacial modules and tracker each add several layers of complexity for the computer programs that make the yield projections. Bifacial panels are solar panels that produce energy from the top and bottom of the panel. Because of the bifacial panels, now the production will be greatly affected by the albedo of the environment around the solar panels, and the albedo is affected by the weather in very complex ways, for example with rain or snow cover (Albedo is the measure of the amount of light reflected by a particular surface). And because of the trackers, there is an element of movement that affects both sides of the module, and is itself subject to complicated variables to predict generation. Despite the research that supports bifacial single axis configuration, the financial sector has occasionally shown a level of skepticism regarding the yield projections in this configuration due to the additional complex variables that contribute uncertainty to the range of outcomes.

Using AI to improve yield projections for bifacial single axis tracker sites is an example how more accurate forecasting could unlock investments and more efficient renewable energy plants. For more information about this topic please see my articles on Bifacial modules, and Solar Tracker Systems, available on Linkedin.

?

B.???? REDUCTION OF COSTS IN PLANNING, DEVELOPING, CONSTRUCTION

Planning, developing and constructing renewable energy plants is costly, difficult, and often quite slow. The high capex has been a deterrent, and the competitive PPAs mean that it is imperative to find was to reduce the cost of planning, developing and constructing these plants.

Site selection.?

For renewable energy plants like solar and wind, the generation and the return on investment is greatly linked to the physical location and layout of the plant. Therefore site selection is of utmost importance to the financial and commercial success of the project. Site selection depends on many complex variables, including weather patterns, the terrain, competing uses of the land, distance to grid connection, local land and environmental regulations, grid congestion, demand forecasts and many other complex variables.

As you can imagine, AI can be used to analyze vast amount of data and find patterns and insights to assist in the site selection of renewable energy plants, and this can have a positive impact on yields and returns and therefore help to accelerate the renewable energy transition.

Using AI, these complexities can be managed more quickly and efficiently, while minimizing project costs. In fact, the United States Department of Energy announced in 2020 an initiative that provided $130 million for research into solar technologies, and $7.3 million was specifically for machine-learning solutions and other AI for solar applications.

This includes funding for a project that will use a “spatiotemporal graph neural network model.” This approach means the goal is to use AI to measure and predict solar production data from around the globe according to the plant’s location on earth, as well as over time taking into account newer versus older installations, and to learn best practices from this exercise. The lessons learned will be used to build a pre-trained computer model that will be able to improve all of the individual PV plants in the group, as well as future systems yet to be built.

Pre-construction planning and design.?

After selecting the site, and before even breaking ground, AI-powered design tools can help contractors to optimize construction plans according to equipment delivery schedules—even taking into account global logistics concerns, labor availability, cash flow, project planning schedules and other complicated variables. This can mitigate the possibility of delays and expensive changes to the project plan and scope. Large and complex projects such as solar and wind power plants can reap significant benefit from AI supported construction optimization which can maximize the use of on-site resources, and this starts at the planning stage.

AI can also provide a “site twin” that can be used to model different scenarios and inform decision making early in the project. This can help to optimize system design to take into account safety standards, minimizing equipment costs, maximize overall efficiency, and to make O&M as easy as possible, which can ensure an efficient and cost-effective installation.

Construction

When construction starts, AI tools can identify options for redeploying resources and keeping efficient use of the budget and calendar by suggesting options for task, equipment, or labor sequencing to keep projects moving forward.

Even as the plan changes and unforeseen events occur like supply chain issues or interconnection delays, the AI tools can adjust very quickly and provide a tremendous advantage in managing complex construction projects.

?

C.??? REDUCTION OF COSTS OF O&M / SAFETY

One of the main use cases today for Ai in the renewable energy industry, is AI aided O&M, which is also related to safety.

Especially when used with a large number of smart sensors throughout the energy ecosystem, AI can provide valuable insights and analysis that can help to reduce the cost of O&M, and to increase the effectiveness of the maintenance tasks, which will thereby reduce downtime and reduce failures and increase safety.

?

By analyzing vast amounts of equipment data, and vast amounts of data about the use of the equipment AI can help lead the transition from calendar based maintenance to condition based maintenance. Calendar-based maintenance is maintenance based on a regular schedule, it does not take into account the actual conditions of use and depreciation of the equipment. It is necessary that calendar based maintenance must be more frequent than is essential because it seeks to detect and prevent disasters and expensive failures before they happen, so it is better to err on the side of caution. This O&M is expensive, and it is not fail-safe. It can only account for maintenance that can be done relatively easily, and it usually requires some amount of downtime. Condition-based maintenance supported by AI can be much more efficient and accurate. It could reduce the frequency of maintenance tasks by looking to key parameters that would indicate the need of maintenance on a certain component, and these parameters can take into account thousands of data sets and patterns in real time thereby increasing reliability, reducing risk, reducing down time, and increasing efficiency and effectiveness of the O&M.

?

Some examples of AI already being used in safety and O&M tasks is that AI is being used to analyze voltage, current, temperature, usage patterns and much more data to detect anomalies in solar arrays and battery packs to identify and prevent fires. When a short circuit occurs, there is a high risk of fire and damage to equipment. AI can analyze the data and identify patterns and detect anomalies to prevent fires. But there can also be signal interference from the electrical devices within and around these components. Therefore AI has been using Machine Learning to get more and more accurate at detecting potential fires while reducing the false alarms.

?

Another example of AI being used in the renewable energy industry in order to improve the efficiency of O&M is the use of drones or fixed-wing aircraft to do surveys of solar plants. A large scale solar plant can often cover hundreds of acres, with hundreds of thousands (even millions) of solar panels. It is very difficult to identify faulty panels within such large arrays. A relatively new technique involves flying an aircraft over these arrays, and using different camera technologies like thermal imaging, to gather data about the panels and their production. AI tools can then analyze the visual data and compare against many other relevant data sets to identify patterns and identify faulty solar panels within the massive arrays. These advanced tools can identify a wide variety of solar panel faults, such as dust cover, cracks in the glass and many others. Some of these faults are invisible to the naked eye and could only be detected with the assistance of thermal imaging or high resolution cameras or some other technology. Once the faulty panels are identified it is quite easy to repair or replace the panels. Using AI is a vast improvement over manual and calendar based maintenance.

?

Another example of AI tools being used in O&M comes from General Electric (GE), one of the leaders in global power generation. GE has already allocated vast resources towards AI maintenance tools. Their software is called ‘Predix, and it combines many AI tools and functions to enhance O&M. One method is known as ‘digital-twin’, which creates a digital replica of a power plant in great detail using real-time and historical big data processing. This allows engineers to run simulations and predictions that make O&M more efficient and safe.

?

D.??? GRID OPERATION AND OPTIMIZATION:

A significant part of the renewable energy transition that is often overlooked and under appreciated is the transmission grid. There are some who may think that we only need to build more renewable energy generation and in that way we will decarbonize the grids. But the transmission grids are a key part of renewable energy planning and operation.

Modern transmission grids are very complicated. They combine multiple sources of energy and from dispatchable sources of power like hydro and thermal plants, as well as variable sources of energy from solar and wind, and sometimes batteries are involved, and they distribute this energy to the distribution network across a wide geographic area. Grid operation involves careful planning, almost constant maintenance, emergency response, as well as monitoring congestion and capacity constraints that are impacted by temperature and weather conditions, and a plethora of other factors.

Nextracker is a leader in solar tracking equipment. In their whitepaper titled “Making Solar Smarter”, they say:

“AI will undeniably be at the forefront of modernizing the grid in order to make it “smarter”. These vanguard technologies will be able to continuously gather and contextualize astonishing amounts of data from millions of sensors to optimize the decision-making process regarding the allocation of energy resources. Consequently, specialized microgrids will supplant enormous regional grids, which can also be combined with advanced battery technologies enabling power to flow continually to and between local communities even in cases of extreme weather.”

By analyzing massive data sets from equipment sensors and many other types of sensors, AI can help to optimize the grids by making instant decisions based on real time information and pattern recognition from voluminous datasets. For example, AI could enhance the utilization of the transmission capacity of a power line by responding to real-time temperature measurements to determine the upper limit that the line can safely carry instead of using static limits based on theoretical and conservative temperature assumptions.

AI also offers an additional benefit to monitoring the grid stability, by inferring and providing missing information that current sensors are unable to provide, through probabilistic interpretations of massive data sets that AI is ideally suited for.

The result will be less downtime, less energy wasted, and more efficient use of transmission assets. Theoretically this could increase the transmission capacity of existing grid infrastructure. This also would lead to reduced curtailment of renewable energy, and therefore reduced consumption of fossil fuels.

AI can also help to streamline interconnection for renewable energy projects, and to assist in planning grid expansion and grid upgrades. Renewable energy generation is often intermittent, and variable throughout the day and throughout the year. This requires careful planning for new projects regarding their location and their scheduling among other energy related projects. Improved forecasting, which we already discussed, can also help to optimize the planning stage, but careful consideration of the transmission network is also crucial to proper renewable energy planning.

The interconnection process is often delayed by problems such as permitting delays and transmission congestion AI’s ability to accurately assess and analyze massive quantities of complex data – combined with predictive abilities that allow it to suggest innovative alternative pathways – can make it incredibly valuable to the interconnection process

E.???? ENERGY DEMAND MANAGEMENT,

The future energy system will be more efficient and more cost-effective if distributed energy generators like rooftop solar sites, and energy consumers can participate in grid balancing and power quality optimization. Today, some large industrial equipment and grid-scale batteries can already provide these services, but it is difficult for smaller devices to participate. Digitalization and AI offer two new ways to operate distributed energy resources as virtual power plants (VPPs):

Aggregation and orchestration:?AI can be used to aggregate and orchestrate small power plants and distributed energy resources to provide grid services that they could not provide on their own.

Automation and autonomy:?AI can be used to automate and control small distributed devices, such as refrigerators and electric vehicles (EVs), so that they can support the grid without disrupting their primary function. For example, an AI-enabled system could manage the charging rate of an EV to minimize costs for the EV owner and benefit the grid operator.

Overall, AI has the potential to make the future energy system more efficient, cost-effective, and reliable by enabling distributed energy resources to participate in grid balancing and power quality optimization.

AI can also be used to analyze solar power usage and forecast future energy demand. This is known as “demand forecasting”. This is important because poor demand forecasting can lead to power outages, brownouts, and renewable energy curtailment. AI systems can identify complex usage patterns and predict potential problems before they happen. By using historical consumption data, AI can provide insights into individual and collective consumer demand, which can be used to optimize the energy system.

Certain consumers could also be good candidates for demand scheduling. This is when certain loads can be diverted or delayed to off peak hours our to certain favorable grid conditions to ease grid congestion and facilitate optimal grid operation. And example of this could be Hyperscale data centers are particularly active in this example of “renewables matching”. In this way AI could help to facilitate the use of renewable energy that may otherwise have been curtailed.

At the consumer level, AI can also help consumers make the most efficient use of their distributed generation assets like rooftop solar. AI tools are already being used to learn a consumers’ habits and respond accordingly to reduce overall demand and increase the proportion of renewable energy consumed.

AI is helping manufacturers to integrate their solar equipment more seamlessly with other technologies in homes like smart thermostats, temperature sensors, and lighting systems. This allows customers to use solar energy more efficiently and effectively throughout their homes, which can save them money.

?

In commercial buildings, AI can optimize the electricity usage of heating and air conditioning units, for example, by using sensor data and computer vision to determine occupancy levels and better understand a building’s thermal behavior and conditions. AI is useful not only in reducing power demand but also in shifting it to match times of high renewables generation, allowing demand to follow supply.

?

An example of a company using AI for demand side management is Huawei, with their AI device known as EMMA. This comes in Huawei’s newest residential solar and storage products. EMMA uses AI to recognize patterns in the homeowners energy consumption in order to optimize the balance of available and predicted solar energy, battery storage, and grid. The AI tool can use these patterns of renewable energy available, and consumption patterns, and grid costs to ensure the optimal balance of free energy from the sun, lowest cost energy possible from the grid to make sure batteries always have enough charge for the expected usage based on consumption patterns. This tool can also potentially provide grid support by loaning power to the national grids from residential batteries or EVs. When many of these intelligent units are connected you get what is known as a Virtual Power Plant, or VPP. This is a simulated power plant created by connecting hundreds or thousands of consumers who can predict and manage their power needs and can be coordinated simultaneously across a wide area to return power to the grid in case of an emergency or need for power balancing, while also maintaining enough available power for the consumption needs.

?

?

F.???? MATERIALS DISCOVERY AND PRODUCT INNOVATION

Artificial intelligence (AI) could be used to design new molecules with specific properties for different applications. This includes materials that are more energy-efficient, can convert heat into electricity, or can be used in solar panels to improve the efficiency of PV production, or batteries with improved performance and durability.

Also, as more consumer devices are connected to the Internet of Things (IOT) there will be an increase in data to understand consumer behavior and how devices are used. The conventional product designs are largely based on static modeling and laboratory results. Furthermore, given the recent advancements in ML, solar hardware could potentially become capable of troubleshooting without human intervention.

This data can be used to improve the design of solar hardware and to develop new AI-powered features, such as self-troubleshooting systems and tracking systems that can adjust to weather conditions.

Overall, AI and machine learning (ML) are poised to make solar equipment smarter and more efficient.

A specific example, is again from NEXTracker. In July 2017, NEXTracker introduced a new tracker control system called ‘TrueCapture’, supported by the predictive BrightBox Technologies, described as “an intelligent, self-adjusting tracker control system that increases typical PV power plant energy by 2-6%”. TrueCapture continually improves the tracking algorithm of each row according to site characteristics and weather conditions, whereas standard systems track all the rows equally.

?

?

4. Economic Value of AI in Renewable Energy

The economic value of AI for the renewable energy transition is difficult to estimate because it has the potential to improve efficiency and reduce costs across the entire value chain, and to enable new revenue streams through new business models. Also, many of the potential benefits of AI that we have discussed will come in the form of avoided costs, such as reducing O&M costs and reducing equipment replacement costs through predictive maintenance.

However, when we consider the levels of investment that will be needed to achieve the global stated energy goals of electrifying most of the population and decarbonizing most of energy generation, we can see that the benefits of AI assisting this transition will be massive. If AI were able to reduce the required investment, or to reduct the peak energy demand by even a small percentage, this would result in billions of dollars of savings.

For example, according to the Bloomberg NEF’s net-zero scenario, there would be $1.3 trillion in savings resulting from every 1% increase of demand side efficiency.

?

5. OBSTACLES

There are certainly a lot of obstacles for adopting AI throughout the renewable energy industry, and other industries as well.

Lack of Knowledge

Most solar installers and operators don't know enough about AI to understand how it can help them. They may not understand what AI can do, or they may think they don't have the resources or people to use it in their businesses.

Cost

Another challenge is the cost of AI technology for solar energy projects. This includes the cost of the hardware and software, as well as the cost of hiring people who can develop and maintain it. This can be a big problem for small solar projects.

The introduction of GPTs as explained earlier has made it more possible than ever before to use AI quickly and cost effectively, so this obstacle will continue to erode away over time.

Explainability

One challenge of using AI in industries with strict rules and regulations is that AI results and insights can be difficult to explain. For example, in the US, banks must explain why they deny someone a loan. But if an AI program makes the decision, it can be hard to explain how it came to that decision, because AI tools use complex relationships between thousands of variables.

Risk

Some sectors are considered high risk to implementing AI due to the fact that small errors can have large consequences. The Energy sector is certainly one of these sectors where automated AI decisions, when wrong, could cause drastic damage and loss of power which in turn has great consequences. Therefore implementation of automated AI might come more slowly than in other sectors with much lower risk, like website design for example. But AI aided decision making is already in use in many ways, providing insights and recommendations to human enegineers.

?

Data Use / Data Privacy

Another obstacle for the use of AI in renewable energy, and other industries, is around the use of data and data privacy. As we have discussed, one of the key enablers of AI to make valuable contributions is to have massive amounts of data available to review and monitor. In the renewable energy industry, however, owners of renewable energy plants have normally been hesitant to disclose too much of their data from their power plants. This could be because such data is crucial for securing current and future investments into the development company. Also manufacturers tend to guard certain data as well, for fear that it could impact their sales and future equipment decisions. In short, there is a general lack of transparency and data sharing among key stakeholders because such information has strong financial implications for those stakeholders. There could be many benefits of data sharing among renewable energy developers and finaniciers and manufacturers, but the fear of the misuse of data, or the data falling into the wrong hands is a current obstacle and will continue to be an obstacle for the development of AI systems until and unless strong and comprehensive regulations about data use and data privacy are in place with the support and consent of stakeholders across the sector.

One possibility for addressing these concerns lies with another disruptive and relatively new technology: the Block chain.

Blockchain technology has the potential to revolutionize the solar energy industry by making it more efficient, transparent, and secure. Here are some of the ways that blockchain can be used in the renewable energy industry, with the potential to reduce data privacy risks, and more:

Peer-to-peer energy trading: Blockchain can be used to create a peer-to-peer energy trading platform where individuals and businesses can buy and sell solar energy directly with each other. This can help to reduce reliance on centralized energy grids and make solar energy

Green certificate trading: Blockchain can be used to track and trade green certificates, which are tradable certificates that represent the generation of a certain amount of renewable energy. This can help to create a more liquid market for green certificates and make it easier for businesses to offset their carbon emissions.

Solar panel financing: Blockchain can be used to create a more secure and transparent way to finance solar panel installations. This can make it easier for homeowners and businesses to afford solar panels and help to accelerate the adoption of solar energy.

Solar energy monitoring: Blockchain can be used to track and monitor solar energy production and consumption. This can help to improve the efficiency of solar energy systems and ensure that they are operating at their full potential.

Solar energy auditing: Blockchain can be used to audit solar energy systems to ensure that they are meeting certain standards. This can help to improve the quality and reliability of solar energy systems.

DCOMM is a company that is pioneering the use of blockchain in the renewable energy space through all these means just mentioned, and through tokenization. DCOMM says:

“Tokenization is the process of converting real-world assets or rights to a digital representation on a blockchain. This process allows for the secure, transparent, and efficient management of these assets in the digital realm”

These are just a few of the ways that blockchain can be used in the solar energy industry. As the technology continues to develop, we can expect to see even more innovative and creative applications of blockchain in renewable energy.

Legal Concerns

AI also raises some serious concerns about legality and liability. If an AI system makes a costly mistake, who should be held accountable? The owner of the equipment? The software developer who provided the AI system? The operator? Perhaps the owner or manufacturer of a sensor that provided erroneous data? There is still a lack of clear legal regulations and liability laws concerning the use of AI in new applications.

Safety / Security

Ai also presents the danger of being susceptible to hackers and other security threats that may be significantly increased over a manual system. Perhaps hackers could change the instructions of an automated AI system and cause great disruption. This could even potentially cause explosions and fires and put lives at risk. There is a need for a well thought out security system and contingency plans before AI can take a more prominent role in the energy industry.

?

Elimination of Jobs

Ai has the potential to eliminate thousands of jobs across the renewable energy industry and other industries. This could be in reducing on-site technicians due to improved O&M, and to reduce the number of designers and data analysts in the planning phase, and many other examples. This is not unique to the renewable energy industry, and it is a complex issue that must be addressed.

?

?

6. CONCLUSION

AI is a truly disruptive technology that has vast potential to revolutionize the renewable energy industry. By analyzing incredible amounts of data and recognizing patterns with speed that is simply impossible for humans, AI can assist the renewable energy transition in many ways. This would result in billions of dollars of cost savings, and increased speed towards a reduced carbon energy industry around the globe.

There are still significant obstacles that must be addressed, but it is clear that Ai will become an ever increasing feature of the renewable energy industry, and it is already being used in a multitude of ways. It is useless to resist this trend. As renewable energy specialists it is our duty to learn about AI and use it to the best of our abilities, and to mitigate the risks and challenges of using this technology.

I am hopeful for a future where renewable energy assets from utility scale to residential, and transmission and distribution assets can be optimized for a cleaner, greener, and more reliable future.

?

Here are some links to articles that I found interesting and which helped me to write this article:

?

https://www.techtarget.com/searchenterpriseai/definition/AI-Artificial-Intelligence

?

?

https://www3.weforum.org/docs/WEF_Harnessing_AI_to_accelerate_the_Energy_Transition_2021.pdf

?

?

https://ratedpower.com/blog/artificial-intelligence-renewable-energy/

?

?

https://www.nextracker.com/wp-content/uploads/2018/01/2018-US-AI-White-Paper-SAMNA.pdf

?

https://www.nrel.gov/docs/fy20osti/77593.pdf

?

https://www.powermag.com/the-power-interview-using-ai-to-optimize-the-power-grid/