By Alex Baldassano, global head of energy transition, Climavision
Energy transitions have been occurring globally since the 1800s. As demand patterns shifted and production technologies matured, the world witnessed an industrial revolution-fueled need for coal. This supply cornerstone gradually gave way to crude oil-based energy production in the early 1900s. Succumbing to the global oil crisis of the 1970s, the 1980s ushered in a transition toward natural gas. And now the term has again been dusted off and pushed into the popular vernacular by political heavyweights such as President Joe Biden and financial wizards like Blackrock’s Larry Fink. 2021’s COP26 created a worldwide platform linking climate change to a green-energy transition; a transition away from the fossil-based fuels that still account for over 80% of global use and toward renewable energy sources, like wind.
Altamont Pass wind farm
Wind is a critical component of the green energy transition solution set, but by no means a new one. Denmark’s 2-MW Tvindkraft wind turbine began producing power in 1978! Similarly, California’s Altamont Pass wind farm, comprised of over 4,900 individual wind turbines, became operational in the early 1980s. However, wind development in 2022 is faced with a new set of challenges reflecting a complicated global patchwork of advanced technologies, governmental subsidies, commodity market dynamics and, perhaps ironically, weather patterns that are bucking historical trends.
Stakeholders participating at different phases of the wind lifecycle are left asking where is the optimal place to locate a wind farm? Will there be enough wind (or too much wind) to produce an optimal power output? How can we maximize our financial return? How do we maintain the physical components of the wind farm to increase longevity and decrease outages?
Now all these critical questions can be answered by relying on math-based solutions. Specifically, the wind industry is relying on machine learning (ML) in all facets of wind development and operations. The effectiveness and accuracy of these adapting algorithms is underpinned by quality data. As the old saying goes, “garbage in, garbage out,” and this is especially true in the wind business. Site-specific historical data such as hourly max and min wind speed, wind gusts, wind direction, average wind speed, wind gradient/sheer, all at different heights is critical in assessing the ideal location of a wind turbine. Naturally, this site-specific data does not exist over every inch of the globe, including oceanic bodies. Instead, wind developers rely on data cultivated from nearby proxy sites, or weather stations such as government controlled automated weather observing systems (AWOS) typically located at airports. This data is aggregated and made available to the public, including wind developers, by government agencies such as the U.S. National Weather Service. Private weather radar, such as Climavision’s North American network, augments publicly available data sets.
At this point, increasing the confidence of where to site a wind farm, based on resulting power output as a biproduct of technological components and meteorological observations, will be achieved by relying on artificial intelligence (AI) machine learning techniques. The proxy data that is being ingested into the ML platform is growing every day, and coupled with recent climate change phenomena, ML becomes the only modern viable method to confidently identify where a wind farm should be sited today, next year and 10 years from now. The importance of constantly refreshed siting models cannot be understated as utility-scale wind farm development is complex and can take years from land/water acquisition, permitting, offtake negotiation, procurement and construction. A prime example is Deepwater Wind’s Block Island wind project, where conception commenced in 2008 and the project was finally operational and producing power in 2016.
Machine learning techniques are not only crucial in completing pre-COD activities like project siting, but also in securing bank financing. There is no shortage of global publicly traded banks that offer long-term debt solutions to wind developers. In fact, the supply of debt products has grown in recent years to include hedge funds, regional banks like Fifth Third Bank, and green banks like the NY Green Bank backed by local and federal governments. A key ingredient for a private wind developer (and any renewable energy developer, for that matter) to secure funds is a long-term power price curve. In other words, over the next 20 years, how much money will each wind-generated megawatt-hour command in the open market? This theoretical offtake energy price (which might include both electricity and renewable energy credits), multiplied by the forecasted generation quantity, will form the revenue foundation of any basic financial model.
Credit: RES
Banks typically require an independent third-party like Wood Mackenzie or DNV to provide a long-term power price curve. In recent years, some more risk-averse banks have even started requiring multiple independent long-term power price curves, which is a reflection of operational wind farm investments not performing according to initially modeled power and price production (think the ERCOT wind generation failure of February 2020 and the UK/Ireland wind generation failure of Summer 2021). The independent consultant groups that have become synonymous with creating long-term power price curves traditionally relied on 10- to 20-year historical power price data, coupled with liquid electricity and gas-forward curves to fuel their forward-looking models. Now the paradigm is shifting toward models incorporating nearer-term historical price data, highly localized weather forecasts from firms like Climavision, and the assimilation and cleansing of this data through machine learning tools.
If today’s wind project does indeed proceed through the siting, permitting, financing and construction development process, once it hits the commercial operation date (COD), machine learning takes on a new relevance for the entire lifecycle of the farm. At the point of COD, profit maximization will now be tethered to the physical ability of the wind farm components to harness the power from the wind and generate electricity. Independent power producers and utilities that own these farms need to ensure the equipment is in good working order, downtime is limited and maintenance occurs at the right intervals for the appropriate parts. Rotors, gearboxes, yaw drives, shafts and blades are in constant motion. Like a car’s engine, these moving parts require regular preventative maintenance to ensure the wind farm’s useful life meets and then exceeds the industry standard 25 years. Combining real-time, onsite SCADA systems from the likes of a GE, ABB or Schneider Electric with machine learning equates to a physical O&M plan enriched by a massive quantity of data points, high-integrity cleansed data, processed at incredibly fast speeds. Operational longevity beyond 25 years is now within a realistic reach.
Our current energy transition is not only witnessing a move from fossil-fired electricity generation to renewable resources like wind, but there is now a sub-transition occurring whereby intermittent renewable generation resources are being paired with nascent storage technologies. The implications of a modernized generation fleet comprised of wind farms paired with batteries is profound from a grid resilience and optimization standpoint. A wind farm can now be transformed from an intermittent to a baseload generation resource. Utilities, independent system operators and balancing authorities can more effectively match load with generation, strengthening grid performance and subsequently supporting low prices. The independent power producer can now generate, store and release electricity onto the grid when wholesale power markets provide real-time and short-term pricing signals. Reaping the rewards of a successful financial optimization strategy is incumbent upon an energy manager like Wood Mackenzie using machine learning to synthesize everchanging regional generation, load, transmission, weather and price data inputs. Machine learning effectively becomes a key ingredient in determining the most profitable quantity, location and time interval to store or release wind-generated electricity.
It is truly amazing that since the recent maturation of artificial intelligence and machine learning techniques, many of the critical wind development processes — including siting, bank financing, physical operations and financial optimization — now rely exclusively on highly sophisticated ML outputs. Machine learning is not only driving the successful development of wind farms, but also providing a critical foundation to the global energy transition.
Alex Baldassano is the global head of energy transition for Climavision, responsible for managing relationships with key clients and developing new opportunities in a rapidly growing market.
Mr. Baldassano is an innovative renewable energy leader with over 15 years of domain expertise. Alex brings a mastery of the complex energy ecosystem and proven business development track record to the team. Before joining Climavision, Baldassano was Wood Mackenzie’s head of energy management for the energy transition practice; a business he co-founded in 2010. In that capacity he led the renewables, electricity and natural gas consulting business. He also formulated restructuring and integration plans for two internal acquisitions, as well as creating green energy product offerings for over 30 North American energy suppliers.
Always the entrepreneur, while at Wood Mackenzie Mr. Baldassano also co-founded Green Resources Group, a utility-scale solar development company. There he created the go-to-market strategy and developed critical vendor partnerships.
In his early career, Mr. Baldassano was a Carbon Portfolio Manager for Natsource Asset Management where he managed all worldwide institutional relationships for a $500 million environmental hedge fund. Previously, Baldassano launched his sales and marketing career at Philip Moris USA where he received numerous accolades.
Mr. Baldassano received a bachelor’s degree in Economics from Haverford College and a master’s degree in Business Administration, Finance (M.B.A.) from Fordham University Graduate School of Business.
Filed Under: Featured