Artificial Intelligence and solar energy generation. The Elephant in the Room.

None of the above solar panels is being monitored individually.?If such was rectified, it would permit far earlier activation of manufacturer guarantees, and thus, improved energy production over many decades. Individual panel monitoring has been delivered in the field for many years, but not yet at scale. The solar industry needs to concentrate on surpassing nuclear and other technologies for BOTH cost and long-term performance. This is feasible today.

Artificial Intelligence, AI (Artificial Intelligence, Machine Learning & Deep Learning), is a very powerful suite of information technologies. It is deployed to interrogate large datasets iteratively to better predict outcomes in the future. It is totally data-type agnostic and thus possesses a literally infinite number of applications. Nonetheless, it is always dependent upon access to data.?

In the case of solar energy production, real-time data collection from hundreds of millions of individual solar panels found across disparate locations around the globe represents the future and Holy Grail of Distributed Energy Generation. Today, these data are neither collected or collated from the vast majority of solar panels. Neither are they being measured in real time. Field-derived baseline data and remote commissioning for these panels are equally non-existent.?

“This represents a splendid business case for solar panel manufacturers, but less so if you are a consumer of their products”.?

Solar panel manufacturing is dominated by Chinese concerns using predominantly coal-fired electricity. Their products are shipped internationally with little regard for ESG (Environmental Social and Governance) investment principals, while only a feeble percentage of panels are being recycled. Most will find their way to landfill following tardive asset replacement (revamping), well in advance of 30 years in the field. These issues have neither slowed or stopped investment in solar technology over the past decade. Indeed, the reverse was true. During this period, more than $2 trillion USD found its way into the sector, namely, a four-fold increase in the level of annual investment (https://www.ehn.org/renewable-energy-growing-2640193068.html ). Notably, the adoption rate of solar technologies is expected to accelerate even further near term.?

The current scenario is far removed from the central tenet of good management practice, such as Six Sigma, for example. Without data collection at the level of individual solar panels, expectations for long-lived, profitable, and dependable solar power production are merely naive and unrealistic. Currently, our efforts are focused on delivery at ‘least cost’.

“The absence of best-in-class management practices further exposes solar financiers to undue risk, particularly as one’s asset base ages. Energy storage is now routinely increasing solar project costs. This further exacerbates exposure to risk for equity-linked investors and credit providers."

A variety of detection technologies coupled with the Internet of Things already exist. Cost alone is a poor rebuttal for not better managing solar infrastructure. Large-scale high-cost infrastructure projects are regularly funded by a multitude of financial constructs. All to the benefit of disparate stakeholders. The world urgently needs a greater contribution to its energy needs from renewable solar energy.? We should not allow the potential for rapid growth and near-term financial gains to obscure underlying inefficiencies. Aging assets will necessarily produce increased levels of underperformance with respect to oft-times overly-optimistic financial spreadsheets. These need to be addressed through higher quality asset management.

The acquisition of millions of data-points in real time is capable of rendering our competition (nuclear, combined-cycle coal- & gas-fired, wind and hydro-) envious of solar's inherent capacity to one day manage their substantial infrastructure correctly. This could assist solar’s overall grid penetration.

AI is revolutionizing our lives, industry and commerce around the globe. Belatedly, the solar sector is starting to adopt these technologies. Unfortunately, this is happening mostly with respect to more spurious applications not directly linked to improving solar energy generation or its reliability. Several dominant trends have been reviewed recently by René Morkos from Stanford University. (URL: https://solarbuildermag.com/featured/artificial-intelligence-can-expand-solar-energy-here-are-7-great-examples/ ). These include: i) Solar site selection; ii) Pre-construction planning and design; iii) Construction cost reduction; iv) Overcoming construction delays linked to supply chain issues, labor shortages, interconnection delays, scheduling and rapid recovery; v) Streamlining interconnection with energy grids; vi) Weather forecasting and real-time adjustments that allow for improved planning, storage, and operational efficiency, elimination of unnecessary power wastage or shutdowns due to weather, environmental hazards, or mismatches in supply/demand, thereby reducing equipment malfunctions and damage; vii) Demand scheduling using historical consumer demand. AI can provide insight into consumer demand (both on an individual and collective basis), revealing data helpful to system optimization. Elsewhere, the World Economic Forum White Paper from September 2021 provides a more detailed overview of current trends, but here too the “Elephant in the Room” is being ignored (URL: https://www3.weforum.org/docs/WEF_Harnessing_AI_to_accelerate_the_Energy_Transition_2021.pdf ).

AI depends upon ‘Learning Sets’ of increasing size accumulated over time to better predict the future and thereby facilitate Informed Decision Making. Project managers of photovoltaic assets distributed around the world as Domestic, or Commercial & Industrial PV-Rooftop installation &/or solar parks of varying sizes on land or on water would welcome such advances. Learning sets featuring focused data collection, low dimensionality and large numbers of observations increasing over time provide the most insightful outcomes. Solar is well-placed to extract maximum benefit from AI.? Solar installations constitute near-perfect experimental design for data interrogation based upon very large populations of similar entities.

It would appear that we are currently not collecting the most relevant solar datasets. Information collected from central &/or string inverters that measure total energy production from hundreds or thousands of solar panels grouped together renders the relevant information 'totally opaque'. Performance measurements are, therefore, being acquired sub-optimally.?

“Our objectives from data generation, collection and interrogation should focus upon delivering : 1) Less down time; 2) Fault prediction, localization and prevention; and 3) Maximization of financial returns and environmental benefit.”??

This is only possible by analyzing the performance of the actual units of energy production, namely, individual solar panels: "The Elephant in the Room"

#mlpe?;?#esg?;?#GREEN?;?#solar?;?#energytransition?;?#solarpv?;?#renewables?;?#sustainability?;?#electricity?;?#climatechange?;?#Energy?;?#globalwarming?;?#climatech?;?#renewableenergy