Why are we not making the most out of big data analysis?

A great series of insights by Everoze partner Ragna Schmidt-Haupt as published in PV Magazine International January 2019 edition.

Why are we not making the most out of big data analysis - Blog by Ragna Schmidt-Haupt Everoze Partners

Although the solar industry still sees itself as young and fancy, its assets are aging. Asset owners seem still to struggle with the complexity of making best use of big data analysis to improve efficiency and profitability of their plants. Many solar assets are reporting high availability but are actually not performing as well as they could. Focus is given to tightening O&M contract terms and putting cost reduction pressure onto the contractors; leaving little room for quality analysis. The author has flagged the untapped values from transforming data into intelligence exactly three years ago as published in the January 2016 PV Magazine edition. In this article, she analyses why this has not changed much and what can or should be improved. Artificial intelligence, advanced data analytics, automated assessments, smart monitoring software: holistic solar asset management starts here.

Why has the situation not changed much?

Exactly three years ago, I flagged as a guest author in PV Magazine the untapped values from transforming data into intelligence. Although some silver lining can be seen at the horizon, there are still many gaps. Vast quantities of data generated by solar assets are becoming available to owners and operators. Much of it remains untouched instead of being transformed into intelligence. Acting on acquired intelligence would lead to smarter operations – maximising energy production, minimising downtime and reducing lifecycle O&M costs.

So why is the transition to smarter operations happening so slowly, despite potential profitability gains of 3% or beyond? How come the digital revolution is still mostly in people’s heads instead of grounded in the fields? We should start perhaps by remembering the bigger picture; solar technology has matured to the point where there is significantly lower failure rates than wind or other comparable energy assets. O&M contractors have also improved efficiency with growing fleets under management, even if this economy of scale may be reaching its natural limit in some cases. Their duties are often narrowly defined in the O&M contracts which are delivered at ever decreasing fees. Many issues only get detected when non-standard assessments or out-of-the-box approaches are tried at the different steps in the asset management value chain.

But despite the obvious gains, we cannot escape the reality that asset management teams manage much smaller budgets then their counterparts in the investment teams. Owners tend to focus more on new investments and less on improving on existing assets. Overwhelming emphasis is placed on cutting costs leaving little room for quality analysis in operations. On top of this, monitoring and communications systems appear fragmented and lacking when it comes to the all-important question of data acquisition. Most O&M contractors offer today a single platform software for monitoring, optimising and control. But as always, the devil is in the detail. Valuable opportunities to make use of digitised data are often missed. The hope that new components with built-in data analytics functions will revolutionise our understanding of the long-term condition of assets, repowering opportunities still face the challenge of significant unbudgeted investment.

I still take comfort though, that at least one of my predictions from three years ago has come true. Back then I said that data analysis from drones would be part of the future tool kit of O&M operators and technical advisors. My team was one of the first to equip an unmanned aerial vehicle with a lightweight thermographic camera to accelerate site inspections of solar power plants and since then this practice has become commonplace in the sector.

Understanding the data

KPI trap and contract flaws: In the previous article, I described the issues of falling into the PR trap, by only relying on one single metric and not considering its evolution over time and under different climatic conditions, or not to break it down to block and array level performance. In addition, there is the availability trap. Advanced data analysis nowadays can easily be employed for several hundreds of inverters installed in one project on the basis of 1-minute SCADA data broken down to string levels. By comparing the different data sets against each other as well as to the metered production data and calculating potential production estimates based on satellite derived irradiance data, the detection of string AND inverter technical unavailability even during times of SCADA data loss is possible. These actual availabilities, including during times of SCADA data loss, can deviate significantly from the availability stated in the O&M reports, and as defined in the O&M contract. The tricky thing is that even after calculating a more accurate availability, not every performance and availability issue is compensated for by warranties due to poor definitions in the O&M contract, especially if it is a very old one.

Monitoring software failure: this leads to another benefit of SCADA data analysis. It is able to pinpoint failures in the monitoring software. In cases of significant SCADA data loss, unavailability can be reported wrongly, because it only reports for SCADA data coverage periods and ignores unavailability during periods of data loss. This data loss may in some cases also not be stated in the O&M reports. Significant data issues affecting key signals can indirectly undermine the value of the asset as they can increase the uncertainty of future P50 forecasts necessary for asset valuation. Therefore complete data coupled with advanced data analytics form an important part of preserving the value of the asset over its life-time, unlocking value during operations and when it comes to refinancing.

Non-updated methodology: automating the data assessments can help to reduce time spent on data analysis. But since fleets are growing and diversifying in terms of technology, countries and regulations, sticking with the old approach is still sometimes the go-to choice, even if it isn’t the most efficient one.

Design, technology or construction flaws or poorly performing contractors: The root cause from issues picked up during big data analysis can also lead to revealing preliminary hidden design or technology flaws or to poorly performing contractors. It shows, that beyond the scope of the daily O&M activities, a comprehensive and holistic approach to solar asset management is key.


While the easiest step might be simply to repair or replace with the same component – be it the monitoring software, module or inverter – asset owners and operators often think twice before upgrading the component in question. But why wait another 5 years or so if the investing today will improve the asset value?

There is also the mine field of renegotiating contractual terms of the O&M contract or insurance. In times of price battles over new acquisitions, the renegotiation of key KPIs, warranties, response times, liquidated damages or even changing the contractor might be a faster route in boosting overall profits.

My fancy pick of today for predicting what will be standard in three years’ time is about the increase of AI (artificial intelligence) to be used in data analysis and failure prediction systems. Even more exciting is the emergence of self-learning algorithms that will enable real-time analysis of price levels. In the times ahead of merchant markets, corporate PPAs and co-locating with storage devices, algorithms that can make immediate decisions based on asset health of a solar storage plant to either store the produced energy or to sell it, seems to me an ultimate fit. Why not hold me accountable on this prediction in three years‘ time?