Digital Twins

Need for best-in-class asset management

By Sharat Singh, Founding CEO and Chief Architect, Quadrical Ai
Dr Hugh Hind, President and Co-Founding CTO, Quadrical Ai

Currently, the installed renewable energy base is roughly 700 GW globally. This is expected to grow to more than 5,000 GW over the next 20 years; this means a 7x growth rate. This growth of so­lar will come in different shapes (utility sc­ale, commercial and industrial, as well as residential systems), and will severely challenge the existing grid and distribution systems. As an intermittent form of energy (like wind), it will likely need to be supported by storage technologies and automation. This growth will need to be more so­phis­ticated, distributed and intelligent. As­set manag­ers for these renewable energy assets are already gearing up and planning for this growth by using data to make it more reliable, available and predictive. We need to go beyond fault codes to make planning more advanced.

Looking at different technologies to assist in the growth of the solar industry, this ar­tic­le goes deeper into its unique challen­ges and explores how digital twins may be the te­chnology solution to add­re­ss these issues.

Disproportional O&M costs

Fast dropping power purchase agreement (PPA) prices along with rising solar panel costs (prices fell quickly initially but are now going up owing to inflation) are forcing O&M teams and asset managers at solar plants to figure out ways to im­pro­ve both operational efficiency and yield at the same time. They need to do this while also dealing with several operational variables, whi­ch include soiling, shading, we­a­ther conditions, measurement accuracy of equipment, installations that are not built as per design and finally degradation of equipment out in the field. This makes their job quite difficult.

Growing scale of problems and shrinking scale of solutions

To put this into perspective, a 1 MW plant would have approximately 70 strings (of 50 panels each) and be spread over 5-6 acres. Often, there would also be so many issues – inadequate instrumentation or cali­­bration quality, along with un­cl­ear ex­pectations of weather, micro-wea­ther, and uneven degradation scenarios across the plant. That list is not finished. There are also issues of shading, soiling and so ma­ny short-term, fixable issues as well. When multiplied by 1,000s for plant and portfolio size, the sheer scale of issues indicates that plant managers are ultimately able to do only bare minimum fixes on time. The­se are limited to contractual preventive maintenance by the assigned O&M company and corrective maintenance, which happ­e­ns when there are alarms during outages or when variances become too high.

Missing tools

While supervisory control and data acquisition (SCADA) systems support the O&M decision-making process, they cannot facilitate the kind of advanced analysis re­quired to significantly improve plant per­formance on the asset manager’s wat­ch. Standard industry benchmarks like PV­Syst are very good for planning to un­derstand the future financial performance. However, they are not completely accurate, granular, or actionable for really im­pro­ving plant performance.

Challenges with traditional contracted (or expected) energy measures result from in­accuracies and noise and blind reliance on pyranometer data. Pyranometers may be very unreliable due to frequent calibration required to keep the equipment accurate, and the number of devices needed to co­ver large plant areas are not available. If we add no consideration of efficiency and be­haviour of individual string combiner boxes (SCBs) at the different sta­ges of ageing – this means that the benchmark may work at an aggregate level (for an entire plant for a longer period of time), but it is ineffective to make actionable, measurable improvements at the SCB le­vel or provide weekly improvement plans for the plant.

Forced errors

Unrealistic yield expectations driven by IPPs and tacitly agreed by investors put O&M teams on a downward spiral as they are unable to make effective investments in technology. Second-guessing O&M te­ams has become routine for IPP management. This often means adding even more costs for preventive maintenance in drone and thermal imaging and I-V testing every 6-12 months.

Digital twins: Enable condition-based maintenance

One of the cutting-edge industry-specific solutions that addresses challenges pos­ed by unrealistic yield expectations and razor-thin O&M margins across the industry is digital twin technology using artificial intelligence (AI). A digital twin is the digital copy of any physical asset, modelled to replicate the asset’s behaviour in real time under any field condition. A unique digital twin is built for each instance. This technology is easily used to create highly acc­u­­ra­te and granular performance benchmarks, personalised for any plant and ea­ch underlying device in the plant, using historical and real-time data. These ben­c­h­marks are for acceptable DC yield, inverter conversion efficiency and AC tra­nsmission efficiency. Any deviations from these benchmarks can then be highlighted and acted upon depending on the se­ve­rity of the issue. The issues can often also be quantified in terms of impacted revenue loss or production loss. All of this is done very quickly using machine learning (ML) and AI. Human be­ings are very good at seeing patterns in small amounts of information, but ML can see patterns in very large amounts of data. Today’s solar plants provide millions and millions of points of data and using those effectively is the key to improved solar plant operations.

Condition-based maintenance: Benefits

This simple digital twin-based benchmark leads to the practice of condition-based maintenance. Replacing preventive ma­in­tenance with prescriptive maintenance and corrective maintenance with predictive maintenance allows the improvement of solar plants by 2-4 per cent with lower costs. This is because O&M resources are utilised much more effectively. Instead of just remedying issues based solely on alarms, remedies are based on proactive, data-driven decisioning, reflecting the true condition of individual plant components. This combined transformation of intellige­nt prescriptive guidance of the current sc­he­dule and predictive planning of remediations results in reduced outages, increa­sed yield, reduced despatches, lower O&M manpower requirement, and better in­ventory planning and management.

Condition-based maintenance: How?

Broadly, there are four steps for condition-based monitoring and asset management:

  • Historical and real-time data acquisition from various data acquisition systems (DAS) like SCADA, dataloggers and thi­rd-party application programming interfa­ce.
  • Data cleaning, harmonisation and standardisation to yield the best possible da­ta quality, which can be used for ac­tu­al data analysis.
  • Creation of digital twins or replicas of ea­ch component of the asset, using both historical and real-time data. The­se begin from the smallest instrumented components from the ground up to the whole plant. Next, these yield ex­pec­tations for each string in real time is added up to create the most accurate estimate (the digital twin benchmark) of the plant’s generation capacity at any time, and under any field condition, with unequalled granularity and precision.
  • Lastly, it involves comparing the real-time output at any or all parts of the pl­ants against their corresponding digital twins to drive condition-based decisions. The deviations from the benchmark should be tracked, characterised, and quantified for weather, structural degradations, shading, soiling, new-fixable issues, and ultimately converted into actionable, revenue-prioritised work orders.

Digital twins: Lots more there

Note that digital twins, as a benchmark driving condition-based maintenance, are the simplest and easiest benefit. One can zoom out to go back in time (time travel) for available data and create digital twins going all the way back to plant commissioning to answer questions for investor due diligence and raise longer-term degradation claims against the manufacturer or the EPC. Similarly, travelling forward, using TMY3 as inputs, the yield forecast using digital twins is far more accurate than traditional static-model based approaches. One can zoom in to find and address classification and quantification of shading, soiling, insulation and anomalies. Finally, as asset owners deploy varying original eq­uip­ment manufacturers’ (OEMs) equipme­nt in varying configurations, digital twins are perfect for uniquely mirroring each as­set accurately, and an ideal tool for overall portfolio management. As multiple battery storage chemistries are being tried by OEMs, digital twins are becoming instrumental in both estimating the current and predicting the state of health as well as corresponding outage prevention.


In conclusion, a continuous, automated, granular and actionable condition-based maintenance allows asset managers to proactively replace their preventive maintenance schedules to improve yield (and eli­minate outages). This also drives O&M efficiency improvement and reduces the frequency and cost of despatch with clustering for owners of many commercial and in­dustrial installations. In addition, the ac­cu­racy of digital twins can give its users stro­ng confidence in their OEM claims and due diligence processes.  The state-of-the-art part of digital twins is that they begin to help immediately like an advanced monitoring system, they are used to optimise as­set performance continually, and they keep getting even better with time. This be­comes true condition-based maintenance guidance. This is be­cause digital twins lea­rn and adapt to the plants’ inherent characteristics over time, and thus keep getting better and better.

Today, in 2022, we are in the very early days of renewable energy growth. Most of our plants were built in the last three to four years, and we all understand that the real grow­th will come in the next 10-20 years. Plan­ning for the management of this gro­wth is what is needed today. If you are an IPP or owner or an O&M te­am, you will ne­ed to scale your assets un­der management 10x over the next few years. To do this effectively, you will need smart (multiple and specialised) digital twins so that your plants can learn to self-manage themselves.


Enter your email address