In the PV industry, AI is now being used to help push solar toward grid parity in markets throughout the world. The International Renewable Energy Agency (IRENA) asserts that such technologies are already playing a “transformative” role in pushing the global energy transition forward, even while acknowledging that some people may already view terms such as AI, big data, the Internet of things (IoT), cloud computing, and blockchain as hackneyed buzzwords.
However, these terms are part of the growing digitalization of electricity generation, distribution, and consumption. And in the world’s biggest solar market, China – where the authorities recently outlined a series of policies to expedite the development of subsidy-free solar projects – these technologies are rapidly driving further reductions in PV plant costs.
The potential of AI in energy has been known for years, and digitalization of the transmission sector is at an advanced stage, the International Renewable Energy Agency says in its report, Innovation Landscape for a Renewable-Powered Future. Although sensors have been installed in solar arrays from almost day one, the digitalization of generation and consumption is still at a relatively early stage. The next step, IRENA argues, will be to modernize power plants and further automate control of the grid.
The role of AI in power systems is rapidly becoming a “necessity,” IRENA asserts, as it can help to more effectively integrate variable renewables into the grid by predicting generation and consumption patterns. “Reductions in uncertainty in power production forecasts and power demand forecasts enable smarter operations,” agrees Elizabeth Traiger, senior researcher at energy consultancy DNV GL. “Unexpected curtailment can be reduced, and planned maintenance can be scheduled at optimal times.”
Investors in new solar projects can use sensor data to provide an accurate picture of the generation possibilities at proposed sites, opening up financing potential. Solar arrays provide enormous amounts of data that are already being used in meteorology applications.
“AI and machine learning algorithms are being proven in forecasting to improve the accuracy of short-term forecasts, up to 48 hours ahead,” explains Traiger. “By combining the wealth of meteorological ground station data, satellite weather data, and local cloud cover imagery, the irradiance, or available solar resource, can be more accurately estimated. The resulting reduction in uncertainty in solar power production helps to ensure a balanced and reliable electrical grid. Variability can be anticipated and alternate dispatch or adjustments to generation can be made more smoothly.”
Scratching the surface
Solar manufacturers are also starting to deploy AI in their factories to streamline production. JinkoSolar, for example, recently revealed plans to use AI at its 400 MW module factory in Florida, starting with electroluminescence inspections. However, the industry is still just scratching the surface of AI’s potential to push solar closer toward grid parity.
“The integration of AI will enhance the ability of solar to integrate in the dynamic energy mix,” says Traiger. “AI enables accurate, informed decisions in all data-heavy areas. For solar, this encompasses the entire value chain: feasibility, development, operations, and decommissioning. Making smart decisions ensures the best outcomes by adding reliability and reducing the variability inherent in renewable resources such as solar.”
That said, the solar industry still has a long way to go in more effectively exploiting data, she adds. “AI and machine learning is based on data. As we gather more data, from historical sources, past weather and production patterns, different spatial resolutions of new satellites, as well as new data from increased ground sensor installations, the tools will evolve,” Traiger says.
As the global cost of solar continues to fall, developers increasingly struggle to reduce risks and improve returns on distributed generation PV projects. The ongoing decline in the cost of energy storage provides a ray of hope, but in order to truly leverage the power of storage, project operators need to utilize AI to provide new sources of value. These value streams include lower demand charges and the provision of higher revenue from virtual power plants (VPPs).
California-based storage specialist Stem, for example, announced plans in late 2017 to install AI-backed storage solutions in Japan, a global hotbed for VPP deployment. Other recent AI-backed storage pilots noted by IRENA include French utility RTE’s Ringo Project, which aims to ease grid congestion without the need for new power lines, by using AI to facilitate dispatch. The future potential of AI-driven solar+storage is huge, with IHS Markit recently predicting that more than 500 MWh of storage capacity will be paired with utility-scale PV projects this year in North America alone, in what the research firm describes as “the first meaningful volumes” to be deployed.
IHS Markit expects an uptick in the use of AI in the solar sector this year, and it sees inverters as playing a “critical role” in this shift. It also expects 11 million PV inverters to be connected to the Internet of energy this year and characterizes the digitalization of the grid as a “mega-trend.”
PV system maintenance, in particular, is becoming more predictive thanks to the growing use of monitoring software, which helps system owners and O&M service providers identify problems before disrupting a solar array’s output. Sensors in inverters facilitate predictive maintenance, and many companies have taken notice of AI’s potential. Huawei, for example, has been exploring smart PV solutions for the past five years, as part of its efforts to accelerate grid parity with AI. Its inverters serve as smart sensors, providing the feedstock needed to continuously improve its AI algorithms, which are used to optimize O&M practices.
Huawei’s AI-driven FusionSolar 1500V Smart PV Solution also boasts a number of features aimed at supporting the push toward grid parity. Intelligent trackers and bifacial modules are backed by AI algorithms, while its Smart I-V Curve Diagnosis and AI recognition technologies facilitate automated O&M of PV projects. Many companies have long used automated drones for inspections to identify defects in advance. But inverters show promise to provide faster, better intelligence. “Algorithms that can give early warning of faults such as inverter failures, or indicate any deviations from normal operating behavior will be instrumental in main-taining even higher levels of reliability for long-term reliance on consistent solar production,” says Traiger.
Tip of the iceberg
But even with solar O&M, there is still a great deal of untapped potential. Data from sensors can help prospective inves-tors with due diligence, while also con-tributing to certification of equipment, for example.“AI can, and will, definitely change and improve solar O&M in many ways and we are only now seeing the tip of the ice-berg of the applications that will be commonplace,” says Jaime Sureda, Director of Digital Transformation for Spanish O&M specialist Solarig. “Price reductions com-ing from the use of AI in the sector will definitely foster the use of solar energy all around the world.”
Solarig has serviced roughly 4.5 GW of solar capacity around the world, and is now in the process of digitally overhauling its operations, with important ramifications for its customized O&M management platform. A key aspect of this shift is the optimization of data collection, integration, and management. It collates data from its monitoring software and uses it to continuously improve its predictive maintenance algorithms. It says such technologies are contributing to significant reductions in global O&M costs, and it says it is using AI to “squeeze” data it collects from a range of sources.
“We are also paying particular attention to the use of cutting-edge technologies in those areas with the highest impact on direct or indirect costs, such as labor intensive activities,” Sureda explains, noting that the company has long used thermographic camera-equipped drones to spot defects in arrays.
While companies such as Solarig have long used sensors to assess the impact of weather conditions on contractual performance requirements, Sureda argues that the challenge ahead is to further integrate such information with weather forecasting models to predict the output of PV plants. “AI applied to this analysis is crucial and will definitely help to improve plant operations in the near future by increasing the reliability of electricity generation models,” Sureda says.
The growth of distributed generation PV and technological advancements in AI have fuelled the proliferation of smart devices and demand-management services in recent years, contributing to significant changes in the global power sector, IRENA says. But the organization believes that developers need to launch more pilot projects in the years ahead to fully grasp the potential of digital solutions.“The disruptive potential is only begin-ning to be understood,” IRENA concludes. “It is far from being fully exploited.”
This content is protected by copyright and may not be reused. If you want to cooperate with us and would like to reuse some of our content, please contact: firstname.lastname@example.org.