AI For Good: Cleaning Up The Environment With Data-Driven Insights

Providing access to clean and cheap energy to the exploding world population is the single most important problem facing humanity.

 – R. Smalley (1996 Nobel Prize in Chemistry)

The Uptake of Clean Energy Globally

The Paris Accord, signed in 2016, was an agreement between countries to ensure a sustainable, low-carbon future. It was a historic moment that shed light on the dire consequences of the future of the planet if we don’t adopt sustainable means of producing energy. Since then, countries world over have accelerated their efforts to move away from fossil fuels and harness the power of natural resources for energy.

Globally, the world produced 5.9 TWh of modern renewable energy in 2016 – a five to six-fold increase since the 1960s. Hydropower remains the most dominant form of renewables consumption with a 70% share. Energy from other clean sources such as biofuels, solar, wind, geothermal and tidal, are steadily growing and outpacing hydro. Countries like India, UAE, Saudi Arabia and Spain among several others are leading in clean energy generation.

With the increase in energy production now, the need for distributed energy systems has risen. Currently, industries, commercial areas, large buildings, municipalities and large communities are facing three main challenges – costs, security of supply and carbon dioxide reduction. It has been deemed possible to turn these challenges into long-term calculable variables across businesses and industrial sectors with the help of local distributed energy solutions.

Optimising Distributed Energy Resources

The availability of inexpensive Distributed Energy Resources (DER) such as solar are revolutionizing the electricity industry. For instance, solar energy will become approximately 25% of India’s generation capacity by 2022. DER are clean, scale-free and reduce the need for expanding power grids. However, they are intermittent resources that are vulnerable to environmental factors (such as dust and heat) and are impacted by usage patterns (batteries survive only a limited number of charge/discharge cycles). Most importantly, they are hard to manage and optimise because a large number of small resources are distributed over large geographies.

These characteristics pose significant challenges to various stakeholders. For instance, electricity suppliers are looking to modulate the amount of electricity injected into  grids so the stability of grids is not impacted; asset operators are interested in condition-based maintenance; solar shortfall insurers need to ensure that the shortfall is due to lack of sunshine.

Many countries, both developed and developing, are working on incorporating distributed energy sources such as solar and storage into their energy portfolio. For instance, India is aiming to incorporate 100 GW of solar generation capacity by the year 2022. Furthermore, solar has reached grid parity (solar can generate power at a levelized cost of electricity (LCOE) that is less than or equal to the price of power from the electricity grid) in many parts of the world.

All these stakeholders need software tools to manage and optimize the assets they are interested in. This requires collecting data (from diverse power electronic equipment such as inverters, weather stations, energy meters, etc), analyzing massive amounts of spatio-temporal data (both on cloud servers and on edge computing devices), controlling the equipment (such as automatic panel cleaning tools, charging/discharging batteries, etc) and presenting the results/insights to appropriate stakeholders through intuitive phone and web interfaces.

To address these needs, Tanuja Ganu Sunil Ghai and Deva P Seetharam established DataGlen Technologies (“DataGlen”) in 2015 to provide IoT  & AI based-software products and solutions for data-driven distributed energy systems.

Solutions Provided By DataGlen & Their Impact

Solar & Storage Optimisation: This can forecast local generation, local consumption and optimisation of energy use within constraints. This has reduced energy bills by 10-50%, and reduced peak demand on the grid.

Predictive Maintenance for Solar Asset Management: This provides an estimation of residuals – an indicator of level of soiling of the panels, provides auto-alerts when the trend crosses the threshold, and sends an optimal cleaning schedule to a plant. Besides, it also provides alerts on under performing sections of the plant and the devices helping increase the capacity utilization of the plant. This has improved generation by 8-10% while reducing operation and maintenance costs, resulting in an increased cash flow of about Rs. 2.5 lakh per MW/year.

Identification of Faulty Rotors, Motors & Drivers: Through frequencies, identification of healthy/faulty motors can be done, clustering algorithms can segregate between defective and healthy devices and auto alerts for maintenance are sent out as well. Based on this system, 90% of faults can be identified in advance, with reduced down time and manual intervention.

The Definitive Impact of Data

DataGlen’s technologies have increased the effective generation capacity of the solar plants by 25 MW. That can be translated to an approximate monetary value of $25 million. By increasing the solar generation, the company has helped reduce 30,000 tonnes (assuming 800 gms of CO2 saved for every unit of generated) of CO2 emissions. Using condition-based monitoring algorithms, DataGlen also helps plant owners reduce the volume of water used in cleaning solar panels without impacting the potential generation.

The company is also conducting several workshops for skill development of O&M staff managing solar plants, and engaging with researchers, professionals and students from prestigious institutions such as the IITs & IIMs through collaborative projects and internships in the CleanTech space.

DataGlen was one of the 10 finalists of AI for Good, conducted by NASSCOM CoE DSAI in association with the Government of Karnataka.

Share This Post

Leave a Reply