“Sustainability has to be a way of life to be a way of business” – Anand Mahindra, Chairman, Mahindra Group.

This vision of the group chairman pretty much says it all – sustainability has to become a way of life for industry. Keeping the wheels of industry moving requires an enormous amount of power generation, for which India is betting big on the sun. As the leader of the ambitious Solar Alliance, India is fast tracking the solar power generation ecosystem from every perspective.

For it to become a way of life, Mahindra Teqo is looking at an innocuous looking aspect which can have a big impact – increasing the productivity of Photo Voltaic (PV) cells simply by keeping them clean. Installing solar power plants is just not enough; unless asset management efficiency improves over time, returns will either stagnate or decline owing to lack of maintenance. Mahindra Teqo, a subsidiary of the Mahindra Group, a new breed of tech-enabled renewable energy asset management solutions provider, assists solar and wind energy asset owners in increasing profitability. They are responsible for around 4.5 GWp of solar PV assets and 7.5 GWp of software.

Understanding WHY it needs to be done …

The upkeep of a solar power plant is fraught with problems, one of which being dust. Dust hinders the photon-to-electricity conversion radiation. According to a study conducted by Mahindra Teqo, the Module Cleaning activity accounts for 30-40% of the overall operation cost of a solar power plant. The easiest controlled losses are soiling and string failure. The planning and monitoring of module cleaning consumed a total of 1,845 man hours each month per location. They figured that by optimizing cleaning cycles and factoring in the string factor, could significantly reduce cleaning time and actually be a game changer. In a market gasping for profitability, this is a bright ray showing the way to high efficiency.

According to Mr. Kapil Panwar, Lead Data Scientist, of Mahindra Teqo, “If our algorithms are applied across the installed solar energy capacity of 40 GW across India, it could result in savings of up to ₹500 crore annually”

In order find out the root cause of the problem, they performed a Why-Why analysis …

Why 1 – Regimented cleaning preformed on site without any focus on optimization

Why 2 – Required complex calculations and extra efforts of site team to plan the process

Why 3 – Relative soiling depends on multiple parameters which keeps changing with time

Why 4 – Automation of cleaning trigger was not available in any SCADA or system

Why 5 – Required predictive and automation methodology was not available

… and came up with strong actionable points. 

Understanding HOW it needs to be done … 

They decided to create a generic module cleaning tool that could be used on any site, based on Artificial Intelligence (AI) models that provide great accuracy and speed. This model could determine when cleaning should be done, which parts of the plant needed to be cleaned and which part should be cleaned first in the event of a water or time constraint in order to enhance generation.

To continually improve the model, a Machine Learning (ML)-based general model was added that uses a high level technology stack consisting of Python, Hadoop, PySpark, FastAPI and PostgreSQL, that learns by itself, optimizes cleaning costs and saves water at no additional cost. It understands the requirements of different PV modules like crystalline and thin file modules, and gives string health information, which dictates the cleaning process - wet, dry, and robotic cleaning. It can also be used to activate cleaning robots which can completely automate the process, requiring no human intervention. 

So how do the different elements come together to deliver efficiency? Through a governance mechanism which tech-enables the end to end process:

• Model Tuning : mapping of what data is required, from which part of the power plant and when
• Data Collection : the actual collation of data from all sources (which is usually a major challenge due to multiple formats and periodicity)
• Processing Data : cleaning and organizing the data to required structures
• AI Model : processing the structured data with algorithms
• Suggested Cleaning : output of suggested cleaning process and schedule, including auto-triggering of cleaning robots
• Feedback from IOT : verification of action taken and comparing with estimated outcomes

The outcome is two-fold :

1. Efficient identification of underperforming modules that require cleaning (without any manpower having to do a physical survey) along with prioritization instead of a full cleaning of all modules
2. Efficient triggering of requirement specific cleaning depending on the type, configuration and relative soiling of the module / string

The team started with a generic model and is continuously improving on it with every process cycle over time, specific to user requirements. Since the model is built on open-source tools, safety of tools and data was a challenge which was overcome through encrypted binary formatting. 

Understanding WHAT the impact can be … 

Mahindra Teqo’s unique model has been able to deliver multi-faceted impact which can be understood best with the People – Planet – Profit (PPP) model, which is again in line with the chairman’s vision of sustainability being a “way of life”

People

• Improve time taken in cleaning planning, prioritization and scheduling from 1845 hr / mth to 30 hr / mth, which can be redeployed for more productive usage elsewhere
• Reduced work complexity through the model will help speed up manpower productivity
• Efficient time utilization will ensure minimal over-runs and therefore improve work life balance
• EPS will also be improved

Planet

• 25% savings in water consumption for module cleaning purposes
• Better asset utilization leading to lower resource consumption

Profit

• 25% cost savings on module cleaning
• ₹ 75 Lac / MW/ year improvement in ROI for clients 
• Increased asset life cycle due to better asset management

Every Bit Matters 

As children, we are often taught how the “small” efforts make a “big” difference. This is especially true of Mahindra Teqo.

Image by ThePictureBox from Pixabay