Smartwatt is taking advantage of Artificial Intelligence tools since 2008. During all these years, we managed to solve real problems from multiple industries, mainly connected to energy, recurring to this technology.
After all, how did it all begin?
Around 2008, Smartwatt was very close to the university and, as it is common in this environment, thinking outside the box is required and our team started to wonder how the present would be influenced by the future – What could we do differently today, if we knew what will happen in the next days?
Stating the obvious, most of us would answer – Well, I would take better decisions and would make less mistakes.– Our team put some thought about it, concluding that if we knew the outcomes of each variable that could influence the result of each decision a company has to make, that will lead the company to make better decisions reducing considerably the risk. –Would this be possible? – Our team put hands to work and it became possible.
Our first challenges
Back in 2008, Smartwatt faced the first challenge, launched by the main Portuguese Distribution System Operator. That company wanted to know the production forecast of a small hydropower plant to manage better that energy system, and so they asked us if it was possible and Smartwatt’s answer was a big yes! It would be possible indeed using predictive algorithms. Then, we made our first partnership with a company, which portfolio was 350 MW distributed by 130 small hydro power plants. We started by collecting the weather data, and exploiting data to find the best approach. Soon we decided that neural networks would be the best approach, and we reduced their uncertainty range from 350 MW to 28 MW, with a chance of error of 8%.
Along the years, we learned, improved our algorithms, gathered data, and after 10 years, we are delivering our clients forecasts for each hour, with a 7 days horizon, with a chance of error of 3%.
In 2015, a Portuguese Wind Operator challenged our team to develop a system to deliver insights for Operation & Management (O&M) optimization and increase the safety of the process for all their portfolio, around 400 wind turbine units.
We faced the challenge by using a methodology that we call – Predictive Thinking Process, inspired by Design Thinking Process, which the main goal would be to add their know-how into our algorithms, transforming the output into an input.
Not only we were delivering a simple power forecast, but we also created key-performance indicators (KPI) to support daily and weekly decisions, such as: weather alerts, energy market forecast to understand which days would be more profitable, intelligent heat maps to manage the technical teams, lightning forecast to reduce health risks for the team, and more KPIs in a sharable platform that made our client’s activity easier. In the end, they were investing 90% of their time in decision-making instead of information seeking.
More recently, in 2017 the Portuguese Distribution System Operator wanted to reduce their downtime periods and improve the management of their maintenance teams, so they challenged us to develop a system able to predict when and where the electrical distribution grid would fail. Their grid infrastructure was formed by 68 000 km of overhead lines spread by the Portuguese territory, and there were three years of historical data available related to energy faults.
Their main concern was the extreme conditions that caused severe energy faults, and there was not enough technical teams to take care of the incident or even enough human resources on the customer support to answer all the calls and requests.
We are still developing and improving the system and we have already saved around 20% of costs to them.
The future for a company that predicts it.
Reviewing the last 10 years, we are aware that this is the beginning of something very powerful that will impact all the industries. We started our path in energy industry by using neural networks, then we created our own tools to improve speed and accuracy using GPU technology. We are developing tailored-made algorithms to improve our client’s activity and solve various problems easily, and we are using their know-how to boost our prescriptive analytics services.
Smartwatt is better at every challenge it faces, so we like to think that the future will be bright to us and all our partners.