HyEnergy: Swiss renewable energy potential in the digital era
Author
Dr Roberto Castello
EPFL
Interview with one of the persons in charge of the NRP75 project „supplying Swiss buildings with a combination of renewable energy sources“.
What are your project goals and what have you already achieved?
Our project aims to develop a methodology to accurately measure the renewable energy potential for Switzerland, in particular by combining solar, wind and geothermal sources. For this, we mostly rely on primary data collected by Swiss Federal Offices and on digital images from satellite and remote sensing measurements. The analysis is done with advanced statistical techniques and by learning from data or, if you like, with the so-called Machine Learning approach.
What does “renewable energy potential” mean?
We could think of it as the amount of energy that can be produced from renewable sources given the current (or future) physical, climatic and topographical configurations of Switzerland. Ideally, as if we could exploit as fully as possible all the available resources. For example, if we were to cover all the buildings roofs with photovoltaic panels or if we were to install wind micro-turbines all over those places where the wind speed is greater than 10 m/s for more than half of the year, etc. Let’s say that this brings to an optimistic estimate by definition. In order to be more realistic, we have to account for other limitations related to urban regulations, social acceptance factors (“I don’t like to see those black blocks on the roof of my house!”) or economic costs, which can actually limit the installation and thus the final potential. Nevertheless, as scientists, we usually tend to be optimistic and we do not give up too easily. We work at best to provide solid scientific results and then we throw the ball to policymakers, companies and government representatives, who take care of translating results into practice.
Why do we use “Big Data” and what does it mean in this context?
The method we are currently developing aims at estimating the solar energy potential at very high temporal and spatial resolution. Think of it as: at every hour of the day and every 200 m of Swiss territory we want to know how many kWh of electricity can be taken from sun. If you do the math (on average 16 hours of sunshine per day x 365 days x the nearly one million of 0.2 x 0.2 km2 pixels covering the Swiss surface) you find that the amount of input and output data at stake is very high. So, we definitely start walking across the field of Big Data, meaning that ad-hoc computational and statistical techniques are required for their analysis.
Which methods do we use?
Sometimes the lack of data and availability of measurements at such high-resolution forces us to guess the solar energy potential for all the pixels on the map starting from those for which we do have a measurement. In order to get this, we use spatial regression techniques based on Machine Learning algorithms. Suppose that we want to predict the incident solar radiation on each pixel. We start from satellite measurements, usually taken at broader resolutions, and we infer the values for lower spatial resolution pixels where we don’t have measurements at all. The principle behind consists in letting the machine learn which are the correlations between the “features”, like the geographical position of the pixel, its altitude, etc… with respect to the “target” which, in our case, is the incident solar radiation on that point. Once the relationship has been learned (after the so-called “training”), for each new pixel and set of features given in input, the machine is able to give in output its best estimate of the incident solar radiation on that pixel, together with its statistical and systematic uncertainty.
To date, we have produced a map of the incident solar radiation throughout Switzerland for every single hour of the year. We also estimated the available area for the installation of photovoltaic panels on each of the nine million roofs. Everything has been done by leveraging the statistical power of Big Data in combination with cutting-edge Machine Learning technology. In this way we have been able to provide, with unprecedented precision, how much electricity we could extract from sun at any hour of the day for every building in Switzerland. Not bad at all, is it?
What are you and your team particularly proud of?
Knowing that our results can become the basis of future political decisions, and perhaps also driving the future economic investments, fills us with responsibility and honour at the same time. The fact that we are helping to inform some of the goals of the Swiss Energy Strategy 2050, aiming at replacing the fraction (about 30%) of electricity from nuclear power plants with renewable sources, is highly motivating.
What changes does your project bring about?
Using cutting-edge Machine Learning and statistical techniques to provide the most accurate measurements is our challenge. Once we measure the energy potentials for solar, wind and geothermal separately we can ultimately combine them together and optimize their proportions within localized “hybrid” energy systems, in order to match the local energy demand. The idea is to move towards a 2.0 model of energy distribution based on decentralized local grids where the energy produced is consumed at the same time, in contrast with the standard centralized distribution. Householders, from being consumers only, will slowly become prosumers of their own energy.
What does NRP 75 mean to you?
It means to be part of a national project, something recognized to be relevant by the majority of Swiss people and therefore of great impact for the future. Personally, having the opportunity to use my scientific expertise to solve such challenges and seeing the direct impact on the future of a country is very rewarding. The possibilities offered by the SNSF, and in particular its PNR75 series of projects, are manifold: from funding to perform our research autonomously to the promotion of cross-cutting activities with researchers working on different topics. This can create original ideas, foster interesting exchanges and often initiate new collaborations.
What would be missing if your project did not exist?
In fact, the aim of our project is precisely to “have something missing”. Aren’t we happy to notice, as an effect of our research work, a few tenths of degree less in the yearly average temperature? Or maybe measuring lower concentrations of CO2 in the atmosphere resulting from lower emissions, thanks to the energy that we now capture with our rooftop solar panels? We work hard for that, and we look forward to achieving the objectives!