Man has been collecting all kinds of weather data for what seems like forever. If only to try and predict the weather itself. Today, slowly but surely, new emerging technologies such as artificial intelligence (AI) and machine learning (ML) are enabling us to make predictions in other areas as well, based on these same historical weather data that we’ve been gathering for ages.
Take the energy sector for example. Not only does the weather determine to a certain extent our behaviour (i.e. whether we stay in or go out) and, more specifically, our energy consumption (whether or not we turn the heating on, for instance), but due to the rise of a number of alternative energy sources, such as wind and solar energy, the weather now also has a growing impact on our energy production.
A complex environment is getting extra complicated
To this day, here in Belgium, nuclear power remains our main or primary source of energy, providing us with a steady supply of electricity. Unfortunately, that supply is not really flexible, since we can’t just switch our nuclear plants off and on again at will, anytime we need to produce less or more electricity. Complicating things further, the electricity carried over our nation’s power grid cannot simply or easily be stored away somewhere either. So whatever electricity we consume has to be produced more or less in real time. Hence the importance of matching our energy production as closely as possible to our energy consumption, predicting both as accurately as possible.
Add to this already highly complex environment the multiple alternative energy sources that form a growing part of today’s energy mix, such as wind and solar plants, and you can see why production forecasting in the energy sector hasn’t exactly become any easier. With these new energy sources, yet another variable has been added to the equation: the weather. For without wind or sunshine, there simply is no electricity production.
Weathering the energy transition with AI
The introduction of smart meters in a smart grid should provide us with the data needed to improve our energy consumption forecasting. To improve our forecasting of green energy that is specifically produced by windmills and solar panels, the required data are largely available already. They are, in fact, the weather data and weather forecasts I mentioned earlier. All that was missing up till now, was the technology that would allow us to analyse those vast amounts of data and build a model that we can learn from, allowing us to do simulations and create multiple scenarios with different degrees of probability.
With artificial intelligence, that technology seems to have arrived at last. And with it come a number of other benefits. To give but one example: once we can more accurately predict the production of alternative green energy, based on weather information and forecasts, then we no longer have to produce as much ‘dark’ energy or buy it, often expensively, from other countries. That is notably the case with Germany: once again in a pioneering role, that country has already analysed the production of its wind and solar energy against its weather. But the weather in Germany is not necessarily the same as in Belgium. And neither is its energy infrastructure. Therefore each country will have to build its own AI model, based on its specific weather conditions, energy resources, and so on.
These are challenging and disruptive times, to say the least. By implementing AI solutions like the one discussed here, we should be able to predict our energy production and consumption much more accurately in the coming years. This in turn will help us manage the energy transition we all need to go through. The required data are available already, so in the end it is just a matter of being able to extract the information that we need from those data. And with the help of AI, we can do just that.