Three Reasons why Blackouts are Inevitable

Following the Spanish blackout earlier this year, our post-mortem analysis has shown how Artificial Intuition could have provided a two-day early warning. Since we only analysed very high level data, like real, estimated and planned consumption, we could only have issued an alarm but without explaining why.

Given the immense economic and reputational damage that such events cause, we thought there was a business case for our unique early warning solution. We contacted a total of six European grid management companies from three countries, with the following proposal: Give us access to your grid data at a granular level – let’s say a few hundred or thousand locations over the entire territory, and for a daily fee we will monitor your energy grid and provide early-warnings of possible power outages.

Obviously, with granular monitoring we are also able to indicate where problems are initiating and potentially even why. By the way, this is called explainability and this is what AI lacks.

However the problem is not money, the problem is data. This may sound difficult to believe in the digital age, but the grid management companies we have contacted claim they don’t have data necessary to monitor their own grids.

Some of these companies said they don’t have sensors to collect the needed information. Others said they still have old equipment within the grid, making data collection impossible. Others still said the process of digitisation will still take years.

So, here are the three plausible reasons blackouts are inevitable and will continue to happen in the near future.

REASON 1: There really is no data. If there is no data, no monitoring and early-warning strategy can be put in place and that ends the story.

REASON 2: There is data but because of its strategic importance, energy is highly politicized meaning inefficiency is the business model. Grid companies cannot have any monitoring and early warning system in place because they would need to take measures in the face of an alarm. They would need to be efficient. With no monitoring system in place all they need to do is react to problems and fix them. However, if the system is sufficiently complex and inefficient, it is very difficult to blame anyone and to identify the culprit when an outage occurs. The system hides behind the immunity induced by its own complexity and inefficiency.

REASON 3: There is data but AI is used to approach the problem. Electricity grids are super complex, especially nowadays, when unreliable sources of energy are being incorporated. These sources lack inertia and increase grid fragility significantly. High complexity implies fragility but the real problem stems from the fact that complex systems tend to fail in unique, non-repeatble ways. Each crisis is unique. This makes it impossible to adopt Machine Learning. There are no patterns to recognise, not enough examples to learn from. Zero explainability. AI is simply the wrong tool here.