How AI engines boost grid energy consumption forecasting accuracy

This audio was created using Microsoft Azure Speech Services

In recent years, patterns of energy production and consumption have changed radically. New variables such as renewable power generation, community microgrids, and rapidly changing energy prices make grids infinitely more complex to manage. To address this challenge, our company Predictive Layer, developed an artificial intelligence (AI) digital solution that is designed to better forecast energy supply and demand variations across an electrical grid. To learn more, read our blog series.

Reuters Event On Demand

How to improve forecasting while managing a variable grid

My clients are concerned with how to improve their forecasting while managing a much more variable grid. Many have similar questions, so here are my responses to some of the most common questions.

Question: How can I maintain my grid stability given the intermittent nature of renewables like wind and solar?

Answer: Renewables make managing the grid more complex because they add an element of unpredictability. It is more difficult to understand where and when energy will be produced and consumed. In order to successfully integrate wind and solar into your grid, the local utility has to tune, program, and schedule your grid in such a way as to be able to adapt, in real time, to these sometime unpredictable changes. Your prediction model cannot afford to just wait for new supply and demand patterns to emerge. It must predict and anticipate the reconfiguration of the grid topology, in order to optimize performance.

To develop this ability, your grid will require two fundamental pillars.

  • First, make sure that the grid infrastructure has the capacity to deliver the energy required and to redirect supplemental flows of energy back to those points in the grid that need it (e.g., when solar or wind power plants are overproducing). It is a matter of optimizing the grid topology to the right capacity at the right moment.
  • Second, controlling the energy flow across the grid is important. Improper tuning of energy flow can degrade the efficiency of the grid by a factor five, 10, or even 20 percent.

Question: Can I scale my predictions?  How can I generate enough data touchpoints to make my grid-wide forecasting more accurate than it already is?

Answer: Not only local grids, but also enterprises and microgrids require more data to deliver an automated, accurate forecasting of the short-term production and consumption. Scalable automated forecasting capabilities are critical. It is not just about having an intelligent artificial intelligence system that will automatically learn the local context and then issue a forecast. It is also about the ability to absorb and evaluate thousands of touchpoints across the grid.

This is almost impossible to implement with only human mathematicians. You cannot send a mathematician out each time you need to readjust your predictive model. You want to implement an automated model, with artificial intelligence and machine learning, to scale with fine precision. If a community microgrid, for example, is doubling its volume of solar panels, the system should learn automatically that seven new panels have been added and instantly factor that into its calculations and predictions. Thanks to new technologies like cloud computing, such scalability is now available at an affordable price. You can rent 100 computers for a couple of hours and accomplish in three hours what an expert human mathematician previously accomplished in three weeks. If we imagine scaling the predictions by a factor of a thousand to meet required accuracy parameters, we can see how a human, using a manual model, would be quickly overwhelmed.

Question: How do clients who have already installed your AI forecasting solution adjust to placing their trust in a machine software algorithm?

Answer: Initially, they are cautious and look over the machine’s recommendations to see if the human needs to override the decision. Then, over time, as the results begin to roll in with more accuracy and precision, the trust builds. Since there is a shortage of data scientists, these experts can let the AI engine take over. Now they are free to tackle bigger problems, like those that require more human thinking than machine thinking.

AI-based energy forecasting tools

Find more information about AI tools to manage grid forecasting

Predictive Layer is a Schneider Electric Technology Partner that supports electrical utility and microgrid managers who are seeking to forecast energy supply and demand across their grids with higher accuracy and speed. To learn more about how AI tools can better manage grid consumption forecasting:

Schneider Electric Exchange

Tags: , , , , , , , ,