Over the last few weeks, solar panels and smart meters reached the front pages of the news. Other than evoking a wave of discussion, it automatically made us think about the possibilities of time series forecasting, or more precisely: predicting energy consumption. 

The principles of time series forecasting have been with us for ages: looking at data over time to predict what will happen in the next time period. With the powers of deep learning knocking on the door, time series forecasting is ready to be revolutionized. And that might thoroughly shake things up for the owners of smart meters, the grid operator and the whole energy industry itself.

Solar struggles

Looking at the bright side, the upcoming smart meter introduces an exciting milestone in grid technology. It provides countless opportunities for innovation, energy efficiency, grid reliability and management. However, for some people, the smart meter comes with complications. 

People with solar panels on their roofs struggle to earn back their investment with the disappearance of the “reversible counter” and the introduction of the smart meter. The smart meter isn’t necessarily a bad thing, but it pushes solar panel owners to consume energy at times when production is high, namely at daylight hours when the owner isn’t home. The unused energy gets transferred back to the grid, but unfortunately the financial compensation isn’t equal to tapping power of the grid. The owner is again forced to use energy from the grid in the evening, making his solar panel investment next to negligible. 

How to get out of this pickle? Naturally, the answer isn’t one-sided, but gaining thorough insights into your energy consumption can make that solar investment worthwhile. That’s exactly where deep learning comes in.

Deep learning unleashed

After reinforcement learning, computer vision and natural language processing, time series forecasting is next in line to be disrupted by deep learning technology. Unlike traditional methods, deep learning models can process tons of data simultaneously and generate a single, general model that’s relevant for a large group.

Let’s project this on energy consumption. A traditional forecasting model would need a month’s worth of data to make an accurate forecast of a single household’s energy consumption. Deep learning, on the other hand, can combine learnings from thousands of households into one general model that can instantly be used for a whole group of households. On top of that, the deep learning model can easily include known categorical features like holidays, seasons or location to make predictions even more accurate.

Making smart meters smarter

Smart meters that use these deep learning models can perform precise energy demand forecasting. When your smart meter is aware of your household’s energy consumption, it could recommend measures to optimize your energy efficiency. Ultimately, it should even show how much you would save by acting differently. 

Thanks to the increasing amount of smart home appliances, you can easily control your devices remotely or make them start at a specific time. The combination of this specific forecasting and remote-controlled smart appliances makes it perfectly possible to consume your energy in the most cost-efficient way. To put it plainly: owners of solar panels could accurately use the energy when they have it, optimizing their energy consumption and saving money.

Benefits for the people, the grid and the industry

Additionally, an energy consumption forecast would be incredibly useful for grid operators. It would enable them to build and optimize a smart control centre that allows them to anticipate accurately on user-demand. When aggregated, the energy demand forecast provides information about when and where energy is needed, resulting in better control operations, better grid monitoring and the prevention of energy waste. 

Energy demand forecasting also acts as an accelerator for innovation. A house battery could be another very efficient way of increasing self-consumption. At this point, its expected lifespan does not yet compensate for the investment. However, energy demand forecasting could drastically improve the efficiency of home batteries, making them a welcome piece of technology for every household in the future. 

The blue ocean of time series forecasting

Energy demand forecasting is only the tip of the iceberg of countless real-world applications with a meaningful impact. We’re seeing a lot of untapped potential in sectors like retail, healthcare and finance. Just think about enhanced sales forecasting or product price monitoring for bankers.

Now that deep learning technology has proven to deliver a substantial increase in prediction performance, time series forecasting will surely be a force to be reckoned with in the future.