Imagine being able to predict the future with such accuracy that it helps you optimize resources, reduce costs, and even ensure the stability of critical infrastructure. While that may sound like a dream in many industries, it’s already becoming a reality in the energy sector thanks to artificial intelligence (AI). A compelling success story is our project with the German distribution grid operator Westfalen Weser Netz (WWN).
Grid operators such as Westfalen Weser Netz (WWN) face the immense challenge of accurately forecasting energy consumption. This is crucial because they must procure the expected energy demand in their grid one day in advance (typically in 15-minute intervals) on the electricity market to ensure optimal grid operation. But this task is becoming increasingly complex.
The reason lies in the growing share of renewable energy sources like wind and solar. These are highly weather-dependent and make predicting both the available energy and actual demand a monumental task. A stable grid requires a balance between supply and demand, and any deviations must be compensated by other sources such as gas, coal, or hydro. Inaccurate forecasts can have far-reaching consequences, ranging from energy surpluses that need to be diverted to costly shortages. In the worst case, an unstable grid can even collapse.
Traditional statistical models have reached their limits. While they can analyze historical data, they often fail to adequately account for external factors that can cause sudden spikes or drops in energy consumption. Take temperature, for example: a sudden cold snap leads to increased use of electric heating and thus higher grid load. These dynamic external variables are essential for accurate forecasting.
This is where AI comes into play. AI-powered models can incorporate such external factors and assess their impact on energy consumption. However, building and optimizing these models can be a significant drain on time and resources, often resulting in months of work for a team of data scientists. Our Demand Forecasting Solution (DFS) addresses exactly this issue by offering a data-driven, objective approach to forecasting.
The tool is designed to simplify and accelerate the complex forecasting process. It automatically evaluates a wide range of models, currently over 70, to identify the optimal one for each use case. This means that you often get a highly accurate forecast right “out of the box.”
A critical aspect is operationalization. Forecasts need to perform reliably, not just once, but daily. For WWN, forecasts had to be delivered within a tight two-hour time window for the following day. On Fridays, forecasts had to cover Saturday, Sunday, and Monday. The tool allows this process to be automated with just a few clicks, and models can be retrained weekly with new data (automatic retraining) to ensure ongoing accuracy. Even for long forecasting horizons, such as 9-day forecasts during holiday periods, the system delivers reliable results.
The forecasting solution integrates seamlessly into existing processes and tools as an additional building block. Beyond the intuitive user interface, the forecast data can easily be exchanged with other systems via an API interface. Given the sensitivity of such corporate data, data security is paramount. Installing the solution within the customer’s Microsoft Azure tenant ensures that no data leaves the customer’s environment.
Using a modern cloud infrastructure like Microsoft Azure offers another major advantage: the resources needed to train the models can be provided flexibly and scaled as needed, without incurring idle costs.
Efficiency and fast value generation were key success factors for WWN as well:
“In just 2 months, the solution was implemented and trained. The initial results already showed a prediction accuracy like our current software but with the potential for even better results through further optimization.” — Julian Schiewer, Specialist for Electricity Balance Group Management
Mr. Schiewer also highlights that the solution stands out thanks to regular model training and the integration of various data sources. The results have even prompted WWN to consider expanding the tool to other areas of the company, a strong signal of the solution’s flexibility and wide applicability.
Introducing new technologies requires trust, especially in critical areas like energy supply. At WWN, this trust was built through a parallel testing phase. Over an extended period, the performance of the AI forecasts was continuously compared with existing solutions. Monthly reviews and analysis of results under changing conditions (e.g., temperature shifts) further strengthened confidence in the AI solution. Human oversight remains an important part of the process, validating the plausibility of forecasts. Over ime, as the model receives more accurate data, the need for manual review decreases.
The benefits are clear: greater accuracy and a simplified, more efficient process for expert departments. The challenges of precise forecasting are not unique to the energy sector; they’re found across many industries. With our Demand Forecasting Solution, taking the first step is easier than you think.
You are currently viewing a placeholder content from Vimeo. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from YouTube. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou need to load content from reCAPTCHA to submit the form. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from Google Maps. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from Google Maps. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from Mapbox. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More InformationYou are currently viewing a placeholder content from OpenStreetMap. To access the actual content, click the button below. Please note that doing so will share data with third-party providers.
More Information