PowerAll: Utilizing Data to Future-Proof the Rail Network

The Challenge: Growing Demand, Static Infrastructure

The anticipated increase in rail traffic creates significant bottlenecks in energy supply. For example, increased traffic raises the thermal load of substations. This could lead to failures and disrupted passenger service. Another example is the voltage on the overhead wires; with more trains on the same infrastructure, the voltage becomes more volatile. This risks damage to the wires when voltage is high and forces reduced speeds when it is low. Expanding infrastructure would solve these problems, but it is expensive.

The Solution: Insight and Forecasting

To make the most impact, ProRail needs to invest in the most imminent bottlenecks first. For this, insight into all assets is required. Lynxx enables this with team PowerAll by gathering data from numerous sources and transforming them into valuable insights and forecasts. Examples include verifying billing accuracy, forecasting substation thermal load and providing insights with an LLM Agent.

Billing Accuracy Verification

One data source that is ingested is the invoices sent by grid operators. Each month, hundreds of ProRail’s grid connections are invoiced. Verifying the correctness of each of these manually is time-consuming, but small mistakes on invoices do add up. Team PowerAll automated this process. As soon as an invoice is received, a data pipeline is executed that verifies consumption, rates and calculations. In the case of any inconsistency, users are notified. This saves money that can be used to improve infrastructure — for example to make overhead wire voltages less volatile.

Forecasting Substation Thermal Load

To prevent thermal overload of substations for new timetables, PowerAll developed a forecasting model. By combining consumption data and realized train services, the team trained a machine learning model. This model is used to determine the risk of substation thermal overload for a proposed timetable. Because the model is fully automated, new predictions are delivered fast and cost-effectively — especially compared to the multi-month projects that were required before. This model therefore saves money, allows for quicker iteration and more out-of-the-box timetable designs.

Insights by an LLM Agent

As all data was stored and transformed according to data engineering standards, PowerAll could quickly capitalize on the evolution of Large Language Models (LLMs). Because it was structured effectively, it was a streamlined process to provide an OpenAI model safe and secure access to the data. This created an AI Agent that enables ProRail specialists without coding skills to perform complex analysis, freeing them from the limitations of premade dashboards. By combining their expertise with the data access of the Agent, ProRail can optimize infrastructure usage more efficiently.

The Result

By unlocking insights from vast amounts of data, team PowerAll enables ProRail to make the most efficient use of its energy infrastructure and budget. Automated pipelines ensure the insights are up to date, quick and cost-effective. Moreover, because the data is structured effectively, new tools or use cases like LLMs can be deployed rapidly. Millions of euros have been saved thanks to the project, enabling improved train service in the Netherlands.