Intelligent Energy Systems

A reliable data foundation is essential for any intelligent energy system. We combine energy monitoring and consumption analysis with non-intrusive load monitoring (NILM), enabling the breakdown of total loads into individual consumers and the creation of benchmarks across machines and systems. This transparency supports targeted efficiency measures and helps companies integrate these insights into energy management systems in line with ISO 50001.

To make data easy to work with, we use AI-based assistants (LLMs). They generate reports in natural language, highlight key findings, and directly transfer recommended actions to downstream systems.

Accurate load forecasting is a key component of intelligent energy systems. We develop methods to generate highly accurate, probabilistic, and hierarchical forecasts, incorporating production data, weather information, and market signals. These forecasts enable predictive control of energy consumption, cost-optimized procurement of electricity and heat, and the reduction of peak loads and balancing energy costs. Companies benefit from lower grid charges and greater flexibility in responding to volatile energy prices.

Alongside forecasting, early anomaly detection plays a critical role. We combine AI-based and physics-based methods to identify unusual patterns in energy use and production processes. This enables, for example, the detection and localization of leaks in compressed air systems. Energy signatures also support condition monitoring of equipment and enable predictive maintenance.

Self-learning models detect recurring patterns and provide the basis for automated analysis. AI-based assistants present results in a clear and intuitive way and integrate them into existing energy management systems. This helps reduce energy losses, prevent failures, and plan maintenance more efficiently.

Leaks in compressed air networks increase energy costs and can disrupt production processes. Our approach combines leak detection, localization, and hotspot analysis to accurately identify leaks. This involves analyzing pressure, flow, and energy signature data, as well as atypical load patterns.

Each detected leak is prioritized based on its potential energy loss and operational criticality. The results are integrated into energy monitoring systems and maintenance planning. This supports the reduction of energy losses and enables targeted allocation of maintenance resources where they provide the greatest benefit.

Linking load forecasting with intelligent energy and load management unlocks new potential for flexibility and cost efficiency. We develop models that evaluate price signals from day-ahead and intraday markets, as well as grid signals and capacity charges, and align them with production constraints.

To this end, we model and optimize the use of battery storage, thermal energy storage (heating and cooling), and flexible loads. These methods generate dispatch schedules for on-site consumption, grid charges, and energy procurement, and, where applicable, enable participation in balancing energy markets.

The results are integrated into energy management systems and continuously evaluated using key performance indicators such as peak loads, energy costs, and CO₂ emissions.

Large language models (LLMs) open up new possibilities in energy management. They process complex energy and production data and summarize it in natural language reports. Our AI-based assistants can prioritize anomalies, analyze root causes, and generate recommended actions.

The reports are integrated into existing energy management processes, while explainable workflows ensure that results remain transparent and traceable. This creates an intuitive interface between data analysis and operational decision-making, supports the continuous optimization of energy and load management, and improves efficiency in practice.