Energy as a Controllable Resource

Conventional load management – reactive cost reduction

The traditional objective is to reduce operational energy costs by limiting peak loads (“peak shaving”). When a predefined threshold is exceeded, loads are curtailed. However, the potential to respond to dynamic market prices remains largely untapped.

Intelligent flexibility management – our approach

We take a proactive, market-integrated approach. Energy and production systems are treated as a connected optimization problem. AI models plan loads, storage, and generation based on forecasts of prices, weather, and processes. The goal is to maximize overall economic value – through cost reduction, revenue generation, and CO₂ reduction.

This approach is built on seamless data integration: production, energy, and market data are combined in a digital twin. This virtual system model enables the simulation, evaluation, and real-time control of flexibility. Control is achieved through a learning feedback loop of forecasting, optimization, and automated execution. The system independently identifies when load shifting is technically feasible and economically viable, and implements decisions automatically via standardized interfaces (e.g., EFMS API).

The result is an intelligent, energy-flexible factory – a virtual power plant that actively responds to market, price, and grid signals while combining energy efficiency, economic performance, and climate impact.

1. Potential analysis & the “energy footprint”

Every optimization starts with understanding. We determine the energy footprint of your production – a key metric that reflects energy consumption per order or product. Based on this, we identify your flexibility potential:

• Inherent storage: We analyze processes with natural thermal inertia (e.g., melting furnaces, cold storage, building envelopes) or process-related buffers (e.g., compressed air storage, material inventories) that enable temporal decoupling of energy use and production.
• Shiftable loads: We identify energy-intensive processes whose start times can be flexibly adjusted within defined limits.
• Quantification: We precisely determine how much power (kW) can be shifted for how long (kWh) without affecting your core processes.

2. Predictive forecasting as a basis for decision-making

Forward-looking control depends on the quality of the forecasts. We develop high-precision, probabilistic forecasting models for all relevant variables:

• Load forecasts: For non-controllable base loads.
• Generation forecasts: For volatile on-site generation (e.g., photovoltaic systems).
• Price forecasts: For relevant energy markets (day-ahead and intraday) to support operational and trading decisions.

3. Mathematical optimization and AI-based decision-making

This is the core of our research expertise. We translate production environments into mathematical models and determine optimal operating strategies.

• Model-based optimization: We formulate processes as stochastic mixed-integer linear programs (MILP). These models identify economically optimal operating schedules while accounting for all technical and process constraints (e.g., switching frequency, temperature limits).
• Learning systems based on deep reinforcement learning (Deep RL): For highly dynamic environments, we develop adaptive control strategies. By training through extensive simulations, the system learns forward-looking strategies to respond to uncertain market conditions and to maximize long-term economic performance.

4. Automated control and implementation

Our expertise extends beyond planning. We develop the logic and interfaces required to implement optimized decisions in real-world operations. This includes controller design, integration with control systems (e.g., PLC, MES), and the establishment of robust, bidirectional communication. 

Our methods have been successfully validated and implemented in practice through research and industry projects:

• Price-oriented production planning: We adjust job sequencing or temporarily interrupt processes to shift energy-intensive steps (e.g., heating a furnace) to periods of low electricity prices. In pilot projects, this has led to significant reductions in energy costs.
• Cross-commodity optimization: In heat planning, our system dynamically selects the most cost-effective energy carrier. It automatically determines whether it is more economical to generate heat using low-cost electricity (e.g., via power-to-heat) or gas.
• Active flexibility trading: Based on price forecasts, our systems generate optimal bids for day-ahead and intraday markets. Identified flexibilities thus become a direct and additional source of revenue.
• Strategic tariff evaluation: We simulate the operation of your systems under different dynamic electricity tariffs. This provides a sound, quantitative basis for selecting the most economically advantageous tariff model.

Our algorithms are not developed in theory alone – they are designed and validated under demanding, real-world conditions. At our unique WanDa facility (Flexible Compressed Air Research Facility), we integrate prototypes into an industry-relevant test environment to evaluate their performance in reproducible scenarios.

In this demonstration setup, compressed air flows either through intact hoses or through hoses with barely visible leaks, bends, and imperfect connections – the most common sources of leakage in industrial systems. Regardless of the path the air takes, the difference is invisible to the naked eye. The actuators carry out their tasks as usual.

However, our system captures what remains unseen. It measures whether the air flows at higher or lower pressure, determines flow rates, tracks actuator positions, monitors valve states, and even records ultrasonic signals. With up to 20 individually controllable actuators, we simulate complex industrial load profiles and generate large volumes of data at sampling rates of up to 100 Hz – amounting to as much as 300 GB per operating day.

This extensive, high-quality dataset forms the basis for developing AI models that perform reliably and accurately even in complex industrial environments.