Initial situation
Cooling supply systems in industrial buildings have traditionally been designed and operated in a conventional manner, without considering multidimensional, time-variable, and stochastic dependencies. In practice, almost exclusively conventional control systems such as two-point or PID controllers are used. However, the operation and efficiency are influenced by many factors, such as weather, internal loads, or energy markets, which are often not taken into account in conventional control. On the other hand, AI methods can be successfully applied, especially when an optimal operating mode in a multidimensional solution space needs to be identified repeatedly in a project-specific manner, and the system to be optimized is subject to high dynamics. This is inevitably the case for cooling technology in industrial buildings: On the source side, dynamic internal building loads, plant waste heat, and transmission and ventilation heat must be dissipated, while the heat sink, particularly the ambient air, is subject to daily and seasonal fluctuations. At the same time, these systems have a multitude of possible control parameters, which have different effects on the overall system depending on the plant topology, technology used, and installation location. Scientific publications and potential analyses conducted with industrial partners indicate high potential for cost and CO2 savings. By using AI algorithms, typically 10-40 % of energy costs can be saved depending on the previous mode of operation.
Objective
The overarching goal of EISKIG is to build an autonomously operating system that independently analyzes and optimizes the operating strategy of relevant systems using AI-based optimization methods to increase energy efficiency and flexibility, minimize implementation effort, and increase user acceptance. In this context, the project aims to achieve at least 15% energy savings in selected cooling supply systems of project partners from the chemical and pharmaceutical sectors, drive and control technology, and the IT industry.
Approach
The EISKIG research project identifies energy wastage or efficiency potentials in building cooling technology through data- and AI-based optimization methods, derives measures for optimized plant operation, and implements these in industrial practice with the project partners.
At the beginning of the project, in TP 1 (System Understanding and Concept), the technical and organizational requirements of the application partners are collected and the system boundaries are defined. Subsequently, in TP 2 (Data Infrastructure), the real plants of the application partners are connected to a central data processing platform via IoT gateways, where the forecasting, simulation, and optimization applications are operated. The analysis of the recorded time series data takes place in TP 3 (Data Analysis). The goal is to gain deeper insights into the dependence of the energy system on external influences and to predict these using statistical methods and machine learning techniques. In TP 4 (Digital Twins), the selected energy systems are then simulated based on the recorded data to test the optimization of plant operation on simulation models of the supply systems. With the insights from TP 3 and TP 4, the optimization of the real plant operation in TP 5 (Operation Optimization) is finally possible. Since the transfer to other companies has so far failed due to the high effort required for the implementation of the procedures, software modules will be developed in TP 6 (Scaling), which will significantly facilitate the application and transfer of the procedures.
Funding source
Project sponsor
Consortium partners