Reduce heating, ventilation and air conditioning power consumption by 10%

Savings for Rheinhold & Mahla

Approximately 80% of HVAC (heating, ventilation and air conditioning) systems use proportional-integral-derivative controllers (PID) for modulated control. PID control logic is easy to implement. However, experience shows that in most cases the technology lacks efficiency in its control method. This is due to the fact that responsive decisions are made reactively.

The rapid development of IoT technologies allows for easy data collection using the HVAC system. This enables building predictive models that help proactively control the system.

In cooperation with the marine industry powerhouse Rheinhold & Mahla, Proekspert devised an HVAC control solution which used model predictive control.

Our tools of wizardry

The analysis of the HVAC systems was done in several stages. Initially, Exploratory Data Analysis (EDA) was implemented to identify the available data and determine its quality. When the EDA was performed, some crew behavior, such as changing setpoints of the chillers, was noted. This resulted in an increase of power consumption. Moreover, throughout the analysis, problems were found with the system design, like low delta T syndrome.

After the EDA was complete, the second stage of the analysis was performed. The second stage focused on the work process, efficiency, and potential improvement of the HVAC system. To increase the efficiency and the effectiveness of the analysis, Proekspert executed interactive visualizations that offered new viewpoints for the existing designed system, which for the first time made visible how the system actually lives and breathes on operational conditions.

To understand how the system reacted to changes, and how different parts of it affect each other, Correlation Analysis was performed in the next phase. Based on the collected knowledge and observations, Proekspert created a model predictive control method. According to this method, grey and black box system identification techniques were used to acquire a model for simulating the future states of the system. The simulation models enabled optimization problem formulation which resulted in HVAC savings.

Results

Proekspert worked on a pilot on two RoPax class ferries. Using R&M’s domain knowledge and Proekspert’s technical experience, a solution was devised that gathers data remotely from the vessels, uses trained models to simulate different control decisions, and sends the optimal decision to a local PLC which communicates the command to a device.

• Despite there being only a few months of operational data, the prediction models already achieve a great deal of accuracy. The temperature in arbitrary rooms can be predicted with an uncertainty of merely 0.2°C.
• Energy use of large devices like chillers is linked to weather and internal configurations. Using predictive models, a control policy was devised that outperforms the current one by 10% of total energy consumption.
• Control decisions (such as room temperature setpoint) are performed autonomously, i.e., not requiring human intervention. However, if the need arises, setpoints can be overridden by the crew.

The HVAC Optimization case is published by Labs Network Industrie 4.0 and publicly introduced by Dr. Dominik Rohrmus (Head of the research group Manufacturing Systems at Siemens Corporate Technology, Germany).