20.02.2025
HVAC Case Study
Heating, Ventilation and Air Conditioning
BOMP is a fast deployable machine learning algorithm for material and process optimizations. Here BOMP is used to optimize the energy recovery of a medium sized office building with workshop areas in Southern Finland. The input data is a printout of historical data during one month in fall 2024 with one minute intervals from building automation software controlling heating ventilation and air conditioning (HVAC) in normal operation conditions.
The data consists of 14 columns (dimensions) such as timestamp, temperatures of rooms as well as pressures and amount of heat recovered of the heat exchange unit. The schematic illustration of the system and reduced example of the data is shown in Figure 1.
The data consists of 14 columns (dimensions) such as timestamp, temperatures of rooms as well as pressures and amount of heat recovered of the heat exchange unit. The schematic illustration of the system and reduced example of the data is shown in Figure 1.
Figure 1: (LHS) Schematic illustration of the HVAC system system and (RHS) example of data.
The standard operation of BOMP is that the user gives the data as an excel sheet, chooses the columns that are controllable and the target column to optimize. Here we choose all temperatures and pressures (columns 2-13) as controllable parameters and recovered energy (column 14) as the column to be optimized. The optimization software takes approximately 10 seconds to produce the resulted document in PDF format as well as the optimal control parameters in machine readable format for building automation interfaces. On average, it takes 5 minutes for a first timer to get optimal parameters starting from the excel sheet to final PDF report.
The main results are illustrated in Figure 2, where the 14 dimensional data is reduced to 2 dimensions for simplicity. The algorithm found 5 operating modes for the building, where one (purple) is recovering energy significantly more than the others illustrated by light background. Two operating modes (red and blue) are consuming energy illustrated by black background while the last two can recover some energy (yellow and cyan). In addition, BOMP algorithm can further explore optimal heat recovery utilizing relations between the 13 control parameters. In this case, the optimal heat recovery would be 130 kW compared to the historical best of 100 kW. Assuming the energy price of 0.10 EUR/kWh, the BOMP algorithm saves 13 euros per hour compared to no control at all and 3 euros per hour compared to simple table seek of historical data.
The main results are illustrated in Figure 2, where the 14 dimensional data is reduced to 2 dimensions for simplicity. The algorithm found 5 operating modes for the building, where one (purple) is recovering energy significantly more than the others illustrated by light background. Two operating modes (red and blue) are consuming energy illustrated by black background while the last two can recover some energy (yellow and cyan). In addition, BOMP algorithm can further explore optimal heat recovery utilizing relations between the 13 control parameters. In this case, the optimal heat recovery would be 130 kW compared to the historical best of 100 kW. Assuming the energy price of 0.10 EUR/kWh, the BOMP algorithm saves 13 euros per hour compared to no control at all and 3 euros per hour compared to simple table seek of historical data.
Figure 2: The operating modes in reduced dimensions show clearly that one operation mode (purple) recovers significantly more energy (100 kW, light background) than others (< 10 kW, dark background).
In conclusion, the BOMP algorithm provides a simple and effective way to improve HVAC performance. It turns detailed building data into clear recommendations, helping to identify the best operating modes and settings for maximum energy recovery. By increasing energy recovery by 30% from 100 kW to 130 kW, BOMP not only enhances system performance but also delivers real cost savings. This case study shows that BOMP can make building operations more efficient and sustainable, offering a practical solution for better energy management.
