The life-cycle costs for compressed air equipment are just under 30 per cent of the capital expenditure – the remaining 70 per cent must be spent on the energy required for operation. Companies are therefore well advised to take advantage of all optimisation opportunities. Various projects in the early 2000s made it clear that there is no single recipe for increasing the efficiency of such systems. The simple reason: no two companies are alike.
The company Xervon in Cologne has now taken another approach, writes Hans-Jürgen Bittermann, an author at the trade journal ‘PROCESS’. The compressor house looked after by Xervon in the Cologne-Merkenich Chemical Park for supplying compressed air to three chemical companies was modernised together with an energy supply company. The core of the conversion work was the investment in new compressors (two turbo compressors and three screw compressors). The redundantly designed compressors were connected to each other via a ring line.
The extensive sensor network was designed as a real-time monitoring system for all relevant measured variables in order to derive an optimal control system. Among other things, the pressure, the amount of air taken off, the air humidity, the ambient temperature and the power consumption are recorded digitally throughout. The control is based on manually programmed templates that determine which compressors should run under which conditions.
Procedures and Challenges for Optimisation
The goal is to optimise the operation of the system overall based on empirical values from the past. For this purpose, the recorded operating data is to be analysed with modern algorithms from the field of machine learning in order to understand correlations in a profound way and to derive perfect control from them.
Specifically, these findings in the field of data science/machine learning will be combined with the plant knowledge of Xervon’s experts. The result is an overall system consisting of compressors from different manufacturers that are also highly dependent on site-related influencing factors. In order to evaluate the potential, two independent strategies were investigated: In the first, the current operation was evaluated using an ideal process, thus determining its efficiency.
In the second, a substitute design was fitted to the data based on the compression models developed by the company.
A special optimisation algorithm then searched for the control parameters with the lowest power consumption. With these strategies, a reduction in power consumption of a good four per cent could be demonstrated.
Artificial Intelligence Strengthens the Digitalisation Trend
In this optimisation challenge, a team of researchers from TU Dortmund University uses current algorithms that have already been tested for the optimisation of related disciplines.
This approach has already proven itself in the field of maintenance: In a past project, the service provider Xervon used artificial intelligence to develop an assistance system that facilitates the operation of the serviced cooling towers in Cologne-Merkenich. The tool developed as part of the project uses this data and a regression model to generate a forward projection for the energetically optimal operation of the cooling towers.
It is quite clear that digital, data-driven services will significantly expand their market share.