SSP Project Summary: Optimising Compressor Lifetime

Student

Supervisors

David Henty, EPCC

Mark Sawyer, EPCC

In the power utilities and the oil and gas industry, networks of turbines or compressors are used to maintain a fluid or gas flow through a system at a specific rate.

These compressors must undergo off-line maintenance to ensure they continue to operate without catastrophic failure. Generally there is a maximum period after which the compressor must be taken off-line for maintenance or else it is liable to catastrophic failure.

The purpose of this SSP is to discover the most appropriate operation schedule for such a networks of compressors. The schedule should minimise maintenance costs whilst maintaining flow at the desired rate through the system.

Different optimisation techniques will be used to find an optimal maintenance schedule for the compressors. The quality of the solutions from the different techniques will be compared to determine if the effort in using a particular technique is worthwhile. The optimisation techniques to be used are genetic algorithms, simulated annealing, random search and hill-climbing. These will be coded using RPL2, a language for programming optimisation problems in terms of genetic algorithms. The other techniques to be investigated can also be programmed using RPL2.

The constraints to the problem are the flow rate the network of compressors must maintain and the maximum number of days between services for a compressor. In the first instance a very simple network of 3 compressors would be used. The resulting schedule should define the output from a compressor for each day of a 28 day month. For each day of the schedule, the combined output from the compressors should match the required flow rate.

A naive solution is available to compare the results against. This naive solution consists of one more than the minimum number of compressors necessary to maintain the flow, where the compressor off-line for maintenance is always rotated.

Once the framework for finding solutions has been formulated, the analysis could be made more sophisticated by incorporating fuel costs into the problem. These compressors require fuel to run and generally have a range of operational parameters over which they operate most efficiently and use the least fuel.

Further sophistication could involve incorporating variable maintenance costs where the longer a compressor has been in continuous operation the longer off-line maintenance period and hence the dearer the maintenance costs.

Webpage maintained by mario@epcc.ed.ac.uk