GAS COMPRESSOR UNIT PERFORMANCE MONITORING USING FUZZY CLUSTERING

This paper concerns development and implementation of an on-line performance monitoring system of a natural gas compressor unit. The system can be used to monitor and evaluate the operation and performance of a gas compressor. The system was built for the use of natural gas network operators and maintenance personnel of Gasum Oy in Finland. This paper focuses on how the reference values of a compressor performance can be obtained if reference is not available. A short overview is made also to the capacity loss due to the gas compressor degradation, compressor operating map and their normalization due to changing process conditions. History data of pressure-, temperature-, and rotational speed measurements of the gas turbine’s axial compressor were collected in order to create the reference value for performance estimation. The monitoring system has to produce reference or expected values of the prime performance parameters in order to reliably evaluate the performance and condition of the compressor. The reference value in this case is the polytropic efficiency of the axial compressor calculated from the history data using a fuzzy cluster model. This paper explains the main principles of how fuzzy cluster model can be built for performance reference value calculation. This method can be used for many types of compressors and applications where reference data is not otherwise available provided that sufficient amount of history data is available. The on-line performance system was developed and installed to one gas compressor unit. The operator interface consists of series of graphic displays presenting the current operating conditions and the reference values of a compressor unit. The validity of the performance calculations was evaluated by third party performance test. The percentual deviations were small. With the help of the performance monitoring system the performance degradation and fouling of the compressor unit can be evaluated and monitored. Verwijderd: is about Verwijderd: , which