The continuous and reliable operation of the compressor fleet is essential for the delivery of gas across the NTS on behalf of customers. The compressor units must start when requested and deliver their duty when required. A major disruption to the operation of the unit occurs if the compressor trips either during start-up or during operation. Depending on the network demand, flow patterns and criticality of the compressor station a machine trip or failure to start may in the worst case lead to network disruption and supply/demand management charges such as buy-backs. This project aims to better predict trips and failure to start (i.e. before they occur) to enable greater corrective action to be taken to prevent the trip and failed start.
DNV GL has conducted a pilot/feasibility study with the cooperation of National Grid, to have an initial look at data analytics on monitored data for single type of compressor unit. This study showed encouraging results for the successful prediction of running trip and failure to start for fuel system issues. If an estimated 75% of such trips were prevented this would have had a significant impact on improving MTBF and start reliability figures for this type of unit.
Machinery malfunctions may often have leading indicators and patterns in data may predict failure, but often the interpretation of the patterns and data required is highly complex and a compressor engineer may not be able to identify these or manage the volume of data unaided. It is proposed that data analytics and pattern recognition may, together with domain expertise (machine problem knowledge) by compressor engineers provide a new and enhanced method of improving compressor reliability.
Further being able to pro-actively address trips and failures to start is a difficult task as maintenance procedures and processes and the age and complexity of some of the equipment can make prioritisation of tasks difficult. By targeting the maintenance to solve the issues which directly affect reliability should increase the efficiency of maintenance budget use. Increased reliability should help reduce operational costs caused by trips.
Benefits
This proposal has a phased approach and depending on the success of the stages determines which phase it continues to;
- To gathering, cleansing and combining relevant sets of the compressor operating data
- To analysis the data and develop algorithms
- To determination of trip and fail to start predictions and test these against the data
- Define the scope of work for roll out to the compressor fleet and a real-time/on-line prediction tool
