| |
|
|
|
|
Big Data: Data Mining Possibilities
The examples of data mining and related results in this paper are part of a huge list of possible information deliverables that can be achieved by many utilities today. Some utilities are still struggling from the paper mode to the electronic mode, but a great number have already made that transition.
The most common first step is the implementation of a CMMS of some kind. That is a very rich body of data for equipment asset management. The CMMS can support work management with asset population age distribution as it relates to maintenance cost, frequency and failure rate. Appropriate coding of work types can be used to help identify "bad actors" through rework monitoring and other very specific equipment problems can be spotted by recurrence of a special work type code (e.g. window replacement may identify locations where design change is needed to prevent projectile damage).
Another common data source is a process data historian. These are essentially servers, usually dedicated, that are fed information from a plant process computer, a SCADA system or a digital control system that the historian then compresses very rapidly and efficiently. While performing the compression on an ongoing basis the historian also responds to specific requests for selected historical data that can be used to assess operational events and their impact on the assets. This data provides the basis for an enormous number of data mining applications that often rely on an understanding of the mechanisms that affect the asset and then perform the assessment on a routine basis. Monitoring the maximum loading of transformers, the number of times an LTC or circuit breaker have operated since their last overhaul, the time since a circuit breaker has last operated and many, many other applications are possible.
Another key information system is the outage database that is used to record interruptions of service, the cause and the assets and customers involved. The ability to combine data from an outage data base with that from the CMMS and the operational data historian can provide very valuable insights into what can be done to improve system performance. Operational data from a historian may never be a direct source for the multidimensional data mining tools due to the manner in which it is stored. Selected portions of the data could be extracted and collected in a data warehouse where multidimensional analysis could be applied. However, for any analysis to be effective, the relationship between the data point ID and the asset will have to be established.
There are many other data systems that have significant value as sources for data mining, including those that store predictive technologies data (e.g. power factor, vibration, temperature, etc.), inventory data and even human resources data. Key to the efforts to establish these systems as useful contributors to the data mining environment is making sure that the data in each one can be easily related to the assets that are being managed.
Next: Data Mining and the Business Process.
Back to Abstract.
Back to Services.
|
|
|
