In the world of Big-Data analytics, quite often, organizations get confused by the vast amount of data that they could potentially harness. One of the easiest ways to make use of this mountain of data is to use it to simply identify the anomalies buried in the data, and alert the user of the anomalies. Anomalies are defined as an incidence or occurrence when the actual result under a given set of assumptions is different from the expected result. In the world of data mining, anomalies are an incidence or occurrence which do not conform to an expected pattern in the dataset. Anomaly detection refers to the problem of finding patterns in data that do not conform to expected behaviour. These non-conforming patterns are often referred to as anomalies, outliers, discordant observations etc. in different application domains.
Anomaly detection techniques have significant application in wind turbines, solar plants, transformers, sub stations and other energy assets which generate vast amount of data. We aim to use anomaly detection to facilitate us in identifying impending failures in sub-systems such as gearbox, generator, main bearing, inverters, string monitoring boxes/combiner boxes, transformers etc. In this post we will look at some approaches to anomaly detection.
Techniques to identify anomalies
The beauty of anomaly analysis lies in identifying the areas of under-performance and over-performance. Using statistical routines to identify anomalies is very easy and can be achieved by most reporting or analysis tool, so let’s look at a few different anomaly detection techniques:
These simple anomaly detection techniques can help identify a range of non-standard behaviour from data of wind farms, solar plants, transformers and other energy assets. Once we identify a deviation or anomaly, we need to determine the correlations and potential causation of the anomalies.
What next once an abnormality is identified?
Once these anomalies are identified, one can now take the next step in the analytics process i.e. in trying to associate meaningful causes or correlations of those anomalies. For example (ref: below image), say for a given set of turbines under same load & other operating conditions, one of the turbine`s Generators drive end (DE) side shows abnormal high temperature. Based on this data, one should go for trend and comparative analysis for a few days. The detailed trend data can then be used to conduct thermography (hot-spot detection test), post which preventive maintenance of the asset can be initiated.
Thus, anomaly detection is a simple way to start any big data analytics journey by deploying some basic statistical techniques to help identify abnormalities in the data that might be indicative of a bigger problem or opportunity.
Hence, we can conclude that the path to business optimization is built on data; the more detailed, fine-grained data, better the position that we are in to identify the anomalies that can start an organization on the analytics journey. This will be the first of many steps to leverage the power of big data for wind farm monitoring and solar PV monitoring.