The winds of digital change in generation
- Jun 1, 2018 9:56 pm GMTJun 1, 2018 8:08 pm GMT
- 3050 views
This item is part of the Special Issue - 2018-06 - Offshore Wind, click here for more
The terms Big Data, Data Analytics, Artificial Intelligence (AI), and Machine Learning (ML) are being used increasingly in articles and publications about energy - but what is the real intrinsic value customers are trying to achieve, and what are some of the specific types of analytics that can have a real impact on generating assets?
Let’s look at where analytics can be used to create value, the types of analytics, and then some specific examples that are impacting the real world of generation. The table below shows some different areas where analytics are applied, and the types of analytics:
Here is a deeper look at the categories of Analytics to consider:
Descriptive Analytics - Events
Most everyone who is using sensor data is familiar with descriptive analytics such as simple threshold based alerts. These have been the foundation of monitoring approaches for many years, and remain the starting point for using sensor data. In order to fully optimize asset performance, however, many performance engineering teams are now moving onto higher value categories, including anomaly detection, and predictive analytics.
Predictive Analytics using Anomaly Detection
An example of how anomaly detection can be very impactful is in the use of anomaly detection to predict when a malfunction is going to happen. At NarrativeWave we built such an analytic while working on bearing temperature data on wind turbines. We noticed that before a bearing temperature would trigger its threshold based alarm, it would start to show temperature behavior different from its neighboring bearings. Using our rapid development tools, we were able to create an analytic in a matter of a few hours that would automatically identify this anomaly, check other factors (such as whether the sensor itself was trustworthy, and the wind speed), and notify operators of a highly probable future fault on the turbine. This use of anomaly detection is very powerful and easy to implement, because it does not rely on advanced Machine Learning techniques - rather just a keen understanding of how the asset should be operating, implemented into an analytic.
Prescriptive analytics are considered by some to be the final step in an Analytic framework, and consist of answering “What is the next step?”. In order to do this effectively, several elements must be incorporated into the analytic:
- What happened?
- Example: A thermocouple crossed a threshold
- What is the holistic situation when it happened?
- Example: the turbine is running at maximum load, the ambient temperature is high, this is second time this has happened in the last 12 hours, the wind speed is high.
- What are the possible causes of it happening?
- Example: The thermocouple is faulty, the wind speed exceeds the limits, or the bearing is showing an anomaly
- What are the recommended next steps?
- Example: Shut down the turbine, replace the thermocouple, monitor turbine closely for next 12 hours.
How can this be used is to automate decisions in a control room setting? A large renewable operator, for example, instead of having technicians manually diagnose certain faults, now uses prescriptive analytics to automate and accelerate decision making via NarrativeWave. This increases the accuracy of decision making, standardizes responses, and improves return to service times for turbines.
There are many new ways that analytics can be used to optimize the performance of generating assets. We at NarrativeWave are big fans of analytics that incorporate domain experts’ knowledge in order to maximize the impact of the analytic and enable the automation of decision making processes, while still keeping engineers and domain experts involved. Whatever your approach, we recommend starting with a Pilot that is used to prove out a focused target area, and then scaling across your operating fleet using the lessons learned.
NarrativeWave empowers domain experts to collaborate and rapidly create analytic models to improve asset performance and operational efficiency. NarrativeWave lets users create decision maps that can analyze events holistically and determine what to do next. Domain experts who know their assets can now apply their knowledge to create analytical models and decision outputs, without needing to be a software developer or data scientist.