Power companies that leverage existing data generation and collection tools for input into predictive analytics software are achieving early warning notification of potential equipment problems days, weeks, or months before failure.
In today’s digital world, an optimally run power plant relies on valuable and accurate data in order to ensure assets are running reliably and with minimal downtime. Streaming in from control systems, SCADA systems, distribution management systems, energy management systems, equipment sensors, plant historians, and many other sources, this data must be organized and available to many resources if optimization is to be effective.
When managing a large fleet of assets across numerous sites, making sense of the streams of operational data coming in from multiple systems and from remote sources can become an impossible task. Most utilities today are already using real-time historians to consolidate disparate data sources for a more comprehensive view of plant performance. Building on that data management foundation, predictive asset analytics solutions can leverage data from the historian to further improve operations by providing early warning detection of equipment issues before they lead to failure.
To achieve the next level of operational efficiency and make the most of the increasing amount of information available, generating companies are investing in predictive maintenance programs, including the implementation of predictive asset analytics software. Predictive analytics software helps organizations to gain the greatest return on every single asset and, in turn, improve overall plant operations.
Data Management Foundation
The first step in implementing a predictive maintenance program is having the data management foundation in place. At most utilities, relevant asset information is typically stored within a historian. The historian captures and archives continuously streaming real-time data from any type of asset or system, regardless of manufacturer. This is beneficial when maximizing the usefulness of time-series data that is often most relevant in real time. A single data management system ensures that all necessary information can be made available and compared with other data streams quickly and accurately. This increases situational awareness and builds the foundation for holistic visibility of the entire enterprise.
Predictive Asset Analytics
After ensuring that a solid data management infrastructure is in place, companies can implement predictive analytics software that uses continuously streaming time-series data for real-time insights. Predictive asset analytics software helps organizations improve operational performance by using predictive algorithms to calculate and predict normal asset behavior. For example, Schneider Electric’s Avantis PRiSM software is based on advanced algorithms that use advanced pattern recognition and machine learning technology, which has been shown to provide weeks to months of early warning notification of deviations in operational performance in power plants (Figure 1).
The software uses empirical data to learn normal performance modes or patterns for the equipment that is being modeled. That empirical data is captured from the data historian and should not require additional investments in specialized sensors. Although equipment modeling can be a complicated task, some solutions are designed with tools and templates that simplify and streamline the process. This allows models to be created in a matter of minutes, rather than days or weeks.
When implementing a predictive asset analytics solution, the organization identifies which assets to monitor based on strategic importance to the business. The models are typically implemented in phases, and assets that have had continuous problems, those that directly affect availability, or those likely to have a quick return on investment may be modeled first.
Predictive analytics software enables utilities to spend less time searching for issues or, worse, waiting for an important piece of equipment to fail. Instead, staff is immediately notified through an alert that an asset is not operating as expected. These insights help engineers and operators better plan and prioritize maintenance activities, such as determining when an asset can continue running as is or needs to be serviced or replaced.
Early Warning Notification
In one example, a large North American power utility with more than 50,000 MW of generating capacity and plants in multiple states had estimated savings of more than $8 million in avoided costs over one year through early warning detection of issues that indicated an asset was headed toward failure.
In a significant catch, plant engineers received an email notification from the predictive analytics software indicating that an aging steam turbine experienced a vibration step change that would not have been detected through standard monitoring practices. Plant personnel verified that a proximity probe and casing vibration had both changed. Further analysis showed a possible loss of mass in the turbine blade path. Based on the unit’s operating history, staff immediately suspected shroud material had been lost. It was determined that the unit could continue to run at a reduced output, under increased observation, until a more convenient and strategic time to bring it offline.
After the unit was brought offline, a borescope inspection verified missing shroud material and revealed several other segments that were close to liberating (Figure 2). Had this issue not been identified with predictive analytics software, it could have caused immediate unplanned downtime, loss of generation, possible catastrophic failure, and danger to personnel. This single early warning notification and the following action resulted in a potential estimated savings of more than $4 million in lost revenue and repair costs, in addition to maintaining the safety of the operating engineers.
Validation: Monitoring Across Multiple Plants
A midsize generation and distribution utility in North America uses the Wonderware eDNA real-time historian to manage generation data across seven sites, and each site has a local redundant history for up to eight years of data. The highly available historian uses lossless data compression to quickly store and retrieve operational data. The data management infrastructure enables efficient sharing of information between systems, applications, and users. The utility uses the historian for electronic discharge monitoring reporting, load forecasting, heat rate analysis, and to build custom display screens. The historian’s display screen tool is used for real-time visualization of operations, including interactive coal-handling operator screens. The historian solution is also integrated with Avantis PRiSM, enabling advanced pattern recognition and machine learning to identify when equipment is not operating as expected during all ambient, loading, and process conditions.
In this application, the predictive analytics software is currently used to monitor more than 200 assets—including coal-fired units, combined cycle units, biomass cogeneration, and wind turbines—at a centralized monitoring center. Assets like boiler feed pumps, turbines, combustion fans, condensers, and feedwater heaters, among many other pieces of equipment, are monitored for mechanical and performance degradation.
In one early warning notification, the predictive analytics software first alerted personnel of abnormal conditions in the Overall Model Residual (an indicator of overall system health) and alerted again on individual vibration sensors. The plant staff was notified, and vibration technicians inspected the issue. Analysis of field data indicated fan support issues, with similar fans experiencing the same issue to a lesser degree. The utility was able to continue running the equipment with increased supervision and scheduled an inspection with the fan manufacturer during the next planned outage. The support system was corrected on all fans during the planned outage.
With the early warning notification, this utility was able to resolve the issue before it substantially damaged other asset components and avoided a forced outage while also improving equipment performance. Other considerable catches at this utility include a water cooler lubrication system failure, a partially plugged control valve, a critical component sensor failure on a pulverizer, and a wind turbine main bearing failure.
With deployment of the historian and predictive analytics software, this utility has:
■ Developed a robust data archive and retrieval architecture for reliable access to real-time and historical data.
■ Created interactive display screens for improved visibility into operations.
■ Developed accurate and standardized reports, reducing the time to produce them.
■ Identified early indications of asset failure while improving reliability and safety.
These results are not uncommon. By leveraging predictive analytics and fault diagnostic technology, Avantis PRiSM customers have reported a 25% reduction in equipment downtime, 25% reduction in operations and maintenance costs, and payback on their investment in three to six months.
Ongoing Operational Benefits
Using predictive analytics software, personnel understand the actual and expected performance for an asset’s current ambient, loading, and operating conditions. They know where inefficiencies are and their impact on financial performance and can use this information to understand the effects of performance deficiencies on current and future operations. This information also helps generators assess the risk and potential consequences associated with each monitored asset and can be used to better prioritize capital and operational expenditures.
— Justin Thomas (firstname.lastname@example.org) is a business development manager for Wonderware eDNA and Avantis PRiSM at Schneider Electric.