Today's wind turbines rely on complex electronic supervisory control and data acquisition (Scada) systems. These capture large amounts of important data that, coupled with performance data analytics, can provide critical tools that enable wind projects to meet or even exceed their performance estimates.
There are more than 200 data points collected continuously for each turbine on a typical wind farm, creating many terabytes of information annually. This Scada data is one of the most overlooked resources available to wind-farm operators.
The data is time-stamped and records important information such as wind speed, generator output, and pitch and yaw angles. The key is knowing how to harness this data to find the underperforming turbines and resolve their issues to maximise the project's energy-generation capacity.
Standard issue Scada systems collect this data and can store it for 30 to 60 days before the system will overwrite the existing stored data due to lack of storage capacity. A method for preserving this data for analytics and historical trending is to export the stored data from each turbine on to a server-based relational database system on a monthly basis.
This stored data, known as "batch" data, will become the living record of turbine production and performance. Performing analysis on the database created by batch downloading can enable owners and operators to understand the relative performance of the turbine.
The batch method is a proven procedure to support performance analytics, but the data is stale - at least 30 days old when it is first analysed. The results are only of academic interest as the problem identified may have already been investigated, and possibly resolved using trial-and-error methods in the field, such as observation and swapping of parts. The batch method rarely catches the problem in a timely manner.
The shift from batch to real time can be arduous, but the benefits outweigh the investment and effort. The implementation of real-time data collection involves the installation of a data historian. The software is relatively expensive and also requires methods and applications for capturing the wind turbine's native Scada data. Converting the native Scada data into the historian database structure then involves complex cross-reference tables. Depending on the size of the wind-farm fleet, this can take months to complete.
The proven data collection methodologies of these data historian products allow terabytes of information to be stored efficiently. With the server speeds and low cost of memory for modern computers, these systems are fantastic tools for performance analytics.
The data is collected within seconds of occurring and transmitted to the historian, creating a database that is very nearly real time. When performance analytics methods are applied to this dataset, the resulting information can be shared with the field and problems can be corrected within hours of their identification.
Like any form of rotating equipment, the mechanical and electrical systems tend to wear and the level of performance falls as the turbine begins to age. The performance analysis most commonly used today is called relative performance assessment (RPA) or, more simply, neighbour comparisons. This method compares the power curve of a single turbine with the site's average power curve and the power curves of its neighbouring turbines.
RPA finds the underperforming turbines. If a turbine is not performing at the site average that in itself may not be a problem as some wind farms are large and situated on varied terrain, so the turbines do not all receive the same wind resource. Comparison with neighbouring machines, however, allows for the next level of investigation. If a turbine is not performing to the site average but is matching that of its immediate neighbours, there may not be a problem. But if it is neither performing to the site average nor its neighbours' then it will be identified as a low-performing turbine.
So what can be done about it? Additional graphs can be made using the data to compare the pitch and yaw systems performance with the turbine's neighbours (see charts, below). Usually, one of these comparisons will show a shortcoming in the pitch and yaw systems of under-performing turbines, which gives the technicians an idea of how to resolve the problem.
Generally, these pitch and yaw issues have root causes in a number of areas. The most common cause relates to mechanical issues with the meteorological instrumentation, specifically the anemometer and wind vane. Next, the control system parameters for pitch and yaw angle coordination can become out of calibration and require to be reset. Finally, the pitch and yaw equipment itself may have mechanical wear that causes it to lag in response to control system commands.
Wind-farm operators that use these performance-analytics gain insight into the root causes of problems. The methods direct technicians to parts and procedures needed to return a turbine to full service without requiring an investigative climb up the tower. This approach reduces the number of climbs and makes each climb as productive as possible.
Maintenance activities can be better planned and aggregated with other turbine problem resolutions, making the entire process more efficient.
Safety is a high priority in the renewable-energy business. This use of big data supports that approach. Eliminating unnecessary climbs reduces injury rates. Problems and solutions are identified on the ground so the technician can go up with the correct parts and procedures instead of going up just to investigate.
The number of wind farms and turbines have grown rapidly over recent years, now amounting to tens of thousands. At approximately 200 data points per turbine, this is an immense amount of information. Even if the data was only of the batch variety, think of the insights, education and analytics that would be created if all this data was being captured.
Most installed turbines conform to about ten basic model types. Many of the biggest wind-farm owners have similar fleets in terms of makes and model. Each owner is working to understand its portfolio's issues and determine solutions to its performance problems.
Entities have been created for owners to share their day-to-day issues and discuss their challenges as well as successes. These entities have originated with groups established through industry-wide organisations and publications focusing on operations and performance, with participation from all players, including owners and operators, manufacturers and third-party service providers. There are national laboratories across the US that also collaborate with the same wide membership.
More recently, a third type of entity, focused exclusively on owners and operators, has been established. This has enabled sharing of ideas and innovations without the participation of manufacturers and third-party service providers. Still, the missing link has been a comprehensive method for sharing the performance data.
Teams from multiple owners could work together to analyse the aggregated data to establish templates for the identification of typical recurring issues. On more complex issues, collaboration could allow for the introduction of new and innovative analyses.
Much of the data collected by wind-farm owners and operators is kept private and secure under the notion of competition. This is unfortunate, as those who have the data may not recognise their opportunity to benefit when combined with other wind farm data and the resulting collaboration. If an owner has, say 1,000 turbines and is impressed by the results of their own fleet analytics, imagine the benefits of combining data from 10,000 or 20,000 units.
The solution to the fear of embarrassing or even damaging data is to create a common database defining standards for what data is to be collected that focuses only on the use of performance analytics. The largest risk of data sharing is choosing not to share. The owner will impede the success of the fleet if data is kept private. We are all on the learning curve together, and we can all help each other with the knowledge that is contained within this data.
The sensible way to combine data from all wind turbines is to create a standardisation that controls what data is collected, together with the time sample, the accuracy of the data and the format. This standard could easily accommodate datasets that are batch collected and may only contain 15-minute samples, and real-time data that is much more granular.
The issue is this: who can create such a database for the use of all those involved and provide the data in a secure, controlled environment for participating members? Hopefully, there will be a solution soon so all owners and operators can benefit from each other's data and analytics.
Judah Moseson is director of operations and maintenance services at Infigen Energy US