Companies worldwide, and certainly in renewables, pride themselves on becoming more data-driven. With the rise in data and our ability to perform large scale analysis, it makes sense that data informs decision-making. Managers and directors know that having data to back up a business decision is essential. Conversely, if a decision backfires, it’s valuable to refer to the data to better understand what went wrong.
But even with high quality input data, a poorly chosen key performance indicator (KPI) can lead to poor outcomes both in the short and long term. In the pursuit of a simple solution for quantifying performance, wind fleet owners and operators should exercise caution in the definition of KPIs and should recognise the pitfalls from the outset.
For many years, availability has been the dominant KPI for wind farms. Simple to understand and relatively easy to calculate. The industry has wisely shifted from primarily time-based to production-based availability, as poorly planned downtime during strong windspeeds will significantly impact the total energy produced. However, even production-based availability may present some difficulties.
For example, if a service contract states a specific target for production-based availability, the service provider will do their best to hit exactly this target – nothing less and probably nothing more. The main incentive is to avoid the downside of contractual penalties.
Of course, the turbine owner will hope to reach the highest availability possible to maximise revenues. But achieving very high availability comes with a disproportionately high increase in operational costs. Therefore, the profit margin of the service providers will shrink if efforts are made to go “above and beyond”.
Timing and Communication
Certain faults trigger an automatic stop in turbines for reasons of safety. To keep turbines running, an operator may need to perform frequent remote restarts. Control centre staff need incentivisation to react quickly but may lack adequate engineering training to differentiate between a minor incident and a chronic fault for which an inspection should be scheduled.
Significant delays often exist between the time when action is taken by the operator, and any subsequent detailed analysis; not to mention the question of who is responsible for performing this analysis? Does the operator have relevant specialists on staff? Does the owner think a full-time data scientist is a worthwhile investment? Does the OEM notice the issue and do they react to it? If so, when are their results communicated to the site or the owner?
The wind turbine’s power performance is another KPI often expressed as a percentage, with 100% equal to the optimal performance. Owners will want as close to 100% as possible to ensure maximised energy yield. But accurate analysis of individual power curves for a large fleet of turbines is time-consuming and the advanced tools required are often costly. Such analysis must be performed with care due to the challenges of dealing with imprecise operational data, windspeeds derived from turbine anemometers, complex environmental effects and a growing number of complex control strategies due to curtailment regulations. Owners may pressure an operator or OEM to perform such detailed analyses; however, in case issues are detected it may be difficult to agree on methodology and results, weeks or months may pass before necessary corrective actions can be carried out.
Annual Energy Production (AEP)
Let’s assume it has been a good year, and AEP targets for a fleet of wind turbines have been reached already midway through Q4. The asset managers will be happy that this will be a profitable year. However, upon closer inspection, we see that typically the main thing this KPI tells us is that wind speeds were above average this year.
Other performance-related factors may be poorly represented by this KPI: power curves may in fact be trending downwards, maintenance actions may have been carried out more slowly, the operator may have frequently performed remote restarts to keep availability high, hiding serial faults which remain unresolved in the drivetrain, and threatening costly failures and downtime in the future. Next year, wind speeds may be lower, and things may not look so positive.
On-site OEM maintenance technicians may not have the correct information and incentivisation to achieve maximum production. For example, turbines may be stopped for maintenance and inspections when electricity prices are high. On some sites, certain turbines may be checked more frequently than others, if they are easier to access. The challenge of performing maintenance during harsh winter conditions may result in delays and lost efficiency.
How to shape smarter KPIs
First, when management sets new KPIs, they should discuss these carefully with their experienced employees in all related areas of the business, Including operations, data analytics, asset management and service.
Second, potentially perverse incentives should be recognised in advance. Avoid KPI’s that inadvertently encourage negative behaviour in some areas of the organisation. Incentives should create an atmosphere where people are encouraged to avoid shortcuts.
Third, understand that if you succeed with one KPI, you still may have a problem with another. Multiple KPIs are often interdependent, and short-term gain may lead to long-term pain.
KPIs are appealing because they simplify our complex world, allowing us to measure the outcome of our complex decisions and actions. Like a flashlight in the dark, they illuminate. Yet taken individually they often do not paint the full picture and planned badly they may incorrectly motivate. Keep this in mind and refine your strategy regularly for best results.