Turbine Analytics

Our analysis increases your ongoing revenue

We use our unique Revenue Leakage Dashboard to target our analytics on areas that could potentially increase revenue. This has previously resulted in:

  • Identification and correction of inaccurate pitch control algorithms, resulting in third-party verified performance improvements of 2-3%
  • Identification and correction of yaw misalignment, increasing production by 4.4%.
  • A reduction in applied dispatch down by 4.6% over a 6 month period, through the identification and removal of inaccurate frequency control curtailment across the entire windfarm, which had been previously been incorrectly implemented by the turbine manufacturer.

Our background in windfarm ownership keeps our analysis focussed on areas where revenue could be increased, meaning that both the short-term and long-term value of your assets is optimised.

We reduce your turbine maintenance expenditure

We reduce your annual turbine maintenance spend by ensuring that contract availability is correctly calculated. This typically results in a reduction in the claimed contractor availability by 1.5%, meaning either a reduction in the upside bonus payable to the maintenance contractor, or facilitating a contractual claim for increased availability compensation.

We do this by identifying common inaccuracies in how the manufacturer SCADA interprets the contractual availability. With over 20 years’ experience, we understand the vagaries of all of the major SCADA systems. We know where each SCADA system typically inaccurately indicates turbine uptime, loss of communications or grid faults, thereby inaccurately outputting higher availability values.

We reduce downtime from major component failures

We reduce turbine downtime and increase turbine revenue by using machine learning to predict faults before the SCADA system identifies them. Early intervention then results in reduced downtime and a planned replacement rather than an extended outage period while replacement components are being source.

Our machine learning algorithms have successfully identified potential component failures 6 months prior to the actual failure. They do this by using live data together with a learned prediction model to forecast specific parameter values based on prior history and live data, and then compare them to the actual parameters.