Turbine monitoring and maintenance
The present invention related to non-thermal renewable energy turbines ( 20,24,34,38,40 ), in particular to the monitoring of turbine performance to identify a loss of performance indicative of faults or component degradation. The method involves comparison of measured power from a target turbine ( 20 ) with a predicted value for same turbine. The predicted value is calculated using the output from a plurality of other turbines ( 24,34,38,40 ) from an array and a predictive model including weightings for the other turbines ( 24,34,38,40 ) based on the strength of correlation of their historical with historical data from the target turbine ( 20 ).
1. A method for turbine fault or component degradation detection comprising:
measuring the power output from a plurality of renewable energy turbines in an array,
performing a comparison of the measured power output from a target turbine with a predicted value for the target turbine, wherein the predicted value is the result of a calculation based on the measured power output from a plurality of other turbines from the array and a predictive model which includes a weighting assigned to each of the plurality of other turbines based on the strength of correlation of their historical data with historical data from the target turbine,
determining a performance change for the target turbine based on the comparison;
outputting an alert signal indicative of a fault or component degradation if a performance change is determined; and
causing repair, maintenance or inspection of the target turbine to be performed based on the alert signal;
wherein a separate predictive model is provided for each of a plurality of defined wind directions, and wherein the method selects the predictive model that corresponds to the wind direction experienced by the target turbine.
2. The method according to claim 1 , wherein the plurality of wind directions are defined as a predefined number of equal segments of a circle.
3. The method according to claim 1 , wherein the calculation is performed using a Deep Neural Network.
4. The method according to claim 1 , wherein the calculation is performed using simple multivariate regression, random sample consensus, or regression trees.
5. The method according to claim 1 , wherein the plurality of other turbines includes all other turbines in the array.
6. The method according to claim 1 , wherein the plurality of other turbines comprises only a subset of other turbines in the array.
7. The method according to claim 1 , wherein the calculation is independent of wind speed and/or devoid of wind speed data as an input.
8. The method according to claim 1 , further comprising an initial step of training the predictive model using historical data for the target turbine and for the plurality of other turbines.
9. A method for turbine fault or component degradation detection comprising:
measuring the power output from a plurality of renewable energy turbines in an array,
performing a comparison of the measured power output from a target turbine with a predicted value for the target turbine, wherein the predicted value is the result of a calculation based on the measured power output from a plurality of other turbines from the array and a predictive model which includes a weighting assigned to each of the plurality of other turbines based on the strength of correlation of their historical data with historical data from the target turbine,
determining a performance change for the target turbine based on the comparison;
outputting an alert signal indicative of a fault or component degradation if a performance change is determined; and
causing repair, maintenance or inspection of the target turbine to be performed based on the alert signal;
wherein the plurality of wind directions are defined as a predefined number of equal segments of a circle.
10. The method according to claim 9 , wherein a separate predictive model is provided for each of a plurality of defined wind directions, and wherein the method selects the predictive model that corresponds to the wind direction experienced by the target turbine.
11. The method according to claim 10 , wherein the plurality of wind directions are defined as a predefined number of equal segments of a circle.
12. The method according to claim 9 , wherein the calculation is performed using a Deep Neural Network.
13. The method according to claim 9 , wherein the calculation is performed using simple multivariate regression, random sample consensus, or regression trees.
14. The method according to claim 9 , wherein the plurality of other turbines includes all other turbines in the array.
15. The method according to claim 9 , wherein the plurality of other turbines comprises only a subset of other turbines in the array.
16. The method according to claim 9 , wherein the calculation is independent of wind speed and/or devoid of wind speed data as an input.
17. The method according to claim 9 , further comprising an initial step of training the predictive model using historical data for the target turbine and for the plurality of other turbines.