Many companies operate large fleets of complex industrial assets which are similar in kind but different in terms of their concrete specifications. The appplication of traditional AI methods for performance monitoring of such fleets leaves a huge potential untapped, as these methods do not exploit important intrinsic characteristics associated to the fleet aspects, i.e. the immense richness of the data originating from the heterogeneity inherent to the fleet aspect, i.e. the immense richness of the data originating from the heterogeneity inherent to a fleet and the hierarchical composition present between the various components in an asset and the varioous assets in a fleet. The presentation below focuses on the opportunities and challenges related to the usage of intelligent data exploitation and modelling techniques to support and facilitate exploitation of fleet-based data for maximising the fleet performance. In particular, individual wind turbines operate in a wind farm and several farms together form a fleet that is monitored by an O&M contractor. Rather than analysing data from a single turbine, it is an opportunity to consider pooling data from all turbines across farms and fleets. In this way, the richness of such fleet data is maximally exploited, i.e., a wider range of operational contexts (e.g. meteorological conditions, operating modes, maintenance history, etc.) is availaible. This allows to derive statistically more significant and more representative models of the prototypical behaviour of a wind turbine in a particular context and operative mode based on usage data from the field rather than manufacturer-provided theoretical specifications. Finally, performance degradation and remaining useful lifetime estimation for individual (components of) turbines can be identified and quatified by comparing them with these prototypical behavioral models.
Interested in this topic? Have a look at the presentation below we gave on 18/10 at the Wind Farm Data Management and Analysis Forum.