Analytics for Asset Management

Key Benefits

Current methods of computing asset risk, as embodied in the National Grid Electricity Transmission (NGET) Network Asset Risk Annex for RIIO-T2, have several shortcomings. These include: (1) Network risk is incorrectly characterized as the sum of individual asset risks. (NARA-T2, Page 15) (2) Asset P(F) values are computed in a manner that may result in anomalous results, such as P(F) values greater than 1. (NARA-T2, Page 16-30) (3) Asset models are based upon a variation of the Weibull distribution, which has shortcomings such as fixed domain in time, unimodality, and non-validation against empirical data. (NARA-T2, Page 25) By implementing these methods, NGET is currently exposed to the following negative consequences: (1) Inaccurate or insufficient understanding of how assets actually operate and fail. (2) Misapplication and inefficient use of limited capital and O&M resources to optimally replace or maintain assets. (3) Exposure to excessive asset failures because asset models may indicate failure later in time than what occurs in reality. (4) Wasted O&M resources due to excessive maintenance of assets that will fail less frequently than what current asset models indicate. To address these shortcomings, ATS proposes to apply advanced data analytics that eliminate the need for legacy asset models such as the Weibull distribution, or broad P(F) approximations that are not validated. Instead, ATS will use the full corpus of data, including conditions data, inspections data, operations data, failure data, loading data, atmospherics data, among other data sources, to build a comprehensive analytical model to more fully and accurately compute P(F) values for assets. Some analytical methods include statistical analysis, training and application of machine learning models and neural networks, and incorporation of quantitatively encoded subject matter expert judgment. This will result in the following benefits: (1) A significantly more complete and accurate understanding of how assets operate and fail, based upon the direct empirical data. (2) Reduction in failure costs by remediating failure risk from better asset models. (3) More efficient allocation of both capital and O&M resources by applying them to the right assets at the right time resulting in an overarching reduction of risk costs. (4) Improved system-level resilience, resulting in higher reliability, mean time between failure, and mean time to repair. This proposed work is an innovation because it represents a significant change from the status quo method of characterizing asset risk, as embodied in the NARA-T2.