Data Importance in Power System Asset Management

Sammanfattning: The current shift towards a higher degree of data-driven decision making in power system asset management highlights the importance of asset data. This thesis identifies, investigates, and proposes methods for data-related research gaps that are encountered by asset managers. These research gaps are in data availability and data quality. It is challenging to generalize the data availability problem on an abstract level. Thus, data availability is studied through three different case studies. Each case study addresses a factor that contributes to data availability problems. Data censoring is modeled as a data quality problem using a Monte-Carlo simulation. Lack of access to and acquisition of data are studied through event tree analysis and multiphysics modelling. These case studies reveal that even in a low data availability environment, informed decision making is feasible. Monte-Carlo simulation techniques are powerful when analyzing the data quality problem. Asset data quality is studied based on two perspectives; namely, maintenance optimization and reliability evaluation. First, using random population studies shows that data quality can have a notable financial and technical impact on maintenance optimization. A critical finding is that missing data can lead to distortions in estimates of the optimal replacement time of a component. It is shown that there exists a certain threshold of missing data proportion beyond which maintenance optimization becomes unreliable. The specific percentage value of this threshold depends on the failure model parameters. Second, incorporating the data quality model in a reliability test system simulation shows that the impact on the annual estimation of system- and energy-oriented reliability indices is nearly non-existent.Finally, this thesis introduces a method to rank component types based on data quality importance. The data quality importance (DQI) ranking is derived from the Weibull function’s sensitivity to data errors. This method indicates that distortions in Weibull parameters have a non-linear impact on maintenance optimization. This leads to a conclusion that investments in data quality must be allocated based on the DQI ranking of a certain component. Reaching the right level of data quality for a component leads to efficient decision making.