Forecasting of Dynamic Thermal Line Rating Under the Conditions of Temporal Discretization and Correlation

  • Author / Creator
    Barton, Tomas
  • Dynamic Thermal Line Rating (DTLR) is a technology that optimizes the utility of overhead power transmission lines by dynamically adjusting the rating according to current ambient conditions. To be truly useful, forecasting of DTLR must be applied within processes governing the function of the electrical system. This thesis focuses on medium term DTLR forecasts on the timescale of days and hours. Two DTLR forecasting systems are developed within this thesis. Both systems are categorized as indirect probabilistic rating systems, where Numerical Weather Predictions (NWP) are processed to estimate the rating. The systems differ in the method used to quantify uncertainty. In the first system, the uncertainty is described by a custom statistical model that is fitted onto the historical forecasts of the model in a scheme called Model Output Statistic. In the other system a Random Forest machine learning model is employed to produce probabilistic output associated with the NWP output. This system uses the regression-via-classification approach, where most processing occurs in the discrete domain even though the inputs and outputs are continuous. Both developed systems provide DTLR predictions that perform better than Static Line Rating and a reference method. The DTLR forecasting systems in this thesis have been specifically designed to produce forecasts of temporally discretized rating. Temporal discretization is a term defined in this thesis as the process of taking the continuous DTLR and turning it into a single value valid over a period. Temporal discretization decreases DTLR benefit and this relationship has been studied and simulated across a vast dataset spanning all of Canada. A figure with the relative benefit of DTLR for different lengths of discretization periods is provided. The effects of temporal discretization on DTLR forecasts are also addressed. A simulation of DTLR forecasts with different discretization period lengths was performed to analyze the sensitivity of DTLR forecasts to this variable. It was revealed that the overall benefit of DTLR forecasts is relatively insensitive to the discretization length. It was argued that temporal correlations within DTLR time series impact the discretized value. It was argued that ignoring these correlations will negatively affect the probabilistic prediction by skewing the distribution toward the extreme value. This was demonstrated on the simulated forecast. A Monte Carlo system with a novel sampling method was proposed to mitigate this problem by generating realistic time series that are used to estimate the rating. This method yields well calibrated results.

  • Subjects / Keywords
  • Graduation date
    Spring 2021
  • Type of Item
  • Degree
    Doctor of Philosophy
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.