Forecasting of wind farm power output based on dynamic loading of power transformer at the substation

Dynamic Transformer Rating (DTR) allows unlocking extra capacity of power transformers using real-time weather data and has been proven to be specifically attractive for application to wind farm substation transformers. In this work, we explore an extreme case where the wind farm expanded to 150% of its original rated power while being connected to the grid with the same transformer to simulate a 1:1.5 ratio between rated generation and rated transformer capacity. The focus of the study is to explore the operational challenges of using such a system in a framework of day -ahead dispatch planning, which is done by building a combined forecasting model for 36 -hour ahead prediction of wind farm generation and the transformer capacity as well as their match together. The goal is to estimate how often the wind farm generation would exceed the available capacity at the substation and would be required to curtail as well as assign accuracy to the curtailment decision. The results indicate that the model shows sufficient prediction accuracy for exceeding the maximum allowable transformer temperature. Main indication of the model accuracy is the ability to correctly predict instances of transformer overheating, which in this case are below 3.5%. However, since the accuracy of correctly ordered curtailment is at 85% for the lower transformer hot spot temperature limit, future studies should focus on improving current results by possibly integrating other time series forecasting models.