| Sumario: | The ENIAN start-up has a renewable energy project database with key metrics collected from open sources. ENIAN’s aim is to use this data to feed a data-driven mechanism (REPSCORE) to assess renewable energy projects. It was reviewed the ENIAN’s database, focusing on the solar PV technology. The dataset was analysed and manually cleaned to fit a linear regression model to predict the unitary cost (specific capital cost) of a project based on three predictors: country, completion year and stage of development. The unitary cost was used to estimate the LCOE (levelized cost of electricity) of a project, to feed REPSCORE. The study was narrowed to South American projects, and the unitary cost and LCOE estimation were focused on the country of Chile. The results show that commercial information is often missing (over 30%), and almost every unusual value needs deep revision before fitting the model. Automatic cleaning of atypical values was not successful. The predictive performance of the model shows a prediction interval of ± 40% or greater predicting the unitary costs beyond 2017. The LCOE prediction is consistent with this uncertainty. Further research can be done to improve the model: add more predictors, consider the data quality in the model, and use imputation methods to recover missing data. Finally, adding more metrics to the ENIAN’s dataset should improve their usefulness.
|
|---|
