Routine maintenance is an essential requirement for the optimal functioning and longevity of any technical system that has been constructed. The issue occurs when the maintenance planning for such a structure becomes necessary. The adverse weather conditions prevalent in the Netherlands contribute to the heightened danger associated with the duty of a maintenance worker. Additionally, it is vital to comprehend the optimal time frame for minimising revenue losses when allocating time towards maintenance activities rather than operational tasks.

The objective of this project is to create a methodology for enhancing and improving the management of asset maintenance planning. This process is carried out by developing two statistical models. The primary objective of this study is to examine the feasibility of utilising API data from wind forecast sites to provide accurate production forecasts for individual wind turbines up to a 7-10 day period in advance. Furthermore, this study aims to forecast the electricity prices for the upcoming week by analysing historical data and recent day-ahead pricing. The outputs generated by these models are subsequently aggregated to yield a single outcome in terms of revenue.

The predictive model for electricity prices utilises data sourced from ENTSOE to generate an aggregate of electricity prices spanning the previous 7-8 years. This aggregate is subsequently adjusted by incorporating the electricity prices observed within the most recent three-week period. The model exhibits high accuracy in predicting day-ahead pricing during weekdays, but its performance is not consistently replicated on weekends.

The wind turbine output forecast model utilises operational archival data and high-resolution 10-day forecasts obtained from the European Centre for Medium-Range Weather Forecasts (ECMWF). A correlation has been established between the archival data and the historical turbine data provided by Green Trust Consultancy for the designated wind farm. The aforementioned correlation is subsequently employed to establish a connection between the output of a turbine and the real-time forecast data. The accuracy of the power generation forecast decreases from the initial day to the tenth day of the projection, resulting in inconsistent outcomes.

The combined outputs of these two models yield a solitary outcome that aids in predicting the potential revenue loss for the selected turbine during the period of maintenance-induced idleness. The limitations inherent in both models contribute to the generation of imprecise outcomes inside the revenue model pertaining to wind turbines. The model accurately predicts outcomes in two out of the four tested scenarios.