The Bachelor's Thesis Wind Farm Power Forecasting Using Deep Neural Networks project aimed to develop models for forecasting the power output of wind farms using advanced artificial intelligence techniques. Deep neural networks were employed, including convolutional neural networks (CNN), recurrent neural networks (RNN) such as Long Short-Term Memory (LSTM), and Gated Recurrent Unit (GRU).
- Data Preparation: Data pertaining to wind farms was collected and processed appropriately, taking into account seasonal and temporal variations in energy production.
- Model Building: Models based on convolutional, LSTM, RNN, and GRU networks were implemented to predict the power generated by wind farms based on historical data.
- Training and Evaluation of Models: The models were trained on the training dataset and then evaluated for their effectiveness on the validation and test datasets. Model parameters were optimized to achieve the best forecasting results.
- The employed deep learning techniques enabled effective forecasting of wind farm power.
- Models based on LSTM and GRU yielded particularly good results, indicating the effectiveness of models based on recurrent structures.
- The results of experiments were thoroughly analyzed and documented, allowing for insights into the effectiveness of different forecasting methods.
Here's a plot comparing the R2 scores of all utilized models:
The project confirmed that deep neural networks, including models based on LSTM, RNN, and GRU, can be effective tools for forecasting wind farm power. Further research efforts can focus on refining these models and their implementation in industrial practice.