A Data-centric Approach for a Day-ahead System Non-Synchronous Penetration Forecast

This paper presents a novel data-centric approach to predict the day-ahead system non-synchronous penetration (SNSP) ratio in the Irish power system. The proposed method uses a deep learning algorithm, based on feedforward artificial neural networks (ANN). The methodology involves a statistical data analysis to determine the temporal features of the target data. It also uses a Pearson correlation analysis to determine the more relevant model inputs. The input data is then normalised using a Min-Max technique to ensure accurate model implementation. Once the model architecture is defined, a random search optimization algorithm is applied using a Keras tuner to select the optimal hyperparameters. The performance of the model is evaluated using three statistical metrics. The results indicate a superior performance of the proposed forecasting model compared to the persistent approach, demonstrating solid and robust predictive capabilities.