Application of k-fold cross validation in photovoltaic power generation forecasting using LSTM Networks

Accurate forecasting of photovoltaic (PV) power generation is crucial for the integration of PV systems into the power system. Long Short-Term Memory (LSTM) networks have emerged as effective tools for time-series data forecasting. This study investigates the application of k-fold cross validation in enhancing the accuracy of LSTM networks for PV power forecasting. By dividing the dataset into k parts and using each part for validation while training on the others, the model's performance across different data segments is evaluated. The empirical results are based on field-collected PV power generation data. It can be inferred that improved machine-learning approaches can enhance the accuracy of PV power forecasting.