Short-term forecasting of global solar irradiance with incomplete data

TL;DR

This paper presents a one-day ahead solar irradiance forecasting pipeline that handles incomplete data using data imputation and compares multiple data-driven models, showing neural networks outperform ARIMA especially in cloudy conditions.

Contribution

Introduces a novel forecasting pipeline for solar irradiance that manages missing data and compares multiple models, highlighting neural networks' superior performance.

Findings

Neural network models outperform ARIMA in most cases.

LSTM performs best in cloudy, unpredictable weather.

The approach effectively handles incomplete historical data.

Abstract

Accurate mechanisms for forecasting solar irradiance and insolation provide important information for the planning of renewable energy and agriculture projects as well as for environmental and socio-economical studies. This research introduces a pipeline for the one-day ahead forecasting of solar irradiance and insolation that only requires solar irradiance historical data for training. Furthermore, our approach is able to deal with missing data since it includes a data imputation state. In the prediction stage, we consider four data-driven approaches: Autoregressive Integrated Moving Average (ARIMA), Single Layer Feed Forward Network (SL-FNN), Multiple Layer Feed Forward Network (FL-FNN), and Long Short-Term Memory (LSTM). The experiments are performed in a real-world dataset collected with 12 Automatic Weather Stations (AWS) located in the Nari\~no - Colombia. The results show that the neural network-based models outperform ARIMA in most cases. Furthermore, LSTM exhibits better performance in cloudy environments (where more randomness is expected).