A Bayesian Approach to Probabilistic Solar Irradiance Forecasting

TL;DR

This paper presents a Bayesian probabilistic approach to forecast solar irradiance, improving grid operation by providing more accurate and uncertainty-aware predictions using publicly available weather data.

Contribution

It introduces a Bayesian method for probabilistic solar irradiance forecasting and demonstrates its effectiveness with real-world data and comparisons.

Findings

Bayesian approach improves forecast accuracy

Method outperforms traditional point forecasts

Provides reliable uncertainty estimates

Abstract

The output of solar power generation is significantly dependent on the available solar radiation. Thus, with the proliferation of PV generation in the modern power grid, forecasting of solar irradiance is vital for proper operation of the grid. To achieve an improved accuracy in prediction performance, this paper discusses a Bayesian treatment of probabilistic forecasting. The approach is demonstrated using publicly available data obtained from the Florida Automated Weather Network (FAWN). The algorithm is developed in Python and the results are compared with point forecasts, other probabilistic methods and actual field results obtained for the period.