A Comparison of the Performance of Supervised Learning Algorithms for Solar Power Prediction

Leidy Gutiérrez,Julian Patiño,Eduardo Duque-Grisales

doi:10.3390/en14154424

Abstract

Science seeks strategies to mitigate global warming and reduce the negative impacts of the long-term use of fossil fuels for power generation. In this sense, implementing and promoting renewable energy in different ways becomes one of the most effective solutions. The inaccuracy in the prediction of power generation from photovoltaic (PV) systems is a significant concern for the planning and operational stages of interconnected electric networks and the promotion of large-scale PV installations. This study proposes the use of Machine Learning techniques to model the photovoltaic power production for a system in Medellín, Colombia. Four forecasting models were generated from techniques compatible with Machine Learning and Artificial Intelligence methods: K-Nearest Neighbors (KNN), Linear Regression (LR), Artificial Neural Networks (ANN) and Support Vector Machines (SVM). The results obtained indicate that the four methods produced adequate estimations of photovoltaic energy generation. However, the best estimate according to RMSE and MAE is the ANN forecasting model. The proposed Machine Learning-based models were demonstrated to be practical and effective solutions to forecast PV power generation in Medellin.

Highlights

The increase in the world’s energy demand is evident, creating a threat of a global energy crisis, which causes adverse environmental effects on our habitat [1]
The primary energy generation methods used in the world come from fossil fuels, reaching an annual rate of consumption in oil, gas and carbon of 3.1 million tons (Mt) [3], representing more than 80% of world consumption [4]
This paper proposes the modeling of PV power production by computational methods based on historical data from a generation system located in Medellín, Colombia

Summary

Introduction

The increase in the world’s energy demand is evident, creating a threat of a global energy crisis, which causes adverse environmental effects on our habitat [1]. The primary energy generation methods used in the world come from fossil fuels, reaching an annual rate of consumption in oil, gas and carbon of 3.1 million tons (Mt) [3], representing more than 80% of world consumption [4]. These sources represent a higher demand due to their low cost, but they negatively affect the environment, considering that they increase carbon dioxide (CO2 ) and greenhouse gas emissions [5], contributing to global warming.

Objectives

Methods

Results

Conclusion