[Paper Review] Computational Solar Energy -- Ensemble Learning Methods for Prediction of Solar Power Generation based on Meteorological Parameters in Eastern India

Debojyoti Chakraborty, Jayeeta Mondal+2 more|arXiv (Cornell University)|Jan 21, 2023

Energy Load and Power Forecasting48 references4 citations

TL;DR

This study proposes an ensemble machine learning framework to predict solar photovoltaic (PV) power generation in Eastern India using meteorological data. By applying Bagging, Boosting, Stacking, and Voting models on a real-world 10kWp solar PV dataset, the authors achieve up to 96% prediction accuracy, with Stacking and Voting models demonstrating superior performance for large-scale solar energy forecasting.

ABSTRACT

The challenges in applications of solar energy lies in its intermittency and dependency on meteorological parameters such as; solar radiation, ambient temperature, rainfall, wind-speed etc., and many other physical parameters like dust accumulation etc. Hence, it is important to estimate the amount of solar photovoltaic (PV) power generation for a specific geographical location. Machine learning (ML) models have gained importance and are widely used for prediction of solar power plant performance. In this paper, the impact of weather parameters on solar PV power generation is estimated by several Ensemble ML (EML) models like Bagging, Boosting, Stacking, and Voting for the first time. The performance of chosen ML algorithms is validated by field dataset of a 10kWp solar PV power plant in Eastern India region. Furthermore, a complete test-bed framework has been designed for data mining as well as to select appropriate learning models. It also supports feature selection and reduction for dataset to reduce space and time complexity of the learning models. The results demonstrate greater prediction accuracy of around 96% for Stacking and Voting EML models. The proposed work is a generalized one and can be very useful for predicting the performance of large-scale solar PV power plants also.

Motivation & Objective

• To address the intermittency and variability of solar energy by improving prediction accuracy of photovoltaic power generation.
• To evaluate the effectiveness of multiple ensemble machine learning techniques in forecasting solar power based on meteorological parameters.
• To develop a test-bed framework for data preprocessing, feature selection, and model evaluation to optimize computational efficiency.
• To provide a generalized, scalable solution applicable to large-scale solar PV power plants in similar climatic regions.
• To validate model performance using real field data from a 10kWp solar PV plant in Eastern India.

Proposed method

• The study employs four ensemble learning methods: Bagging, Boosting, Stacking, and Voting to improve prediction robustness.
• A comprehensive test-bed framework is designed to support data mining, feature selection, and dimensionality reduction to reduce model complexity.
• Meteorological parameters such as solar radiation, ambient temperature, wind speed, and rainfall are used as input features.
• The dataset is derived from a real 10kWp solar PV power plant in Eastern India, ensuring real-world relevance.
• Model performance is evaluated using standard regression metrics, with accuracy reported as the primary evaluation criterion.
• Feature selection techniques are applied to reduce space and time complexity while preserving predictive power.

Experimental results

Research questions

• RQ1 How do different ensemble learning models compare in predicting solar PV power generation using meteorological data?
• RQ2 What is the impact of feature selection and dimensionality reduction on model accuracy and computational efficiency?
• RQ3 Can ensemble methods achieve high-precision solar power forecasts in a region with variable monsoon and weather patterns like Eastern India?
• RQ4 Which ensemble model—Stacking or Voting—delivers the highest prediction accuracy for solar PV output?
• RQ5 To what extent can the proposed framework be generalized for large-scale solar PV power plant performance prediction?

Key findings

• The Stacking and Voting ensemble models achieved the highest prediction accuracy of approximately 96%.
• The proposed test-bed framework successfully reduced both space and time complexity through effective feature selection and reduction.
• Meteorological parameters such as solar radiation, ambient temperature, wind speed, and rainfall significantly influence solar PV power output.
• Ensemble learning methods outperformed individual models in terms of robustness and prediction accuracy.
• The framework is scalable and generalizable, making it suitable for deployment in large-scale solar PV power plants.
• The results demonstrate that ensemble models are highly effective for solar power forecasting in regions with complex and variable weather conditions.