Prediction of PV Maximum Power Output Using Regression and Tree-Based Machine Learning Techniques

Main Article Content

Bhavik Brahmbhatt
Sameer Mansuri
Priyank Patel
Vilashkumar N Chaudhari
Dipak Sahay
Dhaval N Patel
Rajesh T Patel
Manish Chauhan

Abstract

This paper presents a comprehensive evaluation of seven machine learning (ML) models—Linear Regression (LR), Ridge Regression (RR), Lasso Regression (Lasso), Bayesian Regression (BR), Gradient Boosting Regression (GBR), Artificial Neural Network (ANN), and Decision Tree Regression (DTR)—for predicting the Maximum Power Point (MPP) in standalone photovoltaic (PV) systems. Using a MATLAB-simulated dataset of 876,012 samples with irradiance and temperature as inputs, the models predict maximum current (Im), voltage (Vm), and power (Pm). Performance is assessed using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R² metrics. DTR achieves superior results, with R² 0.99, RMSE 2.36 for Pm, and minimal errors, owing to its ability to model nonlinear relationships. The study proposes a DTR-based framework for real-time MPPT, deployed on an STM32F407 microcontroller. Future work includes hybridizing DTR with Particle Swarm Optimization (PSO) to optimize hyperparameters and enhance system responsiveness under dynamic conditions.

Article Details

Section
Articles