| Work Detail |
Researchers have used the chimpanzee optimization algorithm to optimize the hyperparameters of five machine learning models for predicting photovoltaic energy yield. This algorithm is based on the cooperative hunting behavior of chimpanzees in the wild, mimicking the way they work together to hunt their prey.
A scientific group led by researchers from the German Jordanian University has analyzed the effect of the so-called Chimpanzee Optimization Algorithm (ChOA) on different machine learning (ML) models for predicting photovoltaic production.
ChOA is based on the cooperative hunting behaviour of chimpanzees in the wild, mimicking the way they work together to hunt prey, common among small mammals. They typically operate in groups of three or four hunters and initially drive and block prey, then chase and attack it.
The algorithm explores different combinations of parameters to achieve the most promising result. The scientists used it to optimize the hyperparameters of five types of ML models. These include multiple linear regression (MLR), decision tree regression (DTR), random forest regression (RFR), support vector regression (SVR), and multilayer perceptron (MLP).
"The effectiveness of this contribution is verified against data from a real-world case study, while drawing on various performance metrics from the literature, including root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2)," the researchers explain.
Hyperparameters are external settings set before the learning process begins, which govern the learning process and do not change during training. Hyperparameters – such as the learning rate in neural networks – influence the dynamics of training and can therefore significantly affect the effectiveness of models.
The five models, with and without ChOA, were trained on 948 records and tested on 362 records. The records were taken between 2015 and 2018 from a 264 kW photovoltaic system installed on a rooftop of the University of Applied Sciences in Amman, the capital of Jordan. The tilt angle of the installation was set at 11 degrees and the azimuth angle at -36 degrees. Meteorological variables such as wind speed, relative humidity, ambient temperature and solar irradiation were measured at a nearby weather station.
“Amman (Jordan) experiences a Mediterranean climate characterized by hot, dry summers and cool, wet winters,” the researchers add. “The average temperature throughout the year is 17.63 °C, and the average annual global horizontal irradiation is 2040.2 kWh/m2.”
Through this analysis, the scientists found that all models experienced performance improvements as a result of fine-tuning hyperparameters using ChOA.
“The DTR model showed substantial improvements, with the test RMSE decreasing to 1.972 and the R2 increasing to 0.951,” they explained. “The RFR model showed notable improvements, with RMSE values ??decreasing to 1.773 for training and 1.837 for testing, and R2 values ??increasing to 0.964 for training and 0.963 for testing. The SVR model saw the most notable improvement, with the test RMSE decreasing to 0.818 and the R2 increasing to 0.977.”
Following optimization of ChOA, MLP was found to show the best results in predicting PV power output. Specifically, it was able to achieve 0.503, 0.397, and 0.99 in RMSE, MAE, and R2, respectively. “ChOA effectively tuned the parameters, which improved the model fit, reduced overfitting, and enhanced generalization compared to two other widely used optimization algorithms in the literature: particle swarm optimization (PSO) and genetic algorithm (GA),” the team concluded.
The results were presented in “ Enhancing solar photovoltaic energy production prediction using diverse machine learning models tuned with the chimp optimization algorithm ,” published in Scientific Reports . The group consisted of scholars from the German-Jordanian University of Jordan, the University of Jordan, Al-Balqa Applied University, and Tuskegee University in Alabama. |
