Evaluating the Efficacy of the Application of Artificial Neural Network Techniques to Enhance the Modified Flat Plate Solar Water Heating System Performance

Main Article Content

Benjamin Asubam Weyori, Gyimah Kopri, Emmanuel Harris, Opoku Gyabaah, Bright Karim-Abdallah

Abstract

To augment the performance of solar thermal collectors, experimental, analytical, and computational techniques are implemented, which require a lot of time to arrive at an accurate result. On the other hand, theoretical-based results are suitable only for simplified models of practical devices under many simplifying assumptions. Due to their simplicity, high speed, and capability to solve complex and nonlinear relationships among the variables and extracted data, machine learning techniques are becoming the best approach for many thermal system modelling and optimisations. This study uses a real-time dataset to compare several machine learning model’s performance in predicting solar water heaters' thermal performance. The experiment was based on Ghanaian data to test the performance prediction of the Chromagen solar water heater. With the help of a microcontroller-based embedded system, four variables, ambient temperature, inlet water temperature, outlet water temperature, and solar radiation, were collected for simulation in the Python environment using the Google Collaboratory cloud system. Treating the outlet temperature as the target, an EDA was performed with 0.8 c, one of the features and the least 0.4. After training ten (10) Machine learning models, the KNeighbors regression model performed best with MSE = 0.39, R2= 99.33 %, MAE = 0.28, and RMSE = 0.62. With a limited dataset of about 3000, the K-Neighbors is recommended to be the best machine-learning model for such an experiment. Future works in this research are targeted to include two or more datasets from different solar water heaters (with more than 5000 data points) for more generalization and finally deploy the best model on the Web for public usage.

Article Details

Section
Articles