Dust-Induced Power Loss in Solar Plants: A Real-Time Web Application for Degradation Prediction Using Decision Trees and Random Forests Regressions

Solar power plants are an important alternative for implementing sustainable energy. However, their efficiency is often compromised by the presence of dust on solar panels. This research attempts to develop a real-time web system that can monitor and predict power loss in solar panels using environmental data and dust concentration metrics. Two regression models are used, namely Decision Tree Regression (DTR) and Random Forest Regression (RFR), to predict the output of the solar panel. These environmental factors include temperature, humidity, sunlight, dust concentration, and electric current. The efficacy of both models was assessed via Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). In low-dust environments, the RFR model achieved MSE, RMSE, and MAE values of 0.00248, 0.0498, and 0.03912, respectively. Then the DTR model shows values of 0.00261, 0.05116, and 0.04055, respectively. In high dust environments, the RFR model shows superior performance with MSE, RMSE, and MAE values of 0.00173, 0.04159, and 0.04061, respectively. Then the DTR model produces worse values of 0.00181, 0.04263, and 0.04164, respectively. These findings consistently show that the RFR model outperforms the DTR model in both low and high dust conditions. Next, a real-time electric current prediction system is developed, featuring a user-friendly web-based application. The system uses a pre-drilled RFR model to predict future electric current values with more accurate results. By visualizing real-time data, comparing prediction results, and providing insightful insights, the application can be used by users to make timely decisions and improve energy system performance.