Comparative Analysis of Machine Learning Models for Solar Power Prediction Using Time Series Weather Data

Smart Grids preside over traditional power grids due to their efficiency and effectiveness. Smart grid implementation has benefited all components of the power system. These grids usually generate energy from renewable energy resources. However, the stability, reliability, and seamless functioning of the grid are adversely affected by the highly weather-dependent characteristics of renewable energy sources. As a result, accurate energy generation forecasting is an essential requirement for the grid's steady and reliable operation. The forecasting techniques for the quantity of electricity produced by solar power plants are discussed in this study. First, the impact of weather on solar power generation is investigated. Second, a machine learning method for solar panel output power prediction based on weather conditions is proposed. Finally, an ensemble approach utilizing different algorithms for machine learning is utilized to boost the accuracy of the proposed model. For the objective of creating prediction models, we utilized a variety of techniques, including temporal convolutional neural networks, fine trees, support vector machines with different kernel functions, artificial neural networks, and long-short-term memory RNNs. Each model's prediction accuracy has been assessed by comparing and contrasting these different methods. A comparative investigation indicated that the Temporal Convolutional Network-based model trumps the others in terms of efficiency. The proposed ensemble model exceeded the conventional individual models after performance comparison with the standard approaches.