Solar Power Prediction using LTC Models

Abstract

Renewable energy production has been increasing at a tremendous rate in the past decades. This increase in production has led to various benefits such as low cost of energy production and making energy production independent of fossil fuels. However, in order to fully reap the benefits of renewable energy and produce energy in an optimum manner, it is essential that we forecast energy production. Historically deep learning-based techniques have been successful in accurately forecasting solar energy production. In this paper we develop an ensemble model that utilizes ordinary differential based neural networks (Liquid Time constant Networks and Recurrent Neural networks) to forecast solar power production 24 hours ahead. Our ensemble is able to achieve superior result with MAPE of 5.70% and an MAE of 1.07 MW.

Analysis and Selection of Optimal Perovskite/Silicon Tandem Configuration for Building Integrated Photovoltaics Based on Their Annual Outdoor Energy Yield Predicted by Machine Learning

“…This study employed five techniques, namely RT, GPR, ET, SVM, and ANN, to establish prediction models. These algorithms have been previously documented in the literature [ 24,26–28 ] and won't be extensively detailed in this report. The ML models were implemented using MATLAB R2023b software alongside the Statistics and ML Toolbox.…”