Solar Photovoltaic Energy Forecasting Using Machine Learning and Deep Learning Technique

• Published in: 2022 IEEE 9th Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering (UPCON) pp. 1 - 6
• Main Authors: Singh, Prashant, Singh, Navneet Kumar, Singh, Asheesh Kumar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 02.12.2022
• Subjects: Computational modeling; Deep learning; Deep learning (DL); gated recurrent unit (GRU); long short-term memory network (LSTM); Photovoltaic systems; Predictive models; Reliability; Renewable energy sources; root mean square error (RMSE); Support vector machines
• DOI: 10.1109/UPCON56432.2022.9986446

The whole world is going through electrical fuel transition, from traditional to renewable energy (RE) sources. Natural resources like coal, natural gas, fossil fuels are still dominant energy sources to produce electrical energy throughout the world. If the switching towards RE source does not take place, these natural sources will deplete sooner, and heavy energy crises will come into picture. The paper addresses the issue of forecasting short-term renewable energy supply. The stochastic nature of RE sources has an impact on power system planning procedures, lowering the reliability as well as security of power supply for end users [1]. In this paper solar photovoltaic (PV) energy forecasting is performed using two dependent data variables such as (a) solar irradiance and (b) temperature, and past solar PV energy output using machine learning and deep learning (DL) algorithms. DL is a kind of complex learning inspired by human learning. Long Short Term Memory (LSTM) network and Gated Recurrent Unit (GRU) network are the examples of it. The paper investigates the issue of identifying features and determining suitable error metrics. DL model was developed and tested on real solar PV energy produced on MNNIT Allahabad, India campus. The forecasting performance of developed models is evaluated in terms of three important measures, (a) mean absolute error (MAE), (b) mean squared error (MSE), and (c) root mean square error (RMSE).