Solar Prediction in Multi-Area Networks Using Shallow Neural Network

Abstract

In the worldwide electric energy grid, power system operation depend heavily on the accuracy of solar energy forecasts. Thus, it is essential to guarantee that consumers receive a steady and consistent power supply. But because solar energy data is erratic, simple statistical and machine learning techniques are unreliable for predicting. Performance evaluation criteria, weather classification, predicting horizon, and other aspects all affect how well solar irradiance performs. As a result, it offers research on models for forecasting solar irradiance that use shallow neural networks. More than a thousand records of solar energy output are included in the dataset, together with meteorological data covering a full year. Various indicators are used to evaluate these models' performance. The model of the type that was chosen has a mean absolute error (MAE) of 0.24 and a mean square error (MSE) of 0.14.