A Hybrid Forecasting Approach for Solar Power Generation in Smart Grids Using Long Short-Term Memory and Autoregressive Integrated Moving Average

Abstract

An accurate solar energy forecast is important for the efficient operation of smart grids, especially with the increasing penetration of renewable energy sources. This paper proposes a hybrid forecasting approach that combines long short-term memory (LSTM) neural networks with autoregressive integrated moving average (ARIMA) models to improve the accuracy of short-term solar energy predictions. While ARIMA effectively captures linear temporary dependence, LSTM networks are powerful in nonlinear and long-distance pattern modeling. By integrating these two models, the proposed hybrid approach takes advantage of their complementary strengths to stop and address nonlinearity. The model is trained and tested on real-world solar power data collected from the gridconnected photovoltaic system. The evaluation metrics, such as mean absolute error, root mean squared error, and mean absolute percentage error, perform better than stand-alone ARIMA and LSTM models in the hybrid model, outpacing accuracy. Results outline the ability of hybrid intelligent models to increase the prediction of solar energy, contributing to more stable and reliable smart grid operations.

Cite this article as: R. Kanthavel and R. Dhaya, “A hybrid forecasting approach for solar power generation in smart grids using long short-term memory and autoregressive integrated moving average,” Turk J Electr Power Energy Syst., Published online September 3, 2025. doi: 10.5152/tepes.2025.25021.