Forecasting Electricity Consumption Using ARIMA-LSTM Model

Authors P. Chakraborty , S. Kalaivani , A. Ambika , K. Ramya
Affiliations

Department of Electronics and Communication Engineering, B.S.A Crescent Institute of Science and Technology, Chennai, Tamil Nadu, India

Е-mail prernasree@crescent.education
Issue Volume 17, Year 2025, Number 4
Dates Received 18 April 2025; revised manuscript received 21 August 2025; published online 29 August 2025
Citation P. Chakraborty, S. Kalaivani, A. Ambika, K. Ramya, J. Nano- Electron. Phys. 17 No 4, 04011 (2025)
DOI https://doi.org/10.21272/jnep.17(4).04011
PACS Number(s) 07.05.Kf, 07.05.Mh
Keywords Electricity consumption forecasting, ARIMA-LSTM hybrid model, Time series analysis, Machine learning, Deep learning, Energy management systems.
Annotation

Accurate forecasting of electricity consumption is crucial for efficient energy management and planning. This proposed work compares two time series forecasting models – ARIMA (Autoregressive Integrated Moving Average) and an ARIMA-LSTM hybrid model – for predicting electricity consumption. The ARIMA model captures linear patterns, while the ARIMA-LSTM hybrid leverages Long Short-Term Memory (LSTM) networks to model non-linear dependencies. To evaluate performance, three metrics – Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE) – are used. Results show that the ARIMA-LSTM hybrid achieves an MSE of 45.19, RMSE of 6.72, and MAE of 5.80, outperforming the ARIMA model. This demonstrates the effectiveness of integrating statistical methods with deep learning for accurate forecasting. The hybrid model’s ability to handle complex time series data highlights its potential for improving electricity consumption predictions. By modeling both linear and non-linear dependencies, it enhances prediction accuracy compared to traditional approaches. These findings emphasize the significance of combining conventional and advanced techniques in time series forecasting. Future research could refine this model by incorporating additional features optimizing its architecture. Such improvements may further enhance forecasting accuracy, supporting better energy management and planning.

List of References