Prediction of electric power and load forcasting using LSTM technique for EMS
A designed electric load and energy forecasting is proposed for buildings. Two different forecasting models, one for electricity consumption (medium-term load forecasting MTLF) and another for electrical en- ergy (very short-term) are proposed, compared, and interpreted. To feed those prediction mod...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
EDP Sciences
2025-01-01
|
| Series: | EPJ Web of Conferences |
| Online Access: | https://www.epj-conferences.org/articles/epjconf/pdf/2025/15/epjconf_cistem2024_03005.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | A designed electric load and energy forecasting is proposed for buildings. Two different forecasting models, one for electricity consumption (medium-term load forecasting MTLF) and another for electrical en- ergy (very short-term) are proposed, compared, and interpreted.
To feed those prediction models, and depending on dataset quality, two available online websites are chosen. Forecasting models are developed by PYTHON programming language, using various libraries dedicated to machine learning projects, such as ’TensorFlow’, ’Kiras’ and ’Scikit-learn’, and others for the visualization of results and data visualization like ’MatPlotLib’ and ’Seaborn’, and many other libraries. Because of the type of prediction based on time series, we used Long Short Term Memory (LSTM-type) neural network models.
The load forecasting model proposed in this study outperformed a previous engineering work on the same dataset. The proposed model achieved a minimal MAPE of 3.74% and a minimal RMSE that was approximately 20% lower than the engineer’s work. This article presents the proposed work and its results.
Simulation results are presented using Google collaboratory the hosted Jupyter Notebook service. |
|---|---|
| ISSN: | 2100-014X |