Pronóstico de Consumo de Energía Eléctrica de distritos clientes de Hidrandina - Perú

[ulewis]

Abstract:

This study addresses the prediction of energy consumption in districts served by Hidrandina in Peru, using machine learning techniques. A dataset including information on customer energy consumption, district holiday dates, and atmospheric data was analyzed. The main objective was to develop a model capable of accurately predicting monthly energy consumption levels by district.
Various data preprocessing methods were applied, including imputation of missing values and outlier detection. Four machine learning models were evaluated: Logistic Regression with L1 (Lasso) and L2 (Ridge) regularization, Support Vector Machines (SVM), and Random Forest.
Results showed that the L1 Logistic Regression (Lasso) model achieved the best performance, with a mean precision of 0.95, mean recall of 0.95, mean F1 Score of 0.95, and mean ROC AUC of 0.98. This model outperformed more complex alternatives such as SVM and Random Forest, highlighting the importance of considering simpler and more interpretable models in this context.
The study underscores the effectiveness of machine learning methods in predicting energy consumption and the importance of balancing model complexity with interpretability. The findings have significant implications for Hidrandina's energy planning and management, allowing better anticipation of energy needs and more efficient resource distribution. Future research directions are suggested, including the incorporation of deep learning techniques and consideration of long-term factors such as climate change.

Grupo 8: Curso Machine Learning -2024 -1

- De La Cruz U. Lewis
- Gómez Villanueva Kevin
- Boza Gutarra Fernando
- Romero Ramos Yovany

Se adjunta dataset, el notebook lo pueden encontrar en : https://colab.research.google.com/drive ... aR_LmdHoTS , tambien el zip adjunto.