ENERGYLOADPRO: A Transformer-Based Interactive System for Sequential and Probabilistic Electricity Load Forecasting with Visual Analytics Interface

Abstract

Electricity load forecasting has become more complex due to dynamic consumption patterns, renewable energy integration, and the availability of high-frequency data. Traditional forecasting systems often lack flexibility, user interaction, and the ability to provide uncertainty-aware predictions. To address these challenges, this work presents EnergyLoadPro, an interactive forecasting system built on a Transformer-based framework for sequential and probabilistic electricity load prediction.

The system integrates data management, model training, forecasting, and analytics within a unified interface. It allows users to configure datasets, select models, define forecasting horizons, and enable probabilistic prediction intervals such as P10, P50, and P90. The system also supports visualization of prediction intervals and performance metrics, enabling better understanding of model behavior.

A key feature of the system is its modular workflow, which includes data ingestion, feature engineering, model training, forecasting, calibration, and visualization. The interface provides both simple and advanced modes, making it accessible for different user requirements. Additionally, the system includes automation capabilities for scheduled forecasting and supports performance tracking through visual analytics.

The proposed system demonstrates how advanced forecasting models can be integrated into a user-friendly platform, enabling efficient, interpretable, and scalable electricity load prediction suitable for modern smart grid environments.