A Cloud-Integrated Virtual Framework for LSTM-Driven Solar Forecasting and Residential Energy Management

With the increasing dependence on renewable energy-particularly solar power-accurate forecasting and intelligent energy management, it has become essential for reducing grid dependency and optimizing energy usage. This paper presents an AI-powered renewable energy forecasting and scheduling system d...

Full description

Saved in:
Bibliographic Details
Published in:IEEE access Vol. 13; pp. 151055 - 151071
Main Authors: Devanathan, B., Suyampulingam, A., Selvaraj, P., Karuppasamy, Ilango, Ilamparithi, T.
Format: Journal Article
Language:English
Published: Piscataway IEEE 2025
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects:
Accuracy
Algorithms
Alternative energy sources
appliance scheduling
Decision trees
Energy consumption
Energy management
Energy value
Forecasting
Greedy algorithms
Home appliances
home energy management system
Household appliances
Long short term memory
Machine learning
machine learning algorithms
Meteorological data
Optimization
Predictive models
priority-based greedy scheduling algorithm algorithms
priority-based scheduling
Random forests
Real time
Real-time systems
Renewable energy
Renewable resources
Residential energy
Schedules
Scheduling
Smart buildings
Solar energy
solar energy forecasting
Solar power generation
Support vector machines
ISSN:2169-3536, 2169-3536
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:With the increasing dependence on renewable energy-particularly solar power-accurate forecasting and intelligent energy management, it has become essential for reducing grid dependency and optimizing energy usage. This paper presents an AI-powered renewable energy forecasting and scheduling system designed for smart homes. A Virtual platform is developed to predict next-day solar energy generation using real-time weather data obtained via Application Programming Interfaces (API) (such as., Open-Meteo), enabling hourly forecasts without the need for physical sensors. Various machine learning algorithms, including Lasso, Ridge, Support Vector Machine(SVM), Decision Tree, Random Forest, and Long Short-Term Memory (LSTM) networks, were trained and evaluated using historical data. The LSTM-based algorithm demonstrated the highest prediction accuracy and was selected as the core forecasting model. The forecasted solar energy values, at one-hour intervals, are dynamically used to schedule household appliances through a priority-based greedy algorithm. The algorithm prioritizes appliance loads based on criticality and energy availability, ensuring optimal solar utilization and minimizing reliance on grid power. The system is deployed via a user-friendly website with two interactive tabs, namely Forecast and Schedule, where users can select a date, view predicted solar generation, and receive an optimized appliance schedule. Experimental results indicate that intelligent scheduling using LSTM-based forecasting achieves up to 2.81% cost savings compared to conventional non-forecast-based methods. This scalable, sensor-free solution opens avenues for future enhancements, including multi-day forecasting and real-time optimization for broader smart energy applications.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2169-3536
2169-3536
DOI:10.1109/ACCESS.2025.3601722