AI-Enabled Predictive Load Management and Priority-Based Load Shedding for a Campus Solar Microgrid Using Edge LSTM Forecasting and Cloud IoT Monitoring

Author

Ankur R. Ramteke, Aarambh D. Deshmukh, Hrunali I. Dhobale, Sagar G. Tarone, Shreya G. Padalwar, Prof. M. R. Shelke


Abstract

 This paper presents an AI-enabled predictive load management scheme for a campus-scale solar microgrid. The prototype integrates a 30 W solar photovoltaic (PV) source, a 4 Ah lithium-ion battery and an Internet of Things (IoT) monitoring layer. Voltage, current, temperature, light and humidity sensors are sampled by an ESP32-S3 edge-computing unit, which exchanges data with an AI/ML engine. A Long Short-Term Memory (LSTM) network deployed with TensorFlow Lite Micro forecasts solar generation and load demand 24 hours ahead, using weather data from the OpenWeatherMap API. Based on the forecast, a smart relay module keeps critical loads (security lighting) energized and sheds non-critical loads (decorative lighting, fans) when an energy deficit is predicted, protecting the battery from over-discharge. Measurements are published over MQTT to a ThingsBoard CE cloud SCADA dashboard for remote monitoring. The work demonstrates a complete sensing–edge AI–cloud–control pipeline and is expected to reduce energy waste through intelligent load shedding.


Keywords

Solar microgrid, load forecasting, LSTM, edge AI, TensorFlow Lite Micro, load shedding, IoT, MQTT, SCADA, ESP32-S3.



Full Text:

Download Paper PDF


References


[1] S. R. Biswal et al., “An improved IoT based hybrid predictive load forecasting model for a greenhouse integrated with demand side management,” IEEE Access, vol. 13, pp. 108446–108460, 2025.

[2] A. R. V. Babu et al., “Solar energy forecasting using machine learning techniques for enhanced grid stability,” IEEE Access, vol. 13, pp. 93735–93753, 2025.

[3] M. J. Sammar et al., “Illuminating the future: A comprehensive review of AI-based solar irradiance prediction models,” IEEE Access, vol. 12, pp. 114394–114412, 2024.

[4] N. Sakib et al., “Prediction interval in renewable energy forecasting: A comprehensive review of uncertainty quantification methods,” IEEE Access, vol. 13, pp. 185466–185486, 2025.

[5] TensorFlow Lite Micro Documentation, Google Developers, Jun. 2025. [Online]. Available: https://www.tensorflow.org/lite/micro

[6] OpenWeatherMap API Documentation, Jun. 2025. [Online]. Available: https://openweathermap.org/api

Share your valuable work from Social Media Buttons